Themes from Klein: Knowledge, Scepticism, and Justification [1st ed.] 978-3-030-04521-0, 978-3-030-04522-7

Table of contents :
Front Matter ....Pages i-xviii
Introduction (Branden Fitelson, Cherie Braden)....Pages 1-12
Front Matter ....Pages 13-13
Reflective Knowledge (Ernest Sosa)....Pages 15-23
Epistemically Useful Falsehoods (Catherine Z. Elgin)....Pages 25-38
Theoretical Unity in Epistemology (Jonathan L. Kvanvig)....Pages 39-56
Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief (Richard Foley)....Pages 57-70
Knowledge, despite Evidence to the Contrary (Rodrigo Borges)....Pages 71-88
A Causal Aspect of Epistemic Basing (Robert K. Shope)....Pages 89-107
Front Matter ....Pages 109-109
The Moral Transcendental Argument against Skepticism (Linda Zagzebski)....Pages 111-128
Epistemic Humility, Defeat, and a Defense of Moderate Skepticism (Sharon Ryan)....Pages 129-143
Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination (Claudio de Almeida)....Pages 145-168
Front Matter ....Pages 169-169
Finite Minds (Michael Huemer)....Pages 171-187
Finite Minds and Open Minds (Jeanne Peijnenburg, David Atkinson)....Pages 189-196
Some Notes on the Possibility of Foundationalist Justification (Sanford C. Goldberg)....Pages 197-211
A Formal Account of Epistemic Defeat (Matthew Kotzen)....Pages 213-234
Benign Infinity (Matthias Steup)....Pages 235-257
Back Matter ....Pages 259-260

Citation preview

Synthese Library 404 Studies in Epistemology, Logic, Methodology, and Philosophy of Science

Branden Fitelson Rodrigo Borges Cherie Braden Editors

Themes from Klein Knowledge, Scepticism, and Justification

Synthese Library Studies in Epistemology, Logic, Methodology, and Philosophy of Science Volume 404

Editor-in-Chief Ota´vio Bueno, University of Miami, Department of Philosophy, USA

Editors Berit Brogaard, University of Miami, USA Anjan Chakravartty, University of Notre Dame, USA Steven French, University of Leeds, UK Catarina Dutilh Novaes, VU University Amsterdam, The Netherlands

The aim of Synthese Library is to provide a forum for the best current work in the methodology and philosophy of science and in epistemology. A wide variety of different approaches have traditionally been represented in the Library, and every effort is made to maintain this variety, not for its own sake, but because we believe that there are many fruitful and illuminating approaches to the philosophy of science and related disciplines. Special attention is paid to methodological studies which illustrate the interplay of empirical and philosophical viewpoints and to contributions to the formal (logical, set-theoretical, mathematical, information-theoretical, decision-theoretical, etc.) methodology of empirical sciences. Likewise, the applications of logical methods to epistemology as well as philosophically and methodologically relevant studies in logic are strongly encouraged. The emphasis on logic will be tempered by interest in the psychological, historical, and sociological aspects of science. Besides monographs Synthese Library publishes thematically unified anthologies and edited volumes with a well-defined topical focus inside the aim and scope of the book series. The contributions in the volumes are expected to be focused and structurally organized in accordance with the central theme(s), and should be tied together by an extensive editorial introduction or set of introductions if the volume is divided into parts. An extensive bibliography and index are mandatory.

More information about this series at http://www.springer.com/series/6607

Branden Fitelson • Rodrigo Borges Cherie Braden Editors

Themes from Klein Knowledge, Scepticism, and Justification

Editors Branden Fitelson Department of Philosophy & Religion Northeastern University Boston, MA, USA

Rodrigo Borges Department of Philosophy University of Florida Gainesville, FL, USA

Cherie Braden University of Colorado Boulder Boulder, CO, USA

Synthese Library ISBN 978-3-030-04521-0 ISBN 978-3-030-04522-7 https://doi.org/10.1007/978-3-030-04522-7

(eBook)

Library of Congress Control Number: 2019930397 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

The editors of this book, as well as its contributors, submit to you, the reader, this labor of love in honor of Professor Peter D. Klein, notable epistemologist, respected philosopher, beloved professor, oft-mocked infinitist, Rutgers Philosophy Department co-creator/advocate par excellence, and person most likely to roll his eyes at any and all praise contained in these pages. This book was the natural follow-up to the Peter Klein Conference in Epistemology, held at Rutgers University in the spring of 2016 in honor of Peter’s retirement. The co-editors and contributors have all found themselves, in various ways, in the orbit of Peter Klein the philosopher, colleague, and professor. Many of the contributors to this volume count Peter as a dear friend and longtime philosophical interlocutor. The contributions to this volume are a collective testament to the breadth and depth of Peter’s influence on epistemology of the twentieth century and today. The co-editors represent a range of groups, within academia, that Peter Klein has influenced over the years. Branden Fitelson is one of Peter’s former colleagues in philosophy at Rutgers, Rodrigo Borges is Peter’s former Ph.D. student (in fact, his very last Ph.D. student), and Cherie Braden is one of Peter’s former undergraduate students. There are people in the world who have understood the message of adulthood to be one of pure self-interest. To them, what it means to “keep your head when all about you/ Are losing theirs and blaming it on you” is to sidestep the chance of blame altogether by, to put it crassly, refusing ever to stick your neck out. Ah, old colonialist Kipling: his poem “If—” is nice for a quotation, but it isn’t a great source of advice, especially for philosophers, who would hardly wish to think but “not make thoughts [our] aim.” There are people in the world who have understood the message of compassion to be that unless your head is on the chopping block or served up on a sacrificial platter, you are not giving enough to others. There is also Peter Klein, who has learned to keep his own head square on his shoulders, but always (or we could say almost always if we’d like to leave reasonable room for the existence of defeaters beyond our ken—but Peter would object to such a blatant misunderstanding of the v

vi

Preface

significance of defeaters) with a mind to, supporting, promoting, and being a part of the world with the people around him. Peter looks after his own, but with a very generous conception of what it means for you and him to be on the same team. You could share a department, a school, a profession, a love of philosophy, or a love of learning. You could be fighting the same fire on a mountainside. Or sometimes it’s just that you’re both human. One can easily imagine this kind of openness being the downfall of a lesser person. On the other hand, cynicism may well be a hell bursting with middling minds (one of the co-editors, a recidivist cynic, thinks it is), in which case cynicism might not be the fate of the brilliant any more than is naive and ineffectual sacrifice. Peter Klein has greater strength of heart than most people and greater strength of mind than most people, but we should not be surprised that he exists or that these things are true of him. By stipulation, let that argument be sufficient for us to set aside all your unanswered questions concerning the fact of Peter Klein’s existing as he does. So we will not answer, for example, whether Peter is wonderful or quite wonderful. Anyway, it would be his strong preference that we turn to philosophy soon, or “now, immediately, now,” to make an evidential overassertion. Boston, MA, USA Gainesville, FL, USA Boulder, CO, USA

Branden Fitelson Rodrigo Borges Cherie Braden

Acknowledgments

We, the editors, would first like to thank Peter Klein himself, for agreeing to write an intellectual autobiography for this volume. The three of us have been immeasurably inspired by Peter: by his goodness, his kindness, his brilliance, his patience, his rare balance of practical wisdom and high ideals, and of course the unforgettably visible font size in his emails. And we are equally immeasurably grateful, although it is not clear whether any one of us would assent to calling two things without measure either equal or unequal, to know Peter. We would like to thank everyone who organized or attended the Peter Klein Conference in Epistemology at Rutgers in 2016 or the banquet afterward. We thank Larry Temkin for believing in the idea and getting behind it; Andrea Parente for bringing her activist energy to keeping “the Klein Plan” on track; Mike Huemer, Kate Elgin, and Sandy Goldberg for flying out and giving wonderful talks despite our budget; and Bob Shope (who also traveled far), Alvin Goldman, and Susanna Schellenberg for chairing the conference sessions. Thank you to Ernie Sosa, Karen Stubaus, Jane Grimshaw, Ziva Galili, Paul Leath, Dick McCormick, Mike Beals, Brian McLaughlin, and Dick Foley for their wonderful speeches at the banquet (for historical record, we also mention that Rodrigo and Cherie gave speeches; those were probably okay too). For the help at the event itself, we thank Ken Budrow, Susan Dimaio, Chris Hauser, and Janelle Derstine. For the help finding former friends and students of Peter’s, finding pictures and videos, and planning the event, we are especially grateful to Anne Ashbaugh, Kyra Todd, Ernie Sosa, Ziva Galili, Paul Leath, Dick McCormick, Mike Beals, Ann Martin, Susanna Schellenberg, Holly Smith, Brian McLaughlin, Dick Foley, Jane Grimshaw, Alvin Goldman, Rudy Bell, Robert McGarvey, Shari Reiner, David Burrell, and Emily Podhorcer. We thank Barry Loewer, Doug Husak, Howard McGary, Ernie Lepore, Barbara Callaway, Gabrielle Wilders, Martin Kempner, Jerry Balmuth, Dean Zimmerman,

vii

viii

Acknowledgments

Torrey Wang, Paul Clemens, Rob Bolton, Nancy Winterbauer, Harvey Waterman, Lourdes Valdivia, Claudio de Almeida, Guido Melchior, Brian Pollex, Andy Egan, Isabel Nazario, Troy Cross, Ted Warfield, Rogel de Oliveira, Greg Burdea, Robin Jeshion, Amy Cohen, Tim Maudlin, Joe Campbell, Arnold Hyndman, Matt Benton, Mike Huemer, Kate Elgin, Martha Cotter, Rosemarie Dudek, Yvonne Rutty, Bob Audi, Bob Shope, Peter Kivy, Luis Rosa, George Pappas, and Blake Roeber for their work with Andrea Parente to create a memory book and video, the latter with some help from Michael Lacanilao. With only old emails to go by, we have no doubt left names off these lists, and for that we apologize. We are ever grateful to Mercedes Diaz, Pauline Mitchell, and Jean Urteil, three amazing people who helped us in too many ways to list. We are thankful to Peter’s wife, Anne Ashbaugh, for colluding with us to make the event happen after Peter said “Absolutely not,” and to Rutgers administration for offering us funding from the beginning. Finally, we wish to honor beloved faculty members of the Rutgers Philosophy Department who retired soon before Peter or soon after, his comrades in thought: Peter Kivy, Holly Smith, Howard McGary, Robert Matthews, and Alvin Goldman. We are grateful to the whole gang at Rutgers. We may be scattered to the four winds, but you still mean the world to us. We are grateful to our superb collection of contributors. We could not have been more fortunate. Everyone who contributed to this volume did so without reservation. All we had to do was say “for Peter Klein.” While the persuasiveness of Fitelsonlevel enthusiasm cannot in general be overestimated, we are sure that in this case our contributors’ affection and respect for Peter and his work were what brought everyone aboard. (But two of the co-editors insist it be added that Fitelson-level enthusiasm does make everything more fun.) What’s more, no teeth needed to be pulled. No dentist visits were ever considered, at least from the perspective of the co-editors. Our contributors were patient to near saintliness with our delays. The horror stories about edited volumes all seem to involve difficulties extracting the work from contributors, but in this case, the contributors always either kept pace with us or were ahead. Every request from a contributor for extra time gave us an excuse to do our other work, so we appreciate it. Having made that admission, we should also say that we owe a very big thanks to our editors and contacts at Springer—Solomon George, Jeyaraman Hemalatha, Ties Nijssen, Hans van Sintmaartensdijk, Werner Hermens, and Otavio Bueno—who have been nothing but generous, accommodating, and excited about this project. Finally, we thank Bob Pasnau for his clever suggestion regarding the title of the book.

A Note from Peter Klein

I was 14, riding a horse in Montana. . . No, before I get to my story about how and why I got to be a philosopher, I must thank Rodrigo Borges, Cherie Braden, and Branden Fitelson. When I first heard about their intention to put together some articles on my work, I asked them not to do it. I had many reasons for that request, but a primary one was that each of them has their own research and teaching that should take priority over such a project. I knew this would take way too much of their time. But they persisted, and I am very, very pleased and grateful to them. I also thank all the contributors to the volume. I am honored that they thought my work and views are worth critical examination. That’s all a philosopher should hope for. I deliberately have not yet read most of their essays because I knew doing so would keep me awake nights, and I had a few deadlines that I couldn’t postpone. I knew that none of the papers would simply say, “I thought about it, and Klein is right!” That seldom, if ever, happens. If it did, we philosophers would be out of business. Not agreeing with each other is necessary for the survival of homo philosophicus. However, once I study the papers and figure out whether there are good ways to revise my views in response to objections, or whether I should just give some of my views up, I’ll try to find some way to respond in print. ♢ My “intellectual autobiography” could be stated very succinctly. I knew almost nothing that was important. But luckily Plato’s Meno and Descartes’ Meditations were assigned reading in a required college course—“Great Historical Issues” at Earlham College, in 1959—and I came to believe there was something I could learn that was important. And after that my luck continued. That’s what I should talk about rather than talking about “my philosophy” because as Hilary Putnam reportedly said, “Any philosophy that can be put in a

ix

x

A Note from Peter Klein

nutshell belongs in one.”1 In addition, I am confident that Branden, Cherie, and Rodrigo (in the introduction to this collection) and the contributors (in their essays) will thoroughly explore whatever nuttiness there is in my views and, perhaps, even find something worth munching on. So, I was 14, riding a horse in Montana and talking with the guy riding a horse alongside mine. I have forgotten his name, but not the conversation. It was about whether we know what is morally right or wrong, and it went something like this: THE OTHER GUY: We don’t know that kind of thing because all we know begins with our senses, and our senses can’t give us that kind of knowledge. ME: I agree that our senses can’t provide that kind of knowledge, but since we do know that some things are right and other things are wrong, there must be another way we have of getting that kind of knowledge. THE OTHER GUY: What is that other way? ME: I don’t know. I’ll think about it.

Then, we went on riding. Although I have a pretty clear memory of that conversation—including the name of my horse, Red Bud, a half thoroughbred and half quarter horse—it probably wasn’t as clear, articulate, or stark as I just described. And, of course, I didn’t know how prognostic that conversation would turn out to be. Although I have never yet written about moral knowledge, my life in philosophy (as distinct from my U.S. Forest Service life and from my life as an academic administrator) has really been about trying to figure out what that “other” way of getting knowledge is. Turns out that reasoning, if properly done, is that other source of knowledge! And, in a nutshell (oops!), that’s what I’ve been trying to explore. How can reasoning help to generate knowledge? The answer, I think, is that we justify our beliefs by providing good reasons for them. It’s as simple as that. Justifying is like certifying, vivifying, rectifying, magnifying, etc. We make a proposition justified and thereby contribute to the knowledge that it is true. In most cases, we have no control over what is true. But we create knowledge by making some propositions known. The chain of reasons never ends because we can, and often do, challenge any temporary stopping point that serves as a premise of the reasoning we employed to justify the target belief. A great thing about reasoning is that it doesn’t settle matters. Rather, it raises new questions about the truth of the reason just given and about our ability to recognize the truth of the reason just given. Reasons to an epistemic agent are like mountains to a mountain climber. We seek them because they are there. That’s the infinitist position. But reasoning can also produce beliefs that merely look like knowledge to us because they are justified. Not every justified belief is true, and even if it were true as well as being justified, it might not be knowledge. Suppose I am justified in believing there is a horse in the corral because I am justified in believing that Red Bud is in the

Martha Nussbaum reports Putnam’s comment in her wonderful tribute to him. See “Hilary Putnam (1926–2016),” THE BLOG, Huffington Post, 03/14/2016; https://www.huffingtonpost.com/martha-c-nussbaum/hilary-putnam-1926-2016_b_9457774.html. 1

A Note from Peter Klein

xi

corral. And suppose, further, that what I was looking at was a Daedalus-like statue of Red Bud and that behind the statue is a horse that I don’t see. My reasoning led me to a true justified belief (that there is a horse in the corral), but not to knowledge that there is a horse in the corral. I did my best and reasoned well. What else should I have believed, given that I believed that Red Bud was in the corral? The defeasibility theory holds, roughly, that I know that there is a horse in the corral just in case (1) there is a horse in the corral, (2) I believe that there is a horse in the corral, (3) my belief that there is a horse in the corral is justified, and (4) there is no true proposition that defeats my justification. Roughly, a true proposition defeats the justification of any proposition x if and only if the conjunction of the true proposition with the reasons used to justify x is such that x is not justified. That’s the defeasibility theory of knowledge in a fairly large nutshell. Of course, those theories—as stated—are too flat-footed. There are lots of problems with them, and I’m sure that the contributors to this volume will not be shy in stating and supporting their objections. That’s what philosophy is about. But I digress. ♢ Let’s go back to the story of how and why I became a philosopher. The first chapter was about my momentary interest in the source of moral knowledge. I was 14, and it’s now almost 64 years later. There are a few gaps to fill in. Each gap was filled by good luck. The second chapter of the story is brief, and you’ve already read it. Luckily, I was required to take the course with the readings from Plato and Descartes. I enjoyed them a lot but formed no intention of becoming a philosopher. My father was a doctor, and my mother taught nursing. So, of course, I was slated to become a doctor. The slight problem was that I had already discovered that I didn’t like being around sick people. Consequently, I thought I was doomed to earn a living doing what I disliked. Luck intervened again. During the summer break after I took the required course, I was working on roofs as a tinners apprentice. It was difficult work, and it was semidangerous. During a good part of that summer, the roofing company I was working for was installing galvanized sheets of metal on prefabricated gas stations. Unlike the 26–27 gauge corrugated, painted metal sheets used on houses, these sheets were heavy, shiny, and unpainted. We would interlock the metal panels and solder the seams with wooden-handled soldering irons that were heated in a portable charcoal brazier. The sun would beat down on your back, and the reflected heat from the roofing would roast your face to medium-well. It was August. There was the sun, the brazier, and the heated soldering iron you held as you slid the hot soldering bar along the seams. All in all, it was just hot. You’re wondering where the luck is. I’ll get to it. I sometimes got a break and worked on private houses installing or repairing gutters. Some of the houses were three stories high, and I don’t like being high up on a ladder almost as much as I don’t like being around sick people. Needless to say, I didn’t like the work all that much. But I did like the guy, Paul, for whom I was an apprentice. And for some reason he liked me. I think it was because I was so bad at trying to be a good tinners apprentice.

xii

A Note from Peter Klein

Here’s an example: One day we were helping some other guys carry the 90-pound rolls of roofing paper that would eventually be placed, lasagna-like, between layers of hot, sulfur-smelling tar. Or more accurately, Paul was helping them carry some of the rolls. He would put one roll on each shoulder and go up the ladder without using his hands. Did I tell you Paul was 60 400 and looked like a professional wrestler? The other guys were amazed. Paul “suggested” that I stop sitting on my butt (he didn’t quite put it that way!) and carry some of the rolls up the ladder. Then, as I picked up one roll and put it on my right shoulder, I saw the other guys move away from the bottom of the ladder. I was a bit puzzled. But here’s why they did it: They knew that given the way I was clutching the ladder and in order to avoid falling myself, I would have to drop the roll right where they were standing when I stepped off the ladder and onto the roof. Knowing is factive, and I learned that you’d better lean forward as you step off the ladder and let the momentum (which was pretty strong because of the extra 90 pounds) help you step off the ladder onto the roof rather than fall backwards off the ladder. Patience. I’m getting to the lucky part. I just wanted to “contextualize” it. On another blistering hot day on the hot roof with hot soldering iron next to the hot brazier, Paul asked what I was going to be after college. I replied, “I don’t know; maybe a teacher. And I do like reading and studying philosophy.” I added a little bit about what was interesting to me about philosophy. I did not mean to put those two (philosophy and teaching) together to indicate I was interested in becoming a teacher of philosophy. It was a mere conjunction. But, luckily, Paul put them together and said, with a bit of a wry smile, “You mean you’re going to get paid to teach that bullshit.” I replied, “It would sure as hell beat working on these f. . .ing roofs, wouldn’t it?” That’s what I said out loud. But to myself I thought, “Eureka!” That’s it. That’s how I became a philosopher. First, the casual talk about moral knowledge, then a required college course, and, finally, Paul’s question. Thomas Hardy could not have written a better story! Of course, there were some other lucky things that happened. At Earlham, I had two wonderful professors. One was Wayne Booth, a professor of English who studied with Richard McKean— the famous Aristotle scholar and a chief advisor to the United Nations in the writing the Universal Declaration of Human Rights—at the University of Chicago. Wayne returned to Chicago, published his The Rhetoric of Fiction, in which he defended a (roughly) Aristotelian account of literary fiction. Wayne was a second father to me. He taught me how to study and, by example, showed me how to be a good teacher. I hope he would be proud of me. The second wonderful teacher was Grimsley Hobbs (how’s that for a name?). Grimsley was at least 60 600 and was strong as an ox. He and his wife had bought and restored an old water-wheel-powered flour mill. Once, when I was visiting his home, I saw him carrying large sacks of wheat to the top of the mill on flights of stairs that hugged the inside walls of the mill tower. Here was a man who had just given a class on Plato’s Apology and was now carrying those very heavy sacks of wheat. We talked about his two lives, and he explained that accomplishing anything in philosophy took a very long time, and sometimes you would find out that you were completely wrong. But, he said that grinding wheat and producing Hobbs Flour was a way to get some immediate satisfaction (and supplement his measly pay as a college professor).

A Note from Peter Klein

xiii

I got my immediate satisfaction from (1) working various jobs (lookout, firefighter, and dispatcher) in fire control for the U.S. Forest Service for 21 summers and (2) doing various administrative jobs (Chairperson, Vice Dean of the School of Arts and Sciences, and Associate and Vice Provost) for Rutgers University for a combined total of (not all the same) 21 academic years. I could tell if I was successful by just looking at the height of the smoke column or at a list of new faculty hires. But those nonphilosophic lives are (pardon what’s coming!) a horse of a different color and not really part of this story. I was lucky in graduate school at Yale to have Rulon Wells as my thesis advisor. We would meet in his little flat in one of the colleges. He had a large rolltop desk, and each drawer held what must have been hundreds of file cards with his notes about the books and articles he had studied. Each time we met, he would open the drawer and take out some of the cards to check that he was remembering the item correctly, and then he’d tell me to read the articles in the order he was giving them to me and write up a little something about each for the next week’s meeting. He wouldn’t tell me why the order was important. He let me puzzle that out myself. He was a real scholar and teacher. I’ve never been that good, but it was something to aim at. There were lots of other lucky things in my philosophical life. For example, I got to go to an epistemology summer camp sometimes referred to as an NEH Summer Institute. The main lecturer was Roderick Chisholm, who was and still is my role model for what a philosopher should be. If I could write one article, or maybe even a few paragraphs, as well as any of his papers or book chapters, I would accomplish what I aim at in philosophy. More luck: Robert Martin (the semantic category solution to the liar paradox guy), who was a high-school friend of mine and graduate student colleague at Yale, was in the Philosophy Department at Rutgers University in 1970, and, luckily for me, they were expanding the Department. I was hired away from Colgate University by Rutgers and taught there for 46 years. I’ve had great students and great colleagues. They’ve read my work, criticized it, and improved it. Here are some of the colleagues in epistemology I had: William Alston, Branden Fitelson, Richard Foley, Anthony Gillies, Alvin Goldman, Brian McLaughlin, Jonathan Schaffer, Susanna Schellenberg, Ernest Sosa, and Stephen Stich. How’s that for a set of colleagues interested in some of the same issues as I am! I was lucky enough to count Dick, Ernie, and Brian as close friends. And I was also lucky to have Ernie Lepore (a non-epistemologist) as a close friend in my life as an administrator. More luck still: I am married to a philosopher, Anne Ashbaugh, who tries to keep me humble by explaining Plato’s solutions to every issue on which I thought I had something original to say. Maybe it is all a footnote. Okay. That’s my story about how and why I got to be philosopher. The moral of the story is that I couldn’t have been luckier and, for me, it is a hell of a lot better than working on hot roofs. 6/18/2018

Contents

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Branden Fitelson and Cherie Braden

Part I

1

Knowledge

2

Reflective Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ernest Sosa

15

3

Epistemically Useful Falsehoods . . . . . . . . . . . . . . . . . . . . . . . . . . . Catherine Z. Elgin

25

4

Theoretical Unity in Epistemology . . . . . . . . . . . . . . . . . . . . . . . . . . Jonathan L. Kvanvig

39

5

Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Richard Foley

57

6

Knowledge, despite Evidence to the Contrary . . . . . . . . . . . . . . . . . Rodrigo Borges

71

7

A Causal Aspect of Epistemic Basing . . . . . . . . . . . . . . . . . . . . . . . Robert K. Shope

89

Part II

Scepticism

8

The Moral Transcendental Argument against Skepticism . . . . . . . . 111 Linda Zagzebski

9

Epistemic Humility, Defeat, and a Defense of Moderate Skepticism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Sharon Ryan

xv

xvi

Contents

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Claudio de Almeida

Part III

Justification

11

Finite Minds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 Michael Huemer

12

Finite Minds and Open Minds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 Jeanne Peijnenburg and David Atkinson

13

Some Notes on the Possibility of Foundationalist Justification . . . . . 197 Sanford C. Goldberg

14

A Formal Account of Epistemic Defeat . . . . . . . . . . . . . . . . . . . . . . 213 Matthew Kotzen

15

Benign Infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 Matthias Steup

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

Contributors

David Atkinson University of Groningen, Groningen, The Netherlands Rodrigo Borges Department of Philosophy, University of Florida, Gainesville, FL, USA Cherie Braden University of Colorado Boulder, Boulder, CO, USA Claudio de Almeida Pontifical Catholic University of Rio Grande do Sul, Porto Alegre, Brazil Catherine Z. Elgin Harvard University, Cambridge, MA, USA Branden Fitelson Department of Philosophy & Religion, Northeastern University, Boston, MA, USA Richard Foley New York University, New York, NY, USA Sanford C. Goldberg Department of Philosophy, Northwestern University, Evanston, IL, USA Michael Huemer Department of Philosophy, University of Colorado Boulder, Boulder, CO, USA Matthew Kotzen UNC Department of Philosophy, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA Jonathan L. Kvanvig The Department of Philosophy, Washington University in St. Louis, St. Louis, MO, USA Jeanne Peijnenburg University of Groningen, Groningen, The Netherlands Sharon Ryan West Virginia University, Morgantown, WV, USA Robert K. Shope University of Massachusetts, Boston, MA, USA Ernest Sosa Department of Philosophy, Rutgers University, New Brunswick, NJ, USA xvii

xviii

Contributors

Matthias Steup Department of Philosophy, University of Colorado Boulder, Boulder, CO, USA Linda Zagzebski Department of Philosophy, University of Oklahoma, Norman, OK, USA

Chapter 1

Introduction Branden Fitelson and Cherie Braden

Abstract Themes from Klein: Knowledge, Scepticism, and Justification is a collection of essays written to honor retiring philosopher Peter D. Klein, whose work has been and continues to be influential in the ongoing development of contemporary epistemology. Klein has done important work on skepticism, the Gettier problem, the structure of justification, defeasibility theory and defeaters, and his own hotly debated theories, including infinitism and his attempt to systematize what he calls “useful falsehoods.” Klein’s ideas have been influenced, he tells us, by twentieth-century predecessors such as Roderick Chisholm, Ludwig Wittgenstein, and A. J. Ayer. But the reader of Klein’s work will be struck not by whatever similarity it bears to what has come before, nor by the extent to which it departs from the past, but rather by its intellectual integrity. Neither a conformist nor a reactionary, Peter Klein marches to the beat of one drummer: his reasoning. This introduction provides a broad overview of Klein’s ideas on each of the book’s themes (knowledge, skepticism, and justification). Keywords Peter Klein · Defeasibility · Defeasible reasoning · Skepticism · Infinitism · Knowledge defeaters · Structure of justification · Reasoning agent · Epistemology

Themes from Klein: Knowledge, Scepticism, and Justification is a collection of essays written to honor retiring philosopher Peter D. Klein, whose work has been and continues to be influential in the ongoing development of contemporary epistemology. Klein has done important work on skepticism, the Gettier problem, the structure of justification, defeasibility theory and defeaters, and his own hotly debated theories, including infinitism and his attempt to systematize what he calls “useful falsehoods.” Klein’s ideas have been influenced, he tells us, by twentiethB. Fitelson Department of Philosophy & Religion, Northeastern University, Boston, MA, USA e-mail: branden@fitelson.org C. Braden (*) University of Colorado Boulder, Boulder, CO, USA © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_1

1

2

B. Fitelson and C. Braden

century predecessors such as Roderick Chisholm, Ludwig Wittgenstein, and A. J. Ayer. But the reader of Klein’s work will be struck not by whatever similarity it bears to what has come before, nor by the extent to which it departs from the past, but rather by its intellectual integrity. Neither a conformist nor a reactionary, Peter Klein marches to the beat of one drummer: his reasoning. The essays in this volume either directly address Peter Klein’s work or continue a discourse in which his work has played an important role. Klein writes, “the three topics I have written about [are]: What is knowledge? How ought one respond to various forms of skepticism? What is the structure of justification?” (2010). Accordingly, we have organized this introduction and the volume itself roughly along those lines. Without covering the full breadth of Klein’s engagement in answering the three questions, we take a moment to discuss some of his motives and how they shape the sorts of answers he gives.

1.1

Knowledge (Chaps. 2, 3, 4, 5, 6, and 7)

There is a just-so story, familiar to perhaps every epistemologist, that must be rehearsed before we begin. In due keeping with the nature of just-so stories, we retain inaccuracies as best we can. In the early 1960s at some college in the northeast, a philosopher named Edmund Gettier sat philosophically and tweed-enrobed in his office. So lost in philosophical thought was this slacker genius, a.k.a. ideal philosopher, that he had never bothered to publish and was now at risk of being denied tenure. But all was set right when he reached into the desk drawer and retrieved a three-page, carelessly typed paper (or possibly a note spanning four or five napkins) whose glory bathed him in a heavenly light previously reserved only for the personal guardian angels of Sir Thomas Aquinas. Before this moment, epistemology consisted mainly of a centuries-long mistaken definition of knowledge as the sum of three ingredients: justification, truth, and belief (JTB).1 Oh, the horror. The Gettier Paper, with its economy of shiny pennies and a post-industrial Ford owned by someone other than Jones, ushered epistemology out of a dark era characterized by primitive positivists Chisholming away at the walls of moldy caves. Gettier’s paper contained a few short counterexamples to the JTB conception of knowledge, cases in which justification, truth, and belief were all present but knowledge was nowhere to be seen. Epistemologists, Peter Klein and Sir Galahad among them, set out on a quest for the correct analysis of knowledge. Okay, enough of that. A number of analyses have been proposed since the beginning of the Gettier age. Their variability reflects, to some extent, the variability of epistemologists’ priorities in analyzing knowledge. Virtue epistemologists prioritize the moral significance of

1

See Klein (2017, 36–37) if you are ready to give up on fairytales concerning Gettier (1963).

1 Introduction

3

knowledge. Reliabilists prioritize third-person measures of the effectiveness of the knowledge concept in achieving veritistic aims. And there are others. Peter Klein’s analysis of knowledge prioritizes the avoidance of paralysis from skeptical doubt. Thus he must define knowledge at least in part from the perspective of the first-person rational agent, making his an “internalist” analysis. The “externalist,” on the other hand, looks to facts unavailable to the agent in order to decide whether the agent has knowledge or not. Reliabilists, for example, claim that we know a proposition p if p is true, we believe it, and the process by which our belief in p was produced is in fact—that is, externally—reliable. The typical reliabilist does not require that we be able to identify the reliability fact that produced the belief that p, and in many cases he claims that we cannot identify it. Unlike the Kleinian, the typical reliabilist is not so concerned with the inner experience of skeptical anxiety.2 Klein assures us time and again that not every kind of doubt rationally compels us to fear for our knowledge. For example, we might doubt whether our evidence for the propositions we know contains only truths. To preserve knowledge in the face of this doubt, Klein developed a theory of “useful falsehoods” (2008). We might also doubt that our beliefs are consistent with each other and worry that we are irrational if they are not, in response to which Klein has argued that for our beliefs to be “weakly inconsistent” is actually an epistemic virtue (1985). We might doubt that creatures such as we, who rely on evidence that is often misleading or incomplete, can have knowledge at all.3 To quell this doubt, Klein has developed his take on the defeasibility theory of knowledge (see, for example, 1971, 1976, 1983, 1996, 2017), to which we now turn, and which aims to explain how we can be non-omniscient and still both work toward and have knowledge. According to Klein’s earliest formulation of defeasibility theory, a subject S has knowledge of a proposition p at a time t means (i) p is true at t, (ii) S believes p at t, (iii) p is evident to S at t, and (iv) there is “no true proposition such that if it became evident to S at t, p would no longer be evident to S” (1971, 475).4 Condition (iv) is a proto-version of Klein’s defeater condition on knowledge. Defeaters are true propositions unavailable to S (at t) that would undermine S’s justification.5 In Chap. 5, Richard Foley—like Klein an internalist—offers an alternative to defeasibility theory, replacing justification with a new condition. In Chap. 4,

2 Klein himself gives a useful runthrough of differences among theories of knowledge and theories of justification in his 1998a encyclopedia entry. 3 For more on misleading defeaters, see Klein (1979 and 1980). 4 In the 1971 paper, this is given as a definition of “propositional knowledge.” 5 In one later paper, Klein adds two additional conditions to avoid the interference of epistemic luck: S must be an ideal epistemic agent, believing all and only justified propositions for no other reason than that they are justified, and S must believe of each of these beliefs that it is justified (1984, 155–156). The final addition is intended to ensure that S is in fact an epistemic agent and not merely a very accurate machine, but the move is questionable, which may be why Klein prefers the 1971 paper (per email correspondence 2018). A more recent definition of a defeater appears in Klein 2017. Defeaters are discussed a bit more in the final part of this introduction. They are also discussed or refined in a number of the chapters in this book, especially Chaps. 9, 14, and 6, by Sharon Ryan, Matthew Kotzen, and Rodrigo Borges, respectively.

4

B. Fitelson and C. Braden

Jonathan L. Kvanvig deploys a form of defeasibility theory to develop a unified epistemological theory grounded in confirmation. In light of the Gettier question, part of Klein’s quest is to make defeasibility a “Gettier-proof” theory, not subject to defeat by counterexample (2004c, 167). But defeasibility must take care not to be too resilient. Rodrigo Borges, in Chap. 6, argues that Klein’s brand of defeasible knowledge too heartily resists defeat by counterevidence. In Chap. 3, Catherine Z. Elgin takes Klein (2008) to task for using an “etiological” definition of doxastic justification in his defense of useful falsehoods. Etiological, or “causal” theories in epistemology hold that the causal process by which a belief was formed is relevant to determining whether it counts as knowledge. Klein has long opposed causal views (see, for example, 2012, 152–153; 2017, 42).6 In part, this is because Klein thinks causal analyses construe knowledge as something that happens to us rather than something we acquire through our correct reasoning. By contrast, Klein thinks, defeasibility theory passes normative judgment on an agent’s reasons (and reasoning) to determine whether p counts as justified, known, or neither (see 1995, 228–229).7 To have reasons that pass muster—or that can even qualify as “reasons” in the epistemic sense—one must be an epistemically responsible agent. “A responsible epistemic agent [meaning a properly epistemic agent] strives to believe all and only those propositions worthy of belief,” and to capture this striving, justification must have internalist features (2007b, 6).8 Robert K. Shope, in Chap. 7, defends a causal and somewhat normative/agentprioritizing account of the justification condition on knowledge, arguing that at least some of our beliefs are caused in part by our appreciating or deploying reasons for them. In Chap. 2, Ernest Sosa defends the value of what he calls “reflective” knowledge, which is at once causal and normative. But he accepts a non-normative etiology view of the minimally necessary conditions for knowledge,9 believing only that knowledge is meaningfully enhanced when its bearer further supports it through rational reflection. Klein, on the other hand, thinks that consciousness is a precondition of the epistemic: justification and knowledge are fundamentally pursuits of conscious beings.10 Otherwise, we would be equally

6 One of Klein’s favorite points of attack against etiology-of-belief views is their inability to identify which reliable process is relevant to the formation of any given belief, or what causal mechanism produces, in general, beliefs that q from beliefs that p (2012, 161). He divides etiology views into four prevalent approaches: “tracking theories, safety theories, reliabilism, and virtue epistemology” (see 2017, 42 for details). 7 Klein takes all causal/etiology views to be “naturalized” to varying degrees and his own view, by contrast, to be normative (1998b, 362). 8 Contrary to this assumption, Sanford C. Goldberg, in Chap. 13, offers an externalist account of epistemic obligation. 9 Knowledge that meets the minimally sufficient conditions for being knowledge is, according to Sosa, “animal knowledge.” 10 Klein (2007b, 5) grants that Sosa’s non-normative knowledge is also a kind of knowledge, but Klein thinks that animal knowledge is not the properly epistemological kind. Note also that the sort of agency useful for thinking about Klein’s views does not entail doxastic voluntarism (6).

1 Introduction

5

epistemically agential if we were to drift along, aimlessly allowing beliefs to come and go (Klein 1983, 155–156). The confused agent who reasons poorly but effortfully is more appropriately called “epistemic” on Klein’s view than someone who “believes” unreflectively yet, by coincidence, accurately. In recent years, “knowledge-first” epistemologists have worried that any analysis of knowledge fails to recognize that knowledge is more fundamental than epistemic analysis. Knowledge, the knowledge-firster contends, should be taken as an epistemic primitive. Klein, on the other hand, insists that what’s primitive is the reasoner qua reasoner.11 Both views purport to put the Gettier problem to rest, although the knowledge-firster’s path is less arduous. And both Klein and the knowledge-firster wish to avoid throwing away perfectly good knowledge. Unlike the knowledge-firster, however, Klein sees epistemology as driven by a commitment to understanding knowledge not simply as an artifact, but as a reasoner’s pursuit. For Klein, epistemic bedrock is the reasoning agent. The burden of building the reasoner into knowledge’s meaning falls to justification, which will be discussed in the last part of this introduction.

1.2

Skepticism (Chaps. 8, 9, and 10)

“Most, if not all, of our knowledge of the world (if, indeed there really is such knowledge) is not based upon exclusively deductive reasoning from propositions expressing the content of our sense experience” (Klein 2004a, 125; see also 2017, 39, 47). Unless we are mired in skeptical angst, we will accept that we regularly come to know propositions using “less than certain evidence,” “reasoning that is less than absolutely truth preserving,” and “on less than exclusively deductively valid reasoning” (2004a, 126). But the skeptic forces us to entertain that there is a gap between our level of sureness and bird’s-eye-certainty, a gap that can only be occupied by doubt (1987, 270; see also 2004b, 302).12 We set aside the specifics of formulating skeptical arguments, as there are plenty of these to be had in the chapters of Part II themselves. Philosophers have proposed many ways of answering the skeptic. Many of those include giving up various things that we tend to think we know, such as epistemic closure (a principle that allows us to infer that we know propositions that deductively follow from propositions we know) or epistemic humility (the recognition that our evidence is not complete and our reasoning sometimes errs). Klein, perfectly exemplifying the goals of defeasibility theory, gives up none of this intuitive knowledge.

11

Klein himself does not speak in terms of primitives or fundamentality within the epistemic. Defeasibility theory is nonetheless distinct from “reasons-first” views. See Sylvan and Sosa (2018) for a characterization of—and their disagreement with—the reason-firster. 12 Unlike Bertrand Russell, Klein does not require that the falsity of a belief be logically impossible in order for the belief to be certain (Klein 1992, 62).

6

B. Fitelson and C. Braden

The force of the skeptical hypothesis flows from the question, “How can there be knowledge in the presence of doubt?” For a Kleinian, giving a justified answer to that question requires doing some reasoning, because a reasoner who reasons is a necessary part of justification according to Klein. By his own standards, he cannot be permitted to answer the question with an unreflective, “I just know.” Klein has given many answers to the skeptic, and we only mention a few. He has argued that the Dretske-Nozick version of the skeptical argument begs the question, because since we have good reasons to accept the closure principle, the skeptic cannot defend the premise that S doesn’t know that she isn’t in a skeptical scenario (see, e.g., Klein 1987, 274). Elsewhere, he has argued that in global skeptical cases in which all our evidence equally supports our being either in a skeptical scenario or the “real” one, there’s no initial motivation for reasonable doubt (see, e.g., 2010, 159; see also 2004c). According to Klein, the great mistake of the contemporary argument for skepticism is that it asks us to doubt without reason. Descartes, on the other hand, reasonably entertained skepticism because of his own capacity for error due to his imperfect epistemic equipment (Klein 1990, 112). But modern skepticism makes appeal to circumstances beyond the control even of the epistemic agent who has flawless equipment, so that any evidence we could get would be misleading even if our inferences from it were perfect (2010, 159).13 In Chap. 8, Linda Zagzebski develops several “transcendental” arguments against the skeptical hypothesis. These arguments address Cartesian, “practical” skepticism—which threatens to infect our belief systems with genuine doubt—rather than purely theoretical skepticism. But unlike Klein, Zagzebski includes the brainin-a-vat case among skeptical cases that leave us susceptible, at least at first flush, to Cartesian doubt.14 An agent, she thinks, could genuinely worry that perhaps she is wrong about everything. Transcendental arguments save us from skeptical doubt by showing how the skeptical hypothesis is self-undermining. At times, Klein too deploys transcendental responses to skeptical claims: “But even if the epistemic warrant of the proposition that there are tables would be lowered [in a case in which a hitherto unreliable person gives testimony that a skeptical scenario obtains], it would occur only because I continue to be justified in believing that there are people who know things about the world” (1990, 117n30). If Klein’s full answer to skepticism is that the skeptic begs the question or that the skeptic’s question is unmotivated, he appears, like the dogmatist, to be refusing to entertain skeptical doubt.15 Given Klein’s explicit rejection of dogmatism, there must be more to his response (see 2000a for his thorough sendup of dogmatism). The 13

For more Klein on skepticism, see also 1982 and 2002. So too does Claudio de Almeida in Chap. 10. 15 Taking Klein out of context makes it easy to read his response to the skeptic as dogmatic. For example, he admits in a 1986 paper to having defended (in Klein 1981) that “since many of our beliefs are in fact such that there are no defeaters, the Cartesian requirement of indubitability is in fact satisfied” (1986, 381n12). This indeed sounds dogmatic. But Klein has substantially revised his view of defeaters since writing Certainty (for an example, see 1995, 230n5). Klein (2017) suggests we can find his best answers to skepticism in Klein 1981, 1995, and 2004c. 14

1 Introduction

7

answer he would ultimately like to defend seems to be this: Doubt that is reasonable when we are first asked the skeptical question does not remain reasonable after we have done the requisite reasoning. There is reasoning that is requisite for answering the skeptic, since to be doxastically justified16 we need to have reasons and we need to have deployed them. But the skeptic cannot reasonably dictate how we reason against him, and upon reflection we notice that we can reasonably reject the skeptic’s claim that we can’t know we are not in a skeptical scenario, while also admitting that we can’t know this with objective certainty.17 Klein accepts the truth of the skeptical premises, and he accepts the inferences to the skeptical conclusion. We seem to know that we have hands, and we should not want to give up this knowledge. We seem to be able to make inferences from things we already know to get new knowledge, so throwing out epistemic closure would give up knowledge too. And we seem to know that we are not objectively certain that we are not in a skeptical scenario, so throwing out epistemic humility would give up knowledge yet again. Klein takes the most difficult possible route to answering the skeptic, navigating his vessel past Scylla and Charybdis toward the “Pyrrhonian ideal of a middle course between dogmatism, on the one hand, and despairing of acquiring [or keeping] knowledge, on the other” (1985, 132). Whether his strategy works is debatable. In Chap. 9, Sharon Ryan argues that Klein’s account could stand to be enriched with more of the epistemic humility characteristic of the Pyrrhonian. She accepts Klein’s refutation of radical skepticism, but argues that in fact we should adopt a moderate form of skepticism, one that remains unrefuted by Klein’s argument against the radical skeptic. Quite a lot of what we know, Ryan tells us, is false, and there is much we have to doubt. Dismissing this doubt would be dishonest; accepting it makes us better epistemic agents. Claudio de Almeida argues in Chap. 10 that Klein’s response to the skeptic has developed, over the years, into an inconsistent combination of Moorean and Russellian arguments that relies on incompatible elements of classical and non-classical logic. De Almeida skillfully leaves the reader eager to revisit Klein’s original work in search of a rebuttal.

1.3

Justification (Chaps. 9, 10, 11, 12, 13, 14, and 15)

According to Peter Klein, reasoning is an indispensable component of justification. Some epistemologists have wanted to draw a sharp line between the justification we find in the recipe for knowledge and the first-person justification that can result from reason-giving. First-person justification, they warn, is untrustworthy. Sometimes the reasons we offer for believing p are mere rationalizations. Sometimes, they point out,

Doxastic justification is defined in the final section of this introduction. Note that if we counted the truth of the skeptical hypothesis among things we know, reflection would not so easily move us to reject it. 16 17

8

B. Fitelson and C. Braden

we fool even ourselves: we think that some set of reasons motivates us to believe, when in fact we are motivated by unconscious beliefs or biases. Entrusting justification to the reasoner is thus thought to compromise its integrity as a condition on knowledge. But Klein thinks that taking justification out of the reasoner’s hands is throwing the baby out with the bathwater. Sure, we are capable of reasoning poorly, and our evidence might be much more limited than we realize. But we are also capable of collecting evidence and using it well, and what good is an epistemology that makes us mere spectators of knowledge and justification? We should not have to sit on the sidelines hoping we know what we think we know. When we score a goal of knowledge-that-p, it is not a mere coincidence that we have done a good job of marshaling our reasons for believing p: we get the ball in the goal because we kick it in the right direction. Infinitism, Klein’s theory of the structure of justification, takes into account the fact that we are participants in justification and are thereby participants in the production of knowledge.18 Justification is largely up to us: we are justified in believing propositions we appropriately infer in accordance with our evidence. Whether the beliefs we infer using our best reasoning succeed at being knowledge is the part that isn’t up to us. When our justified, true beliefs are knowledge, there is a “tether” (a favorite metaphor of Klein’s from the Meno) that “ties belief to the truth and distinguishes knowledge from the mere conjunction of truth with belief” (2010, 156). According to Klein, this tether consists in “reasoning that is sufficiently resistant to defeat” (156).19 But the details of what it means for reasoning to resist defeat, while crucial for understanding knowledge, are irrelevant to our present discussion of justification, because Klein believes that whether we are justified in believing p does not depend on the strength of the tether (or even the presence of a tether) tying our belief to truth.20 A belief can be justified at the same time that its reasoning is defeated. We cannot step outside ourselves to discover and weigh the defeaters of our reasoning; every reason we have is in principle subject to defeat. But justification plays a role in our epistemic lives independently of whether our reasoning is defeated, because in order to be justified, we must attend to our epistemic obligations. In Chap. 14, Matthew Kotzen uses the standard Bayesian framework to systematize what he calls “evidential defeat”. Kotzen is referring to new information that when made available to us leads us to modify our beliefs. This is a not-unusual repurposing of the term “defeat” (which Klein reserves for the externalist part of knowledge) to talk about information that calls us to epistemic duty. Klein distinguishes between two types of justification: doxastic and propositional. We say that a proposition p is justified for a subject S when there is a set of

18

Compare infinitism to foundationalism, according to which some beliefs (foundational ones) arrive with “justificational juice” that pours or is poured into other beliefs. 19 When there is a tether, it can vary in strength: our reasoning for believing p can be more or less resistant to defeat (Klein 1985, 107). 20 Klein believes that “it is a contingent fact that our justified beliefs tend to be true” (2017, 41).

1 Introduction

9

appropriate reasons for p available to S (see 2005, 156–157; 2007a).21 So, for example, the belief that there is not a unicorn playing Liberace’s piano for you right now was propositionally justified for you before you thought about this sentence, and it continues to be propositionally justified for you now. We say that S’s belief that p is doxastically justified when S is acting in an epistemically responsible way by believing p, which is to say that she inferred the belief appropriately from reasons that propositionally justify p for her (2007b, 6; 2017, 39–40).22 Doxastic justification is what Klein calls “knowledge-grade justification” and what Gettier simply calls justification (Klein 2017, 38). When we are doxastically justified in believing p, it means that we got things right on our end and for our belief to be knowledge, the world must also get things right on the external end. Defeaters are ultimately beyond our control. Meeting our epistemic obligations means consciously doing what we can to form justified beliefs and avoid forming unjustified ones. There are epistemic norms Klein thinks we must obey in order to meet these obligations. For example, we must not reason circularly.23 Recall that the tether necessary for knowledge that p ties our belief to truth. That’s a connection between mind and world. In reasoning, on the other hand, we aim to develop an intra-mind connection between our various beliefs. Imagine for a moment that beliefs are balloons. The connection, established through reasoning, between the belief-balloon with content p and its set of supporting belief-balloons, reasons for believing p, is an internal thought-string (quite a different thing from the external tether) tying each of these reasons to the belief that p. Then Klein thinks that we can’t prevent the balloons from flying away by tying them all only to each other—the whole lot would fly away together. Thus he rejects that justification has a coherentist structure. And he thinks that every belief we reflect on is a balloon that will fly away from the mind if not tied down by thought-strings. There is no “foundational balloon” that is somehow already tied down. In other words, Klein rejects that justification has a foundational structure. In Chap. 13, Sanford C. Goldberg adopts an externalist

21

These reasons can be occurrent beliefs or available dispositionally. But a known belief that p relies on a reason that has actually been deployed and subjected to reflection (see Klein 1984, 157; 1999, 300, 322; 2012, 152). 22 Klein does say that “[d]oxastic justification is parasitic on propositional justification” (2007b, 8), but he does not explicitly demand that the reasons necessary for propositional justification be allthings-considered reasons. Suppose you have some memories of seeing blackbirds and some memories of seeing bluebirds. Willfully ignoring your blackbird memories to focus only on your set of bluebird memories does not mean you are justified in believing the proposition “All birds are blue.” But the definition of propositional justification given above would allow that you could be doxastically justified in believing “All birds are blue” on the basis of your bluebird memories if your blackbird memories have understandably slipped your mind. Sometimes we forget. In more recent work, Klein avoids the ambiguity by referring to a single set of reasons R available to S that contains “all and only (1) the propositions that are the content of S’s beliefs and also (2) the propositions that are rational extensions of S’s beliefs” (2017, 47), but p (the proposition R supports) is a rational extension of S’s beliefs, so the newer definition is problematic. 23 The prohibition on reasoning in circles, even very large ones, is something he in most places takes for granted, although he does provide an argument in Klein 2003 (85).

10

B. Fitelson and C. Braden

notion of epistemic responsibility to argue that Klein’s rejection of foundationalism begs the question. But if Klein does succeed in disallowing foundationalist and coherentist structures for justification, there are no beliefs that upon reflection need not be tied to other beliefs, non-circularly, in order to be kept.24 Infinitism is the view that if S believes p then S is justified in believing p only if p is supported by an infinite chain, or regress, of non-repeating reasons each of which could be justified by the next reason down the chain. Peter Klein is, notoriously, infinitism’s primary proponent.25 It is no surprise, then, that three chapters of this book are on that topic. In Chap. 11, Michael Huemer argues that infinitism cannot be the structure of justification, because it cannot be accommodated in the limited capacity or cognition of our finite minds. In Chap. 12, Jeanne Peijnenburg and David Atkinson accept the finite-mind objection but agree with Klein that justification has a chain-like structure. The impossibility of infinite chains is, they argue, no impediment to justification, because there is a point in every chain at which additional justification is probabilistically superfluous.26 In Chap. 15, Matthias Steup rejects the finite-mind objection to infinitism, conceding to Klein that “in principle” we can give infinite reasons justifying our beliefs. But Steup argues that for any given belief the chain of reasons is required to extend exactly four steps, after which additional steps are harmless but add no additional justification. If we take Klein’s account as-is, a question is raised: why aren’t the infinite balloons all flying away at once? The issue is similar to what Klein calls the Ancients’ “no starting point” objection that justification has to start somewhere if it is to be transferred. That objection, he tells us, betrays a pro-foundationalist bias toward thinking of justification as a property of propositions (2007b, 16). But according to Klein, knowledge-grade justification is a property of beliefs, as it is something that can only belong to agents who fulfill their epistemic obligations. It can’t, for example, belong to a sentence written on a page. Some of infinitism’s supposed advantages are that it is not circular, that with it inquiry never ends (suggesting it fits well with how science seems to work), and that it is “not committed to the view that at base our reasons are arbitrary” (Klein 2000b, 203). As for the final advantage, one might wonder whether it is worse to be arbitrary at base or both baseless and arbitrary. But not to worry, because “infinitism, when

24

Klein’s own proposal for reconciling infinitism and foundationalism appears in Klein 2014. The thought-string metaphor, again not to be confused with the tether between belief and truth, should not be taken too seriously, beyond what it is used to illustrate here. For one, Klein does not think of justification as something that is transferred along a chain of beliefs, so these strings are not “conductors”, so to speak, of justification. Furthermore, the beliefs in a chain of justification should not be understood as contiguously connected; reasoning does not cause beliefs on Klein’s view. 25 Klein points out that Paul Moser used the term before him and John Post first characterized the position (1998c, 919n1). 26 Peijnenburg & Atkinson’s chapter (Chap. 12) and Matthew Kotzen’s chapter (Chap. 14) are a testament to the relevance of Klein’s ideas to a generation informed by what we could call the “formal turn” of epistemology in recent decades.

1 Introduction

11

properly understood, can provide a basis for true skepticism—rational belief without assent” (2000a, 12); it entitles us, as much as infinitism allows that we can ever be entitled, which is only provisionally, to have provisional justification for our beliefs as long as we refuse to assent to them. It (provisionally) lets us say I’ll believe p if I can’t help it—after all, anything else would be irrational (2000a, 20–21, 24). Here the foundationalist might object that Klein has confused justifying a belief with justifying keeping the belief, but Klein can counter that the latter is just what justification is: it is responsibly deliberating over which beliefs we must keep and which ones we must throw away if we are to be honest with ourselves. Deliberating over whether we think a belief is true or not is exactly the same thing, as there is no such thing as believing p while believing p is false.

References Gettier, E. L. (1963). Is justified true belief knowledge? Analysis, 23, 121–123. Klein, P. (1971). A proposed definition of propositional knowledge. Journal of Philosophy, 67(16), 471–482. Klein, P. (1976). Knowledge, causality, and defeasibility. Journal of Philosophy, 73, 792–812. Klein, P. (1979). Criticism and comments: Misleading ‘misleading defeaters’. Journal of Philosophy, 76(7), 382–386. Klein, P. (1980). Misleading evidence and the restoration of justification. Philosophical Studies, 37 (1), 81–89. Klein, P. (1981). Certainty: A refutation of scepticism. Minneapolis: University of Minnesota Press. Klein, P. (1982). Reply to professor Odegard. Philosophical Books, 23(4), 198–203. Klein, P. (1983). Review of Keith Lehrer: Profiles. Philosophy and Phenomenological Research, 43 (3), 409–419. Klein, P. (1984). Real knowledge. Synthese, 55(2), 143–164. Klein, P. (1985). The virtues of inconsistency. The Monist, 68(1), 105–135. Klein, P. (1986). Radical interpretation and global scepticism. In E. Lepore (Ed.), Truth and interpretation (pp. 369–386). Oxford: Basil Blackwell. Klein, P. (1987). On behalf of the skeptic. In S. Luper-Foy (Ed.), The possibility of knowledge: Nozick and his critics (pp. 267–281). Totowa: Rowman & Littlefield. Klein, P. (1990). Epistemic compatibilism and canonical beliefs. In M. D. Roth & G. Ross (Eds.), Doubting: Contemporary perspectives on skepticism (pp. 99–119). Dordrecht: Kluwer. Klein, P. (1992). Certainty. In J. Dancy & E. Sosa (Eds.), A companion to epistemology (1st ed., pp. 61–64). Oxford: Basil Blackwell. Klein, P. (1995). Skepticism and closure: Why the evil genius argument fails. Philosophical Topics, 23(1), 213–236. Klein, P. (1996). Warrant, proper function, reliabilism and defeasibility. In J. Kvanvig (Ed.), Warrant and contemporary epistemology (pp. 97–130). Lanham: Rowman & Littlefield. Klein, P. (1998a). The concept of knowledge. In E. Craig (Ed.), The encyclopedia of philosophy (Vol. V, pp. 266–276). Routledge. Klein, P. (1998b). Epistemology. In E. Craig (Ed.), The encyclopedia of philosophy (Vol. III, pp. 362–365). Routledge. Klein, P. (1998c). Foundationalism and the infinite regress of reasons. Philosophy and Phenomenological Research, 58(4), 919–925. Klein, P. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13(Epistemology), 297–325.

12

B. Fitelson and C. Braden

Klein, P. (2000a). The failures of dogmatism and a new pyrrhonism. Acta Analytica, 15(24), 7–24. Klein, P. (2000b). Why not infinitism? In R. Cobb-Stevens (Ed.), Epistemology: Proceedings of the twentieth world congress in philosophy (Vol. V, pp. 199–208). Bowling Green: Philosophy Documentation Center. Klein, P. (2002). Skepticism. In P. K. Moser (Ed.), The Oxford handbook of epistemology (pp. 336–361). Oxford: Oxford University Press. Klein, P. (2003). How a pyrrhonian skeptic might respond to academic skepticism. In S. Luper (Ed.), The skeptics: Contemporary essays (pp. 75–94). Aldershot: Ashgate Press. Klein, P. (2004a). Knowledge is true, non-defeated belief. In S. Luper (Ed.), Essential knowledge: Readings in epistemology (pp. 124–135). New York: Longman. Klein, P. (2004b). There is no good reason to be an academic skeptic. In S. Luper (Ed.), Essential knowledge: Readings in epistemology (pp. 299–309). New York: Longman. Klein, P. (2004c). Closure matters: Academic skepticism and easy knowledge. Philosophical Issues, 14(1), 165–184. Klein, P. (2005). Infinitism’s take on justification, knowledge, certainty and skepticism. Veritas: Perspectives in Contemporary Epistemology (special edition), 50(4), 153–172. Klein, P. (2007a). How to be an infinitist about doxastic justification. Philosophical Studies, 134(1), 25–29. Klein, P. (2007b). Human knowledge and the infinite progress of reasoning. Philosophical Studies, 134(1), 1–17. Klein, P. (2008). Useful false beliefs. In Q. Smith (Ed.), Epistemology: New essays (pp. 25–61). Oxford: Oxford University Press. Klein, P. (2010). Self-profiles: Peter Klein. In J. Dancy, E. Sosa, & M. Steup (Eds.), A companion to epistemology (2nd ed., pp. 156–163). Oxford: Wiley-Blackwell. Klein, P. (2012). What makes knowledge the most highly prized form of true belief. In T. Black & K. Becker (Eds.), The sensitivity principle in epistemology (pp. 152–169). Cambridge: Cambridge University Press. Klein, P. (2014). Reasons, reasoning, and knowledge: A proposed rapprochement between infinitism and foundationalism. In J. Turri & P. D. Klein (Eds.), Ad infinitum: New essays on epistemological infinitism (pp. 105–124). Oxford: Oxford University Press. Klein, P. (2017). The nature of knowledge. In R. Borges, C. de Almeida, & P. D. Klein (Eds.), Explaining knowledge: New essays on the Gettier problem (pp. 35–56). Oxford: Oxford University Press. Sylvan, K., & Sosa, E. (2018). The place of reasons in epistemology. In D. Star (Ed.), The Oxford handbook of reasons and normativity. Oxford: Oxford University Press.

Part I

Knowledge

Chapter 2

Reflective Knowledge Ernest Sosa

Abstract It might be thought that reflective knowledge is of interest only to puzzlesolving intellectuals in their armchairs, that the knowledge of general interest is the “animal” sort, so that reflective knowledge deserves by comparison little more than a footnote. But this overlooks the fact that the difference between animal and reflective knowledge is not a sharp difference in kind. Rather is it a difference of degree. Apart from the intrinsic interest of reflective knowledge, there is a substantial historical reason for distinguishing it: namely, that only thus can we fully understand the power and traditional importance of philosophical skepticism. For the Pyrrhonists it is not enough that one’s sources be in fact reliable, even if one has no notion that this is so. Thus, they would not dignify with the title of knowledge any information acquired through sources that happen to be reliable beyond one’s ken. I do not join in thinking only such reflective “knowledge” worthy of the title. But I do find it, other things equal, an accomplishment on a higher plane than mere unreflective, “animal” knowledge. Enlightened belief requires a perspective on the reliability of one’s cognitive faculties. “Real” knowledge, found at this higher level, requires that one’s belief be reliably produced, but also that one see one’s belief as reliably produced. Keywords Reflective knowledge · Knowledge in degrees · Enhanced knowledge · Knowledge as achievement · Pyrrhonism · Metajustification · Knowledge of reliability · Epistemic competence · Epistemology

Our epistemic example is a special case of the idea that properly guided activity is admirable, that the causal basis of one’s conduct matters to its evaluation. Apt conduct that attains success under the guidance of competence is to that extent admirable, as is the sequence of movements that constitute a dance by a dancer in control, by contrast with a sequence that derives from helter-skelter movements by someone who produces an identical sequence of movements just by luck, through no E. Sosa (*) Department of Philosophy, Rutgers University, New Brunswick, NJ, USA © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_2

15

16

E. Sosa

guidance aimed at beauty or elegance or unified, interesting, and coherent flow. Almost anything that involves guidance towards some dance desideratum would surpass in interest and worth just a helter-skelter, accidentally brought together sequence that turns out to be identical geographically. That gives a generally applicable reason why knowledge is enhanced by being not just animal but reflective. And that helps explain why its attainability is worth defending against the traditional skeptics with their Agrippan and other Pyrrhonian doubts. Through the dinner conversations that have sustained our friendship over many years, Peter Klein and I have argued whether any such knowledge is in the cards, including dialectic over what it would be to have it. This piece continues our conversation.

2.1

Enhancing Knowledge

An immigrant—recently arrived from a primitive culture—who has never seen a car, may eventually trust someone who tells him about fuel gauges. And this may enable justified belief that his tank is full, based on a fuel gauge reading. Justification being a matter of degree, suppose the subject to be just barely justified in so believing, and barely to know the truth of what he believes. Let’s say he does know it, but not as well as do those familiar with that instrument and others like it. The owner of that car, for example, will know very well what the immigrant barely knows. Given his broader and better basis, the owner is likely to be more confident, and appropriately so, which fits the fact that a belief with that enhanced basis is so much more likely to be right. Accordingly, we do well to exploit the flexibility offered by English locutions like “You know as well as I,” or “As you know very well,” etc. We can thus compare the epistemic situations of the two believers, saying that although both know the tank to be full, the owner knows it better. Among the factors that bear on the epistemic quality of a belief is its rational basis (how good are the reasons on which the belief is based?). In particular, as our immigrant improves his knowledge about instruments, and more specifically about fuel gauges, and about that fuel gauge in particular, so improves the epistemic quality of his trust in the gauge’s deliverances. And this generalizes: any belief’s epistemic standing can be boosted through the believer’s enhanced knowledge of its sources. How then does one get to know about the sources of one’s own beliefs? It is not through any direct “bootstrapping,” not when it is so obviously vicious to collect data from that very fuel gauge, with no independent means to verify their truth, and to conclude inductively that the gauge is reliable. Three stages are involved in such direct bootstrapping: first, to accept the deliverances of a certain source; second, to attribute those deliverances to that source; and, third, to conclude by simple, enumerative induction, based exclusively on that data base, that the source is reliable.

2 Reflective Knowledge

17

Such reasoning is vicious when it is used in an attempt to determine whether a fuel gauge is reliable. One cannot escape initial doubt about the reliability of a source by direct bootstrapping. That cannot be a way to acquire all of one’s justification for believing the gauge to be reliable. For, at some level one must have believed it already, given the trust required for accepting the gauge’s deliverances in the first place. That trust can’t have been entirely unjustified if it was to yield justified belief of its delivered data. This applies both to external instruments and to our natural basic sources: to our perception, for example, in any of its modalities, to our memory, and to any other such source. Fortunately, one can reason circularly without falling into deplorably direct bootstrapping. Any empirical study of our memory and its reliability will depend on our memory. Any empirical study of our perception will depend on our perception. Et cetera. Think of psychology, the social sciences, the history of science. Individually or in concert, these might reveal how reliable are various faculties and modes of inquiry, including the most fundamental. Can we seriously countenance that any such study is doomed from the outset? Can we philosophers dismiss such studies as viciously circular in one fell swoop? That seems ludicrous. The possibility remains open, therefore, that with better knowledge of our sources, including the most fundamental sources, we can enhance the epistemic quality of our beliefs. How so? In just the way in which we come to know better by trusting the deliverances of our senses and memory, with increasing knowledge of their reliability, when we broaden our basis for such trust. Take, for example, two parallel line segments with the familiar inverted arrows, and suppose one is perceptibly a bit longer than the other, just barely enough to allow a normal perceiver to have knowledge of that fact. Someone ignorant of the Müller-Lyer illusion might possibly know those line segments to be incongruent, but he will not know it as well as someone who is in the know about the illusion. When thus ignorant, one may fail to believe through competence, since one trusts a mechanism too liable to mislead. Those in the know are better placed, since their mechanism is sensitive to the misleading effects of the illusion. With increasing knowledge of what affects the reliability of our ostensible perceptions or memories, comes epistemically improved acceptance of the relevant deliverances of perception or memory.

2.2

Praxis

Knowledge based on such a reflective perspective may for convenience be distinguished as “reflective” knowledge. This is not to suggest that ‘knowledge’ is ambiguous, meaning sometimes “reflective knowledge,” sometimes “unreflective” or “animal” knowledge. Rather, we just distinguish two varieties of knowledge—the reflective, and the animal—without supposing this to be mirrored by any ambiguity in the relevant words of English. We do distinguish indefinitely many varieties of knowledge in that sort of way. There is knowledge distinguished by subject matter, for example, and so on to quite

18

E. Sosa

trivial classifications. Such varieties of knowledge vary in epistemological significance. Far less trivial would be the a priori / a posteriori distinction, or the idea of perceptual knowledge, or introspective knowledge, or self-knowledge, and so on. Whatever may be the prospects for such interesting varieties of knowledge as the empirical, the scientific, and the rest, these have traditionally seemed important to distinguish and understand as distinctive varieties of knowledge. Why think that reflective knowledge belongs with the epistemologically interesting varieties, and not with the epistemologically insignificant? Apart from the intrinsic interest of reflective knowledge, there is a substantial historical reason for distinguishing it: namely, that only thus can we fully understand the power and traditional importance of philosophical skepticism. What best accounts for that—what we find in a tradition that extends from before the Pyrrhonists through the moderns, especially Descartes, and beyond—requires a conception of reflective knowledge, or so I would argue. It is such knowledge that is mainly threatened by the Pyrrhonian problematic, with its Agrippan trilemma. The reason why we seem in danger of being denied any knowledge on the reflective level is that any attempt to attain it seems to founder on viciously circular reasoning. Here lies the main source of Pyrrhonian skepticism, for example, in a tradition renewed powerfully by Descartes. It might be thought that reflective knowledge is of interest only to puzzle-solving intellectuals in their armchairs, that the knowledge of general interest is the “animal” sort, so that reflective knowledge deserves by comparison little more than a footnote. But this overlooks the fact that the difference between animal and reflective knowledge is not a sharp difference in kind. Rather is it a difference of degree. Already in early infancy, the knowledge of humans is essentially reflective to some important degree, not purely animal. Let us see how and why this is so.

2.3

Rational Self-Correction

Some perceptual faculties are best considered not belief-forming mechanisms, but modules that yield, rather, inclinations to believe, along with intellectual attractions or seemings. When we face Müller-Lyer lines, for example, we are inclined, or attracted, to believe the lines incongruent, but we form no such belief. Our background information in fact leads us to the opposite belief—that despite seeming incongruent, the lines are really congruent. Eventually no deliberation or pondering is required: we come to believe the truth immediately on sight. We are then able to jump to our conclusions. They are indeed no longer conclusions but rather beliefs prompted so strongly and swiftly by the sensory input that the perceptual attraction is outpulled immediately by the relevant background information. Having earlier been prompted to believe the lines incongruent, we are now prompted to think them congruent. We thus replace a commitment to a

2 Reflective Knowledge

19

certain prompting pattern with a commitment rather to an opposing pattern. While still attracted to think the lines incongruent, we are no longer prompted to believe without delay. How does this happen? It seems no great mystery. We find through experience that the earlier pattern is incorrect, that it is unreliable or even counter-reliable. That is why we replace it with the new, more reliable pattern. It is thus awareness of unreliability, and belief about improved reliability, that rationally install the new pattern. Such cognitive procedure amounts to a kind of perspectival reasoning. An earlier commitment to a “reasoning” pattern, to a pattern whereby we based belief in the incongruence of the seen lines, is replaced by a different, and indeed an opposing, commitment. What we do is rationally motivated, not arbitrary. In rejecting the earlier commitment and adopting the new, we are motivated by reasons. And somehow, at some level, we must have some awareness of these reasons, if they are relevantly to “motivate” us, if we are to act for those reasons. Which reasons? Reasons that involve the unreliability of the earlier pattern and the enhanced reliability of the new. Such self-correction is essential to the rational animal. An inflexible automaton unable to learn from experience thus theoretically (and, in a similar way, practically, so as to improve not only in knowledge but also in wisdom) would hardly count as rational. We must distinguish here the deliverances of modules from beliefs formed on their basis. A rational belief is never just an inflexible reflection of some module’s deliverance. This is one reason why beliefs that flow inflexibly from a brain lesion or from a clairvoyance module would not be adequately rational, good candidates for knowledge. A rational belief will always be better integrated than that. It will reflect not just the deliverances of an isolated module, but also anything else in the subject’s relevant background information. Crucial here will be whatever the subject knows about the reliability of the module, about its reliability in the pertinent circumstances. And now we are already at the reflective level, where vicious circularity threatens our perspective on the reliability of our delivering modules no less than it threatens our perspective on the reliability of the deliberation or pondering that goes beyond the deliverances of those modules. Such responsiveness to reasons seems also essential to agency, properly so-called. Automata do not act except in the very thin sense in which a thermostat acts when it triggers a furnace. Intentional action is reason-involving. There are reasons why agents act, and reasons for which they act, motivating reasons. This is what requires a level of sophistication and complexity that includes the ability selfcorrectively to learn from experience, theoretically and practically, with changed weights to various delivery modules, and changed commitments in one’s forming of beliefs and intentions, in judgments and in choices. The distinction between animal knowledge and reflective knowledge is hence a matter of degree. If the animal’s beliefs constitute knowledge, while dependent on his ability to learn from experience, it will be a knowledge to some degree reflective, since to some degree based on an affirming epistemic perspective about its own

20

E. Sosa

delivery systems and modes of forming beliefs. And it will then constitute better reflective knowledge the richer the affirming perspective.1 That is one way in which agency involves enlightened reflection. We may more clearly see how through a striking Pyrrhonian metaphor: Let us imagine that some people are looking for gold in a dark room full of treasures . . . [N]one of them will be persuaded that he has hit upon gold even if he has in fact hit upon it. In the same way, the crowd of philosophers has come into the world, as into a vast house, in search of truth. But it is reasonable that the man who grasps the truth should doubt whether he has been successful. (Sextus, 27 M VII 52)

Compare the good fortune of those who strike it rich in the dark with the enlightened discovery of those whose deliberate search for gold, guided by good eyesight in clear light, is rewarded with success. Enlightened discovery is creditable to the agent; luck in the dark is not. Enlightened belief requires a perspective on the reliability of one’s cognitive faculties. “Real” knowledge, found at this higher level, requires that one’s belief be reliably produced, but also that one see one’s belief as reliably produced. When one gathers information under reflective guidance, this is, on the Pyrrhonian account, like acquiring gold in the light of day. Unreflectively acquiring that information is, by contrast, like grasping gold in the dark. In each case the former is more admirably attributable to the agent as his own doing. For the Pyrrhonists it is not enough that one’s sources be in fact reliable, even if one has no notion that this is so. Thus, they would not dignify with the title of knowledge any information acquired through sources that happen to be reliable beyond one’s ken. Only sufficiently enlightened “knowledge” is supposed worthy of the title.2

2.4

Consequences for Belief

I do not join in thinking only such reflective “knowledge” worthy of the title. But I do find it, other things equal, an accomplishment on a higher plane than mere unreflective, “animal” knowledge. The privileging of reflective knowledge is said to doom us to vicious regress or else arbitrariness, an epistemic disaster either way. But this overlooks the fact that the quality of one’s knowledge is a matter of degree along various dimensions. One dimension is that of reflective defensibility—of how defensible is one’s belief under reflective questioning. Even if the pursuit of such questioning to higher and higher 1 But it is crucial to recognize that the affirming perspective need not be explicitly conscious, at least not wholly. It can be largely constituted rather by implicit acceptance that one’s faculties are reliable enough, an acceptance manifest in one’s continuing use of those faculties, and one modifiable through reasons to doubt their reliability. 2 It is of course assumed in the example that one does not have some other way to distinguish the gold, say through its smoothness or weight, which would enable enlightened acquisition even in the dark.

2 Reflective Knowledge

21

levels must inevitably peter out, this does not imply that its successful pursuit is therefore worthless at any level.3 It is, moreover, clearly of interest in epistemology whether such knowledge, reflective knowledge, is necessarily beyond our reach. It is knowledge of this kind that is threatened by epistemic circularity. The knowledge more characteristic of the lower animals, unreflective knowledge, requires no endorsing perspective, and is for that reason evidently safe from the threat of circularity. Epistemic circularity would not have figured as a threat taken so seriously except by those interested in knowledge that is reflective, not just animal. Of course, it remains a threat in any case to knowledge that one knows, but the threat looms larger when we see the effect that being blocked from such ascent would have on our knowledge generally, and not just on our knowledge that we know. I would like to avoid quibbling over words. My focus is therefore not so much on whether the English word ‘knowledge’ is ambiguous. My focus is rather on a state beyond the words, the state of believing and its epistemic evaluation.4 A belief seems evaluable along two (at least two) distinguishable dimensions: an unreflective dimension involving its “safety” and the truth-reliability of its sources, and a second dimension involving its rational defensibility in the arena of reflection, where, for example, either its safety or the reliability of its sources is put in question. Once such defensibility is seen to affect the epistemic quality of a belief itself and not just the quality of the relevant beliefs about that belief, it should be clear that vicious circularity threatens not just our knowledge that we know, but the quality of our knowledge generally.

2.5

Responding Appropriately

Compare now the rational defensibility of a belief when the question arises of how well justified it is epistemically. It is better epistemically for a belief to be rationally defensible than not, better along that dimension, and to that extent. That seems trivial, and holds good whether the belief is someone else’s or one’s own. Someone else’s belief attains no enhanced epistemic status, however, in virtue of one’s acquiring the information that it is defensible, indeed defensible in a specific way. Does that change when the belief is one’s own? Suppose one knows a belief of one’s own to be defensible in a specific way. One knows it to have a certain source, for example, which one knows to be highly reliable. Does this add to the epistemic standing of that belief? If one’s belief is 3

One might even entertain the hypothesis that returns from such pursuit will diminish rapidly with ascent beyond a level or two, so that the additional epistemic value of reflection beyond initial levels may even just approach a limit asymptotically. 4 This focus on the state beyond the words comports with a view of philosophy as largely valuebased, a study of commodities that we wish not only to understand but also to possess (see Sosa 2000).

22

E. Sosa

now sustained at least in part by its being a deliverance of that source, which one knows independently to be reliable, one then bases one’s belief on independently supported premises. And this is paradigmatically a way to enhance the epistemic standing of a belief. Plausibly, then, it is epistemically better, ceteris paribus, to hold a belief based on such a basis than not so based. Once having placed that belief in the arena of reflection, moreover, by asking whether its source is so much as minimally reliable, you are better off if able to answer yes, better than if you must answer either no or maybe. The latter two answers, if consciously and sincerely given, would import epistemic incoherence. The more coherent answer is that the source is at least minimally reliable. Consciously thinking that source reliable, moreover, might help enhance your coherence. Assurance that the source is highly reliable opens up two possibilities: either that the assurance has no supportive bearing on the belief itself, or else that it does. Whether it has such bearing matters to the epistemic coherence of your mind. If you see clearly the high reliability of what you consider the source of your belief, but this leaves you cold, with no effect on your holding it, you suffer a loss of epistemic coherence. Your knowledge of the high reliability of your source requires an appropriate response on your part. You now have a further potential basis that you are required to take into account. You fall short epistemically if you just ignore this newfound reason. Sources deliver deliverances. How one receives such deliverances, how confident one is of their truth, should comport with one’s view of the source’s reliability. When one learns a source to be significantly more reliable than previously appreciated, or significantly less, one must adjust one’s credence on that basis, on pain of rational incoherence. Our epistemic example is a special case of the idea that properly guided activity is admirable, that the causal basis of one’s conduct matters to its evaluation. Apt conduct that attains success under the guidance of competence is to that extent admirable, as is the sequence of movements that constitute a dance by a dancer in control, by contrast with a sequence that derives from helter-skelter movements by someone who produces an identical sequence of movements just by luck, through no guidance aimed at beauty or elegance or unified, interesting, and coherent flow. Almost anything that involves guidance towards some dance desideratum would surpass in interest and worth just a helter-skelter, accidentally brought together sequence that turns out to be identical geographically. That gives a generally applicable reason why knowledge is enhanced by being not just animal but reflective. And that helps explain why its attainability is worth defending against the traditional skeptics with their Agrippan and other Pyrrhonian doubts.

2 Reflective Knowledge

23

References Sextus Empiricus. (1990). Sextus empiricus: Against the logicians. Translation by Jonathan Barnes, in The Toils of Scepticism, translated by Jonathan Barnes, 138–139. Cambridge University Press. Sosa, E. (2000). Skepticism and contextualism. Philosophical Issues X, Supplement to Noûs, 10, 1–1.

Chapter 3

Epistemically Useful Falsehoods Catherine Z. Elgin

Abstract In “Useful False Beliefs,” Peter Klein argues that the justification required for knowledge can contain a false belief essentially. When this happens, the agent arrives at her conclusion via a chain of inference that includes a false belief. He illustrates his argument with cases that depend on apparent memory, testimony, recorded empirical evidence, and observation-based calculation. If the agent’s inferential path is close enough to a route that contains only truths, Klein maintains, her conclusion is justified. Still, he intimates, reliance on a falsehood is an epistemic defect, even if not a fatal one. All things considered, it is preferable to take the epistemic high road and rely only on truths. I will argue, however, that sometimes inferring via a falsehood is an epistemic strength. Reliance on a falsehood enables the agent to marshal epistemic resources by excluding from her reasoning irrelevant complications that would blind her to the relation between her evidence and her conclusion. Keywords Useful falsehoods · Inference chains · Knowledge from falsehoods · Counter-closure · Defeasibility · Hyperintensionality · Falsity-makers · Klein · Epistemology

Pummeled by a relentless barrage of Gettier cases, epistemologists tend to conclude that false beliefs never figure essentially in the justification of a bit of knowledge. Peter Klein demurs. He agrees that Gettier sequences do not result in knowledge. And he agrees that the reason they do not is that a false belief plays an ineliminable role in the agent’s inference. But, he urges, not all inferential paths that involve false beliefs essentially lead to Gettier cases. There are, he maintains, epistemically useful

I am grateful to Samuel Elgin, Robert Shope, and Cherie Braden for useful discussions of the issues raised in this chapter. C. Z. Elgin (*) Harvard University, Cambridge, MA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_3

25

26

C. Z. Elgin

falsehoods; they figure essentially in the justification of bits of knowledge. A chain of inference contains a falsehood essentially (or ineliminably) just in case the inference would not be valid if the false statement were eliminated. Klein develops an account of how and when a path involving a falsehood yields knowledge. It is, in many respects, a remarkably open-minded account. He takes pains to show that it is compatible with foundationalism and coherentism, as well as, of course, infinitism. This is not one of his more radical theories. Still, there are elements of it that give me pause. So after explicating the theory and identifying those elements, I will sketch an alternative way of arriving at the same conclusion. My goal is not to show that he is wrong or that my approach is better. It is to suggest that a variety of disparate epistemological theories will and should have a place for epistemically useful falsehoods. In short, you can’t get out of this by simply objecting to the details of Klein’s position. Here I discuss the position Klein develops in “Useful False Beliefs” (2008). My goal is not to ask how the position meshes with the rest of his epistemology.1 It is not even to applaud or object to the fine-grained details of Klein’s proposal. It is, rather, to ask whether he has supplied good reason to think falsehoods can ever figure ineliminably in the justification needed for inferential empirical knowledge. I think he has. Klein provides several cases. For expository purposes, I’ve modified them slightly, while preserving what I take to be the epistemological features that are relevant to my discussion. APPOINTMENT: On the basis of an apparent memory, Paul believes that his secretary told him on Friday that he has an appointment on Monday. From that belief, he infers that he has an appointment on Monday, which in fact he does. Although his secretary told him about the appointment, she told him on Thursday, not on Friday. Nevertheless, it seems, Paul knows that he has an appointment on Monday. SANTA CLAUS: On Christmas Eve, Virginia’s parents tell her that Santa will put presents under the tree that night. She believes her parents and infers that there will be presents under the tree on Christmas morning. Klein contends that Virginia knows it. AVERAGE RAINFALL: Weatherman believes that the average annual precipitation in northwest Montana is about 13 inches, because he believes that accurate records have been kept for over eighty years and during that period the average comes to 13 inches. In fact, the average rainfall is about 13 inches, but accurate records were kept for only seventy-nine years. Still, it seems, he knows the average rainfall. PTOLEMAIC ASTRONOMER: On September 2, 1203, a Ptolemaic astronomer uses the deferent and epicycle orbits of the Earth, Mars, and the Sun and what he takes to be their (then) current positions, to predict that, ceteris paribus, Mars will be 1

In particular, I am not concerned here to defend or challenge Klein’s defeasibility analysis of knowledge (see Klein 1976, 1981). Whether or not a defeasibility analysis is correct, it is worth asking whether inferences that lead to empirical knowledge can contain falsehoods essentially.

3 Epistemically Useful Falsehoods

27

visible from Earth on September 2, 2003. (The ‘ceteris paribus’ controls for the possibility of clouds and the premature annihilation of the relevant celestial objects, but not for the possibility that Ptolemaic astronomy is incorrect.) If the assigned orbits and then-current relative positions of the three bodies allow for sufficiently accurate extrapolations, Klein maintains, the astronomer knows that, ceteris paribus, Mars will be visible on September 2, 2003 (Klein 2008, 37–38). These cases—call them Klein cases—cover a range of grounds for knowledge. APPOINTMENT rests on (apparent) memory; SANTA CLAUS, on testimony; AVERAGE RAINFALL, on recorded empirical information; PTOLEMAIC ASTRONOMER, on observation and calculation.2 The range suggests that if Klein is right, epistemic reliance on useful falsehoods is not a local or parochial matter. We should expect to find them in any realm where empirical knowledge stems from inference. We don’t have to accept all of Klein’s cases. Some may seem more plausible than others. But if any strike us as cases of knowledge, the conviction that ineliminable falsehoods in the inferential path are anathema to empirical knowledge needs to be reconsidered. Klein, of course, does not rest his conclusion on the plausibility of these examples. Their role is to illustrate the phenomenon he is interested in and to elicit intuitions. They should, he hopes, suggest that despite the plethora of Gettier cases, it is not so obvious that false beliefs in epistemically ambitious empirical inferences are inevitably pernicious. Some, if not all, of the cases are apt to strike us as instances of knowledge gleaned from inferences that essentially involve false beliefs.

3.1

Klein’s Account

Assuming that propositions are the contents of beliefs (Klein 2008, 27), Klein distinguishes between two sorts of justification. Propositional justification is what justifies the content of a belief—the proposition that occurs in the that-clause of a belief ascription. Doxastic justification is what justifies the agent in believing that content. A proposition h is propositionally justified for S just in case S has an epistemically adequate basis for h (28). In the cases that concern us, the epistemically adequate basis is either another proposition that S believes or a proposition that is available to S, where to be available to S is either to be one of the contents of S’s mental states or to be appropriately connected to S’s mental states. Ideally, one would like to hear more about what it takes to be appropriately connected. For our purposes, however, it suffices that entailment is an appropriate connection. If the content of S’s belief that g entails the content of her belief that h, then g affords her an epistemically adequate basis for h.

2

In Klein’s presentation, the Ptolemaic Astronomer has his students make the calculation, based on what they have been taught. If teaching is testimony, this is an instance of useful falsehoods’ figuring in knowledge based on testimony as well. I simplified the case. Placing the inferential burden on the students is, for my purposes, an unnecessary epicycle.

28

C. Z. Elgin

Propositional justification is keyed to the epistemic agent, in that the resources it draws on are resources available to her. So considerations that are inaccessible to S cannot propositionally justify h for her, even if in an absolute sense they justify h. Thus, for example, prior to the development of calculus, a conclusion that required solving a differential equation was not justified for anyone, even though it was, in an absolute, mathematical sense, propositionally justified. But a consideration that is accessible to an agent need not actually be accessed by her. An important consequence is that h can be propositionally justified for S without S’s being aware that it is. She might fail to draw the needed inference. Indeed, she might be incapable of drawing the inference. Maybe it requires an extended but straightforward inference that she will not live long enough to carry out. Nevertheless, if h is entailed by S’s belief that g, her belief that h is propositionally justified for S. Doxastic justification pertains to beliefs themselves, not (or anyway not directly) to their contents. According to Klein, a belief is doxastically justified when it has an appropriate causal pedigree (2008, 28). How S came to harbor the belief that h or what considerations now sustain the belief that h, determines whether that belief is doxastically justified. Doxastic justification is the sort of justification that is at issue in determining, when S holds a belief, whether S is justified in holding that belief. It is the justification needed for justified true belief. Propositional justification and doxastic justification are mutually independent. A subject is doxastically justified in believing that g if she came to that belief in the right way. But a belief whose causal pedigree is impeccable can still fail to be propositionally justified because there is information available to her that renders the belief content unjustified. Moreover, a proposition that she does not believe can be propositionally justified for her. On the flip side, she might have a propositionally justified belief j that is in fact appropriately supportable by other justified beliefs. But if she arrived at it by reading tea leaves and never even attempted to infer it from her other justified beliefs or from her other evidence, she lacks doxastic justification for j. In Gettier cases as in Klein cases, a false belief plays an essential role in arriving at a true conclusion. S’s belief that someone in the class owns a Ford is caused by her belief that Nogot, who is in the class, owns a Ford. Paul’s belief that he has an appointment on Monday is caused by his belief that his secretary told him on Friday that he has an appointment on Monday. Virginia’s belief that there will be presents under the tree is caused by her belief that Santa Claus will put presents under the tree. Weatherman’s belief that the average rainfall is about 13 inches is caused by his belief that the average over the more than 80-year record of annual rainfalls is 13 inches. The astronomer’s belief that Mars will be visible on September 2, 2003 is caused by his belief that celestial objects orbit and will continue to orbit the Earth in accordance with the laws of Ptolemaic astronomy. We might conclude then that Klein cases just are Gettier cases. If so, then since Gettier cases fail to yield knowledge, the intuition that Klein cases yield genuine knowledge is wrong. This does not seem plausible. The difference between standard Gettier cases and Klein cases is that in Klein cases the false beliefs seem to be false in ways that are irrelevant to their epistemic role. Thursday? Friday? Who cares when Paul’s secretary told him about the appointment? What matters is that she told him. Seventy-nine

3 Epistemically Useful Falsehoods

29

years? Eighty years? Who cares exactly how many years they have been keeping rainfall records, given that they have been doing so long enough to establish a reliable record. And so forth. The issue is how to capture this difference. Aside: even if we want to be hard-nosed and insist that Klein cases are Gettier cases—that they yield true belief but not knowledge—there is something epistemologically interesting going on. Unlike standard Gettier cases, they certainly seem to yield knowledge. However we accommodate them, we should acknowledge the fact that they seem quite different from standard Gettier cases. Klein calls the false beliefs that figure in Klein cases ‘useful falsehoods’. Rather than leading us astray, they point us in the right direction. He suggests that: 1. The belief that uf is a useful falsehood to S (for acquiring knowledge that h) by producing a doxastically justified belief that h iff: 2. uf is false. 3. The belief that uf is doxastically justified for S. (That is, S believes uf and uf has an appropriate causal pedigree.) 4. The belief that uf is essential in the causal production of the belief that h. (Without it, there would be no causal chain leading to S’s belief that h). 5. uf propositionally justifies h. 6. uf entails a true proposition, t. 7. t propositionally justifies h. 8. Whatever doxastically justifies the belief that uf for S also propositionally justifies t for S. (Klein 2008, 49) Klein’s account depends on the fact that a subject’s doxastic justification and her propositional justification are distinct. In the cases of interest, her doxastic justification for h is flawed because she arrives at a truth via a false belief. That false belief would ordinarily be deemed a defeater. But in cases where the falsehoods are useful, Klein suggests that they have a distinctive property: each such falsehood entails a relevant truth that is propositionally justified. ‘Paul’s secretary told him on Friday that he has an appointment on Monday’ entails ‘Paul’s secretary told him he has an appointment on Monday.’ ‘Santa Claus will put presents under the tree’ entails ‘Someone will put presents under the tree.’ The subject need not believe or even entertain the truth in question. What is at issue is propositional justification, not doxastic justification. But because, owing to the entailment, the road not taken is close enough to the inferential route actually traversed, the subject knows.3 It is important that the road not taken is not utterly alien to S. Because true proposition t is propositionally justified for S, it consists of considerations that are available to her even though she did not actually draw on them in coming to her belief. Whether or not this shows that useful falsehoods contribute to the production of knowledge, it at least yields a plausible explanation of why Klein cases are apt to strike us as cases of knowledge.

3

Or is in a position to know. We are here talking only about the justification condition on knowledge.

30

C. Z. Elgin

The mere existence of such a road not taken might seem inadequate to secure knowledge. If S does not believe that t, the fact that t is entailed by something she believes may seem too flimsy a basis for claiming that uf provides knowledge. We might seek to shore it up. Perhaps we can connect uf and t by insisting that because S believes uf, she believes t. Then we could say that t is the real link in the chain of doxastic justification. This is surely false. t might never cross her mind. In that case, it is no part of the doxastic chain. And, it is worth noting, this would make uf merely a harmless falsehood, not a useful one. For in that case, uf could be eliminated without epistemic loss. Perhaps we should say, ‘Because S believes uf, S would believe t, if she were to entertain it.’ As a general matter, this too is false. Were the Ptolemaic astronomer to entertain the idea that, for purposes of prediction, it is just as if Ptolemaic astronomy was true, he might well balk. As Klein notes, he could be incapable of ‘just as if’ reasoning. Alternatively, the astronomer might resist using it in this case. He might think that if Ptolemaic astronomy is false, there is no reason whatsoever to think that planetary positions will be just as they would be if the theory were true. Virginia could balk as well. Given the number of days on which her parents have failed to shower her with gifts, she might well think, ‘It’s either Santa Claus or nothing!’ It is then implausible that S has another belief that is the real link in the chain of doxastic justification. Klein concedes that the relation between uf and t might be looser than entailment. Perhaps rendering t highly plausible would suffice. We might also worry that entailment is too generous. If t follows from uf by an obscure and convoluted (but logically impeccable) route, we might doubt that t supplies the needed backup. I’m not going to address these concerns. If need be, we could restrict the requirement to obvious entailments and expand it to include other suitably strong relations of support. Here I am interested in seeing whether knowledge can be underwritten by useful falsehoods in anything like the way that Klein suggests. If it can, we can worry about how or whether we need to tweak things.

3.2

Concerns

Klein’s general account of what it is to be a useful falsehood is plausible if his assumptions are sound. He has argued for them elsewhere, and they are not unduly radical. Still, I have qualms about the underlying metaphysics. I will raise them and go on to suggest an alternative route to the same conclusion. My goal is to suggest that there is something deeply right about Klein’s conclusion, something that cannot be easily evaded. Klein (2008) maintains that doxastic justification depends on the causal history of a belief. I have a couple of worries about this. The first is this: A subject might come to the true belief that p unjustifiably. She overlooked contravening evidence, believed an unreliable informant, engaged in sloppy reasoning, or jumped to conclusions. Suppose, however, that, having formed the belief, she continues to amass

3 Epistemically Useful Falsehoods

31

evidence. The additional evidence enables her to hold that p for good and sufficient reasons. The belief’s disreputable causal ancestry is then swamped by trustworthy supports. Tabloid readers came to believe that O. J. Simpson abused his wife because it was so reported in the National Enquirer, a notoriously unreliable scandal sheet. Subsequently, more reputable news sources uncovered strong evidence of his abuse. The credulous tabloid readers should not be permanently barred from being doxastically justified simply because they originally formed their belief on the basis of an unreliable source, given that they now take trustworthy news reports as the support for their conviction. This sort of objection has been raised against causal theories in the past. The obvious and sufficient response is to recognize not just generating causes but also sustaining causes. Then the tabloid readers who now base their beliefs on reputable news sources are in a position to know. There is, however, more to worry about than occasional epistemic sloppiness. According to psychology, the etiology of many beliefs is not particularly estimable. In forming beliefs, we overlook, over- or under-emphasize, cut corners, and distort what we find in order to bring new evidence into accord with our presuppositions. We are guided (even blinded) by stereotypes. We are prey to wishful thinking, selfdeception, and confirmation bias. We seem to go out of our way to ignore base rates. We are frequently in denial. We suffer from implicit bias, stereotype threat, and all manner of repressions and distortions. And that’s if we are normal, mentally healthy cognizers! If doxastic justification turns on how we actually come to our beliefs, much of what we take to be knowledge falls short. Despite what you may think of it, your conviction that Bernie Sanders is more liberal than Hillary Clinton is not caused by your appreciation of their voting records. It is largely due to unresolved issues from your childhood. Your grandfather always brought you presents; your grandmother told you to sit up straight. Your non-custodial father was generous; your mother, with inadequate child support, counted pennies. This is not to deny that we reason inferentially. But one person’s modus ponens is another’s modus tollens. If S believes that p (and wants to be consistent), her recognition that p entails q presents her with an option. She can either endorse q or repudiate p. What factors she considers salient, how she weighs the evidence, how she assesses plausibility, and a host of other matters, some of them epistemically quite dubious, influence her choice. The cases of heuristics and biases that have dominated the literature are of course cases in which, on account of these shortcomings, the agent was not justified in the belief she formed (Gilovich et al. 2002). But, psychologists insist, many of the heuristics and biases are valuable. Although they occasionally engender false beliefs, they often lead quite efficiently to true ones. I’m not going to argue that we should just validate them and consign the failures to the realm of epistemic misfortune. Even if we are tolerant of a measure of epistemic luck, that would be going too far. Rather than depending exclusively on the sequence of considerations that led to the adoption of a belief, or on the full network of considerations that sustain it, I suggest, doxastic justification depends on those elements of the network that endow the belief with credibility (see Elgin 2017). Which elements these are is a normative, and not merely a causal matter.

32

C. Z. Elgin

Recently, philosophers have urged that causation is not the only sort of metaphysical dependence (Rosen 2010). Some relations are grounded in others; some states of affairs obtain in virtue of others; some facts hold because others do. If this is so, it opens the door to the idea that a belief’s doxastic justification might turn on its standing in an appropriate dependence relation, which need not be (but may be) causal. I am not going to develop this idea here. Much needs to be worked out before we have a clear sense of what the non-causal dependence relations are and of what functions they perform. My point is simply this: Doxastic justification is the agent’s justification for believing. Klein needs there to be justification for believing (and not just for the proposition believed). He may have concluded too quickly that the cause of the belief is the only thing that could provide that justification. I suggest then that we should say that doxastic justification is that which justifies an agent in holding a belief, and for now remain neutral about whether it consists of causes, grounds, or collateral support of a different kind. My other metaphysical worry concerns Klein’s contention that the contents of beliefs are propositions.4 Standardly, propositions are held to be that which truth-apt sentences express. And two sentences express the same proposition just in case they are co-intensional. Moreover, it is held, the co-intensionality condition is sometimes satisfied; some pairs of sentences are co-intensional. If ‘vixen’ and ‘female fox’ are synonymous, then ‘Vixens are fierce’ and ‘Female foxes are fierce’ are co-intensional. They express the same proposition. Any world in which vixens are fierce is one where female foxes are fierce; any world in which female foxes are fierce is one where vixens are fierce. But the human capacity for ignorance and error is legion. Even if ‘vixen’ and ‘female fox’ are synonymous, it is entirely possible for Harry to believe that vixens are fierce without believing that female foxes are fierce. Even if ‘London is beautiful’ and ‘Londres est belle’ express the same proposition, Pierre can believe one but not the other (Kripke 2011). Propositions, as standardly characterized, are not sufficiently fine-grained to reflect belief contents. Such contents are evidently hyperintensional. Some, such as Bealer (1998), reject the co-intensionality criterion and maintain that propositions are hyperintensional too. But the criterion is so deeply entrenched that it is preferable, I think, to introduce new terminology. Let us call the content of a belief its hyperintension, and say that two beliefs share a content just in case they are co-hyperintensional. It seems plausible that the beliefs of two agents could share a content. The agents might both believe that vixens are fierce. What is unclear is

“Some might prefer to limit the scope of propositions to the contents of beliefs because, they claim, the contents of perceptions and memories are neither true nor false. Those contents might be deemed accurate or inaccurate, rather than true or false. While understanding that reluctance to take my rather permissive view of the scope of ‘proposition’, I will use ‘proposition’ to designate the full class of mental contents, for the sake of ease of presentation. Nothing in the argument depends upon that choice” (Klein 2008, 27n4). I am not convinced that construing considerations that are not truth-apt as propositions is so innocuous. But here my point concerns propositions that uncontroversially are truth-apt. My question is whether the standard criterion for individuating those propositions is sufficiently fine-grained to properly characterize belief contents like them.

4

3 Epistemically Useful Falsehoods

33

whether a single agent could harbor distinct mental representations—that p and that q—which are co-hyperintensional. In that case, she would believe the very same fine-grained content under different modes of representation. The relation between hyperintensional contents p and q is stronger than necessary equivalence. If belief contents are hyperintensionally equivalent, it is impossible in principle to believe that p without thereby believing that q and conversely. A more plausible alternative is that different modes of representation express different hyperintensions. Then Harry’s belief that vixens are fierce and his belief that female foxes are fierce have different contents, even though he realizes that ‘vixen’ and ‘female fox’ are synonymous.5

3.3

Alternative Proposal

Initially, Klein’s strategy for evading Gettier cases was an anti-defeater strategy (Klein 1976). Knowledge, he maintained, is justified true belief, where the justification in question is doxastic and there are no genuine defeaters in the doxastically operative inferential chain. A defeater is a false belief that is ineliminable from the inferential chain. A defeater is genuine if, in addition to being ineliminable form the causal chain, it is not itself defeated. On this account, ‘Paul’s secretary told him on Friday . . .’, ‘Santa will bring presents’, and ‘They have kept accurate records for eighty years’ are defeaters. Thinking that APPOINTMENT, SANTA CLAUS, AVERAGE RAINFALL and PTOLEMAIC ASTRONOMER result in knowledge despite involving defeaters, Klein introduced useful falsehoods. Now he holds that S knows that h only if: S’s doxastic justification for h is a causal chain that contains no genuine defeaters essentially

or S’s doxastic justification is a causal chain that contains essentially only defeaters that entail relevant propositionally justified truths.

The first disjunct must be satisfied for standardly justified beliefs, and the second for useful falsehoods. Disjunctive analyses are not dreadful, but they are unattractive. We would prefer to avoid them. Moreover, as we saw earlier, a falsehood’s entailing a propositionally justified truth does not obviously compensate for its defectiveness as a link in a justificatory doxastic chain.

5 Another aside: Although I have put the point in terms of contemporary metaphysics, one can get the same fine-grained using Israel Scheffler’s inscriptionalism. It is not obvious that we need to inflate our metaphysics to get the resources we need to properly individuate belief contents (see Scheffler 1955). Nor is it obvious that we need to inflate our metaphysics to get non-causal dependency relations. The resources developed in Goodman’s Structure of Appearance suffice (see Goodman 1977).

34

C. Z. Elgin

I suggest then that rather than focusing on what they entail, we take a closer look at the useful falsehoods themselves. The discussion of hyperintensionality highlights the idea that beliefs and other hyperintensional contents have a fine-grained internal structure. Perhaps we can evade the difficulties by exploiting that structure. In “Truth about Jones,” Ullian and Goodman (1977) argued that a false statement (or proposition, although they would never use the term) can, despite its falsity, be true about something. ‘Sam is in New Haven and Gareth is in Delhi’ is false, since Gareth is in Mumbai. But it is true about Sam, for he is in New Haven. Insofar as the truth value of the sentence depends on where Sam is, it is true. This, I suggest, is what makes the falsehoods in Klein cases useful. They are true about the factors that bear on or figure in the agent’s inference. They owe their overall falsity to factors that do not matter. Philosophers frequently focus on truth-makers. Here we are interested in their counterparts—falsity-makers. Let’s look at Klein’s cases. What makes Paul’s relevant belief in APPOINTMENT false is that his secretary told him about the appointment on Thursday rather than on Friday. What makes Virginia’s relevant belief in SANTA CLAUS false is that it is her parents rather than Santa Claus who will put the presents under the tree. Although they make the propositions in question false, these factors are inferentially idle. They do no work in the agents’ inferences. That ‘on Friday’ plays no inferential role in APPOINTMENT seems obvious. The phrase could simply be eliminated from the statement of Paul’s belief, and the inference would still go through. SANTA CLAUS is a bit trickier. To just drop the term ‘Santa Claus’ leaves an open sentence. For Virginia to have a mental content capable of being believed, a suitable noun phrase must replace it. But beyond the requirement that the replacing term putatively denote an agent or entity seemingly capable of putting presents under a tree, it doesn’t really matter what the replacement is. To see this, imagine that Virginia’s parents mumbled. Virginia didn’t quite catch who exactly they said would put presents under the tree. For Virginia’s inference, this doesn’t matter. Both ‘on Friday’ and ‘Santa Claus’ are, in Strevens’s terms, not difference-makers (Strevens 2008). It is less clear, however, that the falsity-making factor is inferentially idle in AVERAGE RAINFALL. Weatherman’s inference requires that the average rainfall was recorded for a long time. Whether the duration was greater than 80 years doesn’t matter, but that it was a considerable number of years does. Just how many years is not obvious. It depends on how variable rainfall is from year to year. If it is quite variable, then a good many years will be required to get an average we can trust; if it is pretty constant, relatively few years will do. Luckily, that is a problem for meteorology. We don’t have to solve it. We have a different problem: the fact that the falsity-maker seems to be doing some work. It could not simply be eliminated from Weatherman’s inference, in the way that the falsity-makers in APPOINTMENT could. What work is it doing? I suggest that it is facilitating the inference because ‘They’ve been recording the weather for eighty years’ is true enough (see Elgin 2017). Eighty years, the duration mentioned in the sentence, is a good many years. In Weatherman’s statement, the term ‘eighty years’ not only denotes number in the extension of the predicate ‘a good many years’, it exemplifies that it does so. It

3 Epistemically Useful Falsehoods

35

makes the fact that there is plenty of evidence salient. The precise number of years mentioned in Weatherman’s statement is inferentially idle. That it belongs to the extension of ‘a good many years’ is not. PTOLEMAIC ASTRONOMER seems to present an even harder problem. Here we do not have an infelicitous phrase that could be suitably paraphrased or eliminated from an otherwise estimable sentence. The Ptolemaic astronomer’s driving assumption—the geocentric model of the heavens—is wildly incorrect. How could the poor man possibly know what planets would be visible in 800 years when he is dead wrong about what is circling what? The amazing thing about Ptolemaic astronomy is that it works. Even today, celestial navigators both on the sea and in the desert adopt a Ptolemaic perspective because for getting around in vast, trackless realms, it is practical to treat the earth as motionless. And because the Ptolemaic astronomers were very good at their job, to a surprisingly precise degree Ptolemaic astronomy preserves the relevant appearances. Contemporary celestial navigators get where they want to go. To determine what will be seen in the future, it is appearances that need to be preserved. Hence here too we can resort to the true-enough strategy. If all we care about is the appearances, then what is true enough is what correctly describes and predicts the appearances. I suggest then that the astronomer knows whether Mars will be visible from Earth 800 years in the future because, in his reasoning, Ptolemaic astronomy functions as no more than a calculational device. He may use it because he believes the constitutive claims of the theory. But in his reasoning, the constitutive Ptolemaic assumptions do no more than supply an inference ticket that brings him from accurate inputs to accurate outputs. Despite its being very far from the truth, it is true enough to underwrite the inference. If this is right, a constructive empiricist like van Fraassen could maintain that he knows the consequences of scientific inferences. We do not, the constructive empiricist maintains, know which empirically adequate theory is true. But if useful falsehoods lead to knowledge, it simply doesn’t matter whether he arrived at his belief that p via a true theory or via a usefully false, empirically adequate alternative (see van Fraassen 1980). We can test my suggestion by seeing what happens when we vary some of Klein’s examples. Begin with APPOINTMENT. Suppose, that Paul’s secretary (although excellent in other respects) is a bit of a scatter-brain. She is, in general, less than reliable about telling him what is in his calendar. But recognizing that with the weekend impending it is important to make sure Paul knows his schedule for Monday, she tends to take more care and be far more reliable in what she tells him on Fridays than she does about what she tells him on other days of the week. If that were so (and Paul were aware of this fact), it is plausible that he would know that he has an appointment if she told him on Friday, but not if she told him on Thursday. In that case ‘on Friday’ would not be idle in his inference. It would be a differencemaker. Consider now AVERAGE RAINFALL. Suppose the annual rainfall in Northwest Montana varies a lot from year to year. There are trends, and in the long run they smooth out. But an average taken over only a few years would likely be misleading.

36

C. Z. Elgin

Meteorologists have concluded that they need a minimum of 80 years to get numbers they can trust. In that case, Weatherman’s conclusion, based on only 79 years, would not qualify as knowledge. That 79 years is, and exemplifies that it is, a considerable number of years would not make Weatherman’s grounds true enough in this context. It may seem a mite churlish to insist that evidence from merely 79 years is too sparse if 80 years suffices. But a slippery slope threatens. If meteorologists were justified in setting the threshold where they did, then failing to meet that threshold is good reason to think that Weatherman does not (quite) know. The account I offer does not rely on a road not taken. It focuses on the beliefs and belief contents that the agent actually has and uses in her inferences.6 It exploits the fact that only some aspects of the belief contents figure in her inferences. If she is right about them (where this includes their being true enough, when that is all that is required), she is on solid epistemic ground. Moreover, my account is not disjunctive. I do not give one condition for standard justified true beliefs and another for useful falsehoods. Rather, the position I have sketched enables us to say, as Klein originally did, that S knows that h only if S’s doxastic justification contains no genuine defeaters essentially. This is so because not every ineliminable falsehood is a defeater. If the falsehood is true enough—if it is either true or it is false in ways that make no difference to the function it is performing in the agent’s doxastic economy—the agent knows. Klein considers relying on useful falsehoods an epistemic defect, albeit not a fatal one. It would be preferable, all things considered, if Paul had simply reasoned from ‘My secretary told me I have an appointment on Monday’ to ‘I have an appointment on Monday.’ It would be preferable if Weatherman had reasoned from ‘They’ve been keeping accurate records for 79 years’ (or ‘for a considerable number of years’) to ‘The average rainfall is about 13 inches.’ Klein may be right. His assessment handles his cases and comports with the veritistic spirit of most epistemology. But I’m not so sure. The sort of cases I normally focus on concern idealizations and simplifying assumptions. In these cases, the utility of the falsehood lies precisely in that it prescinds from features of actual situations that do not matter. Consider Galileo’s famous thought experiment about falling bodies. According to Aristotelian physics (and common sense) heavier bodies fall more quickly than lighter bodies. So, Galileo says, consider a cannonball and a musket ball. The cannonball is heavier and should fall faster than the musket ball. Tie the two together with a rope. The new, composite object, is heavier than the cannonball, for we’ve added the weight of the rope and the musket ball. Imagine dropping an untethered cannonball (of the same size, shape, and material as the tethered cannonball) and the composite object. Which will fall more quickly? Aristotelians answer: the composite object. But, Galileo notes, part of the composite object is the musket ball, which 6

Normally I rely on L. Jonathan Cohen’s notion of acceptance rather than belief. An agent accepts that p just in case she is willing to use p as a basis for assertoric inference or action when her ends are cognitive (Cohen 1992). Here I speak of beliefs in order to bring my position into contact with Klein’s. But belief seems not to be the relevant attitude when we ask how thought experiments and idealizations function epistemically. So talk of belief is rather out of character for me.

3 Epistemically Useful Falsehoods

37

Aristotelians say, must fall more slowly. As it falls, it serves as a brake, slowing the rate of fall of the composite object. That being so, the composite object should fall more slowly than the cannonball alone. But it cannot fall both more quickly and more slowly than the cannonball. So we have a contradiction. It follows that the rate of fall of an object must be independent of its weight. This is a familiar, beautiful thought experiment. It is conclusive. It would not be better if rather than constructing the thought experiment, Galileo did a bunch of tests, dropping things from various towers and measuring their rate of fall. Had he done so, he would have needed to show how he accommodated potentially confounding factors—air resistance, limitations on the accuracy of his measuring devices, and so forth. He controlled for them in the thought experiment simply by leaving them out. Their introduction would only mislead. However lovely, it might seem that this example is not about inferential empirical knowledge. It is an imaginary exercise about the relation between two magnitudes. It seems rather a priori. Indeed, its conclusion is a modal claim. Rate of fall and weight must be independent. But from the necessity claim, a claim about actuality immediately follows. If rate of fall and weight must be independent, then rate of fall and weight are independent. The inference yields knowledge about a matter of fact. The thought experiment is what I have called a felicitous falsehood (see Elgin 2017). The difference between my felicitous falsehoods and Klein’s useful falsehoods may be just terminological. Perhaps not. In Klein’s discussion, useful falsehoods seem to be second-class epistemic citizens. We grudgingly admit that they yield knowledge; but to be honest, we’d rather not. I suggest that there’s nothing second class about felicitous falsehoods. They are typically epistemically preferable to the unvarnished truth, for they prescind from factors that threaten to mislead (see Elgin 2017). In the cases Klein considers, useful false beliefs involve some sort of mistake. I’m suggesting here that some falsehoods—the felicitous ones—involve no mistake. Moreover, their utility lies not in their facilitating inferences despite their falsity, but in their facilitating inferences because of their falsity. They omit confounding factors that would make inferences confusing, difficult, or impossible to carry out. The issues are complicated and take us into neighborhoods where Klein may not want to venture. But once you open the door to false beliefs’ playing a legitimate justificatory role, the question ‘How large a role?’ emerges.

3.4

Conclusions

I’ve made a number of suggestions that Klein and his followers may be loath to accept. Following Klein, I allowed that doxastic justification is what justifies the epistemic agent in holding her belief. It is about her believing that p, not (directly) about p itself. Contrary to Klein, I suggested that the dependence relation needed for doxastic justification need not be causal. Another sort of metaphysical dependence might suffice.

38

C. Z. Elgin

I argued that belief contents are exceedingly fine-grained. I suggested that the fine-grainedness figures not just in propositional justification (or perhaps we should say hyperintensional justification), but also in the affordances a belief provides for doxastic justification. In making an inference, the agent draws on only some features of her belief content. Some features are not difference-makers. They perform no function in her inference, so can be ignored. If her belief is false only on account of those features, it is a useful falsehood. As I see it, we need not detour through a road not taken. Rather, we need only attend to the operative elements of the inferential road we are actually on. One final point: It may seem that Klein has replaced his defeasibility theory with a hybrid account. I disagree. Rather, I suggest, he has discovered that not all false beliefs that figure essentially in chains of inference that give rise to beliefs are defeaters.

References Bealer, G. (1998). Propositions. Mind, 107, 1–32. Cohen, L. J. (1992). An essay on belief and acceptance. Oxford: Clarendon Press. Elgin, C. (2017). True enough. Cambridge, MA: MIT Press. Gilovich, T., Griffin, D., & Kahneman, D. (Eds.). (2002). Heuristics and biases. Cambridge: Cambridge University Press. Goodman, N. (1977). The structure of appearance. Dordrecht: Reidel. Klein, P. (1976). Knowledge, causality, and defeasibility. Journal of Philosophy, 73, 792–812. Klein, P. (1981). Certainty: A refutation of scepticism. Minneapolis: University of Minnesota Press. Klein, P. (2008). Useful false beliefs. In Q. Smith (Ed.), Epistemology: New essays (pp. 25–62). Oxford: Oxford University Press. Kripke, S. (2011). A puzzle about belief. In Philosophical troubles (pp. 125–161). Oxford: Oxford University Press. Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In B. Hale & A. Hoffman (Eds.), Modality: Metaphysics, logic and epistemology (pp. 109–134). Oxford: Oxford University Press. Scheffler, I. (1955). On synonymy and indirect discourse. Philosophy of Science, 22, 39–44. Strevens, M. (2008). Depth. Cambridge, MA: Harvard University Press. Ullian, J., & Goodman, N. (1977). Truth about Jones. Journal of Philosophy, 74, 317–338. Van Fraassen, B. (1980). The scientific image. Oxford: Oxford University Press.

Chapter 4

Theoretical Unity in Epistemology Jonathan L. Kvanvig

Abstract Epistemology is more than the theory of knowledge. It involves reflection on and theorizing about cognitive successes from a purely theoretical or intellectual point of view, one that brackets other concerns such as practical, moral, and aesthetic ones. Knowledge is certainly one success of this sort, but not the only one. In addition, there are the great achievements of understanding and wisdom, as well as the ordinary accomplishments of having rational opinions and justified beliefs. Multiplicity, however, can lead to fragmented theorizing. Fragmentation can arise by ignoring much of the theoretical landscape, as when one limits one’s theorizing in epistemology to the theory of knowledge. Fragmentation can also arise by addressing more (or even all) of the landscape, but not in a unified fashion. For example, one might be a Chisholmian foundationalist about justification while also being a virtue epistemologist about knowledge. Fragmentation is disappointing. Just as it is disappointing to be told that the laws of terrestrial motion have little in common with the laws of celestial motion, so too is disunity in epistemology something to lament. It is worthwhile, then, to compare some possible ways of avoiding such fragmentation, to see whether there is any hope for greater unification regarding the variety of items in our epistemological landscape. My goal will be to argue in a very tentative way for the superior prospects for unity found by taking the theory of confirmation as fundamental to epistemological theorizing, at least when compared to the extant rivals of a knowledge-first approach and a reasons-based account. Keywords Fragmentation · Holism in epistemology · Epistemic primitives · Confirmation relation · Normative epistemology · Meta-epistemology · Knowledgefirst · Defeasibility · Epistemology

J. L. Kvanvig (*) The Department of Philosophy, Washington University in St. Louis, St. Louis, MO, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_4

39

40

4.1

J. L. Kvanvig

Introduction

Epistemology is more than the theory of knowledge. It involves reflection on and theorizing about cognitive successes from a purely theoretical or intellectual point of view, one that brackets other concerns such as practical, moral, and aesthetic ones. Knowledge is certainly one such kind of success, but not the only one. In addition, there are the great achievements of understanding and wisdom, as well as the ordinary accomplishments of having rational opinions and justified beliefs. This list isn’t meant to be exhaustive: we might also mention beliefs that are safe for purposes of getting to the truth, or sensitive, or ones that track the truth well, or are warranted, or confirmed by our experience, or supported by reasons (of a truthrelated kind), or beliefs that partake of some type of positive epistemic status not yet mentioned. Moreover, there are character traits that fall within the territory in question, whether in terms of native endowments such as vision and other senses that generally get us to the truth, or developed character traits such as openmindedness and intellectual courage. This multiplicity can lead to fragmented theorizing. Fragmentation can arise by ignoring much of the theoretical landscape, as when one limits one’s theorizing in epistemology to the theory of knowledge. Fragmentation can also arise by addressing more (or even all) of the landscape, but not in a unified fashion. For example, one might be a Chisholmian foundationalist about justification while also being a virtue epistemologist about knowledge. Such fragmentation is disappointing. Just as it is disappointing to be told that the laws of terrestrial motion have little in common with the laws of celestial motion, so too is disunity in epistemology something to lament. It is worthwhile, then, to compare some possible ways of avoiding such fragmentation, to see whether there is any hope for greater unification regarding the variety of items in our epistemological landscape. My goal will be to show the superior prospects for unity found by taking the theory of confirmation as fundamental to epistemological theorizing, at least when compared to the extant rivals of a knowledge-first approach and a reasons-based account. We can develop this argument by organizing the theoretical tasks we face when constructing a complete epistemology, after which we will look at some competing approaches for generating the kind of unity we seek, beginning with a knowledge-first approach.

4.2

The Main Tasks

The issue of unity in epistemological theorizing is a by-product of the recognition that the study of cognitive success from a purely intellectual point of view encompasses more than just the theory of knowledge. It involves a domain of intellectual accomplishments, including knowledge, understanding, and wisdom, and it also

4 Theoretical Unity in Epistemology

41

includes the domain where normativity is central, including the more atomistic areas concerning the theory of confirmation and the theory of defeasible reasons and reasoning, and the more holistic areas involving the theory of rationality, justification, and warrant, all within the general territory of normatively positive epistemic status. It also includes the notion of intellectual virtue, both in terms of natural excellences as well as developed traits. We can begin a search for unity in our epistemology by organizing each domain in hopes of finding something fundamental in each. For example, there is a longstanding tradition of making knowledge basic when it comes to the great intellectual goods. In this tradition, understanding is glossed in terms of knowledge of causes, while wisdom is understood in terms of knowledge of what is important or what matters, knowledge of the significance of things. This tradition thus ties understanding to the kinds of accomplishments characteristic of modern science to generate explanatory knowledge of the various systematic features of our universe, while wisdom will include, among others things, the intellectual accomplishments deriving from ethical reflection as well as value studies conceived more generally. Though this tradition encounters difficulties,1 it offers a picture of the kind of unity in theorizing that I wish to investigate further here, at least with respect to the part of epistemology that involves these great intellectual goods. It is obvious, however, that unity on the side of the great intellectual goods is not the whole story, however, for there remain the topics noted above that fall within the domain of normative theory. An initial glance at our list (the list that includes theories of confirmation, defeasible reasons and reasoning, and justification) suggests, as noted above, the following. There are both more holistic and more atomistic items on the list. The theories of justification, rationality, and warrant require more holistic treatment (since, in principle, anything in the relevant domain can generate failures of justification, rationality, or warrant), while the theory of reasons and reasoning itself calls for a more atomistic approach (since we reason from individual pieces of information to other pieces of information, and the normative theory should tell us when these transitions are legitimate from a purely intellectual point of view). The theory of confirmation involves a two-place relation on information contents (represented as C( p, q)), while a theory of defeasible reasons involves a three-place relation between a person, a reason, and what the reason supports (represented as R (r, q, s)). Finally, a theory of justification involves a target of justification (either some informational content or some cognitive state such as belief), something that

1

There is, for example, the growing body of literature on whether understanding requires knowledge (see, e.g., Kvanvig 2003b, 2009a, b, c); Pritchard (2009, 2010); Elgin (2007, 2009); Khalifa (2011, 2013); Kelp (2015); Grimm (2011); Morris (2012); Rohwer (2014), Sliwa (2015); and Janvid (2012, 2014). Though the literature on wisdom is not as extensive, the general approach derived from Aristotle that treats wisdom as a kind of knowledge has been challenged recently by Sharon Ryan. Her early work on wisdom treated it as (or as involving) a kind of knowledge (Ryan 1999; see also 2014), but Ryan (2012) argues for a view of wisdom in terms of rationality rather than knowledge. Further literature can be gleaned from Dennis Whitcomb’s annotated bibliography on the subject (Whitcomb 2010).

42

J. L. Kvanvig

generates the justificational status in question (let’s agree to call it “evidence” for now), and the cognizer relative to whom the item in question is justified. Among the locutions that need accounting for, then, are propositional justification (“p is justified for S”, or J(p, S)), doxastic justification (“S’s believing p is justified”, or J*(BS, p)), and personal justification (“S is justified in believing p”, or J{(S, Bp)). The two areas here where the distinction between the two is easiest to elide are the theory of confirmation and the theory of defeasible reasons. For it is easy to think of the theory of confirmation as involving the notion of evidence, and it is tempting to identify your evidence with the reasons you have.2 But caution is in order here. First, reasons can be reasons for acting as well as reasons for believing, and the theory of defeasible reasons that is relevant to epistemology will be a theory about epistemic reasons, not practical or pragmatic ones. But making the point in this way flouts appropriate circularity prohibitions, so something more instructive must be said here. Perhaps one could say that epistemological projects include the theory of confirmation and the theory of epistemic reasons, where the latter are reasons appropriately tied to the intellectual goal of getting to the truth and avoiding error. Proceeding in this way may incline us to distinguish the theory of defeasible reasons from the theory of defeasible reasoning, for even if there can be defeasible reasons for believing that are not epistemic reasons, it may not be possible for defeasible reasoning to proceed from anything other than epistemic reasons. That is, even if there can be non-epistemic reasons for believing a given claim, such reasons might not be ones that can function in defeasible reasoning to the claim in question. Instead, we might want to insist that a theory of defeasible reasoning separate epistemic reasons from practical ones so that defeasible reasoning from practical reasons can only support believing the claim in question, whereas defeasible reasoning from epistemic reasons supports the claim itself. If we categorize epistemological tasks in this way, it is easier to identify the theory of defeasible reasoning with the theory of confirmation, but the relationship between the theory of defeasible reasons and the theories of confirmation and defeasible reasoning would still remain in doubt, since any simple identification of reasons with evidence is suspect. Second, one might also think that evidence is relevant to the propositional content of a belief, whereas a reason is relevant, not to the content, but to the state of believing itself, in which case a reason is an appropriately intellectual reason for believing. In short, it is a pre-theoretical option to think of confirmation as a relation between propositions and reasons as a relation between mental states (or between facts or propositions and mental states). When two relations are relations between different relata, any identification of the two relations is suspect. Finally, some may be inclined to think of the confirmation relation as significantly different from intellectual reasons and reasoning, thinking (perhaps) that one of the two is more holistic than the other. One way in which this approach might be pursued is by claiming that confirmation is a two-place relation whereas the reasons relation needs more places in it. For example, Thomas M. Scanlon (2014) claims that the reasons relation is a four-

2

For an instance of such an identification, see McCain (2014). See also Lord (2018).

4 Theoretical Unity in Epistemology

43

place relation that obtains between a proposition p, an agent x, a set of circumstances c, and an action or attitude a: R(p, x, c, a).3 Mere difference in the number of relata would be sufficient grounds for refusing to identify the two relations. In short, we should assume from the start that there are (at least) three tasks within normative epistemology, refusing to elide the distinctions given here. It is worth noting in passing, at least, the way in which typical externalist theories of knowledge fail to address what is demanded of a complete and unified approach within the territory of normative epistemology. Proper Functionalists, Virtue Epistemologists, and Process Reliabilists all present their theories as, at most, theories of justification within this landscape, showing insufficient efforts to develop the theories of confirmation and defeasible reasons in terms of the fundamental notions they claim are involved in the theory of knowledge. And Modal Epistemologies, characterizable in terms of safety, sensitivity, and truth-tracking, are not always sensitive, so to speak, to the various parts regarding which we look for a good theory.4 Without an adequate survey of the territory to be covered, it is easy to develop an epistemology that is inattentive to the question of whether the resources of one’s theory have any hope of generating a complete epistemology that has the kind of theoretical unity that we hope for. These points about ways in which fragmentation can result from ignoring much of the theoretical landscape help with the other main part of that landscape, the domain of the intellectual virtues. One might try to make the virtues fundamental to epistemological theory, but the prospects for such are rarely pursued and tend to be dim. Linda Zagzebski (1996), for example, attempts to explain both knowledge and justification in terms of the virtues, but when pressed,5 she counsels paying less attention to the notion of justification: “. . . the concept of justification has been made to serve too many purposes, a situation that provided one of my motives for proposing that we shift our focus away from justification to traits of persons” (Zagzebski 2000, 213). To the extent that virtue theories actually ignore, or propose to ignore, regions of the epistemological landscape, they engender fragmentation rather than unity.

3 For some reason I can’t detect, Scanlon holds that this account of reasons doesn’t make reasons relative to the agent in question, but rather holds that “something is a reason for an agent only if it is also a reason for any other agent in similar circumstances” (Scanlon 2014, 32). First, as Ralph Wedgwood (forthcoming) points out in a review of Scanlon’s book, this claim doesn’t follow from the characterization of reasons Scanlon gives. Moreover, if it did follow, then there is no need for a four-place relation that includes reference to a person. Instead, it would be sufficient to characterize the Scanlon view of reasons in terms of a three-place relation between a proposition, a set of circumstances, and an attitude or action. 4 There are exceptions, of course. Nozick (1981) attempts to use truth-tracking to generate a theory of evidence, which might be used to give a theory of confirmation, as does Roush (2005). The same can’t be said for other such approaches, such as Sosa’s (2007) reliance on a safety condition or Pritchard’s (2005) anti-luck theory. 5 By her critics in the Philosophy and Phenomenological Research discussion of her book: Alston (2000), Greco (2000), Kornblith (2000), Kvanvig (2000), and Rorty (2000).

44

J. L. Kvanvig

The best attempt to avoid such fragmentation is in John Greco (1993), where he attempts to employ virtue notions to define knowledge and the kind of subjective justification that internalists prize, but it is clear from the discussion above that such an approach falls well short of covering the landscape of normative epistemology. An alternative route would be to understand the virtues in question as dispositional states toward various items in the epistemological landscape canvassed above, thereby making them theoretically derivative rather than fundamental. Since the prospects for a virtue theory of defeasible reasoning or a virtue theory of confirmation are plausibly thought to be dim at best, I will assume here that the character traits in question are best thought of derivatively, to be accounted for after we have developed a unified approach to the great intellectual goods together with the domain of normative epistemology.

4.3

Knowledge-First Epistemology

Once we reach this point, we are well positioned to appreciate the attention that has been paid to Timothy Williamson’s knowledge-first approach to epistemology (see, especially, Williamson 2000). For this approach offers some hope of unifying the normative domain with the domain of the great intellectual goods, and this quick, initial glance at the normative domain gives us at least a glimpse of why such a knowledge-first approach is both original and promising when it comes to the quest for unity in our epistemology. Williamson doesn’t discuss understanding or wisdom, but if one accepted the ancient tradition concerning them described above (in terms of knowledge of causes and knowledge of significance, respectively), there is a clear way in which knowledge is the basic concept when it comes to these great intellectual goods. What remains, then, are the topics of normative epistemology, and Williamson attempts to show that knowledge is the basic concept here as well. He insists, for example, that knowledge is unanalyzable and thus should not be understood in terms of true belief plus some suitable combination of normative elements. Instead, he hopes to reverse the order of explanation, attempting to use the notion of knowledge itself to illuminate this normative domain. Most notorious here is his insistence that the notion of evidence itself must be identified with knowledge. Once evidence has been identified with knowledge, we have gone some distance toward explaining the normative domain in terms of the fundamental notion of knowledge. Notice from the above characterization of the three tasks described, we have the following relations to investigate: • • • •

C( p, q) R(r, q, S) J(e, p, S) J*(e, BS, p, S)

4 Theoretical Unity in Epistemology

45

where C is the confirmation relation, R the defeasible reasons relation, J the propositional justification relation, and J* the doxastic justification relation. There is good reason for thinking that the last two are interdefinable.6 Moreover, it is tempting to side with Evidentialism of some sort in thinking that both doxastic and propositional justification are to be understood in terms of the fundamental notion of evidence,7 and it is natural to think that the theory of confirmation and defeasible reasoning might be explained in terms of the concept of evidence. Given the Williamsonian identification of evidence with what a person knows thus yields a great deal of unity in our theorizing by making knowledge central to much of the work to be done in generating a complete epistemology. Of course, the work is not done, and full unity has not been achieved until and unless we can show that all the normative notions in question can be understood without recourse to epistemic notions other than knowledge. How might such unity be achieved on a knowledge-first approach? Consider, for example, the connection between the theory of confirmation and the theory of defeasible reasoning. It would be nice for these two theories to reduce to the same, evidence-based theory, and for this evidence-based theory to be basic to the theory of justification (or rationality or, more generically, positive epistemic status). Note, however, that endorsing the claim that evidence is knowledge won’t get us remotely close to the harmonization envisioned, and it is worth seeing why. Begin with the idea of confirmation itself, the C relation. Our question is about this relation, and part of a full theory of confirmation will be some explanation of how many places there are in this relation and the nature of each relatum. But our theorizing about confirmation can’t be completed simply by attending to the relata. Our fundamental question is about the C relation itself, not about the relata in question, and we can’t answer the question about the C relation itself just by noting that the evidence involved as one of the relata is to be identified with knowledge. Asking about the nature of the C relation is asking about the relation itself, not about the relata. For vast stretches of the history of epistemology, this distinction was ignored because of the infallibilist assumptions being made about the kind of evidential support needed for knowledge. If knowledge-level support requires infallibility, and if infallibility is understood so that it requires entailment, then the only question left is to address the question of what one’s evidence is. But once fallibilist approaches to knowledge and justification arise, giving an account of evidence does not automatically give us a theory of the C relation, and so we need an account of that relation in addition to an account of its relata.

6 For discussion of how and why, see Kvanvig and Menzel (1990) and Kvanvig (2014, Appendices A and B). These sources also include discussion of the notion of personal justification, which I ignore in the text for convenience. Personal justification is attributed when we describe a person as being justified in believing a certain claim, and the sources just mentioned show how to explain this notion in terms of doxastic or propositional justification. 7 For defense and discussion, see Feldman and Conee (1985, 2004). Further discussion can also be found in Dougherty (2011).

46

J. L. Kvanvig

On this score, Williamson offers the following account (if we identify confirmation with evidence in the usual way, so that e confirms p iff e is evidence for p): def

Cðe; pÞ for S ¼ e 2 K ðwhere K ¼ fx : S knows xgÞ and PðpjeÞ > PðpÞ: We can think of this account as a Williamsonian refining of what we might call the Orthodox Approach, according to which the C relation is to be understood in terms of the mathematical theory of probability, where the confirmational quantity is the conditional probability on the evidence when that conditional probability strictly exceeds the unconditional probability. That is, we define qualitative degree of confirmation (Cd) as follows: def

ðQCÞCd ðe; pÞ ¼ PðpjeÞ > PðpÞ: And then we partially endorse what Fitelson (2005) calls the “Received View” of quantitative confirmation (Cq): def

ðRVÞCq ðe; pÞ ¼ PðpjeÞ: The partiality in question arises for the Orthodox Approach insisting that (RV) holds just where (QC) is satisfied: def

ðOAÞCq ðe; pÞ ¼ PðpjeÞ $ Cd ðe; pÞ It is worth noticing as well what Williamson adds that isn’t a part of (OA). Note that (OA) treats the C relations as two-place relations between propositions, but Williamson adds a relativizing clause to the account. Thus, Williamson’s account is not an account of any confirmation relation that holds between propositions, but rather an account of a confirmation relation between propositions relative to some cognizer or other. Such an emendation of (OA) plays well into a project that wishes to align the theory of confirmation with that of defeasible reasons, but anyone attracted to (OA) should hesitate here. If our first-glance approach to confirmation identifies it with a two-place relation (an approach, by the way, that fits nicely with thinking of the conjunction of the premises of a deductively valid argument as providing confirmation for the conclusion of that argument), then we need some discussion and argument for moving from this two-place relation to a three-place relation that relativizes confirmation to some cognizer or other. My claim here shouldn’t be understood as an objection to this move, but rather as a complaint against any attempt to generate the full unity in theorizing that we might have hoped for from a knowledge-first approach when it relativizes in this way without substantiating argumentation. As we will see later, there may be good reasons for thinking that the C relation is not really what it appears to be at first glance—that it is not really a two-place relation—so we may find reasons for

4 Theoretical Unity in Epistemology

47

thinking that this justificational lacuna isn’t a severe difficulty for a knowledge-first approach. I flag it, though, because it helps us see what is needed if this approach is to fulfill its promise. Perhaps, though, we should understand the second clause of Williamson’s account as an account of a subject-independent confirmation relation, specifying what the evidence is, with the first clause telling us when the evidence is possessed by the subject in question. In such a case, we have no account of subject-independent confirmation in terms of knowledge, and hence there is some loss of unity to the theory. It is also important to notice that (OA) isn’t best thought of as a theory of the C relation, but rather a condition of adequacy on a theory of the C relation, much as the T-schema is a condition of adequacy on a theory of truth.8 In order to develop a theory that meets this condition of adequacy, we need, among other things, an explanation of what kind of function P is—in particular, whether it is a probability function—and exactly which function it is.9 Telling us exactly which probability function is being used is a lacuna in Williamson’s treatment. To give such a treatment, one has to show where the probabilities in question come from. Williamson is clear that these probabilities aren’t degrees of belief or even ideally rational degrees of belief, but saying what they aren’t isn’t good enough.10 The lesson here is, then, that a knowledge-first approach, conceived of in terms of providing full unity to our epistemological theorizing, has issued promissory notes that stand in need of redemption. If, for example, one could defend the idea that the probabilities employed are logical, in the way attempted by John Maynard Keynes (1921) and Rudolf Carnap (beginning with Carnap 1945, 1949, through Carnap 1971, 1980), we’d be able to use such a theory in a knowledge-first approach that didn’t ignore the C relation by talking only about its relata. For the only distinctively epistemic concept involved in the C relation would be knowledge, with the full account of that relation explicated by knowledge plus logic. With a bit of theoretical luck thrown into the mix, we might be able to defend the move from the two-place C relation involved in (OA) to the three-place relation that Williamson defines, and then perhaps the theory of defeasible reasons and reasoning could then be

8

As required in Alfred Tarski’s seminal work on the theory of truth in Tarski (1944). A philosophical rant: I weary of the philosophical tack of claiming to have given a theory of X by identifying it with a function from Y to Z. Nice start. Now finish the job! Tell us which function (not to mention the difference between identity and identification). 10 Note that, in the absence of completing this task, not only don’t we know which probability function is involved in the Williamsonian account, we don’t even know that it is a probability function at all. And if P isn’t a probability function, then it is just another resident in the stable of normative epistemological concepts—an apparently non-mathematical, non-formal, normative notion of epistemic probability—rather than some independent notion that can serve as a grounding, even if only partial, of the relevant normativity. To the extent that Williamson’s approach relies on other normative epistemic notions, the knowledge-first approach fails to achieve the hoped-for unity in our epistemology theory. 9

48

J. L. Kvanvig

constructed on top of the theory of confirmation, identifying (epistemic) reasons with evidence. Finally, we might then have hope for developing the theory of justification in terms of a holistic proportioning of belief to totality of evidence. There is much in this working diagram of epistemological machinery that remains to be constructed, but the most significant part, to my mind, is the way in which the promise of this version of knowledge-first epistemology depends on results concerning the relation of confirmation that have not yet been developed. The threat to knowledge-first epistemology is threatening indeed. First, it is hard to see how to understand the confirmation relation itself in terms of knowledge (as opposed to identifying one of its relata in terms of knowledge). One might, perhaps, be tempted toward the idea of understanding confirmation in terms of being in a position to know, but that approach could only be plausible when talking about the threshold notion of confirmation (as opposed to the incremental notion). Such difficulties might suggest taking the notion of confirmation as primitive, but doing so would eliminate the unity in theory promised by the knowledge-first approach. If both the notions of confirmation and knowledge are primitive notions, it is neither a knowledge-first approach nor a confirmation-first approach. Something similar occurs when we try to achieve the desired unity by using the resources of reasons-first epistemology. Showing how and why is the task of the next section.

4.4

A Reasons-Based Alternative

It is worth exploring as well a non-confirmational approach, one that takes the notion of reasons to be fundamental. On this approach, we define the notion of evidence in terms of the notion of an epistemic reason, where we identify an epistemic reason as a reason constitutively tied to the goal of getting to the truth and avoiding error.11 We could then give a theory of defeasible reasoning by insisting that such reasoning always proceed via epistemic reasons. We can then treat epistemic justification (or rationality or warrant) in terms of the relation between the pro tanto and the all-things-considered evaluation: justification arises from an account of the weight of reasons for and against an option, so that to say that a belief is justified is to say that some appropriate measure on all the reasons favors believing it over its alternatives. Perhaps, as in Lord (2018), we say that a belief is justified when the weight of the reasons in favor of belief is at least as weighty as the weight in favor of any other (competing) option.

11

I note here that such a definition faces significant challenges arising from the fact that reasonsbased epistemology needs to treat the reasons relation as a relation between (token) mental states. If so, however, such a reasons-based approach will not be able to give a proper account of the distinction between there being evidence and one’s having evidence. I pursue this criticism of such approaches (Kvanvig 2018).

4 Theoretical Unity in Epistemology

49

Such an approach might seem to mesh nicely with standard defeasibility approaches to the theory of knowledge, identifying knowledge with justified true belief that is not defeated by any relevant information one does not possess. To so mesh, we need an account of defeat that explains it in terms of reasons, and we’ll have to develop such an account for both internal defeaters—the kind that undermine justification—and external defeaters—the kind that undermine knowledge but not justification. To characterize this distinction in types of defeaters, we need a distinction between reasons that exist and reasons in one’s possession, between (in the case of epistemic reasons) the evidence there is and the evidence one has. Things get tricky here, however, for one might cash out the idea of having reasons in terms of knowledge (Williamson 2000), being in a position to know (e.g., Gibbons 2013), or rational or justified belief (e.g., Neta 2008), and then the unity sought will have been undermined. But perhaps we could adopt the better-entrenched perspective that imposes a mere psychological constraint on the having of reasons: to have a reason is for there to be a reason that is psychological realized in cognition, in terms of either belief or the contents of experience. The generic notion of a defeater is that of a piece of information that prevents a set of reasons from being weighty enough to justify believing a given claim.12 Whether these pieces of information are facts or truths or beliefs or takings isn’t of primary concern at this point. What does matter at this point is that these things have to have epistemic significance, and the significance in question had better be explicable in terms of the language of reasons, on pain of losing the meta-epistemological unity we seek. So, what it is for a piece of information to prevent a set of reasons from being weighty enough? We might try to characterize this idea in the way John Pollock does in Pollock and Cruz (1986), where he distinguishes rebutters from undercutters. A rebutter is a reason to disbelieve, and an undercutter is a reason to believe that the reasons one has for believing a claim are not indicative of truth in the case in question: where R is a reason for p, an undercutter is a reason for ~(R p) (where ‘ ’ is the symbol for counterfactual implication).13

Compare Lord (2018): “Defeaters are facts that prevent a set of reasons from being weighty enough to confer some all things considered epistemic standing.” 13 I note in passing that this account of an undercutter is vexed. One would have thought that the intuitive idea of an undercutter is a further piece of information that attacks the epistemic connection between R and p: it is a piece of information that confirms, or is a reason for thinking, that (in the context in question) R isn’t a weighty enough reason to justify believing p. The problem with such an account is that it undermines the theoretical unity sought, since we have to be able to explain the notion of a defeater in order to use the language of reasons to explain the notion of justification in terms of reasons weighty enough to support belief. But if we clarify an undercutter in terms of the language of reasons and justification in the way imagined, we have circularity in the account. It is perhaps for this reason that Pollock clarifies an undercutter as he does. The problem, however, is that a piece of information can attack the epistemic connection p). The latter is but one species of the former between R and p without being a reason for ~(R genus. To see this, just note the variety of ways in which skeptics can undermine our confidence that our reasons are good enough. One way is to give us reasons for thinking that our reasons are truth12

50

J. L. Kvanvig

Pollock’s complete theory of internal defeat results from adding that one’s reasons for belief are defeated when there is a non-overridden undercutter or rebutter, since the weightiness of a set of reasons can be restored by adding further information that disarms the defeaters that are present. Restorers can be characterized in terms of the language of reasons as well. Where D is a defeater of the support relation between a set of reasons R and a belief with content p, R0 is a restorer just in case R & D & R0 is a reason to believe p. Besides internal defeaters as characterized above, however, a reasons-based, defeasibility approach to the domain of epistemological normativity cannot be extended to the great intellectual goods of knowledge, understanding, and wisdom without developing a defeasibility approach to the Gettier problem. To do that, we need not only the notion of an internal defeater, as above, but an external defeater as well. In order to provide the kind of unity sought here, it is important that the notion of defeat be explicable in terms of the notion of a reason, and for internal defeaters, an intuitive idea is that a defeater is a piece of information which, in conjunction with a reason in favor, is no longer a reason in favor. External defeaters also need to be understood in terms of the fundamental notion of a reason, and a first pass at such an account is a counterfactual one: there is no fact such that if one were aware of that fact (or came to justifiably believe that fact), one would have an internal defeater for any reason one had for believing p.14 One difficulty with such an approach is the Shope conditional fallacy, the takeaway lesson of which is that one should resolutely avoid any conditional analysis of anything (see Shope 1978).15 A further problem, however, is that the proposal overreaches, since it encompasses both misleading and non-misleading defeaters. The former do not undermine knowledge, whereas the latter do. So, without an account of the distinction between misleading and non-misleading defeaters, an

sensitive (i.e., ~(R p)); another is to give us reasons for doubting that our reasons are truthp)). A further way is to get us to agree that if we sensitive (i.e., for thinking it unlikely that ~(R can’t show that our reasons are truth-sensitive once we start reflecting on the matter, we are no longer in a position for our reasons to be good enough. (This last point has to be made carefully to avoid falling prey to William Alston’s concerns about level confusions in epistemology (see Alston 1980), but I argue in Kvanvig (2014, §3.4) that it can be done.) All these skeptical maneuvers are ways of undercutting the connection between our reasons and what they are reasons for, and they force a Pollockian to find a way of saying how many different ways of undercutting there can be and giving a characterization of each in terms of the language of reasons (without relying on the notion of justification, which is supposed to be defined in terms of reasons plus some no-defeater clause). Whether such can be done in a way that preserves theoretical unity is a vexed issue indeed. 14 Compare Peter Klein’s account in Klein (1971, 475): “there is no true proposition [p] such that if it became evident to S at t1, p would no longer be evident to S.” 15 For resistance to the take-away lesson, see Bonevac et al. (2006), where it is argued that some careful restrictions on conditional analyses can avoid the fallacy. These authors are correct that the take-away lesson overreaches, since there is no ground for thinking that no conditional analysis can succeed. I am unconvinced that the restrictions offered are the correct ones, however, though the present venue is not the place to go into the matter.

4 Theoretical Unity in Epistemology

51

account using no normative epistemic notions other than the notion of an epistemic reason, a defeasibility account cannot succeed. There is a further reason for dissatisfaction with a reasons-based epistemology. Such an account takes as primitive the relation of support that exists between a reason and what it is a reason for. Just as in the case of a knowledge-first epistemology, where the relation of support (or confirmation) was ignored in favor of a focus on one of the relata of that relation, just so with a reasons-first epistemology. There is a relation of support, and a reasons-first epistemology tells us that one of the relata for this relation is to be understood in terms of the notion of a reason. As before, however, this leaves us with an important complaint: we want to know not just about the relata of the supports relation, but also about the relation itself. Without an account of the relation itself, the epistemology will be incomplete, and just as with the knowledge-first approach, it isn’t clear at all how to turn the account of one of the relata into an account of the relation. Of course, one can always take a crucial unexplained element of one’s approach as primitive, and the same could be claimed here. Such a maneuver leaves us with an unfortunate element of normativity in our account, of just the sort that might make one suspicious of using epistemic notions in developing, defending, and characterizing our best scientific accounts of how the world works. On this score, it is worth noting how one might try to avoid such an approach without being in a position to characterize the reasons relation directly. One might model such an approach on Chisholm’s epistemology. That epistemology can be viewed as attempting to provide the kind of unity sought by building the entire theory of knowledge on the fundamental relation of one thing being a reason to believe another. Central to the account are Chisholm’s levels of epistemic appraisal, and the principles he endorses start with the idea that self-presenting states themselves achieve some positive standing: they count as epistemically acceptable. One can also achieve this status for beliefs about how one is appeared to, and the remainder of the story concerns how one climbs the ladder of appraisal to the level of the evident, which is required for knowledge. The story of climbing this ladder is the story of the role of coherence in the theory of knowledge, and the theory can be completed by noting that nothing gets to the level of the evident except by the path already articulated. One might claim that such an approach can elucidate the relation in question indirectly, on analogy with the way recursive definitions can elucidate a notion without defining it directly. The idea, then, would be to see the foundations of knowledge in terms of principles that connect self-presenting states and beliefs about how things appear to levels of epistemic appraisal, and then to characterize how collections of items at any given level climb to the level needed for knowledge, noting at the end that there is no other way to get such status. If carried out successfully, such an account can tell us what the non-normative ground of normativity is: it is found in coherence relations involving self-presenting states and beliefs about how one is appeared to. It is important to note, however, that there are two problems with this Chisholmian attempt to characterize the relation in question. The first is that the

52

J. L. Kvanvig

foundational role played by self-presenting states involves a state that is characterized epistemically: a self-presenting state is one that is necessarily such that if you are in it, the claim that you are in that state is evident for you (the level of epistemic appraisal required for knowledge). There is an important caveat to this story, for all the principles in question include a grounds-for-doubt clause. The analogy with recursive definitions, however, requires that foundational beliefs be characterized in non-epistemic terms, just as, when we say in first-order theory that an atomic formula is a well-formed formula, we don’t define an atomic formula in terms of well-formedness. Second, all of Chisholm’s epistemic principles include a groundsfor-doubt clause, so that one can say that believing that you are appeared to redly is a reason to believe that something is red, but only when there are no grounds for doubt concerning the belief in question. Notice, however, that the grounds-for-doubt idea is not characterized in terms of a relation between mental states where one of them is a reason for the other. Instead, Chisholm characterizes this relation in terms of the notion of confirmation, a relation between propositional contents. As a result, the promise of unity disappears unless and until a defender of such a Chisholmian approach can find a way to explain the confirmation relation in terms of the notion of reasons, and until a way is found to characterize the relation between reasons and what they are reasons for that doesn’t rely on this inappropriate analogy with recursive definitions. In both cases looked at to this point—a knowledge-first approach and a reasonsfirst approach—there is a relation left unexplained. Moreover, it is a relation that is fundamental to the entire epistemological project, and once its fundamentality is appreciated, it is easy to see a third approach to generating the desired unity in our epistemological theorizing. Such an approach focuses on this fundamental relation itself and attempts to use this notion to explain the entire landscape of intellectual success from a purely cognitive or theoretical point of view. We can use the traditional language of confirmation to specify this relation, and our task in the next section is to show the promise of a confirmation-first approach for generating the desired unity in epistemological theorizing.

4.5

A Confirmational Approach

An introductory sketch of such an approach is relatively simple, since its basic outline mirrors what was just noted regarding a knowledge-first approach. The difference is that a confirmational approach treats the task of clarifying the C relation itself as the fundamental task, not ignoring it in favor of attention to its relata. If we have an adequate account of this relation, we can then proceed to build a theory of defeasible reasons on top of it either by identifying reasons with evidence, which will be the first relatum of the C relation, relativized to the agent in question for whom that evidence counts as a reason, or by thinking of reasons in terms of a psychological realization of evidence where this psychological realization is

4 Theoretical Unity in Epistemology

53

something that pro tanto justifies the formation of a further psychological state (with a specific content) in virtue of the confirmational relationship that obtains between the two contents in question. We then build the theory of justification on top of these theories by describing a proposition or belief for which a person has reasons that are not ultimately defeated by other information that the person has, and then we hope to use this same defeasibility notion to provide an account of knowledge in terms of justified true belief, where there are no (non-misleading) defeaters not in the possession of the individual in question (defeaters which meet some requirement of relevance, such as being reasonable to believe to be false).16 Building a theory in this way involves recognizing the theoretical priority of propositional justification over doxastic justification,17 where the former is clarified in terms of justificational grounds characterizable in terms of the confirmational notion of evidence that is fundamental to this approach. Once the theory of knowledge is in place in this way, if we accede to the ancient tradition of identifying understanding with knowledge of causes and wisdom with knowledge of significance, we have the hoped-for promise of unity in our epistemological theory. And even if we resist this ancient tradition, identifying understanding and wisdom with some combination of factivity and justification or rationality (as in Kvanvig 2003b and Ryan 2012, respectively), we are still wellpositioned to offer an approach to epistemology displaying the full unity in theorizing that we seek. Such an approach makes confirmation fundamental to epistemological theorizing. Such an approach will want to include some account of the relationships between the qualitative notion of confirmation—identifying whether a piece of information confirms, disconfirms, or does neither with respect to a given claim—and the idea of the quantity of confirmation as well as the degree to which a given piece of evidence confirms one hypothesis more than another. At a minimum, we will want to take one of these notions as primitive and clarify the others in terms of it, thereby generating a unified approach to epistemological theorizing which is confirmationfirst. If we also wish to avoid irreducible normativity in our theory, we will need to re-examine the early twentieth-century interest in the nature of confirmation, trying to find a way to elucidate or clarify the fundamental notion of confirmation in 16

This brief summary elides many important issues in the development and defense of a defeasibility theory of knowledge, but one deserves special mention. I ignore in the text the distinction between incremental confirmation (typically expressed in terms of there being some degree of support for a claim) and a threshold concept (typically expressed in terms of the evidence in question showing that a given claim is true). If we adopt an approach which treats incremental confirmation as primary here, we will need an account about how to connect incremental confirmation to threshold confirmation. The standard story here is the Lockean one: threshold confirmation is a perhaps contextually determined level of incremental confirmation that includes all the information possessed by the cognizer. For defense and discussion of this Lockean thesis, see, e.g., Foley (1992); Sturgeon (2008). 17 For discussion and defense, see Kvanvig and Menzel (1990) and Kvanvig (1992, 1996, 2003a, 2005, 2007a, b, and, most recently, 2014).

54

J. L. Kvanvig

non-normative terms. Pursuing those issues, however, takes us beyond the scope of the present essay, so I will only note the importance of the project here and pursue it at another time.

4.6

Conclusion

Our discussion to this point reveals a central lacuna common to both knowledge-first and reasons-first epistemology, a lacuna addressed directly by a confirmation-first approach. The direction our discussion points, then, is this: there is little hope for unity in epistemological theorizing without first saying what confirmation involves. And if one is able to say what confirmation involves, one is well-positioned to characterize the remainder of the epistemological landscape as well, leaving no need for approaches that advertise themselves as alternatives to a confirmation-first approach. Though this argument on behalf of confirmation-first approaches is not decisive—in large part because I have looked only at extant competitors to it—we can at least say that the main alternatives to a confirmation-first approach have some explaining to do.

References Alston, W. (1980). Level-confusions in epistemology. Midwest Studies in Philosophy, 5, 135–150. Alston, W. (2000). Virtue and knowledge. Philosophy and Phenomenological Research, 60(1), 185–189. Bonevac, D., Dever, J., & Sosa, D. (2006). The conditional fallacy. Philosophical Review, 115(3), 273–316. Carnap, R. (1945). The two concepts of probability: The problem of probability. Philosophy and Phenomenological Research, 5(4), 513–532. Carnap, R. (1949). Truth and confirmation. In H. Feigl & W. Sellars (Eds.), Readings in philosophical analysis (pp. 119–127). New York: Appleton-Century-Crofts. Carnap, R. (1971). A basic system of inductive logic, part I. In Jeffrey (Ed.), Studies in inductive logic and probability (pp. 34–165). Los Angeles: University of California Press. Carnap, R. (1980). A basic system of inductive logic, part II. In Jeffrey (Ed.), Studies in inductive logic and probability (pp. 2–7). Berkeley: University of California Press. Dougherty, T. (2011). Evidentialism and its discontents. Oxford: Oxford University Press. Elgin, C. (2007). Understanding and the facts. Philosophical Studies, 132(1), 33–42. Elgin, C. (2009). Is understanding factive? In Haddock (Ed.), Epistemic value (pp. 322–330). Oxford: Oxford University Press. Feldman, R., & Conee, E. (1985). Evidentialism. Philosophical Studies, 48, 15–34. Feldman, R., & Conee, E. (2004). Evidentialism. Oxford: Oxford University Press. Fitelson, B. (2005). Inductive logic. In S. Sarkar & J. Pfeifer (Eds.), The philosophy of science: An encyclopedia (pp. 384–394). New York: Routledge. Foley, R. F. (1992). The epistemology of belief and the epistemology of degrees of belief. American Philosophical Quarterly, 29(2), 111–121. Gibbons, J. (2013). The norm of belief. Oxford: Oxford University Press.

4 Theoretical Unity in Epistemology

55

Greco, J. (1993). Virtues and vices of virtue epistemology. Canadian Journal of Philosophy, 23(3), 413–432. Greco, J. (2000). Putting skeptics in their place. Cambridge: Cambridge University Press. Grimm, S. R. (2011). Understanding. In D. Pritchard & S. Bernecker (Eds.), The Routledge companion to epistemology (pp. 84–94). New York: Routledge. Janvid, M. (2012). Knowledge vs. understanding: The cost of avoiding Gettier. Acta Analytica, 27 (2), 183–197. Janvid, M. (2014). Understanding understanding: An epistemological investigation. Philosophia, 42(4), 971–985. Kelp, C. (2015). Understanding phenomena. Synthese, 192(12), 3799–3816. Keynes, J. (1921). A treatise on probability. London: Macmillan. Khalifa, K. (2011). Understanding, knowledge, and scientific anti-realism. Grazer Philosophische Studien, 83(1), 93–112. Khalifa, K. (2013). Understanding, grasping, and luck. Episteme, 10, 1): 1–1):17. Klein, P. D. (1971). A proposed definition of propositional knowledge. Journal of Philosophy, 68 (16), 471–482. Kornblith, H. (2000). Linda Zagzebski’s virtues of the mind. Philosophy and Phenomenological Research, 60(1), 197–201. Kvanvig, J. L. (1992). The intellectual virtues and the life of the mind: On the place of the virtues in contemporary epistemology. Savage: Rowman & Littlefield. Kvanvig, J. L. (1996). Plantinga’s proper function theory of warrant. In J. L. Kvanvig (Ed.), Warrant and contemporary epistemology (pp. 281–306). Savage: Rowman & Littlefield. Kvanvig, J. L. (2000). Zagzebski on justification. Philosophy and Phenomenological Research, 60, 191–196. Kvanvig, J. L. (2003a). Propositionalism and the perspectival character of justification. American Philosophical Quarterly, 40(1), 3–18. Kvanvig, J. L. (2003b). The value of knowledge and the pursuit of understanding. Cambridge: Cambridge University Press. Kvanvig, J. L. (2005). On denying a presupposition of sellars’ problem: A defense of propositionalism. Veritas, 50, 173–190. Kvanvig, J. L. (2007a). Propositionalism and the metaphysics of experience. Philosophical Issues, 17, 165–178. Kvanvig, J. L. (2007b). Two approaches to epistemic defeat. In D.-P. Baker (Ed.), Alvin plantinga: Contemporary philosophy in focus (pp. 107–124). Cambridge: Cambridge University Press. Kvanvig, J. L. (2009a). Knowledge, assertion, and lotteries. In D. Pritchard & P. Greenough (Eds.), Williamson on knowledge (pp. 140–160). Oxford: Oxford University Press. Kvanvig, J. L. (2009b). Responses to critics. In Haddock (Ed.), Epistemic value (pp. 339–353). Oxford: Oxford University Press. Kvanvig, J. L. (2009c). The value of understanding. In Haddock (Ed.), Epistemic value (pp. 95–112). Kvanvig, J. L. (2014). Rationality and reflection. Oxford: Oxford University Press. Kvanvig, J. L. (2018). McCain on propositionalism. In K. McCain (Ed.), New essays on evidentialism (Synthese Library). New York: Springer. Kvanvig, J. L., & Menzel, C. P. (1990). The basic notion of justification. Philosophical Studies, 59, 235–261. Lord, E. (2018). Epistemic reasons, evidence, and defeaters. In D. Star (Ed.), The Oxford handbook of reasons and normativity. Oxford: Oxford University Press. McCain, K. (2014). Evidentialism and epistemic justification. New York: Routledge. Morris, K. (2012). A defense of lucky understanding. British Journal for the Philosophy of Science, 63(2), 357–371. Neta, R. (2008). What evidence do you have? British Journal for the Philosophy of Science, 59(1), 89–119. Nozick, R. (1981). Philosophical explanations. Cambridge, MA: Harvard University Press.

56

J. L. Kvanvig

Pollock, J., & Cruz, J. (1986). Contemporary theories of knowledge. Totowa: Rowman & Littlefield. Pritchard, D. H. (2005). Epistemic luck. Oxford: Oxford University Press. Pritchard, D. H. (2009). Knowledge, understanding, and epistemic value. Royal Institute of Philosophy Supplement, 84, 19–43. Pritchard, D. H. (2010). Knowledge and understanding. In A. Haddock, A. Millar, & D. Pritchard (Eds.), The nature and value of knowledge: Three investigations (pp. 3–90). Oxford: Oxford University Press. Rohwer, Y. (2014). Lucky understanding without knowledge. Synthese, 191(5), 945–959 Published online: 29 August 2013. Rorty, A. O. (2000). Distinctive measures of epistemic evaluation: Character as the configuration of traits. Philosophy and Phenomenological Research, 60(1), 203–206. Roush, S. (2005). Tracking truth: Knowledge, evidence, and science. Oxford: Oxford University Press. Ryan, S. (1999). What is wisdom? Philosophical Studies, 93(2), 119–139. Ryan, S. (2012). Wisdom, knowledge, and rationality. Acta Analytica, 27(2), 99–112. Ryan, S. (2014). Wisdom. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Winter 2014 ed.). Scanlon, T. M. (2014). Being realistic about reasons. New York: Oxford University Press. Shope, R. K. (1978). The conditional fallacy in contemporary philosophy. Journal of Philosophy, 75, 397–413. Sliwa, P. (2015). IV—Understanding and knowing. Proceedings of the Aristotelian Society, 115 (1.1), 57–74. Sosa, E. (2007). A virtue epistemology. Oxford: Oxford University Press. Sturgeon, S. (2008). Reason and the grain of belief. Noûs, 42(1), 139–165. Tarski, A. (1944). The semantic conception of truth: And the foundation of semantics. Philosophy and Phenomenological Research, 4(3), 341–376. Wedgwood, R. (Forthcoming). Review of Being realistic about reasons, by T. M. Scanlon. Philosophical Quarterly. Whitcomb, D. (2010). Wisdom bibliography. Oxford Bibliographies Online. Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press. Zagzebski, L. (1996). Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge: Cambridge University Press. Zagzebski, L. (2000). Responses. Philosophy and Phenomenological Research, 60(1), 207–219.

Chapter 5

Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief Richard Foley

Abstract Peter Klein’s defeasibility theory has considerable appeal, but it cannot deal with cases in which someone’s beliefs are as accurate as humanly possible but don’t satisfy all the requirements of epistemic justification. Since this person would seem to know things that the rest of us don’t, it’s best to abandon the notion that knowledge requires justification, and to instead think about it in terms of sufficient information, where sufficient information is understood negatively. Whenever an individual has a true belief but seems not to have knowledge, there is some important aspect of the situation about which she lacks true beliefs. This view is conspicuously similar to a defeasibility view. When someone intuitively lacks knowledge despite having a true belief, both views recommend looking for a truth about the situation that the person lacks. They differ only on whether to link the missing truth to a justification requirement. The gap between the two can be further narrowed by noting that justification comes in degrees. One is then in a position to argue that one can have knowledge only if one’s overall belief system meets at least minimal standards of reasonability. Since any belief system that is maximally accurate and comprehensive will meet such standards, the result is an approach to understanding knowledge that retains much of the spirit of the defeasibility theory while avoiding problems that arise from importing an overly strong standard of justification into the conditions of knowledge. Keywords Knowledge without justification · Justification requirement · Accuracy · Maximally accurate · Epistemic information · Information gaps · Alternatives to defeasibility · Knowledge conditions · Epistemology

R. Foley (*) New York University, New York, NY, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_5

57

58

5.1

R. Foley

The Basic Claim

There’s no end to the truths associated with any situation, even the simplest. There’s a blue ball on the table. The ball has a precise circumference and weight. It’s made of rubber. The rubber is a particular shade of blue. The table has its own distinctive properties. It has a certain height, width, and weight, and it’s made of oak. There are historical truths about the situation as well. The ball was purchased at a store in Brooklyn six months ago. The owner of the store is a Jamaican who has three children, two sons and one daughter. The table has three previous owners, and the wood from which it was made came from a tree in Pennsylvania. There are also negative truths. The ball is smaller than a basketball; there’s not a hat on the table, nor is there a cell phone on it. Truths about this and every other situation radiate out in all directions. By contrast, the information we have about any situation is constrained. There are always gaps. But imagine creatures, not as limited as we humans, who have maximally accurate and comprehensive information about some event, or at least as close to this as possible for finite creatures. The event might even be one as significant as the beginnings of life on Earth. There’s almost no information they lack about when earthly life first appeared, what the conditions on Earth were then, what the processes were that produced the first living cells, and so on. These creatures would appear to know a lot more about the origins of life on Earth than we humans do, but, odd as it seems, most contemporary accounts of knowledge imply that this needn’t be so. Why? Because according to them, for beliefs to count as knowledge, special conditions beyond accuracy and comprehensiveness have to be met, ones that the beliefs of these creatures may well lack. What are these conditions? Opinions differ. Many say that if true beliefs are to rise to the level of knowledge, they have to be justified, where beliefs are justified only if they are embedded in an appropriate structure, although, to be sure, there are again different views about what this structure is. Some say it has to be tiered, with a set of basic, self-justifying beliefs providing support that makes additional beliefs justified. According to others, the structure is more web-like. A belief is justified only if it coheres with the rest of what one believes, where these beliefs must also cohere and where coherence in a set of beliefs is (roughly) a matter of each belief being supported by the rest taken together. But since any number of rival belief systems can be mutually supportive, proponents of these accounts point to various other virtues that the systems must have—for example, conservativeness and simplicity. Conservatism is a matter of preserving, to the extent possible, prior opinions, and simplicity is a function of the number of entities, laws, and assumptions implicit in the beliefs. Whatever the required structure, however, there’s no necessity that the beliefs that our hypothetical creatures, despite being maximally accurate and comprehensive, have it. There is no necessity that their beliefs about the origins of earthly life can be adequately defended in terms of a set of self-justifying basic beliefs, nor any necessity that their overall belief systems satisfy conditions of mutual support,

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

59

simplicity, and conservatism. But then, on these accounts of knowledge, these creatures may not know anything about the beginnings of life on Earth, surprising as this sounds given that they would be able to accurately answer virtually any question that might be posed to them about these matters. Other epistemologists find themselves pushed towards the same conclusion but by a different route. They insist that the special conditions needed to turn true beliefs into knowledge are ones about the reliability of the processes that produced the beliefs. When asked what makes a process reliable in the required sense, they reply that it’s not enough that it happens to generate true beliefs in the actual situation. It also has to be the case that it tends to generate mostly true beliefs in relevantly close possible situations. What counts as “relevantly close” turns out not to be so easy to say, but however spelled out, it’s again possible for our imaginary creatures to have no knowledge about the beginnings of life on Earth. Why? Because despite being maximally accurate and comprehensive, their beliefs may have been produced by a combination of curious processes and unlikely events that in most other situations would not have produced accurate beliefs but in this case did. My recommendation, in contrast, is to go with the obvious and reverse the direction of these arguments. Because these creatures have such detailed, exhaustive information about the origins of life on Earth, they know a lot more about these matters than we do, and hence there must be something amiss with any account of knowledge that finds itself forced to deny this. I’ll be returning to this claim later and qualifying it a bit, but I want first to take a quick look at some recent history in epistemology, which will help set the stage for the later discussion.

5.2

Some Recent History

In his landmark paper “Is Justified True Belief Knowledge?”, Edmund Gettier used the following story to illustrate that justification added to true belief isn’t sufficient for knowledge. Smith believes that the person about to be hired for a new position has ten coins in his pocket because he believes that Jones, who has ten coins in his pocket, will get the job. Smith is unaware, however, that it is he, not Jones, who will be hired, and is also unaware that he too has ten coins in his pocket. His belief that the new hire has ten coins in his pocket is thus true, but he doesn’t know this to be case (Gettier 1963). This and another case that Gettier presented convinced most philosophers that justified true belief isn’t enough for knowledge, but the cases also changed how epistemology was done. In hopes of eliciting additional intuitions that would further clarify how best to think about knowledge, other epistemologists became occupied with narrating their own tiny stories in which a subject has a true belief but seems to lack knowledge. They became storytellers, and a number of the stories achieved the status of stock cases that everyone in the field is now expected to know.

60

R. Foley

One such story is about a tourist in farm country who is charmed by the old barns he is seeing. He stops his car on the side of the road to admire the latest barn he has happened across. He’s unaware that the local tourist board has populated the region with highly realistic façades of old barns and that up until now he has been seeing façades, not real barns. By chance, however, he has stopped in front of one of the few genuine old barns remaining in the area. As he looks out of his car window, he believes that he is looking at an old barn, and he is. Yet, his true belief doesn’t rise to the level of knowledge (see Goldman 1976; Ginet 1988). Here’s another such story. Beth has bought a ticket in a lottery of a million tickets. The winning ticket has been chosen but not yet announced. Beth believes that her ticket is not the winner, and she is correct, but still, she doesn’t know this to be the case because of the non-zero probability that her ticket could win (Hawthorne 2004; Vogel 2004). These and numerous other stories like them posed the question, what has to be added to true belief in order to get knowledge? Prior to Gettier’s article, the standard answer had been that what is needed is something like an argument in defense of the belief—a “justification”, to use the term common in epistemology—but because Gettier’s examples convinced most philosophers that justification on its own is not enough, the relevant question became, what else is needed? Many said that a distinct kind of justification is required. They differed about its precise nature, but among the most influential answers was one articulated by Peter Klein. In his explication of a defeasibility theory of knowledge, he said that the justification needed to be indefeasible, meaning that there cannot be any truths that would defeat it (Klein 1981).1 I’ll return to this suggestion in a moment, but other epistemologists, struck by the fact that most of us would be hard-pressed to provide adequate defenses of much of our everyday knowledge, came up with their own stories to illustrate that something less explicitly intellectual than justification is required for knowledge—something such as reliability (Goldman 1988), or truth tracking (Nozick 1981), or proper functioning (Plantinga 1993), or some variation of these ideas.2 All these suggestions, despite their differences, share an assumption. They all assume that what had to be added to true belief in order to get knowledge is something related to true belief but distinct from it—non-defective justification, undefeated justification, reliability, truth tracking, proper functioning, or whatever. My counter suggestion is that what has to be added are more true beliefs. In particular, more true beliefs about the situation in question. The best explanation for why the creatures considered earlier have knowledge about the origin of life on Earth

1

Also see Audi (1993), Lehrer (1974), Pollock (1986), and Swain (1981). See, for example, Armstrong (1973), Dretske (1981), Lewis (1996), and Ernest Sosa (1991), especially Chapters 13–16. 2

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

61

is that they have such maximally accurate and comprehensive beliefs about it. There’s no important information they lack. Correspondingly, the best explanation for why the characters in the above stock stories lack knowledge is that, by contrast, they do lack important information about the situation in question. They’re unaware of key truths. Why does Smith not know that the person about to be hired has ten coins in his pocket, and the tourist not know that he’s looking at an old barn, and Beth not know that her lottery ticket is not the winner, even though their beliefs are true? The most compelling answer is the simplest. Smith isn’t aware that he, not Jones, will get the job; the tourist isn’t aware that there are lots of barn façades in the area he is touring; and Beth isn’t aware which of the other tickets have lost the lottery and which has won. Post-Gettier epistemology overflows with stories in which the subject has a true belief but seems to lack knowledge. All these stories, from the bizarre to the ordinary, can be analyzed in the same way. Each focuses attention on some aspect of the situation about which the main character of the story is ignorant, and each is told, so as to suggest that this gap in the character’s information is important. To the degree that we the audience are convinced that the missing information is in fact significant, our intuition is that the character, despite having a true belief, doesn’t have enough of the right information for knowledge. Something important is missing, and that’s why the subject lacks knowledge. There’s an important gap in his or her information. What makes a gap important? There’s no single test for determining when a piece of missing information is important, any more than there is a single, simple test for the importance of anything else—important news, important people, important discoveries, etc. All our values potentially come into play when assessing the importance of something, which means it’s never an easy matter to determine what’s important and what isn’t. So too it is for important information. The key point for my purposes here, however, is that it is our views about the importance of missing information, however formed, that drive our intuitions about knowledge. Test this out on yourself. Select any story in the literature in which the subject has a true belief but the author of the story claims that the subject lacks knowledge. If you share the author’s intuition, look for some feature of the situation, as it is described, about which the subject is ignorant and which you think constitutes an important gap in the subject’s awareness of the situation. My contention is that you will be able to find such a gap, which then can be used to explain why you have an intuition that the subject lacks knowledge.3

3

For more details on this way of thinking about knowledge, see Foley (2012).

62

5.3

R. Foley

Knowledge, Justification, Reliability

Defeasibility theorists make a strikingly similar claim. When dealing with stories in which a subject seems to lack knowledge despite having a true belief, they too recommend looking for a truth about the situation of which the subject is unaware, but because they are committed to justification being a necessary condition of knowledge, they link the subject’s ignorance of this truth with the justification requirement. The individual lacks knowledge because the missing truth, if believed, would defeat her justification for the target belief. But a simpler explanation is possible. The subject doesn’t know because there are important gaps in her information. Earlier I imagined creatures that have as complete as possible information about the beginnings of life on Earth and suggested that it would be ludicrous to deny that they have knowledge about these beginnings. After all, if we were able to question them, they would be able to answer all our questions about when and how life began and would be able to do so in as detailed a manner as we wanted. Yet, defeasibility accounts and most of the other familiar approaches to understanding knowledge are forced to admit that these creatures might well lack knowledge. Let’s now go back and take another look to see how and why they are pushed into this corner. Begin with reliability accounts. Their problems in dealing with this kind of case arise because there’s nothing inconsistent in imagining that the beliefs of these creatures, despite being maximally accurate and complete, were produced by some combination of extraordinary processes and events that in most relevantly similar circumstances would not have produced accurate beliefs but did so in this case. If our creatures were unaware of how they came to have their beliefs, this gap in their second-order beliefs would be available to potentially explain why they lack knowledge, but let’s stipulate that there is no such gap. They are fully aware of what processes and events led them to believe what they do, fully aware as well that these processes and events normally couldn’t be relied on to produce accurate beliefs, but fully aware also of how and why in their particular circumstances, these processes did result in maximally accurate and comprehensive beliefs. If so, I say that they know. Moreover, they know even if their second-order beliefs, albeit completely accurate, were themselves produced in ways that in relevantly similar circumstances wouldn’t have been accurate. What about justification-based approaches? According to coherentists, an individual’s beliefs about an issue are justified only if they stand in mutually supportive logical and probabilistic relations with the rest of what the individual believes, and in addition the overall belief system has various other virtues, such as being as being simpler and more conservative than rival possibilities. There can be different views about how to unpack the notions of mutual support, simplicity, and conservativeness, but whatever these details are, there’s no inconsistency in imagining that our creatures’ beliefs about the beginnings of life on Earth are maximally accurate and comprehensive without it also being the case that their overall belief systems are the simplest and most conservative set of mutually supportive beliefs that they could have.

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

63

Foundationalist accounts fare no better. They maintain that beliefs are justified only if they are appropriately supported by basic beliefs. There are competing proposals about both what makes a belief basic and what kinds of support relations are needed, but however the characteristics of basicness and the support relations are fixed, there will be no inconsistency in telling the story about our creatures in such a way that, by stipulation, their beliefs about how life came to exist on Earth—albeit completely accurate and comprehensive—cannot be adequately defended in terms of the required kinds of basic beliefs and support relationships. The story about our imaginary creatures illustrates that neither justification-based nor reliabilist views of knowledge can be quite right, but still, one might wonder whether they are completely off the mark. Might not there be some looser, less exacting relationship between knowledge on the one hand and justification and reliability on the other? If so, perhaps there’s a fallback position in the general neighborhood of these views. As a way of beginning to explore this possibility, consider the following question about the beliefs of our hypothetical creatures: Is it really possible for their beliefs to be maximally accurate and comprehensive and yet not be at least relatively reasonable ones for them to have? In using the expression “relatively reasonable,” I’m relying on the widely accepted assumption that opinions, behavior, policies, and so forth can be more or less reasonable. Reasonability comes in degrees. When this assumption is kept in mind, it fairly invites the question, is there some less demanding standard of reasonability that would inevitably be met when beliefs about an issue are maximally accurate and comprehensive? And if the answer is “yes,” it invites the further question whether in more ordinary cases of knowledge a looser standard of reasonability is likewise always met. After all, if reasonability comes in degrees, if it’s not all or nothing, why shouldn’t this be acknowledged in treatments of knowledge? Moreover, perhaps there’s an analogous fallback position to be reached with respect to reliability. Namely, even if it’s possible for beliefs to be maximally accurate and yet not be caused by processes that meet the specific conditions that reliability theorists have standardly insisted upon, might not such beliefs nonetheless meet a less stringent standard of reliability? If so, then the point again might be extended to more ordinary cases of true belief that rise to the level of knowledge. Perhaps they, too, inevitably meet some modest standard of reliability. Let’s reconsider the issues with these possibilities in mind.

5.4

Knowledge, Justification, Reliability Reconsidered

Reliabilists go to considerable lengths to insist that processes that in fact produce only true beliefs aren’t necessarily reliable in the way required to turn true beliefs into plausible candidates for knowledge (Goldman 1988, 106–109). This is why the beliefs of our creatures, despite being maximally accurate and complete, might not qualify as knowledge. On the other hand, it may be that the thinly described story

64

R. Foley

about these creatures cannot be filled in without assuming an orderly universe that they have interacted with in systematic ways. If so, the curious processes that produced their beliefs might nonetheless satisfy a more relaxed standard of reliability. In thinking about this possibility, recall Donald Davidson’s views about situations of “radical interpretation,” where it’s an open question whether the subjects being interpreted have beliefs at all. He maintained that it’s not possible to attribute beliefs to the subjects without assuming that the salient features of their environment are in general both the causes and the contents of their utterances and the beliefs these utterances express (1973–74; 1986). Davidson hoped to use this point to construct an argument against the possibility of any belief system being fundamentally in error and to in this way counter skeptical worries. Even if he overstated the relevance of his point for dealing with skepticism (Foley 1993, especially Chapter 2), it’s still an important point on its own, and important in particular for the issues under consideration here. For it suggests that in order for our creatures to have beliefs at all, much less to have maximally accurate and comprehensive ones, there must have been a critical mass of law-like interactions between them and their surroundings. This may be a wedge for introducing a modest notion of reliability that has to be present despite the strange circumstances that produced their beliefs. Even if all this is so, it wouldn’t affect the main point I’ve been defending, since the simplest and best explanation of why the creatures know isn’t that their beliefs are true and the products of processes that are reliable in a barebones sense. After all, all sorts of true beliefs would meet these conditions without being plausible candidates for knowledge. What such an argument does suggest, however, is the possibility of there being a link between knowledge and reliability after all, even if it’s a considerably stripped-down version of the one that reliabilists have standardly insisted upon. Rather than dwell on this possibility, however, I want to move on to a closely related but, as I see it, more intriguing possibility. For just as it may be possible to argue that altogether accurate and comprehensive beliefs inevitably satisfy modest standards of reliability, so too it may be that they always meet modest standards of reasonability. Davidson’s views about radical interpretation are again potentially relevant. He argued that because beliefs come and go in “clumps,” not atomistically, in any situation of radical interpretation, the attributed beliefs must largely cohere with one another (Davidson 1986). The nature of belief itself ensures that this is so, which in turn implies that the beliefs of our imaginary creatures about the origins of life on Earth must likewise be part of an overall system of beliefs that is largely coherent. The qualifier “largely” is noteworthy. It’s a recognition that the coherence comes in degrees. So, the conclusion of the argument here is only that the belief system of our creatures, like any other belief system, must be such that its elements are sufficiently mutually supportive. Otherwise, it wouldn’t qualify as a belief system at all. Once again, this conclusion wouldn’t affect the main point I’ve been arguing for. The best explanation of why our imaginary creatures know so much about the origins of life on Earth is not that their beliefs are true and part of an overall belief

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

65

system that is largely coherent. For there are numerous beliefs that meet these conditions without being credible candidates for knowledge. The view does have the potential, however, to begin narrowing the differences between the approach I’ve been advocating and more traditional, justification-based approaches to knowledge. This narrowing can be further accelerated by noting that even if the nature of belief on its own doesn’t ensure high degrees of coherence, the world may do so. More exactly, it may do so to the extent that the beliefs are also accurate and comprehensive. For then the claim about our hypothetical creatures can be that it’s not possible for their beliefs about the origins of life to be maximally accurate and comprehensive without accurately reflecting whatever orderly structure the world actually has, which in turn ensures that their beliefs will also be highly coherent with one another. It’s to this narrowing that I’ll now turn.

5.5

Loosening the Link

I talked earlier about severing the link between knowledge and justification, but once it’s acknowledged that justification along with its sibling notion of reasonability are not all-or-nothing affairs, that they come in degrees, it becomes plausible to speak of loosening the link as opposed to altogether severing it. Begin with the above claim that if one’s beliefs about an issue are close to being wholly accurate and comprehensive, they inevitably will also display a high degree of coherence with one another. This still doesn’t mean that they are necessarily the most reasonable beliefs to have. They may not fit with the rest of what one believes as ideally as they might, and even if they do cohere well with one’s other beliefs, as mentioned earlier, any number of rival overall belief systems can be equally coherent even though, given the circumstances, some are more reasonable than others. On the other hand, having coherent opinions about an issue isn’t immaterial either. It indicates that one’s views about the issue fit together well. There might be ways in which the opinions nonetheless could be improved, even pretty obvious ways, but likewise they also could have been much worse. One’s beliefs about the issue might have been strongly at odds with one another, and as such positively unreasonable ones to have, but they aren’t. They meet at least minimum standards of reasonability. This point might then be extendable to everyday cases of knowledge as well. One has knowledge about a particular situation only if there aren’t important gaps in one’s information, but if one’s beliefs about the situation are accurate and comprehensive enough to satisfy this condition, they must fit together well. Once again, there might still be some fairly obvious problems with them. They may not cohere with the rest of what one believes as well as they could. But even if they are less reasonable than they could be, they still may be good enough, given the circumstances, to meet at least minimal standards of reasonability. The result is a view that has a good deal in common with the defeasibility approach to knowledge championed by Peter Klein and others. According to both

66

R. Foley

approaches, whenever one has a true belief but lacks knowledge, there’s something significant about the situation that one isn’t aware of, that one doesn’t “get.” Likewise, according to both, whenever one does have knowledge about a situation, one’s beliefs about it also meet a certain standard of reasonability. The two approaches differ on how demanding this standard is, but an even more fundamental difference is that on the view I’m recommending, the primary focus, when dealing with questions of knowledge, should be on whether the beliefs of would-be knowers are accurate and comprehensive enough—in particular, on whether there is something important they aren’t aware of. When beliefs meet this condition, however, they will also satisfy modest standards of reasonability. So, there’s a link between knowledge and reasonability, but it’s a derivative one. It falls out of the requirement that the beliefs be accurate and comprehensive. Much more would need to be said about what these modest standards are and the ways in which beliefs that meet them can depart from what is most reasonable given the circumstances, but for purposes here, the point I want to stress is a general one about the advantages of developing an epistemology that fully acknowledges there are fine gradations in the reasonableness of opinions and then uses this acknowledgement to relax the tie between knowledge and reasonability. Such an epistemology is better positioned to do justice to the complex realities of our intellectual lives. Among these realities is that we are no more settled and undivided intellectually than we are practically. It’s a common observation that our wants and preferences can change with the circumstances, and they often enough can be in some tension with one another. Even so, we are usually able to live with the fluidity and strains. We make do. We make do because we don’t need perfectly coordinated and stable preferences, only ones that are steady enough and coherent enough to allow us to get on with our everyday lives adequately. So too it is with opinions. They aren’t always stable. With respect to some issues, we can be doxastically fickle, believing at one moment what we didn’t believe previously and what we shortly may again not believe.4 In addition, there are almost always tensions to be found amongst our opinions about various issues. This isn’t the exception. It’s the rule. There are endless gradations in how thorough and reflective we are in arriving at our views. So, it shouldn’t be surprising that there are equally fine gradations with respect to the coherence and reasonableness of the resulting opinions. Even when things go well, the support that our opinions enjoy from the rest of what we believe comes in different degrees. And when it happens, as it invariably does, that some of what we believe is in tension with other matters we consider settled, this again can be a matter of degree. The tensions can be of varying intensities, some being deep and serious and others being more superficial and minor. An unfortunate byproduct of some justification-driven approaches to knowledge is that they can encourage overly simplistic assessments of opinions. Opinions either are justified or not; if they are, they can be plausible candidates for knowledge; if not, they cannot be plausible candidates. Distinguishing spontaneous and other largely

4

Jane Friedman makes this point (Friedman forthcoming a).

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

67

unreflective beliefs from reflective ones is a step in the right direction (see, e.g., Sosa 2011), but it’s only a step—only a step because it can also be more or less reasonable to be reflective about an issue. Sometimes the context and issue are such that it’s not appropriate to spend much time or energy in gathering evidence on it, deliberating about the evidence, and considering how well one’s presumptive opinion about it fits with what else one believes. And even when it is reasonable to devote a good deal of time and energy to inquiry and deliberation, there are still endless gradations in how seriously to do so, and correspondingly, there are fine gradations in the reasonableness of the subsequent opinions. Complexities of this sort have repercussions for both the theory of knowledge and the theory of rational belief. For the theory of knowledge, the complexities illustrate why it’s important to resist any temptation to insist that if one is to have knowledge about an issue, one’s opinions about it have to be purged of tensions with the rest of what one believes. This is far too rigid. As with wants and preferences, we can make do with an overall set of opinions that we realize don’t fit perfectly with one another. Moreover, this is compatible with our having knowledge about lots of matters as long as our opinions about them are serviceable enough, in particular, accurate and comprehensive enough that we don’t lack critically important information. With respect to the theory of rational belief, the complexities indicate why there can be no categorical requirement to unearth problems in our intellectual practices and the opinions they generate, or for that matter even a requirement to always ameliorate problems once we become aware of them. Whether the problems are ones of inconsistency, or ones having to do with the use of less than optimal evidence, or even ones involving biases, the affected opinions may be about matters that are of little or no significance, in which case it won’t be worth the time and effort to identify and correct flaws in them. We’ll have more pressing matters to attend to. Nor would there be such a categorical requirement if we were to imagine ourselves as purely intellectual beings, all of whose time and energy could be devoted to having accurate and comprehensive opinions. There still would be too many issues, too little time, and too little capacity to be able to investigate and deliberate about everything. As we are finite creatures, questions of resource allocation would still arise. Choices would still need to be made about what it’s important (or unimportant) to have accurate beliefs about. Consider an extreme example. Suppose someone were to suggest that I should spend all my waking hours in the houses of friends, counting the number of grains of salt in their saltshakers. And the further suggestion is that if I eventually run out of saltshakers (or friends), I should then go on to counting the number of leaves on trees in Central Park. By following this advice, I would be acquiring information I now lack, but the project would be ill conceived and even disturbing, since the value of the information being acquired is so out of whack with the time and effort expended in obtaining it. Moreover, it doesn’t help to imagine all my non-intellectual ends somehow being immediately met so that I am free to devote all my time to intellectual pursuits. The project would still be ill conceived, since the truths being obtained would still be worthless—worthless because they aren’t connected in any apparent way with any of my practical ends, but in addition because they have no

68

R. Foley

apparent intellectual value for me, either intrinsic or instrumental. I’m not curious about them, and they wouldn’t be helpful to me in forming opinions about issues that I am curious about. Moreover, this isn’t an isolated problem limited to truths about grains of salt and leaves on trees. There’s no end to the valueless propositions about which I could acquire information. Nor is there even an end to the valueless propositions supported by information I already have in hand. If I have information indicating that p is highly likely to be true, then the disjunction ( p or q1) is also likely to be true. Ditto for ( p or q2), ( p or q3),. . ., ( p or qn), regardless of what q1, q2, q3,. . ., qn are. But just because I have information indicating that these propositions are likely to be true, there is no requirement on me to believe each. If I failed to believe one or more of them, this needn’t indicate that I have been in any way unreasonable.5 Similarly, there’s no general requirement to revise opinions that my information indicates are unlikely to be true. For, just as there are worthless truths, so too there are harmless falsehoods—ones that do no significant intellectual or practical harm and whose correction would produce no good. I have been arguing that questions of knowledge and considerations of what is important information are inextricably linked. If one has a true belief p but lacks knowledge, there’s always some important truth about the situation that makes p true but which one fails to believe or take into account. The point now is that in dealing with issues of what it is reasonable to believe, considerations of what is of value again have to be introduced. One cannot get around this by imagining that we are purely intellectual beings. In actuality, of course, we are not such beings. We have all sorts of ends—practical and ethical as well as intellectual. These ends give direction to inquiry and deliberation. They determine which issues it makes sense to spend time on. They also establish constraints on how much time and effort we spend. Because a variety of such ends are in play in any real-life situation, the pressing intellectual issues confronting us are always ones about how to be appropriately careful and thorough, where what is appropriate varies with the importance of the issue, the particularities of the situation, our social roles, and any number of other such factors. If accounts of knowledge and rational belief are to be relevant to our intellectual lives, they too must operate within this territory. It’s a territory where the reasonability of what we believe is a function of a complex nexus of information, values, and needs. It is a territory where, as the stakes of having accurate opinions about a set of issues go up, so too do the standards that must be met in order to have responsible opinions, but where the reverse is also the case. As the stakes go down, it’s reasonable not to spend so much time and so many resources in inquiry and deliberation. As a result, it’s perfectly possible for responsible inquirers to have opinions which they realize don’t mesh well with other opinions they have. And, it’s possible for such tensions to co-exist with knowledge. To have knowledge, one’s beliefs about the matter in question have to be accurate and comprehensive enough

5

Compare with Harman (1988) and Friedman (forthcoming b).

5 Accurate Enough, Comprehensive Enough, and Reasonable Enough Belief

69

that one isn’t missing crucial information. When this condition is met, one’s beliefs about the matter will be mostly coherent and hence also fairly reasonable, but they don’t have to fit perfectly well with everything else one believes, and moreover one can be aware of this. The direct tie between knowledge and reasonability is thus much looser than is often assumed. On the other hand, like friends of friends, the two do have indirect ties with one another through “third parties.” In particular, both knowledge and reasonable opinion are sensitive to the full range of our values—pragmatic, ethical, intellectual, etc.—and sensitive also to the particularities of the situations in which we find ourselves. These third-party ties produce secondary relationships between knowledge and reasonable opinion that, like a magnetic force, tend to draw the two toward one another. Here’s one example of the attraction. The more important the issue, the higher are the standards that must be met if we are to have responsible opinions about it. There are greater demands on us to be thorough in our evidence-gathering and deliberations, which when all goes well increase the accuracy and extent of our information about the issue, which in turn are the key ingredients of knowledge. Or we can come at the attraction from the reverse direction. The more important an issue is, the greater the pressure there is on us to have knowledge about it, but the same factors that create this pressure also create pressures to have something more than minimally reasonable opinions about it. These pressures push the standards of responsible belief upward. The most general lesson here, however, is an altogether familiar one, but in philosophy as in life, it’s often useful to be reminded of the familiar. Namely, insofar as we aspire to be responsible believers and knowers, the target shouldn’t be perfection. With respect to the management of opinions, as with much else, the ideal can be the enemy of the good. The focus should be on having opinions about issues that are good enough for the circumstances. They should be accurate and comprehensive enough in the sense that we are aware of the important information, where importance is a function of the kind of issue in question and the context; and in addition, our opinions should be reasonable enough in the sense that we have been appropriately thorough and careful in gathering evidence and reflecting on the issues, where what counts as appropriate is again dependent on the relative importance of the issues and the context.

References Armstrong, D. M. (1973). Belief, truth, and knowledge. Cambridge: Cambridge University Press. Audi, R. (1993). The structure of justification. Cambridge: Cambridge University Press. Davidson, D. (1973–1974). On the very idea of a conceptual scheme. Proceedings and Addresses of the American Philosophical Association, 17, 5–20. Davidson, D. (1986). A coherence theory of truth and knowledge. In E. Lepore (Ed.), The philosophy of Donald Davidson, perspectives on truth and interpretation (pp. 307–319). London: Basil Blackwell.

70

R. Foley

Dretske, F. (1981). Knowledge and the flow of information. Cambridge, MA: MIT Press. Foley, R. (1993). Working without a net. New York: Oxford University Press. Foley, R. (2012). When is true belief knowledge. Princeton: Princeton University Press. Friedman, J. (Forthcoming-a). Inquiry and belief. Forthcoming in Noûs. Friedman, J. (Forthcoming-b). Junk beliefs and interest driven epistemology. Forthcoming in Philosophy and Phenomenological Research. https://doi.org/10.1111/phpr.12381. Gettier, E. L. (1963). Is justified true belief knowledge? Analysis, 23, 121–123. Ginet, C. (1988). The fourth condition. In D. F. Austin (Ed.), Philosophical analysis: A defense by example (pp. 105–117). Dordrecht: Kluwer. Goldman, A. (1976). Discrimination and perceptual knowledge. Journal of Philosophy, 73(20), 771–791. Goldman, A. (1988). Epistemology and cognition. Cambridge, MA: Harvard University Press. Harman, G. (1988). Change in view. Cambridge: MIT Press. Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press. Klein, P. (1981). Certainty: A refutation of Scepticism. Minneapolis: University of Minnesota Press. Lehrer, K. (1974). Knowledge. Oxford: Oxford University Press. Lewis, D. (1996). Elusive knowledge. Australasian Journal of Philosophy, 74(4), 549–567. Nozick, R. (1981). Philosophical explanations. Cambridge, MA: Harvard University Press. Plantinga, A. (1993). Warrant: The current debate. Oxford: Oxford University Press. Pollock, J. (1986). Contemporary theories of knowledge. Totowa: Rowman & Littlefield. Sosa, E. (1991). Knowledge in perspective. Cambridge: Cambridge University Press. Sosa, E. (2011). Reflective knowledge: Apt belief and reflective knowledge (Vol. 2). Oxford: Oxford University Press. Swain, M. (1981). Reasons and knowledge. Ithaca: Cornell University Press. Vogel, J. (2004). Speaking of knowledge. Philosophical Issues (Epistemology), 14, 501–509.

Chapter 6

Knowledge, despite Evidence to the Contrary Rodrigo Borges

Abstract Can new evidence against what one knows defeat one’s knowledge? It depends on whom you ask. Strong views of knowledge claim that knowing cannot be defeated by counterevidence. Weak views of knowledge claim that knowing can be defeated by counterevidence. I discuss recent versions of those views of knowledge—Maria Lasonen-Aarnio’s and Peter Klein’s, respectively—and argue that both are wanting. Strong views have to account for the defeat intuition (i.e., the

I wrote this chapter to honor Peter Klein, the best philosopher I have ever met, and one of the best people I have ever met. Peter was my thesis advisor at Rutgers; he is also a mentor, and the best friend my work has ever had. Peter made me believe I could (sometimes) do philosophy, and that philosophy was something worth doing. During the eight years I was at Rutgers, Peter and I met once or twice every month during the academic year to discuss my work. We discussed every issue in this chapter (and more, much more) innumerable times over that period. We didn’t always agree, but Peter always taught me something valuable. I cherish those lessons, and I will be forever grateful to Peter for having given them to me This chapter benefıted from the feedback of many people. Some of them read different drafts and were kind enough to send me comments. Among those, I am especially thankful to Peter Klein, Claudio de Almeida, Joao Fett, Felipe Medeiros and Gregory Gaboardi. Others let me pick their brains in conversations about some of the issues in the chapter. Those include Mike Veber, Nicola Salvatore, Fred Adams, and Rogel de Oliveira. Many audience members asked really smart and insightful questions about the presentations based on different parts of this chapter. I am especially thankful to the audiences at the following events: a meeting of the East Carolina University Philosophy Club, the 1st Colloquium on Contemporary Debates in Epistemology, the XIII Epistemology Colloquium, and the XVII Meeting of the National Association of Graduate Programs in Philosophy. I also gave a version of this paper during a visit to East Carolina University in January 2018. I am grateful especially to the faculty in the audience: George Bailey, Josh Collins, Nicholas Georgalis, Jay Newhard, and my host Mike Veber were particularly generous with their comments. I am grateful for their help. Part of the research for this chapter was funded by the Sao Paulo Research Foundation through a post-doctoral research grant. I am grateful for their support. Finally, many thanks to the two other editors for this volume, Cherie Braden and Branden Fitelson, for their support and patience as I failed to meet (almost) all the deadlines for this chapter. Cherie also sent me detailed written comments on a previous draft and saved me from making many mistakes. I am grateful for her help making this chapter better. R. Borges (*) Department of Philosophy, University of Florida, Gainesville, FL, USA © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_6

71

72

R. Borges

inclination some of us have, at least some of the time, to think that knowledge is defeated in particular cases). Weak views have to solve the indeterminacy problem (i.e., explaining, in a principled way, when and why knowledge is defeated). I argue that Lasonen-Aarnio’s view cannot account for the defeat intuition, and that Klein’s view cannot account for the indeterminacy problem. I then offer a novel version of the strong view of knowledge, one that improves on the version previously discussed. Keywords Defeasibility · Knowledge defeaters · Counterevidence · Knowledgefirst · Epistemic defeat · Psychological defeat · Lasonen-Aarnio · Klein · Epistemology

Plato claimed, in the Meno (97a–98c), that knowledge is more valuable than true belief in virtue of it being more stable than true belief. Knowledge, Plato argued, is a more reliable guide to action than true belief: the traveler who knows the way to Larissa is less likely to make a wrong turn (or to give wrong directions) than the traveler who has a mere true belief of the way to Larissa, because known truths are— and merely believed truths aren’t—tethered to one’s mind1 by an explanation of why they are true (aitias logismos).2 In other places (e.g., in Republic 412e ff.) Plato also seems to have claimed that a truth, once tethered, is not easily (or perhaps ever) untethered by counterevidence.3 The main goal of this chapter is to understand when, if ever, knowledge is, to use Plato’s expression, untethered by evidence suggesting that what one knows is false. The idea that knowledge is not defeated by counterevidence enjoys some popularity among contemporary philosophers. For example, Norman Malcolm (1952, 185–186) wondered whether there is any evidence that should be allowed to lower his confıdence in a proposition he says he knows; that there is an ink-bottle in front of him. Malcolm concluded his rumination thus:

1

Plato’s discussion of the difference in value between knowledge and true belief famously relates knowledge and true belief to Daedalus statues, which, if not tethered, were said to run out and escape their base. A true belief, argued Plato, is like an untethered Daedalus statue—beautiful but prone to escaping its owner. An instance of knowledge, on the other hand, is like a Daedalus statue that is tethered—beautiful and not likely to escape its owner. 2 I am aware of the pitfalls of trying to translate ‘aitias logismos,’ but I will not discuss them here. For an excellent discussion of some of the main exegetical issues surrounding the translation this expression, see Fine (2004). 3 For example, in Republic (534b), Plato seems to argue that knowledge is immune to counterevidence, while mere true belief is not, because the latter, but not the former, may be essentially based on false beliefs. Because mere true beliefs can depend essentially on false beliefs, they may be undermined by new truths; this is not the case with knowledge, which depends exclusively on what is true. See Fine (2004, §viii) for a discussion of this point.

6 Knowledge, despite Evidence to the Contrary

73

. . . I should regard nothing as evidence that there is no ink-bottle here now. . . [This] describes my present attitude towards the statement that here is an ink-bottle. . . . My present attitude toward that statement is radically different from my present attitude toward . . . other statements (e.g., that I have a heart). I do now admit that certain future occurrences would disprove the latter. Whereas no imaginable future occurrence would be considered by me now as proving that there is not an ink-bottle here. These remarks are not meant to be autobiographical. They are meant to throw light on the common concepts of evidence, proof, and disproof. (Malcolm 1952, 181–182; emphasis his)

According to Malcolm, the case highlights the strong sense of ‘know.’ He contrasts a strong sense of ‘know’ with a weak sense of that word. Central to Malcolm’s distinction is the idea that the weak sense of ‘know’ allows for the possibility of refutation, while the strong sense does not. For example, when one says that one knows that 92 times 16 is 1,472, but one is not sure this is the case (say, because one did the calculation in one’s head), ‘know’ is being used in its weak sense; when one says that one knows that 2 plus 2 is 4, one is sure this is the case and ‘know’ is being used in its strong sense.4 I will make a similar suggestion about knowledge below. However, unlike Malcolm, who focused on the ordinary use of ‘know,’ my focus will be on the concept of knowledge.

6.1

Headache and Weather

We can easily reproduce cases that seemingly show that knowledge may be retained in the face of counterevidence. Consider: HEADACHE: Liz has a throbbing headache. On her way to the medicine cabinet she bumps into her father, a psychiatrist, who proceeds to tell Liz that she does not have a headache. He argues that Liz’s subconscious mind is playing tricks on her in order to allow her to deal with the sudden death of her dog, Fifı.5

4 Malcolm is making another distinction with his case. Although he is primarily interested in the synchronic features of the case (i.e., he is interested in what it is now rational for him to think about possible counterevidence to his knowledge that there is an ink-bottle in front of him), he acknowledges that one could also take an interest in the diachronic features of the case (i.e., one could be interested in what Malcolm would do, in the future, if he were confronted with evidence suggesting there was no ink-bottle in front of him). Malcolm says that he is not talking about the diachronic features of his situation, and he concedes that he cannot confıdently predict now how he would behave in the future were he to encounter evidence suggesting that there is no ink-bottle in front of him. For all he knows, says Malcolm, he would ‘become mad’ or ‘fall into a swoon’ if someone he trusts were to say that ink-bottle hallucinations are quite common for people in his circumstances. See Kripke (2011) and Borges (2015) for a discussion of how heeding this distinction may prevent one from misunderstanding the charge of dogmatism that is usually leveled against strong views of knowledge such as the one discussed by Malcolm. 5 I don’t mean this case to presuppose that headaches are luminous states (i.e., states that are such that, if one is in one of them, then one is in a position to know one is in them). Maybe headaches are luminous; maybe they aren’t. The claim that they are is irrelevant for my case. Clearly, we are sometimes in a position to know we have a headache; I mean this case to be about one of those times. I mention this because Timothy Williamson (2000) has persuasively argued that there are no luminous states. His argument does not preclude the claim that one is sometimes in a position to know one is in a certain mental state, though. But see Greenough and Pritchard (2009) for criticism of Williamson’s argument against luminosity.

74

R. Borges

Although the testimony of Liz’s father provides her with counterevidence to her knowledge that she has a headache, that does not defeat the justifıcation she has for believing she has a headache. The case illustrates Malcolm’s point that (strong) knowledge is not defeated by counterevidence. Here’s another case that seems to make the same point: WEATHER: Liz is looking at a sunny afternoon through the window in her study while listening to a DVRed Weather Channel forecast from last night. The forecaster said, then, and with a fair amount of certainty, that it would be raining right now (i.e., at the time Liz is listening to the recording).

The fact that the forecaster said it would be raining at the time Liz is watching the recording amounts to counterevidence to Liz’s knowledge that it is sunny outside. Moreover, she knows that the forecaster said that it would be raining by the time she is listening to the recording. This does not seem to change the fact that Liz is justifıed in believing that it is sunny outside, and, since it is true that it is sunny outside, she knows that it is.6 In short, Liz seems entitled not to change the confıdence she places in the proposition that it is sunny outside even though she encountered counterevidence for the claim that it is. Like Plato and Malcolm, I also feel attracted to the claim that knowledge is sometimes justifıably retained in the face of counterevidence to what one knows— HEADACHE and WEATHER tell me this much.

6.2

Marbles

In contrast with the Plato-Malcolm tradition that takes knowledge to be indefeasible, a different (and at least as influential) tradition takes knowledge to be defeasible. Gilbert Harman, Timothy Williamson, and many others have given voice to this tradition.7 For instance, here is Williamson making quite succinctly the point that knowledge is defeasible: The assumption [that knowledge can be added but not subtracted over time] is obviously false in practice, because we sometimes forget. But even if the model is applied to elephants, idealized subjects who never forget, the assumption that [knowledge] cannot be lost is implausible. On any reasonable theory of [knowledge], an empirical proposition which now counts as [knowledge] can subsequently lose its status as [knowledge] without any forgetting, if future evidence casts suffıcient doubt on it. (Williamson 2000, 206)

6

Of course, I am also supposing that there is no funny business of the Gettier type going on here. The same is true of all the other cases in this chapter. 7 For example, Jaakko Hintikka (1962, 20), claims that ‘whoever says “I know that p” proposes to disregard the possibility that further information would lead him to deny that p.’ Peter Klein (2017), and Roy Sorensen (2012), also endorse the view that knowledge is defeasible. I will discuss Peter Klein’s views below.

6 Knowledge, despite Evidence to the Contrary

75

This point is usually backed up by examples such as this8: MARBLES: Mary puts a red and a black marble in a bag. She shakes the bag well and proceeds to make draws with replacement while carefully taking notes of the result of each draws. Each of Mary’s fırst one thousand draws produce a red marble.9

The natural reaction to MARBLES seems to be something like this: although Mary knows, at fırst, that there is a red and a black marble in the bag, after drawing a red marble one thousand times Mary ceases to know that this is the case, even though, by stipulation, this is true. Her evidential situation changes as the case progresses, however. She starts the case with excellent evidence for the claim that the bag has both a red and a black marble, but each time Mary draws a red marble from the bag, she acquires a small piece of counterevidence to the claim that the bag has both a red and a black marble. By the one-thousandth time she draws a red marble, the small pieces of counterevidence have added up to more counterevidence than her knowledge-level justifıcation could bear. Mary’s knowledge that there is a red and a black marble in the bag is thus defeated; its justifıcation dies a death by one thousand counterevidential cuts.10 Knowledge, MARBLES suggests, is not completely tethered, or fıxed; it is in fact somewhat loose and may, as it were, be pushed out of one’s mind. The line of reasoning in this reaction to MARBLES is intuitive, and apparently in tension with the idea that knowledge is tethered or stable. We will get back to this tension below. Here one might feel inclined to complain and say that the case is too implausible to support the epistemic point I am making. After all, what is the chance that someone will draw the same red marble one thousand times in a row?!11 In order to assuage these fears, we may consider an analogous case that takes care of this issue. MARBLES*: Things are as before except for the fact that Mary’s draws are red or black more or less at the rate one would expect (i.e., more or less half of the time). After doing this for a couple dozen times, Mary’s reliable friend, Larry, walks into the room and says, ‘You probably drank tainted water. Two in every three bottles of water in the kitchen have been tainted with a hallucinogenic substance that causes people to hallucinate the wrong color of objects—they think non-red objects are red! You have been drinking from a bottle you got from the kitchen, right? So, it’s likely that you drank the tainted water.’ Suppose that Larry’s

8

Gilbert Harman (1973, 192) does just that in the following passage: That undermining evidence . . . is relevant . . . to the acquisition of knowledge [and] also to its maintenance is clear from the . . . example concerning Tom and the library detective. The detective knows that he saw Tom steal the book and so he testifıes to the Judicial Council. After he leaves the hearing, Tom’s mother fabricates her story about Tom’s twin brother. Her lying testimony convinces the Judicial Council but is unknown to the detective back at his post in the library. [O]nce Tom’s mother has. . . testifıed, it is no longer true that the detective knows that Tom stole the book. 9 MARBLES is a variation of a case discussed by Williamson (2000, 205). 10 N.B., I am not making a claim about exactly what impact each draw has on her justifıcation; the impact of each draw is plausibly really small. The point, rather, is that the sum of all one thousand draws intuitively does have a major impact on her justifıcation. Thanks to Cherie Braden for discussion here. 11 Answer: small—0.51000.

76

R. Borges testimony is true and that Mary has no reason to doubt him—quite the opposite, Mary has all the reason to trust him, since he is a good reliable friend and Mary knows that. Suppose also that Mary did not drink from the tainted water.

Larry’s testimony amounts to counterevidence to Mary’s knowledge that there is a red and a black marble in the bag. What is more, although we are supposing it to be true that there is a red and a black marble in the bag, it seems that Larry’s testimony defeats the justifıcation Mary had for being confıdent that there is a red and black marble in the bag. One way to put this is to say that Larry’s testimony made it implausible for Mary to neglect the possibility that she drank the tainted water and that she might be hallucinating having put a red marble in the bag. It seems that this is enough to lower the justifıcation Mary had for this claim below the threshold required for knowledge. While the Plato-Malcolm tradition and cases such as WEATHER and HEADACHE suggest that counterevidence cannot lower one’s justifıcation below the threshold required for knowledge, the Harman-Williamson tradition and cases such as MARBLES/MARBLES* suggests the exact opposite.

6.3

Weak and Strong Views of Knowledge

The clash between these two views of knowledge is captured in the following two claims: (C1) Knowledge is never defeated by counterevidence. (C2) Knowledge is sometimes (but not always) defeated by counterevidence.

A word of warning is warranted before we progress. I do not mean for the discussion so far to be taken as ‘conclusive evidence’ or even as ‘strong evidence’ for (C1) and (C2). Rather, the discussion so far illustrates what epistemologists who accept (C1) and (C2) have in mind when they argue for those claims. The Plato-Malcolm tradition seems to think that the indefeasibility of knowledge is what explains (at least in part) why knowledge is more valuable than true belief. The Harman-Williamson tradition, on the other hand, seems to believe that the defeasibility of knowledge is a consequence of subjects taking evidence seriously. In a sense, the fırst tradition emphasizes the difference in quality between the grounds for knowledge and grounds for mere true belief, while the latter tradition emphasizes what those grounds have in common. (C1) and (C2) cannot both be true, of course. However, both claims can be false at the same time: if knowledge is always defeated by counterevidence, then it is false that it is never defeated or that it is sometimes but not always so defeated. Although there are costs and benefıts associated with accepting one of those claims and rejecting the other, I take it that WEATHER and HEADACHE show quite conclusively that the idea that knowledge is always defeated by counterevidence is not very plausible. But, more generally, if one’s epistemology allows for the partial defeat of evidential support, then knowledge is not always defeated by counterevidence, even if evidential support is always so defeated. Let me explain.

6 Knowledge, despite Evidence to the Contrary

77

Since what gets defeated by counterevidence is the epistemic support one’s belief receives from one’s evidence (i.e., one’s justifıcation), if one can know something even if the probability of the proposition believed given one’s evidence is lower than 1, then counterevidence will not always defeat knowledge: if S can know that p when the probability that p is true given S’s evidence is, say, .95, then the probability that p is true conditional on S’s evidence may decrease from, say, .98 to .96 in virtue of some counterevidence without S ceasing to know that p.12 There are at least a couple of costs associated with accepting (C1). One of them is having to explain away the intuition that Mary’s knowledge in MARBLES/MARBLES* is destroyed by counterevidence.13 The clear benefıt of accepting (C1), on the other hand, is arguably methodological. (C1) is simple (i.e., there is no need to fınd a principled way to distinguish between situations in which knowledge is defeated by counterevidence from situations in which knowledge is not defeated by counterevidence) and easily falsifıable (i.e., one case in which knowledge is defeated by counterevidence is suffıcient to falsify C1). (C1) also fıts nicely with the Platonic idea that knowledge is stable in a way that true belief is not, for, according to (C1), knowledge is not only stable but it simply cannot be moved by evidence to the contrary. As for (C2), the obvious cost associated with this view of knowledge is one of providing a principled way to distinguish situations in which counterevidence defeats knowledge from situations in which it does not. We could get rid of this cost by emending (C2) and suggesting that counterevidence always destroys knowledge, but that emendation would, I take it, make (C2) way too implausible—as HEADACHE and WEATHER suggest. On the other hand, the benefıt of accepting (C2) does not seem to be methodological simplicity, but something like fitness, since it seems to fıt our intuitive judgment about all the cases we discussed so far—it takes those cases as data to be explained rather than to be explained away) and, if correct, accounts for the intuitive judgment that knowledge is not defeated in WEATHER and HEADACHE, and the judgment that knowledge is defeated in MARBLES/MARBLES*. Before we move forward, it is useful to distinguish between two different ideas we might want to convey when we say that someone’s knowledge is defeated by counterevidence. On the one hand, when we say that S’s knowledge that p is 12

Be that as it may, all cases discussed in this chapter deal with categorical, rather than partial, defeat. That is, with respect to all cases discussed in this chapter, I am supposing that the potential defeating effect of counterevidence is always such that it lowers the subject’s degree of justifıcation below the threshold required for knowledge. 13 There is another, less obvious, potential cost of accepting (C1). That is the danger of dogmatism and the rejection of an intuitive notion of intellectual humility: if knowledge is indefeasible, why should one pay attention to counterevidence to what one knows? Isn’t all counterevidence to what one knows misleading evidence? This problem for views that accept (C1) was put forward by Saul Kripke in a talk in the early 1970s. Harman (1973) made a similar problem popular. Kripke (2011) includes Kripke’s talk as well as an appendix explaining the difference between his version of the problem and Harman’s. Since I dealt with this issue in Borges (2015), I will not dis cuss it here again.

78

R. Borges

defeated by some counterevidence, e, we may mean to say that S’s learning that e is the case caused her to stop believing that p. On the other hand, when we say that S’s knowledge that p is defeated by some counterevidence, e, we may mean to say that S’s learning that e is the case lowers S’s degree of justification for believing that p below the threshold required for knowledge. One can lose knowledge because of counterevidence in either of those two ways: by ceasing to believe or by having one’s justifıcation lowered below the threshold required for knowledge.14 What is more, although those two ways in which knowledge may be defeated are clearly related (e.g., an awareness of the fact that counterevidence lowered one’s justifıcation below the threshold required for knowledge may cause one to stop believing), it is important to keep them apart when assessing the plausibility of (C1) and (C2). In particular, one may distinguish two different versions of each claim: (C1B) Knowledge is never defeated by counterevidence that causes one to stop believing. (C1J) Knowledge is never defeated by counterevidence that lowers one’s degree of justifıcation below the threshold required for knowledge. (C2B) Knowledge is sometimes (but not always) defeated by counterevidence that causes one to stop believing. (C2J) Knowledge is sometimes (but not always) defeated by counterevidence that lowers one’s degree of justifıcation below the threshold required for knowledge.

Once we distinguish between these different versions of (C1) and (C2), we make room for a more nuanced approach to cases like the ones we have been discussing. For example, one may now argue that, although Mary’s justifıcation for believing that there is a black and a red marble in the bag is not lowered by her coming to know that she drew a red marble one thousand times, this counterevidence may well cause her to lose her belief in the proposition that there is a red and a black marble in the bag. This would involve accepting both (C1J) and (C2B). Call this type of view a strong view of knowledge. Alternatively, one may insist that the counterevidence in MARBLES/MARBLES* does lower Mary’s justifıcation for believing that there is a red and a black marble in the bag below the threshold required for her to know that fact; she also loses her belief in that proposition, but the epistemically signifıcant event is

14

For simplicity’s sake, in drawing this distinction I am presupposing that knowing that p entails believing that p. I think that even if one rejected this presupposition one could still draw the relevant distinction, although doing so would be more complicated. What I mean is that the distinction is compatible with the view, sometimes exposed by knowledge fırsters, that knowledge cannot be analyzed in terms of belief, truth, etc. The reason why this view is compatible with our distinction is that one may hold it and still think that whenever someone knows that something is the case that person also believes that it is the case; that is, one may consistently hold the view that knowledge cannot be analyzed in terms of belief while, at the same time, argue that knowing that p is always accompanied by believing that p (or, alternatively, that the latter state is always a by-product of the former state). If one had this kind of knowledge fırst view, one could consistently hold the distinction between what I will call below ‘psychological’ and ‘epistemic defeat.’ (Incidentally, Fine (2004) discusses a version of the knowledge fırst type of view I am sketching here and suggests one might attribute it to Plato.) I do not have the space to defend these claims here, however. I will come back to them some other day.

6 Knowledge, despite Evidence to the Contrary

79

that her justifıcation gets defeated. This approach to the case would involve accepting both (C2J) and (C2B). Call this type of view a weak view of knowledge. Epistemologists who endorse strong views of knowledge believe that knowledge defeat is exclusively a psychological phenomenon in the sense that counterevidence to what one knows leads one to lose (at most) one’s belief; they reject the claim that knowledge defeat is an epistemic phenomenon, however. That is, they reject the view that sometimes counterevidence to what one knows lowers one’s justifıcation below the threshold required for knowledge.15 Strong views of knowledge elegantly explain what is going on in our four cases: although the counterevidence can never epistemically defeat one’s knowledge, subjects in cases such as MARBLES/MARBLES* have their knowledge psychologically defeated. The challenge for strong views is to explain away the intuition that Mary’s knowledge in MARBLES/MARBLES* suffers epistemic defeat, not only psychological defeat. Following Maria Lasonen-Aarnio (2010), I will call the intuition that the knowledge subjects have in cases such as MARBLES/MARBLES* is epistemically defeated, defeat intuition. Let us explore this point a little further, for it will become decisive below when we assess a recent version of the strong view of knowledge. The defeat intuition emerging in relation to cases such as MARBLES/MARBLES* amounts to an intuition about how justified the target proposition is for the subject given her new total evidence (i.e., once she updates her degree of belief in light of the new evidence she acquires). This means that a theory will do justice to this intuition only if it tells a story about justification (i.e., that which accounts for a good chunk of the epistemic difference between true belief and knowledge), rather than if it tells a story about some other—albeit related—epistemic desiderata.16 In section fıve I will argue that a recent version of the strong view of knowledge is unable to account for the defeat intuition in a way that does justice to the story connecting justifıcation and defeat. I will then offer an alternative account of the defeat intuition that does justice to this story. According to epistemologists who endorse the weak view of knowledge, however, knowledge is vulnerable to both epistemic and psychological defeat. This type of view also offers an explanation of what is going on in our cases: while Mary’s knowledge in MARBLES/MARBLES* suffers epistemic and psychological defeat, the knowledge of subjects in WEATHER and HEADACHE suffers neither. Weak views of knowledge have a hard time giving a principled account of when counterevidence is strong enough to lower one’s justifıcation below the relevant threshold for knowledge. This is an important challenge for a few reasons but no less because we seem to have a better intuitive grip on whether a subject knows than we have on whether a subject has a certain degree of epistemic justifıcation and whether the justifıcation 15

For versions of the strong view of knowledge, see, among others, Dretske (1971, 1981), LasonenAarnio (2010), and Baker-Hytch and Benton (2015). 16 This is true not only of more traditional versions of the weak view of knowledge (i.e., weak views that take knowledge to be analyzed in part by justifıcation), but also of knowledge-fırst views such as the one in (Williamson 2000, Chapter 10), which says that knowledge is not analyzable but is nonetheless susceptible to epistemic defeat.

80

R. Borges

she has is affected by counterevidence in a particular way. Knowledge-fırst versions of the weak view may have an advantage over belief-fırst versions of the weak view, for knowledge-fırst versions of the view presuppose that we have a better grip on knowledge than on justifıcation.

6.4

Defeasibility and Weak Knowledge

In this section we look at Peter Klein’s version of the weak view of knowledge—the defeasibility theory of knowledge. Peter has defended a version of the defeasibility theory of knowledge for over 45 years now.17 His account accepts the defeat intuition and is designed to accommodate it. Consider: DEFEASIBILITY: S knows that p iff (1) p; (2) S believes that p; (3) S is justifıed in believing p; (4) there is no truth, d, such that the conjunction of d and S’s justifıcation, j, fails to justify S in believing that p.

Condition (4) is a no-defeat condition in line with (C2J) above. According to (4), certain truths (i.e., defeaters) are such that they lower one’s justifıcation below the threshold required for knowledge. So, according to Defeasibility, Mary’s knowledge that there is a red and a black marble in the bag is defeated in this way by the truth that Mary drew a red marble one thousand times.18 Defeasibility is a fıne exemplar of the weak view of knowledge, with seemingly a lot going for it. For one, the view seems to provide a neat account of cases where one has a justifıed true belief but no knowledge; for example, in the Gettier cases: in those cases there is a truth that is such that the conjunction of the subject’s justifıcation with this truth fails to justify the subject in believing the target proposition. This take on Gettier cases also allows Peter to account for epistemic luck: it is a matter of luck that the gettiered subject ended up with a justifıed true belief, for she could have easily believed the defeater and failed to know. There are clouds on the horizon, however. Recently Peter focused on accommodating the possibility of inferential knowledge that is essentially based on a false premise (a ‘useful falsehood,’ as he calls it). As he points out, given condition (4) the problem for Defeasibility is that the negation of the false premise seems to be a defeater for the subject’s justifıcation. Here’s one such case (adapted from Warfıeld 2005): HANDOUTS: Ted carefully counts the number of people present at his talk and reasons: ‘There are 53 people at my talk; therefore my 100 handout copies are suffıcient.’ His premise is false. There are 52 people in attendance—Ted double-counted one person who changed seats during the count.

17 18

See, among many others, Klein (1971, 1981, 2008, 2017). As far as I know, Peter also accepts (C2b), but I will not deal with that aspect of his view here.

6 Knowledge, despite Evidence to the Contrary

81

At fırst glance, Defeasibility seems to give the wrong result about HANDOUTS (at least to the extent that we feel inclined to attribute knowledge to Ted): the truth ‘It is not the case that there are 53 people in attendance’ seems to be just the kind of truth (4) says is a defeater and, hence, a truth that lowers one’s justifıcation below the threshold required for knowledge. The problem is that this verdict dispensed by Defeasibility clashes with people’s intuition that Ted knows. Klein (2017) has replied to cases like HANDOUTS in the following way: in those cases, the truth that seems to defeat the subject’s justifıcation is itself defeated by something else the subject knows and that is entailed by the alleged defeater. So, Ted knows that he has enough handouts because the epistemic damage ‘It is not the case that there are 53 people in attendance’ could do to his justifıcation is preempted19 by his knowledge that there are approximately 52 people in attendance. I think this reply to HANDOUTS is promising.20 The key feature is that something other than the falsehood is what is turning the conclusion in Ted’s inference into knowledge.21 There might be problems for Defeasibility when we compare its response to HANDOUTS with its response to MARBLES/MARBLES*. Why can’t we say that something Mary knows prevents her knowledge that there is a red and black marble from being (epistemically) defeated, like we did with Ted’s knowledge that his one hundred handouts are enough? After all, is it not the case that Mary saw that she

19 In Klein (2017, 54) Peter says that this latter truth ‘restores’ Ted’s justifıcation for believing that he has enough handouts. I fınd that confusing—if ‘It is not the case that there are 53 people in attendance’ is not really a defeater, then there is no damage to be restored by this truth. Instead, what seems to be going on here is that the destructive effect of ‘It is not the case that there are 53 people in attendance’ is preempted by Ted’s knowledge that there are 52 people in attendance. 20 I have given a similar reply to HANDOUTS myself in Borges (2017). See de Almeida (2017) for criticism of Peter’s way of handling cases of useful falsehoods. For alternative ways of dealing with HANDOUTS as well as with other cases of knowledge from non-knowledge, see, among others, Ball and Blome-Tillmann (2014), Montminy (2014), Luzzi (2014), and Schnee (2015). 21 It may be suggested that Peter should not have been so quick to assume that Ted is ‘epistemically close’ to what is obviously entailed by something he knows. After all, closure-deniers such as Fred Dretske and Robert Nozick can reasonably reject the claim about epistemic proximity Peter relies on. This is an important worry about Peter’s strategy, but I think it can be mitigated. I do not have space to give it a full treatment here; so, a couple of quick remarks will have to do—at least for now. The key point is this: the thing pushing the proposition beyond the subject’s epistemic reach in counterexamples to closure (e.g., Dretske’s zebra-in-the-zoo case) is not present in cases of useful falsehood, suggesting that closure does not fail in the latter cases (even if it fails in the former cases). According to Dretske, knowledge closure fails in his zebra-in-the-zoo case because, although the zoo-goer has evidence that is good enough to give him knowledge of ‘That’s a zebra’ (e.g., there’s a plaque saying that the animals in the pen are zebras), his evidence is not good enough to give him knowledge of something that is entailed by this truth, namely ‘That’s not a cleverly disguised mule.’ For Dretske (and, in related ways, for Nozick as well) knowledge closure fails in cases where the subject’s evidence, although good enough to produce knowledge of the entailing proposition, is not good enough to produce knowledge of the entailed one. That is the one thing pushing the entailed proposition beyond the subject’s epistemic reach. But Ted’s situation is different. The evidence he has for the falsehood ‘There are 53 people in attendance’ is also good enough to produce knowledge of the truth it entails (i.e., ‘There are approximately 53 in attendance’). Many thanks to Claudio de Almeida for discussion here.

82

R. Borges

put a black and a red marble in the bag? Why can’t we take this knowledge to preempt the defeating effect of the counterevidence coming from the drawings? Saying that Defeasibility can treat those cases differently because the subject knows the target proposition in one case but not in the other is utterly unsatisfactory. Supposedly, the analysans in a correct analysis of the concept of knowledge is more informative than its analysandum. But, if in some cases we need fırst to judge whether the analysandum is instantiated before we can learn whether the analysans is instantiated, then the analysis under consideration is failing to display that core feature of correct analyses of the concept of knowledge. This issue generalizes. For example, supposing we want Defeasibility to accommodate the intuitive judgments that the subjects in HEADACHE and WEATHER know, while the subject in MARBLES/MARBLES* does not know, then we need an account of why the counterevidence in the latter case defeats the subject’s justifıcation while the counterevidence in the fırst two cases does not. Given that in all three cases the counterevidence is encapsulated in a truth that could, at least in principle, be taken to defeat the subject’s justifıcation, it seems arbitrary for Peter to say that justifıcation gets defeated in one case but not in the other. It seems that, in general, there is nothing in Defeasibility itself that allows us to state, in a principled way, when counterevidence epistemically defeats knowledge and when it does not. Call this the indeterminacy problem for Defeasibility. Peter has, of course, addressed the indeterminacy problem in a number of his writings, but I am not convinced his main strategy is fully satisfactory. Let me reconstruct this strategy here.22 When considering some particularly tricky cases where the subject has a justifıed true belief but it is not so clear whether there is a defeater preventing her from knowing, Peter suggests that it is a virtue of Defeasibility that it neither necessitates nor precludes the existence of a defeater.23 To illustrate this point consider the following case24: NEWSPAPER: A political leader is assassinated and the local newspaper reports the event accurately. The whole island, including Smith, reads about it. He reads the newspaper and believes, on that basis, that the leader has been assassinated. However, the leader’s party, fearing a coup, later convinces an otherwise reliable TV channel to televise the false story that a member of the security team, not the leader, has been killed. Although everyone on the island believes the televised piece of fake news, dismissing the ‘rumor’ that the leader was assassinated, Smith is unaware of the broadcast.

Does Smith know that the political leader was assassinated? Peter argues that it is a virtue of Defeasibility that it is compatible with a positive and a negative answer to this question. According to Defeasibility, which answers we give depends on whether we take the true proposition a reliable TV channel said that a member of Peter used to call this problem the ‘problem of misleading defeaters’ but he has recently dropped this description of the problem, and for a good reason: since, according to Peter, misleading evidence against what one knows is just counterevidence that does not defeat knowledge; calling a misleading defeater a ‘defeater’ is just confusing, for defeaters, by defınition, do defeat and the so-called ‘misleading defeaters’ do not defeat (see Klein 2017). 23 See, for example, Klein’s discussion of the Mr. Magic case in Klein (2004). 24 This is a version of a case presented by Harman (1973, 143–144). 22

6 Knowledge, despite Evidence to the Contrary

83

the security team, not the leader, has been killed to be a defeater of Smith’s justifıcation. If we think Smith knows, then we should not consider it to be a defeater; if we do not, then we should not. As I said before, this kind of reply on behalf of Defeasibility is less informative than one would have hoped for. We expect that analysans in a correct analysis of knowledge to be more informative than its analysandum and that this means, in turn, that fınding out whether the latter is satisfıed in a case should take explanatory priority over whether the former is satisfıed in that case, not the other way around.25 But this expectation is frustrated by Peter’s approach to the indeterminacy problem.26 I do not take this problem to be a knockdown argument against Defeasibility. Rather, I wanted simply to illustrate the kind of challenge any weak view of knowledge faces.27 25

Objection: It is vague whether Smith knows or not in this case, and Defeasibility does exactly what we want it to do—it explains clear cases of knowledge/ignorance. Reply: The relevant accusation (what I mean by ‘less informative than one would have hoped for’) is that we look for defeaters only after we decide whether there’s knowledge in a relevant case. This gets things backwards, explanatorily speaking. As in science, we want philosophical theories that predict what type of evidence we are likely to fınd (whether we will fınd knowledge in a case or not) and not theories that only account for evidence we already have (whether cases that we already know have knowledge/ignorance conform to the theory). My claim is that Defeasibility has poor predictive power. Consider. The claim ‘All emeralds are green’ tells me that, for any x, if x is not green, then x is not an emerald. This conditional exploits the analysans (so to speak) of ‘emerald’ and suggests a method for fınding emeralds (i.e., looking at the color of things and disregarding things that are not green). Similarly, ‘All knowledge is undefeated belief’ suggests that, for any x, if x is defeated (if there is a defeater), then x is not knowledge. This conditional exploits the analysans of ‘knowledge’ and suggests a method for fınding knowledge (i.e., looking for defeaters and disregarding beliefs for which there are defeaters). However, while I can look for non-green things even if I do not know whether there are emeralds present or not, I do not see how I can look for defeaters without fırst knowing whether there is knowledge present or not. This is not good. It might not ‘refute’ Defeasibility, but it’s a problem that should be addressed, I think. Also, it is not so clear that it is vague whether the subject knows or not in NEWSPAPER. Nozick (1981, 177), for example, thought it was a virtue of his proposed defınition of knowledge that it explained ‘why we are reluctant to say [the subject in NEWSPAPER] knows the truth.’ According to him, condition 4 in his defınition of knowledge (i.e., if p, then S would believe that p) predicts the right result in that case— that the subject fails to know. What this tells me is that Nozick did not think the subject knew, that he thought that his theory explained why this was the case and that this conformed with what he took to be people’s general inclination about the case. Harman (1973, 145) also argued that Smith did not know and that this was a problem for his account. Thanks to Joao Fett for discussion here. 26 A similar point applies to some versions of the Grabit case, which was originally presented in Lehrer and Paxson (1969). See Klein (1981) for Peter’s treatment of the Grabit case. 27 In fact, one might even say that this way of looking at the indeterminacy problem makes DEFEASIBILITY circular, for whether condition (4) is satisfıed depends in part on whether the subject knows. In my view, this limitation of belief-fırst versions of the weak view of knowledge gives credence to a knowledge fırst version of the weak view: instead of explaining knowledge in terms of defeat (i.e., what explains why one knows, when one does, is the fact that one’s justifıcation for believing is undefeated), we should explain defeat in terms of knowledge (i.e., what explains why one’s justifıcation is undefeated is the fact that one knows). I only mention this here, however, as I do not have the space to elaborate on this approach to defeat. See Baker-Hytch and Benton (2015) for a similar criticism of reliabilist versions of weak views.

84

6.5

R. Borges

The Defeat Intuition

I now turn to a recent version of the strong view of knowledge. I will suggest that it too faces important limitations. I will then briefly suggest one way in which these limitations can be remedied. Maria Lasonen-Aarnio (2010) has recently defended a version of the strong view of knowledge. As such, the view states that counterevidence can never epistemically defeat knowledge. She does, however, want to explain why some have the defeat intuition (i.e., the inclination to think that subjects in cases such as MARBLES/ MARBLES* have their knowledge epistemically defeated by counterevidence). According to Lasonen-Aarnio, although “these intuitions are mistaken, subjects who retain knowledge in defeat cases are genuinely criticizable.” In other words, although Lasonen-Aarnio does not think that knowledge can be epistemically defeated, she wants to give an account of why subjects who retain knowledge when confronted with counterevidence deserve to be criticized. According to her, those subjects are criticizable because they do not follow beliefforming policies that are rational for someone with the goal of acquiring knowledge to adopt. Quite the opposite, says Lasonen-Aarnio; to the extent that those subjects manifest the disposition to retain a belief in the face of counterevidence, they are following an epistemic policy that often prevents false belief from being extirpated by new evidence, for this subject is also disposed to retain false beliefs in the face of new evidence. Hence, for Lasonen-Aarnio, the mistaken intuition that there is epistemic defeat in cases such as MARBLES/MARBLES* is explained by the fact that we have a tendency to think that when a subject is not acting reasonably, she lacks knowledge (Lasonen-Aarnio 2010, 15). That is, we mistakenly take a subject’s unreasonability in believing to mean that this subject fails to know what she believes. Lasonen-Aarnio (2010, 16) says that asking whether a belief is reasonable is a “useful heuristic” for determining whether this belief amounts to knowledge, but she rejects the notion that one’s belief’s being reasonable is a necessary condition on knowledge. Even though this view is initially attractive, its account of the defeat intuition is in tension with central features of any strong view of knowledge. Consider the claim that being disposed to retain knowledge in the face of counterevidence is an indication that one is also disposed to retain belief in situations where one is confronted with counterevidence but one has a false belief. Why think that this is the case? That is, why should we think, with Lasonen-Aarnio, that someone who is disposed to retain knowledge in a circumstance in which counterevidence is present is ipso facto disposed to retain a false belief in similar circumstances? For one thing, it is logically possible for a reasonable person to have one disposition without having the other (at least for some cases or classes of cases). For instance, I may be disposed to retain knowledge of simple arithmetic truths in the face of counterevidence while, at the same time, being disposed not to retain false beliefs about the same subject matter, for the mistakes are easy to spot with the help of new evidence provided, for example, by a calculator.

6 Knowledge, despite Evidence to the Contrary

85

Perhaps what is behind Lasonen-Aarnio’s suggestion is something like the following line of reasoning. Justifıcation is fallible; that is, it is possible for S and S* to be equally justifıed in believing that p even though S knows that p and S* has a justifıed but false belief that p. Since justifıcation is fallible and one’s disposition to retain knowledge and belief is to a good extent determined by how justifıed one is, a disposition to retain knowledge indicates a disposition to retain justifıed false belief, for, given the fallibility of justifıcation, there need not be any difference between how justifıed one is when one knows that p and when one has a justifıed false belief that p. The problem with this argument appears when we try to conjoin the claim that justifıcation is fallible with the claim that knowledge cannot suffer epistemic defeat. Lasonen-Aarnio’s view can have one or the other claim, but not both. If knowledge is what the strong view of knowledge says it is, then it cannot be the case that S is as justifıed as S* in believing that p but S knows and S* falsely believes that p. This entailment holds for the simple reason that, according to any strong view of knowledge, the justifıcation of someone who believes falsely that p may be epistemically defeated by new evidence but the justifıcation of someone who knows that p cannot. But, if S and S* were equally justifıed in believing p, this difference in how resilient to defeat their justifıcation is should not exist. But it does. Hence, if knowledge is what the strong view of knowledge says it is, then justifıcation is not fallible. The upshot is that Lasonen-Aarnio’s account of the defeat intuition loses considerable support. Her account of this intuition relies on the claim that justifıcation is fallible, a claim that does not jive well with strong views of knowledge.

6.6

Strong Knowledge and Defeat

I now suggest a way to provide strong views of knowledge with an explanation of the defeat intuition that does not incur the diffıculties I mentioned in the previous section. I too will presuppose that, although quite strong, this intuition is ultimately unsound. My account gives a central role to justifıcation, but it does so in a way that coheres with strong views of knowledge. Some authors in the knowledge-fırst literature28 (and elsewhere) have pointed out that ‘justifıcation’ most likely does not express one single concept. From our argument in the previous section it follows that a strong view of knowledge has to say that the false beliefs about hands formed by handles BIVs are not as justifıed as our knowledge of hands. What this shows is that if we want to hold a plausible strong view of knowledge and, at the same time, say that the BIV’s hand-belief is justifıed, then ‘justifıed’ must mean one thing when applied to the BIV’s hand-belief and another when applied to our knowledge of hands.

28

For example, Sutton (2007), Williamson (2011), Littlejohn (2012), among others.

86

R. Borges

One way to capture the relevant different senses of ‘justifıed’ in play here is the following. First, we may say that the sense in which I am justifıed is the same sense in which I know: I see that these are my hands, and this entails the truth of my belief that I have hands. In fact, it is hard to think of a better reason to believe one has hands than having seen that some things are one’s hands!29 Secondly, we may say that the sense in which a BIV is justifıed in believing she has hands is the sense in which she has an excuse to believe she has hands, and that she should therefore not be blamed for doing so—given how things seem to her, she should be highly confıdent that she has hands. In fact, it is hard to think of a better excuse to believe one has hands (when one has no hands) than its seeming to one as if one has hands!30 According to this picture, if we are talking about justifıcation as the thing that is epistemically distinctive about knowledge, then we are talking about ‘being justifıed’ as having what Fred Dretske (1971) called a ‘conclusive reason,’ a reason that would not be the case unless the target belief were also the case. I would not have seen that I have hands unless I had hands. Justifıcation, in the sense in which we say that the BIV is justifıed in believing she has hands, is an entirely different beast, however. For one, it is not the case that its seeming to her that she has hands is a conclusive reason for her to believe she has hands (i.e., it is not the case that things would not seem to her the way they do unless she had hands). If we use this distinction (i.e., ‘justifıed’ as knowing vs. ‘justifıed’ as having an excuse) in an explanation of why some of us have the defeat intuition, we get the following simple story: in cases where knowledge is allegedly defeated by counterevidence, it seems to the subject as if what she knows is false; this change in how things seem to her, in turn, leads her to blamelessly form a belief in a false proposition (i.e., the negation of what she knows) and to stop believing what she knows. There are two crucial differences between those two senses of ‘justifıed.’ Firstly, having a conclusive reason to believe is a necessary condition on knowledge, but having an excuse isn’t—I do not need to have an excuse to believe that p if I have a conclusive reason to believe that p. It is true, however, that there is something wrong, epistemically speaking, with someone who retains her knowledge that p when she has a strong enough excuse to believe that ~p—cases like MARBLE/MARBLE* show this much. Mary knows at fırst that there is a red and a black marble in the bag. It is because of the dynamic evolution in how things seem to Mary that she is (at fırst) justifıed in believing that there is a red and black marble in the bag—in the ‘having an excuse’ sense of that term—and, then, as she draws a red marble one thousand times she stops being justifıed. Secondly, as MARBLES/MARBLES* makes clear, justifıcation, understood as having an excuse, can be defeated by new evidence, but justifıcation, understood as having a conclusive reason, cannot. As Mary draws a red marble one thousand times, it stops seeming to her as if there is a red and a black marble in the bag. This is not what

29

This is similar to some versions of epistemological disjunctivism. For one such view, see Pritchard (2016). 30 Williamson (Forthcoming) and Littlejohn (2012) make a similar distinction.

6 Knowledge, despite Evidence to the Contrary

87

happens to Mary’s reason: the reason she has for believing that there is a red and a black marble in the bag is the same throughout the case (i.e., she saw herself putting those marbles in the bag). The changes in how things seem to the subject is what explains why counterevidence sometimes defeats knowledge: when counterevidence to what one knows is strong enough, it makes the known proposition seem false to one, and this, in turn, causes one to decrease one’s confıdence (and to lose one’s belief) in the known proposition; and, at the same time, causes one to increase one’s confıdence (and to form a belief) in the denial of the known proposition. As promised, this version of the strong view of knowledge takes counterevidence to defeat knowledge only via the loss of belief, not via the loss of one’s justifıcation to believe. In other words, there is only psychological defeat. The defeat intuition is an illusion: we take Mary’s knowledge to be epistemically defeated because we recognize that, from Mary’s perspective, it seems as if there is only a red marble in the bag. In general, the degree to which one is inclined to think that subjects in situations such as Mary’s have their knowledge epistemically defeated by counterevidence roughly matches how excusable one thinks it is for subjects in those situations to ‘change their mind’ about the relevant proposition.

References Baker-Hytch, M., & Benton, M. A. (2015). Defeatism defeated. Philosophical Perspectives, 29(1), 40–66. Ball, B., & Blome-Tillmann, M. (2014). Counter closure and knowledge despite falsehood. The Philosophical Quarterly, 64, 552–568. Borges, R. (2015). On synchronic dogmatism. Synthese, 192(11), 3677–3693. Borges, R. (2017). Inferential knowledge and the Gettier conjecture. In R. Borges, C. de Almeida, & P. D. Klein (Eds.), Explaining knowledge (pp. 273–291). Oxford: Oxford University Press. Borges, R., de Almeida, C., & Klein, P. (Eds.). (2017). Explaining knowledge: New essays on the Gettier problem. Oxford: Oxford University Press. De Almeida, C. (2017). Knowledge, benign falsehoods, and the Gettier problem. In R. Borges, C. de Almeida, & P. D. Klein (Eds.), Explaining knowledge (pp. 292–311). Oxford: Oxford University Press. Dretske, F. (1971). Conclusive reasons. Australasian Journal of Philosophy, 49(1), 1–22. Dretske, F. (1981). The pragmatic dimension of knowledge. Philosophical Studies, 40(3), 363–378. Fine, G. (2004). Knowledge and true belief in the Meno. In D. Sedley (Ed.), Oxford studies in ancient philosophy (Vol. XXVII, pp. 41–81). Oxford: Oxford University Press. Greenough, P., & Pritchard, D. (Eds.). (2009). Williamson on knowledge. Oxford: Oxford University Press. Harman, G. (1973). Thought. Princeton: Princeton University Press. Hintikka, J. (1962). Knowledge and belief. Ithaca: Cornell University Press. Klein, P. D. (1971). A proposed defınition of propositional knowledge. The Journal of Philosophy, 68(16), 471–482. Klein, P. D. (1981). Certainty: A refutation of scepticism. Minneapolis: University of Minnesota Press. Klein, P. D. (2004). Knowledge is true, non-defeated belief. In S. Luper (Ed.), Essential knowledge: Readings in epistemology (pp. 124–135). New York: Longman.

88

R. Borges

Klein, P. D. (2008). Useful false beliefs. In Q. Smith (Ed.), New essays in epistemology (pp. 25–63). Oxford: Oxford University Press. Klein, P. D. (2017). The nature of knowledge. In R. Borges, C. de Almeida, & P. D. Klein (Eds.), Explaining knowledge (pp. 35–56). Oxford: Oxford University Press. Kripke, S. (2011). Philosophical troubles. Oxford: Oxford University Press. Lasonen-Aarnio, M. (2010). Unreasonable knowledge. Philosophical Perspectives, 24, 1–21. Lehrer, K., & Paxson, T. (1969). Knowledge: Undefeated justifıed true belief. The Journal of Philosophy, 66(8), 225–237. Littlejohn, C. (2012). Justification and the truth-connection. Cambridge: Cambridge University Press. Luzzi, F. (2014). What does knowledge-yielding deduction require of its premises? Episteme, 11 (3), 1–15. Malcolm, N. (1952). Knowledge and belief. Mind, 61(242), 178–189. Montminy, M. (2014). Knowledge despite falsehood. Pacific Philosophical Quarterly, 44(3–4), 463–475. Nozick, R. (1981). Philosophical explanations. Cambridge, MA: Belknap Press of Harvard University Press. Pritchard, D. (2016). Epistemic angst: Radical skepticism and the groundlessness of our believing. Princeton: Princeton University Press. Schnee, I. (2015). There is no knowledge from falsehood. Episteme, 12(1), 53–74. Sorensen, R. (2012). Epistemic paradoxes. In N. Z. Edward (Ed.), The Stanford encyclopedia of philosophy. Winter 2011 Edition. Sutton, J. (2007). Without justification. Cambridge, MA: MIT Press. Warfıeld, T. A. (2005). Knowledge from falsehood. Philosophical Perspectives, 19(1), 405–416. Williamson, T. (2000). Knowledge and its limits. Oxford/New York: Oxford University Press. Williamson, T. (2011). Knowledge first epistemology. In S. Bernecker & D. Pritchard (Eds.), The Routledge companion to epistemology (pp. 208–218). New York: Routledge. Williamson, T. (Forthcoming). Justifıcations, excuses, and sceptical scenarios. In J. Dutant & F. Dorsch (Eds.), The new evil demon. Oxford: Oxford University Press.

Chapter 7

A Causal Aspect of Epistemic Basing Robert K. Shope

Abstract An epistemologist who treats one’s having a justified belief that p as a requirement for one’s knowing that p typically maintains that at least sometimes that justified status arises in virtue of (possessing/appreciating/deploying) reasons for which one believes that p. It continues to be controversial whether there must be a causal aspect of the relationship—call it the epistemic basing relationship—between (possessing/appreciating/deploying) those reasons and believing that p. Previous causal accounts of epistemic basing have overlooked a candidate for such a causal aspect that I shall explore. I shall argue that by requiring this aspect we avoid certain puzzles faced by prior accounts formulated in terms of causes, including Peter D. Klein’s worry that those accounts run an empirical risk of allowing general skepticism, since we still understand so little about what causes believing. I shall also consider how the requirement that I propose relates to several well-known attempts by Keith Lehrer to construct counterexamples to various conditions of basing that mention causation. My goal is not to present a full analysis of epistemic basing but only to highlight the merits of one candidate for a necessary condition that specifies a causal aspect of it, that is, an aspect whose description refers at some point to causation. Keywords Epistemic basing · Deploying reasons · Causal basing · Lehrer · Privations · Etiology · Anscombe · Internal reliability · Epistemology

An epistemologist who treats having a justified belief that p as a requirement for knowing that p typically maintains that in at least some cases such a justified status arises in virtue of one’s possessing (appreciating/deploying) good reasons for which one believes that p. It remains controversial whether there must be a causal aspect of the relationship—call it the epistemic basing relationship—between such possession and one’s believing that p.

R. K. Shope (*) University of Massachusetts, Boston, MA, USA © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_7

89

90

R. K. Shope

Previous causal accounts of epistemic basing have overlooked a certain candidate for such a causal aspect. I shall argue that by requiring it we can avoid certain puzzles faced by accounts formulated in terms of the involvement of causes. For instance, Peter D. Klein worries that such accounts run an empirical risk of allowing general skepticism, because we presently understand very little about the causes of beliefstates. I shall also consider how the requirement that I propose relates to several wellknown attempts by Keith Lehrer to construct counterexamples to various conditions of epistemic basing that speak of the causes of believing. My goal is not to present a full analysis of epistemic basing but only to highlight the merits of a schema that specifies an aspect of it that is causal in the sense that its description mentions causation (albeit without mentioning actual causes of the belief). My proposal differs in two important ways from prior causal accounts of epistemic basing. First, the aspect of causation that I shall characterize implies fewer restrictions than usual concerning the nature of the causes of believing that p. Second, I shall draw upon a non-Humean perspective on causation that has been advocated by, among others, G. E. M. Anscombe.

7.1

The Importance of Being Superstitious or Prejudiced

As a challenge to possible causal requirements concerning epistemic basing, Keith Lehrer proposed an example concerning a lawyer who is a gypsy and who forms a completely firm belief in the innocence of his client because he superstitiously reads the cards as telling of that innocence (see Lehrer 1974, 124–125). The belief in the client’s innocence causally leads to the lawyer’s discovery of a conclusive but quite complex line of reasoning from the evidence, in virtue of which the lawyer’s beliefstate concerning the innocence is for the first time justified. At that point, the lawyer claims—in Lehrer’s opinion quite correctly—to know that his client is innocent on the basis of the genuine evidence, even though others reject his line of reasoning. This portion of the example shows that the belief in innocence does not originate from discovery of the conclusive line of reasoning, which occurred after the acquisition of the belief. Yet further details of Lehrer’s example need consideration, since causal accounts of epistemic basing have typically distinguished efficient causes of belief from ‘sustaining causes’ of it. Lehrer includes in the example the relevant detail that the lawyer’s awareness of the line of reasoning could not by itself resist the pressure of emotions, which he shares with the general public, inducing a fervent hope that the genuine culprit has been found. Even the lawyer would be unable to believe in the client’s innocence “were it not for the [supposed] fact that the cards told him it was so” (1974, 125). The latter point, however, does not rule out the possibility that awareness of the reasoning forms part of what in some relevant sense sustains belief in the client’s innocence, with the remaining part being superstitious trust in the card reading. Lehrer apparently aims to rule out this possibility by stipulating that what the lawyer read in the cards is what “nurtures” the lawyer’s conviction of the innocence, so that

7 A Causal Aspect of Epistemic Basing

91

the recent good reasoning “in no way supports, reinforces, or conditionally or partially explains why he believes as he does” (1974, 125). In the editions of his subsequent Theory of Knowledge, Lehrer stopped referring to the case of the gypsy lawyer, without explaining why he did so, and instead challenged causal accounts by describing the case of Mr. Raco, who becomes completely convinced of something out of racial prejudice rather than superstition. Perhaps Lehrer shifted to this example because he recognized that the earlier one might be understood as a situation in which the causal path from the gypsy’s general superstitious attitude toward card readings to his conviction in the client’s innocence runs through the intermediate formation of a belief that c: ‘The cards reliably indicate that the client is innocent.’ The example might seem analogous to a situation where one trusts testimony that h from someone whom one does not realize is an unreliable testifier, and one responds by believing the falsehood that the person reliably testifies that h, which belief in turn leads one to believe that h. Under such an understanding, the falsehood that c seems to count as either (1) a non-superfluous member of the set of statements expressing that upon which the lawyer bases belief in the client’s innocence, or (2) at least part of what the lawyer should sincerely report as belonging to that set. On either interpretation, the involvement of the falsehood may mean that, as Lehrer describes the case, the lawyer is not making a proper claim when he states that he knows that the client is innocent on the basis of the considerations that he cites. In Lehrer’s newer case, Mr. Raco has a true conviction, call it believing that d, which originates purely out of racial prejudice, to the effect that the members of a certain race are susceptible to a disease that members of his own race are not (see Lehrer 1990, 170). Here there may be no intermediary akin to the lawyer’s believing that c, and it is clear that at this stage in the example Raco’s believing that d is not justified. Lehrer adds that after Raco becomes a medical expert, his ongoing conviction that d leads him to research the disease in question so that he is led to have, appreciate, and deploy in reasoning what counts as conclusive evidence that what he believed is true. While admitting that a sufficiently precise account of causal sustenance is lacking, Lehrer stipulates that in this example Raco has come to know that d even though “the evidence justifying Mr. Raco’s beliefs does not in any way explain or causally sustain his belief” (see Lehrer 1990, 170).

7.2

Originating versus Sustaining

In response to such examples, some epistemologists have allowed that only in some cases of reason-based knowledge that p is possession (appreciation/deployment) of one’s (good) reasons for believing that p what originates one’s believing that p (call these TYPE 1 cases). The remaining cases (call them TYPE 2) are those in which believing that p is rendered justified by one’s possessing a collection of reasons, yet that possession did not originate one’s believing that p. The epistemologists in question maintain that in TYPE 2 cases, possession of the reasons causally sustains

92

R. K. Shope

one’s believing that p, rather than constituting its efficient cause. Everyday examples that have been offered of the causal sustainment relationship include the way each leg of a table resting on multiple legs helps to sustain the tabletop’s distance from the floor, or the way one of two cords running from the ceiling to a plant-holding basket helps to sustain the plant’s distance from the ceiling (see Dretske 1981, 89; cf. Armstrong 1973, 79–85, 94–95). Several concerns about this appeal to causal sustainment should be noted. Not only does it leave us with a disjunctive formulation of a causal requirement for epistemic basing, but, according to Peter D. Klein (2012), it involves an unwise risk in presuming that epistemic basing involves causes. He maintains that it is preferable to seek a way of characterizing basing without commitment to an ‘etiology view’ of the belief-state involved in one’s knowing that p. Klein employs the term, ‘etiology,’ in the broad sense of what concerns “the causes and origins” of something, so that to take the etiology view of the belief-state in question is to regard the belief as having a certain type of causal origin (152n1). Klein cautions that we should presently avoid etiological requirements concerning knowing, since they “presuppose some very risky empirical claims—so risky that any explanation that employs them cannot be, at least at this point, the best explanation of our acquisition of inferential knowledge” (160). If we imagine Lehrer’s case of the gypsy lawyer as a real happening, it illustrates how this problem might arise, since the lawyer’s belief that the client is innocent originates upon reading the cards and yet, as Klein might put it, “only later are good reasons located for the content of the belief-state that contribute to converting the belief-state to knowledge” (160). Klein’s elaboration of his general warning regarding this type of case is puzzling. He says that a defender of the etiology view is committed to the risky empirical claim that “the cause of the belief-state has changed” and must hold that “the ‘basis’ on which the belief-state is now held is a new cause of the belief-state” (2012, 161). Apparently, Klein is saying that the etiology view must hold that—in a real gypsy lawyer case—an obtaining of the state of affairs S’s-believing-that-p originates in a way that leaves the belief-state without epistemic justification yet is later replaced by a different occurrence of that state of affairs, originating as a result of the acquisition of the good reasons for believing (cf. 161n15). Klein goes on to speak of one version of the etiology view, standard reliabilism, as committed to holding that S’s “belief-states containing the reasons became the sustaining causes of the belief-state [that p],” and Klein asks, “But isn’t it possible that wasn’t what happened? Perhaps the cause never changed” (compare to Klein 2012, 161). Klein’s mention of sustaining causes does not seem to be in accordance with the usual motives for utilizing that label when responding to examples such as that of the gypsy lawyer. Klein seems to regard the schema, ‘x is the sustaining cause of y,’ as having been introduced into the debate so as to entail that x (is at least part of what) originated y. I take it that philosophers have wished to speak of efficient causes and sustaining causes as types of causes, yet to treat a phrase of the form, ‘x was a/the cause of y,’ as being about an instance of the former type of cause.

7 A Causal Aspect of Epistemic Basing

7.3

93

Speculation and Introspection

A separate thread in Klein’s objections to the etiology view argues that we have little evidence about what causes our belief-states and about whether having a reason and deploying it is a cause of a belief-state (see Klein 2014a). Klein maintains that with respect to a given instance of S’s basing believing that p on believing that q, what is needed in order for us to have enough empirical evidence to claim that the former was caused by the latter is for us already to have applied a way of identifying believing-that-p states and a way of identifying believing-that-q states so as to obtain “enough evidence to show that beliefs of the one type are causes of beliefs of the other type” (2014b, 113). Klein questions whether we possess such a way of identifying belief-states. He admits that some might maintain that we do have introspective evidence of causation with respect to some relevant instances, yet he nonetheless asserts that “what we discover from introspection is that we believe that p and that we believe that q. We do not discover the causal relation” (2014b, 113). In responding to this point, I shall draw upon anti-Humean views concerning causation, especially those presented by G. E. M. Anscombe, Rom Harré, and Edward H. Madden.1

7.4

Apprehending Causation

Anscombe (1981) maintains that we lack awareness of the constant conjunctions demanded by Humeans for making a proper causal judgment. Instead, we acquire a concept of causation by abstraction from our having learned to use members of a set of expressions, such as ‘cut,’ ‘drink,’ ‘push,’ ‘wet,’ ‘scrape,’ ‘knock over,’ ‘keep off,’ or ‘hurt,’ where we have experienced some of the situations to which the terms apply. She maintains that learning to use these verbs is part and parcel of learning to apply words—such as ‘wind,’ ‘body,’ or ‘fire’—and terms for various natural kinds of stuff. She further maintains that the abstracted concept of causation is not subject to independent introduction. Although we cannot perceive every instance of causation, perception of causation does not involve any characteristic sensation (in contrast, say, to the perception of red). Similar examples of perceiving causation had been proposed by Harré and Madden, e.g., seeing the waves eating away at the shore, the axe splitting the wood, or the avalanche’s destroying the village (see 1975, 49–50, 53–54, 60, 113).2

1 Exposition and discussion of Anscombe’s position appears in Haldane (2015), Hornsby (2011), Makin (2000), Osborne (2007), and Teichmann (2008). I critically consider Harré and Madden’s views in Shope (1988, 1999). 2 Earlier, E. H. Madden and Peter H. Hare (1971) maintained that we sometimes perceive causation, and they referenced the work of the psychologist Albert Michotte (1963). Harré and Madden point

94

R. K. Shope

Anscombe describes some examples involving what she calls a ‘mental cause.’ For instance, hearing a sound gives one a fright, or hearing it makes one jump, and one apprehends this causation independently of having data about constant conjunctions (see Anscombe 1957, 17–18). I would add that there are cases of one’s being aware of having reasons for believing something and being aware of that state’s causing one to form the belief. These are cases that one might report by saying, e.g., ‘Careful consideration of the evidence finally made me believe that h,’ or ‘My realizing that k got me to believe that h,’ or ‘My awareness of those facts caused me to believe that h.’ Concerning the introspecting of causation reported in such comments, we should say what Harré and Madden say about experiences of causation in our environs, namely, that “there is no reason to doubt the authority of an ordinary experience,” which “carries its own warrant,” and that one’s claiming to be directly aware of causation by the particulars involved in such examples “requires no further ‘justification’ in order to be acceptable” (1975, 56). Harré and Madden distinguish having prima facie warrant or prima facie acceptability for such causal judgments from what is provided when we “finally justify” or provide “verification” of the judgment by supplying an understanding of causal mechanisms that are involved. They stress that in pursuing the latter task we sometimes realize that when we made the initially acceptable causal judgments, we were “mistaken in supposing that what looked like causal production really was” (1975, 162, 93).

7.5

A Non-disjunctive Condition of Basing

Let us consider the possibility of explaining an aspect of epistemic basing that is involved in each example of reason-based knowledge by means of the following proposal, whose various elements I shall gradually clarify and refine: When one possesses the reasons in a set {r1,. . ., rn}, a necessary condition for the combination of the members of {r1,. . ., rn} to constitute a basis for which one believes that p is the following: (C) The possession of the members of {r1,. . ., rn} is enough to prevent the presence of a privation of one’s believing that p.3

7.6

Factoring in Differentiating Factors

Anscombe proposes that “causality consists in the derivativeness of an effect from its causes. This is the core, the common feature, of causality in its various kinds. Effects derive from, arise out of, come of, their causes” (Anscombe 1981, 136). Such a view

out (Harré and Madden 1975, 49, 60, 113; cf. 68n13) that ‘radical empiricists’ such as Sterling P. Lamprecht (1967) had defended the position. 3 A reason for employing the apparent double negative will be explained below.

7 A Causal Aspect of Epistemic Basing

95

need not be restricted to efficient causation, at least not if we restrict the concept of an efficient cause to that of a triggering factor—for instance, a loud noise that makes a horse bolt and run out of a barn. The view is compatible with what some philosophers call causal selection—roughly, our common practice, in light of salient contextual interests, of picking out as the cause (or one of the causes) of some effect only one (or only a few) of the factors relevantly involved in leading to that effect, where what we select may, but need not be, a triggering factor. Discussions of this practice have pointed out that when a person who has a particular set of interests picks out, for example, the loud noise as the cause of the horse’s being out of the barn, someone having a different set of interests may regard the cause as being the failure of a stable hand to shut the barn door in accordance with the stable owner’s rules. In light of Anscombe’s perspective, we can say that in each context the causal judgment presupposes a causal nexus from which to select certain factors, that is, presupposes causation in the form of a situation where various factors derive from, arise out of, come of others (e.g., the horse’s bolting derives from hearing the loud noise and the horse’s getting outside comes of its bolting in the direction of the door) even though some factors within this nexus are only conditions contributing to or making possible such relationships and are not the cause, a cause, or one of the causes. What is sometimes labeled the ‘differentiating-factor’ or ‘difference-maker’ account of causal selection maintains that in selecting x as (one of) the cause(s) of y, we have in mind—given salient interests—a certain intentional object (‘the contrast situation’) which we think of as a situation not including x and as developing in a way that does not lead to the occurrence of y. Furthermore, we judge the presence of x in the actual situation to have made the difference between that actual situation and the contrast situation. Here, making a difference is to be construed not merely as explaining why x’s combination with salient elements of the contrast situation was causally relevant to the existence of y.

7.7

Privations and Causation

Applying the differentiating factor perspective to the horse example, we may say that someone may have interests in light of which that person regards the failure to shut the barn door as the difference-making factor and the cause of the horse’s being outside the barn. Granted, nobody can perceive that failure’s causing the effect, and the failure is, indeed, a privation. Yet it is not a privation in the sense of a mere nothing, but in the sense explained by John Haldane (2007), according to which a privation is the obtaining of a certain item’s having a certain characteristic. Haldane offers as an example the way that a window that is a simple hole in the wall is a privation of wall materials, a privation consisting in their actually being distributed throughout a certain portion of space. In the present horse example there is a privation of dutiful behavior by some person, such as a stable hand, consisting in the person’s actions running contrary to mandated practices. I take Haldane to have

96

R. K. Shope

persuasively argued that privations can be causes.4 The privation in the horse example is what makes the difference between the actual situation and a contrast situation that is imagined not to include that privation and not to develop so that the horse eventually goes outside the barn. Thinking of cause x as making the difference with respect to effect y can sometimes be thinking of what I once called an imagined ‘interaction’ of x with some portions of the contrast situation in accordance with ‘standards of development’ that are contextually salient for the combination of x with those portions (see Shope 1967).5 The term, ‘interaction,’ is appropriate insofar as the development in question concerns what is, from Anscombe’s viewpoint, the manner in which a process leading to y arises out of, comes of, or derives from x (and, perhaps, in which some of what is involved in that process arises out of, comes of, or derives from other factors within it). I shall allow those brief considerations to suffice in favor of regarding the antiHumean views in question as compatible with the broad outlines of major philosophical and psychological research concerning the pragmatics of causal selection, research that is often free of the objectionable Humean insistence on universal generalizations. A useful upshot of the preceding considerations is that they have led us to consider the presence of privations in a causal nexus. If some privations can be causes, then some privations can be effects. Indeed, some privations can count as possible effects that are prevented—which is the type of prevention mentioned in condition C.

7.8

Prevention, Sustainment, and the Possibility of Existential Inertia

Consider prevention in regard to Lehrer’s case of the gypsy lawyer. When philosophers called the lawyer’s eventual possession and appreciation of the evidence a sustaining cause, they may have been comparing it to the way that a leg of a table helps to prevent the tabletop’s tilting or falling, or to the way that a suspending cord helps to prevent a plant basket’s sagging or dropping. Can we maintain that in an analogous fashion the gypsy lawyer’s acquisition and appreciation of the evidence helps to prevent the lawyer’s belief-state from weakening or being lost?

Anscombe may be regarding a privation as a cause when she discusses “causalities of various kinds” that may contribute to a traveler’s arriving (or not arriving) at his destination, and writes, “The causalities will for example include negations. Because this man did not know this language, he went this way rather than that. . .” (Anscombe 1983, 190). 5 These standards presumably require that the interaction does not involve what philosophers call wayward or deviant causal chains. I have taken a stab at characterizing such chains in Shope (1999). 4

7 A Causal Aspect of Epistemic Basing

97

Sympathy with such an analogy seems to be implied by Socrates’ remarks in Meno when he considers why knowledge is much more valuable than mere true belief. Socrates suggests that we can view one’s being in a true belief-state as analogous to possessing a statue made by the sculptor Daedalus, or possessing a runaway slave. Plato scholars report that this sculptor’s statues were so lifelike as to be rumored to wander off during the night. Analogously, the runaway slave might be imagined as still intending to escape. Socrates suggests that in order to retain possession of the statue or the slave, so that they can be of significant value to one, they need to be tethered. He speculates that since all belief-states have a tendency to depart, then for a true belief-state to be of much value to one, it needs to be related to something like a tether—namely, an ‘account.’ The situation of its being thus ‘tethered’ might constitute one’s possessing knowledge. When offering these analogies, Socrates leaves unspecified how possession of the statue or slave originated. For present purposes, we can imagine their being tethered as analogous to what I have called a TYPE 2 case of reason-based knowledge, insofar as the possession predates the tethering. Such a comparison helps to reveal a problematic detail related to everyday cases of causal sustainment. In cases such as those concerning the table leg and the suspending cord, gravity operates contrary to an upward mechanical force exerted by the leg on the tabletop or by the cord on the plant holder. If the leg does not quite touch the tabletop, then it is not helping to sustain it, and if the cord is slack, then it is not helping to sustain the basket’s position. From Klein’s cautious perspective, it is a risky empirical speculation to follow Socrates’ analogy by assuming that every true belief-state is subject to something pressing upon it to depart.6 An alternative speculation is that at least some of the belief-states involved in a person’s coming to have knowledge possess “existential inertia”, that is, continue to exist for some time after they originate without being sustained against the influence of something else, and so are in no obvious sense sustained at all (cf. Beaudoin 2007). A way to connect causal sustainment to considerations about prevention can be illustrated by contrasting the manner in which different households rely upon a backup electrical generator. In house A, a piece of equipment, E, is a generator and is in a state of readiness such that it will turn on in response to the house’s amperage dropping below a certain level. In house B, device E is present but not activated, and what monitors the home’s circuits is a separate device that is in a state of readiness to turn on E in response to any significant drop in amperage. It is not implausible to maintain that it is only in house A that E is helping to prevent a loss of current, and that in house B the device will help to prevent a loss if a drop begins to occur. Why

6

A presumption of opposing forces is involved in Kant’s examples of simultaneous causes, which Jay Rosenberg (1998) speaks of as sustaining causes.

98

R. K. Shope

suppose that in TYPE 2 cases of reason-based knowledge one’s possession of the good reasons is analogous to the presence of device E in house A rather than in house B?7

7.9

Enough

A way to avoid risky speculations about whether something is actually being prevented can be found by considering further details of Anscombe’s anti-Humean approach. She sides with the physicist Richard Feynman concerning an example of what she calls a non-necessitating cause, a cause “that can fail of its effect without the intervention of anything to frustrate it” (Anscombe 1981, 144). In the example, someone has rigged a bomb to explode if a Geiger counter registers a certain reading, and the counter “is so placed near some radio-active material that it may or may not register that reading” (145). Anscombe maintains that if the reading did occur and the bomb went off, there would be no doubt that the cause of the reading was the radioactive material’s emitting particles in such a way as to register the relevant reading on the Geiger counter. Moreover, the emission of the particles is caused inasmuch as “it occurs because there is this mass of radio-active material here” (145). Anscombe’s agreement with Feynman accords with her question for those Humeans who demand what they call ‘sufficient’ conditions for an effect and who sometimes speak of such supposed ‘necessitating’ conditions as enough for the effect: “May there not be enough to have made something happen—and yet it not have happened?” (135).8 One might apply this insight to the Feynman example by saying that the presence of the mass of radioactive material is enough for an emission of the sort in question even in a variant of the example where no such emission occurs while the setup endures. Upon considering how we might speak of what is enough with respect to the generator examples, it is appropriate to say that in both house A and house B the presence of device E is enough to prevent a privation of adequate amperage, just as we may say that the presence of the tether or shut door is enough to prevent a privation of the slave or horse, whether or not an attempt to run away is made. Analogously, for the lawyer to have knowledge of the client’s innocence, the eventual possession of good reasons needs to be enough to prevent a privation of a

7

If we imagine a slave who never plans to run away, or a horse that never tries to get outside the barn, it seems proper to say that the tether or shut door will prevent escape if an attempt is made, but not that they are preventing escape. 8 There are, in addition to examples that I shall discuss, a wide variety of situations that illustrate Anscombe’s point. Consider, for instance, relevant contexts for the following comments: ‘The roughness of the road was enough to set off the nitroglycerine, so we were lucky on the drive;’ ‘What you just said is enough to make me angry, but I’m controlling my temper;’ ‘You’re smart enough to have realized it, so why didn’t you?’ and ‘In that academy, most of the behavior of the students is enough to get them expelled, but the administration is so clueless that it isn’t aware of what is going on.’

7 A Causal Aspect of Epistemic Basing

99

belief in that innocence even in the face of a weakening of the belief-state or a radical drop in the degree to which the lawyer is motivated by superstition regarding the cards. Although the possession of the reasons is enough to prevent the presence of a privation of that type of belief-state, such prevention might fail to occur. There may be a real possibility of the lawyer’s losing faith in the client’s innocence, for instance, in the face of an especially powerful and clever presentation from the prosecutor.

7.10

The Explanatory Upshot

I have now assembled the components needed to explain condition C, provided that we recognize a certain point concerning its application to TYPE 1 cases, namely, that the possession of reasons originating one’s believing that p entails that the possession of the reasons was enough to cause one’s being in that state. This, in turn, entails that the possession was enough to prevent the presence of a privation of such a belief-state, assuming that we view privations from Haldane’s perspective. For the privation of the belief that p is the possession of a certain characteristic by the collection of one’s belief-states. It is the characteristic of having a membership consisting in beliefs whose contents are articulable without making statements whose conjunction entails that p.9 Roughly put, the collection does not include the belief-state that p.10 Originating a member whose content is that p is an instance of causally preventing the presence of such a privation. The judgment that a specific TYPE 2 case of believing that p satisfies C may be justified by similarities that the case bears to other examples which we already justifiably judge to involve causation of belief-states by the possession of reasons.11 This is analogous to the way in which, in thinking about Feynman’s example, we draw upon our prior rational judgments concerning actual radioactive samples and emissions. Klein might respond that part of what guides us regarding that example is quantum theory, and he might point out that we do not possess any analogous theory

9

Compare Haldane’s example concerning the window, where the domain throughout which the wall materials extend is specifiable without making statements about the spatial domain that is the hole. 10 Understanding the horse examples requires considering a privation that consists in the barn’s having the characteristic of possessing contents none of which is the horse. Again, when engaging in causal selection and judging x to be (one of) the cause(s) of y, we imagine a specific causal contrast situation as a privation—in Haldane’s sense—where the situation is individuated as involving such-and-such a variety of details and has the characteristic of possessing contents none of which in the course of its development are x or y. 11 The wording of C employs the phrase, ‘a presence of a privation,’ rather than merely ‘a privation,’ in order to discourage an impression that in every instance satisfying C one’s possession of the reasons postdates one’s believing that p. Such an impression would construe C as unfulfilled in TYPE 1 cases.

100

R. K. Shope

about believing. Yet why should this dissimilarity cancel the reasonableness of an admittedly fallible judgment about the satisfaction of C in a relevant TYPE 2 case?

7.11

Is ‘Enough’ Enough?

I am content here to utilize an ordinary sense of statements of the form ‘e is enough to a’ without committing myself to an analysis of them. However, some may desire more of an indication of the sense in which those statements employ the term ‘enough.’ The best that I can do at present is to suggest that it is a sense in which such a statement is tantamount to one of the form, ‘There is a real possibility that were a-ing to obtain/occur/exist, it would be in virtue of the obtaining/occurring/ existence of e.’

7.12

Similarities to Disagreements over Causal Selection

Such an understanding helps to explain a similarity between affirming something to be a cause of something and affirming it to be enough for something, namely, in each case, disagreement may arise over a given affirmation because of differing presumptions regarding the situation being contrasted with the actual situation. As noted above, people may not agree concerning what is to be selected as (one of) the cause(s) of an effect because they have in mind different causal contrast situations. Similarly, intuitions about whether to make a statement of the form ‘e is enough to a’ may arise from differences in what is imagined as the situation in which a-ing is thought of as a possibility. For instance, consider the Russian monk Rasputin, who supposedly was fed, by conspirators, cakes laced with enormous amounts of strychnine, yet who was not killed (at that point and by that means). He is said to have been one of those rare persons whose stomach fails to produce hydrochloric acid, which is a chemical interacting with strychnine so as to generate cyanide. Someone inclined to make the statement that the amount of strychnine that Rasputin ingested was enough to kill him might be imagining a contrasting situation wherein Rasputin has a normal digestive system, and so imagining that there is a real possibility that, were he to be killed in that situation upon ingesting the strychnine, it would be in virtue of ingesting it. Whereas, someone else might be disinclined to make the statement in question because of imagining a contrasting situation that still includes Rasputin’s gastric peculiarity, so that it is not a real possibility that, were Rasputin to be killed in that situation upon ingesting the strychnine, it would be in virtue of the latter occurrence. People might also differ in their statements as to what is enough to prevent a privation. For instance, judgments might differ as to whether the lock on the barn door is enough to prevent the horse’s absence from the barn. One person might think

7 A Causal Aspect of Epistemic Basing

101

of a contrasting situation in which the horse is imagined as calmly wandering up to and pushing open the unlocked door, whereas another person, concerned about a rash of recent lightning strikes in the area, may be imagining a contrasting situation in which the horse is invigorated by fear brought on by such a strike and is wildly beating on the door with blows having a quite unusual degree of force. Some differences in intuitions concerning the case of the gypsy lawyer might be explained in such a fashion. One person might imagine a contrasting situation in which the lawyer is cognitively normal and not subject to superstitious beliefs about clients, and that person might think of the lawyer as being in a situation such that, were prevention of a privation of a belief in the client’s innocence to occur, it would be in virtue of the possession of the evidence. A different reader might be struck by Lehrer’s having stipulated that if the lawyer were to lack his superstitious belief in the cards, then because of both the complexity of the reasoning from the evidence and the high emotions surrounding the crime, he would be led to believe that the client is guilty. Lehrer’s full description of those emotional factors did not make them seem cognitively abnormal. So, hesitancy to state that the evidence is enough to prevent privation of the belief-state might be comparable to the hesitancy of the person who declines to say that the lock on the door is enough to prevent the horse’s absence, since that person is thinking of the contrasting situation involving a lightning strike and ensuing equine frenzy. It is also helpful to notice a certain similarity between an assertion of the form, ‘the prevention of y is in virtue of x,’ and an assertion of the form, ‘w is (one of) the cause(s) of z.’ The latter presupposes causally relevant factors or background conditions that co-operate with w in explaining the effect, even though the assertion does not entail a description of those factors. Similarly, an assertion of the former sort presupposes but does not entail a description of factors that co-operate with x in explaining y’s absence. For instance, the assertion, concerning the Feynman example, that the presence of the radioactive material is enough for an explosion of the bomb, presupposes but does not entail a description of contextual details of the setup concerning the Geiger counter and its connection to the bomb.12 Again, one’s reasons may be enough for the prevention of a privation of belief even though co-operating causal factors involved in the prevention include one’s going through a line of reasoning deploying those reasons. Although enoughstatements are relative to contextual considerations, the salient interests surrounding attributions of knowledge that p require us to understand the application of condition C as relative to an imagined contrast situation within which the person retains the

12

I am tempted to construe satisfaction of condition C as involving a causal power to prevent the obtaining of a privation of the belief that h, a power belonging to one’s having the relevant reasons, and to consider satisfaction of C as involving what Harré and Madden call a state of readiness to manifest such a power (cf. Shope 1988). Nonetheless, I anticipate Klein’s raising the concern that taking a stand on those issues involves risky empirical speculations. Although people do understand mechanics sufficiently to tell whether a lock on the barn door is in a state of readiness to prevent the horse’s getting out, it is unclear what similar information we have concerning the mental nexus involved in epistemic basing.

102

R. K. Shope

cognitive characteristics that in the actual situation are needed for the justified status of believing that p. In some cases in which believing that p is based on reasons r1,. . ., rn, the relevant justified status arises not merely in virtue of the possession of r1,. . ., rn, but also in virtue of appreciating those reasons as reasons, and perhaps even partly in virtue of deploying them in a line of reasoning to the conclusion that p. Even though such cognitive details are not mentioned within the relevant instantiation of condition C, an understanding of C involves understanding how it is applicable to varying situations, which allows the wording of C to be at least a start on providing a succinct, non-disjunctive formulation of a rough schema for a causal aspect of epistemic basing.13

7.13

The Unimportance of Superstition and Prejudice

Up to this point, I have found it useful to regard the content of the gypsy lawyer and Raco examples as dealing with belief-states of those individuals. It is time to consider how Lehrer came to characterize the content of those examples in a manner that actually made them unsuitable for his purposes and unsuitable for appraising requirement C. Nonetheless, a consideration of his characterization will invite construction of an example that is more useful for present purposes. I have been discussing the manner in which Lehrer deals with the gypsy lawyer example in his book, Knowledge. He revisited the example in Theory of Knowledge after deciding to distinguish believing from a special kind of acceptance, and having come to maintain that it is the latter state that is requisite for knowing (see Lehrer 1990, 10–11). I shall follow him in employing the root, ‘accept,’ in order to speak of that admittedly special kind of acceptance. Lehrer tempted commentators to misinterpret him and to think that he was defining acceptance as a species of believing when he wrote near the beginning of the first edition of Theory of Knowledge, “We may, however, consider the appropriate kind of acceptance to be a kind of belief. . . . We gain some continuity with tradition as well as some expository simplification by considering acceptance to be a special kind of belief” (1990, 11). That Lehrer only temporarily sought such continuity and expository simplification emerged later in the book when he reconsidered the distinction in question and allowed that only for the most part is what we accept also something that we believe (see 113).14 Accordingly, Lehrer

13

This is the extent to which I take account of the view, of Keith Allen Korcz (2015, 9), that when seeking to specify conditions of epistemic basing, it would be odd to suppose that “the conscious evaluation and acceptance of the evidential import of a potential reason for a belief was, in principle, completely irrelevant to whether the belief was based on the reason.” Even though the statement of condition C does not mention such evaluation and acceptance, spelling out the ways of applying condition C to some examples will mention it. 14 In clarifying this matter in the second edition, Lehrer speaks of the usefulness of considering belief as accompanying acceptance, adding, “but we shall recall that it is acceptance aimed at truth

7 A Causal Aspect of Epistemic Basing

103

spoke of knowledge as a species of acceptance consisting in the recognition of information or the combination of such recognition with a positive evaluation of it (see 1990, 34; 2000, 39). The relevant claim that Lehrer wished to dispute became the claim that the origin of a state of acceptance is part of what converts it into knowledge.15 It is unclear whether Lehrer could regard a version of the gypsy lawyer case concerning acceptance as a plausible example of knowledge. In his original presentation of the case, Lehrer spoke of the lawyer’s response as being for the wrong reason, namely, “his faith in the cards,” which has “nothing to do whatever with his having evidence” (Lehrer 1974, 125). Again, it is unclear whether the Raco case could be consistently revised so as to describe racism as what originates Raco’s acceptance of the hypothesis about the disease. For in a later remark concerning the individuals in his original examples, Lehrer wrote that superstition or racism “is not the correct causal account of why they professionally accept what they do. They accept what they do because they follow cogent canons of reason and evidence” (Lehrer 2012, 168). It is also unclear that a suitably altered version of the Raco case could have details concerning acceptance that illustrate something parallel to a claim that Lehrer makes concerning belief, namely, that “evidence that justifies a person’s belief may be evidence she acquired because she already held the belief, rather than the other way around” (Lehrer 1990, 169; cf. 2000, 196). Perhaps Lehrer’s position should be appraised instead in relation to an acceptance variant of the following case concerning belief: THE SUPPLEMENTARY BASIS: Ms. Supp is a researcher who comes to know that a previously unsuspected phenomenon exists, thanks to finding a quite complex line of reasoning from her evidence. Having such a basis for her belief in the phenomenon motivates her to pursue additional research in order to find some evidence that can be used in a simpler way to prove its existence. The effort pays off, causing her to find such new evidence, thereby supplementing her original basis for believing in the phenomenon with an additional basis.16

This example accords with my account, which has explored the traditional perspective of treating knowledge that p as a species of believing that p, or at least as sometimes requiring such belief. Supp’s belief in the relevant phenomenon originated from possession and appreciation of the reasons that constituted her initial

that is genuinely required for knowledge, not the belief that accompanies it” (Lehrer 2000, 14; cf. 124–125). 15 Instead, Lehrer writes that what he disputes is the claim that “it is what originates a belief that converts it into a justified belief and knowledge” (Lehrer 1990, 168–169; cf. 2000, 195). 16 It is not one of my present goals to explore the intricacies of Lehrer’s views concerning what he calls acceptance, so I leave open whether his account does permit an acceptance variant of this case. Perhaps in his terminology, any change in what counts as a basis for Supp’s accepting the existence of the phenomenon by definition involves a new instance of accepting it, replacing the prior instance, so that when her supplementary basis is acquired, there exist two, mutually independent, causes of the new instance of the relevant acceptance state, whether or not the latter did possess existential inertia.

104

R. K. Shope

basis for believing. Since that combination of possession and appreciation did prevent the presence of a privation of a belief in the phenomenon, it was enough to prevent the presence of such a privation. In light of the similarity between that combination and Supp’s appreciation and possession of her supplementary basis, my intuition is that each basis satisfies requirement C.17

7.14

A Needed Refinement

I have spoken of C as a rough schema because it places no restrictions on the manner of prevention that it concerns. Yet certain restrictions may be needed. In order to allow for this, the schema could be amended as follows: (C*) S’s possession of the members of {r1,. . ., rn} is enough to prevent in such-and-such a manner the presence of a privation of believing that p.

Let us see why the italicized refinement might be needed, while admitting that epistemologists have continued to disagree concerning what conditions should supplement the analysans within a JTB account of knowing (in order to exclude various examples of inferential ignorance, including Gettier-type examples). Relative to a given epistemologist’s improved analysans, let us label ‘A’ whatever the analysans includes as that supplement. In some proposals, A has been concerned with TYPE 1 cases of reason-based knowledge but not with TYPE 2 cases. So merely adding the relevant instantiation of requirement C to such an analysans will render the expanded analysans too broad. It will be satisfied by those TYPE 2 cases where the only imaginable manner of prevention available to S resembles one that had to be excluded from some TYPE 1 examples, for instance some TYPE 1 Gettier example. It seems likely that an epistemologist who finds the latter sort of TYPE 1 example intuitively unacceptable to classify as knowledge will have a similar intuition concerning the TYPE 2 case in question. This concern can be illustrated by a variant of the original Nogot case. Suppose that S forms the belief that q: ‘Someone in the office owns a Ford’ for bad reasons and later acquires the evidence from the previously reliable officemate Nogot who is described in the original example (evidence such as Nogot’s driving a Ford in front of S and displaying a purported certificate of ownership), where S does not realize that Nogot is shamming and it is instead some other officemate who owns the Ford. Further suppose that at that point S’s possession of the evidence is enough to prevent the presence of a privation of the belief that q but only in a way that mirrors the inference pattern in the usual Nogot case, e.g., S might come to believe in light of the

17

Beliefs fail to satisfy the condition if they are psychologically insulated from relevant evidence in a way that renders it really impossible that were a privation of such a belief to be prevented, it would have been by evidence or information encountered by the person. This still leaves open the possibility of such a “teflon” belief’s “changing, or even being given up upon the receipt of some information unrelated to the content of the belief” (Lipson and Savitt 1993, 57n3).

7 A Causal Aspect of Epistemic Basing

105

new evidence the falsehood that Nogot owns a Ford, and consequently infer q. The mere detail that in this variant the belief had the earlier origin should not be sufficient to shift S into the position of knowing that q. I expect that treating C* rather than C as the schema for a causal requirement of knowing will allow us to deal with this difficulty thanks to the qualification in C* concerning the manner of prevention of the relevant privation. We may anticipate that an epistemologist who has provided an analysans adequate to cover every TYPE 1 case will be able to draw upon the conceptual materials involved in such considerations so as to apply those materials to TYPE 2 cases in a fashion analogous to, albeit somewhat different from, the manner in which they apply to TYPE 1 cases.

7.15

An Illustration: The Possibility of an Extended Reliabilism

In order to illustrate the preceding possibility, consider an objection that the reliabilist Alvin Goldman has made to Lehrer’s gypsy lawyer example: “To the extent that I clearly imagine that the lawyer fixes his belief solely as a result of the cards, it seems intuitively wrong to say that he knows—or has a justified belief—that his client is innocent” (Goldman 1979, 22n8). Keith Allen Korcz remarks that although Goldman does not elaborate this objection, “the concern is that the lawyer has not taken his good reason into proper account when forming his belief that his client is innocent” (Korcz 2015, 4). However, this reference to the formation of the belief takes us back to the time of the card reading, when the lawyer did not have the good reasons. It is instead possible that Goldman regards ‘fixing’ a belief that h as entering into a state of quite firmly believing that h. Goldman may be concerned that Lehrer portrays the lawyer’s acquisition of good reasons as neither an efficient cause nor a sustaining cause of either the lawyer’s believing in the client’s innocence or of that belief’s being a firm one. In dealing with Goldman’s objection, my perspective allows us to distinguish two versions of the gypsy lawyer case. In one version, the possession of the good reasons occurs in a context in which the lawyer appreciates them and they are enough for the preventing of a privation of belief in the innocence of the client. In a second version, the context does not include such appreciation, nor is the possession of the reasons a sustaining cause of the belief-state. My intuition is that only in the first version does the lawyer know that the client is innocent. Standard forms of reliabilism will not attribute the origin of the lawyer’s belief in either version to a reliable process. Yet it does not stretch the spirit of reliabilism unfairly to contemplate a species of reliabilism that construes the first version of the case as one in which the lawyer’s good reasons are enough to prevent in a suitably reliable way the privation of belief in the client’s innocence. The belief in some sense would then be fixed, and knowing would still have a reliabilist aspect.

106

7.16

R. K. Shope

The Future of an Illusion?

One who sympathizes with Klein’s caution might ask, “What if the judgments about causation that you treat as being about TYPE 1 cases involve an illusion of the causing of belief by possession of reasons, and there is only a common cause of both the belief and the possession of reasons, presumably some hidden brain-state?” I grant that if we are subject to such an illusion then it is safer to treat all relevant cases of epistemic basing as being of TYPE 2. Nonetheless, when aiming to instill in someone the knowledge that p, it will still make sense to go through the process of providing the person with good reasons to believe that p. The process will, if successful, include eliciting the hypothesized common cause, and so the acquisition of the reasons will be enough for the person not to have lacked the belief that p. Accordingly, I propose a further refinement in my schema for a causal condition of epistemic basing: (C**) The possession by S of the members of {r1,. . ., rn} is enough for the presence of a privation of S’s believing that p to be prevented in such-and-such a manner.

A relevant instantiation of the schema will be satisfied whether or not we are subject to the illusion in question in the face of the alleged common cause.

References Anscombe, G. E. M. (1957). Intention. Oxford: Basil Blackwell. Anscombe, G. E. M. (1981). Causality and determination. In Metaphysics and the philosophy of mind, collected philosophical papers of G. E. M. Anscombe (Vol. II, pp. 133–147). Minneapolis: University of Minnesota. Anscombe, G. E. M. (1983). The causation of action. In C. Ginet & S. Shoemaker (Eds.), Knowledge and mind: Philosophical essays (pp. 174–190). New York/Oxford: Oxford University Press. Anscombe, G. E. M., & Geach, P. T. (1961). Three philosophers. Oxford: Basil Blackwell. Armstrong, D. M. (1973). Belief, truth and knowledge. Cambridge: Cambridge University Press. Beaudoin, J. (2007). The world’s continuance: Divine conservation or existential inertia? International Journal for the Philosophy of Religion, 61(2), 83–98. Dretske, F. I. (1981). Knowledge and the flow of information. Cambridge, MA: Bradford Books. Goldman, A. (1979). What is justified belief. In G. Pappas (Ed.), Justification and knowledge: New studies in epistemology (pp. 1–23). Dordrecht: Reidel. Haldane, J. (2007). Privative causality. Analysis, 67(3), 180–186. Haldane, J. (2008). Gravitas, moral efficacy and social causes. Analysis, 68(1), 34–39. Haldane, J. (2015). Philosophy, causality, and god. In R. E. Auxier, D. Anderson, & L. Hahn (Eds.), The philosophy of Hilary Putnam. The library of living philosophers volume XXXIV (pp. 683–701). Chicago: Open Court. Hare, P., & Madden, E. (1971). The powers that be. Dialogue, 10(1), 12–31. Harré, R., & Madden, E. (1975). Causal powers. Oxford: Basil Blackwell. Hornsby, J. (2011). Actions in their circumstances. In A. Ford, J. Hornsby, & F. Stoutland (Eds.), Essays on Anscombe’s intention (pp. 105–127). Cambridge, MA/London: Harvard University Press.

7 A Causal Aspect of Epistemic Basing

107

Klein, P. D. (2012). What makes knowledge the most highly prized form of true belief. In T. Black & K. Becker (Eds.), The sensitivity principle in epistemology (pp. 152–169). Cambridge: Cambridge University Press. Klein, P. D. (2014a). No final end in sight. In R. Neta (Ed.), Current controversies in epistemology (pp. 95–115). New York: Routledge. Klein, P. D. (2014b). Reasons, reasoning, and knowledge: A proposed rapprochement between infinitism and foundationalism. In J. Turri & P. D. Klein (Eds.), Ad infinitum: New essays on epistemological infinitism (pp. 105–124). Oxford: Oxford University Press. Korcz, K. A. (2015). The epistemic basing relation. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Fall 2015 ed.). http://plato.stanford.edu/archives/fall2015/entries/basing-episte mic/. Lamprecht, S. P. (1967). The metaphysics of naturalism. New York: Appleton-Century-Crofts. Lehrer, K. (1974). Knowledge. Oxford: Oxford University Press. Lehrer, K. (1990). Theory of knowledge. Boulder/San Francisco: Westview Press. Lehrer, K. (2000). Theory of knowledge (2nd ed.). Boulder/Oxford: Westview Press. Lehrer, K. (2012). Cognition, consensus and consciousness: My replies. Philosophical Studies, 161 (1), 163–184. Lipson, M., & Savitt, S. (1993). A dilemma for causal reliabilist theories of knowledge. Canadian Journal of Philosophy, 23(1), 55–74. Makin, S. (2000). Causality and derivativeness. In R. Teichmann (Ed.), Logic, cause and action: Essays in honor of Elizabeth Anscombe (pp. 59–71). Royal Institute of Philosophy, Supplement 46. Cambridge: Cambridge University Press. Michotte, A. (1963). The perception of causality. London: Methuen. Osborne, T., Jr. (2007). Rethinking Anscombe on causation. American Catholic Philosophical Quarterly, 81(1), 89–107. Plato. Meno, translated by Adam Beresford, 2006. New York: Penguin Classics. Rosenberg, J. F. (1998). Kant and the problem of simultaneous causation. International Journal of Philosophical Studies, 6(2), 167–188. Shope, R. K. (1967). Explanation in terms of ‘the cause’. Journal of Philosophy, 64(10), 312–320. Shope, R. K. (1988). Powers, causation, and modality. Erkenntnis, 28(3), 321–362. Shope, R. K. (1999). The nature of meaningfulness: Representing, powers, and meaning. Lanham: Rowman & Littlefield. Shope, R. K. (2015). The state of affairs regarding true assertions. In R. E. Auxier, D. Anderson, & L. Hahn (Eds.), The philosophy of Hilary Putnam. The library of living philosophers volume XXXIV (pp. 365–384). Chicago: Open Court. Teichmann, R. (2008). The philosophy of Elizabeth Anscombe. Oxford/New York: Oxford University Press.

Part II

Scepticism

Chapter 8

The Moral Transcendental Argument against Skepticism Linda Zagzebski

Abstract Transcendental arguments against skepticism argue that the skeptical argument depends on, as a necessary condition for its formulation, the falsehood—or, more weakly, the unbelievability—of the skeptical hypothesis. In this chapter, I offer a series of arguments that the skeptic’s hypothesis needs to presuppose the moral or practical rationality of the subject, requiring either the existence of an external world with certain features (strongest arguments), the falsehood of the skeptical hypothesis (strong arguments), or the subject’s belief in such a world (weak arguments). In all the proposed transcendental arguments, the states of the agent requiring the falsehood or unbelievability of the skeptical hypothesis are not representational states such as the beliefs the skeptic asks the subject to doubt. I begin with an argument from rational agency, and I then investigate the sense of moral obligation, moral motives, and virtues in the vat, and argue that although the skeptic needs to presuppose the rational and moral agency of the subject, the skeptical hypothesis denies or undermines the subject’s agency. I end by considering whether the skeptic can retreat to Pyrrhonian skepticism to save the skeptical project, and I conclude that he cannot. Keywords Kant · Transcendental arguments · Responses to skepticism · Cartesian skepticism · Skeptical hypothesis · Practical rationality · Moral agency · Rational agency · Epistemology

8.1

Introduction

Transcendental arguments aim to reveal the necessary conditions for something that is not in dispute. In transcendental arguments against skepticism it is argued that a necessary condition for something the skeptic needs to formulate his skeptical argument—e.g., having a concept, being aware of a sequence of mental states, L. Zagzebski (*) Department of Philosophy, University of Oklahoma, Norman, OK, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_8

111

112

L. Zagzebski

being the subject of an experience—is the falsehood or, more weakly, the unbelievability of the skeptical hypothesis. What makes transcendental arguments particularly interesting is that they attempt to show that the mental states whose existence requires the existence of an outer world are not states directed towards such a world. In Kant’s classic version of this form of argument, one’s awareness of the temporal order of one’s states of mind requires a temporally ordered world outside the mind, and thus to regard a temporal ordering of the world as a dubitable hypothesis is to use that same order to deny its existence.1 In recent philosophy, it has been proposed that there are yet other mental states or capacities presupposed by the skeptic that require an external world.2 All these arguments propose that there is a path from mental states to an external world other than the path from representational states to what they are allegedly representations of. Most of the arguments I will try out in this chapter are in that category. Transcendental arguments come in different degrees of strength. Barry Stroud argued in an influential paper that they succeed at best in showing the necessary conditions for thought— roughly, they show how we have to think about things, not how things have to be (1968). Nevertheless, Stroud and others maintain that a weaker kind of transcendental argument is still useful (see 1999, 2002). Some of the arguments that follow are weak transcendental arguments. They aim at demonstrating that a person cannot consistently think of the skeptical hypothesis as grounds for doubting her own beliefs. Others have the more ambitious aim of demonstrating that the skeptic’s position is inconsistent; it entails two propositions that cannot both be true. However, the latter is not the strongest anti-skeptical argument, since it is possible that the skeptical hypothesis is self-inconsistent even though there is no external world. Perhaps the non-existence of an external world arises from something independent of the skeptical hypothesis. At least one of the arguments that follow aims at demonstrating the existence of an external world. The arguments of this chapter therefore have three degrees of strength. Since transcendental arguments involve a careful look at the skeptical hypothesis (SH), it is important to get clear on what the skeptic is proposing. In a general way, the strategy of the SH is clear: It is possible that our access to the external world is veiled. Everything could be exactly as it seems to me even if there is nothing behind the veil. I could be living in a virtual reality machine—or in standard philosophical parlance, I could be a brain-in-a-vat (BIV)—and if I were a BIV I would have no way of finding out that I am a BIV. The way in which the skeptic moves from the

See Immanuel Kant, “Refutation of Idealism,” in Critique of Pure Reason (2nd edition). Kant did not apply the term “transcendental argument” to the form of argument he presents there, but it is common to call it a transcendental argument in contemporary philosophy. 2 A famous variation of the Kantian-style argument is Hilary Putnam’s anti-skeptical argument in “Brains in a Vat” (1981). Putnam uses his semantic externalism to argue that a being that can raise the question, “Am I a brain in a vat?” is not a brain in a vat. When we think “brain” or “vat” we are thinking about brains and vats in an external world because we are causally connected to that world. But a being that has never had contact with external objects is not asking a question about those objects when thinking “brain” and “vat.” 1

8 The Moral Transcendental Argument against Skepticism

113

possibility that I am a BIV to doubting my beliefs varies. Sometimes the SH is used to attack my knowledge of an external world; sometimes the SH is used to attack my justification for believing an external world exists. In any case, the intent is to undermine the supposed epistemic status of my beliefs in such a way that the appropriate response is to doubt them. There is a connection between understanding the skeptic’s argument and accepting it, and between accepting it and doubting. Descartes maintained that the reason for doubt grounded in the hypothesis of a deceiving God “is very tenuous and, so to speak, metaphysical” (Meditation III.36). Presumably, a tenuous reason for doubt grounds only a tenuous doubt. Nonetheless, a doubt is a doubt; the skeptical hypothesis would not be interesting were it not for the fact that we are supposed to change our epistemic attitude towards our beliefs in response to it. I assume, then, that skepticism in its more interesting form is a practical problem.3 In any case, that is the form I will be considering. We need one other assumption to set up the moral transcendental arguments I want to explore in this chapter. Given that the skeptic wants me to doubt, the skeptic assumes I am a subject capable of entertaining the skeptical hypothesis and doubting. At the outset, we may assume that I am the thinnest subject capable of understanding the SH and doubting, but we will need to investigate the issue of how thin the subject can be, consistent with the SH. In proposing the SH, the skeptic must have a conception of the subject whose position in the world (or the vat) is given by the SH. This includes the conditions under which a subject such as I can perform the mental acts and operations the skeptic acknowledges I perform. The issue is whether my doing that is compatible with its being the case that my beliefs about an external world are not of such a world. The purpose of the next section will be to identify the minimal conditions for being a subject that are needed to use the skeptical hypothesis as grounds for doubt, and to use these conditions as a first step in undermining the skeptical project.

8.2

The Transcendental Argument from Rational Agency

When the skeptic proposes the SH to get me to doubt my beliefs, the skeptic makes a number of assumptions, some of which are as follows: I have beliefs, I can understand what the SH is, I can understand that the SH possibly applies to me now, I can

3

There are widely differing views about the practical relevance of skepticism. At one extreme, Greco (2000) argues that the real skeptical problem is purely theoretical and has nothing to do with doubt. But doubt is the focus of most ancient and modern skepticism, although there is dispute about the relevance of skeptical doubt to ordinary life. Some commentators claim that for the ancient skeptics, doubt was a practical matter, whereas it was purely methodological for Descartes. Gail Fine disputes this position in “Descartes and Ancient Skepticism: Reheated Cabbage?” (Fine 2000), arguing that Descartes and the Pyrrhonists were much closer to each other in their views than is generally realized. Both understood skepticism as affecting practical life, and both attempted to prevent skeptical doubts from leading to inaction.

114

L. Zagzebski

understand the connection between the possibility that the SH is true and doubting my beliefs, I ought to believe that I ought to doubt my beliefs, it is true that I ought to doubt my beliefs, and I can doubt my beliefs. So I am a being who is able to think rationally to the degree necessary to perform the above mental operations, and who has beliefs. Some of the above assumptions rely upon a further assumption about the nature of belief. Unlike such states as pain, hunger, sensations, and passing thoughts, I am not the passive subject of my beliefs. This is not to say that I choose my beliefs. Nonetheless, the way I have a belief differs from the way in which I have a pain or a passing thought. If the skeptic assumes that a belief is the sort of thing it can be reasonable to doubt, the skeptic must also assume that it is the sort of thing it can be reasonable (in principle) not to doubt. There are norms for believing and doubting, whereas there are no norms for feeling pain or hunger. The skeptic assumes that norms for belief exist and that I am capable of grasping those norms and their application to my own beliefs. In proposing the SH, then, the skeptic assumes that I have a degree of both theoretical and practical rationality. This is not an optional assumption that the hypothesis merely permits and which could be altered without serious harm to the SH. Granted, there are many things about myself and my own mental states about which I could be mistaken and which could be eliminated from the class of beliefs in front of the veil without jeopardizing the skeptic’s strategy. Perhaps I can be mistaken in believing that I have certain sensations, emotions, desires, or intentions. If so, the set of beliefs in front of the veil shrinks. But there must be something left that is sufficient to permit me to understand the skeptic’s hypothesis and to act in response to it in the way the skeptic regards as rational. Notice that it is not the content of the SH that requires that I be rational, since we can imagine a scenario in which the beliefs in front of the veil shrink to less than the minimum necessary for me to grasp the hypothesis, thereby making it the case that I am not rational. Rather, it is the use of the SH by the skeptic to get me to doubt my beliefs or to alter my beliefs or my attitude towards my beliefs in some other way that requires that I satisfy the conditions for undertaking the project. In fact, for my initial argument in this chapter, it is not even required that I believe that I satisfy these conditions. The skeptic, however, is committed to the assumption that I satisfy these conditions. Here, then, is my initial transcendental argument. In proposing the SH as a reason to get me to doubt my beliefs or to change my behavior in some way as a rational response to the SH, the skeptic presupposes: (1) I am a rational agent. The sense in which the skeptic is committed to (1) can be minimalist. The skeptic has no reason to assume I must be a rational agent in a robust sense. But a being to whom oughts apply and who is capable of understanding rational grounds for believing and doubting is a certain kind of being. When I say the skeptic assumes I am a rational agent, I just mean that the skeptic thinks I am that kind of being. According to the SH, I should do a particular thing:

8 The Moral Transcendental Argument against Skepticism

115

(2) I ought to doubt my beliefs about the external world. Why should I doubt my beliefs, according to the SH? If the reason is that I ought to have true beliefs and on the SH I do not have true beliefs, then since “ought implies can,” it follows that I can have true beliefs, and hence I am not a BIV. This answer gives us a short transcendental argument against skepticism. To block that route, what the skeptic must have in mind is that I ought to doubt my beliefs not because I ought to have true beliefs, but because I ought not to have false ones, and I can avoid having false beliefs by doubting. This seems to be Descartes’s position when he says, “even if it is not within my power to know anything true, it certainly is within my power to take care resolutely to withhold my assent to what is false.” (Meditation I.24). Descartes seems to think, then, that I can avoid falsehood and that I ought to do so by withholding assent. Now the skeptic will admit that the logic of ought requires that sometimes if I ought to do x, I ought to do y, and if I am a rational agent, I understand that. For instance, there are logical connections between the ought that applies to beliefs and the ought that applies to acts, and this is no less true if the only acts I can perform are mental acts. This point can be put in terms of justification. I am not justified in acting unless I am justified in the beliefs upon which my act is based. In any case, (1) requires: (3) I ought not act on any belief I ought to doubt. (3) needs to be qualified. I can rationally act on what I ought to doubt if I can rationally judge that there is much to be gained and little to be lost by acting on a dubitable belief. When I am in a desperate situation, for instance, one in which it appears almost certain that I will die no matter what I do, it can be rational to attempt to save myself by acting in a way that has an extremely low chance of success. So it is rational to act for ends under the assumption that I can weigh the probability of success as well as the value of what is gained and what is lost in acting. But when I entertain the SH, I am not in that situation. It is not that I judge the SH to be probable and I am desperate, but neither can I judge the SH to be improbable, since one of the things the SH does to me is render me unable to judge the probability of anything external to the mind. Doubt need not rest on a judgment of probabilities. But since the skeptic wants me to conclude that the appropriate response to the SH is to doubt my beliefs, the skeptic must also think that I should doubt that my acts have the external consequences I intend. But that is not all I should doubt. When I act, my act is not only based upon beliefs about the ends of the acts, it is also based upon beliefs about what it is I am doing, any practical or moral norms governing the act, and the trustworthiness of any other state motivating my act, such as an emotion of trust, fear, compassion, or admiration. If I engage in practical reasoning before acting, there will also be beliefs used in such reasoning. Almost all these states are beliefs about an external world or emotions directed at such a world. So:

116

L. Zagzebski

(4) On the SH, virtually every act I perform except the act of doubting my beliefs is based upon beliefs about the external world or emotional states whose intentional object is putatively something in the external world. It follows from (2) and (4) that: (5) On the SH, virtually every act I perform except the act of doubting my beliefs is based upon beliefs I ought to doubt or untrustworthy emotional states. Either I do act on the beliefs the skeptic argues I should doubt, or I do not. If I act, I am violating (3) and I am not a rational agent because I am not rational. If I do not act, I am not a rational agent because I am not an agent. It follows that according to the SH: (6) I am not a rational agent except in the performance of the acts specifically required to entertain the SH and to doubt my beliefs. But (6) is highly problematic. The problem is that the skeptic must explain the conditions under which an ought applies to me without also committing himself to oughts that apply to me at other times. At a minimum, the SH must be described in a way that removes the puzzling feature of how it is that the skeptic can propose the SH to someone who, by the skeptic’s lights, is not a rational agent in advance and will not be a rational agent once she follows through with the project the skeptic proposes to her. The skeptic’s position at this point is mysterious, although perhaps not incoherent. At a minimum, its peculiarity is something the skeptic should address. The SH should include a determinate conception of me as the subject of doubt that is not itself in doubt, in addition to a determinate conception of that which I am called to doubt on the hypothesis. There is also a weak version of the above argument. One could argue that whether or not I can consistently be a rational agent and a BIV, I cannot consistently think of myself as a rational agent and as a BIV. In entertaining the SH, I am entertaining the possibility that I am not a rational agent. But if I think of myself as not a rational agent, I cannot simultaneously think of myself as a being who is required by rationality to doubt, as a being who ought to doubt. If I cannot consistently entertain the SH and think of myself as a being who ought to doubt, the skeptic fails in his purpose. This argument does not show that I am not a BIV, since we have not ruled out the possibility that this is a world in which I am a BIV who should not doubt. It is even possible that I am a BIV who is a skeptic of a certain kind. I have argued that the SH appeals to the rational agency of a person, but the modes of the Pyrrhonists do not. The natural outcome of thinking through the ancient modes is a state that they found desirable, even pleasant, and arguably they found it pleasant precisely because it is a state of non-agency. They were freed from the anxiety of making and acting on judgments. Consistency might constrain our conception of ourselves even when we are not attending to it, so it is not clear that the Pyrrhonian strategy will succeed, but I am not ruling it out. Since this volume is in honor of Peter Klein, who has had a

8 The Moral Transcendental Argument against Skepticism

117

special interest in Pyrrhonism, I will return to the topic at the end of this chapter as an attempt to make skepticism immune to the arguments of the chapter.

8.3

Vat Morality

The argument of Sect. 8.2 gets us some distance towards identifying the necessary conditions for being the kind of being to whom the skeptic proposes his argument, but it does not uncover some of the most important features of rational agency that the skeptic cannot ignore. Many philosophers have thought that there is a strong connection between rational agency and moral agency. A being that is not a moral agent is not a rational agent, is not the kind of being to whom the skeptical argument is addressed. In this section, I want to investigate what happens to moral agency in the vat. What I mean by value externalism is the view that a state/motive/intention/act of a person gets its value at least in part from something outside the person’s mind. What I mean by value internalism is the position that value-making conditions are wholly internal to the person’s mind. If I am a BIV, then according to the usual vat scenario, my acts/states/motives have no externalist moral value. I assume that the connections between my mind in the vat and the external world are not the right ones to give my mental states/acts externalist value. Vat designers may not all have the same purposes, but the interesting vat scenarios are ones in which the world either is not at all what I think it is, or it is not related to me in the way I think it is. So if my acts/ motives/states in the vat have any value, it is internalist. The issues I want to raise next are these: What are the conditions under which my acts, intentions, and motives have internalist value? Can the skeptic consistently propose the SH to me as grounds for doubting my beliefs without destroying vat morality along with morality? Can the skeptic do that consistently while maintaining my rational agency?

8.3.1

Moral Obligation

Let me begin with some observations about morality. I assume that to be a moral agent includes being under moral obligation, and although there are many views about the scope of moral obligation, everyone agrees that it is not limited to the obligation to have future mental states, the only thing I can do in the vat. If I am in a vat, I believe I have obligations to act in a world outside the vat, but I cannot do so. I may think that I have performed an act that is a moral duty or a virtuous act by, say, helping a suffering person, but I clearly have not, because there is no suffering person and I have not helped anyone. The only acts I can perform in the vat are mental acts. I can form intentions, but I cannot be obligated to form the intentions I

118

L. Zagzebski

can form in the vat since they are the wrong ones, and they are the wrong ones whether or not semantic externalism is correct.4 Suppose first that semantic externalism is the correct interpretation of what the vat inhabitant means when she thinks that she ought to help this suffering person and forms the intention to do so.5 In that case, what she means is what we (the non-vat persons) mean by saying she ought to have certain brain images (or other vat events). But it is false that she ought to have certain brain images. Hence, she thinks something false when she thinks she ought to help this suffering person and she is mistaken in what she directs her intentions towards when she forms the intention of helping the person. It cannot be the case that she ought to form intentions to do something that it is false that she ought to do. Suppose instead that when she thinks to herself that she ought to help this suffering person, she means what we call helping another person. Again she is mistaken if that is what she thinks, because she cannot help anybody and she is not obligated to do what she cannot do. And again, she is not obligated to even form the intention to help a suffering person, since she is not obligated to intend to do what she cannot do. To do it she would have to climb out of the vat. So whether or not semantic externalism is true, it is false that she ought to do what she thinks she ought to do and it is false that she ought to form the intention to do so. In the vat, all her beliefs about acts she morally ought to do overtly (in an external world) are false, and there is no overt act she morally ought to do. For the same reason, she cannot violate a moral obligation to act overtly or to intend to act overtly in the vat. But there is more to moral obligation than obligation to do or to intend to do specific overt acts. Kant argued that the only thing valuable in itself is a certain state of the inner self. The idea is that the truly morally valuable thing cannot depend upon any external circumstances, but only upon what I control—myself, stripped of all contingency. There is a self that a tyrant cannot reach, nor, presumably, can Descartes’ Evil Genius or the vat designer. What is valuable in itself and knowable as valuable a priori is a good will—fulfilling what that self commands me to do, to obey the Categorical Imperative (CI). An imperative can be right or wrong, but not true or false. I assume that this is what Kant meant by autonomy, which he considered an intrinsic feature of rationality. This move asks us to take the controversial position that moral value is internalist, but my interest in this chapter is not to argue against what I have called value internalism. Instead, I want to examine the issue of whether the skeptic can use Kantian internalism to save morality in the vat and preserve my rational agency. Suppose that the skeptic admits that the moral ought in the vat does not apply to overt acts or to intentions to perform particular overt acts, but he maintains that there

4 What I mean by semantic externalism is the view that the meaning of a term is determined in part by factors external to the speaker’s mind, such as the way the world is. Two persons could be in the same mental state and use the same term in thought or speech, yet they could be thinking about and talking about different things. 5 This option is modeled on the argument of Hilary Putnam mentioned in footnote 2 of this chapter.

8 The Moral Transcendental Argument against Skepticism

119

is still moral obligation in the vat. The skeptic might maintain that vat morality is the same as morality; both are internalist. Alternatively, the skeptic might agree that morality is externalist, but maintain that vat morality is internalist. Vat morality is fundamentally a matter of having a Kantian Good Will. Perhaps I can have a Good Will in the vat since I can will to be moral for the sake of the moral even though I cannot act on it. Since the moral ought is a command, not something that can be true or false, it is not subject to the doubt that applies to beliefs or to any propositional attitude threatened by the skeptical hypothesis. I apprehend the Categorical Imperative in a way I cannot doubt. The CI commands that I follow the CI motivated by my understanding of what the CI is. As Kant understands the CI, it commands me to act on universalizable maxims, to respect rationality wherever I find it, including rationality in myself, and to form a kingdom of ends with other rational agents. The CI commands autonomy: the adoption of the CI as a governing principle of acting. Surely I can do that in the vat. Or can I? The BIV can never correctly will any instance of the CI involving overt action, although it can have the general will to do its duty from a motive of duty. If that is sufficient to have a Good Will as Kant understands it, then Kant’s formulations of the CI should apply to it. Even though the vat inhabitant does not perform overt acts, what she takes to be overt acts have maxims which may or may not be universalizable, and she can will to treat what she takes to be rational beings as ends in themselves. But this brings out some ambiguities in the Categorical Imperative. Consider first Kant’s Formula of Humanity, his second formulation of the Categorical Imperative. There is more than one way to interpret this formula. On one interpretation, a being has a Good Will if, given that someone is a rational being, she wills to treat him as an end in itself. On this interpretation nobody can have a Good Will in the vat. The only rational being the BIV encounters in the vat is herself, and it is even doubtful that she can treat herself as an end, given that one of her duties to herself is to get morality right, and she cannot do that. On the second interpretation, a being has a Good Will if, whenever she believes someone is a rational being, she wills to treat that being as an end. She can do that in the vat. On this interpretation, then, the BIV can have a Good Will provided that she believes that she encounters rational beings in an external world. The same ambiguity appears in the formula of universalizability, Kant’s first formulation of the CI. On one interpretation, the formula reads: given that a person performs overt acts, she has a Good Will just in case she can will the maxim of her act to be a universal law. On this interpretation, the BIV cannot follow the formula of universalizability, since she does not perform overt acts. On the second interpretation, the formula reads: Given that a person believes she performs overt acts, what she takes to be overt acts have maxims, and she has a Good Will just in case she can will such a maxim to be a universal law. The BIV can satisfy this formula as long as she believes there is an external world and she acts in that world. In order to save moral obligation in the vat, suppose someone maintains that all the CI obligates us to do is to have the general will to do what is morally right because it is morally right, not to do any particular thing at any particular time. Systematic and radical ignorance about the external world would make most of what

120

L. Zagzebski

I will in particular cases miss its moral target, but perhaps my will would be autonomous, not because its maxims are universalizable or because it is a will to treat other particular persons in certain ways, but just because my will is directed towards non-specific duty. The BIV could will something like this: If there are any rational beings, I will to treat them as ends in themselves. Similarly, she can will: If I perform any overt acts, I will that their maxims be universalizable. That turns the Categorical Imperative into a Hypothetical Imperative, the antecedent of which is never satisfied in the vat. If that is sufficient for a Good Will, then the will of a BIV can be good. But there are some serious problems with taking this line. For one thing, a will that is good in this non-specific sense cannot be part of a kingdom of ends. Kant’s crucial fifth formulation of the CI, which sets out the point of morality from the Kantian perspective, cannot be followed in the vat. The vat inhabitant is not part of a community of rational beings. She is cut off from them entirely. But more importantly, the general “Good Will” is not the will of an agent if it is never the governing principle of any act. If it is the governing principle of an act, then in some particular case the BIV determines that this is a rational being or this is an act governed by the principle of universalizability. In that case, the CI reduces to one of the other two interpretations. Either she has a good will because she treats actual rational beings as ends in themselves, or she has a good will because she treats beings she believes to be rational beings as ends in themselves, and the parallel point applies to the formula of universalizability.6 I conclude that acts and intentions to act have no externalist moral value in the vat, and whether the BIV can have a Kantian Good Will depends upon the way we interpret the Categorical Imperative. If the Kantian Good Will is externalist in value, the BIV does not have a Kantian Good Will. If the BIV can have a Kantian Good Will, its value is internalist, but its having internalist value depends upon her belief that there is externalist value. So she cannot have a Good Will unless she believes the SH is false.

8.3.2

Virtue

Moral obligation does not exhaust morality. Motives and virtues are also bearers of moral value. Can a BIV have good or bad motives? Good or bad traits of character? Motives are the best candidate for bearers of internalist moral value, and for that reason they are the best candidate for a correspondence between morality and vat morality. What I mean by a motive is not an aim, but an emotion state that is operating to bring about an act, and I believe that emotion dispositions are

6 The argument above applies to any interpretation of the Categorical Imperative that is de dicto. For example, I might will the following: Help suffering people. But even though my will in that case is good in a sense—clearly it is better than having the de dicto will that suffering people be harmed— nonetheless, such a will is not the will of an agent unless it governs particular acts.

8 The Moral Transcendental Argument against Skepticism

121

components of virtues and vices.7 Examples of motives are envy, anger, fear, love, compassion, and pride. This section will address the value of a BIV’s emotions and traits of character. Although I will make some assumptions about the structure of emotion and virtue, the general approach does not depend upon these assumptions. Every account of the value of emotion and virtue requires that its value be either externalist or internalist. My arguments are designed to cover both possibilities and a wide range of views about the nature of virtue. I take it that emotions have an intentional structure. They are states of feeling a characteristic way about something before the mind. The vat inhabitant’s emotions are directed towards her own mental states, although she takes those states to be states of something outside her mind, and she has emotions directed at such objects. The character of her compassion is shaped by her taking the object to be in an external world. If she discovered that what she thought was outside her mind was really her own mental image, she would no longer feel the same compassion. The same thing holds for fear, love, and any emotion that she has under the assumption that the object of the emotion is something in the external world. This is not to deny that emotions can be directed towards fictional characters or make-believe creations of the imagination, as in movies and in children’s play, but the BIV is not making believe. She is making a mistake because her emotion is directed towards something she takes to exist in the external world and she is making a mistake about that. Her emotion is the wrong one for the circumstances, and she would agree if she became aware that there is no external object. One way of proceeding would be this. There is no value or disvalue in fearing a non-existent threat, being angry at a non-existent person, or being generous towards non-existent beings, and a trait of character in which a person characteristically fears the non-existent, is generous towards the non-existent, and so on, is not a virtue or vice. In addition, if we assume that a person does not have a virtue or vice unless he reliably brings about the aim of the motivational component of the virtue or vice, there is another reason why the BIV would not have a trait of character.8 Hence, on an externalist account of motives and virtues and vices, a BIV has no character. But as I mentioned, internalist intuitions about value are most likely to appear in an account of the value of emotions and their operation as motives. It does seem as if the BIV’s motives can be good and that she can have a good character even though she does not have virtues and good motives in the sense attributed by the externalist to persons outside the vat. If the BIV has motives that are good, given the way the world appears to her, surely there is a sense in which her motives are good. She deserves moral credit for them. We can even imagine a scenario in which the world is a skeptical scenario by design. Suppose God decides to judge people by their behavior in the vat world. This could even be a solution to the problem of moral evil, because if we are all vat

7 I have argued that a virtue has a motivational component that is an emotion disposition in several places, including Zagzebski (1996,2004). 8 I defend this view in Zagzebski (1996).

122

L. Zagzebski

inhabitants, nobody does anything that hurts other people. It just seems that way.9 If so, our eventual judgment by God could depend upon what we do in the vat. One could argue that what is vat-good is not good, and what is vat-evil is not evil, but whether or not vat morality coincides with morality, vat morality on this scenario is sufficient to ground what is often taken to be one important purpose of morality— reward and punishment. Notice that what makes my motives vat-good is ignorance. I cannot have vat-good motives while I entertain the SH and conclude that it succeeds in undermining the justification for my beliefs. The internalist value of my motives depends upon my believing that there is an external world to which I respond in my emotions and the acts they motivate. I argued in Sect. 8.3.1 that I cannot have a Kantian Good Will in the internalist sense of a Good Will if I think of myself as a BIV. For the same reason, I cannot have motives good in the internalist sense if I think of myself as a BIV. This leads to a weak moral transcendental argument that I will present in the next section. To summarize Sect. 8.3, if I am a BIV, my acts and intentions have no externalist value or disvalue. What I take to be overt acts are not overt acts, and my intentions to perform overt acts are not intentions directed at what has externalist moral value, so my intentions lack externalist moral value. All beliefs I have about what I am obligated to do overtly are false, and there are no true propositions about what I am obligated to do overtly. I can have a Kantian Good Will in the sense that I can have the general will to do whatever morality commands, but I cannot correctly will any instance of the CI that involves overt acts. If it is sufficient for a Good Will that the maxims of what I take to be my acts are universalizable and that I treat those I take to be rational beings as ends in themselves, then I can have a Good Will provided that the value of a Good Will is internalist and I believe that the value of my acts is externalist. Similarly, if motives and traits of character are good only if they have externalist value, I do not have good motives in the vat and do not have virtues or vices. My motives can have internalist value, and I can have that part of virtue that includes vat-good motive-dispositions, but only if I believe that the intentional objects of my motives are outside my mind. I must believe that the objects of my emotions as well as the acts to which those emotions give rise are in an external world. Internalist value depends upon the belief in an external world.

9 Matthew Hodge suggested to me that it could also be a partial solution to the problem of natural evil. The only natural evil any of us would have to worry about in the vat would be our own.

8 The Moral Transcendental Argument against Skepticism

8.4 8.4.1

123

The Moral Transcendental Arguments The Weak Moral Transcendental Arguments

The skeptic allows that I think of myself as a moral agent. In fact, I must think of myself as a moral agent if I am a rational agent. Believing that I am a moral agent includes the belief that there are moral oughts that apply to me. Moral agency is an intrinsic component of my sense of the self who allegedly ought to doubt by reflecting upon the skeptical hypothesis. This would be the case even if I have a perverse sense of morality. The skeptic does not deny that. So the skeptic accepts: (1) I think of myself as a moral agent. From Sect. 8.3 we get: (2) If I am a BIV, no externalist moral properties apply to me. I can have at best internalist moral properties such as a Good Will (on one interpretation) and vat-good motives. If there is vat-morality, it is internalist. (3) But I have internalist moral properties only if I believe that there are externalist moral properties. I must believe that there is an external world in order to have the motives and intentions that have internalist moral value. (4) Hence, I cannot consistently think of myself as a moral agent who is a BIV. The argument supports an even stronger conclusion: I cannot think of my beliefs about the external world as systematically false. I cannot consistently think of myself as a moral agent and as a BIV, and if I think of myself as a moral agent, I cannot consistently think of my belief in the existence of an external world as false for some reason other than the SH. It follows that the skeptic cannot use the SH as a way to get me to doubt my beliefs without destroying vat morality as well as morality. Since at least vat morality is required for me to think of myself as the subject of doubt, and since the skeptic cannot succeed if he prevents me from thinking of myself as the subject of doubt, the skeptic cannot succeed in using the SH as a way to get me to doubt my beliefs. A parallel argument can be given about vat practical rationality. I need vat practical rationality in order to doubt, but vat practical rationality is destroyed by the SH. So the skeptic cannot use the SH as a way to get me to doubt my beliefs. There is a second weak transcendental argument against skepticism from the assumption that I think of myself as an agent capable of acting out of a sense of the moral or practical ought. To the extent that I take the SH seriously, I lose my motivation to act, and hence I lose my agency. The problem is that there is not only an alethic connection between the rationality or justification of acts and the rationality or justification of the beliefs upon which the acts are based, as argued in Sect. 8.2; there is a psychological connection as well. We are not motivated to act unless we have confidence in the beliefs upon which the act depends. The beliefs upon which an act depends include beliefs about the way the world is and how a given act will causally affect that world. Without confidence in the truth of these

124

L. Zagzebski

beliefs, we will not be motivated to act. Further, as argued in Sect. 8.3, unless I believe in an external world, I will not be motivated to act out of motives whose appropriateness to the circumstances depends upon an external world, nor will I be motivated to act for ends whose value depends upon the existence of an external world. As a rational agent, I must act. In order to act I must be reasonably confident that the SH is false. Paralysis in action is not compatible with rational agency. Hence, it is not rational to take the SH as a ground for doubt. In the second weak moral transcendental argument, if I take skepticism seriously I lose my moral agency by losing my agency. In the first argument, if I take skepticism seriously I lose my moral agency by losing the moral value of my mental states and acts. It is not rational to lose my moral agency. Hence, I should not take skepticism seriously.

8.4.2

The Strong Moral Transcendental Argument

We can see next that the skeptical hypothesis is inconsistent with what the skeptic herself takes to be the truth about me or the subject of the hypothesis. According to the picture the skeptic is presenting to us, if I am a BIV, virtually none of my beliefs about the external world is true. The skeptic assumes that truth is an externalist value and falsehood an externalist disvalue, since that is why doubt is appropriate, but there is no externalist value in the vat. So either there is no value in the vat, or it is internalist. But now the skeptic has a dilemma. She must decide whether the value of the subject’s motives, acts, and aims is externalist or internalist. If it is the former, she must think that under the SH no state or act of the subject of the hypothesis has value. If it is the latter, she must think that whereas the value of the subject’s beliefs depends upon an external world, the value of her motives, acts, and aims is wholly internal to her mind.10 Suppose first that the skeptic thinks of beliefs, acts, emotions, intentions, and aims as all having externalist value if they have any value at all. That is, their value depends, at least in part, upon there being the right connection between the subject and an external world. On the SH, none of the subject’s states has value because none of them has externalist value. Recall that the argument of Sect. 8.2 relied on the idea that acts rest upon beliefs, so doubt about the latter ought to lead to doubt about the former. The issue I am bringing up here does not depend upon that assumption. The question is whether the skeptical scenario gives the subject a reason to doubt her beliefs on the grounds that they might be false if nothing she does, believes, feels, intends, or aims at has any value on the SH. How can should apply to her if nothing she can do has or results in either positive or negative moral or practical value? In fact, if no state of the agent has value on the SH, then it is very hard to see how the

10

Of course, beliefs have value other than their truth value, but it is the truth value that is the focus of the skeptical attack.

8 The Moral Transcendental Argument against Skepticism

125

subject can be either a practical or a moral agent. If the skeptic does think she is an agent, more needs to be said in describing the skeptical scenario. Suppose instead that the skeptic takes the other option. Can the skeptic consistently think of the SH as grounds for the subject to doubt because the value of beliefs is externalist, while also maintaining that the subject’s acts, emotions, intentions, and aims have only internalist value? To take this line the skeptic must say that the value of true belief is independent of the value of non-epistemic states, since the latter do not depend for their value on anything external. But notice that if the skeptic says the subject’s reasons for action are internalist in value, then the truth of a belief has no bearing on whether she should act on the belief or whether she is right or wrong in her emotions/motives, or whether she is aiming at the right thing, or whether her act expresses a good will. If the skeptic takes this option, she is saying the subject has no reason to take truth into account when interacting with the world she takes to be external. The skeptic would have to say to the subject, “It’s good to be in tune with external reality and I am proposing to you a hypothesis in which you are not in tune with external reality. But that has nothing to do with your practical or moral judgment about your acts or the appropriateness of your emotions.” If the subject cares about the value of truth, then she will care that her beliefs might be false, but the skeptical hypothesis thus interpreted is a far cry from the type of hypothesis that most philosophers consider a threat.

8.4.3

The Strong (Non-transcendental) Argument from the Categorical Imperative (CI)

According to Kant, a rational agent is necessarily a moral agent. Many philosophers have been persuaded of this position, whether or not they are Kantians. Given this position, if the skeptic allows that the subject is a rational agent, or at least is rational enough to know whatever can be known a priori, then if Kant is right that the Categorical Imperative can be known a priori,11 we can formulate a different strong moral argument against skepticism. This argument is not transcendental. It can be varied in many different ways for those who are not Kantian but who believe that a significant number of the demands of the moral law can be known a priori. I will give only the Kantian form of the argument here. (1) I apprehend the CI by reason alone. The CI obligates me unconditionally. I ought to obey the CI no matter what I think or believe and no matter what I feel or choose. (2) If I ought to obey the CI, I can do so. (3) Therefore, I am able to obey the CI.

11

Since the Categorical Imperative is not a proposition, the sense of knowledge used here must be broader than propositional knowledge.

126

L. Zagzebski

(4) If the CI applies in some situation, I must be able to truly believe that the situation has the features that make it such that the CI applies (e.g., This is the kind of being who deserves to be treated as an end in itself; If I say p, I am lying; S is suffering and in need of aid, and so on.) (5) Therefore, I am obligated to obey CI, and if CI obligates me in some situation, all the factual propositions that are necessary conditions for the application of CI in that situation are true and I must be able to believe them. (6) If CI obligates me in some particular situation, I must also be in a position to have the degree of confidence in the truth of these propositions necessary for the capacity to act upon them. (7) Therefore, if morality comes in the form of a Categorical Imperative, either the CI never applies in any particular circumstance, or I must be able to believe the CI, to have a number of true factual beliefs, and I must be in a position to be confident of their truth. (8) The CI applies in many particular circumstances if it applies at all. The CI is not the hypothetical imperative: If I ever encounter a situation of a certain kind, I am obligated to do x. (discussed in Sect. 8.3.1). (9) Therefore, I must be able to have many true beliefs about the external world and I must be in a position to be confident of their truth. This argument suggests that my awareness of myself as a rational agent under a categorical imperative is simultaneously an awareness of myself as engaged with a world outside my mind. The skeptic can abandon the latter awareness only by abandoning the former. This point also appears in different guises in the transcendental arguments. In brief, it is this: The skeptic cannot consistently endorse the subject’s sense of practical necessity (to which she appeals to get the subject to doubt her beliefs) while also denying the truth of the beliefs and the appropriateness of the motivating emotions and sense of obligation that underpin practical action. I have chosen to give the argument of this subsection in terms of a Kantian Categorical Imperative because it is so clear and well known, but the argument could be recast in terms of obligations of virtue or other obligations purportedly known a priori. My purpose is to call attention to the connection between the a priori judgments of moral agents and the capacities of those agents to which the skeptic appeals.

8.5

Can the Skeptic Retreat to Pyrrhonism?

In this essay, I have proposed a series of arguments for discussion. These arguments have been directed at modern skepticism, the type of skepticism intended to lead to massive doubt about our beliefs. But there are other historically important forms of skepticism. The ancient Pyrrhonists used skeptical arguments not to convince a person that she ought to doubt her beliefs, but to put her in a state of ataraxia, or quietude, a state thought to be blissful because the agent no longer feels the anxiety

8 The Moral Transcendental Argument against Skepticism

127

that accompanies the need to make judgments (see Sextus, I:25–30). The skeptical modes do not appeal to the rational agency of the subject, but function as a sort of medicine for the soul. When the medicine works, the skeptic lives a life “in the appearances,” without the commitment to the truth of the appearances that characterizes the behavior of ordinary persons. The Pyrrhonian skeptic therefore lives a very different kind of life than the normal one; indeed, that is the point of Pyrrhonian arguments. The Pyrrhonian skeptic may act, but she is untouched by the sense of practical necessity that motivates ordinary people to act in the characteristic way we describe as “out of duty” or out of a sense of what ought to be done. Is Pyrrhonian skepticism immune to the arguments of this chapter? It depends upon whether the Pyrrhonist makes judgments at the meta-level. He may suggest that we ought to make no judgments because we can never tell when we have got a judgment right. Arguably, the Pyrrhonists used the “ought implies can” rule to argue from the fact that we cannot tell reality from appearances to the judgment that we ought not to try. That is why we ought to live a life in the appearances. A philosopher ought to go through their standard modes and end up in a state of ataraxia. We who go through the skeptical modes may not be asked to conclude that we ought to doubt; nonetheless, the Pyrrhonists produce the modes because they think we ought to doubt. We ought to take our medicine. If this is what the Pyrrhonists are doing, they would be fair targets of the arguments of this chapter. But perhaps they are not doing that. The Pyrrhonian skeptic may not think that anybody ought to live a life in the appearances. They may simply see it as a pleasant life, one they would like to see others live also. If it is possible to live such a life and to desire it for others without a sense of ought, then we can live a type of skeptical life free from doubt and also, I have argued, free from agency. The psychological possibility of such a life and its coherence raise some very interesting questions, but the pertinent question for this essay is whether we can escape morality and practical necessity even if we can make ourselves believe that there is no such thing. If there is either a Categorical Imperative or hypothetical imperative whose antecedents are satisfied by the Pyrrhonian skeptic, or if identity-conferring practical necessities of the kind described by Bernard Williams apply to the Pyrrhonist,12 then the transcendental argument against skepticism or the argument from the Categorical Imperative cannot be avoided by the Pyrrhonist. If a contemporary Pyrrhonist thinks that that form of skepticism is immune to the arguments of this essay, he should make it clear that he is proposing moral nihilism along with skepticism. And, of course, the nihilism is not only in the realm of the moral, but extends to one’s entire practical life. I don’t know if it is possible to live without morality, but it is worth noticing that a person who lives without morality must also live without radical doubt. But whether or not it is possible for someone to live without morality, we should not think that morality is undermined by the skeptical hypothesis. I am operating with the assumption that the sense of moral or practical necessity does not derive from any states the veridicality of which can be attacked by the skeptical hypothesis.

12

See Williams (1993), especially Chapter 4.

128

L. Zagzebski

My position is that morality is not a deliverance of our epistemic states or any operation of theoretical reason. I think that it follows that its source is something independent of the states that are the object of attack in the standard skeptical arguments. If I am wrong about that, then the strong forms of the arguments of this essay cannot get going, but even if I am wrong and morality depends upon states that are vulnerable to skeptical attack, the skeptic must defend that position in his argument. The skeptic must walk a very thin line, addressing the skeptical hypothesis to a rational agent, but one whose rational and moral agency is in jeopardy if that being takes the hypothesis seriously. However, if I am right, morality is immune to the moves by which reason subverts itself in the best of the skeptical challenges. The truth of many beliefs is required by the authority of practical necessity; the latter does not derive from the former. Giving up morality is not an option, and it is our moral and practical sense that demands that there be true epistemic states.13

References Descartes, R. (1979). Meditations on first philosophy, translated from the Latin by Donald A. Cress. Indianapolis, Indiana: Hackett Publishing Co Inc. Fine, G. (2000). Descartes and ancient skepticism: Reheated cabbage? Philosophical Review, 109 (2), 195–234. Greco, J. (2000). Putting skeptics in their place: The nature of skeptical arguments and their role in philosophical inquiry. Cambridge: Cambridge University Press. Putnam, H. (1981). Brains in a vat. In Reason, truth, and history. Cambridge: Cambridge University Press. Sextus Empiricus. (1993). Sextus empiricus, outlines of pyrrhonism, translated by R. G. Bury. Loeb Classical Library No. 273. Cambridge, MA: Harvard University Press. Stroud, B. (1968). Transcendental arguments. Journal of Philosophy, 65(9), 241–256. Stroud, B. (1999). The goal of transcendental arguments. In R. Stern (Ed.), Transcendental arguments: problems and prospects (pp. 203–223). Oxford: Oxford University Press. Stroud, B. (2002). The quest for reality: Subjectivism and the metaphysics of colour. New York: Oxford University Press. Williams, B. (1993). Shame and necessity. Berkeley/Los Angeles: University of California Press. Zagzebski, L. (1996). Virtues of the mind. Cambridge: Cambridge University Press. Zagzebski, L. (2004). Divine motivation theory. Cambridge: Cambridge University Press.

13

I wrote an earlier version of this chapter many years ago and it was never published. I remembered how much Peter Klein liked to talk about the Pyrrhonists, and so I decided that completing this chapter with reference to Pyrrhonian skepticism would be a fitting tribute to him.

Chapter 9

Epistemic Humility, Defeat, and a Defense of Moderate Skepticism Sharon Ryan

Abstract This chapter will argue for a moderate form of skepticism. Moderate skepticism is the view that human beings know far less than is normally credited to them. Although I believe Peter Klein has refuted a radical form of skepticism that serves as familiar fodder for philosophical debate, the moderate form of skepticism I defend here is immune to Klein’s objections. This form of skepticism is not only plausible and interesting, but I believe it is supported by an argument with premises Klein ought to accept. My argument for moderate skepticism will rely on a defeasibility analysis of knowledge, an evidentialist theory of epistemic justification, the fact that we all suffer from epistemic limitations, and the claim that epistemic humility is a virtue. The argument maintains that when it comes to the important and interesting questions human beings ponder, we all suffer from epistemic limitations including insufficient evidence, imperfectly reliable memories, cognitive biases, epistemic peers who disagree with us, and underdeveloped cognitive skills. For the vast majority of human beings, these limitations serve as defeaters that undermine our capacity to know many of the things we wonder about and investigate. The virtue of epistemic humility requires us to be honest about our epistemic shortcomings and to embrace moderate skepticism and all that it suggests about learning, teaching, communicating, civil discourse, and the pursuit of wisdom. Keywords Pyrrhonism · Epistemic humility · Moderate skepticism · Epistemic limitations · Knowledge defeaters · Intellectual humility · Epistemic virtue · Klein · Epistemology

S. Ryan (*) West Virginia University, Morgantown, WV, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_9

129

130

9.1

S. Ryan

Introduction

Peter Klein’s book, Certainty: A Refutation of Scepticism, is a convincing argument against direct skepticism, the familiar and irritating brand of skepticism that is sometimes called Cartesian, radical, global, or external-world skepticism. Direct skepticism “claims that no person, S, can know that p, where ‘p’ stands for any proposition ordinarily believed to be knowable” (Klein 1981, 6). It is the kind of skepticism that denies that we have knowledge of any proposition about the external world. I will advocate for another, more moderate form of skepticism that Klein’s argument does not challenge. What distinguishes moderate skepticism from direct skepticism is that a moderate skeptic, as I’m using the term, does not claim that knowledge is impossible. In addition, the moderate skeptic allows for knowledge of simple claims about the external world. The moderate skeptic claims that we don’t know very much when it comes to the complex and interesting subjects we ponder. A moderate skeptic agrees with Klein that we know we have hands, that we are not mere brains in vats, that other people exist, and that we know plenty of other simple and easily accessible facts. Yet if moderate skepticism is true, we don’t know much about the distant past; we don’t know much about how our lives and our relationships will unfold; we don’t know the answers to difficult philosophical questions; and the answers to currently debated scientific wonders are beyond the scope of knowledge for most, if not all, inquirers.

9.2

An Argument for Direct Skepticism, and Klein’s Objection

Before developing my argument for a moderate form of skepticism, I will review, very briefly and simply, a standard argument for direct skepticism, along with the heart of Klein’s objection to the argument. I am over-simplifying here, but in summary, Klein’s argument against direct skepticism uses an evidentialist theory of epistemic justification and a very simple and plausible version of a closure principle—a version at use in many arguments in support of skepticism—to challenge direct skepticism. The skeptical argument challenged by Klein goes as follows (1981, Chap. 2): A Standard Argument for Direct Skepticism 1. If S is justified in believing p, then S is justified in believing the skeptical hypothesis is false. 2. S is not justified in believing the skeptical hypothesis is false. ∴ S is not justified in believing p. This argument is intended to apply to any person (S), and p can be any proposition about the external world. If the argument is sound, then any knowledge of the

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

131

external world is impossible because none of our beliefs about the external world are justified. For example, like Descartes, you might wonder if your body is real or just a pure figment of your imagination. You might wonder if the person you seem to be having a conversation with is real or just a hallucination. The skeptical hypothesis is any of a number of scenarios where p is false yet appears to S to be obviously true. There are many familiar versions of the skeptical hypothesis, ranging from the claim that you are just dreaming, that you are a mere brain in a vat being stimulated to think you have hands, that you are actually in the matrix rather than in the real world, etc. Pick your favorite. So, basically, PREMISE (1) claims that if you are justified in believing p is true, then you are justified in believing ~p is false. Klein concedes the hypothetical, first premise to the skeptic. The second premise of this standard argument for direct skepticism asserts that you are not justified in believing you are not a mere brain in a vat (or in ruling out any of the other skeptical hypotheses). The fact that you are fallible and your evidence does not rule out that you are deceived shows that you are not justified in believing you are not a mere brain in a vat. The point of the deception is to trick you with very realistic evidence! Klein’s response to the argument is to accept PREMISE (1), but then use it, and the rationale behind it, against the skeptic. Because most of us do have plenty of evidence supporting our belief that p is true, we do, via the evidence we have for p and the fact that the truth of p obviously entails that ~p is false, have plenty of reason to reject the skeptical hypothesis. Therefore, PREMISE (2) can be rejected, and the problem of direct skepticism is solved.1 As long as the other conditions for knowledge are satisfied, we are not only justified in believing what we ordinarily think we are justified in believing, but we know pretty much what we ordinarily think we know. Again, let me acknowledge that my one-paragraph summary of Klein’s intricately argued, book-length treatment of skepticism is over-simplified, but that’s his basic view in a nutshell. I think Klein’s book provides a convincing defense of the claim that many of our ordinary and obvious perceptual beliefs are justified. When you are seated by the fire (in a well-lit room) in your dressing gown staring down at your hands and feeling the sensation of having hands, without any reason to think that you are just dreaming, in a vat, etc., you are justified in believing you have hands. Even if you are a mere brain in a vat without hands, your belief that you have hands is justified by your overwhelming evidence. You don’t know that you have hands, but that is not because your belief fails to be extremely well justified by your evidence—it is. You will lack knowledge only because your belief is, despite being sufficiently justified, false. Klein’s argument puts direct, or radical, skepticism to rest.

1

Ned Markosian makes a similar point (Markosian 2014).

132

9.3

S. Ryan

Internalism and Evidentialism

There is, however, another more moderate form of skepticism that is defensible and immune from Klein’s objections to direct skepticism. Moreover, I believe moderate skepticism is justified by several plausible premises that Klein either does accept, or ought to accept. The argument for moderate skepticism will begin with an internalist and evidentialist theory of epistemic justification. As noted above, Klein makes use of internalism and evidentialism in his refutation of direct skepticism. Here, I will use Richard Feldman and Earl Conee’s well-known statement of the view: (EJ) Doxastic attitude D toward proposition p is epistemically justified for person S at t if and only if having D toward p fits the evidence S has at t (Feldman and Conee 1985).2

So, as long as your evidence supports believing p, belief is the epistemically justified attitude for you to take toward p. If your evidence concerning p is balanced or neutral, suspension of judgment is the appropriate doxastic attitude to have toward p. When your evidence goes against p, disbelief is the doxastic attitude that is justified for you. This is an internalist theory of epistemic justification because what matters is the evidence you possess, not facts or other forms of evidence external to your awareness. In response to the standard argument for direct skepticism, remember that Klein asks us to consider how things seem to us. Consider the evidence you have. Even if, in fact, there’s a vat of nutrients surrounding you and you are nothing but a brain, it seems exactly like you are a human being with a body, sitting in a chair, wearing clothing, enjoying a crackling fire in your fireplace, etc. I will not defend evidentialism or internalism in this chapter. Rather, I will assume that both are true and show how, if we accept these along with other plausible premises, we will be led to a moderate form of skepticism that many of us, including Klein, ought to embrace. Of course, if this internalist version of evidentialism is false, then Klein’s alleged refutation of direct skepticism fails along with it.

9.4

Epistemic Limitations and Epistemic Humility

The argument for moderate skepticism will assume that epistemic humility is (at least) an epistemic virtue, and this is another view Klein endorses (1985). Epistemic humility may well also be a moral virtue, but that is not a contention I need to defend here.3 We can understand epistemic humility as: an awareness of, and a proper response to, our epistemic limitations. An obvious example of an epistemic limitation is a lack of evidence. When I wonder if my neighbor, about whom I know absolutely nothing, is a racist, I suffer from a serious epistemic

2

t is a time. I have defended the claim that we have a moral obligation to believe in accord with our evidence in Ryan (2015). 3

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

133

limitation. I have no evidence on the matter. If an epistemically humble person considers a proposition p about which she has no evidence, she will refrain from believing p. An epistemically humble person restricts his or her beliefs to what her evidence supports—her doxastic attitudes are epistemically justified. But epistemic humility is more than this. If she is epistemically humble, she will also refrain from making proclamations about p and she will refrain from pretending to know p, and if the topic is important, she will attempt to learn more about the subject matter. Lacking evidence about p is one epistemic limitation, but there are many others. Even someone with the best evidence in the world has a serious epistemic limitation if she has trouble drawing basic inferences from her evidence. An epistemically humble epistemic agent with this type of cognitive disability will be aware of her limitation and refrain from believing (or trusting) her inferences. She’ll also refrain from pretending to know the results of her inferences, and she’ll refrain from asserting the results of her inferences to others. If possible, she’ll work to improve her skills. Yet another familiar epistemic limitation is having some decent evidence and believing exactly what the evidence you have supports, but the evidence you have is a small sample of the total evidence that is out there and beyond your current grasp. Perhaps this is because the subject matter is vast and complex and you have just scratched the surface. An epistemically humble person in such circumstances will acknowledge their situation, believe what their evidence supports, but proceed with caution and a mind open to learning much more. They won’t claim to have knowledge, and if asked about the subject matter, will respond appropriately, perhaps saying “well, I don’t know, but here’s what I believe, and my reasons for so believing are x, y, and z. . .”. Or, perhaps the subject is not especially complicated or vast, but you are just beginning to understand it. For example, someone could listen to a speech condemning NFL players who take a knee when the National Anthem is played before a football game. Without looking into why players are doing so, or understanding what exactly it is that they are protesting, an epistemically humble person will not conclude that such players are bad people who ought to be fired, and so on. Before coming to any other conclusions, she will find out why players are kneeling. In providing an account of intellectual humility, a closely related concept, Dennis Whitcomb, Heather Battaly, Jason Baehr, and Daniel Howard-Snyder describe intellectual limitations by citing the following paradigm examples: “gaps in knowledge (e.g. ignorance of current affairs), cognitive mistakes (e.g. forgetting an appointment), unreliable processes (e.g. bad vision or memory), deficits in learnable skills (e.g. being bad at math), intellectual character flaws (e.g. a tendency to draw hasty inferences), and much more besides” (Whitcomb et al. 2015, 517). A person with intellectual humility, according to Whitcomb, Battaly, Baehr, and HowardSnyder, is a person who takes an appropriate stance toward his or her limitations. She strikes the mean between arrogance and servility:

134

S. Ryan

On our view, humility partly consists in a disposition to be aware (even if just implicitly) of one’s limitations, for them to come to mind when the occasion calls for it. In this connection, notice that the paradigmatically arrogant person is often oblivious to his limitations; they don’t show up on his radar. The paradigmatically servile person, however, hardly sees anything else; his radar is perpetually peppered with his limitations. On our view, humility lies in the mean between these extremes. When life calls for one to be mindful of a limitation, then, and only then, will it appear on the ideally humble person’s radar. And what goes for humility in general goes for [intellectual humility] in particular. (Whitcomb et al. 2015, 517–518)

My view is a departure from their view because I believe epistemic humility is something that the quest for knowledge always demands. Epistemic humility, on my view, is an ideal virtue. An ideally epistemically humble person is always mindful of their epistemic limitations. As I understand it, epistemic humility does not fall on a scale between servility and arrogance or obliviousness and obsession. A servile person concedes authority to others excessively and never trusts herself. An arrogant person rarely concedes to others and is excessively confident. On my view, an epistemically humble person might concede to others as infrequently as an arrogant person or as frequently as a servile person. The way one ranks oneself in comparison to others is not central on my view. On my view, what matters is how aware, honest, and responsive one is to one’s evidential situation. The testimony of others can count as evidence, so it is not as if an epistemically humble person is indifferent to the epistemic strengths or status of others. But how one rates their epistemic status relative to others is not where epistemic humility is determined on my view. Being aware of and responding appropriately to the quality and quantity of one’s evidence is what matters. On my view, our epistemic limitations should always be on our radars. That is, it is always appropriate to be aware of our epistemic limitations. Our awareness of our epistemic limitations should not make us servile, but should make us pursue the truth with honesty and appropriate caution. I cite Whitcomb, Battaly, Baehr, and Howard-Snyder here because I think that despite our differences, their view is interesting, carefully developed, and helpful in explaining my own, similar view. Peter Klein also believes that epistemic humility is a virtue. In his “The Virtues of Inconsistency,” Klein discusses the preface paradox and reveals his endorsement for epistemic humility (1985). Epistemic humility is, after all, what leads to the preface paradox. The preface paradox is a puzzle that arises when you have good reason to think you’ve made a mistake—in this case, it is somewhere in your book, for which you are writing the preface. In the preface of your book, you include the admission that “No doubt I’ve made mistakes in this book, and I take full responsibility for them.” You haven’t identified the mistakes (otherwise you would remove them before publication), and yet you are sure you have made some. The belief that you have no doubt made a mistake is justified according to Klein. This belief is justified because you know that you are fallible and there are a lot of sentences in your book that took years to carefully research and write. Moreover, some of the conclusions you draw in your book rest on very complex ideas and logical inferences. However, since you’ve carefully researched your book and you have strong evidence

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

135

supporting each sentence, you are allegedly justified in believing each sentence of your book is true as you read the book slowly one last time before sending it to the publisher. The problem is that if you are justified in believing each sentence is true, you seem to have as much, or even more evidence, considering the coherence of the whole book, for believing ‘all the sentences in my book are true.’ And that is a problem if we accept that your appreciation of your own fallibility justifies you in believing ‘it is not the case that all the sentences in my book are true.’ Klein’s solution to this paradox (and the lottery paradox, a similar puzzle about the possibility of justified inconsistent beliefs discussed later in this chapter) is to deny the conjunction principle. That is, he denies that being justified in believing each of the individual sentences is adequate justification for believing the conjunction ‘all the sentences in the book are true.’ Acknowledgment of your fallibility serves as a defeater of that conjunction on Klein’s view, but acknowledgement of your fallibility, under the circumstances, does not defeat any one of the individual conjuncts. Thus he avoids the conclusion that one can be justified in believing an obvious contradiction. Yet, as the title of his paper reveals, he believes that one can be epistemically justified in believing claims that are obviously inconsistent. Klein endorses the view that an awareness of your epistemic limitations can downgrade, or defeat, the epistemic status of some of your beliefs.

9.5

Defeasibility and the Analysis of Knowledge

Following the publication of Edmund Gettier’s “Is Justified True Belief Knowledge?” (1963), Peter Klein, among others, defended versions of a defeasibility analysis of knowledge. The theory has evolved over the years, and although there is still no consensus on any one theory of knowledge—on whether defeasibility is the right approach to an analysis of knowledge; on what it takes for one proposition or belief to defeat another; or even on the hope that knowledge can be analyzed at all—the idea that knowledge is undefeated, justified, true, belief remains a serious proposal worth developing. We can state the general idea of the defeasibility theory of knowledge as follows: (DK) S knows p if and only if p is a justified, true, and undefeated belief for S.

Defeasibility theorists face several challenges. A major challenge is to provide precise conditions under which a justified, true belief is, or is not, defeated. Taking a quick look at one of Gettier’s examples will help to explain the basic theory, the concept of a defeater, and the worry about how to understand what counts as a genuine defeater. In one Gettier case, we have Smith believing ( p): ‘Jones owns a Ford’. He believes p on the basis of Jones’s testimony, a certificate that lists Jones as the owner of the Ford, etc., all of which together make Smith’s belief justified. Although he has no idea of the whereabouts of his friend Brown, Smith, a real logic whiz, believes the disjunctive claim, (q): ‘Either Jones owns a Ford or Brown is in Barcelona’ because his belief in the first disjunct is so strongly justified by his

136

S. Ryan

evidence, and because a disjunction is true if at least one disjunct is true. Strangely, Brown just so happens to actually be in Barcelona, but Jones does not actually own the Ford he claims to own. Since Smith has no idea that Brown is in Barcelona, he is right about q ‘by accident.’ His belief is ‘right for the wrong reason.’ Thus, although q is a justified, true belief for Smith, he does not have knowledge. Defeasibility theories were introduced to explain the difference between a genuine instance of knowledge and a justified, true belief that falls short. The idea is that when a person knows p, there is no defeater lurking outside of the believer’s set of evidence that prevents the belief from being knowledge despite being true and adequately justified. What has proven difficult is coming up with an adequate way to understand what is a genuine defeater. A very simple answer, but ultimately inadequate for reasons I will explain shortly, is the following: A defeater is any true claim d, such that S is not aware of d and if S were made aware of d, she would no longer be justified in believing p.

It is important to note that these kinds of defeaters do not actually undermine the believer’s evidence, because the believer is not aware of them. They are factual defeaters, external to the set of evidence possessed by the believer. The defeasibility approach to analyzing knowledge grants Gettier that justified, true belief is not knowledge. The view adds a counterfactual fourth condition—the defeater condition—to the traditional analysis of knowledge that distinguishes genuine cases of knowledge from the justified, true beliefs that fall short of knowledge. In the Gettier case just described, there is a defeater, namely, the fact that (d ): Jones does not own a Ford. If Smith were made aware of this fact, he would no longer be justified in believing q. This explains, allegedly, why in the original scenario Gettier imagines, Smith is actually justified in believing a true claim that he nevertheless does not know. As we know from all of the vast and fascinating post-Gettier literature, the view needs some refinement. The proposed account of a defeater is not right because it is not sensitive enough to separate genuine defeaters from defeaters that are misleading (or defeated defeaters). Lehrer and Paxon’s famous Tom Grabit example (1969), taken up by Klein and others, led to questions about what it takes for something to be a legitimate defeater as opposed to a misleading defeater. The original Tom Grabit example goes something like this. Imagine that you are working in a very small library and you are very familiar with all the students who study there every day, including one student named Tom Grabit. You carefully observe a student who looks exactly like Tom stealing a copy of Peter Klein’s Certainty: A Refutation of Scepticism. You get a very good look at the chap and conclude that ( p): ‘Tom Grabit stole the book.’ Given your detailed familiarity with Tom and your eagle-eyed observation of the thief, your belief is extremely well justified. While you are busy writing up your detailed police report, unbeknownst to you, one of Tom’s parents is on the phone notifying the lead investigator that Tom is out of the country, but that his kleptomaniac twin brother, John, who is also a student at the university, was in the library at the time of the theft. If you were made aware of the fact that (d ): ‘Tom’s parent said that Tom is out of the country but his brother

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

137

John, who is a kleptomaniac, was pilfering books at your library today’, you’d save your time and stop writing out the report. You’d back off and no longer believe p. Although you don’t realize it because you have no clue about the testimony of the parent, your justified, true belief has been defeated by d if a defeater is to be understood as a true claim that would undermine your justification for believing p if you were made aware of it. But this case has a twist that makes it a counterexample for a defeasibility theory given the proposed suggestion about what counts as a defeater. Here is the twist. Your observation is correct. Tom really did steal the book. There is no twin brother, John, although the parent did say that Tom has an identical twin etc., making d, which is a statement about the parent, true. But, the parent is demented (or a pathological liar). Thus, it seems you really do know that Tom Grabit stole the book even though d is true, and if you were made aware of d, you would no longer be justified in believing p. The so-called defeater is misleading, or defeated. Thus, we need a better way to specify what counts as a genuine defeater. Moreover, a justified, true belief can be defeated as knowledge even if the relevant fact that is external to the believer’s body of evidence would not, if acquired by the believer, undermine their justification. For example, consider a modified version of the Tom Grabit case. In this modified version, there is no demented or lying parent. You know Tom very well and he spends a lot of time at the library. Suppose Tom did steal the book and you saw him steal the book. Suppose also that although you don’t know this fact, Tom really does have an identical twin brother who was in town on the day of the theft. Let (d ) be: ‘Tom has an identical twin brother who is in town on the day of the theft.’ What if you were made aware of d? With no additional reason to think the twin is in the library or that he looks exactly like Tom, and with lots of reason to believe Tom is in the library as usual, we’d say that you still have excellent reason to believe p: Tom Grabit stole the book. If you were made aware of d, your justification would not be as strong in the counterfactual scenario as in the actual scenario, but your belief would still be highly justified. Given how well you know Tom and how often he is in the library, your belief that Tom stole the book would remain justified. In order for your justification to be undermined, you’d need some reason to think Tom’s twin was in the library at the time of the theft, that he looks and dresses just like Tom, etc. Without that additional evidence, you should continue believing Tom stole the book. However, with or without information about the whereabouts and appearance of Tom’s twin, d is a type of defeater that prevents you from having knowledge. It downgrades you from a knower without downgrading you from having a very well justified, true belief. You have a justified, true belief that is not knowledge. Your belief that Tom stole the book is a justified, true, but defeated belief. The famous barn-façade example could be understood as another such case of a justified, true belief that is not knowledge where the relevant defeater would not undermine the believer’s justification. To elaborate, in the barn-façade case, we have Lucy, a world-renowned barnologist, driving through the countryside. Lucy is on a research trip counting the number of barns within a two-mile radius of her university, and she counts 100. In fact, despite her incredible knowledge and skills, 99 of her

138

S. Ryan

identifications are incorrect. She has observed 99 incredibly realistic-looking fake barns. Only one of them is a real barn. Yet in all 100 cases, her evidence strongly supports her belief ‘There is a barn.’ In the one case where her belief is true, most people have the intuition that Lucy does not know ‘There is a barn’ even though she has a justified, true belief. A defeasibility analysis of knowledge with an adequate conception of defeaters could account for this by noting a defeater along the lines of ‘There is exactly one real barn in the two-mile radius, and the other 99 barn-looking objects are fakes.’ But notice that while this defeater undermines her claim to knowledge, it doesn’t undermine (or undermine significantly) her evidence. It is still true that each barn-looking object looks exactly like a real barn. If so, and if an evidentialist theory of justification is correct, then Lucy’s belief remains justified in the face of the defeater. One might object to this description of what we should conclude from the barnfaçade case and insist that the defeater works just as the defeaters work in Gettier’s cases. That is, Lucy’s true belief would no longer be justified if she learned about the defeater. Or one might dismiss the barn façade case as a counterexample to the traditional analysis of knowledge. One might think that Lucy does know ‘There is a barn’ in the one case in which she is really looking at an actual barn. My argument for moderate skepticism does not depend on my interpretation of the barn-façade case. It is simply an illustrative example. The preface paradox (discussed earlier) and the lottery paradox make use of a similar kind of defeater. The difference is that in the lottery paradox and the preface paradox, the defeater is a crucial part of the believer’s actual body of evidence—not a fact external to her body of evidence. In the lottery paradox, we consider the epistemic situation of a person in a fair, one-million-ticket lottery that has exactly one winning ticket. She considers her ticket and the odds of winning, and justifiably believes that her ticket is not the winning ticket. But since she knows that one ticket will definitely win, she does not know that her ticket is a loser. She has a knowledge defeater but not a justificatory defeater. It is worth noting that epistemologists disagree about what are the correct solutions to the lottery paradox and the preface paradox. An alternative position to the one I’ve stated here is to claim that all but one believer does know that her ticket is a loser when she purchases her ticket. Another position is to deny that a person in the lottery scenario described above is justified in believing her ticket is a loser despite the overwhelming odds that her ticket is a loser.4 I won’t defend the position that she does not know her ticket is a loser but is justified in believing her ticket is a loser. I will simply note that if it is the correct solution, then a defeasibility analysis of knowledge needs to do a better job of articulating the concept of a defeater. Before turning away from the lottery, we should take note that the case is not a typical Gettier example where a person reasons through a false premise or winds up with a justified, true belief that is right for the wrong reasons. Nevertheless, let me note again that if the lottery presents us with a

4

I defended the view that the ticket holder is not justified in believing her ticket is a loser in Ryan (1996). Since publishing that paper, I’ve changed my view.

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

139

case of a justified, true belief that is not knowledge, a defeasibility analysis of knowledge needs a more nuanced characterization of defeat. Thus, for the many reasons just mentioned, an adequate defeasibility theory of knowledge needs some fine-tuning. The task does not seem hopeless, and I believe Klein and others think the defeasibility theory is basically correct—we just need to do a better job in working out the details. Assuming that the theory is on the right track, we can now work up the argument for moderate skepticism.

9.6

Epistemic Limitations and Defeat

Defeaters, whether one is aware of them or not, are epistemic limitations. Moreover, epistemic limitations, whether one is aware of them or not, can be defeaters. We can understand the lottery ticket holder as someone who suffers from an epistemic limitation. Her evidence, no matter how strong, is unable to give her knowledge. Her epistemic situation leaves her vulnerable and prevents her from knowing. In my revised Tom Grabit case, the librarian suffers from an epistemic limitation. The defeater leaves her with an epistemic liability that prevents her from having knowledge. In the original Gettier case, Smith has an epistemic limitation in that he is unaware of a devastating fact, namely, Jones does not own a Ford. In the barn case, the lottery, and the preface limitations are defeaters. Thinking beyond contrived puzzles like the lottery and the examples that cropped up in the Gettier literature, I think we should acknowledge that nearly all of us, with respect to many beliefs we have, suffer from epistemic limitations that prevent us from knowing. In the next section of this chapter, I will argue that many of our beliefs about complex issues are defeated. In some cases, our limitations prevent us from having justified beliefs. In others, the limitations, while allowing for our beliefs to remain justified, prevent us from knowing. Our obligation to be epistemically humble requires us to pay attention to those limitations, and that, I will contend, will lead to a skeptical outlook and all that it involves. An epistemically humble person will not pretend to have knowledge or downplay her ignorance. For example, if she is aware that she has not sufficiently investigated an issue, she will not assert that she knows the truth about that issue or make assertions that will be interpreted by others to be credible claims to knowledge. With respect to questions on which there is serious research, complex bodies of evidence, deep disagreements, controversy, and the like, there are defeaters. When we are in such circumstances, the fact that we suffer from so many epistemic limitations either defeats the justification of our beliefs or else prevents our justified beliefs from being knowledge. Thus, I will argue that we do not have knowledge about many of the most interesting propositions in these areas of inquiry. The humble are aware of this problem. The rest lack knowledge—they just don’t realize it.

140

9.7

S. Ryan

An Argument for Moderate Skepticism

It is time to put all of this together and present my argument for moderate skepticism. For the sake of simplicity, and at the cost of clarity, I will use ‘interesting topics’ to include those that intellectually curious and educated people wonder about, debate, research, and study. I acknowledge that this characterization is rough. The argument I will present, if successful, will show that knowledge is much rarer than we normally think. If the argument is successful, it shows that the subjects people attend universities to gain knowledge about are subjects in which many of the most interesting claims are not known to be true. If successful, the argument will show that wise people do not know the answers to the questions we consult them about. This moderate form of skepticism is important because many human beings spend most of their doxastic energy wondering about topics on which they will not obtain knowledge. I don’t wonder if the coffee cup I’m staring at is real or if the liquid inside is really coffee. I just sip the coffee while I wonder about a reporter’s assertion that America is a deeply divided nation. What does that mean? Is she claiming that Americans are divided on fundamental principles? Is that true? How could we know if it is true? Even if social scientists could pull off a great study, will I pore over the research and understand the truth myself? I wonder if the frustration, violence, and hate we see in America is based on disagreements about fundamental principles or if it is based on conflicting views about how to apply fundamental principles that are widely shared. I wonder if what appears to be a division of principles or applications of principles is actually, in many instances, simply non-rational, habitual allegiances rearing their ugly heads. If there were serious studies figuring out the answers to these questions; if I read a lot more and a lot more carefully; if I listened more openly and carefully to a much wider variety of people; if I were a lot smarter; if my experience of humanity were much deeper and more thorough than it is right now; I might obtain a better understanding. But as it is, and as I am, I am extremely ignorant about these issues. I submit that most, if not all, of my fellow humans are equally ignorant. Our epistemic limitations, whether we note them or not, undermine our ability to know the answers to the interesting questions we actually ponder. The topics are not relegated to the ivory tower and dusty libraries. Moderate skepticism shows up on your wedding day when you are thinking about your faithful, fulfilling, lifetime of love and kindness. Our epistemic limitations undermine knowledge when we wonder how we should invest our money in preparation for retirement. We are in such circumstances when we wonder whether our favorite job candidate will really turn out to be as great as we hope. We are in such circumstances when we are wondering if, given our actual driving practices, buying a hybrid car is the best environmental option for us. We are in such circumstances when we are non-experts trying to decide which of several alternative medical treatments will be best for us. We are in such circumstances when we form beliefs about a subject based on a small sample of a vast body of literature.

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

141

If moderate skepticism is true, nobody knows whether or not a god exists. Nobody knows what is the best ethical theory. Nobody knows what justice is. Claims within entire fields of inquiry—such as philosophy—are unknown. Here is a statement of the argument: Epistemic Limitations Argument for Moderate Skepticism (1) Our epistemic limitations defeat our most interesting and contentious beliefs on interesting subjects. (2) If (1), then human beings know far less than is normally credited to them. ∴ Human beings know far less than is normally credited to them. PREMISE (1) is supported by taking stock of the many limitations we all suffer from when attempting to get at the truth in any area of inquiry that requires advanced cognitive skills and a body of evidence that is vast, loaded with complexity, and steeped in disagreement. All the subjects studied at the university level, for example, are subjects for which all of us, even the experts in those fields, have epistemic limitations. The most obvious limitation for most of us is that we are not experts. For many questions, and for many of us, our evidence is so sketchy and weak that belief isn’t even the correct justified doxastic attitude to hold toward the claims we consider. Other times, we are in the more fortunate situation of understanding, at best, a very small sample of a vast amount of available evidence. Those who have quite a bit of evidence, and qualify as experts, often have other experts who disagree with them. Thus, even if our beliefs are justified by the evidence we do have, we very frequently have one or more epistemic limitations that serve as defeaters that prevent us from knowing. Here are a few such defeaters, any one of which is typically enough to prohibit one from knowing: (D1): There is a vast amount of important evidence that you do not possess on the subject matter. (D2): You have epistemic peers who disagree with you. (D3): The experts on this subject matter disagree with one another. (D4): Your evidence is very sketchy. Limiting ourselves to just these four very common defeaters, we can see that even if the justification condition is satisfied for our beliefs, there are factual defeaters that prevent us from knowing. Thinking about the epistemology of disagreement debate is helpful. I am sympathetic to the view that when an epistemic peer disagrees with you this significantly weakens your justification. Both parties in a peer disagreement ought to back off. But even those who think there is rational peer disagreement and that disagreement alone gives neither side a reason to budge should be sympathetic to the view that until you can see the error in your peer’s thinking or come to downgrade them as an epistemic peer, you do not have knowledge. Because (D1), (D2), (D3), and/or (D4) are true on a regular basis, many of our most controversial beliefs on interesting subjects are regularly defeated. It is important to note that PREMISE (1) is based on epistemic limitations and the response that the virtue of epistemic humility demands. It is not based on the old and

142

S. Ryan

worn-out claim that we are fallible or that our evidence falls short of complete certainty. As I noted early on in this chapter, I agree that we have knowledge despite our fallibility and uncertainty.5 My belief that I have hands is both fallible and uncertain. Epistemic limitations pose a threat that goes far beyond mere fallibility and uncertainty. PREMISE (2) assumes that we need to have undefeated, justified, true beliefs in order to have knowledge. But when it comes to issues steeped in disagreement, issues for which there are complex and fragmentary bodies of evidence, and issues on which one must have more sophisticated cognitive skills than one has, our beliefs are regularly unjustified or defeated. Yet it is on precisely these issues that educated people often proclaim to have knowledge and use that so-called knowledge to influence others, devise a curriculum, or defend public policies. One might object to PREMISE (2) by claiming that we don’t normally credit people with knowing the complex and interesting propositions ruled out by moderate skepticism. A lot hinges on how ‘we’ and ‘normally’ are interpreted. I will avoid this objection by noting that plenty of people claim to have knowledge on the sort of propositions this argument rules out. If the argument is correct, and I believe that it is at least worthy of serious consideration, there are a lot of knowledge proclamations that are undermined. Perhaps many philosophers will accept this argument. If so, it opens up an opportunity to explain to non-philosophers why skepticism is worth thinking about as much more than a philosophical exercise. Skepticism is worth embracing as a way of life.

9.8

Epistemic Humility and the Implications of Moderate Skepticism

I think this argument for moderate skepticism is strong. I also think Peter Klein ought to embrace it since each premise is supported by some of Klein’s philosophical commitments over the years. Klein’s refutation of direct skepticism does no damage to moderate skepticism. Unlike our belief that we have hands, we don’t typically have enormous amounts of high-quality evidence supporting our controversial beliefs on complex and interesting topics. While we have no reason at all to think we are mere brains in vats, we always have very good reason to think we are epistemically limited when we ponder the tricky questions of philosophy, history, psychology, investment strategies, relationships, and the like. Moreover, the defeaters I pose in support of moderate skepticism are logically consistent with the beliefs they challenge. For example, my belief that ‘Compatibilism is the correct stance to take on the free will problem’ is logically consistent with the facts

5

Klein disagrees. A major point of Klein’s (1981) book is to show that knowledge does require certainty, and we possess certainty when we look at our hands and form the belief that we have hands.

9 Epistemic Humility, Defeat, and a Defense of Moderate Skepticism

143

describing all my limitations on this topic. Thus, a simple closure principle is useless against moderate skepticism. Moreover, we don’t have to be aware of defeaters, or actually believe them, for them to effectively prevent us from having knowledge. Moderate skepticism allows for a lot of knowledge about the external world. But the complex issues that we debate and really wonder about are issues on which we have limited knowledge according moderate skepticism. Since epistemic humility is a virtue, we all should respond appropriately to this argument. It does not show that we know nothing about interesting subjects. We might know a lot of simple and uncontroversial facts within a complex field. However, if the argument for moderate skepticism is correct, we don’t know any of the most controversial and complex claims within a field. An acceptance of moderate skepticism should impact the ways we communicate, teach, learn, and engage in civil discourse. Even though a defender of moderate skepticism can acknowledge that some people are epistemically better off than others, a moderate skeptic insists that everyone is limited and has much to learn. If understanding reality matters to us, moderate skepticism and epistemic humility should motivate us to devote more time and resources to education. Education should focus on identifying and addressing our epistemic limitations. It should focus on cultivating a culture of inquiry and epistemic humility. It should focus on learning how to ask good questions, on developing comfort with not knowing, and on developing a thirst for wisdom. Finally, our attempts to understand wisdom need to be updated. Theories that require the wise to have knowledge about the most important questions in life need to be abandoned once and for all.6

References Feldman, R., & Conee, E. (1985). Evidentialism. Philosophical Studies, 48(1), 15–34. Gettier, E. (1963). Is justified true belief knowledge? Analysis, 23(6), 121–123. Klein, P. (1981). Certainty: A refutation of scepticism. Minneapolis: University of Minnesota Press. Klein, P. (1985). The virtues of inconsistency. The Monist, 68(1), 105–135. Markosian, N. (2014). Do you know that you are not a brain in a vat? Logos and Episteme, 5(2), 161–181. Ryan, S. (1996). The epistemic virtues of consistency. Synthese, 109(2), 121–141. Ryan, S. (2015). In defense of moral evidentialism. Logos and Episteme, 6(4), 405–427. Whitcomb, D., Battaly, H., Baehr, J., & Howard-Snyder, D. (2015). Intellectual humility: Owning our limitations. Philosophy and Phenomenological Research, 94(3), 509–539.

6

I’d like to thank Cherie Braden for her challenging, insightful, thorough, and extremely helpful comments on this chapter. No doubt this chapter still contains many mistakes for which I take full responsibility.

Chapter 10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination Claudio de Almeida

Abstract The effort to understand Peter D. Klein’s work on (so-called) Cartesian skepticism is simply not optional to anyone wishing to become familiar with stateof-the-art scholarship on the problem. Nearly four decades ago, Klein developed the invariantist pro-closure response to Fred Dretske’s counterexamples to epistemic closure principles, thus stealing some of the thunder from the nascent, Dretskeinspired contextualist views in epistemology. Since then, he’s added important theoretical elements to his analysis of the central problem of our debate about skepticism and closure: the problem of how a closure defender might refute the Cartesian skeptic without turning into a purveyor of ‘easy knowledge.’ According to Klein, “a proper understanding of closure shows that the [Cartesian] skeptic cannot provide a good argument for her view and helps to show that there is no genuine problem of easy knowledge” (see “Closure Matters: Academic Skepticism and Easy Knowledge” 2004). Indeed, inattention to Klein’s work accounts for some of the most important fallacies perpetrated in the recent literature on epistemic principles that may be of use to the skeptic. Still, on close inspection, Klein’s case against Cartesian skepticism proves optimistic. The chapter briefly reviews the main features of 34 years of Klein’s work on the problem in 15 publications, from his 1981 book on Certainty to his 2015 Stanford Encyclopedia of Philosophy entry “Skepticism”. It finds that his case against the skeptic is a distinctive mixture of Mooreanism and Russellianism and concludes that none of his main objections to Cartesian skepticism is tenable. Keywords Pyrrhonian skepticism · Cartesian skepticism · Evidential underdetermination · Epistemic closure · Klein · Brueckner · Skeptical argument · Mooreanism · Epistemology

C. de Almeida (*) Pontifical Catholic University of Rio Grande do Sul, Porto Alegre, Brazil © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_10

145

146

10.1

C. de Almeida

Introduction: To Catch an Anti-skeptical Tidal Wave

Peter D. Klein’s work on (so-called) Cartesian skepticism is as important as it is intricate. It has more moving parts than the more popular analyses of the problem, in contrast with which it is offered as the superior alternative. But the effort to understand it is simply not optional to anyone wishing to become familiar with state-of-the-art scholarship on the problem. Thirty-seven years ago, Klein (1981) developed the invariantist response to Fred Dretske’s (1970) counterexamples to epistemic closure principles, thus stealing some of the thunder from the nascent, Dretske-inspired contextualist views in epistemology. Since then, he’s added important theoretical elements to his analysis of the central problem of our debate about skepticism and closure: the problem of how a closure defender might refute the Cartesian skeptic without turning into a purveyor of what Stewart Cohen (2002, 2004) called ‘easy knowledge’, the fallaciously derived claim that one knows that a given skeptical hypothesis is false, or, in Klein’s hands, the problem of exposing the fallacy in the skeptical reasoning. According to Klein (2004c, 182), “a proper understanding of closure shows that the [A]cademic [or ‘Cartesian’] skeptic cannot provide a good argument for her view and helps to show that there is no genuine problem of easy knowledge.” There is, indeed, much to be learned from Klein’s argument for this conclusion. It is by now well-known that inattention to his work accounts for some of the most important fallacies perpetrated in the literature on epistemic principles that may be of use to the skeptic. Still, on close inspection, Klein’s case against Cartesian skepticism proves optimistic. As we shall see, in addition to seemingly insurmountable problems on the closure front, Klein’s work can be challenged on another, decisive front. My main task in this chapter is to examine a key element of his anti-skeptical views, an argument against what has become known as the ‘Underdetermination Principle’. The problematic argument originated in his 1981 book on Certainty and has figured prominently in most of his writings on the topic since then, including the 2015 version of his entry on “Skepticism” in the Stanford Encyclopedia of Philosophy. Klein’s case against that key principle is, inter alia, challenged here. But, before we get to Klein’s case against that skeptical principle, I’ll have a number of warnings to take notice of what seem to me to be some very interesting mistakes in Klein’s work on Cartesian skepticism. Some features of Klein’s work on the issue that I cannot discuss here at any length are important to the literature and are distinctively his own. I should just like to mention three of those distinctive features in passing. One of them is Klein’s fondness for Pyrrhonism. Stewart Cohen once remarked that we would naturally regard belief in a skeptical hypothesis as “crazy” (Cohen 1988, 112). Klein responded by exempting Pyrrhonism from any ‘craziness’ charge. For Klein, the Pyrrhonian claim that “reasoning cannot settle matters” (Klein 2003a, 80) is interpreted in a way that leads to an interesting form of epistemological infinitism: “I [...] argue that the Pyrrhonians did not push their insights quite far enough,” he

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

147

notes (2000b, 8).1 And he further claims that his Pyrrhonian form of infinitism holds an important lesson for those who are enamored of Cartesian skepticism (2003a). Be that as it may, I’ll have nothing to say about it, partly because I can’t see that the lesson Klein draws from his response to Agrippa’s Trilemma applies in any special way to any important argument for Cartesian skepticism. Nor will I offer a discussion of the related Kleinian claim that, when compared to the Cartesian skeptic, the Pyrrhonian skeptic holds the more defensible view, namely, the view that we should suspend judgment with regard to whether there can be empirical knowledge. I won’t go into that either, partly because it is not clear to me that there can be anything deserving to be called a stable ‘Pyrrhonian view’ that stops at justified suspension of judgment and does not ultimately collapse into the Academic (Cartesian) claim that there cannot, after all, be any such knowledge.2 The second distinctively Kleinian claim that I cannot extensively discuss here is the one according to which, if the validity of a skeptical modus tollens (which characterizes the skeptical argument of Sect. 10.2 below) were put into question, we would be led down a path where “reasoning would apparently come to a complete standstill” (Klein 2015, § 4). To my knowledge, while we all have been aware that his target here is, indeed, very unpopular, nobody else has put the point in such dramatic terms. According to Klein, If one were to deny that modus tollens is a valid form of inference, one would also have to deny the validity of (i) disjunctive syllogism and (ii) modus ponens or contraposition, since it is easy to transform modus tollens arguments into ones employing the other forms of inference. Hence, if this alternative [response to skepticism] were chosen, reasoning would apparently come to a complete standstill. That, presumably, is why no one has ever seriously considered this alternative. (Klein 2015, § 4)

But denying that either modus tollens, or modus ponens, or Contraposition, or Disjunctive Syllogism is a valid rule of inference is not the extravagant move that Klein would have us believe it is. Ernest Adams (1988), Jonathan Bennett (2003), Edwin Mares (2004), Vann McGee (1985), Stephen Read (1988), David Sanford (2003), Robert Stalnaker (1975), among many others (including the whole project we call ‘relevance logic’) all claim that the unrestricted application of those classical rules (notwithstanding disagreements about such restrictions) is objectionable and then try to devise a set of reasonable restrictions.3 Thus it is far from clear that anyone intending to relate those ‘relevance-friendly’ efforts with the analysis of skeptical arguments would be engaged in a hopeless project.

1

Incidentally, I am not confident that there is good history of Pyrrhonism underwriting Klein’s decision to label his brand of infinitism a ‘new Pyrrhonism’. 2 While I admit that I find Klein’s case for infinitism very attractive, I also see room for disagreement as regards some of the interesting details, such as Klein’s distinction between belief and assent (Klein 2000b), but this distinction is not a core element of the view. 3 There are also those who think that some of those objections to classical rules are untenable but deserve to be discussed, e.g., Walter Sinnott-Armstrong et al. (1990), Frank Jackson (1987), John Burgess (2005), and Roy Sorensen (1988), plus Sorensen supporter John Hawthorne (2004), among others.

148

C. de Almeida

While that issue regarding validity should not, in full generality, be pursued here, since it leads us to problems that Klein did not tackle in his work on skepticism, it will briefly resurface at a crucial juncture below.4 The third distinctive aspect of Klein’s work that I will steer clear of in what follows concerns his analysis of the differences between an exegetically defensible reconstruction of the Cartesian argument in the first Meditation and the argument from Cartesian skepticism as it is most often discussed in the contemporary literature on the issue, focusing on epistemic closure principles. Here, Klein’s work—as represented in § 3 of his Stanford Encyclopedia entry and in Klein (1987)—is, I think, a case of unalloyed success. As interesting as all those three distinctive features of Klein’s work on skepticism are, they are not among the most important aspects of his work on Cartesian skepticism specifically. It is to those more important aspects that I turn in what follows. My last introductory remark concerns terminology. Klein is in the habit of using the terms ‘Academic skepticism’ and ‘Cartesian skepticism’ interchangeably. To better integrate excerpts from his work with my commentary, and to be parsimonious, I defer to his terminology.5

10.2

Academic Skepticism: Finding a Baseline

There is much in Klein’s work on Academic skepticism with which I must disagree. So, there is no avoiding the impression that this short review will read like a laundry list of complaints, with much of what I like in his work on the topic being pushed to the background. That is the nature of this task, I’m afraid. And, in order to make it easier for the reader to understand why I must so vigorously disagree with Klein here, I should begin by putting forward my own understanding of what I will call ‘the master argument’ for Academic skepticism as discussed in the literature to which Klein’s work on the problem belongs. Even though the terms ‘knowledge’ and ‘certainty’ are so often used in discussions of Academic skepticism, as they were by Descartes himself, Klein has led the way to an understanding of the issue that puts the focus exactly where it belongs: on the notion of epistemic justification (or ‘epistemic entitlement’, often called, after Firth (1978), ‘propositional justification’). There is a fundamental theoretical reason 4

I discuss the issue at length in de Almeida (forthcoming b). But I am not persuaded that it is accurate. In any case, Klein (1995, 566) is certainly right when he suggests that fundamental aspects of the Cartesian problem of the first Meditation are prefigured in ancient writings. “In the Academica (2:26),” Klein notes, “Cicero points out that some people do have identical twins who cannot be distinguished on the basis of perception alone. In addition, he claims that the ‘single case of resemblance’ will ‘have made everything doubtful’.” That’s a powerful reminder that something fundamental in Cartesian skepticism was, after all, prefigured in ancient sources. See also Klein (1987). 5

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

149

for the emphasis on justification. Klein has it nailed down from the outset in his involvement with the issue: a key premise in the skeptic’s master argument is an instance of an epistemic closure principle, and the relevant principle concerns epistemic justification (or ‘epistemic entitlement’), understood as a necessary condition of knowledge. For Klein and Anthony Brueckner (1985, 1994, 2010b), the two authors whose bodies of work have contributed most to our understanding of this form of skepticism, that master argument may schematically be represented as follows:6 (1) If S (our doxastic agent) is justified in believing that p (where ‘p’ ranges over propositions representing empirical states of affairs) at a given time t, then S is justified in believing that ~sk (where ‘sk’ ranges over skeptical hypotheses which are incompatible with p) at t. (2) But S is not justified in believing that ~sk at t. (3) Therefore, S is not justified in believing that p at t.7 This formulation of the skeptic’s master argument makes clear that acceptance of the skeptic’s first premise relies on a closure principle concerning epistemic entitlement, as emphasized by Klein in numerous writings (for instance, in 2003a, 78). The range of what you are, or fail to be, justified in believing includes every proposition representing external-world states of affairs, not just the ones that have actually been or will ever be contents of the beliefs you have had or will ever have in your life. The only relevant closure principle (justification-closure) capturing that notion may be expressed as follows: JC: You are (ultima facie) epistemically entitled to believe (i.e., you have an ultimately non-overridden justification for believing), at a given time t, all the logical consequences of any proposition(s) that you are (ultima facie) epistemically entitled to believe at t.8

Think about it in these terms: Suppose you are now (ultima facie) entitled to believe that p and that p
q. Regardless of whether you actually believe any of those propositions (the conditional or its antecedent), and assuming that we accept that they entail that q, you are (ultima facie) entitled to believe that q.9 Entailment infallibly and invariably transmits (ultima facie) epistemic entitlement (justification)

6 Speaking of the strongest influence on our understanding of Academic skepticism prompts us to acknowledge the third member of a triumvirate, Barry Stroud. I have discussed an aspect of his work on skepticism (in de Almeida 2016) that is also relevant to the Klein reader in a way that I will specify below. 7 See Klein (2003a) and Brueckner (1985), where the argument appears in similar terms. In order to simplify matters a bit and express the argument as a modus tollens, I ignore the quantifiers and the claim that the relevant instance of p logically implies—or ‘entails’ (taking these as synonyms)— ~sk, as is so often done in the relevant literature. 8 Klein does not accept JC in full generality, but the disagreement over his proposed restrictions would be irrelevant to present purposes. I discuss some of the sticky points regarding closure principles in de Almeida (forthcoming b). 9 When Klein uses the term ‘justification’ while discussing closure principles, he is undoubtedly concerned with ultima facie justification.

150

C. de Almeida

between entailing proposition(s) and entailed proposition.10 That’s what PREMISE (1) is supposed to depend on. We then turn to the principle on which PREMISE (2) is supposed to depend. Both Klein and Brueckner agree that, to most of us (myself not included), PREMISE (2) is where the battle against Academic skepticism will be fought.11 Most of us, when looking at the master argument, would expect an explanation of why we cannot be justified in believing the negation of any of the relevant skeptical hypotheses. Here, the desired explanation crucially appeals to what Jonathan Vogel called an ‘underdetermination principle.’12 Following Brueckner (2010b) very closely, we can express that principle as follows: UP: A body of evidence e justifies the belief that p for a given S, at a given time t, only if e justifies S in believing, at t, the negation of any proposition that is incompatible with the proposition that p.13

Given our acceptance of UP, the skeptic may then provide us with the following argument for PREMISE (2).14 The first premise in the sub-argument for PREMISE (2) might look like the following conditional: (1a) If S’s evidence for believing that p at a given time t is not evidence for the belief that ~sk at t, the hypothesis that sk is arbitrarily deemed false (or improperly ‘eliminated’, or improperly disregarded) at t.

PREMISE (1A) is supposed to be sustained by UP. A moment’s reflection should convince you that we tacitly apply UP in everyday judgments about what is good evidence for what. For instance, I believe that I am in Brazil right now. Obviously, my being in Brazil (or my being where Brazil is in the actual world) is incompatible—in this case, nomically incompatible—with my simultaneously being in Singapore (or with my being where Singapore actually is). So, naturally, whatever is regarded as good evidence for believing that I am in Brazil must be regarded as good evidence for believing that I am not in Singapore (regardless of whether I, the agent, entertain any thoughts about such matters). The claim rests on a stable, time-tested

10 But bear in mind that nomic entailment (or ‘nomic implication’) is also represented in examples, since the policy here is simply one of avoiding unnecessary circumlocution. For instance, we’ll let ‘I’m now in Brazil (i.e., where Brazil is in the actual world)’ entail the negation of ‘I’m now in Singapore (i.e., where Singapore is in the actual world)’, since it is contextually understood that we might properly add ‘but not in Brazil’ to the latter sentence. 11 To be more precise, I have suggested that the skeptic can do her damage with principles that are weaker than JC. But I do think JC is false. See de Almeida (2011), Sect. 10.6 below, and de Almeida (forthcoming b). 12 In correspondence just before he passed away, Brueckner told me it was his former Ph.D. supervisee Jonathan Vogel who put the label ‘underdetermination principle’ into circulation. 13 While retaining the essence of Brueckner’s formulation, I mean to improve on it. I regret not having the space to compare the formulations here. Thanks are due to Cherie Braden for interesting objections to my formulation of the principle in correspondence. 14 The following four paragraphs are closely based on a passage in de Almeida (2016).

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

151

intuition concerning the evidential support relation. Its denial is deeply puzzling. Like JC, UP should seem unassailable.15 Given an acceptance of UP, the argument for PREMISE (2) in the skeptical modus tollens might proceed as follows: (2a) S’s evidence for believing that p at t is not evidence for the belief that ~sk at t (because, if S were in an sk-scenario, S would still, but then falsely, believe that p).16 (3a) Therefore, sk is arbitrarily deemed false at t. (4) But, if sk is arbitrarily deemed false at t, S is not justified in believing that ~sk at t. (5) So, S is not justified in believing that ~sk at t.

It should be clear that the skeptical rationale for PREMISE (2A) is not fully displayed in the above schematic presentation of the sub-argument for PREMISE (2). The dialectic employing, for instance, a dream-hypothesis expects it to be obvious to you that, if you were only dreaming that p, you would still, but then falsely, believe that p. To JC and UP, we need only add a misleading-evidence hypothesis, something like the claim that what one experiences in a vivid dream is qualitatively indistinguishable, to the doxastic agent, from what one experiences in one’s waking moments. But there are the other familiar hypotheses concerning misleading evidence, such as one’s being fed misleading evidence in demon worlds, brain-in-vat worlds, spectacularly elaborate conspiracies, etc.17 We may, of course, want to tweak the language of the skeptical sub-argument for PREMISE (2) in a number of ways. But I trust we have seen enough to understand the appeal of the skeptical modus tollens. There is no significant feature of those two arguments, the master argument and its sub-argument, that Klein fails to address. So, in what follows, I’ll take the set of those two arguments as Klein’s target and discuss his objections to them.

10.3

A Question-Begging Skeptic?

One of Klein’s distinctive anti-skeptical claims is that the master argument is epistemically question-begging (1995, 2000a, 2003a, 2004a, b, c, 2010, 2015). According to him, the master argument “adds no credibility to [its] conclusion because all the work is done in the sub-argument for [PREMISE (2)]” (2010, 160). The claim here is that “the conclusion [of the master argument] is reached before [the master argument] begins” (160; emphasis added). More precisely, Klein thinks he can show that, on close inspection, the master argument is such that “either it begs

15

I’d be remiss if I didn’t note that I’m one of those for whom UP is the only unassailable principle of the two. 16 I elaborate on the parenthetical point in what follows. 17 Apparently, one would have to go as far as to deny that humans experience misleading evidence—maybe as far as to hold that the very notion of misleading evidence is philosophical fabrication—in order to stop the skeptical sub-argument.

152

C. de Almeida

the question or contains a false premise” (2000a, 115). That possibly false premise is 18 PREMISE (1), the premise that rests on JC. We can zero in on the essence of his argument as follows. Suppose we like PREMISE (1) of the master argument because we think that a justified belief that p is good evidence for a justified belief that ~sk. (We obviously should expect no less from anyone who finds JC attractive.) And assume that having good evidence for believing a given proposition is the very minimum we require for one to know the fact represented by that proposition. Now, according to Klein, if we assume that PREMISE (1) is true, the task of defending PREMISE (2) is the task of showing that there is no good evidence for believing that ~sk. In order to accomplish that goal, the sub-argument must show that we are not justified in believing that p, i.e., the argument must show that the antecedent of PREMISE (1) is false. But that is the conclusion of the master argument. Ergo, either we have no good reason to believe PREMISE (2) in the master argument or the master argument is epistemically questionbegging, since the case for one of its premises evidentially depends on its conclusion. Since Klein would be ready to grant that there is a compelling argument for PREMISE (2), he feels entitled to conclude that the master argument is questionbegging. Here’s an excerpt where the argument is put forward: If the [...] skeptic thinks that closure holds in this case because one must arrive on the inference path at the proposition that there is a table [p] before one arrives at the denial of the skeptical hypothesis [sk], then in arguing for the second premise the skeptic must show that we cannot arrive at the proposition that there is a table, because if we did, we could get to the denial of the skeptical hypothesis (since that is what this instantiation of closure maintains). But if the sub-argument for premise 2 shows that we can’t arrive at the proposition that there is a table, then that sub-argument already establishes the conclusion. (Klein 2003a, 90)

Notice: Klein very clearly claims that the master argument is epistemically questionbegging because “all the work is done in the sub-argument for [PREMISE (2)]” (2010, 160), that is, “the argument for premise 2 alone establishes the conclusion” (2003a, 90, emphasis added). There are at least three problems with Klein’s case for epistemic circularity against the skeptic. Two of the problems have been identified by Brueckner (1985, 2000, 2005, 2010a). First, as Brueckner (1985, 303) noticed when commenting on an early version of Klein’s argument (in Klein 1981), the argument confuses inferential justification with justified reasoning (though Brueckner doesn’t put it in these terms). According to the objection, Klein mistakenly sees temporal ordering between premises and conclusions at the level of entitlement transmission, that is, at the level of logical relations (on which epistemic support supervenes) between propositions, whereas any such ordering might only be appropriate at the level of belief-states, as when an episode of reasoning occurs. The point should come across very clearly in Brueckner’s own words, as follows:

Elsewhere, he represents the second horn of the dilemma as the one according to which “the first premise must be unmotivated” (Klein 2010, 160), rather than false. For present purposes, we can safely ignore the discrepancy.

18

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

153

[A]ny argument which shows that [PREMISE (2) is true] will show that [the antecedent of PREMISE (1) is false]. [...] But the foregoing fact, which is guaranteed simply by modus tollens, obviously does not establish that in order to show that [PREMISE (2) is true], the skeptic must first show that [the antecedent of PREMISE (1) is false]. To show that [PREMISE (2) is true] is, by modus tollens, to show that [the antecedent of PREMISE (1) is false] (given [our JC]). But nothing whatsoever follows about the order in which the logically related propositions must be made plausible by the skeptic. (Brueckner 1985, 303)

Second, as Brueckner later claimed (Brueckner 2005, 170–171; 2010a, 372), Klein’s argument applies to any case of modus tollens. On Klein’s understanding of the master argument, modus tollens turns out to be epistemically circular (or ‘question-begging’). To my mind, Brueckner’s point is definitive, but it doesn’t go as far as it could have gone. Wouldn’t modus ponens also turn out epistemically circular according to Klein’s argument? Recall that, in order for Klein to get his conclusion (that the master argument is question-begging), he expects us to assume that the first premise is true (by justification-closure). If you make that assumption, shouldn’t you think, along Kleinian lines, that any argument that justifies you in believing the antecedent of the conditional premise in a case of modus ponens also ensures that a belief in its consequent is justified? As a simple matter of epistemic entitlement, it certainly does. But, for an episode of reasoning to yield justified belief, the reasoner must be epistemically entitled to believe the premises before, as Klein would put it, the reasoner arrives at the conclusion. But, even extending the charge of epistemic circularity to modus ponens stops short of what we may be entitled to conclude from Klein’s argument. Notice the apparent sleight of hand in Klein’s wording of his argument as represented in the above excerpt from Klein (2003a, 90)—the very same move occurring in each of the many versions of his argument19: Although he wants you to believe that the problem with the master argument is that “the argument for premise 2 alone establishes the conclusion” (90; emphasis added), he’ll clearly have no case unless you also assume that PREMISE (1) is true. Here’s another version of the same move made just before he claims that “all the work is done in the sub-argument for [PREMISE (2)]”: “Suppose the skeptic says that what makes premise [(1)] true with regard to the relevant pair of propositions is that ‘I have a hand’ [our p] is in the evidential ancestry of ‘I am not in a switched world’ [our ~sk] and is sufficient to justify it (2010, 160).” But, if you must assume the conditional premise in either a modus tollens or a modus ponens, in addition to believing the other premise, in order to derive a charge of epistemic circularity against an application of those inference rules, you must be fast on your way to making (mutatis mutandis) the same charge against any valid rule! For instance, if you assume the disjunctive premise in a case of Disjunctive Syllogism, won’t any case for the negation of one of the disjuncts, together with your disjunctive assumption, make you epistemically entitled to believe the conclusion before you ‘arrive’ at said conclusion? Entitlement to believe a conclusion is attained when all 19

I don’t at all mean to suggest that there is intentional sleight of hand in what certainly is just an unfortunate, but very significant, slip on Klein’s part. I simply mean to suggest that the move looks like a verbal trick.

154

C. de Almeida

(evidentially essential) premises are believed, but none of this impugns the epistemic legitimacy of your ‘arrival’. Third, it should also be noticed how little Klein’s claim against the master argument would have accomplished even if it were defensible. To anyone who admits that “all the work is done in the sub-argument for [PREMISE (2)]”, the claim that the master argument is question-begging boils down to the claim that the skeptic’s case in the set consisting of master argument and sub-argument is redundant. I think we can all agree that redundancy is, at worst, a crime against elegance.

10.4

Skepticism, Mooreanism, and Russellianism

Klein’s attack on Academic skepticism is a clear case of both Mooreanism and Russellianism. Both views are fundamental to the current state of play in discussions of Academic skepticism, and neither view is easily refutable, even by those of us (myself included) who think neither is tenable. Russellianism is the view that no proposition for which you have no evidence (to believe)—that is, not even a bit of evidence justifying the state of tending to believe the proposition—deserves to be an object of doubt.20 The epistemologically important consequences of the view become clear when you consider that we tacitly assume that, if p is incompatible with q and you are justified in doubting that (or whether) q at a given time t, you are not justified in believing that p at t. We naturally assume that p must also be an object of doubt for you while q is.21 Thus, for instance, your being justified in believing that you are now reading a book is obviously incompatible with your being justified in believing that you are now a disembodied soul (without eyes) who cannot read anything but can be misled into falsely believing that you are now reading a book. Are you justified in, now, suspending judgment as to what you are doing? “Hell, no!”, answers the Russellian, because, the Russellian would claim, you don’t have a shred of evidence that you are now that pitiful disembodied soul. In contemporary philosophy, it was Bertrand Russell who made that kind of approach to skepticism as popular as it has become. In his immensely influential 1912 book on The Problems of Philosophy, Russell makes the point as follows: No logical absurdity results from the hypothesis that the world consists of myself and my thoughts and feelings and sensations, and that everything else is mere fancy. [...] But

20

Notice that the claim is trivialized if the agent already is fully justified in believing the proposition, or fully justified in disbelieving it. What matters here is one’s having some evidence favoring belief, as opposed to disbelief, but not yet justifying belief. 21 I eschew the contextually irrelevant details that a more precise, or more neutral, description of the phenomenon would involve, such as countenancing graded doxastic attitudes, where I speak of a justified tendency to believe a given proposition that falls short of one’s being justified in believing it. The philosophical point would remain unaltered by such details. I thank Cherie Braden for helpful discussion here.

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

155

although this is not logically impossible, there is no reason whatever to suppose that it is true. (Russell 1912, 22)

And, on that basis, he thought we could safely disregard this kind of skeptical hypothesis while admitting that the hypothesis is not refutable. For the Russellian, we find a good-enough response to skepticism when we notice that we are justified in ignoring the skeptic. Russell was still advocating the very same view when he wrote his 1959 book My Philosophical Development. While addressing testimonial justification, he writes, You might have experiences in a dream which would be equally convincing while you still slept, but which you would regard as misleading when you woke. Such facts warrant a certain degree of doubtfulness, but usually only a very small degree. In the immense majority of cases, they justify you in accepting testimony if there is no evidence to the contrary. (Russell 1959, 144–145)

The view is clear; its charms, undeniable. It is hard to think of a more popular view regarding Academic skepticism in contemporary philosophy.22 And this is exactly what we find in Klein’s work, with lots of excerpts for us to choose from.23 Here is one of them: Recall Dretske’s oft-cited Zebra-in-the-Zoo Case. [...] We are standing before some zebralooking animals and someone says, “How do you know that these animals are not very cleverly disguised Seventh-Day Adventists who dress up as zebras in order to catch you off guard so that they can begin their spiel?” Or “How do you know these are not aliens from another galaxy who have disguised themselves as zebras in order to observe large numbers of humans as they visit the zoo?” Pretty silly. Pretty unmotivated. (Klein 2004a, 119)

Klein’s point here is that the mere conversational salience of a skeptical hypothesis should not suffice for anyone to be concerned with, as he often puts it, “eliminating” that hypothesis “before” one can be justified in believing that one is looking at zebras. The reason for the claim of ‘silliness’ betrays his Russellianism: Now, if Seventh-Day Adventists or aliens have resorted to such disguises, or perhaps, even if you just thought they had, then such an alternative might need to be eliminated with epistemic “priority”, as Sosa puts it... Why should anyone think that the mere raising of the skeptical scenario imposes any obligation upon us to eliminate it in order to have the highest form of human knowledge? Where does that obligation come from? Not our ordinary practices. (Klein 2004a, 119)24

You may be thinking, as I think you should: Who, after all, has the data of our ordinary practices: Dretske (1970, 2014), the contextualists, and Brueckner (1985,

22

For more on the Russellian strategy against Academic skepticism, see de Almeida (forthcoming a). 23 But Klein seems to ignore that Russell is a major source for the view. To my knowledge, he does not anywhere describe his view as Russellian. 24 Klein’s Russellianism is very clearly on display in Klein (2002, 352; 2010, 161; 2015, § 6).The Russellian character of his view is sometimes (as in his SEP entry) obscured by references to Wittgenstein, whose anti-skeptical views differ from Russell’s. This is not the venue for a discussion of the contrast.

156

C. de Almeida

288) (their disagreements notwithstanding) on the one hand, or Klein on the other? Count me among those who would have regarded as pathological the reaction of the zoo visitor who shrugs his shoulder to the hypothesis that those animals are disguised mules, not for a moment noticing the threat to her belief that the animals are zebras, given the evidence available to her on that occasion.25 But, no, ultimately, I don’t think Klein’s Russellianism can easily be dismissed on the basis of such a dispute. Maybe our ordinary practices do sanction the dismissive zoo visitor’s reaction. Maybe the sticky point is precisely what the Russellian claims it is: It all depends on our having some evidence, however minimal, for the skeptical hypothesis in order for it to be epistemically effective, that is, for it to induce rational doubt. And, as Cohen (2000, 134) notes in reply to Klein, maybe we can epistemologically discriminate between cases of relatively mundane deception and radical demon-style deception for which we don’t have a shred of evidence. Maybe the jury is out on the data after all. (Klein is still too sanguine for his own good in this regard.) A more disturbing consequence of Russellianism is the fact that it is a form of epistemological conservatism. Why must a challenge to some of your beliefs be sanctioned, as it were, by some other of your beliefs? More generally, why must the reasonableness of a doubt depend on one’s having some evidence that the propositional object of doubt is true? To many of us (myself included), epistemological conservatism is a deeply counterintuitive view.26 It amounts to a form of epistemological chauvinism that many of us reject, even if rejecting it leaves the door open for the skeptic. The Russellian finds it reasonable to react to perceived evidential underdetermination with the serene preservation of the doxastic status quo.27 Russellianism is a form of conservatism, and Klein is a hardcore Russellian with regard to Academic skepticism. But here’s the threat to consistency: Klein is also a Moorean. And Mooreanism is an extreme form of conservatism. To the Moorean, you’re not only entitled to assume that your beliefs are ‘innocent until proven guilty’, as the Russellian would maintain, you’re also invited to believe any of the logical consequences that you care to draw from the (putatively) justified beliefs you now hold. So, again, let’s suppose you believe you’re now reading a book. Mooreanism is the view that— assuming your belief about your present activity is somehow justified (how exactly

If we could equate the ‘silliness’ of the skeptical zoo hypotheses, as Klein sees them, with Dretske’s charge of ‘irrelevance’, Dretske and Klein would agree in the ways they characterize the data of our ordinary practices. But, as we shall shortly see, for Klein, the ‘silliness’ of those skeptical hypotheses should be equated with their obvious falsehood. Dretske (1970, 2014) thought that we need an explanation of why the non-skeptical zoo visitor should resist the opportunity to perform modus tollens and admit ignorance with regard to the animals’ being zebras. The Dretskean response to the data does not warrant describing those skeptical hypotheses as ‘pretty silly’ or ‘pretty unmotivated’. (In what follows, I address the tension between a Russellian stance and the view that skeptical hypotheses are silly because obviously false.) 26 For more on conservatism, see Hamid Vahid (2004). 27 Notice that the Russellian may still consistently balk at the prospect of claiming to know the negation of a skeptical hypothesis. 25

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

157

not mattering to Mooreanism)—you are now entitled, by JC and classical logic, to infer that you are not, after all, a disembodied soul (without eyes) who can’t be reading anything. The Moorean’s dispute with the skeptic then becomes the one over whether you are ultima facie justified in believing you’re deception-free, with the Moorean saying “yes” by modus ponens from the same conditional premise, PREMISE (1) in the master argument, from which the skeptic derives her conclusion by modus tollens.28 A catchy label for inferential knowledge acquired in Moorean fashion was coined by Stewart Cohen (2002, 2004): ‘easy knowledge’, he called it. He motivates the charge by asking us to think about whether one can come to know that a given table that looks red is not a white table that’s being illuminated by red lights, given the fact that there is no evidence that the table is any color other than red. In such a scenario, can you come to know that the table is not white and illuminated by red lights by inference from your putative knowledge that the table is red? Klein reacted to the question as follows: [O]ur question is this: Can someone come to know that the table is not white illuminated by a red light solely by reasoning from the table looks red to the table is red and then to the table is not white illuminated by a red light? Since the problematic step is the first one, not the second one, that question becomes: Can someone come to know the table is red by reasoning from the table appears red? (Klein 2004c, 179)

He then goes on to explain to Cohen that one can never come to know that p solely by reasoning to p from ‘it appears that p’. Your justified belief that p, he explains, must also be one whose justification is undefeated—and that, he concludes, is not so easy after all. Something has gone very wrong in Klein’s reply to Cohen. The ‘easy knowledge’ problem was never the problem of what must be added to justified true belief in order for the belief to be a case of knowledge. It was, from inception, the problem of expanding one’s empirical knowledge (if any) too easily, i.e., illegitimately, in a Moorean fashion, by valid deduction from what one putatively knows to the negation of skeptical hypotheses incompatible with what one putatively knows.29

28 In de Almeida (forthcoming a) you’ll find a brief discussion of Mooreanism that highlights an aspect of the view that has confounded some influential Mooreans. 29 Elsewhere in his reply to Cohen, Klein does correctly frame the problem, and, consistently for a Moorean, fails to give any weight to the anti-Moorean intuition. But, surprisingly, after presenting Cohen’s red table case, Klein goes on to say: “I agree. The son [the skeptical challenger in Cohen’s scenario] should not be satisfied with this response [i.e., his father’s Moorean response in the case] because what the son is challenging is not the move from the table is red to the table is not white being illuminated by a red light, but rather the son is challenging the move from the table looks red to the table is red” (Klein 2004c, 178–179). That’s a surprising way of agreeing with Cohen! Compare to Cohen (2004, 19): “The problem of easy knowledge arises in two ways. The first exploits the deductive closure principle. [...] [I]f one has no prior knowledge that there are no red lights shining on the table, it seems counterintuitive that one could acquire it subsequently in this way.” The problem from the Moorean use of closure was precisely what Cohen (2002) had in mind when the red table case was put forward. (Cohen’s other route to ‘easy knowledge’ is irrelevant to present concerns.)

158

C. de Almeida

In any case, the above excerpt is very clear: the problematic inferential step, for Klein, is not the one from ‘the table is red’ to ‘it is not white but illuminated by red lights’. This is one of the recent passages in the Kleinian corpus where he reaffirmed his commitment to Mooreanism. In the first round of his exchange with Cohen, before the ‘easy knowledge’ splash in the literature, Klein had been more explicit: “I know I am not in the skeptical scenario,” he then wrote, “because I know that I have a hand” (Klein 2000a, 115).30 Somehow, from that Moorean stance, Klein transitioned, in the very same paper, to the following claim: The “solution” to global skeptical worries is quite different from the “solution” to ordinary incredulousness. In the latter case, the doubt can be removed by locating more evidence. In the former, careful reflection about the standard argument [i.e., the master argument] shows that either it begs the question or that it is impossible for the skeptical hypothesis to be true and for there to be evidence for its truth. Such hypotheses need not be ruled out. (Klein 2004c, 181–182; emphasis added)

Although it is not entirely clear in the above excerpt what would count as a case of ordinary incredulity, in his Stanford Encyclopedia entry, he describes the distinction more fully as follows: Ordinary incredulity arises within the context of other propositions of a similar sort taken to be known, and, in principle, the doubt can be removed by discovering the truth of some further proposition of the relevant type. On the other hand, philosophical skepticism about a proposition of a certain type derives from considerations that are such that they cannot be removed by appealing to additional propositions of that type—or so the skeptic claims. (Klein 2015, § 1)

The question that immediately arises here is: How can this distinction give rise to anything we might be tempted to call a ‘solution’ to skepticism (with or without scare quotes)? What is the basis for saying that what determines the legitimacy of a doubt is the effortlessness involved in removing it? Are you prepared to hold that, if a problem is very hard, then it must be a pseudo-problem? The harder it gets the less legitimate it is? That’s a peculiar metric for legitimacy. It gets worse. We have seen that, for Klein, the legitimacy of a skeptical doubt is compromised by its extreme implausibility. But that extreme implausibility is not independent from the apparent impossibility of removing the skeptical doubt by appeal to the kind of evidence we ordinarily bring to bear on ordinary, putatively legitimate doubt. These two ideas go hand-in-hand: this is illegitimacy by apparent insolubility. However, if I understand him, Klein should be holding that, instead of being so difficult to remove, philosophical doubt can easily be removed by the simplest of inferences! He invites us to believe he is a Moorean. Anti-skeptical knowledge—knowledge that skeptical hypotheses are false—is only a quick valid argument away for the Moorean. So, why the talk of impossibility in removing

You’ll find a defense of Klein’s reply to Cohen in the work of fellow Moorean Steven Luper (2016, § 4.1). 30

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

159

skeptical doubts on the basis of the kind of evidence on which empirical knowledge depends in ordinary circumstances? It is tempting to think that Mooreanism implies Russellianism (but not conversely). The Moorean holds both that you do have some empirical knowledge and that, given the knowledge you already have, anti-skeptical knowledge can be gained by valid deduction. Russellianism is the weaker view that hypotheses for which you have no evidence need not be dwelled upon. Mooreanism seems stronger than Russellianism in that, for the Moorean, skeptical hypotheses can be proven false, whereas, for the Russellian, you have the epistemic right to shrug your shoulders to such hypotheses. So, if you do prove them false, you go beyond what is required from an epistemologically sound response to the challenge; you have done even better than expected. Accordingly, you should expect a Moorean to make the stronger assertion, the Moorean one, rather than the weaker Russellian claim. (Normally, it is misleading to make the weaker claim of two possible claims when you see yourself as justified in making the stronger one.) So, why has Klein put so much emphasis on his Russellianism? While ‘officially’ remaining a Moorean—since he never explicitly retreated from the view—he has lately, since 2004, been emphasizing his Russellianism. Has he been retreating from Mooreanism? The absence of a bold Moorean claim in both Klein’s autobiographical piece (2010) and his Stanford Encyclopedia entry (2015) may suggest that he is. But, if there has been a retreat from Mooreanism in Klein’s work, it has been a very quiet one, with all the conceptual machinery for Mooreanism left intact.

10.5

Skepticism and Gettierization

In his review of Keith Lehrer’s epistemology, Klein puts forward the following claim about how Academic skepticism may be seen as related to the Gettier Problem: Those two problems [the Gettier Problem and skepticism] are not unrelated. The Gettier Problem pointed to the felicitous, coincidental fulfillment of the justification and truth conditions of knowledge in rather mundane situations. Skepticism—at least of the [A] cademic as opposed to the [P]yrrhonian variety—is motivated by envisioning scenarios in which we have fulfilled our epistemic responsibilities to accept only those beliefs that pass the requirements for justification but nevertheless fall short of knowledge because even if the beliefs are true, it is a lucky accident because, given what justified our beliefs, it could easily be that our beliefs are false. If the demon were sleeping, perhaps there would be hands on some occasion when we are justified in believing that there are hands. But that would be a lucky break for us. (Klein 2003b, 282)

The very same perspective on how the problems may be related can be found in Duncan Pritchard’s work: The enduring appeal of this debate [about Academic skepticism] owes a great deal to our underlying commitment to some version of the ‘epistemic luck’ platitude—the intuitive thought that knowledge is incompatible with luck. For what the skeptic claims to be

160

C. de Almeida

highlighting to us is how our putative ‘knowledge’ is in fact acquired in an extremely lucky fashion. [...] For example, one might think that one is currently seated by the fire, but one could just as well be merely dreaming that one is. [...] [T]he point is that it is merely a matter of luck, in some sense to be specified, that circumstances are as we take them to be, and thus that we cannot have knowledge of what we believe about the world, even if these beliefs are in fact true. (Pritchard 2005, 15)

And notice, to complete a trio of influential authors converging to the same view, how Linda Zagzebski sees the epistemic luck problem (most obviously involved in Gettier-type scenarios) in connection with Academic skepticism: “Coherentism about justification,” she says, “would not avoid [Academic] skepticism since the latter is not about justification, but about the connection between justification (or evidence) and truth” (Zagzebski 2009, 44). Lastly, notice how Klein explains the notion of epistemic luck that is relevant to all three authors: “The impermissible kind of luck has to do with the accidental connection between the belief and the truth condition (if you’re a reliabilist) and/or the justification and the truth condition (if you’re a defeasibility theorist)” (Klein 2005, 163; emphasis added). The view put forward in the above excerpts is clear enough. According to the three authors, you’re in the clear to regard the victim in the typical skeptical scenario as one who has a Gettierized belief. In both cases, justified true belief fails to be a case of knowledge because luck accounts for the coincidental fulfillment of—or accidental connection between—the truth and justification conditions. Now, I have no doubt that it would be very interesting to see how the above remarks on the Gettierization-skepticism connection might fit within the context of all else that both Pritchard and Zagzebski say about Academic skepticism in their writings. But that is not my object here. My object is how the Gettierization-skepticism connection (involving epistemic luck) fits within the broader context of Klein’s work on Academic skepticism. It does not. The envisaged Gettierization-skepticism connection seems to be a case of flat-out inconsistency in Klein’s work. When explaining Academic skepticism, Klein routinely represents the skeptic as one for whom no belief about the external world can be justified. It’s what the master argument purports to show. It’s what Klein tells us, for instance, here, in no uncertain terms: “The basic issue at stake [in the debate between skeptics and their opponents] is whether the justification condition can be fulfilled” (2015, § 1). But, by Klein’s own lights—in fact, by anyone’s lights—your belief cannot be Gettierized if the justification condition is not fulfilled. So, the envisaged Gettierization-skepticism connection seems clearly inconsistent with well-established Kleinian views. It’s a mirage.31 The victim in a skeptical scenario simply cannot be thought of as one whose relevant belief is Gettierized. Looking at the two issues from Klein’s perspective, it should be clear that there is no theoretical use for the notion of epistemic luck, as it applies to Here’s an earlier version of the mirage: “[T]he [Cartesian] sceptical puzzle is just this: What more needs to be added to true, justified belief to make it knowledge? The answer suggested by the sceptic is ‘certainty’” (Klein 1990, 105). Epistemic—as opposed to psychological—certainty is what Klein’s no-defeaters condition prescribes as an anti-Gettier condition. See Klein (1981).

31

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

161

Gettierized belief, in any interesting discussion of Academic skepticism. The demon-world scenario that occupies Klein in the above excerpt (2003b, 282) is an ordinary Gettier-type scenario affording no lesson as regards Academic skepticism. But the idea has made inroads in the literature on skepticism, which attests to its superficial appeal.

10.6

Klein and Closure

One of the best-known Kleinian views, one that has stood the test of time and has attracted very few opponents, is his defense of JC in the face of Dretske-style counterexamples.32 It arises from the observation that Dretske-style counterexamples to justification-closure derive their impact from confusion between JC and the following, stronger principle, that I have labeled ‘evidential closure’: Evidential closure (EC): You are ultima facie epistemically entitled to believe (i.e., you have an ultimately non-overridden justification for believing), on body of evidence e, all the logical consequences of any proposition(s) that you are ultima facie epistemically entitled to believe on e.

The distinction clearly helps us dispose of Dretske-style counterexamples. Once you are at the zoo under normal conditions, the evidence you have for believing you’re looking at zebras may be deemed inadequate for you to believe you’re not looking at perfectly disguised mules in a way that poses no obvious threat to JC. The case derives its anti-closure impact from the feeling that you need much stronger evidence in order to rule out the disguised-mule hypothesis. But that should leave the door open for the Moorean claim that you are epistemically entitled to believe that those animals are not disguised mules on the basis of a belief, namely, the belief that the animals are zebras (and that no zebra is a mule), that you are entitled to hold on a different basis. This is Klein as an occasional ally of the Academic skeptic. As we have seen above, like the skeptic, the Moorean depends on JC to launch her anti-skeptical attack. The success of Klein’s 1981 pro-closure counterattack remains the basis for his 2015 Stanford Encyclopedia claim that the old move should suffice to ward off any threats to JC. But the latter-day claim should strike us as optimistic in view of counterexamples which seem immune to the distinction between EC and JC, such as the following case:

32 The two more vocal opponents are Dretske (2014) and Brueckner (2000, 2010a). Brueckner calls the move “mysterious”. Dretske went as far as to call it “verbal hocus pocus”. The critics were wrong. Klein’s move is a definitive development to any reasonable reaction to Dretske-style attack on JC. As Klein (1995) acknowledges, the move was initiated by Irving Thalberg (1974). But it was Klein who made it crystal-clear and compelling. I discuss its limitations in both de Almeida (2011) and de Almeida (forthcoming b).

162

C. de Almeida

Napoleon: While preparing a talk on Napoleon, you double-check your sources and conclude that you can safely claim that (~p) Napoleon was not Italian (not Genoese). You expect trouble at the talk because Napoleon was born in Corsica, and you have a colleague who insists that Corsica belonged to Italy (to Genoa) when Napoleon was born.33 It didn’t. It had just been acquired by France. In any case, what most matters to your talk is that you have knowledge-level justification for believing (and claiming at the talk) that (q) Napoleon was ruthless. You’re about to leave for your talk when a paralyzing thought crosses your mind. That Corsica-obsessed colleague of yours will certainly find a way to give you a hard time at the talk. Given that you’re entitled to believe that ~p and ~p entails that p
~q, then, according to your epistemology teacher, you’re entitled to believe that if Napoleon was Italian, he was not ruthless. And you’re now wondering how you might convince your audience that your claim about Napoleon’s character can’t possibly depend on when the Treaty of Versailles was signed!34

We know how relevance logicians would react to the scenario. They would claim that your reasoning (of the form ‘~p ‘( p
~q)’) is invalid. Klein, a classicist, does not have that option. To protect JC, he would have to provide us with an explanation of your apparent lack of entitlement to believe the conditional ‘p
~q’ in the Napoleon case, one that is solely based on his epistemological apparatus, especially on the distinction between EC and JC. But he seems ill-equipped for the task. And the evidence for thinking that we, classicists, are ill-equipped for the task if we simply rely on Klein’s pro-JC move soon becomes overwhelming. Consider the following case, adapted from a famous example by Ernest Adams.35 JFK: Suppose new evidence puts all conspiracy theories to rest and establishes that John Kennedy was, indeed, killed by the shots fired at him in Dallas (and not, say, because he was poisoned on his way to the hospital), and the new evidence also establishes that only Oswald could have fired those shots. So, we’re now justified in believing that, if someone fatally shot Kennedy in Dallas, Oswald was the shooter. Why, then, are we seemingly not justified in believing, by Contraposition, that, if Oswald didn’t fatally shoot Kennedy in Dallas, no one else did?

Or consider this old favorite, one of Vann McGee’s counterexamples to modus ponens: [Reagan:] Opinion polls taken just before the 1980 [U.S. presidential] election showed the Republican Ronald Reagan decisively ahead of the Democrat Jimmy Carter, with the other Republican in the race, John Anderson, a distant third. Those apprised of the poll results believed, with good reason:

33

Assume that Napoleon’s birthdate is not disputed by the troublemaker. Notice, you would naturally be thinking of the problematic conditional in subjunctive terms—if Napolean were Italian..., or if Napoleon had been Italian... But you obviously can’t refuse to engage with one who puts that hypothesis in weaker terms—if Napoleon was Italian... —in order to challenge your claim that Napoleon was ruthless. Cf. Michael Woods (1997, 53–54): “Even in a context in which it is a presupposition accepted by speaker and hearer that I have not been misinformed, I can hardly decline to accept ‘If Mary has left-wing views, I have been totally misinformed.’” And he adds: “Note that in such a case I should not say, ‘If Mary had had left-wing views, I would have been totally misinformed.”’ 35 Here, I take a cue from David Sanford (2003, 227), who appropriated Adams’s example to make a point about indicative conditionals. 34

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

163

If a republican wins the election, then if it’s not Reagan who wins it will be Anderson. A Republican will win the election. Yet they did not have reason to believe If it’s not Reagan who wins, it will be Anderson. (McGee 1985, 462)

Again, how does the distinction between EC and JC protect the inference to what looks like an obviously unjustified conclusion? It’s easy to lose sight of what’s at stake here. I’ve put the problem to Klein as the challenge to account for apparent JC failure on the basis of his epistemological apparatus, plus classical logic, since he is a classicist.36 That may be an unfair demand, even in view of how he attracts a challenge in those terms with his optimistic claims in favor of the EC/JC distinction. Maybe the classicist moves in the literature on this kind of phenomenon—be it Gricean (1975), or Jacksonian (1987), or Sorensenian (1988), or some other—will help dispose of those apparent threats to JC without specific recourse to Klein’s distinction between EC and JC. In that case, the distinction would still remain useful (for its use against Dretske-style counterexamples), and both JC and classical logic would (ceteris paribus) remain unscathed. That should count as a Klein win. The JC-denier, on the other hand, will prevail if and only if, in at least one case, any successful response to apparent JC failure must involve the claim that appearances are not, after all, deceiving. That seems to be a fair assessment of the dispute, and I’ll leave it here, simply suggesting that life is looking increasingly easier for the JC-denier.37

10.7

Klein and Underdetermination

Even if all the other anti-skeptical moves in Klein’s repertoire fail to stop the master argument when all is said and done, Klein will still have scored a definitive victory over Academic skepticism if this last argument succeeds. The argument in question is aimed at UP, the principle on which the sub-argument for PREMISE (2) in the master argument depends. Klein offers us his anti-UP argument (not identified as such) in at least seven papers (2003a, b, 2004a, b, c, 2010, 2015), but it originated in Klein 1981 (100–104). For an elegantly put example: If it were required [by the skeptic] that the evidence, e, for some hypothesis, h, must contain the denials of all the contraries of h, it is clear that e would have to entail h. To see that, note that (~h & p) as well as (~h & ~p) are contraries of h, and it is not possible for both ~(~h & p) and ~(~h & ~p) to be true and h to be false. Thus, if the skeptic were to adopt [the Eliminate All Contraries First Principle], the evidence for h would have to entail h. [...] [I]n so far as

36

In various writings, Klein introduces restrictions to JC that should not concern us here, restrictions specifically aimed at ex falso quodlibet and verum ex quodlibet—that is, aimed at delivering the results that contradictions don’t justify anything and that not every belief justifies belief in a given necessary truth. So, the marriage of classical logic and JC is not entirely trouble-free according to him. But no theory of epistemically degenerate cases is put forward while JC is defended. 37 I discuss the options for classicist JC-advocates in de Almeida (forthcoming b).

164

C. de Almeida

skepticism remains an interesting philosophical position, the skeptic cannot impose such an outrageous departure from our ordinary epistemic practices. (Klein 2015, § 4.2)

There is a defect in Klein’s reasoning.38 Addressing the argument when it first appeared (in Klein 1981), Brueckner castigated Klein for confusing levels: in order to know that p, Brueckner (1985, 287–288) rightly claimed, S’s “justifying evidence for p must be strong enough to justify the negations of p’s contraries”, as per UP (not identified by the label back then); but, in order for S to know that she knows that p she would, indeed, have to explicitly run the contraries in her head, as it were, and consciously believe that her evidence counted against each of them. Brueckner described the confusion as “a simplistic objection [...] which we must quash” (287). I find it hard to argue with Brueckner here. Klein never explicitly rid himself of the level confusion.39 But it doesn’t strike me as a disqualifying objection, not in the context of a discussion of UP anyway. Is UP the place where skepticism rears its ugly infallibilist head, as Klein suggests? That is the kernel of Klein’s challenge here, although he never explicitly addressed UP.40 So, think of Klein’s challenge as follows: If UP is a necessary condition of justification on an evidentialist view of the matter, S’s evidence for believing a given p must imply the negation of every ‘contrary’ of p (meaning ‘every proposition that’s incompatible with p’). That’s the natural reading of his talk of the evidence containing the negations of all of p’s contraries. But the set of all negated contraries of p entails p. So, the UP skeptic is an infallibilist, for requiring that the evidence for believing p entail p. Is this true? Here, I must ask you to look at UP again (in Sect. 10.2 above). Notice that UP does not require that the negation of each contrary of p be entailed by the evidence you have for p. It only makes the much weaker requirement that justifying evidence

38

The defect in question is more evident in some of the other six sources referenced in the text. The excerpt from Klein’s SEP entry is chosen here for the clear and elegant expression of its main point. In the paragraph that follows the quoted passage in his SEP entry, Klein clearly invites his reader to think of the ‘elimination’ of contraries as something that would be performed by the reasoner, not as something that the believer’s evidence should be expected to do on its own, as required by UP. 39 In de Almeida (2016), I discuss level confusions in Stroud’s work on skepticism that are relevant to the problem we see here. In Klein’s case, the corresponding confusion can also be seen in his discrepant expressions of the master argument. Sometimes, he frames the argument in the firstperson: ‘If I’m justified in believing that p...’. Sometimes, he correctly frames it in the third-person: ‘If S is justified in believing that p...’. He has failed to see that level confusions are engendered by the first-person formulation—confusions that severely affect our understanding of Academic skepticism—as have many of those who write on skepticism, such as Keith DeRose (1995), for instance. This kind of level confusion is discussed at length in de Almeida (2016). 40 Frankly, the fact that Klein never explicitly addressed UP, in spite of its prominence in the literature on Academic skepticism, is a disappointing aspect of his published work on the issue. But, if you look just under the surface, the challenge to UP is clearly there. It is certainly more charitable to think that he thought he was, in his own language, addressing UP than to think that he simply ignored the literature on UP. I submit that his opposition to an ‘Eliminate All Contraries First Principle’ should be understood as applying to UP. Otherwise, we’ll be looking at a gaping hole in his work on Academic skepticism.

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

165

for p be justifying evidence for the negation of any given contrary of p.41 My evidence for believing that I am in Brazil (now) is not entailing evidence. Why should it entail that I am not in Singapore (now)? But it certainly would be stupefying if my evidence for believing I’m in Brazil weren’t good evidence for believing I’m not in Singapore. It’s not entailing evidence for either I am in Brazil or I’m not in Singapore. And, if it’s not entailing evidence, Klein doesn’t have a case against UP. Indeed, the conjunction of every negated contrary of p entails p (as you can easily see from Klein’s choice of contraries). The good evidence I have for believing that p is just as good for believing the negation of each contrary of p.42 So, in the Brazil/Singapore case, I have good inductive evidence for believing the infinitely long conjunction that implies ( p) I am in Brazil. But it should clearly seem fallacious to infer, from that, that I have, in my belief system, entailing evidence for believing that p. One who, for instance, has good inductive evidence to believe that p and good inductive evidence to believe that q will, if she forms the beliefs that p and that q, have acquired entailing evidence for believing that p & q, but the entailing evidence that p & q is not her evidence for p together with her evidence for q; the entailing evidence are her beliefs that p and that q. The evidence for the conjuncts will still be inductive. If brought to bear against UP, Klein’s case against an ‘Eliminate All Contraries First Principle’ is designed to convict the UP-skeptic of the crime of infallibilism.43 The verdict is in: it’s a case of the straw man fallacy.

10.8

Concluding Remarks: Unfair to the Skeptic

It should be noted that each of the Kleinian views discussed here appeared at least twice in his publications. None of them can plausibly be seen as having resulted from a fleeting thought. I have tried to provide evidence for each of the following claims. (A) Klein has suggested that a skeptical scenario may be thought of as one where the victim’s justification for the relevant belief is only luckily connected to the truth of the belief, if the belief is true, as in cases of Gettierization, but we’ve seen that the claim is inconsistent with some of the tenets of his analysis of Academic skepticism. (B) He has also aligned himself with both Mooreanism and Russellianism, but neither describing himself as a Moorean nor acknowledging Russell’s role in shaping this 41

Beware of confusing this consequence of UP with the objectionable consequence of EC. The latter requires that justifying evidence for p be justifying evidence for any of p’s logical consequences. UP is weaker than EC. 42 As Brueckner (1985, 288) suggests, you can go ahead and think of any extravagant contrary of p you may care to bring up. Good evidence for believing that p must be good evidence to believe the negation of each extravagant contrary of p. 43 Interestingly, Klein’s defeasibilism is infallibilistic. I discuss the problem and offer a fallibilistic alternative for defeasibilists, in de Almeida (2017).

166

C. de Almeida

brand of anti-skeptical conservatism. (C) On the most plausible reading of those two views, Russellianism is logically weaker than Mooreanism. (D) It is confusing to see that, in his latest writings, Klein’s emphasis is on his Russellian view, the weaker view of the two, which makes us ponder the possibility that he has quietly become an ex-Moorean. But, (E) one doesn’t consistently become an ex-Moorean while subscribing to views that imply a Moorean stance. (F) Klein’s defense of justificationclosure against Dretske-style counterexamples is a major contribution to contemporary epistemology. But (G) Klein’s (Moorean) combination of justification-closure with classical logic is vulnerable to objections from classicist closure-deniers. (H) Brueckner scored twice against Klein: he rightly accuses Klein of level confusions in the analysis of Academic skepticism, and he also rightly notes that Klein’s circularity claim against the skeptic can be seen as a covert circularity charge against modus tollens. (I) Brueckner’s reasoning in this regard may justify an even stronger claim concerning other inference rules. (J) Klein’s case against the sub-argument for PREMISE (2) in the skeptic’s master argument can be seen as an argument against the so-called ‘Underdetermination Principle’; the case, if effective, puts an end to Academic skepticism as we know it. But (K) the closest thing to a Kleinian case against UP is fallacious. And (L) UP is never explicitly addressed in Klein’s work on skepticism, in spite of the fact that it is widely discussed in the literature on the problem. Disagreements notwithstanding, Klein’s work on Academic skepticism touches on just about every important aspect of the problem, and I admire it for broadening our horizons on the issue, and especially for making a conservative response to skepticism so enticing. The optimist in me loves to think that Klein will prevail against the skeptic. But the philosopher in me has no use for optimism.44

References Adams, E. W. (1988). Modus Tollens revisited. Analysis, 48(3), 122–128. Bennett, J. (2003). A philosophical guide to conditionals. Oxford: Oxford University Press. Brueckner, A. (1985). Skepticism and epistemic closure. Reprinted in Brueckner, A.: Essays on skepticism, 281–305. Oxford: Oxford University Press, 2010. Brueckner, A. (1994). The structure of the skeptical argument. Reprinted in Brueckner, A.: Essays on skepticism, 319–326. Oxford: Oxford University Press, 2010. Brueckner, A. (2000). Klein on closure and skepticism. Reprinted in Brueckner, A.: Essays on skepticism, 327–336. Oxford: Oxford University Press, 2010. Brueckner, A. (2005). Cartesian skepticism, content externalism, and self-knowledge. Reprinted in Brueckner, A.: Essays on skepticism, 163–173. Oxford: Oxford University Press, 2010.

44

I’m grateful to J. R. Fett, Gregory Gaboardi, and John N. Williams for helpful comments on drafts of this chapter. A very special thank-you goes out to Cherie Braden, who supplied me with a large number of insightful comments on the penultimate draft of this chapter. I must also thank the editors, Branden Fitelson, Cherie Braden, and Rodrigo Borges, for honoring an author to whom all of us who love epistemology are deeply indebted, Peter Klein.

10

Klein, Skepticism, Epistemic Closure, and Evidential Underdetermination

167

Brueckner, A. (2010a). ~K~SK. In A. Brueckner (Ed.), Essays on skepticism (pp. 367–381). Oxford: Oxford University Press. Brueckner, A. (2010b). Skepticism and closure. In J. Dancy, E. Sosa, & M. Steup (Eds.), A companion to epistemology (pp. 3–12). Malden/Oxford: Wiley-Blackwell. Burgess, J. P. (2005). No requirement of relevance. In S. Shapiro (Ed.), The Oxford handbook of philosophy of mathematics and logic (pp. 727–750). Oxford: Oxford University Press. Cohen, S. (1988). How to be a fallibilist. In J. E. Tomberlin (Ed.), Philosophical perspectives volume 2, epistemology (pp. 91–123). Atascadero: Ridgeview. Cohen, S. (2000). Replies. Philosophical Issues, 10(1), 132–139. Cohen, S. (2002). Basic knowledge and the problem of easy knowledge. Philosophy and Phenomenological Research, 65(2), 309–329. Cohen, S. (2004). Structure and connection: Comments on Sosa’s epistemology. In J. Greco (Ed.), Ernest Sosa and his critics (pp. 17–21). Chichester: Wiley. De Almeida, C. (2011). Epistemic closure, skepticism and defeasibility. Synthese, 188(2), 197–215 2012. De Almeida, C. (2016). Stroud, skepticism, and knowledge-claims. Sképsis, 7(14), 40–56 Open access: http://philosophicalskepticism.org/skepsis/numero-14/. De Almeida, C. (2017). Knowledge, benign falsehoods, and the gettier problem. In R. Borges, C. de Almeida, & P. Klein (Eds.), Explaining knowledge: New essays on the gettier problem (pp. 292–311). Oxford: Oxford University Press. De Almeida, C. (Forthcoming-a). On our epistemological debt to Moore and Russell. In S. Hetherington & M. Valaris (Eds.), Knowledge in contemporary philosophy. London: Bloomsbury. De Almeida, C. (Forthcoming-b). Epistemic closure and post-gettier epistemology of reasoning. In S. Hetherington (Ed.), The Gettier Problem. Cambridge: Cambridge University Press. DeRose, K. (1995). Solving the skeptical problem. Reprinted in Skepticism: A contemporary reader, edited by K. DeRose, & Ted A. Warfield, 183–219. Oxford: Oxford University Press, 1999. Dretske, F. I. (1970). Epistemic operators. The Journal of Philosophy, 67(24), 1007–1023. Dretske, F. I. (2014). The case against closure. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (2nd ed., pp. 27–40). Oxford: Wiley-Blackwell. Firth, R. (1978). Are epistemic concepts reducible to ethical concepts? In A. Goldman & J. Kim (Eds.), Values and morals (pp. 215–229). Dordrecht: Reidel. Grice, H. P. (1975). Logic and conversation. Reprinted in Conditionals, edited by F. Jackson, 155–175. Oxford: Oxford University Press, 1991. Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press. Jackson, F. (1987). Conditionals. Oxford: Basil Blackwell. Klein, P. D. (1981). Certainty: A refutation of scepticism. Minneapolis: University of Minnesota Press. Klein, P. D. (1987). On behalf of the skeptic. In S. Luper-Foy (Ed.), The possibility of knowledge: Nozick and his critics (pp. 267–281). Totowa: Rowman & Littlefield. Klein, P. D. (1990). Epistemic compatibilism and canonical beliefs. In M. D. Roth & G. Ross (Eds.), Doubting: Contemporary perspectives on skepticism (pp. 99–119). Dordrecht: Kluwer. Klein, P. D. (1995). Skepticism and closure: Why the evil genius argument fails. Reprinted in Epistemology: Contemporary readings, edited by M. Huemer, 552–574. London: Routledge, 2002. Klein, P. D. (2000a). Contextualism and the real nature of academic skepticism. Philosophical Issues, 10(1), 108–116. Klein, P. D. (2000b). The failures of dogmatism and a new Pyrrhonism. Acta Analytica, 15, 7–24. Klein, P. D. (2002). Skepticism. In P. K. Moser (Ed.), The Oxford handbook of epistemology (pp. 336–361). Oxford: Oxford University Press. Klein, P. D. (2003a). How a Pyrrhonian skeptic might respond to academic skepticism. In S. Luper (Ed.), The skeptics: Contemporary essays (pp. 75–94). Hampshire: Ashgate.

168

C. de Almeida

Klein, P. D. (2003b). Coherence, knowledge and skepticism. In E. Olsson (Ed.), The epistemology of Keith Lehrer (pp. 281–297). Dordrecht: Kluwer. Klein, P. D. (2004a). Skepticism: Ascent and assent? In J. Greco (Ed.), Ernest Sosa and his critics (pp. 112–125). Chichester: Wiley. Klein, P. D. (2004b). There is no good reason to be an academic skeptic. In S. Luper (Ed.), Essential knowledge: Readings in epistemology (pp. 299–309). New York: Pearson Education. Klein, P. D. (2004c). Closure matters: Academic skepticism and easy knowledge. Philosophical Issues, 14(1), 165–184. Klein, P. D. (2005). Infinitism’s take on justification, knowledge, certainty and skepticism. Veritas, 50(4), 153–172. Klein, P. D. (2010). Self-profiles: Peter Klein. In J. Dancy, E. Sosa, & M. Steup (Eds.), A companion to epistemology (pp. 156–163). Malden/Oxford: Wiley-Blackwell. Klein, P. D. (2015). Skepticism. In E. N. Zalta (Ed.), The stanford encyclopedia of philosophy, Summer 2015 Edition. https://plato.stanford.edu/archives/sum2015/entries/skepticism/ Luper, S.. (2016). Epistemic closure. In E. N. Zalta (Ed.), The stanford encyclopedia of philosophy, Spring 2016 Edition. https://plato.stanford.edu/archives/spr2016/entries/closure-epistemic/. Mares, E. D. (2004). Relevant logic: A philosophical interpretation. Cambridge: Cambridge University Press. McGee, V. (1985). A counterexample to modus ponens. The Journal of Philosophy, 82(9), 462–471. Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press. Read, S. (1988). Relevant logic: A philosophical examination of inference. Oxford: Basil Blackwell. Russell, B. (1912). The problems of philosophy. London: Williams and Norgate. Russell, B. (1959). My philosophical development. London: George Allen & Unwin. Sanford, D. H. (2003). If P, then Q: Conditionals and the foundations of reasoning (2nd ed.). London: Routledge. Sinnott-Armstrong, W., Moor, J., & Fogelin, R. (1990). A defence of modus tollens. Analysis, 50 (1), 9–16. Sorensen, R. A. (1988). Dogmatism, junk knowledge, and conditionals. The Philosophical Quarterly, 38(153), 433–454. Stalnaker, R. (1975). Indicative conditionals. Reprinted in Conditionals, edited by F. Jackson, 136–154. Oxford: Oxford University Press, 1991. Thalberg, I. (1974). Is justification transmissible through deduction? Philosophical Studies, 25(5), 347–356. Vahid, H. (2004). Varieties of epistemic conservatism. Synthese, 141(1), 97–122. Woods, M. (1997). Conditionals. Oxford: Oxford University Press. Zagzebski, L. (2009). On epistemology. Belmont: Wadsworth.

Part III

Justification

Chapter 11

Finite Minds Michael Huemer

Abstract Infinitism claims that a belief is justified only if it stands at one end of an infinite series of available reasons. I argue that this condition cannot be satisfied by any human mind. In general, as one expands the list of propositions that a subject is said to believe, one must either add new basic evidence, add increasingly complex propositions, or include propositions that are ever more similar to each other. But I argue that (a) for a proposition to be available to one as a reason, one must either believe it or be disposed to acquire the belief without the need of acquiring new evidence, (b) there is a limit to the complexity of the propositions that a human mind can grasp, and (c) there is a limit to the capacity of a human mind to distinguish propositions. Thus, no human being can have infinitely many propositions available as reasons. Keywords Infinitism · Epistemology of the infinite · Finite mind · Finite mind objection · Cognitive limitations · Complexity of propositions · Availability of propositions · Klein · Epistemology

11.1

Infinitism and the Finite Mind Objection

According to epistemological infinitism, we have an infinite series of available reasons standing behind any justified belief—that is, a reason for the belief, and a reason for that reason, and so on (Klein 2007; Turri 2009; Aiken 2014). The most obvious objection is the finite mind objection: human minds are finite and so cannot contain infinitely many beliefs; thus, assuming that an available reason must be something believed, we cannot have infinitely many available reasons (Williams 1981; Audi 1993, 127–128). Peter Klein’s initial response to this objection was to suggest that (i) an available reason need only be dispositionally believed, not occurrently believed, (ii) an infinite series of available reasons therefore only requires infinitely many dispositional M. Huemer (*) Department of Philosophy, University of Colorado Boulder, Boulder, CO, USA © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_11

171

172

M. Huemer

beliefs, and (iii) we can have infinitely many dispositional beliefs, even though we have only finitely many occurrent beliefs (Klein 1999, 300, 307–309).1 Later, he backed away from requiring that the members of an infinite series of reasons should be even dispositionally believed, suggesting that, for a proposition p to count as available to a subject S, it might be enough for S to have some epistemic practices that would lead S to acquire the belief that p (2007, 13). In the literature on infinitism, attention has drifted away from the finite mind objection to other, more subtle objections.2 I believe, however, that a version of the finite mind objection succeeds. I shall argue that human minds are finite in ways that prevent us from having infinitely many beliefs (whether occurrent or dispositional), or in any other way having infinitely many reasons available to us.3

11.2

When Is a Reason Available?

11.2.1 Availability through Belief I propose to leave aside the question of what makes a proposition count as a reason for another proposition, focusing instead on what makes a reason count as available to a given subject.4 Let us allow that if S occurrently believes that p at time t, then p is thereby available to S at t (taking availability in a purely psychological sense).5 That is the simplest case. Also, if S dispositionally believes that p at t, then p is thereby available at t. But what is required for one to count as dispositionally believing p? I

1 He further distinguishes ordinary dispositional beliefs that involve already-formed dispositions to (occurrently) believe something, from dispositional beliefs that involve second-order dispositions to form the first sort of dispositions. He proposes that the infinitist may rely on the latter sort of dispositional beliefs. For more on the nature of dispositional belief, see Sect. 11.2.1 below. 2 See the recent papers in Turri & Klein 2014, which, despite many criticisms of infinitism, only briefly mention the finite mind objection. Fumerton (2014, 77), though otherwise critical of infinitism, seems to grant that the finite mind objection fails. But note that Kvanvig (2014, 137–141) and Podlaskowski and Smith (2011) advance variants of the finite mind objection. 3 In Huemer 2016, I argue that it is metaphysically impossible for anything to possess an infinite intensive magnitude. I think having infinitely many available reasons might require one to instantiate an infinite intensive magnitude (for instance, infinite intelligence, or infinite information density). But I shall not pursue that suggestion here. Here, I am content to argue that having infinitely many available reasons is psychologically impossible for actual human beings. 4 Klein (2005, 136) draws this distinction. Earlier (Klein 1999, 300), he distinguished “subjective availability” from “objective availability”; in those terms, I am concerned with subjective availability. 5 One might think that “availability” should be treated normatively—e.g., perhaps a reason should be considered available only if it is justified, or even known. I leave this point aside, since I want to focus on the unsatisfiable psychological demands of infinitism. As Fumerton (2014, 80–81) observes, normative conditions on availability create additional problems for an infinitist.

11

Finite Minds

173

assume that this cannot be a brute fact; there must be some other facts about one, in virtue of which one counts as dispositionally believing that p. There are, in my view, three ways of dispositionally believing that p. The first way is to have occurrently believed p at some time in the past, having neither forgotten nor changed one’s mind about p, and consequently remaining disposed to occurrently believe p again should the question arise. (Perhaps there are further conditions required to qualify as believing p, but this is good enough for now.) For example, suppose I point at Peter Klein while he is asleep and say, “There is one of the few philosophers who believes infinitism.” What makes this true? Though he is presumably not thinking about infinitism at the moment, he has on many occasions during his waking life consciously entertained infinitism, and at these times, he affirmed it. He has not changed his mind since the last time he affirmed it, nor has he forgotten his view on infinitism, and if the question were posed to him again, he would again affirm it. That is enough for him to now qualify as a believer of infinitism. Second, one may dispositionally believe that p in virtue of having an (occurrent) appearance that represents that p (that is, a mental state whereby it seems to one that p), where one is disposed to take this appearance at face value (again, there might be additional conditions required). For example, let’s say Peter enters an ordinary classroom, seeing it and its contents normally as he walks to the front of the room. Does he at this time know that the room has a ceiling? It would be very odd to deny that he knows this, although it would also be very strange if he occurrently entertained the thought (“Ah, good, another room with a ceiling!”). Since knowledge entails belief, he believes that the room has a ceiling, albeit nonoccurrently. What makes this true? Well, his visual experience represents the room as having a ceiling, and, so I assume, he has the normal attitude of trust in his visual experiences. He is not, for example, a skeptic who thinks vision is not to be trusted, nor does he have some special doubt, or reasons for doubt, concerning ceilings, or concerning this ceiling in particular. He thus would affirm the ceiling’s existence if the issue arose. All this enables him to qualify as dispositionally believing that the room has a ceiling. Third, one may dispositionally believe p by having some other (assertive6) representational mental state that obviously entails p, while being disposed to affirm this implication if the issue arises (again, other conditions may be needed). For example, Peter now believes that 47 is more than 2, even if he is not presently thinking about the numbers 47 and 2. Why does he count as believing this? Because he believes other things about the number system that obviously entail that 47 is more than 2. For instance, he thinks that the right-most position in a numeral is the “one’s place,” the position second from the right is the “ten’s place” (representing the number of tens contained in the number referred to), that ten is more than one,

6

Assertive mental states are those that represent their contents as true or actual. For instance, perception, intuition, memory, and belief are all assertive; however, imagination and desire are non-assertive (see Huemer 2001, 53–54).

174

M. Huemer

and so on. Those beliefs, in turn, may also be dispositional (he probably is not occurrently thinking of the fact that the position second from the right is the ten’s place, and so on). That example illustrates that one can dispositionally believe that p through having a chain of beliefs—that is, one might believe that p in virtue of believing that q, where one believes that q in virtue of believing that r, and so on. However, there are limits to such chains—they must be reasonably short, and the implications reasonably obvious. Consider, for example, Plato’s argument for the doctrine of recollection in the Meno. Plato has Socrates leading the slave boy down a chain of reasoning for the conclusion that the length of the diagonal of a square equals the square root of two times the length of the side (Plato, Meno, 84a–85d).7 At each stage, Socrates merely asks the slave questions, and the slave answers according to his own perception, but pffiffiffi the questions are so ordered that at the end, the slave sees that the diagonal is 2 times the side. wants us to conclude that the slave pffiffiPlato ffi already believed that the diagonal was 2 times the side but had merely “forgotten” it. But this of course is false. What Socrates does in the dialogue is to cause the slave to acquire this belief by inducing him to attend to the right facts. The lesson is that if the chain of implications that would lead from one’s current beliefs to a certain conclusion is too long or non-obvious, then the conclusion is not something one presently believes (not even dispositionally) but merely something one can come to believe.

11.2.2 Availability through Epistemic Practices As noted above, Klein suggests that p might count as available in virtue of one’s merely having epistemic practices that would lead one to affirm p if the question as to whether p arose. For instance, [Helena is the capital of Montana] might now count as available to S by virtue of the fact that S’s epistemic practices are such that if the question arose as to what is Montana’s capital, S would consult a reliable almanac that in fact lists Helena as the capital (Klein 2007, 13). In fairness, Klein does not say that this is in fact the correct view of availability, but only that he does not want to rule out this liberal view of availability. I, however, want to rule it out. This view of availability is far too liberal. Why? Our theory of availability should reflect the role that “available reasons” are supposed to play in our epistemological theory. The theory of an infinite series of available reasons is supposed to solve a problem that allegedly plagues foundationalism. The problem is supposed to be that putatively foundational

7 More precisely, the conclusion is that given a square ABCD, the square constructed on the diagonal AC has twice the area of the original square.

11

Finite Minds

175

propositions are arbitrary, and hence unjustified, because one lacks reasons for them (Klein 1999, 299).8 Infinitism is supposed to avoid this sort of arbitrariness by positing an available reason for each belief. Therefore, the sense in which a reason needs to be “available” is whatever sense would be required to help answer a charge of arbitrariness and secure justification. The liberal view of availability does not satisfy this requirement—it does not help to answer a charge of arbitrariness or to secure justification. For example, suppose Sue believes that unicorns exist. She has no actual evidence for this in any ordinary sense—she has never seen a unicorn, has never heard credible testimony about one, does not know of any observations that would be best explained by hypotheses about unicorns, and so on. Her unicorn belief is purely arbitrary. Now add a bit more to the story. Unicorns, as chance would have it, actually exist, even though no one has ever had any evidence of this. Just now, bizarrely enough, there happens to be a baby unicorn hiding under Sue’s bed. Suppose, further, that if someone were to ask Sue for justification for her belief in unicorns, she would start looking around for a unicorn. She would start by looking under the bed, whereupon she would see the baby unicorn, which she would happily cite as evidence that unicorns exist. But alas, no one ever asks Sue for justification, so she never looks for and never sees any unicorns. End of story. Is Sue justified in believing that unicorns exist? Clearly not. Is her belief in unicorns arbitrary? Clearly so. This is about as clear a case of an arbitrary and unjustified belief as one could come up with. Of course, if Sue were to look under the bed and see the unicorn, she would then be justified in believing in unicorns. But that doesn’t make her belief actually justified, or even non-arbitrary, in a world in which she never does look. Briefly, here is what I take this case to illustrate: for a reason to count as available in the epistemologically relevant sense—the sense that precludes arbitrariness and secures justification—it must at least be a reason that one could and would deploy without acquiring new evidence. That is, if we allow propositions one is disposed to believe but does not yet believe to count as available, the activation conditions for one’s disposition may not include one’s acquisition of new evidence. Seeing new objects is acquiring new evidence, so if one must look around until one sees a new object in order to acquire the belief that p, then p is not now available. What exactly counts as “acquiring new evidence” is vague. For example, if one discovers a new proof from previously known axioms, does this qualify as acquiring new evidence? Perhaps it does if and only if the proof is sufficiently non-trivial. But this vagueness is not a problem. We don’t need to have all cases definitively classified; it is enough that there are some clear cases of acquiring new evidence, and that, for example, looking under the bed clearly counts for Sue as acquiring new evidence.

8

This is a slight misstatement, since having non-inferential justification for p is compatible with also having inferential justification for p. But consider an allegedly foundational belief that does not have any inferential justification. That belief, according to Klein, is arbitrary.

176

11.3

M. Huemer

Some Putative Infinite Sets of Beliefs

11.3.1 Numbers The main reason for thinking that we have only finitely many available reasons is the sheer difficulty of thinking of any way in which we might plausibly have infinitely many available reasons—each of the obvious ways one might think of turns out not to be viable. For example: I believe that 2 is more than 1, that 3 is more than 1, that 4 is more than 1, and so on. So it looks as though I have infinitely many beliefs, and thus infinitely many psychologically available propositions (cf. Fumerton 2014, 77). Right? No, not really. There are numbers in the series of natural numbers that I am unable to think about or refer to.9 At some point in the series, there comes a number that is too large for me to comprehend; in some sense, I cannot think about that number. At some later point in the series, there comes a number too large for me even to perceive its conventional representation (e.g., because the numeral would not fit in my visual field, even when the characters were written at the smallest size that I could discern), much less use that representation in a sentence. Even on a generous reading of “ability to refer to x,” I am not able to refer to such a number.10 What if I were to invent a new system for representing numbers, one that is much more compact than the conventional decimal representation system? Still, there must be limits to how many distinct numbers I could individually represent, however clever my representational system. Presumably, for me to use a representational system, I must at least be able to distinguish the numerals representing different numbers from each other. If the symbols are visual, then the limits of my visual discrimination capacities will limit how many symbols I can meaningfully use; there are only so many scenes that I can visually distinguish. The same goes for auditory, tactile, or any other sort of symbols I might use. The same point applies even if the symbols that I employ exist purely in my imagination. What if I employ representations that are extended over time? I might adopt a representation system in which the symbols for larger numbers become ever longer (like the conventional decimal system we are already familiar with), and in which

9 One way to appreciate this point is to read about Graham’s Number, once said to be the largest finite number ever used in a serious mathematical proof. I won’t here try to explain what this number is, since doing so requires considerable time (see Gardner 1977). The considerable work required to describe Graham’s Number helps to bring home how increasingly large numbers demand greater cognitive resources to refer to them. 10 Don’t worry about the apparent self-referential paradox, i.e., that I have apparently just referred to numbers that I can’t refer to. I mean that I can’t refer to these numbers individually, i.e., I can’t pick out specific ones. I can only refer to this general category of numbers. You might also wonder about the paradox generated by the expression “the smallest number that I can’t individually refer to.” I don’t know the solution to this sort of paradox, but I assume that the solution is not to declare that I can individually refer to every number.

11

Finite Minds

177

these increasingly lengthy symbols simply need to be reviewed over ever longer periods of time. However long a numeral is, there will be some amount of time such that I could read the entire numeral if given that amount of time. Now, there will still be some symbols that I could not read in my entire lifetime, even if I read them at the maximum speed of which I am capable. But one might argue that the numbers denoted by such symbols still count as representable by me, since I could represent them, using the mental capacities I actually possess, if only I lived long enough. Right? Again, no, not really. The problem is that my memory is limited. If I read a numeral containing a million digits, by the time I reach the end of it, I will have forgotten the beginning. At no time will I have a single mental state that takes in the entire numeral; therefore, at no time will I have a mental state that refers to the specific number picked out by the symbol. To bring out this point, imagine that sometime after the cure for aging is discovered, I decide to read a very long mathematical statement. I spend 500 years reading a long numeral, which is followed by the symbol “¼,” followed by another numeral that requires 500 years to read. At the end of the millennium of reading, I am asked whether the statement I just read is true. Given my actual mental capacities, I would be unable to answer. If I even remembered that there was an “¼” sign somewhere in the statement, I would still have no idea whether the statement was true, as I would be unable to remember any of the digits I had read more than 500 years ago. Now, you might think that I could refer separately to the two long numbers, by using such expressions as “the first really long number I read” and “the second really long number I read,” and thus I could form distinct beliefs about the two numbers. Still, there would be only finitely many numbers I could refer to in this way. Notice that these descriptions themselves invoke the ordinal numbers “first” and “second,” so this sort of description faces the same limitations as the numeral system in general. As n increases, the expression referring to the nth number becomes increasingly unwieldy. Perhaps I could just point at parts of the paper and say “that number” and “that number” (referring to the numbers denoted by the inscriptions pointed at). In this case, I would be relying on the seemingly unlimited capacity for referring to distinct things using indexical expressions. This approach will be discussed in Sects. 11.3.3 and 11.4.4 below. For now, I limit myself to the observation that only finitely many numbers can be denoted using non-indexical symbols. In sum, to form a belief about a particular number, I must be able to refer to that number. There are only finitely many numbers that I can refer to (without using indexicals). So there are only finitely many numbers about which I can form (non-indexical) beliefs. Propositions about other numbers are unavailable to me, even on Klein’s view of availability.11

11

Audi (1993, 127) makes a similar point, which Klein (1999, 307) appears to grant.

178

M. Huemer

Note that all of this is consistent with the fact that I can have beliefs with infinitely many implications or instances. For example, I presently believe that every natural number has a successor. This belief has infinitely many instances (namely, that 1 has a successor, that 2 has a successor, and so on). Nevertheless, it is but one belief. Nor do I count as believing each of the infinitely many propositions it entails, considered separately, since most of them are propositions I cannot grasp. To vindicate infinitism, I must not merely have beliefs with infinitely many instances or other implications; I must believe (or otherwise have psychologically available to me) infinitely many distinct individual propositions.

11.3.2 Colors Here is a second try. Suppose I look up at the sky. I see it to be a specific shade of blue. I may not have a name for that exact shade, but that needn’t preclude my referring to it, at least in thought. Because I have a general attitude of trust in my visual experiences and there is nothing unusual about this particular experience, I count as (dispositionally) believing, of that portion of sky, that it is that exact color— or so one might argue. Now, the sky’s being that exact color directly and obviously entails, for each other possible (non-overlapping) color, that the sky is not that other color. So I dispositionally believe, of every other possible color, that the sky is not that other color. All of this is in accord with the account of dispositional belief given in Sect. 11.2.1. But, one might claim, there are infinitely many such other colors, because, given any point in the color solid, there are infinitely many regions not containing that point. So it looks as though I have infinitely many dispositional beliefs about the color of the sky. Again, not really. We may theoretically model the color space as a threedimensional continuum of points, but a human mind cannot individually refer to each of continuum-many colors. The eye has limited discriminatory capacity, and thus, while there may not be a determinate answer to “how many colors can the eye discriminate?”, it definitely is not the case that the eye can discriminate infinitely many colors. (Even smaller is the number of colors for which we have concepts.) So, even if in some sense there are infinitely many possible colors, I lack the ability to individually refer to all these colors. Perhaps I have sold short my conceptual abilities. Perhaps, for example, colors are certain classes of spectral reflectance distributions, and I can refer to these spectral reflectance distributions using mathematical concepts. I can use decimal numbers to represent increasingly precise wavelengths of light. But at some point, this runs into the same sort of problem as that encountered in referring to very large natural numbers. There are not infinitely many decimal numbers to which I can refer, because at some point, a decimal expansion becomes too complex for me to comprehend, or even to take in a representation of.

11

Finite Minds

179

So again, even if we accept that there are infinitely many colors, most of them will be colors to which I cannot refer. About these colors, I have no ability to form beliefs. I can only form beliefs about finitely many colors. A similar point applies to shapes, sounds, sizes, and any other class of properties you care to name.

11.3.3 Indexicals Klein does not offer either of the above two suggestions. Instead, he offers a third way in which a person might have infinitely many beliefs available. Suppose I have the concept of red, and I have the ability to refer to things that are presented to me using the indexical “that.” So if I am presented with an object, I can believe of the object that it is red, which I would express with the sentence, “That is red.” Now, there could be infinitely many objects with which I could be presented, in which case I would be capable of believing each of infinitely many propositions, each of the form “That is red” (Klein 2005, 138). Alas, these propositions could not all be available to me in an epistemically relevant sense. Recall that the relevant sense of “availability”—the sense that enables availability of reasons to answer a charge of arbitrariness and confer justification—requires that “available” propositions be available to one without one’s needing to acquire new evidence. At any given point in time, a human being can only have finitely many objects available to refer to. Space may be infinitely divisible, but we cannot actually distinguish infinitely many objects, since each of our senses has limited discriminatory capacity. We may extend consideration from objects presently perceived to objects perceived in the past; however, we each have only a finite life history, and there is some minimum time that an object can be presented and still be registered by a human mind (if objects are flashed before one’s eyes too quickly, one cannot distinguish them). Therefore, there can be only finitely many objects that we have ever been in a position to refer to. Even if a human were to become immortal, there would never come a point in time at which the human had perceived and could distinguish infinitely many objects. Klein’s example turns on a very liberal notion of availability: there could be infinitely many objects such that each of them could be presented to one, in which case one could then form a belief about the object presented. But this is a notion of availability we have already rejected: for me to be presented with some new object which I have never before perceived would be for me to acquire new perceptual evidence. The propositions that I would then believe are not now available to me. More specifically, propositions describing the colors of objects that I have in fact never seen are not now available to me as reasons—they are not made presently available to me by the mere fact that if I were to see those objects, I would then have perceptual beliefs about their colors. To think so would be akin to holding that [there is a unicorn under the bed] is available to Sue before she looks under the bed.

180

11.4

M. Huemer

Entertainable Propositions

So three attempts to show how we might have infinitely many available propositions fail. Is there a way to show in general that no attempt can succeed? There may be no rigorous proof of that conclusion, but the conclusion becomes highly plausible when we think about what is involved in entertaining a proposition. It seems that there are exactly four ways in which a being might wind up with infinitely many entertainable propositions: (i) one might be in a position to refer to infinitely many things, (ii) one might be able to ascribe infinitely many predicates, (iii) one might be able to grasp infinitely many propositional structures, or (iv) one might have thoughts or sentences with context-sensitive semantics, and one might have infinitely many available contexts, such that a thought or sentence with a given character could be mapped to infinitely many propositions.12 Let us examine these possibilities in turn.

11.4.1 Infinite Terms First, if there were infinitely many things to which a person could refer, then there might be infinitely many propositions that that person could entertain. But it is hard to see how a person could ever be in a position to refer to each of infinitely many objects. As discussed above, one’s perceptual experiences from birth up to the present can only position one to refer to finitely many things. We can also refer to some further, imperceptible things, such as numbers, universals, and other abstract objects, using abstract concepts. But a human being can have only finitely many concepts at any given time, so this only finitely extends the number of things one can refer to. Concepts can also be combined to produce new concepts, thus further extending the number of abstract objects to which one can refer. For instance, given the concepts RED and TRIANGULAR, and the operation of disjunction, I can form the concept RED OR TRIANGULAR, and thus refer to the property (if such there be13) of being either red or triangular. But there is still a limit to the complex concepts that I can grasp: too complex, and I can no longer understand them. So again, this only finitely extends the range of things to which I can refer. But perhaps, as Klein suggests, I have the ability to form new basic (simple) concepts.14 Is there any plausible way in which this ability might be unlimited? Not

12

I here make use of the character/content distinction, as explained in Kaplan (1989, 500–507). Armstrong (1978, 20) rejects such arbitrary disjunctive properties. 14 Klein (1999, 308–309) suggests that “when our vocabulary and concepts fall short of being able to provide reasons, we can develop new concepts.” I assume that he has in mind new simple concepts, since the possibility of new complex concepts constructed out of concepts one already possesses would not help. 13

11

Finite Minds

181

without my acquiring new evidence. Thus, suppose I were to experience some new quale15 that I have never before experienced—say, the sensation of a fourth primary color. Then I would form one or more new basic concepts. But the propositions that those new concepts would enable me to entertain are not propositions now available to me in any epistemically relevant sense. They do not count as reasons of a sort that might justify any of my present beliefs. Another way of forming new basic concepts might be to notice new patterns among the things one has previously been aware of (it is debatable whether such concepts would be basic, but let’s not worry about that). But since there are only finitely many such things and one starts with a limited ability to discriminate properties of those objects, it is unclear how an unlimited number of new patterns could be apprehended. Surely there could not be an unlimited number of simple patterns in such finite data. Perhaps infinitely many patterns may exist in some sense, if one allows patterns of increasing complexity—but then a point would come at which patterns would be too complex for one to perceive or understand. In the purely abstract realm, mathematicians have devised many new concepts during the last few centuries—the concepts of quaternions, hyperreal numbers, tensors, and so on. This might give the infinitist hope. But notice that these new concepts are increasingly abstract, complex, and in general difficult to understand. Most objects in advanced mathematics are already incomprehensible to most people. While mathematicians are still able to understand their creations/discoveries, there must be limits to even a professional mathematician’s ability to grasp complex abstractions. It therefore does not seem that infinitely many concepts are graspable by any human being, even a mathematician. Furthermore, even if the range of concepts in principle graspable by a human being were unlimited, it would still be implausible to hold that this makes an unlimited range of propositions now available to us in an epistemically relevant sense. For example, it is not plausible that the Continuum Hypothesis (CH) was psychologically available to Aristotle. Aristotle had the cognitive capacities in principle to grasp the CH, but the concepts required to grasp it were not to be developed for 2,000 years. Those concepts are sufficiently far from any concepts that he actually possessed, and the path to forming them is sufficiently long and complicated, that no view on the CH (or its negation) can be something that Aristotle had available to him as a justifying reason for anything. If Aristotle had any beliefs whose justification would have required citing the proposition that the cardinality of the continuum is ℵ1, then those beliefs would have to have been unjustified (even if in fact CH is true, and even if CH is somehow provable). The fact that in principle, under some radically different circumstances, Aristotle could have understood the CH is irrelevant here. The point here is that propositions that would need to be grasped using concepts fundamentally different from one’s actual concepts are propositions that are not presently available to one.

Qualia (singular: “quale”) are the intrinsic, qualitative features of experience, the features that define what the experience feels like to the person having it. 15

182

M. Huemer

We need not settle what exactly is required for a proposition to be available. We need not settle, in particular, how far a concept may be from one’s current conceptual scheme for the propositions that concept would enable one to grasp to now be “available” to one. It is enough to say that there are some limits—at some point, the concepts are too advanced. That is what is illustrated by the Aristotle/CH example. This renders it extremely implausible that any human being has infinitely many concepts available in the relevant sense.

11.4.2 Infinite Predicates Having addressed the first putative way in which a person might have infinitely many entertainable propositions, the remaining three possibilities will be easier to address. The second possibility was this: if there were infinitely many predicates that a person could ascribe, then there might be infinitely many entertainable propositions. But again, it is hard to see how one could have infinitely many predicates available. The same considerations apply here as in the preceding discussion—the finite discriminatory capacity of the senses, the finite number of concepts possessed by any human being, the finite level of complexity that any given human being can deal with, and so on.

11.4.3 Infinite Structures If there were infinitely many propositional structures that one could grasp, then one could have infinitely many entertainable propositions. For instance, given a grasp of some proposition, A, and an understanding of negation, perhaps one is in a position to understand the propositions A, ~A, ~~A, and so on. Again, this is not plausible. There is no apparent way of multiplying propositional structures without end without resorting to ever more complex structures. At some point, the propositional structures will be too complex for a human mind to grasp or even refer to, in the same way that very large numbers cannot be referred to. After that point, propositions having those structures are not available to one.

11.4.4 Infinite Contexts Finally, if one had thoughts or sentences with context-sensitive semantics and infinitely many contexts were available to one, then there might be infinitely many entertainable propositions. Is this plausible? Keep in mind our constraint on available reasons: for a reason to be available in the epistemically relevant sense, it must be available to one without one’s needing to

11

Finite Minds

183

acquire new evidence; if citing reason r requires gaining new evidence, then r is not now available. Now suppose some new context positions me to cite new propositions as reasons that I could not cite before. This new context might, for example, involve my encountering new objects or observing new things happening to objects previously encountered. This new context is either subjectively distinguishable by me from all the previous contexts I have experienced, or it is subjectively indistinguishable from one or more previous contexts. If the new context is subjectively distinguishable, then it introduces new evidence. For example, if I encounter a new object that I can tell is distinct from any object I have previously encountered, then I thereby have new evidence; if I observe a property of an object that I can tell is distinct from any property of that object that I have previously observed, then I likewise acquire new evidence. As a result, the reasons that the new context makes me able to cite count as newly available reasons; they are not reasons that were available to me before the new context arrived. If, on the other hand, the new context is subjectively indistinguishable by me from some earlier context, then it does not position me to cite new reasons. It won’t enable me to have any thoughts that are subjectively distinguishable, by me, from thoughts I could have had earlier. By hypothesis, due to context-sensitive semantics, I am able to have thoughts that express distinct propositions from any propositions I could previously have expressed, but since these thoughts are subjectively indistinguishable by me from some thoughts I could have had earlier, I will be unable to tell that I am expressing new propositions. And this means that I cannot rationally rely on the new thoughts in any context in which I could not rely on the old thoughts. To illustrate, suppose that I encounter a bird, B1, which I observe to be red. “That is red,” I think to myself, instantiating thought token T1. Later, I encounter bird B2, which I also observe to be red. “That is red,” I think, instantiating thought token T2. Now, assume that, even though B1 and B2 are in reality distinct birds, my experiences of them are subjectively indistinguishable from each other. I cannot tell B1 and B2 apart; I cannot even tell that there were two distinct objects. (If I could at least tell that there were two distinct objects, then the second experience would have given me new evidence.) Let’s grant that T1 and T2 express distinct propositions because the demonstrative “that” picks out different things in the two thoughts. Nevertheless, I have no way of knowing that T1 and T2 express distinct propositions. In this case, it seems to me, I could not deploy T2 as a reason in support of any belief for which I could not have deployed T1. If, in a certain context, deploying T1 as a reason would have been question-begging, irrelevant, or otherwise fallacious, then citing T2 must appear equally question-begging, fallacious, or irrelevant to me, since I can’t tell that T2 is different from T1. In that case (appealing now to an internalist intuition), I think that I cannot rationally appeal to T2. Therefore, the arrival of T2 does not really enable me to extend the regress of reasons. In short, the possibility of infinitely many distinct reference-determining contexts does not make infinitely many reasons now available. The contexts that one has not in fact experienced but could experience either (i) would introduce new evidence, in which case the reasons they would make available do not count as presently available, or (ii) would not introduce new evidence, in which case they could not be used to extend the regress of reasons.

184

11.5

M. Huemer

Absolute versus Conditional Availability

The infinitist posits an infinite chain of reasons that are all in some sense available to the subject. There are two ways of understanding this availability. On the first understanding, each proposition in the series must be available to the subject at time t, if the subject is to have a justified belief at t (call this condition “absolute availability”). For example, for me now to be justified in believing that squirrels are furry, I must now have available to me the 1000th proposition in the series of reasons standing behind the proposition [squirrels are furry]. On the second understanding, each proposition in the series (including the initial belief whose justification is provided for by the series of reasons) must be such that if that proposition were to be questioned, the succeeding proposition in the series would then be available (call this “conditional availability”). Thus, for me now to be justified in believing that squirrels are furry, it must be the case that if [squirrels are furry] were called into question, I would then have the reason for [squirrels are furry] (whatever that reason is) available to me; and if that reason were questioned, I would then have its reason available to me, and so on. I need not now have the 1000th reason in the series available, but I must be such that, if the first 999 reasons had been questioned, I would then have the 1000th reason available. Klein has indicated a preference for the latter understanding of availability, which makes the conditions for justification less demanding.16 Does this affect the preceding arguments? I believe not. The sort of cognitive limitations I have appealed to would affect the conditional availability of propositions as much as their absolute availability. Consider, for example, the sort of mathematical propositions that I am unable to entertain due to the difficulty of referring to very large numbers. I would not be rendered able to entertain these propositions by my having previously entertained other propositions that they could serve as reasons for. More generally, as the regress of reasons continues, the propositions it involves must grow ever more complex (or become indistinguishable, as discussed in Sect. 11.4.4 above). At some point, this complexity puts the propositions beyond my ability to grasp. My having entertained (and called into question) less complex members of the series would not render me able to grasp propositions of unlimited complexity. You might suppose that my practice in entertaining less complicated propositions might somewhat expand the range of propositions I could grasp—but there is a limit to how far this cognitive expansion could go. No amount of practice, for example, would enable me to grasp a proposition whose most perspicuous written expression requires a million pages of text.

Klein communicated this during an exchange at the conference where this paper was first presented, at Rutgers University, April 29, 2016. 16

11

Finite Minds

11.6

185

The Indeterminacy Problem

The preceding discussion has proceeded as though there were some particular number, n, that is the answer to “How many beliefs do you have?” or “How many reasons are available to you?” But one might think there is no such number. There are two reasons one might think this. First, “available” is a vague predicate, in the sense that it has borderline cases. On some reasonable views of vagueness, vagueness entails indeterminacy: there is no fact of the matter as to whether a borderline case falls inside or outside of the category. Other views reject this, holding that the correct classification of borderline cases is merely unknown to us (Williamson 1994). To make things interesting, suppose we take the indeterminacy view. Then the question “How many F’s are there?” may lack a determinate answer when “F” is vague. For this reason, “How many reasons are available to us?” lacks a determinate answer. Second, there may be no definite answer to how many beliefs we have. Consider that one way of dispositionally believing something is to have an appearance that one is disposed to take at face value—the things that thus appear to one to be the case may thereby qualify as dispositionally believed. But there is no definite answer to how many things appear to one to be the case. Consider your present visual experience. How many things does it represent to be the case? One could report the approximate content of one’s visual experience using a large number of sentences, but there is no uniquely correct way of doing this, no unique way of dividing up the experience into discrete pieces of information. So there is no particular number that is the answer to how many dispositional beliefs one has by virtue of one’s having a normal visual experience. Similar arguments can be made regarding other kinds of belief. All of this complicates the treatment of the question, “How many reasons do you have?” I obviously cannot provide a definite natural number to answer that question, and there may not be any such number. This, however, does not preclude our endorsing the finite mind objection. The reason is that the finite mind objection requires only that the answer to “How many reasons do you have?” is definitely not “infinity.” And that is surely the case, regardless of the above observations. Even if we regard the boundaries of the category of “available” reasons as indeterminate, they are indeterminate only within a certain range, and that range includes only finite numbers of propositions. There is no reasonable interpretation of “available” that makes infinitely many propositions available. To illustrate, consider philosophers’ favorite vague predicate, “bald.” There is (according to the view of vagueness we’re presently entertaining) no determinate answer to “How many hairs must one have on one’s head in order not to be bald?” However, there is a determinate answer to “Must one have infinitely many hairs on one’s head in order not to be bald?” The answer to that is definitely “No.” Similarly, though there is no definite answer to how many reasons are available to you, the answer to “Are there infinitely many reasons available to you?” is definitely “No.” Of course, that claim is less obvious than the corresponding claim about baldness . . . but that is why I have produced the arguments of this chapter.

186

M. Huemer

Likewise, though there is no determinate answer to “How many things does your visual experience represent as being the case?”, the answer is definitely not “infinity.” Although there is no single, privileged way of dividing up the content of a visual experience into discrete propositions, there is no admissible way of dividing it up into infinitely many propositions, given the facts about human beings’ limited capacity for making visual discriminations. But perhaps we should reject the whole notion of counting the number of pieces of information contained in a visual experience. We might then likewise reject the idea of counting the number of beliefs a person has. We might say an individual’s mind contains a mass of information, but this information does not consist of discrete atoms. As a related point, perhaps (some or all of) the mind’s information is in analog rather than digital form. This is an interesting suggestion, but note that it will not help the infinitist. The infinitist holds that a belief is justified (at least in part) by virtue of its being at one end of a series containing infinitely many available reasons. This requires a discrete conception of the information available to one—in what sense could one speak of an infinite series of propositions, unless one believed that there were in fact discrete propositions available? How many reasons, then, do we have? It is not clear that any particular finite number is the correct answer to this question; it is not even clear that “finitely many” appropriately answers the question. But it is clear that “infinitely many” is not the correct answer. That is enough to sustain the finite mind objection to infinitism.17

References Aiken, S. (2014). Knowing better, cognitive command, and epistemic infinitism. In J. Turri & P. D. Klein (Eds.), Ad Infinitum (pp. 18–36). Oxford/New York: Oxford University Press. Armstrong, D. (1978). A theory of universals. Cambridge: Cambridge University Press. Audi, R. (1993). The structure of justification. New York: Cambridge University Press. Fumerton, R. (2014). Infinitism. In J. Turri & P. D. Klein (Eds.), Ad Infinitum (pp. 75–86). Oxford/New York: Oxford University Press. Gardner, M. (1977). Mathematical games. Scientific American, 237(5), 18–28. Huemer, M. (2001). Skepticism and the veil of perception. Lanham, MD: Rowman & Littlefield. Huemer, M. (2016). Approaching infinity. New York: Palgrave Macmillan. Kaplan, D. (1989). Demonstratives. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press. Klein, P. D. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13 (Epistemology), 297–325. Klein, P. D. (2005). Infinitism is the solution to the regress problem. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 131–140). Malden, MA: Blackwell.

17

I would like to thank Peter Klein, Cherie Braden, and the participants of the conference honoring Peter Klein at Rutgers University in April, 2016, for their many thoughtful and helpful comments on this chapter. Peter Klein is not responsible for any errors in this chapter. He and I have agreed that all responsibility for any mistakes falls on Cherie Braden.

11

Finite Minds

187

Klein, P. D. (2007). Human knowledge and the infinite progress of reasoning. Philosophical Studies, 134(1), 1–17. Kvanvig, J. (2014). Infinitist justification and proper basing. In J. Turri & P. D. Klein (Eds.), Ad Infinitum (pp. 125–142). Oxford/New York: Oxford University Press. Plato. 2012. Meno (Trans. Cathal Woods). Available at http: //ssrn.com/abstract¼1910945. Accessed 5 October 2016. Podlaskowski, A., & Smith, J. (2011). Infinitism and epistemic normativity. Synthese, 178, 515–527. Turri, J. (2009). An infinitist account of doxastic justification. Dialectica, 63(2), 209–218. Turri, J., & Klein, P. (Eds.). (2014). Ad infinitum: New essays on epistemological infinitism. Oxford: Oxford University Press. Williams, J. (1981). Justified belief and the infinite regress argument. American Philosophical Quarterly, 18(1), 85–88. Williamson, T. (1994). Vagueness. London: Routledge.

Chapter 12

Finite Minds and Open Minds Jeanne Peijnenburg and David Atkinson

Abstract One of the most persistent complaints about Peter Klein’s infinitism involves the finite mind objection: given that we are finite, how can we ever handle an infinite series of reasons? Klein’s answer has been that we need not actually produce an infinite series; it is enough that such a series be available to us. In this chapter, a different reply is presented through the reconstruction of epistemic justification as a trade-off. In acting as responsible agents, we are striking a balance between the number of reasons that we can handle and the level of precision that we want our beliefs to have. If we are unable or unwilling to manage a large number of reasons, then we have to pay the price in terms of justificatory inexactitude and thereby of accepting relatively untrustworthy beliefs. As well as being intuitively attractive, this idea of a trade-off is warranted by the mathematics of epistemic justification, understood as involving probabilistic relations. Keywords Finite mind · Infinitism · Inference chains · Justification · Structure of justification · Probabilistic justification · Degrees of justification · Klein · Epistemology

When Peter Klein first gave a talk on what later became known as infinitism, he was rudely interrupted. “You are kidding, aren’t you?”, chimed in one of the listeners, who, to judge from subsequent approving chuckles, was clearly not alone in his discombobulation. Fortunately, Peter belongs to the kind that is challenged rather than deterred by such a reaction. Questioning received wisdom not only appeals to his open and unorthodox mind, but it also seems to tickle his well-developed feel for the absurd. At any rate, he unflinchingly continued his thoughts on the subject and developed them into an epistemological program that keeps generating fruitful and inspiring discussions.

J. Peijnenburg (*) · D. Atkinson University of Groningen, Groningen, The Netherlands e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_12

189

190

J. Peijnenburg and D. Atkinson

A frequently raised complaint about Klein’s infinitist program is the age-old finite mind objection. Given that humans are finite, how can they forever go on citing reasons for reasons for what they believe? In his early papers, Klein argued that the finite mind objection is based on the “Completion Requirement,” according to which a belief can only be called justified if the agent has actually completed the process of reasoning. This requirement, says Klein, is contrary to the spirit of infinitism, and moreover it is too demanding: Of course, the infinitist cannot agree to [the Completion Requirement] because to do so would be tantamount to rejecting infinitism. More importantly, the infinitist should not agree because the Completion Argument demands more than what is required to have a justified belief. (Klein 1998, 920; Cf. Klein 1999, 314)

Klein sees epistemic justification as being essentially incomplete; it is provisional at heart and always open to further improvement. In later work, he has fleshed this out by means of two distinctions: that between propositional and doxastic justification, and that between objective and subjective availability. Propositional justification depends on the objective availability of reasons in an endless chain, where objective availability means that one proposition is a reason for another, even if we are not aware of it. Doxastic justification, on the other hand, hinges on an availability that is subjective: the agent must be able to actually “call on” a reason in the endless chain.1 Although in its entirety the chain can never be subjectively available to a finite mind, the agent can take a few steps on the infinite trajectory. How many steps are taken, or need to be taken, is a pragmatic matter and depends on the context: We don’t have to traverse infinitely many steps on the endless path of reasons. There just must be such a path and we have to traverse as many as are contextually required. (Klein 2007, 13)

We sympathize with this view, but tend to approach the subject in a somewhat different way. Where Klein denies that infinite epistemic chains can be completed, we assert that there is a sense in which they can. Moreover, we have a different opinion about what it means that justification is context dependent. As Klein sees it, we follow a path of reasons and stop at a point where a reason is sufficiently obvious or very likely to be true. Or in the words of Nicholas Rescher, In any given context of deliberation, the regress of reasons ultimately runs out into ‘perfectly clear’ considerations which are (contextually) so plain that there just is no point in going further . . . Enough is enough. (Rescher 2010, 47)

On our view, by contrast, the fact that a reason is sufficiently plain or clear or highly likely or even self-evident is irrelevant for any decision about stopping or continuing. What is relevant for such a decision is the size of the contribution that the reason makes to the probability of the belief that we aim to justify by means of an

1

For the distinction between objective and subjective availability, see Klein (1999, 299–300); Klein (2003, 722); Klein (2005, 136). For the difference between propositional and doxastic justification, see Klein (2007, 6–11).

12

Finite Minds and Open Minds

191

epistemic chain. If this contribution is small enough to be neglected, then we simply ignore the reason in question, no matter how high or low its probability is. These divergences from Klein stem from our framing epistemic justification as something that involves probabilistic relations, rather than, for example, entailments. Consider the following infinite epistemic chain: A0

A1

A2

...

We call A0 the target proposition or target belief (disregarding for the moment the distinction between propositional and doxastic justification). An An + 1 means that An + 1 is a reason for An, or that the (belief in) proposition An + 1 epistemically justifies the (belief in) proposition An. In earlier work we have argued that infinite epistemic chains are vicious if the relations between the propositions or beliefs are those of entailment. However, if justification is probabilistic, in the sense that An + 1 makes An more probable, then the chains are generally benign. In particular, the following two statements can be proven: 1. The target A0 may have a unique and non-zero probability, notwithstanding the fact that it is justified by an infinite chain. 2. The effect of distant propositions on the unique and non-zero probability of the target diminishes as their distance from the target increases, and an infinitely distant proposition has no effect at all. Claim 1 goes against the idea that the probability of the target in an infinite probabilistic chain must be either zero (sometimes known as the probability diminution argument) or indeterminate. Notable representatives of the former position are David Hume (1739/2000/2006) and C. I. Lewis (1929), while the position that it is indeterminate has been defended by, for example, Rescher (2010). Claim 1 of course also gainsays anyone who believes that a grounding proposition is needed for determining the probability of the target. In this sense it nullifies another prominent argument against infinitism, dubbed by Peter Klein ‘the no starting point objection’ (2000, 204).2 Claim 2 is particularly interesting for finite chains. Effectively, it states that the further away the grounding proposition is from the target, the smaller is its contribution to the latter’s probability. Applied to infinite chains, Claim 2 means that in the limit the impact of any grounding proposition will vanish completely.3 Together, the claims indicate how we differ from Peter Klein. The first claim means that a probabilistic chain can be completed, in that it yields a well-defined and positive probability value for the target proposition. The second one articulates the fact that the influence of a particular reason on the probability value of the target 2

Laurence BonJour, for example, raises this objection when he remarks that in an infinite chain “justification could never get started and hence no belief would be genuinely justified” (BonJour 1976, 282). Carl Ginet makes similar remarks, but uses the term “structural objection” (Ginet 2005). 3 In this chapter, we are talking about subjective probability, since we are dealing with beliefs, but in fact our formalism applies also to objective probability. It can, for example, be used in the analysis of causal chains, on condition that causality is interpreted probabilistically.

192

J. Peijnenburg and D. Atkinson

lessens as the distance between the target and this reason increases. At a certain point, the influence of the reason on the target will be small enough to be neglected. Where exactly that point is located depends on pragmatic considerations; but, as we will further explain below, we can make these considerations as precise as we wish. Elsewhere we have formally demonstrated Claims 1 and 2 on the basis of the probability calculus.4 Interestingly enough, it is precisely this formal demonstration that appears to have triggered the most resistance. Not that the demonstration itself is flawed—everybody appears to agree that it is not. Rather, the complaint is that the actual enterprise of justifying beliefs is not properly modelled by our formal approach. Even if one assumes, as most epistemologists do, that the phrase ‘An+1 justifies An’ implies that An + 1 makes An more probable, the consequences of the probability calculus may not be applied lock, stock, and barrel. Thus Jeremy Gwiazda has complained that what we call the completion of an infinite justificatory regress is in fact merely the computation of the limit of a convergent series (Gwiazda 2011). In a similar vein, Adam Podlaskowski and Joshua Smith have argued that although “valuable lessons” can be drawn from our results, such as 1 and 2, it is “entirely unclear” that these results meet a basic requirement, namely “providing an account of infinite chains of propositions qua reasons made available to agents” (Podlaskowski and Smith 2014, 212). Podlaskowski and Smith call this ‘the availability problem’: [A] demonstration that finite agents can actually calculate the probability of a proposition’s truth—even if it belongs to an infinite chain of reasons—does not thereby show that each reason is equally available to a finite agent.5 (Podlaskowski and Smith 2014, 216)

In a word, the criticism is that the finite mind objection still applies. The point is well taken, moreover it is one that is familiar. Jonathan Cohen famously argued that what he calls “Pascalian probability” (read: probability according to the calculus) is not particularly suited to reasoning in court, everyday life, or even science (Cohen 1977). Cohen’s argument about probability could equally well be applied to logic, and in fact has been applied that way. Historically, logical systems have often been criticized for being too formal and too far removed from actual human reasoning. This goes for mediaeval scholastic systems exploiting Aristotelian syllogistics as much as for modern forms of mathematical logic. In his valedictory lecture, Johan van Benthem rightfully reminded us that a heedful logician typically goes back and forth between normative and descriptive considerations: Logical theory that ignores actual behavior seems dangerously empty, lacking focus. On the other hand, I do not want to lose the potential of normative thinking either, that can help us improve performance, or design better ways of dealing with the world and with one another (van Benthem 2014, 18–19).

4

Peijnenburg (2007); Atkinson and Peijnenburg (2010); Peijnenburg and Atkinson (2014a, b). Atkinson and Peijnenburg (2017, Chapter 5, especially §5.3). 5 Michael Rescorla in this connection even uses the scare term “hyper-intellectualism” (Rescorla 2014).

12

Finite Minds and Open Minds

193

Probability is in this respect like logic, and so is epistemic justification: on the one hand, we ponder abstractly about how we ought to reason when justifying our beliefs, and on the other hand, we have to keep an eye on how we actually do go about finding reasons for what we believe. The fact itself that epistemic justification has a normative and a descriptive side is presumably uncontroversial. Controversies are rather about the question as to how much weight should be apportioned to each side. An approach like ours, which authorizes the completion of infinite epistemic chains, may seem to unduly stress the normative part. The above, however, insufficiently takes into account the consequences our view has for everyday reasoning and for finite chains. Of course it is true that people cannot forever continue giving reasons for their beliefs—most of us already lose track after three or four steps. The point is that our approach appertains to ordinary finite chains as well, since it enables us to determine how long an epistemic chain needs to be, even without any knowledge of a foundational proposition. Here is how it works. Imagine the shortest chain there is, a belief A0 is justified by A1: A0

A1

In this finite chain, A0 is the target and A1 is the ground. The arrow is interpreted as before, namely as implying, as a necessary but not sufficient condition, that A1 makes A0 more probable. It is important to realize that the unconditional probability of A0, namely P(A0), is not solely a function of the conditional probability of A0 given A1, that is, of P(A0|A1). In determining the unconditional probability of A0 we must also take into account what this probability would be if A1 were false: P(A0|ØA1). Now P(A0) must lie between P(A0|A1) and P(A0|ØA1). If neither of these conditional probabilities is zero, the unconditional probability of A0 cannot be zero either. Suppose that the value of P(A0|A1) is x and the value of P(A0|ØA1) is y, and let x and y differ greatly; for example x is very close to one and y is very close to zero. Assuming we don’t know the unconditional probability of A1, we face a great deal of uncertainty as regards the value of P(A0). The only thing we can be sure of is that this value lies somewhere in the wide interval between x and y. However, our results are of help here, for they imply that the interval shrinks as the chain is lengthened. To illustrate this, let us first add one link to the chain, by giving a reason for what originally was the ground A1: A0

A1

A2

Here A1 is justified by A2, which now does duty as the new grounding proposition. The unconditional probability of A1 must lie between the conditional probability of A1 given A2 and that of A1 given ØA2. If neither of these conditional probabilities is zero, then the unconditional probability of A1 cannot be zero. Importantly, this has the effect of further restricting the interval in which the unconditional probability of A0 must lie: it will now be strictly smaller than it was in the absence of A2. A similar story would apply if we were to expand the series further, by adding A3 as a new

194

J. Peijnenburg and D. Atkinson

ground. The probability of A3 would restrict the domain in which the probability of A2 may lie, which in turn diminishes the interval in which the probability of A1 may lie, which thereby further narrows down the interval for P(A0). The more propositions there are, the smaller is the interval within which the probability of A0 must lie. In the limit of an infinite probabilistic regress, this interval has shrunk to a point. The probability of A0 has then been determined uniquely in terms of all the conditional probabilities along the chain. All of this means that one can determine in advance how many reasons an agent needs in order to approach the true probability of the target within a given error margin. If this number of reasons happens to be too large to fit into the agent’s finite mind, then she will have to relax the level and be content with a degree of justification that is less accurate. But if the number of reasons is rather small, so that they are mentally encompassed with ease, then the satisfaction level can always be tightened up and brought closer to the target’s true probability. Epistemic justification thus boils down to striking a balance. In acting as responsible epistemic agents, we are instigating a trade-off between the number of reasons that we can handle and the level of accuracy that we want to reach. If we are unable or unwilling to manage a large number of reasons, we have to pay in terms of a lack of precision and hence of trustworthiness of (our belief in) proposition A0. Taking the short route thus comes at a price, but in situations where precision is not important, we can take it easy and should do so on pain of exerting ourselves unnecessarily. The point is a general one, and it can be made in qualitative or quantitative terms. In the latter case, we may choose between using precise numbers or intervals possibly with vague boundaries. Here is an example with precise numbers. Suppose we have taken a few steps in the chain and then stopped. We first set the probability of the ground equal to one and then to zero. Assuming we know the values of the conditional probabilities, we can now determine the minimum and the maximum value of the probability of A0, let them be 0.65 and 0.67 respectively. Then our best estimate of P(A0) will lie in the middle, at 0.66, for in that manner we have minimized the possible error that this estimate can have, to wit 0.01. The unknown true limiting probability of A0, whatever its precise value is, cannot deviate from 0.66 by more than 0.01, since we know for a fact that it lies somewhere between 0.65 and 0.67. If we proclaim ourselves satisfied with a number that deviates by no more than 0.01 from the true value, then we need go no further in inquiring as to any support that the target might have beyond the minimal required to reach this error of 0.01. This is because any extension of the chain, obtained by adding a proposition that supports the erstwhile ground, would only decrease the error. So in this case we know exactly how many reasons we need in order to approach the true value of the target to a level that satisfies us. Since we are content with a value that deviates no more than 0.01 from the true value, we require no more than the reasons we already have. And if our mind is capacious enough to store these reasons, then we have accomplished our task: we have justified A0 to a satisfactory level, staying neatly within the limitations of our finite mind.

12

Finite Minds and Open Minds

195

Why can we be so sure that any extension of the chain will always decrease the error margin surrounding the probability of A0? How do we know that the margin will never widen, or that no fluctuations will occur further up in the chain? The answer is that these features follow directly from the fact, articulated in our Claim 2, that distant reasons are less important than those that are nearby.6 The structure of the probabilistic epistemic chain is such that it enables us to say how many reasons we need to call on in order to approach the probability of the target to a satisfactory level. To do that, we do not need to know the length of the chain; we need not even know whether it is finite or infinite. Nor do we have to know the probability of the ground. The only thing we need are the (precise or imprecise) values of a certain number of conditional probabilities (sometimes more, sometimes fewer, depending on the speed of convergence) that suffice to take us to within a desired level of accuracy with respect to the true, but unknown probability of the target. Once we are there, we can safely ignore the rest of the chain.

References Atkinson, D., & Peijnenburg, J. (2010). The solvability of probabilistic regresses. A reply to Frederik Herzberg. Studia Logica, 94, 347–353. Atkinson, D., & Peijnenburg, J. (2017). Fading foundations: Probability and the regress problem. Dordrecht: Springer Open Access: https: //link.springer.com/book/10.1007%2F978-3-31958295-5. Bonjour, L. (1976). The coherence theory of empirical knowledge. Philosophical Studies, 30, 281–312. Cohen, L. J. (1977). The probable and the provable. Oxford: Oxford University Press. Ginet, C. (2005). Infinitism is not the solution to the regress problem. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 140–149). Malden: Blackwell. Gwiazda, J. (2011). Infinitism, completability, and computability: Reply to Peijnenburg. Mind, 118, 1123–1124. Hume, D. (1739/2000/2006). In Norton, D. F & Norton, M. J, (eds.), A treatise of human nature. Oxford: Oxford University Press. Klein, P. D. (1998). Foundationalism and the infinite regress of reasons. Philosophy and Phenomenological Research, 58(4), 919–925. Klein, P. D. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13(Epistemology), 297–325. Klein, P. D. (2003). When infinite regresses are Not vicious. Philosophy and Phenomenological Research, 66(3), 718–729. Klein, P. D. (2005). Infinitism is the solution to the regress problem. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 131–140). Oxford: Blackwell. Klein, P. D. (2007). Human knowledge and the infinite progress of reasoning. Philosophical Studies, 134(1), 1–17. Lewis, C. I. (1929). Mind and the world-order. An outline of a theory of knowledge. New York: C. Scribner’s Sons. Peijnenburg, J. (2007). Infinitism regained. Mind, 116, 597–602.

6

For details see Atkinson and Peijnenburg (2017).

196

J. Peijnenburg and D. Atkinson

Peijnenburg, J., & Atkinson, D. (2014a). Can an infinite regress justify anything? In J. Turri & P. D. Klein (Eds.), Ad Infinitum: New essays on epistemological infinitism. (pp. 162–179). Oxford: Oxford University Press. Peijnenburg, J., & Atkinson, D. (2014b). The need for justification. Metaphilosophy, 45, 201–210. Podlaskowski, A. C., & Smith, J. A. (2014). Probabilistic regresses and the availability problem for Infinitism. Metaphilosophy, 45, 211–220. Rescher, N. (2010). Infinite regress. The theory and history of varieties of change. New Brunswick: Transaction Publishers. Rescorla, M. (2014). Can perception hold the regress of justifications? In J. Turri & P. D. Klein (Eds.), Ad Infinitum: New essays on epistemological infinitism. (pp. 179–200). Oxford: Oxford University Press. Van Benthem, J. (2014). Fanning the flames of reason. Valedictory lecture delivered on the occasion of his retirement as University Professor at the University of Amsterdam on 26 September 2014.

Chapter 13

Some Notes on the Possibility of Foundationalist Justification Sanford C. Goldberg

Abstract In a good deal of his work on epistemic justification and the regress problem, Peter Klein has argued that foundationalist replies cannot work, as they suffer from an in-principle inability to reply to the regress problem. In this chapter, I argue that his argument begs the question. To do so, I do not argue that foundationalism is true; only that there are versions which, on some arguable (but not uncontroversial) assumptions, would constitute an adequate response to the regress problem. The conclusion will be that if foundationalist replies are unacceptable, it is not because they can’t respond to the problem, as Klein alleges. The significance of my argument lies in its meta-epistemological orientation: I will be arguing for the possibility that there is a kind of doxastic (epistemic) responsibility that is not a matter of having adequate reasons. If this claim is plausible, it opens up the prospects for a certain kind of (reliabilist and socially-oriented) foundationalism. The irony should not be lost on us: externalist accounts of justification are standardly thought to run afoul of considerations of epistemic responsibility, whereas I will be suggesting that they are a core part of our best hope to be able to make sense of epistemic responsibility. Keywords Foundationalism · Foundationalist justification · Epistemic responsibility · Externalist epistemology · Regress problem · Klein · Reliabilist foundationalism · Structure of justification · Epistemology

With thanks to Cherie Braden, for very helpful comments on an earlier draft; and to Peter Klein and the members of the audience at Rutgers’s one-day conference in his honor in May 2015, at which I presented an earlier version of this paper. S. C. Goldberg (*) Department of Philosophy, Northwestern University, Evanston, IL, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_13

197

198

13.1

S. C. Goldberg

Introduction: Klein’s case against Foundationalism

In various papers, Peter Klein has argued that foundationalism cannot answer the “infinite regress of reasons” problem. His claim is not merely that the attempt fails— as would be the case if he allowed that some form of foundationalism might succeed (although none to date has done so). Rather, his claim is a much stronger one: by their very form, accounts of this kind are incapable of succeeding. This is a strong claim, and in this section I review his reasons for making it. By Klein’s lights, the “infinite regress of reasons” problem arises in the attempt to provide “an acceptable account of rational belief” by characterizing what he calls the “structure of justificatory reasons” (1998, 297). In this he follows the ancient skeptics themselves, who had noted that a problem arises in the attempt to characterize this structure. Klein himself brings out the source of the problem in terms of what it takes to vindicate the justification of a subject’s belief. Doing so is, at least in part, a matter of the subject’s providing reasons in support of her belief. However (Klein notes), the cited reasons serve to justify a belief, or render it rational, only insofar as they are good reasons. Since a reason is good only if the proposition that constitutes it is (likely to be) true,1 we then need a reason (r2) to think that the original reason (r1) is (likely to be) true. But of course r2 is itself a reason, and so has to be certified as a good reason. And of course the same can be asked of the reason (r3) for regarding r2 itself as a good reason, and we are off on the regress. (Hence the “infinite regress of reasons.”) This is a problem since it is not clear whether this task can be adequately addressed, and so it is not clear how to vindicate the rationality or justification of any particular belief. Klein also follows the ancients in thinking that, skepticism aside, there are three main types of reply to this line of reasoning, corresponding to which are the three main accounts of the structure of justification. One reply is to argue that a certain class of reasons is “foundational,” and so is not in need of certification by other reasons. Corresponding to this approach is the foundationalist account of the structure of justificatory reasons. A second reply is to maintain that the line of reasons themselves ultimately turns back on itself, as one and the same reason will recur in the course of the subject’s attempt to justify her belief. On this view the structure of reasons itself forms something like a circle; this corresponds to the coherentist account of the structure of justificatory reasons. A final reply is to claim that “the reasons that justify a belief are members of a chain (perhaps branching) that is infinitely long and non-repeating” (Klein 1998, 919). Corresponding to this third approach is the infinitist account of the structure of justificatory reasons, which is the one Klein himself defends (see 2003, 2007a, b).2

1 In using the part in parentheses—(likely to be)—I am following Klein’s own formulation. See Klein (1998, 924). 2 More recently, Klein has allowed that there can be a “rapprochement” between foundationalism and infinitism (Klein 2014). However, his proposal there appeals to a contextualist account of ascriptions of justification; I will ignore this suggestion in this chapter, as I aim to show that there is a defensible version of foundationalism which does not depend on any contextualism about justification.

13

Some Notes on the Possibility of Foundationalist Justification

199

On various different occasions, Klein has spelled out why he thinks foundationalist approaches simply can’t be made to work as a reply to the regressof-reasons problem. Here is what Klein said against Richard Fumerton’s very traditional sort of foundationalism (developed in Fumerton 1995): Suppose that some proposition, say F, is offered as a putative foundational one. . . . Either F has some characteristic which makes it such that it is (highly likely to be) true or it doesn’t. If it does, then the possession of that characteristic can be and should be offered as a reason for thinking that it is true—and the regress continues. If it doesn’t, then there is nothing that distinguishes it from non-foundational propositions and it becomes arbitrary to treat it as foundational. (Klein 1998, 924)

Klein’s point here, of course, is a perfectly general one that holds against any and all forms of foundationalism. And so, attempting to summarize “what IS wrong with foundationalism” in a more recent paper, Klein writes that even if foundationalism is true [as an account of the structure of justificatory reasons], it cannot solve the regress of reasons problem. . . . [O]nce the regress of reasons begins, a selfconscious foundationalist is forced into a form of unacceptable arbitrariness in her reasoning. (Klein 2004, 167; emphasis added)

Klein articulates the sort of arbitrariness he has in mind in the form of a principle, which in earlier work he had called the Principle of Avoiding Arbitrariness: PAA: For all x, if a person, S, has a justification for x, then there is some reason, r1, available to S for x; and there is some reason r2, available to S for r1; etc. (Klein 1999, 299).3

The in-principle problem with foundationalism, then, is that by the lights of PAA, the foundations themselves will inevitably be arbitrary. When it comes to the impossibility of a foundationalist reply to the regress-ofreasons problem, it should be clear that PAA itself is the heart of the issue: if PAA stands, foundationalist replies to the regress-of-reasons problem will be unacceptable in principle—which is to say that no such reply can be made to work, and Klein’s strong claim about the impossibility of foundationalist replies to the regressof-reasons problem will have been vindicated. But what is the rationale behind PAA itself? As best I can tell, Klein thinks that it is only by satisfying PAA that one can avoid the charge of stopping the regress at an arbitrary place. This reflects what the reasoning cited above tries to capture: a given reason is an acceptable place to stop the regress only if the reason is good, and so only if the proposition constituting that reason is (likely to be) true4; and since for any candidate reason we can ask whether it

3

It is PAA, together with a principle he calls the Principle of Avoiding Circularity PAC: For all x, if a person, S, has a justification for x, then for all y, if y is in the evidential ancestry of x for S, then x is not in the evidential ancestry of y for S. (Klein 1999, 298) that leads Klein to conclude that the infinitist account must be correct. 4 Insofar as the allegedly foundational reason r is itself a reason supporting some other (non-foundational) belief b, r must render the truth of b likely; I take it that this is something Klein himself builds in to the notion of something’s being a reason for b. But insofar as foundational reasons themselves do not depend for their epistemic status on being supported by other reasons, they must have some property—something other than being supported by

200

S. C. Goldberg

fulfills this condition, for any candidate reason we need yet another reason to think that it does fulfill this condition; in which case the allegedly foundational reason is no such thing. In short, PAA appears to be advanced as a requirement on avoiding one horn of a dilemma that Klein would foist onto the proponent of foundationalism: allegedly foundational reasons are either arbitrary or non-foundational. They are arbitrary if PAA is unsatisfied, since in that case no reason will have been given for thinking that they satisfy the condition on being an acceptable place to stop the regress; and they are non-foundational if such a reason is given. While I think Klein is right to think that any foundationalist will have to face up to this dilemma, I think he is wrong to think that there is no acceptable response to it within the foundationalist picture. In what follows I will argue that there are ways of responding to this dilemma. In particular, I submit that there are non-arbitrary yet epistemically proper points at which to stop the regress of reasons. I will be arguing that Klein can resist this response only by begging one or another question against the foundationalist.

13.2

Constraints on an adequate Foundationalism: Truth-aptness and Entitlement

Ultimately I want to defend foundationalism against Klein’s contention that no version is capable of responding to the regress-of-reasons problem. To do this, I will be using what I regard as the strongest version of foundationalism: externalist, and in particular reliabilist, foundationalism. It is important to me to address the regress-of-reasons problem as Klein himself sees it, so it will be helpful to begin by characterizing what I take to be common ground: what it takes for a reason or a proposition to be properly foundational. Foundations are supposed to be regress-stopping propositions. To play this role, they are conceived to be propositions (i) which are prima facie justified for the subject and (ii) whose status as prima facie justified for the subject is independent of any other reasons that the subject has.5 How might there be such propositions? A plausible view (which I myself endorse) is this: there are certain belief-forming processes which confer such justification on those propositions which the subject comes to believe through those processes. (These are the so-called basic beliefforming processes.) Now, it is a central task of foundationalist epistemology to identify such processes and to offer a meta-epistemological defense of the hypothesis that these processes are basic in the relevant sense. But in advance of delving into such details (for which see Sect. 13.4), one point is worth underscoring at the outset: one of the motivations for externalist epistemology is that it offers the prospect for another reason—that renders them likely to be true. (I thank Cherie Braden for indicating the need for this clarification.) 5 We might add: they are ultima facie justified so long as the subject has no reasons to doubt the propositions in question.

13

Some Notes on the Possibility of Foundationalist Justification

201

the relevant sort of meta-epistemological defense. To put it crudely, one might think that part of what makes a process basic in the relevant sense is that it is a token of a reliable type of belief-forming process—that is, it is of a type that amounts to a reliable route to true beliefs. For in that case we might think that the reliability of the process-type is what confers prima facie justification onto the belief formed (or the proposition believed) through such a process. For his part, Klein is well aware of attempts to provide externalist accounts of foundational belief, and he thinks they are no advance over alternative accounts of foundational belief. On this score his exchange with Michael Bergmann (in Klein 2004) is telling. Bergmann himself had offered a detailed externalist (broadly proper functionalist) version of foundationalism. In discussing Bergmann’s proposal (developed at greatest length in Bergmann 2006), Klein allowed that there are foundational reasons. That is, he allowed (if only for the sake of argument) that there are reasons which enjoy what he calls “autonomous warrant.”6 What Klein questions is whether such (autonomous-warrant-enjoying) reasons can provide a foundation by appeal to which a subject can stop the regress of reasons. Klein denies this. On this score it is worth quoting him at some length: Once we start the reason-giving game can we achieve the intended purpose of providing reasons by stopping the regress with a proposition that has some autonomous warrant? Why does the reason-giving game begin? Presumably because Fred [the one who is trying to answer the challenge to his belief] accepts Sally’s challenge that some belief, p, requires a reason in order to have the appropriate kind of epistemic warrant—warrant required for Sally to be epistemically responsible in believing p to some degree higher than she would were she not to have a reason. . . . [D]oes the [mere] truth of X [some proposition ascribing to Fred’s belief a feature in virtue of which that belief is allegedly non-inferentially justified] give Sally (or Fred) any reason for thinking that b [the belief itself] is true? Fred has three (and only three) possible answers: Yes; No; I do not know. If he says that autonomously warranted propositions are likely to be true in virtue of having the property, F, then Sally can point out that he has a (very good) reason for thinking that [the belief] b is true and the regress has not stopped. If he says that basically warranted propositions are not likely to be true in virtue of being basically warranted, then Sally should point out that from her point of view Fred has not given her any reason to believe that p . . . [in which case] believing anything based upon b, including p, would be arbitrary. If Fred withholds judgement regarding the truth-conductivity of propositions that are basically warranted, then Sally should again point out that Fred’s reasoning has not provided any basis for believing that p is true. Perhaps . . . there is a good sense of “arbitrary” in which believing b is not arbitrary. After all, Fred is entitled to believe b . . . because b has F. True, but that will not help in resolving any misgivings about p’s truth unless one thinks that Fred’s entitlement to believe that b is connected to b’s truth. By way of analogy: Descartes saw that even though he was entitled to believe some proposition because it was clear and distinct, he had to show that clarity and

6

I take it that warrant for Klein is a positive epistemic status of some sort or other, and that it is “autonomous” when its presence does not depend on the presence of any other reason supporting the belief that enjoys this status. I should add that while Klein speaks of propositions as having such warrant, I find it more natural to speak of beliefs or reasons as having such warrant. I do not think that any confusion will result from this shift.

202

S. C. Goldberg

distinctness was truth conducive before he should accept that the proposition was true. . . . Descartes’ worry applies even if one is a modest foundationalist because the point of the why-game is to give Sally a basis for believing p. Ending with what from Sally’s point of view is a mere assertion of b and a claim that b is warranted in the absence of a reason cannot provide Sally with a reason for believing p unless being so warranted is a reason for Sally to believe that b is true. (Klein 2004, 169–171)

From the foregoing quotation it would seem that Klein’s main beef with Bergmann’s externalist foundationalism—and presumably his main beef with any version of externalist foundationalism—is the very problem he associated with all versions of foundationalism: the allegedly foundational (regress-stopping) reasons these foundationalist accounts postulate are either not really foundational at all (they don’t stop the regress), or they are arbitrary (and hence don’t amount to an epistemically appropriate reply to the regress-of-reasons problem). Now, I think that Klein ultimately begs the question against the externalist foundationalist. However, it will require some care in bringing this out in a clear way. Let me begin with a point of agreement, which concerns the nature a reason would have to have if it is to be properly foundational. A truly foundational reason, if such there be, would have to be both (likely to be) true by its very nature and such that the subject is entitled to regard it thus. In being (likely to be) true by its very nature—call this truth-aptness—such a reason would meet one of the conditions on providing the sort of epistemic support that is expected of the reasons to which we might appeal in the hope of rendering our belief rational or justified. In being such that the subject is entitled to regard it as truth-apt, such a reason would be an appropriate candidate to which the subject might appeal as she sought to respond to the regress-of-reasons challenge in a fashion that was epistemically responsible. Together, these two features of a reason—truth-aptness, and the subject’s entitlement to suppose as much—render the reason epistemically kosher. I submit—and on this point I think Klein can agree—that foundational reasons (if there are any) just are epistemically kosher reasons whose status as epistemically kosher does not depend on the subject’s having still further reasons. (I will call these reasons kosher-foundational, or k-foundational for short.) If they exist, k-foundational reasons can stop the regress of reasons. At the same time, if there are no k-foundational reasons, it would appear that there can be no foundationalist reply to the regress-of-reasons problem. As the foregoing definitions make plain, there are exactly two ways one can criticize a proposed foundational reason. One might criticize such a reason on the grounds that it is not truth-apt in the first place, and hence is not such as to provide the relevant sort of epistemic support for beliefs based on it. (Call this the truthaptness challenge.) Alternatively, one might criticize such a reason for being such that without further ado—e.g., without having in her possession still further reasons—the subject herself is not entitled to regard that reason as truth-apt. (Call this the entitlement challenge.) One can fail to meet the entitlement challenge because one lacks an entitlement to regard one’s reason as truth-apt in the first place; or, alternatively, one can fail to meet the entitlement challenge because, while

13

Some Notes on the Possibility of Foundationalist Justification

203

one may be so entitled, one’s entitlement depends on one’s having still further reasons (in which case one’s original reason fails to be k-foundational). Now on the face of it, it would seem that certain versions of foundationalism can address the entitlement challenge, but founder (or appear to founder) on the truthaptness challenge. A paradigmatic example is the one Klein himself mentions above: Cartesian foundationalism. The Cartesian construes foundational reasons as beliefs that are clear and distinct. Klein points out that if this is what foundational reasons amount to, it is far from clear that such reasons are truth-apt, and so it is far from clear that such reasons can meet the truth-aptness challenge. I think that Klein is right to say this. As an aside, I note that his point on this score does not depend on the subject’s own ability to cite reasons for thinking that clarity and distinctness are indications of (likely) truth; the point as I see it is that it is unclear whether clarity and distinctness are indications of (likely) truth in the first place. Even if a subject were to have good, albeit misleading, reasons to think that clarity and distinctness are indications of (likely) truth, she’d be wrong to think that her reasons did in fact meet the truth-aptness challenge. It seems to me that any thoroughgoingly internalist foundationalism is likely to face difficulties in connection with the truth-aptness challenge.7 But while the truth-aptness challenge bears against a foundationalism like Cartesian foundationalism (and arguably any internalist foundationalism), I would have thought that non-internalist versions of foundationalism can address that challenge; if they face a difficulty, it is in connection with the entitlement challenge. Consider in this light an externalist foundationalism. (Here I will move from Bergmann’s to my own favored view, a process reliabilist foundationalism.) According to such a view, a (reason-constituting) belief is foundational only if it is produced by a reliable process-type. Being produced by a reliable process-type is defined in such a way that any belief so produced is sufficiently likely to be true.8 So questions about its satisfaction of the truth-aptness challenge can be swiftly met: it does so by definition. Of course, this is not to say that beliefs/reasons that satisfy such a process-reliability criterion are invariably regress-stopping: on the contrary, one can still question whether the subject whose belief/reason fulfills this condition is entitled to assume as much. This, I take it, was the point of BonJour’s (1980) clairvoyant, Lehrer’s (1990/2015) Truetemp, Lackey’s (2007) case of the Denver Virus, and the problem of easy knowledge as raised by Fumerton (1995), Cohen (2002), and Vogel (2008). Since Bergmann’s foundationalism is an externalist foundationalism, it would seem that the best way to challenge this sort of foundationalism—in fact, the only way to challenge it—is to question whether there is any plausible account of the entitlement one can have for regarding one’s reason as satisfying a truth-aptness criterion, where the account does not require still additional reasons.

7

This claim ought to be qualified a little, albeit for reasons having nothing to do with my present argument. See Goldberg (2015) for a discussion. 8 We would have to add: when it is produced in normal circumstances by a properly functioning process. I disregard this here.

204

S. C. Goldberg

Unfortunately, Klein’s criticism above seems to level the truth-aptness challenge rather than the entitlement challenge. For while it appears as though he levels the entitlement challenge—after all, in the last paragraph of the long quotation cited above, he is explicitly levelling the allegation of arbitrariness—even so, the “analogy” that Klein uses to bring out the problem allegedly facing Bergmann’s account on this score is the analogy with Cartesian foundationalism. And Klein goes on to note that “Descartes’ worry [regarding truth-aptness] applies even if one is a modest foundationalist . . .”. But in response, I submit that, when it comes to externalist foundationalisms, the analogy with Descartes is simply not apt. On the contrary: if an externalist foundationalism is going to be convicted of arbitrariness (and hence of failing to meet the entitlement challenge), it should not be by appeal to any difficulty meeting the truth-aptness challenge. The claim must be, rather, that even if externalist foundations are truth-apt, the believing subject herself is not entitled to assume as much (unless, that is, she has reasons for so doing). Now it might be thought that this flaw in Klein’s argument is easily rectified: simply disregard his appeal to Descartes and the truth-aptness challenge, and read him instead as raising the entitlement challenge. Indeed, there is some textual evidence to think that this is what Klein had in mind in his criticism of Bergmann. Call a proposition metajustificational relative to S’s belief that p when it ascribes to that belief a feature in virtue of which that belief is non-inferentially justified. In the extended passage cited above, Klein does question whether “the [mere] truth” of some metajustificational proposition relative to Fred’s belief that p “give[s] Sally (or Fred) any reason for thinking that b [the belief] is true,” and this can make it seem as though Klein meant to raise the entitlement challenge. But here it is worth noting that, while it is one thing to question whether the mere truth of such a proposition gives Fred or Sally a reason to think that the belief in question is true, it is another thing altogether to question whether the mere truth of such a proposition entitles Fred or Sally to regard the belief as (likely to be) true. For the purpose of determining the viability of a foundationalist reply to the regress-of-reasons problem, it is the latter question that is crucial. And while Klein might well think that an affirmative answer to the latter question depends on an affirmative answer to the former question, I will argue that it is distinctly question-begging to assume as much, since in effect this is to assume that one can be entitled to endorse a proposition (in the sense relevant to the regress-of-reasons problem) only if one has good reasons to think that the proposition is true. This is precisely what the foundationalist denies; and in what follows I will argue that Klein has no non-question-begging argument on his side. (I will not go so far as to defend any particular account of the entitlement in question, but I will be suggesting several possibilities in Sect. 13.4.)

13.3

Against the Regress of Reasons

I just argued that, given that the reliability condition ensures that beliefs formed through reliable processes meet the truth-aptness challenge by definition, the only challenge that can be plausibly raised against an externalist (reliabilist)

13

Some Notes on the Possibility of Foundationalist Justification

205

foundationalism is the entitlement challenge. To raise this challenge is to ask, of the allegedly foundational reasons, whether the subject is entitled to regard those reasons as (likely to be) true. As noted, this is precisely the sort of challenge that is familiarly raised against reliabilist accounts of justification. When it is raised in connection with the regress-of-reasons challenge, a foundationalist response must be able to account for the entitlement in question without appealing to other reasons that the subject herself has. In light of this, suppose Klein were to question whether the mere fact that S’s belief that p satisfies a process-reliability condition entitles S to regard the belief as (likely to be) true. Surely he would be correct to do so. Or at least most people would agree; that was the point of Truetemp, the reliable clairvoyant, the Denver virus case, and the problem of easy knowledge. But I submit that this is not an in-principle problem for the foundationalist. On the contrary, it merely puts on notice those who offer an externalist (here, process reliabilist) account of foundational reasons: proponents of such views owe an account of the entitlement in question. That is, their account must ensure that the beliefs (or reasons) it regards as foundational are such as to make the following claim true: (*) S is entitled to regard her belief that p as (likely to be) true.

Nor is this all. As Klein notes (and here I agree with him), the entitlement in question must be “connected to” the truth of the belief (2004, 71). As I understand this additional requirement, it must be the case that the entitlement itself must derive from the feature(s) X in virtue of which the belief is (likely to be) true. I will formulate this as the requirement to ensure that the following holds: (**) S’s belief that p has some feature(s), X, in virtue of which (i) S’s belief is (likely to be) true, and (ii) S is entitled to assume as much.

In developing this line of thinking, I am essentially developing Klein’s case against the foundationalist. Where I disagree with him is over the ability of foundationalism to meet this demand: I think that there are versions of foundationalism that can ensure the truth of (**). More importantly, I submit that unless Klein has some reason to think that such an account cannot be offered, he is wrong to think that this is an insuperable difficulty—and so he would be wrong to think that foundationalist replies cannot answer the regress-of-reasons problem. Does he have a reason, then, to think that no foundationalist can offer an account of the entitlement in question, on which (**) is true for all foundational beliefs? In what follows I want to argue that, while Klein might take himself to have such a reason, the most plausible candidate reasons—or, rather, what from his perspective are the most plausible candidate reasons—beg the question against the foundationalist. Suppose Klein’s reason for thinking that no foundationalist account can secure the truth of (**) is that he thinks that: KR: The entitlement in condition (ii) of (**) must derive from a prior reason S has.

If KR is true, this would ensure that the regress continues, and so the foundationalist reply does not succeed. However, I hope it is clear that if KR is the only reason Klein

206

S. C. Goldberg

has for thinking that no foundationalist account can secure the truth of (**) for its foundational beliefs, he has begged the question against the foundationalist. This is because the foundationalist denies KR (compare Ginet 2005). Now I am not sure whether KR is the only reason Klein might have for thinking that no foundationalist account ensuring the truth of (**) is forthcoming. But he sometimes speaks as if it is. For Klein uses ‘arbitrary’ when he wants to designate the situation in which the subject lacks the sort of entitlement described in the condition (ii) of (**): when he describes an alleged regress-stopping-point as arbitrary, he appears to mean that the subject is not entitled to regard that stopping-point as (likely to be) true. But it is noteworthy that he explicitly construes the sort of arbitrariness in question in terms of reasons. Recall the Principle of Avoiding Arbitrariness: PAA: For all x, if a person, S, has a justification for x, then there is some reason, r1, available to S for x; and there is some reason r2, available to S for r1, etc. (as cited in Klein 2003, 719)

In Klein (1998, 2003), PAA is the only principle Klein cites which makes trouble for the foundationalist. The trouble is that this appeal to PAA is itself question-begging, as it is based on the assumption of KR. We can bring this out as follows. If we substitute the belief that p for ‘x’ in PAA, PAA implies that if S has no reason for her belief that p, then S has no justification for this belief.9 Now insofar as a belief formed through a reliable belief-forming process can’t fail to be justified for failing to satisfy the truth-aptness condition, it would appear that if Klein is right to suppose that a reliably-formed belief nevertheless fails to be justified—that is, if the subject “has no justification” for the belief—it must be because the subject fails to satisfy the entitlement condition with respect to this belief. But if this is how Klein would have us criticize the reliabilist foundationalist—and I can see no other way to ground his criticism—it reveals that his appeal to PAA commits him to the thesis that one is entitled to regard a proposition as (likely to be) true only if one has a prior reason to do so. In short, it reveals that Klein’s commitment to PAA commits him to KR—and so begs the question against the foundationalist. Still, if Klein has positive reasons that he can offer in defense of PAA and KR, perhaps he can claim that he isn’t assuming PAA or KR—he’s argued for them. So what is his argument for PAA or KR? Unfortunately, on this score Klein is not as explicit as one might hope. Nevertheless, we can get a sense of what he thinks supports PAA—which he understands to embody the demand that “every reason stands in need of another reason” (1998, 299)—by considering what he has to say of those theories, such as foundationalism, that violate it. The thought is that they are “unacceptable because [they] advocate accepting an arbitrary reason at the base, that is, a reason for which there are no further reasons making it even slightly better to accept than any of its contraries” (297; emphasis added). But once again I fear that this claim can be made out only if Klein makes the question-begging assumption of KR. I assume that this talk of ‘having a justification’ is stand-in for talk of doxastic justification. After all, doxastic justification is the sort of justification tied to knowledge, and it is this that Klein claims to be focusing on in the regress-of-reasons challenge. 9

13

Some Notes on the Possibility of Foundationalist Justification

207

The basis for my present allegation should already be clear. I grant that any adequate foundationalism will need to secure the truth of (**). But I also note, what I hope should be clear in any case, that if there is an account of foundational reasons that secures the truth of (**), then the case for foundationalism, as a reply to the regress-of-reasons problem, is alive and well, since in that case we will have shown that there can be foundational reasons that meet all of the individuallynecessary-and-jointly-sufficient conditions on being epistemically kosher. For this reason, Klein must deny that there can be an account of foundational reasons which secures the truth of (**). Now above I have been imagining a reliabilist foundationalism, where the feature in question, X, is that of having been produced by a reliable process. Such a foundationalism, I argued, can ensure that condition (i) of (**) holds of any belief of S’s with feature X: S’s belief is (likely to be) true. So the only way for Klein to resist that (**) holds is to deny that (when X ¼ the property of having been produced by a reliable process) condition (ii) of (**) holds: S is entitled to assume that her belief is (likely to be) true. Now, as I have been noting, Klein is in good company to think that the mere fact that a belief is produced by a reliable process does not entitle a subject to assume that the belief is likely to be true: this is the point of the examples of BonJour’s reliable clairvoyant, etc. But Klein’s view must be stronger. He must hold that there is no way to supplement the reliabilist account of justification so as to get condition (ii) of (**) to hold. And securing this stronger claim is what I think requires him to assume KR: without assuming KR he cannot resist what I will now argue are various ways to supplement reliabilism—various ways to engineer an appropriate value for X—so as to ensure that condition (ii) of (**) holds in any case in which the belief in question has property X.

13.4

An (Externalist) Entitlement to Foundational Belief

The notion I am pursuing—that of an entitlement to regard one’s own belief as (likely to be) true—is an abstract one, and it is a technical one to boot. It will be important to ensure that I am begging no questions against Klein in my understanding of that notion. To this end, it will be helpful to embrace points Klein himself has made, when he discusses the nature of the challenge that the regress-of-reasons problem presents to us. He writes that “the regress problem is about what kind of reasoning can satisfy the norms of epistemic responsibility” (2007a, 5). In line with this, let us assume that: (A) One is entitled to regard one’s own belief as (likely to be) true only if in believing as one does, one satisfies “the norms of epistemic responsibility.”

208

S. C. Goldberg

What is more, Klein also argues that the regress problem generates worries for the foundationalist about arbitrariness at the foundation. In line with this, let us assume further that: (B) One is entitled to regard one’s own belief as (likely to be) true only if in believing as one does, one’s belief is not arbitrary in any sense that implies that the proposition believed is epistemically no better off than its contrary.

(Here I speak of a proposition’s being epistemically better off than its contrary, and I avoid speaking as Klein does of reasons, since I am questioning (KR)‘s insistence on reasons, and I will be arguing that the requirement of non-arbitrariness can be satisfied even in the absence of reasons.) Finally, I will assume that if the consequent conditions of both (A) and (B) are satisfied, then S is entitled to regard her belief as true. (That is, I am assuming that these consequent conditions are not only individually necessary but also jointly sufficient for the entitlement in question.) It is worth making clear that my assumption on this score, that the consequent conditions of (A) and (B) are individually necessary and jointly sufficient for the entitlement, doesn’t beg any question against Klein. For one thing, if these two conditions are jointly satisfied, then his central demand (in Klein 2004, 71) is met: the very condition X that ensures that the subject was epistemically responsible in how she formed the belief in question, and so the very condition X that ensures that the subject satisfies (A)’s consequent, is also a condition that ensures that the subject’s belief is likely to be true (and so satisfies (B)’s consequent). For another, Klein can read into the “norms of epistemic responsibility” anything he wants short of begging the question (as he would do if he were to assume, for example, that epistemic responsibility always requires additional reasons whenever one has reasons at all). Here, then, is the task for the reliabilist: supplement the property of being produced by a reliable process in such a way that the fact that a belief has that (reliability + supplemental) property ensures that one would be entitled to regard one’s belief as (likely to be) true, which is to say that in believing as one does, one both satisfies the norms of epistemic responsibility and avoids arbitrariness. Here I can think of a variety of different ways to supplement the property in question. Consider the property of being produced by an environmentally sanctioned process (ESP), where: ESP: A belief-forming process φ is environmentally sanctioned ¼df (i) it is part of the species’ evolutionary endowment, (ii) it is reliable when working properly in normal conditions, (iii) humans have a sensitivity to the range of (abnormal) conditions under which it is not reliable and they reliably refrain from relying on it in those conditions, and (iv) conditions (ii) and (iii) explain why (i) holds (i.e., why the process-type has persisted throughout the generations).

Alternatively, consider the property of being produced by a socially sanctioned process (SSP), where: SSP: A belief-forming process φ is socially sanctioned ¼df (a) it is environmentally sanctioned, (b) the most reliable and epistemically responsible subjects in one’s community

13

Some Notes on the Possibility of Foundationalist Justification

209

endorse (or would endorse) φ as a belief-forming process, and (c) condition (a) partially explains condition (b).

Now I do not want to defend either ESP or SSP (although see Goldberg (2018), where I defend something in the vicinity of both at great length). My present point is only this: these two properties both do what the reliabilist needs to do to fulfill the task I noted at the outset of this paragraph. In particular, both properties supplement the property of being produced by a reliable process in such a way that the fact that a belief has that enhanced (reliability + supplemental) property ensures that one would be entitled to regard one’s belief as (likely to be) true. To establish this, I will now argue that if the reliabilist identifies X with the property of having been acquired through use of an environmentally sanctioned process, then a subject whose belief has feature X is such that in believing as she does (1) she satisfies the norms of epistemic responsibility and (2) she avoids arbitrariness. (If so, the same would hold true if the reliabilist identifies X with the property of being a socially sanctioned process.) Suppose that X ¼ the property of having been acquired through use of an environmentally sanctioned process. Then there are clear grounds for thinking that a subject whose belief has feature X is such that in believing as she does she satisfies the norms of epistemic responsibility. What might those norms be? We can imagine that they include some combination of evidential norms, coherence norms, reasoning norms, and norms pertaining to the sources/processes that are appropriate to rely on (perception, memory, and reasoning are OK; entrail-reading, clairvoyance, and astrology are not; etc.). But it would seem that insofar as the belief-forming processes that are environmentally sanctioned are all among those that are acceptable according to the source norms, anyone who forms beliefs through accepting the outputs of properly functioning processes of this sort would ipso facto be forming beliefs in an epistemically responsible way. The beliefs would be reliably formed; they would be in accord with the (perceptual, memorial, etc.) evidence; they would not give rise to cases in which the subject believes despite having grounds for thinking that she shouldn’t be relying on the process in those circumstances; the subject herself would be sensitive to conditions of having such grounds; the reasoning processes she employed themselves are among the approved processes; etc. It would seem, then, that beliefs formed in these ways are formed in an epistemically responsible fashion. To be sure, such beliefs don’t involve further reasons. But insisting on further reasons, as a condition on epistemic responsibility, would be begging one of the very questions at issue: no foundationalist should or will grant that epistemically responsible belief requires an endless set of reasons. In addition, if X ¼ the property of having been acquired through use of an environmentally sanctioned process, then there are also grounds for thinking that a subject whose belief has feature X is such that in believing as she does, she avoids arbitrariness. Recall Klein’s original construal of arbitrariness: a reason is arbitrary in this sense when it is “a reason for which there are no further reasons making it even slightly better to accept than any of its contraries” (Klein 1998, 297; emphasis added). Now I submit that there are two ways of taking this criterion, one of which is

210

S. C. Goldberg

question-begging, the other of which can be satisfied by the foundationalist. It would be question-begging if we simply assume that a subject’s reason is arbitrary unless the subject herself has further reasons that make it at least slightly better to accept this reason than any of its contraries. This is question-begging since it assumes something that the foundationalist denies, namely, that no reason (in a subject’s possession) is epistemically better than its contraries unless it is backed by further reasons (in the subject’s possession). The foundationalist denies this, even as she can allow that there are reasons that can be offered for thinking that the subject’s reason is epistemically better than its contraries: namely, it was produced by an environmentally sanctioned process, and an environmentally sanctioned process is one that is reliable—hence one that yields a preponderance of truths—when working properly in normal conditions. Indeed, this captures the sense in which the foundationalist can meet the anti-arbitrariness requirement. If we say that a reason is arbitrary in the absence of any reasons—whether they are possessed by the subject or not—then the foundationalist can point to the reliability considerations themselves as favoring the subject’s reasons over the contraries. The point is simply that the subject herself need not cite, or even be aware of, these reasons. And so I conclude that on the non-question-begging construal of the anti-arbitrariness requirement, the foundationalist has no difficulty at all meeting this requirement. In sum, when a subject’s belief is formed through an environmentally sanctioned process, it is formed in a way that is both epistemically responsible and relevantly non-arbitrary. What is more, the belief produced in such cases is epistemically kosher in the sense characterized above: such a belief is both likely to be true and is also such that the subject herself is entitled to assume as much. And I have been arguing that for this class of beliefs—namely, those produced by environmentally sanctioned processes—the foundationalist has an account of the belief’s status as epistemically kosher, which account does not appeal to still further reasons. The result, then, is that the foundationalist has an account of foundational beliefs. My claim has been that Klein can resist this account if—and it would appear only if—he makes one or another question-begging assumption. He might assume that the truthaptness requirement can be met only if there is an infinite chain of reasoning. Alternatively, he might assume that the entitlement requirement can be met only if there is an infinite chain of reasoning. Finally, he might assume that a belief is formed in an epistemically responsible fashion only if it is backed by an infinite chain of reasoning. But in each case, to make the assumption in question just is to beg the question against the foundationalist. After all, the foundationalist has given us an independent account of the satisfaction of those two requirements, as well as an account of the conditions on epistemic responsibility. If foundationalism fails, it must be because one of these accounts is false. If Klein thinks this is so, he owes us an argument. Short of that, we have grounds to conclude that Klein’s central thesis is false: it is not true that no foundationalist account of reasons could possibly work as a reply to the regress-of-reasons problem.

13

Some Notes on the Possibility of Foundationalist Justification

13.5

211

Conclusion

In this chapter my aim has been rather modest. I claim to have shown that Klein’s contention—that foundationalism cannot reply to the regress-of-reasons problem— can be sustained only if Klein makes one or another question-begging assumption. In a way, this merely points out that more work needs to be done: if there really is to be a viable version of foundationalism, more must be said about what it is that entitles a subject to regard the foundational beliefs as (likely to be) true. This meta-epistemological task is one I leave for another occasion (see, e.g., Goldberg 2018).

References Bergmann, M. (2006). Justification without awareness: A defense of epistemic externalism. Oxford: Clarendon Press. Bonjour, L. (1980). Externalist theories of empirical knowledge. Midwest Studies in Philosophy, 5 (1), 53–74. Cohen, S. (2002). Basic knowledge and the problem of easy knowledge. Philosophy and Phenomenological Research, 65(2), 309–329. Fumerton, R. (1995). Metaepistemology and skepticism. Maryland: Rowman & Littlefield Publishers. Ginet, C. (2005). Infinitism is not the solution to the regress problem. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 140–149). London: Blackwell. Goldberg, S. (2015). What is the subject-matter of the theory of epistemic justification? In D. Henderson & J. Greco (Eds.), Epistemic evaluation: Purposeful epistemology (pp. 205–223). Oxford: Oxford University Press. Goldberg, S. (2018). To the best of our knowledge: Social expectations and epistemic normativity. Oxford: Oxford University Press. Klein, P. (1998). Foundationalism and the infinite regress of reasons. Philosophy and Phenomenological Research, 58(4), 919–925. Klein, P. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13(Epistemology), 297–325. Klein, P. (2003). When infinite regresses are not vicious. Philosophy and Phenomenological Research, 66(3), 718–729. Klein, P. (2004). What is wrong with foundationalism is that it cannot solve the epistemic regress problem. Philosophy and Phenomenological Research, 68(1), 166–171. Klein, P. (2007a). Human knowledge and the infinite progress of reasoning. Philosophical Studies, 134(1), 1–17. Klein, P. (2007b). How to be an infinitist about doxastic justification. Philosophical Studies, 134, 25–29. Klein, P. (2014). Reasons, reasoning and knowledge: A proposed rapprochement between infinitism and foundationalism. In P. Klein & J. Turri (Eds.), Ad infinitum: New essays on epistemological infinitism (pp. 105–124). Oxford: Oxford University Press. Lackey, J. (2007). Why we don’t deserve credit for everything we know. Synthese, 158(3), 345–361. Lehrer, K. (1990/2015). Theory of knowledge. New York: Routledge. Vogel, J. (2008). Epistemic bootstrapping. The Journal of Philosophy, 105(9), 518–539.

Chapter 14

A Formal Account of Epistemic Defeat Matthew Kotzen

Abstract The goal of this chapter is to disentangle several related—but importantly distinct—notions of evidential defeat. The broadest distinction in the literature on epistemic defeat is that between rebutting and undercutting defeat; very roughly, the idea is that rebutting defeaters provide a “positive” reason to disbelieve the conclusion, whereas an undercutting defeater merely “blocks” existing reasons to believe the conclusion. In this chapter, I formalize these two notions and explore some related (and under-discussed) phenomena such as “hybrid” defeat (where a single defeater can both rebut and undercut), “bidirectional” defeat (where some information that serves as evidence for a conclusion can become a defeater in the presence of another piece of evidence for the conclusion), and “redundant” defeat (where an undercutting effect is generated by the non-independence of two pieces of information, rather than by the “blocking” phenomenon that occurs in more typical cases of undercutting). Keywords Formal epistemology · Degrees of justification · Evidential defeat · Belief revision · Justification defeat · Bayesian epistemology · Types of defeaters · Bayesian approaches to defeaters · Epistemology

14.1

Introduction

It is standard in epistemology to distinguish between two different kinds of defeaters of S’s evidence for her belief that q—“undercutting” defeaters and “opposing”1 defeaters. Very roughly, an undercutting defeater undermines the effect that the This latter class is more commonly referred to as “rebutting” defeaters; for reasons that will become clear later, I prefer the term “opposing.” The term is due to Jim Pryor. 1

M. Kotzen (*) UNC Department of Philosophy, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_14

213

214

M. Kotzen

evidence has on q, without directly providing evidence against q.2 An opposing defeater, on the other hand, serves as “direct” evidence against q. Suppose, for instance, that Julio tells me that it is raining outside and that I believe on that basis that it is raining outside. If I were to learn that Julio in fact has no idea what the weather is like right now but likes to make up claims about the weather and confidently share them with others, that would plausibly serve as an undercutting defeater of my evidence that it’s raining. This new information about Julio isn’t “directly” evidence that it’s not raining outside; rather, it merely undermines the evidential force that I originally took his testimony to have. If, however, I were to go outside and see for myself that it’s sunny, that experience would serve as an opposing defeater of my evidence that it’s raining; here, my new experience does serve as “direct” evidence against the proposition that it’s raining. One thing we might wonder about is whether we can draw this distinction in a more principled and precise way—and, if so, how we should go about doing it. This is particularly important because some philosophers—for instance, Jim Pryor (see Manuscript and 2013)—have expressed doubts about whether the distinction can be captured using the standard Bayesian probabilistic model of partial belief. If this doubt is justified, then it provides some reason to pursue a richer formal model that can capture the distinction; after all, we want a formal model of belief to capture those distinctions that play a crucial role in epistemology. In this chapter, I will give a formal account of epistemic defeat generally, and also of the distinction between undercutting and opposing defeaters, using the standard Bayesian apparatus. I will also introduce a third kind of defeat that seems to have gone unnoticed in the epistemological literature, which I call “bidirectional” defeat. My account will also handle cases of “partial” defeat, where the evidential effect under consideration is only partly undermined, as well as cases of “hybrid” defeat, where the same defeater plays both an undercutting and an opposing role (or an undercutting and a supporting role). I will close by addressing a concern that my account cannot properly distinguish undercutting defeat from evidential redundancy.

14.2

A First Stab

Consider again the case where I take Julio’s testimony that it’s raining to be some evidence that it’s raining. Let E be Julio’s testimony that it’s raining out and let H be the proposition that it is raining out. Let’s assume that, given the background information that I have at the 2 There has been a lot of discussion in the literature recently on so-called “higher-order” defeaters, which work by inducing doubts about the reliability of the cognitive process(es) that produced a belief. See, e.g., Christensen (2007a, b, c, 2009, 2010), Elga (unpublished), Kelly (2010), LasonenAarnio (2014), and Schechter (2011). I consider these sorts of higher-order defeaters to be undercutting defeaters, though I will for the most part focus on lower-order kinds of undercutting defeat in this chapter.

14

A Formal Account of Epistemic Defeat

215

time that I hear Julio’s testimony, E is evidence (for me) for H. In Bayesian terms, this amounts to the assumption that (my) p(H|E) > p(H ). Call my credence in H before collecting E, p(H ), my prior credence in H. Call my credence in H after collecting E but before acquiring any defeaters, p(H|E), my evidential credence in H.3 Call my credence in H after collecting E and after acquiring defeater D, p(H| E & D), my defeated credence in H. In my rough characterization above of the difference between undercutting and opposing defeaters, I suggested that while opposing defeaters constitute direct evidence against the relevant proposition, undercutting defeaters do not. Undercutting defeaters merely undermine the evidential force of some other evidence for the proposition, and therefore do not constitute direct evidence against the proposition. One might think, then, that we can characterize opposing defeaters as those propositions which, if learned, lower our credence in H (on the assumption that E is known): OPPOSING IFF CREDENCE-LOWERING: D is an opposing defeater for the evidence that E provides for H just in case p(H| E & D) < p(H|E).

But this clearly will not work. However we should understand the distinction between the “directness” of opposing defeaters and the “indirectness” of undercutting defeaters, both kinds of defeaters are defeaters, and both can therefore lower our credence in H. Take, for instance, the information that Julio is a highly unreliable testifier about the weather—a paradigm case of an undercutting defeater (of Julio’s testimonial evidence that it’s raining). Once I learn this, it seems clear, I ought to become less confident that it’s raining than I was before finding out how unreliable Julio is. Thus, in this case, it’s true that p(H| E & D) < p(H|E) even though the D in question is an undercutting defeater. Therefore, OPPOSING IFF CREDENCE-LOWERING can’t be used to uniquely characterize opposing defeaters. Still, it’s somewhat plausible that the condition that p(H| E & D) < p(H| E) characterizes defeaters in general, even if it doesn’t distinguish between opposing and undercutting defeaters. It’s plausible that, when E’s evidential effect on H is defeated, the agent’s defeated credence in H is lower than her evidential credence. So let’s provisionally accept: DEFEATER IFF CREDENCE-LOWERING: D is a defeater for the evidence that E provides for H just in case p(H| E & D) < p(H|E).

One advantage of DEFEATER IFF CREDENCE-LOWERING over several extant theories is that it accounts for cases of partial defeat which do not cross the “threshold of binary belief”—i.e., cases where the defeater lowers the agent’s credence in H from its evidential value, but does not lower it from a point above the threshold to a point below the threshold. Since several other theories of defeat focus only on binary beliefs, they are insensitive to this kind of partial defeat.

3 Here and throughout, I will assume Conditionalization—i.e., I will assume that the new rational credence for me to have in H, after collecting exactly evidence E, is my old p(H|E).

216

M. Kotzen

For example, in the course of developing a “No True Defeaters”-style solution to the Gettier Problem, Klein (1971, 1976) develop an account of defeat according to which a (true) proposition D defeats4 S’s justification to believe p just in case, “if S were to add d to whatever justified p for S, p would no longer be justified for S” (1976, 802). Similarly, in Contemporary Theories of Knowledge (Pollock and Cruz 1999, 37), Pollock and Cruz give the following characterization of defeaters5: If E is a reason for S to believe H, D is a defeater for this reason if and only if D is logically consistent with E, and E & D is not a reason for S to believe H.

Chisholm provides an account that is similar in spirit to both Klein’s and Pollock and Cruz’s, though Chisholm sets things up in terms of defeat of some evidence’s “tendency to make a hypothesis evident”: D defeats E’s tendency to make H evident ¼df E tends to make H evident; and D& E does not tend to make H evident. (Chisholm 1989, 53)

Since these accounts focus on binary belief, they do not capture cases where the agent’s defeated credence in H is lower than her evidential credence in H, and yet where either both or neither of those credences is above the threshold for belief. Of course, this is not an objection to any of these accounts; it is a perfectly legitimate epistemological project to give an account of the defeat of binary beliefs, and I have no particular claim on the term “defeat” to characterize the more general phenomenon that I’m interested in. But it is also a legitimate epistemological project to characterize this more general phenomenon, and an advantage of DEFEATER IFF CREDENCE-LOWERING is that it is able to do so. Even if DEFEATER IFF CREDENCE-LOWERING is correct, we still have the problem of distinguishing opposing defeaters from undercutting defeaters. One natural idea is that, since undercutting defeaters merely (perhaps only partially) undermine the evidential effect that E has on H, the net effect of acquiring some undercutting defeater should be to push the agent’s credence in H back toward the value it had before the agent acquired E, so that her defeated credence in H is lower than her evidential credence in H but no lower than her prior credence in H. By contrast, since an opposing defeater actually gives us independent evidence against H, it has the capacity to push the agent’s credence in H below the value it had before the agent acquired E, leading to a defeated credence that is lower than the prior credence. Thus: UNDERCUTTING IFF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE: D is an undercutting defeater for the evidence that E provides for H just in case p(H )  p(H| E & D) < p(H|E). OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE: D is an opposing defeater for the evidence that E provides for H just in case p(H| E & D) < p(H ) < p(H|E).

Klein uses the term “disqualifying proposition” in Klein (1971), but he uses the term “defeater” in Klein (1976). 5 Here and throughout, I have changed the notation of the theories I’m discussing for the sake of consistency. 4

14

A Formal Account of Epistemic Defeat

217

Notice that each of UNDERCUTTING IFF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE’s condition and OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition individually entail DEFEATER IFF CREDENCE-LOWERING’s condition; since both undercutting defeaters and opposing defeaters are defeaters, this is desirable. Notice too that DEFEATER IFF CREDENCE-LOWERING’s condition entails the disjunction of UNDERCUTTING IFF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE’s condition and OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition; if you think that all defeaters are either undercutting defeaters or opposing defeaters, then this is desirable as well. However, just because some opposing defeaters can push the agent’s credence below what it was before acquiring E, it clearly doesn’t follow that all opposing defeaters have that effect. The opposing defeater in the rain example above—my going outside and seeing for myself that it’s sunny out—is a particularly strong opposing defeater; barring skeptical scenarios and elaborate tricks, the experience I have when I walk outside and seem to see sunny and cloudless skies makes me virtually certain that it’s not raining out, regardless of whose testimony about the weather I’ve previously heard. Thus, assuming that I found it at least moderately credible that it was raining out before hearing Julio’s testimony, the combined effect of Julio’s testimony and the opposing defeater of that testimony (i.e., my sunny experience) will be to make me less confident that it’s raining out than I originally was, satisfying OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition. But there surely are weaker opposing defeaters possible. For example, suppose that instead of going outside and observing the weather for myself, I had instead come across Jill, someone in whom I have the same confidence that I do in Julio. If Jill were to tell me that it’s not raining out, it’s fairly plausible that this would serve as a (weaker) opposing defeater of Julio’s testimony. But since I have equal confidence in Julio and Jill, it’s plausible that my defeated credence in H would (approximately) equal my prior credence; after all, there’s a natural sense in which these two pieces of testimonial evidence “cancel each other out.” Similarly, if I had slightly less confidence in Jill’s reliability than I do in Julio’s, her testimony that it’s not raining would still serve to lower my credence somewhat that it’s raining, though not all the way down to my prior credence. These are both counterexamples to the necessity of OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE, and also to the sufficiency of UNDERCUTTING IFF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE. So far, nothing that I have said rules out the sufficiency of OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE. Even if we grant that not every opposing defeater generates a defeated credence in H lower than the subject’s prior credence in H, still it might be true that only opposing defeaters can do so. And indeed, it’s at least somewhat plausible that p(H| E & D) < p(H ) < p(H|E) does specify a sufficient condition for opposing defeat. After all, if D merely undermines the effect of E, how could D leave us with a defeated credence in H that is lower than our prior credence in H? It seems that in order to end up with a defeated credence lower than our prior credence, we’d need some independent reason to disbelieve H; a reason merely to doubt the evidential force of E (i.e., an undercutting defeater) seems

218

M. Kotzen

like it couldn’t be such an independent reason to disbelieve H. Thus, only opposing defeaters look to be capable of generating a defeated credence which is lower than the prior credence. So we can fix the problem presented by the counterexamples to the necessity of OPPOSING IFF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE by simply dropping the “only if”: OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE: D is an opposing defeater for the evidence that E provides for H if p(H| E & D) < p(H ) < p(H|E).

And we can fix the problem presented by the counterexamples to the sufficiency of UNDERCUTTING IFF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE by dropping the “if”: UNDERCUTTING ONLY IF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE: D is an undercutting defeater for the evidence that E provides for H only if p(H)  p(H| E & D) < p (H|E).

However, we’d like to be able to say more. We’d like well-motivated necessary and sufficient conditions both for undercutting and for opposing defeat. And we’d like a way to distinguish between those cases where p(H ) < p(H| E & D) < p(H|E) and where D is an undercutting defeater from those cases where p(H ) < p(H| E & D) < p (H|E) and where D is an opposing defeater. Neither OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE nor UNDERCUTTING ONLY IF EVIDENTIAL CREDENCE IS NO LOWER THAN PRIOR CREDENCE gives us any guidance here.

14.3

A Different Approach

In order to pursue a different tack, let’s look at how Pollock and Cruz characterize opposing defeaters: If E is a defeasible reason for S to believe H, D is a[n opposing] defeater for this reason if and only if D is a defeater (for E as a reason for S to believe H ) and D is a reason for S to believe ~H. (Pollock and Cruz 1999, 37)

The idea behind Pollock and Cruz’s account of opposing defeaters can’t be that opposing defeaters are defeaters that motivate us to lower our credence in H (i.e., to have a defeated credence that is lower than our evidential credence); as already discussed, undercutting defeaters do that too. Nor can it be that opposing defeaters leave us with a low credence in H; if your prior credence in H was low, then a fairly strong undercutting defeater can do that too. Rather, a more plausible reading of their account is that opposing defeaters are reasons to lower our credence in H even when we ignore E. In other words, a natural thought here (regardless of whether it is Pollock and Cruz’s thought or not) is that the core difference between an opposing defeater and an undercutting defeater is that, since an undercutting defeater merely undermines (perhaps only partially) the evidential effect of E on H, an undercutting defeater would have no effect on H once E is removed from consideration. By contrast, an

14

A Formal Account of Epistemic Defeat

219

opposing defeater provides some general reason to think that H is false; thus, its effect on H isn’t mediated by E. In other words, an undercutting defeater reduces our credence in H only on the assumption of E, while an opposing defeater reduces our credence in H regardless of whether we assume E or not. To distinguish the two types of defeaters, then, we need to take a look at the effect of the defeater on H when we haven’t already taken E into account. This motivates the following: UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT: D is an undercutting defeater for the evidence that the E provides for H just in case [p(H| E & D) < p(H|E)] & [p(H|D) ¼ p(H )]. OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT: D is an opposing defeater for the evidence that E provides for H if and only if [p(H| E & D) < p(H|E)] & [p(H|D) < p(H )].

Obviously, each of UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition individually entails DEFEATER IFF CREDENCE-LOWERING’s condition. Moreover, notice that UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition are mutually exclusive; again, this is desirable if we think that undercutting defeat and opposing defeat are mutually exclusive. I’ll say more about this in Sect. 14.4. It would be noteworthy if the condition specified in OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE entailed the condition specified in OPPOS6 ING IFF DEFEATER HAS INDEPENDENT EFFECT, given our background assumptions. As already argued, it’s plausible that OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE specifies a sufficient condition for opposing defeat. And if OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT is true, and the condition specified in OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE entails the condition specified in OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT, then it follows that OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE is true too. This would be a prima facie mark in favor of OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT. The condition specified in OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE might look to entail the condition specified in OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT. OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition entails DEFEATER IFF CREDENCE-LOWERING’s condition, which is the first conjunct of OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition. After all, OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition says that p(H| E & D) < p(H ), and we’re assuming throughout this discussion that p(H) < p(H|E). By transitivity, p(H| E & D) < p(H|E), which is the first conjunct of OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition. What about the second conjunct? It might seem as though OPPOSING IF EVIDENTIAL CREDENCE IS LOWER THAN PRIOR CREDENCE’s condition does entail the second conjunct of OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition. After all, OPPOSING IF EVIDENTIAL CREDENCE IS

6

Namely, that p(H|E) > p(H ).

220

M. Kotzen

LOWER THAN PRIOR CREDENCE’s condition says that p(H| E & D) < p(H )—i.e., that E & D lowers the agent’s credence in H to below her prior credence in H. But we’re assuming that p(H|E) > p(H )—i.e., that E alone raises the agent’s credence in H to above her prior credence in H. So if E alone raises the agent’s credence in H, and the conjunction E & D lowers the agent’s credence in H, isn’t the only explanation for this that D alone must lower the agent’s credence in H (and by more than E raises it)? After all, if D raised (or was neutral to) the agent’s credence in H, how could the combined effect of E and D be to lower the agent’s credence in H, given that E alone raises it? The above argument, however, is invalid. The somewhat plausible-sounding principle CONJUNCTIONS OF CONFIRMERS ARE CONFIRMERS (CCC): If A is evidence for H, and B is evidence for H, then A & B is evidence for H.

is false. Here’s a counterexample: Suppose that someone is applying for a job in your department, and that you don’t know which graduate program she’s coming from. You know that Graduate Program X produces above-average percentages of Metaphysics students (compared to other graduate programs), and also above-average percentages of Logic students. However, suppose that you also know that Program X never produces students who do both Metaphysics and Logic; students are forced to choose between these areas in their first year, and aren’t permitted to work in both. However, this policy is unique to X; though this happens somewhat infrequently, other programs do produce students who work in both Metaphysics and Logic. Suppose first that you find out only that the applicant does Metaphysics (it’s left open when she works in other areas too). Since you know that X produces an aboveaverage number of Metaphysics students, this is some evidence that the applicant comes from X. If, instead of finding out that she does Metaphysics, you had instead found out that she does Logic, that too would have been some evidence that she comes from X. But if you learn both that she does Metaphysics and that she does Logic, this is clearly conclusive evidence that she is not from X, since X doesn’t produce any students who do both Metaphysics and Logic. Thus, the fact that the applicant does Metaphysics is a confirmer that she’s from X, and the fact that she does Logic is a confirmer, but the joint effect of these confirmers is to disconfirm. Hence CCC is false. Suppose that, in this case, you learn first that the applicant does Metaphysics. As argued above, this is some evidence that she’s from X. When you learn that she also does Logic, you become certain that she’s not from X. It’s quite natural, then, to characterize the information that she does Logic as a defeater of your evidence that she’s from X. But there’s something a bit strange going on here. If you hadn’t already learned that she does Metaphysics, then the information that she does Logic would have been evidence for the hypothesis that she comes from X. Thus, whether the information that she does Logic is a confirmer or a defeater of the hypothesis that she’s from X seems to turn on the order in which you learn her specialties. I don’t know how to answer the question whether, in the case above, the information that the applicant does Logic is an undercutting defeater or an opposing

14

A Formal Account of Epistemic Defeat

221

defeater. I’m hesitant to classify it as an undercutting defeater, since (given that we also know that she does Metaphysics) the defeated credence is lower than the prior credence. And I’m hesitant to classify it as an opposing defeater, since it actually confirms H if we assume that we don’t know that she does Metaphysics. I don’t think that our pre-theoretic notions or intuitions are clear enough to deliver an unambiguous verdict here. I propose that we classify this case as involving a new type of defeater, which I call a bidirectional defeater, characterized as follows: BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER: D is a bidirectional defeater for the evidence that E provides for H if and only if [p(H| E & D) < p(H|E)] & [p(H| D) > p(H )].

So where are we now? We have the following tripartite distinction: UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT: D is an undercutting defeater for the evidence that E provides for H just in case [p(H| E & D) < p(H|E)] & [p(H|D) ¼ p(H )]. OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT: D is an opposing defeater for the evidence that E provides for H if and only if [p(H| E & D) < p(H|E)] & [p(H|D) < p(H )]. BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER: D is a bidirectional defeater for the evidence that E provides for H if and only if [p(H| E & D) < p(H|E)] & [p(H| D) > p(H )].

We retain DEFEATER IFF CREDENCE-LOWERING: D is a defeater for the evidence that E provides for H just in case p(H| E & D) < p(H|E).

as a general characterization of defeaters, since the three conditions above individually entail it (and it clearly entails the disjunction of the three conditions). Now let’s see whether this gets the cases that we started with right. Take the case where Julio’s testimony that it’s raining out is undercut by the information that Julio is an unreliable testifier about the weather. Does this case satisfy UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition? Clearly, p(H| E & D) < p(H|E) here. When I first heard Julio’s testimony, I became more confident that it was raining out. When I learned that he was unreliable, I became less confident that it was raining out. Notice that this inequality holds regardless of whether the defeater is that Julio is highly unreliable or that he is only somewhat unreliable. Either way, it’s plausible that the defeater induces a defeated credence lower than my evidential credence. What about the second conjunct? Before hearing Julio’s testimony, I had my prior credence that it’s raining out, p(H ). How would this credence be affected by learning only that Julio is unreliable? Intuitively, not at all. After all, I haven’t heard Julio’s testimony yet, so I don’t even know what his testimony is, and I certainly haven’t taken it into account yet. I might not have even heard of Julio, so the information that he tends to get weather reports wrong shouldn’t seem particularly relevant to me. Thus, my p (H ) ¼ p(H|D), so both conjuncts of UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition are satisfied. Now take our paradigm cases of an opposing defeater, where Julio’s testimony that it’s raining is defeated either by my own sunny experience outside or by Jill’s

222

M. Kotzen

testimony that it’s not raining. Is OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition satisfied? Again, it’s fairly clear that p(H| E & D) < p(H|E); both of these defeaters give me reason to lower my credence in H on the assumption that I’ve already taken E into account. Moreover, it’s fairly clear that the second conjunct of OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition is also satisfied. Since each of these defeaters intuitively constitutes direct evidence against H, it should follow that acquiring the defeaters would force me to lower my credence in H, even if E hasn’t already been taken into account. Even if I haven’t heard Julio’s testimony or don’t know who Julio is, still a sunny experience or Jill’s testimony that it’s not raining constitute evidence that it’s not raining. Accordingly, p(H|D) < p(H ), so OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition is satisfied.

14.4

Hybrids

I think that DEFEATER IFF CREDENCE-LOWERING, UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT, OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT, and BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER do a fairly good job characterizing ordinary types of epistemic defeat. However, these principles do less well when we consider hybrid cases of defeat. Consider the following case. I have a visual experience as of Ada writing a large check to a worthy charity, and on that basis my credence increases that she is a morally upstanding person.7 Then, I hear testimony from a reliable friend that Ada surreptitiously put a visual hallucinogen in my coffee this morning.8 Intuitively, two reactions seem appropriate here. First, I ought to (at least partially) discount the confirming effect that I took my visual experience as of Ada’s donation to have on the proposition that Ada is morally upstanding. Since I’ve acquired some evidence (in the form of the auditory testimony about Ada’s behavior this morning) that the visual experience I had as of Ada’s donation was a hallucination, I ought to regard that visual experience as a significantly less reliable indicator of Ada’s moral upstandingness than I previously thought. Second, it seems, I ought to “directly” decrease my credence that Ada is morally upstanding. After all, surreptitiously putting drugs in people’s coffee is a morally bad thing to do; since I have some

7

Some positions in metaethics—notably, some versions of non-cognitivism—entail that my attitudes about Ada’s moral upstandingness are non-cognitive in nature, and hence do not involve my being related to a proposition about Ada’s moral upstandingness. For the purposes of simplicity, I simply ignore these positions here; I assume in the text that being confident that Ada is morally upstanding is simply a matter of having a high credence in the proposition that Ada is morally upstanding. However, nothing essential turns on this choice, and the example could be modified (at the cost of simplicity) to avoid this complication. 8 For the purposes of this example, suppose that the hallucinogen at issue causes visual, but never auditory, hallucinations; thus, there is no reason for concern about whether your experience as of your friend speaking to you is veridical.

14

A Formal Account of Epistemic Defeat

223

reliable evidence that Ada performed this morally bad act, I have “direct” evidence against the proposition that Ada is morally upstanding. The trouble here is that UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition are incompatible, since UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition entails that p(H|D) ¼ p(H ) and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition entails that p(H|D) < p(H ). Thus, if these principles are correct, then hybrid cases of undercutting and opposing defeat are impossible. In the case under consideration, it’s plausible that p(H|D) < p(H ); even if I haven’t had the visual experience as of Ada writing the check, my friend’s testimony that Ada drugged my coffee is some evidence against the proposition that she is morally upstanding. So, OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT counts this as a case of opposing defeat. But that’s only part of the story; intuitively, this case is also a case of undercutting defeat, and UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition isn’t satisfied (indeed, necessarily, it can’t be satisfied if OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition is). If this is right, then UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition can’t be a necessary for undercutting defeat (though the hybrid cases don’t put any pressure on the claim that it is a sufficient condition for undercutting defeat). I think that this sort of hybrid case motivates a reconsideration of our strategy for characterizing undercutting defeat. Though I think that it is false, Pollock and Cruz’s account is again instructive here. Here is how they characterize undercutting defeaters: If believing E is a defeasible reason for S to believe H, D is an undercutting defeater for this reason if and only if D is a defeater (for believing E as a reason for S to believe H ) and D is a reason for S to doubt or deny that E would not be true unless H were true. (Pollock and Cruz 1999, 37)

This characterization is a bit odd. First, there are notorious problems with counterfactual analyses, and there are the usual wrinkles related to overdetermination, pre-emption, etc. in this case. But more importantly, suppose again that Julio tells me that it’s raining out (E), and that I believe on that basis that it is raining out (H ). According to Pollock and Cruz, D is an undercutting defeater of Julio’s testimony (for me) if only if (D is a defeater and) D gives me a reason to doubt or deny that: Julio wouldn’t say it’s raining unless it were raining. In some cases, this account works just fine: a paradigm undercutting defeater like information that there is no correlation whatsoever between Julio’s weather reports and the actual weather would indeed give me a reason to deny that Julio wouldn’t say it’s raining unless it were raining. However, suppose that I knew in advance that Julio is very unlikely to say that it’s raining when it’s not in fact raining. Now, suppose I learn that Julio is also very unlikely to say that it’s raining when it is raining; in fact, we can even suppose that he’s equally as unlikely to say that it’s raining regardless of whether it’s raining or not. This information should clearly be a defeater for Julio’s testimony about the weather, since it entails (given our background knowledge) that Julio is just as likely

224

M. Kotzen

to say that it’s raining when it is raining as when it’s not. But this information does not give us any reason to doubt or deny that Julio wouldn’t say it’s raining unless it were raining; we knew all along that Julio is very unlikely to say that it’s raining when it’s not raining, and the information in question doesn’t change our attitude about that. How should we fix things up? We need to weaken UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition, since hybrid cases show that it’s not necessary for undercutting defeat. And we need the weakened condition to be compatible with OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT (or whatever alternative condition we formulate to characterize opposing defeat), so that we can get hybrid cases of undercutting and opposing defeat. Pollock and Cruz’s characterization of undercutting defeat seemed to get at the idea that undercutting defeaters are such that, when they are assumed to be true, the evidential connection between E and H is interfered with; however, as argued above, their counterfactual formulation of that idea is flawed. Intuitively, in the hallucinogen case above, the reason that my reliable friend’s testimony serves as an undercutting defeater is that it’s some evidence that my visual experience as of Ada donating money was a hallucination. Thus, once I hear the testimony that Ada drugged my coffee, my visual experience as of her donating money no longer confirms the hypothesis that she is morally upstanding as much as that visual experience did before the testimony gave me reason to doubt the veridicality of my visual experiences. In other words, the degree of confirmation that E confers on H is lower once we assume D. Let dc(E, H, K) be a real-valued function of three variables E, H, and K, which quantifies the degree of confirmation that E confers on H relative to background information K. The above discussion motivates: UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING: D is an undercutting defeater for the evidence that E provides for H (relative to background information K ) just in case dc(E, H, K) > dc(E, H, K & D).

In other words, an undercutting defeater is such that the amount of confirmation that the evidence confers on the hypothesis is lower when you assume the truth of the undercutting defeater than when you don’t. It’s controversial how best to measure degree of confirmation.9 Here are a few candidates: dc1(E, H, K ) ¼ p(H|  p(H|K )
E & K)  pðH│ E&K Þ 10 dc2(E, H, K ) ¼log pðHjK Þ

9

See Fitelson (1999) and Eells and Fitelson (2000) for good surveys of various candidates. One purpose of taking the log of these quantities is so that the measure counts as a so-called “relevance measure,” where the measure is positive if E confirms H, negative if E disconfirms H, and 0 if E is neutral to H. Another purpose is to ensure scale-invariance. For our current purposes, the log can be ignored. Since log is a monotone increasing function, it will follow from the fact that 10

14

A Formal Account of Epistemic Defeat


dc3(E, H, K) ¼ log

 

pðH j E&K =pðeHj E&K Þ pðH jK =pðeHjK Þ

225




 ¼ log

pðE j H&K

  (The former is the

pðE j eH&K log of the “Bayes Factor” and the latter is the log of the “likelihood ratio.”)

For reasons that are beyond the scope of this chapter to address, I like dc3(E, H, K) as a measure of degree of confirmation.11 But for my purposes here, I wish to remain agnostic about which measure of confirmation is best. Whichever account is right, we can plug that account into UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING to yield a precisified account of undercutting defeat.12 For example, if you also like dc3(E, H, K), then UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING becomes: UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING*: D is an undercutting defeater for the evidence that E provides for H (relative to background information K ) just in case 

 pðH j E&K =pðHj eE&K Þ pðH j E&D&K =pðeHj E&D&K Þ  log > log . pðHjK Þ=pðeHjK Þ pðH j D&K =pðeHj D&K Þ

Two points are worth mentioning here. First, notice that UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition entails UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING*’s condition, given the assumptions that I’ve been making. UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition entails that p(H|D) ¼ p(H ), which entails that p(~H|D) ¼ p(~H), so the denominators of the expressions on both sides of the inequality can be ignored. And we’re supposing that p(H| E & D) < p(H ), since we’re supposing that D is a defeater, which entails that p(~H| E & D) > p(~H ). It follows that UNDERCUTTING IFF DEGREE-OFCONFIRMATION LOWERING*’s condition is satisfied. So, UNDERCUTTING IFF DEGREE-OFCONFIRMATION LOWERING* entails that UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition is a sufficient condition for undercutting defeat, which is desirable. Moreover, this result doesn’t essentially depend on my selection of dc3 as the measure of degree of confirmation. Presumably, any reasonable measure of confirmation is going to be a “relevance measure” (i.e., a measure that is positive if E confirms H, negative if E disconfirms H, and 0 if E is neutral to H ) and one such that if p(H|E1) > (H|E2) then E1 confirms H more than E2 does. If UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition holds, then p(H ) ¼ p(H|D), and hence on any reasonable measure dc(E, H, K) > dc(E, H, K & D) will hold iff the linear distance between p(H|E) and p(H ) (which is equal to p(H|D)) is greater than the linear distance between p(H| E & D) and p(H ). But that holds iff p(H| E & D) < p A > B that log A > log B (and conversely). So if we want to compare two degrees of confirmation, all we need to do is to compare the argument of the log. 11 For some reasons to accept dc2(E, H, K), see Milne 1996 (though see also Pollard 1999). For some reasons to accept dc3(E, H, K), see Eells and Fitelson (2000). 12 Of course, if there is no “one true measure” of degree of confirmation but rather just a plurality of different measures, then my account entails that there will be many different notions of undercutting defeat—one relative to each of the confirmation measures. But I think that this is precisely the right result; if there is no one privileged way to measure evidence, then I don’t think that there can be one privileged way to measure undercutting of evidence either.

226

M. Kotzen

(H|E), which is just our condition on defeat (and which UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT’s condition entails anyway). Second, and more importantly, notice that UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING*’s condition and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition are compatible with each other. A defeater D can simultaneously reduce the amount of confirmation that E confers on H and serve as evidence against H in the absence of E. Indeed, in the hallucinogen case, my friend’s testimony that Ada drugged my coffee seems to do precisely this. Since the testimony is some evidence that I’m hallucinating, it reduces the degree of confirmation that my visual experience as of Ada writing a check confers on the hypothesis that she is morally upstanding. And since the testimony is also some evidence that Ada has committed a morally bad act, it’s also “direct” evidence against the claim that she is upstanding. Somewhat more concretely: Let E be my visual experience as of Ada donating money, let H be the hypothesis that Ada is upstanding, and let D be my trusted friend’s testimony. Clearly, p(H|E) > p(H ) and p(H| E & D) < p(H|E). If we ignore the visual experience as of the donation, the testimony is still evidence against the hypothesis that Ada is upstanding, so p(H|D) < p(H ), so OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition is satisfied. It’s a little trickier to see that UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING*’s condition is satisfied, but it is. Let’s fill in the details of the case a bit to see that. Suppose that before having the visual experience, my credence that Ada is upstanding is .5 (so p(H ) ¼ .5) and my credence that Ada is non-upstanding is .5. After having a visual experience as of Ada donating money, my credence that Ada is upstanding goes up to .8 (so p(H|E) ¼ .8). If I had just heard my friend’s testimony that Ada drugged my coffee, without having the visual experience as of Ada donating money, my credence that Ada is upstanding would have gone down to .1 (so p(H|D) ¼ .1). The combined effect of acquiring both E and D isn’t to push my credence in H all the way down to .1; after all, there is still some non-trivial chance that my friend is mistaken or lying about Ada drugging my coffee, in which case I (probably) really did see her donate money to a worthy charity. But, since I don’t know much about Ada and trust my friend, my credence in H after acquiring both E and D should still be significantly lower than p(H ); let’s suppose that p(H| E & D) ¼ .15. UNDERCUTTING   IFF DEGREE  -OF-CONFIRMATION :8=:2 :15=:85 LOWERING* will then hold just in case log :5=:5 > log :1=:9 , which will hold just in case log 4 > log 1.588 (approximately). So UNDERCUTTING IFF DEGREE-OFCONFIRMATION LOWERING*’s condition holds. Clearly, this argument would have gone through even if I had used slightly different numbers.

14.5

A Second Taxonomy

Now that we’ve abandoned UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT in favor of UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING, should we modify OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT, BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER, or DEFEATER IFF CREDENCE-LOWERING

14

A Formal Account of Epistemic Defeat

227

as well? I see no reason yet to modify OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT; as argued above, it is consistent with UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING’s condition, which is desired, and it seems to capture our intuitions about opposing defeat quite naturally. But once we accept UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING, that begins to cast some doubt on DEFEATER IFF CREDENCE-LOWERING’s condition. As I’ll argue next, D can lower the extent to which E confirms H, and yet can fail to lower a rational agent’s evidential credence (so p(H| E & D)  p(H|E)). Is this a problem for UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING, or rather a problem for DEFEATER IFF CREDENCE-LOWERING? I think it’s the latter. Just as there can be hybrid case of undercutting defeat and opposition defeat, I think that there can also be hybrid cases of undercutting defeat and evidential support; in some such cases, the evidentially supporting effect of D can be strong enough to neutralize or outweigh D’s undercutting effect on E, in which case the net effect of D can be to make it the case that p(H| E & D)  p(H|E), in violation of DEFEATER IFF CREDENCE-LOWERING’s condition. But since this is still a case where D lowers the degree of confirmation that E confers on H, I still think that this sort of cases deserves to be classified as a case of (undercutting) defeat. To see such a case, we can make a few changes to the scenario involving Ada from above. This time, let H be the hypothesis that Ada is a morally bad person. And this time, let E be my visual experience as of Ada doing something mildly morally bad—say, cutting someone in line at the grocery store. Clearly, given suitable background assumptions, it’s plausible that p(H|E) > p(H ). As before, let D be a trusted friend’s testimony that Ada put a visual hallucinogen in my coffee this morning. If I have reason (from any source) to believe that I’m visually hallucinating, then clearly my visual experience as of Ada cutting in line does less to confirm the hypothesis that Ada is a morally bad person than it would if I didn’t have any such reason, so UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING entails (properly, in my view) that D is an undercutting defeater of the evidence that E provides for H in this case. But, in this case the effect of D’s “direct” evidential support for H is much stronger than the effect of D’s undercutting E’s support for H; seeming to see Ada cut in line is only very mild evidence that she is a morally bad person, whereas being told by a reliable friend that Ada drugs people without their knowledge is much stronger evidence that she is a morally bad person. So, it’s very plausible in this case that p(H| E & D) > p(H|E), in violation of DEFEATER IFF CREDENCE-LOWERING’s condition.13 If you agree with me that this is still a case of (undercutting) defeat, then we should abandon DEFEATER IFF CREDENCE-LOWERING. What about BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER? As I characterized bidirectional defeat above, it is a phenomenon where E and D are each evidence for H separately, but if we already know E, then D “flips” to being evidence against H. (The fact that the candidate does Metaphysics is evidence that

13 Clearly, we could play with the details here so that p(H| E & D) ¼ p(H|E), which would also violate Defeater IFF Credence-Lowering’s condition.

228

M. Kotzen

she’s from X, and the fact that she does Logic is evidence that she’s from X; but if you already know that she does Metaphysics, then the fact that she does Logic is evidence against her being from X.) BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER was a natural characterization of a third kind of defeat in the context of the acceptance of DEFEATER IFF CREDENCE-LOWERING, UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT, and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT; bidirectional defeaters, on that taxonomy, were just defeaters that were neither undercutting nor opposing. But once we abandon UNDERCUTTING IFF DEFEATER HAS NO INDEPENDENT EFFECT and DEFEATER IFF CREDENCE-LOWERING, it’s no longer clear what to say about the cases (like the philosophy job applicant case) that motivated BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER. Moreover, another clue that focusing on whether p(H|D) > p(H ) was the wrong strategy is that it is straightforward to construct cases quite similar to the job applicant case above where p(H|D) ¼ p(D)—say, by changing the case so that X produces an average number of Logic students, but still no students who do both Metaphysics and Logic. Of course, the issue here is at least partly stipulative; as far as I know, “bidirectional” defeaters aren’t discussed anywhere in the literature on epistemic defeat, and I certainly don’t think that we have clear intuitions about when a given D is serving as a bidirectional defeater. But, BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER fit into our old taxonomy in a way that made bidirectional defeat look to be a distinctive third kind of defeater, so it is natural to wonder how to fit bidirectional defeat into a taxonomy that accepts UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT. I propose that the crucial feature of the cases that motivated BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER is not (as BIDIRECTIONAL IFF E FLIPS D FROM A CONFIRMER TO A DISCONFIRMER claims) that D goes from confirming H to disconfirming H when we assume E, but rather that E goes from confirming H to disconfirming H when we assume D. This happens in the original job applicant case, where the fact that the candidate does Metaphysics goes from confirming that she’s from X to disconfirming that she’s from X when we assume that she also does Logic. And this also happens in a modified job applicant case where X produces an average (rather than an above-average) number of Logic students. So I propose to characterize bidirectional defeat as follows: BIDIRECTIONAL IFF D FLIPS E FROM A CONFIRMER TO A DISCONFIRMER: D is a bidirectional defeater for the evidence that E provides for H just in case p(H|D) > p(H| E & D).

Since we’re assuming that p(H|E) > p(H ), the satisfaction of BIDIRECTIONAL IFF D FLIPS E FROM A CONFIRMER TO A DISCONFIRMER’s condition entails that E’s positive relevance to H turns into negative relevance to H once D is assumed as background information.

14

A Formal Account of Epistemic Defeat

229

As it turns out, BIDIRECTIONAL IFF D FLIPS E FROM A CONFIRMER TO A DISCONFIRMER’s condition entails UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOW14 ERING’s condition, so this taxonomy has it that all bidirectional defeaters are undercutting defeaters (but not vice versa). Again, bidirectional defeat is a partly stipulative matter, so I’m perfectly happy with this result. On the current taxonomy, undercutting defeaters lower the amount that E confirms H when they’re assumed as background. Bidirectional defeaters, then, are just the special subclass of undercutting defeaters that lower the amount that E confirms H so much that that quantity becomes negative; E disconfirms H when a bidirectional defeater is assumed as background. How should we replace DEFEATER IFF CREDENCE-LOWERING as a general characterization of defeat? The answer to this question depends on whether all opposing defeaters turn out to be undercutting defeaters or not. I argued in Sect. 14.4 that there are some cases of “hybrid” defeaters that are both undercutting and opposing, but that clearly doesn’t settle the question of whether all opposing defeaters are undercutting defeaters. But if UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING is true, then there is some good reason to suspect that they might be. After all, if source S provides me with evidence that q is true, and I then acquire reason to believe that q is actually false, then I seem to have acquired at least some evidence that S is an unreliable source.15 To take the paradigm case of an opposing defeater from Sect. 14.1: when Julio tells me that it’s raining out and then I acquire reason to believe that it’s not raining out (either from my own observation or from Jill’s testimony), it’s plausible that I have (at least typically) acquired some new reason to put less stock in Julio’s testimony—in other words, that I have acquired information which serves as an undercutting defeater of Julio’s testimony. If it turns out that all opposing defeaters are indeed undercutting defeaters, then (since on my taxonomy all bidirectional defeaters are undercutting defeaters too) it follows that all defeaters are undercutting, and therefore that UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING characterizes defeat in general, and not just undercutting defeat. “Pure” undercutting defeaters, on this picture, would just be (undercutting) defeaters that do not also count as opposing or bidirectional. By contrast, it may turn out that not all opposing defeaters are undercutting defeaters. To return to our case of Ada, where E is my experience as of Ada donating money and H is the proposition that Ada is a morally good person, perhaps the information that Ada surreptitiously put a vision-enhancing drug in my coffee this

14 For any relevance measure of confirmation, since p(H|E) > p(H ), dc(E, H, K) > 0. Similarly, for any relevance measure, if p(H|D) > p(H| E & D), then dc(E, H, K & D) < 0. So, if Bidirectional IFF D Flips E From a Confirmer to a Disconfirmer’s condition holds, then dc(E, H, K) > 0 > dc(E, H, K & D), so Undercutting IFF Degree-of-Confirmation Lowering’s condition holds. 15 Note that while this argument provides some reason to believe that all opposing defeaters are undercutting defeaters, it provides no reason to think that all undercutting defeaters are opposing defeaters.

230

M. Kotzen

morning could serve as an opposing defeater of the evidence that E provides for H without also serving as an undercutting defeater of that evidence.16 If opposing defeat without undercutting defeat is possible, then I see no obvious or natural characteristic that UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING’s condition and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition have in common. On this picture, undercutting and opposing defeat are more distinct than they may at first have seemed, rather than being two instances of the same naturally characterizable type. Undercutting and opposing defeaters may be similar in that they both “count against” a hypothesis in some sense, but there might not be any non-disjunctive way to formalize any shared sense in which they both count against the hypothesis. So, if there are indeed some opposing defeaters that are not undercutting defeaters, then I think that all that we can usefully say is that D is a defeater of the evidence that E provides for H just in case it’s either an opposing or an undercutting defeater (we don’t need to include bidirectional defeat here, since all bidirectional defeaters are undercutting defeaters, on my taxonomy). Thus, we would accept: DEFEATER IFF UNDERCUTTING OR OPPOSING:: D is a defeater for the evidence that E provides for H (relative to background information K ) just in case either UNDERCUTTING IFF DEGREEOF-CONFIRMATION LOWERING’s condition or OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT’s condition is met.

One might worry that, just as there are hybrids of undercutting defeat and evidential support, so too might there be hybrids of opposing defeat and evidential support, which would make problems for DEFEATER IFF UNDERCUTTING OR OPPOSING and OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT. But I don’t think that there really are such hybrids. Take a putative hybrid of opposing defeat and evidential support, such as the information that Julio says it’s raining but Mary says it’s not; if that’s too conjunctive-sounding to you, imagine a light that goes on just in case both Julio says that it’s raining and Mary says that it’s not, and consider the evidential significance of the light going on. We might be inclined to count this is a hybrid because it is a conjunction (or, in the case of the light, it entails a conjunction) of a proposition that supports the hypothesis that it’s raining (the proposition that Julio says it’s raining) and a proposition that opposes that hypothesis (the proposition that Mary says it’s not raining). But if that’s a sufficient condition for such a hybrid, then we will end up with far too many hybrids. Consider the information that Julio said only that it’s raining. That’s just plain evidential support for the proposition that it’s raining, and certainly not any kind of opposing defeater for the hypothesis of rain. But Julio said only that it’s raining is a conjunction of Julio said it’s raining or Julio

16

The idea here would be that this information is an opposing defeater, since it’s morally bad to put drugs in people’s coffee without their consent, even if the drug has a vision-enhancing effect. However, this information wouldn’t be an undercutting defeater of the evidence that my visual experience provides for the proposition that Ada is morally good; if anything, this information would tend to strengthen the impact of my visual experience on that hypothesis, since it reduces the probability of a visual error.

14

A Formal Account of Epistemic Defeat

231

said that the barometric pressure is falling and Julio didn’t say that the barometric pressure is falling. Since the former conjunct is evidential support for rain, and the latter conjunct (under appropriate suppositions) is opposing defeat for rain, we end up with the result that Julio said only that it’s raining is a hybrid opposing defeater for rain, which is unacceptable.17 I take this to strongly suggest that there can’t be a non-trivial account of hybrids of opposing defeat and evidential support. Thus, I think it’s a welcome result that OPPOSING IFF DEFEATER HAS INDEPENDENT EFFECT entails that opposing defeaters can’t provide overall evidential support for H, and that DEFEATER IFF UNDERCUTTING OR OPPOSING doesn’t allow for pure18 hybrids of evidential support and opposing defeat.

14.6

Redundancy and Undercutting

Unfortunately, there is a potential wrinkle with UNDERCUTTING IFF DEGREE-OF-CONLOWERING. UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING says that D is an undercutting defeater just in case it lowers the degree to which E confirms H. The problem is not with the necessity of this condition; as far as I can tell, everything that we would intuitively count as an undercutting defeater does satisfy the condition. Rather, the problem is with the sufficiency; there is a broad class of propositions that do lower the degree to which E confirms H, and yet may not deserve to be counted as undercutting defeaters. The class that I have in mind is the class of propositions that are somehow redundant of E. Suppose, for example, that I know that my friend Rex is a fairly reliable predictor of the weather. Each night, Rex communicates to me his predictions for the weather the following day. Since Rex wants to make sure that I receive his predictions, he both sends me an email and also leaves me a voicemail with his predictions (the predictions are always identical in the email and the voicemail). One night, I receive Rex’s email with a prediction of rain the following day. Since I take Rex to be reliable about these matters, the email is evidence that it is going to rain. Next, I listen to Rex’s voicemail with (of course) the exact same prediction. Intuitively, the voicemail is redundant or perhaps irrelevant once I’ve already read the email, but it may be somewhat strained to call it an undercutting defeater of the evidence that the email provided for rain tomorrow. After all, the voicemail doesn’t cast any doubt on the reliability of the emailed report; it’s not like the voicemail said “The email that I sent you earlier was mistaken!” or anything like that. But UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING entails that the voicemail is an undercutting defeater for the evidence that the email provides for

FIRMATION

17

I’m assuming here that if p and q are logically equivalent, then p is a hybrid opposing defeater iff q is a hybrid opposing defeater, but I think that’s overwhelmingly plausible here. 18 By “pure,” I mean to refer to defeaters that aren’t also hybrid undercutting defeaters.

232

M. Kotzen

rain. When we ignore the voicemail, the email is good evidence for rain, since I know that Rex is reliable about the weather. Suppose that the email justifies a credence in rain of .9. But when we assume the voicemail as background information, the email is no longer any evidence for rain; if I’ve already heard the voicemail and thus already increased my credence in rain to .9, the email is completely redundant and thus doesn’t do anything to further confirm the hypothesis that it’s going to rain. Let E be the email, let H be the hypothesis that it’s going to rain tomorrow, and let D be the voicemail. Then, dc(E, H, K) is high, since the email is good evidence for rain relative to my background information K. But dc(E, H, K & D) is 0; once I’ve already taken D into consideration, E does nothing further to confirm H.19 So UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING’s condition is easily satisfied; if UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING accurately characterizes undercutting defeat, this entails that the voicemail is an undercutting defeater for the evidence for rain provided by the email, which is counterintuitive. What should we say about this sort of case? One option, of course, is to just accept that redundant evidence does undercut, since it reduces the amount of confirmation that the original evidence confers on the hypothesis. Another is to try to formally distinguish undercutting from redundancy. I’ve tried a number of different ways that this latter project might go; unfortunately, I do not have space here to fully develop the various possibilities or their advantages and liabilities. One possible fix is to modify UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING by adding in a term to the right-hand side of the inequality which quantifies the extent to which D confirms H in virtue of being redundant of E. When there is no redundancy between E and D this term should simply go to 0, whereas when there is “complete” redundancy (as in the Rex case) this term should be equal to dc(E, H, K) itself, with the value of this term monotonically increasing as the extent of the redundancy increases. One (though certainly not the only) way to implement this is to compare the values of p(E| H & D) and p(E|H) to characterize the extent to which D is redundant of E (on the assumption of H ). When D is not at all redundant of E (on the assumption of H ), p(E| H & D) ¼ p(E|H), whereas when D is totally redundant of E (on the assumption of H ), p(E| H & D) ¼ 1. So one natural  way to characterize the pðE j H&D pðEjH Þ

extent of D’s redundancy of E (assuming H ) is ; this term can be 1pðEjH Þ understood to quantify the fraction of the distance between p(E|H) and 1 that D increases E’s probability (on the assumption of H ). NEW PROPOSAL: D is an undercutting defeater for the evidence that E provides for H (relative  to background information K ) just in case dc(E, H, K) > dc(E, H, K & D) þ  dc(E, H, K).

pðEj H&D pðEjH Þ 1pðEjH Þ

19 If dc is a relevance measure (which I’ve been assuming it is), then dc(E, H, K & D) ¼ 0 when p(H| D) ¼ p(H| E & D). But it’s clear here that p(H|D) ¼ p(H| E & D) ¼ .9.

14

A Formal Account of Epistemic Defeat

233

In the original rain case where Julio’s testimony that it’s raining and my learning that Julio is an unreliable testifier about the weather are probabilistically independent (on the assumption of rain), p(E| H & D) ¼ p(E|H), so the third term in NEW PROPOSAL goes to 0, and NEW PROPOSAL makes the same predictions as UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING. In the Rex case, the voicemail saying that it’s going to rain tomorrow makes the email saying that it’s going to rain tomorrow certain, so p(E| H & D) ¼ 1, so the third term is equal to dc(E, H, K). The second term, dc(E, H, K & D), equals 0, since the email doesn’t confirm rain on the assumption of the voicemail as background information (which is the whole reason that we’re now looking for an alternative to UNDERCUTTING IFF DEGREE-OF-CONFIRMATION LOWERING). So NEW PROPOSAL entails that there is undercutting defeat just in case dc(E, H, K) > dc(E, H, K), which of course is false, so NEW PROPOSAL entails that there is no undercutting in the Rex case, as desired. However, NEW PROPOSAL may stumble with certain cases where D has both a redundancy effect on E and also an undercutting effect. For example, if we let D ¼ E & U, where U is some genuine undercutting defeater, then NEW PROPOSAL looks to wrongly entail that D doesn’t undercut. Since D ¼ E & U, p(E| H & D) ¼ 1, so the third term in NEW PROPOSAL becomes dc(E, H, K). And because D entails E, dc(E, H, K & D) ¼ 0. So NEW PROPOSAL’s condition again becomes dc(E, H, K) > dc(E, H, K), which is never satisfied, so NEW PROPOSAL entails that there is no undercutting defeat. But, intuitively, D ¼ E & U is an undercutting defeater here, since it entails some information (i.e., U ), which is by hypothesis undercutting. I leave the project of assessing and addressing this worry to future work.

14.7

Conclusion

On this approach, we get: In this chapter, I have tried to give a framework for analyzing how notions of evidential defeat that have been deployed mostly in the context of binary belief should be generalized to partial belief contexts. Part of my goal has been taxonomic, but I don’t think that the primary issue here is just the terminological one of whether this or that piece of information deserves classification as an “undercutting” or “opposing” defeater. Rather, I think that these notions of defeat play important roles in our epistemological theorizing, and I hope that I’ve made some progress on tracing the contours of these notions. My approach has been unabashedly Bayesian, and though I share the skepticism of some philosophers about how much of our epistemic lives can be modeled in standard Bayesian terms, I do think that Bayesianism has had some truly remarkable successes and that there is a lot to be learned from seeing how many of our epistemic concepts and norms can be modeled in Bayesian terms. To the extent that the proposals above fall short, I hope that their shortcomings will give us some useful clues about which Bayesian or non-Bayesian approaches we might fruitfully pursue with regard to evidential defeat in the future.

234

M. Kotzen

References Chisholm, R. (1989). Theory of knowledge (3rd ed.). Englewood Cliffs: Prentice-Hall. Christensen, D. (2007a). Epistemic self-respect. Proceedings of the Aristotelian Society, 107, 319–337. Christensen, D. (2007b). Does Murphy’s law apply in epistemology? Self-doubt and rational ideals. Oxford Studies in Epistemology, 2, 3–31. Christensen, D. (2007c). Epistemology of disagreement: The good news. Philosophical Review, 116(2), 187–217. Christensen, D. (2009). Disagreement as evidence: The epistemology of controversy. Philosophy Compass, 4(5), 1–12. Christensen, D. (2010). Higher-order evidence. Philosophy and Phenomenological Research, 81 (1), 185–215. Eells, E., & Fitelson, B. (2000). Comments and criticism: Measuring confirmation and evidence. Journal of Philosophy, 97, 663–672. Elga, Adam. (Unpublished Manuscript). Lucky to be rational. Available at https://www.princeton. edu/~adame/papers/bellingham-lucky.pdf Fitelson, B. (1999). The plurality of Bayesian measures of confirmation and the problem of measure sensitivity. Philosophy of Science, 66(Supplement), S362–S378. Kelly, T. (2010). Peer disagreement and higher-order evidence. In R. Feldman & T. A. Warfield (Eds.), Disagreement (pp. 111–174). Oxford: Oxford University Press. Klein, P. D. (1971). A Proposed definition of propositional knowledge. Journal of Philosophy, 68 (16), 471–482. Klein, P. D. (1976). Knowledge, causality, and defeasibility. Journal of Philosophy, 73(20), 792–812. Lasonen-Aarnio, M. (2014). Higher-order evidence and the limits of defeat. Philosophy and Phenomenological Research, 88(2), 314–345. Milne, P. (1996). log[Pr(H|E\B)/Pr(H/B)] Is the one true measure of confirmation. Philosophy of Science, 63, 21–26. Pollock, J. L., & Cruz, J. (1999). Contemporary theories of knowledge (2nd ed.). Lanham: Rowman & Littlefield. Pollard, S. (1999). Milne’s measure of confirmation. Analysis, 59, 335–338. Pryor, J. (2013). Problems for credulism. In C. Tucker (Ed.), Seemings and justification: New essays on dogmatism and phenomenal conservatism (pp. 89–131). Oxford: Oxford University Press. Pryor, J. (Manuscript). Uncertainty and undermining. Available at http://www.jimpryor.net/ research/papers/Uncertainty.pdf Schechter, J. (2011). Rational self-doubt and the failure of closure. Philosophical Studies, 163(2), 429–452.

Chapter 15

Benign Infinity Matthias Steup

Abstract According to infinitism, all justification comes from an infinite series of reasons. Peter Klein defends infinitism as the correct solution to the regress problem by rejecting two alternative solutions: foundationalism and coherentism. I focus on Klein’s argument against foundationalism, which relies on the premise that there is no justification without meta-justification. This premise is incompatible with dogmatic foundationalism as defended by Michael Huemer and Jim Pryor. It does not, however, conflict with non-dogmatic foundationalism. Whereas dogmatic foundationalism rejects the need for any form of meta-justification, non-dogmatic foundationalism merely rejects Laurence BonJour’s claim that meta-justification must come from beliefs. Unlike its dogmatic counterpart, non-dogmatic foundationalism can allow for basic beliefs to receive meta-justification from non-doxastic sources such as experiences and memories. Construed thus, non-dogmatic foundationalism is compatible with Klein’s principle that there is no justification without meta-justification. I conclude that Klein’s rejection of foundationalism fails. Nevertheless, I agree with Klein that when in response to a skeptical challenge we engage in the activity of defending our beliefs, the number of reasons we can give is at least in principle infinite. I argue that this type of infinity is benign because, when we continue to give reasons, we will eventually merely repeat previously stated reasons. Consequently, I reject Klein’s claim that the more reasons we give the more we increase the justification for our beliefs. Keywords Infinitism · Klein · Non-dogmatic foundationalism · Reasons for reliability · Benign regress · Metajustification · Foundationalist infinitism · Internalist reliabilism · Epistemology

M. Steup (*) Department of Philosophy, University of Colorado Boulder, Boulder, CO, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7_15

235

236

M. Steup

15.1

Three Types of Regress

According to Peter Klein, infinitism is the correct solution to the regress problem.1 The problem arises when we consider an inferentially justified belief, b1, and ask how its justification comes about. Since b1 is inferentially justified, it received its justification from another belief, b2. But only justified beliefs can transmit justification via inference to other beliefs. So, if b1 receives its justification from b2, then b2 is justified in turn. Unless b2 is justified non-inferentially, its justification comes from a third belief, b3. Unless b3 is justified non-inferentially, its justification comes from a still further belief, and so forth. About the chain of beliefs on which b1’s justification rests, there are the following options: (i) (ii) (iii) (iv)

it terminates in an unjustified belief it terminates in a non-inferentially justified belief forming a circle, it loops back to b1 it continues ad infinitum

Each of these options is puzzling: (i) because it seems impossible for a belief’s justification to come ultimately from an unjustified belief; (ii) because non-inferential justification is mysterious; (iii) because it seems impossible that a belief can receive its justification ultimately from itself, even if there are many intermediate links, and (iv) because it’s hard to see how finite minds can help themselves to infinite chains of reasons. The problem is that if indeed each of options (i)–(iv) is unacceptable, then the allegedly justified belief we started out with turns out to be unjustified. Deservedly, option (i) has received little support, if any. Foundationalists recommend option (ii) and coherentists option (iii). Klein defends infinitism on the ground that neither foundationalism nor coherentism can solve the regress problem. I will focus on Klein’s rejection of foundationalism and argue that it doesn’t succeed. I will also argue in favor of a certain kind of infinitism. Initially, it seems that foundationalism and infinitism are incompatible. However, the appearance of incompatibility disappears when we recognize that there are three different types of regress. The correct response to one of them need not be the correct response to the others. First, there is the doxastic regress: it consists solely of an initial, inferentially justified belief b1 and the further beliefs b2,. . ., bn from which b1 is inferred. Here is an example of a two-step doxastic regress: b1: The dog ate my cookie. b2–4: The dog is on the couch, the dog looks guilty, and my cookie is gone.

Beliefs b2–b4 are the premise beliefs from which b1 is inferred. The regress is doxastic because it consists solely of beliefs.

1

See Klein (1999, 2010, 2014a, b). See also Ginet (2014a, b).

15

Benign Infinity

237

Second, there is what we may call the evidential regress. If we reject the circle-ofbeliefs thesis and agree that beliefs are not the only things that can be reasons or evidence, then we will allow for two kinds of reasons: doxastic and non-doxastic.2 Non-doxastic reasons are perceptual, introspective, memorial, and intuitional experiences or seemings.3 The example above of a doxastic regress turns into an evidential regress when we add the perceptual experiences that may be viewed as the reasons that justify beliefs 2–4: e1–3: Visual experiences with the content that the dog is on the couch, that the dog looks guilty, and that the plate that had my cookie on it is now empty.

Unlike the previous doxastic regress, this enlarged regress includes both beliefs and visual experiences. Finally, there is what we may call the argumentative regress. Whereas an evidential regress consists of a series of reasons a subject has, an argumentative regress is made up of reasons a subject gives to defend a belief that’s been challenged. Obviously, in ordinary cases there is nothing argumentative about the other two types of regress, which merely trace where—whenever a belief is justified—its justification comes from. An argumentative regress arises only when a subject engages in a certain activity—that of giving reasons—in response to a request for providing a justification. I will argue that, whereas foundationalism is correct about the doxastic and the evidential regress, infinitism is correct about the argumentative regress.4

15.2

Basic Beliefs

To assess whether foundationalism is a successful solution to any of the three regress types mentioned above, we had better take a closer look at what kind of a view foundationalism is supposed to be. Here is a textbook version of it. According to foundationalism, all inferentially justified beliefs owe their justification to a foundation of basic beliefs that receive their justification not from any further beliefs but

Davidson claims that “nothing can count as a reason for holding a belief except another belief. The circle-of-belief thesis’s partisan rejects as unintelligible the request for a ground or source of justification of another ilk. See Davidson (1983, 426). For discussion, see Pryor (2014, 206ff) and Steup (1996, 140f). 3 An experience or seeming that p, Sp, is a reason for believing that p, Bp, because Sp has the same content as Bp, and because Sp asserts its content assertively, as compared with, for example, imagining that p or wishing that p (see Huemer 2001, 99f). Why think that experiences can terminate the regress? They can do so because they can confer justification without needing any justification themselves (see Steup 1996, Chap. 7). 4 The distinction between the doxastic and the evidential regress on the one hand, and the argumentative regress on the other hand, is closely related to Alston’s distinction between a belief’s property of being justified and the activity of justifying a belief (Alston 1989, 82). 2

238

M. Steup

instead from experiences.5 According to foundationalism thus understood—experiential foundationalism, as we might call it—basic beliefs can be defined as follows: A belief b is basic iff (i) b is justified and (ii) b does not depend for its justification on any other beliefs.6

Advocates of experiential foundationalism must explain how it’s possible for an experience to justify a belief. I will address this issue further below. They must also respond to Laurence BonJour’s well-known argument for the conclusion that there aren’t any basic beliefs (see BonJour 1985, 30f). Since this argument is closely related to Klein’s argument for infinitism, it will be fruitful to examine it. I will follow Richard Feldman’s clear presentation of it, and I will also follow Klein in introducing a scenario involving two protagonists: foundationalist Fred, and Doris the doubter (Feldman 2003, 75f; Klein 2014a, 276). Fred, let us suppose, believes bATE IT: The dog ate my cookie.

Doris challenges Fred to explain why he thinks that bATE IT is justified. Fred responds by saying that he inferred bATE IT from three other propositions he believes: the dog is on the couch, the dog looks guilty, and the plate that had Fred’s cookie on it is now empty. Suppose Doris selects the first of these beliefs for further questioning and asks Fred why he thinks that the following proposition is justified: bDOG: The dog is on the couch.7

Fred replies that bDOG is basic; it is justified not by further beliefs but instead solely by: eDOG: A visual experience with the content that the dog is on the couch.

At this point, BonJour’s argument against basic beliefs kicks in. The key thought of his argument is that it’s not possible for eDOG by itself to justify bDOG. A justificatory contribution from further beliefs is needed.

15.3

BonJour’s Argument against Basic Beliefs

BonJour’s argument can be construed as a sequence of two arguments. The first rests on the thought that from the point of view of the subject, justification must be a path to truth: a belief is justified if, and only if, given what the subject has to go on, the 5 For examples of this type of view, see Audi (1993), Chisholm (1966), Feldman (2003, 70ff), Pollock (1986 Chap. 2), Huemer (2001 Chap. 5), and Pryor (2000). For a recent collection of essays discussing experiential foundationalism, see Tucker (2013). 6 Feldman (2003, 50) defines basic beliefs as beliefs that are “justified but . . . not on the basis of any other beliefs.” 7 Of course, she could instead have chosen one of the other two premise beliefs, or asked Fred to justify all three. I’ll focus on bDOG, which is why I didn’t introduce ‘bGUILTY’ and ‘bEMPTY’ as names for Fred’s other two premise beliefs.

15

Benign Infinity

239

belief is probably true, that is, has a probability of being true that is greater than .5 (see BonJour 1985, 8).8 So, if Fred claims that bDOG is justified in the basic way— that is, without owing its justification to any other beliefs—he must argue that bDOG has a certain truth-indicative property, P, by virtue of which this belief is probably true. The pattern for such an argument is as follows: THE TIP ARGUMENT (1) bDOG has a truth-indicative property P. (2) Beliefs having property P are probably true. ∴ bDOG is probably true. The second argument is based on the thought that, since justification is supposed to be a path to truth, having a justified belief requires having a TIP argument. A TIP argument provides what we may call a ‘meta-justification’. Fred tells Doris that eDOG is the reason that justifies bDOG. At this point, as Klein puts it, Doris goes meta and ask Fred why he thinks that eDOG is a reason for bDOG (2014a, 277). In response, Fred might argue that eDOG has the property of being a clear and vivid visual experience, and that beliefs based on experiences having that property are probably true. According to BonJour’s argument, if Fred provides such a meta-justification for his claim that is eDOG a reason for bDOG, he admits that bDOG is in fact not basic. BonJour’s second argument, then, runs thus: BONJOUR’S META-JUSTIFICATION ARGUMENT AGAINST BASIC BELIEFS (4) All justification requires having a TIP argument. (5) If (4) is true, then bDOG is not basic. ∴ bDOG is not basic. Since BonJour’s meta-justification argument can be applied to any putatively basic belief, we get the general conclusion that basic beliefs are impossible. The key thought on which BonJour’s argument is based is what we may call: THE PRINCIPLE OF META-JUSTIFICATION: There is no justification without meta-justification.

Klein, too, endorses the principle of meta-justification, though not in quite the same way as BonJour does. Next, I’ll turn to Klein’s rejection of foundationalism, and then I’ll defend foundationalism against both BonJour and Klein.

8

BonJour says that a justified belief is one that is highly likely true, but this claim conflicts with the rather basic point that justification comes in degrees, which is to say that a belief can be justified— albeit to a low degree—if the subject’s reasons make it somewhat likely that it is true.

240

15.4

M. Steup

Klein’s Rejection of Foundationalism

Klein bases his rejection of foundationalism on the dialectic that ensues in the dialogue between Doris and Fred. Essentially, his argument mirrors BonJour’s: when Doris goes meta, Fred must accede to her demand for meta-justification and therefore give up his claim that bDOG is basic, thus abandoning foundationalism (Klein 2014a, 276f). Doris goes meta when Fred asserts that eDOG is a reason for bDOG. She asks Fred why he thinks that basing bDOG on eDOG is truth-conducive. In other words, she wants to know from Fred whether eDOG has a property that makes it probable that bDOG is true. As an experiential foundationalist, Fred will answer Doris’s question by offering the following meta-justification: MJ: eDOG has the property of being a clear and vivid visual experience, and beliefs based on such experiences are probably true.9

In response to Doris’s request for a meta-justification, Klein would say Fred faces a trilemma. He has three options: (i) assert MJ, (ii) deny MJ, or (iii) suspend judgment about MJ. Options (ii) and (iii) are bad. If Fred were to exercise either one of them, he would in effect be saying that he takes eDOG to be a reason for bDOG although he has no justification at all for thinking that basing bDOG on eDOG is truth-conducive. That looks like a breakdown of epistemic rationality. Hence Klein dismisses options (ii) and (iii), taking them to induce a kind of arbitrariness that’s incompatible with the assumption that bDOG is indeed a justified belief. I think Klein is clearly right about that. So unless Fred is prepared to abandon his claim that eDOG is his reason for bDOG, he had better exercise option (i) and assert MJ. However, Klein would argue, if Fred exercises option (i), he must then retract his claim that bDOG is basic. Next, I will argue that neither BonJour’s nor Klein’s attempted refutation of foundationalism succeeds.

15.5

Foundationalism Defended

As I mentioned above, BonJour’s anti-foundationalist argument is based on the principle of meta-justification: there is no justification without meta-justification. It’s important to see that there are two ways to understand this principle. The first is doxastic, and that’s the reading BonJour has in mind:

9 Like BonJour, Klein argues that if a belief is basic, it must have a special property because of which it is highly likely to be true (Klein 2014a, 276f). He does not, however, focus on a specific instance of that property. In exploring what transpires when Fred responds to Doris’s request for meta-justification, I let Fred respond as I think he should, namely by offering what I consider a plausible candidate for the special property in question: eDOG’s being a visual experience that is clear and vivid.

15

Benign Infinity

241

THE PRINCIPLE OF DOXASTIC META-JUSTIFICATION (DMJ): There is no justification without beliefs providing meta-justification.

If DMJ is true, then basic beliefs are indeed impossible. Hence, to defend foundationalism against BonJour’s argument, DMJ must be rejected. And rejecting it is well advised, since DMJ is a recipe for radical skepticism. First, since metajustificatory beliefs must themselves be justified, it generates the consequence that justification requires an infinite number of beliefs.10 Second, in ordinary situations, people just don’t have any meta-justificatory beliefs (see Ginet 2014a, 287). So if DMJ were true, then ordinary perceptual beliefs would be unjustified. That is why BonJour’s argument doesn’t amount to a plausible challenge to foundationalists. Consider the first premise of his argument: (4) All justification requires having a TIP argument. BonJour insists on the following strong reading of (4): (4A) All justification requires believing the premises of a TIP argument.

If (4) must be understood as (4A), then the following is clearly true: (5A) If (4A) is true, then bDOG is not basic.

However, for the reasons mentioned above, (4A) is not plausible (see Feldman 2003, 77). Now, it might be thought that when foundationalists reject (4A), they must insist that justification does not require meta-justification in any form. But that thought would be a mistake. There is an interpretation of the principle of meta-justification that allows for the existence of basic beliefs, namely: THE PRINCIPLE OF EVIDENTIAL META-JUSTIFICATION (EMJ): There is no justification without possessing evidence providing meta-justification.

Accordingly, (4) can be read alternatively as: (4B) All justification requires possessing evidence for the premises of a TIP argument.

If premise (4) is understood as (4B), the next premise must be: (5B) If (4B) is true, then bDOG is not basic.

But (5B) is false. If the needed evidence for the TIP argument comes solely in the form of perceptual, introspective, memorial, and intuitional experiences, Fred’s meta-justification for bDOG doesn’t involve any beliefs and thus is compatible with bDOG’s being basic. EMJ, therefore, poses no threat to the foundationalist claim that inferentially justified beliefs receive their justification ultimately from basic beliefs. Next, I will respond to Klein’s rejection of foundationalism. Like BonJour’s, it can be seen as a sequence of two arguments. The first is Klein’s trilemma argument: if Fred wants to avoid unacceptable arbitrariness, he must oblige Doris’s demand for a meta-justification and explain why he takes eDOG to be truth-conducive. According to the second argument, defending the truth-conduciveness of eDOG is incompatible with bDOG’s being basic: 10 For the infinite mind objection, see Audi (1993, 127–128), Huemer’s essay in this volume, and Klein (2014a, 281).

242

M. Steup

KLEIN’S META-JUSTIFICATION ARGUMENT AGAINST BASIC BELIEFS (1) Fred asserts MJ: eDOG has the property of being a clear and vivid visual experience, and beliefs based on such experiences are probably true. (2) If Fred asserts MJ, then bDOG is not basic. ∴ eDOG is not basic. Why think that (2) is true? Klein might say that by asserting MJ, Fred concedes that bDOG is in part justified by three additional beliefs: (i) eDOG is a reason for bDOG, (ii) eDOG is a clear and vivid visual experience, and (iii) clear and vivid visual experiences are truth-conducive. Now, it would indeed be odd of Fred to tell Doris that even though he responds to her request for a meta-justification by offering (i)– (iii), he doesn’t really believe these propositions. Here, Klein’s trilemma argument applies. If Fred didn’t believe (i)–(iii), he would be guilty of an unacceptable kind of epistemic irrationality. However, from the fact that Fred believes (i)–(iii) after Doris’s request for a meta-justification, it doesn’t follow that he already had these beliefs before Doris’s request. Let’s distinguish between two times: tbefore and tafter. The former is the time period between the formation of bDOG and Doris’s request for a meta-justification. The latter is the period during which the dialogue takes place. At tbefore, Fred did not yet have the meta-justificatory beliefs (i)–(iii). So Fred can reasonably claim that, at tbefore, bDOG was basic: justified without receiving justification from any further beliefs. Hence, even if Klein succeeded in showing that, at tafter, bDOG isn’t basic, his argument wouldn’t show that bDOG wasn’t basic at tbefore.11 And Fred could further defend this claim by pointing out that, for the sake of avoiding an unacceptable kind of skepticism, we should reject the thought that our perceptual beliefs are only justified when we verbally defend the premise that our perceptual experiences are a source of justification. Alternatively, Klein might defend (2) by endorsing the evidentialist reading of the principle of meta-justification. He might argue that when Fred asserts MJ, he admits that, in addition to eDOG, he has further reasons—not beliefs but experiences—from which his justification for bDOG derives. If that is what Klein were to say, I would be in fundamental agreement with him. However, from the premise that Fred’s justification for bDOG doesn’t come from eDOG alone but derives from additional perceptual, introspective, memorial, and intuitional experiences providing Fred with reasons for believing that eDOG makes it highly likely that bDOG is true,12 we don’t get the conclusion that bDOG isn’t basic. We get that conclusion only if it’s impossible to have a meta-justification without having at least one meta-justificatory belief.

11

It is not obvious that bDOG isn’t basic even at tafter. Fred’s justification for bDOG does not depend on providing Doris, or anybody for that matter, with an argument for why eDOG is a reason for bDOG. Hence, if foundationalists define basic beliefs in the way I have in this chapter, bDOG remains basic even after Fred has provided Doris with a meta-justification for it. 12 Note that this premise doesn’t require Fred to form the belief that eDOG makes it highly likely that bDOG is true.

15

Benign Infinity

243

The evidential version of the principle of meta-justification doesn’t require any such belief. Appealing to EMJ is not, therefore, an effective way of arguing against the possibility of basic beliefs. Although I do not think Klein’s reasoning against foundationalism succeeds, I nevertheless think it supports an important and rather significant conclusion. What the dialogue between Doris and Fred shows is that the need for meta-justification is not easily dismissed. In other words, Klein’s argument provides strong support for EMJ: the evidential interpretation of the principle that there is no justification without meta-justification. But if EMJ is true, then what follows isn’t that foundationalism is wrong. Rather, what then follows is that dogmatic foundationalism is wrong.

15.6

Meta-justification and Dogmatic Foundationalism

Michael Huemer and Jim Pryor have defended versions of foundationalism that are versions of a view that is now called dogmatism (see Pryor 2000 and Huemer 2001).13 According to this view, having an experience as of p is sufficient for having defeasible justification for believing p.14 Having a meta-justification—a reason for thinking that the experience as of p is a reason for believing p—is not needed. The hallmark of dogmatic foundationalism, then, is the rejection of the principle of metajustification in any form. About Fred’s belief that the dog is on the couch, dogmatists would say that eDOG by itself is sufficient for making bDOG justified. When responding to Doris’s request for a meta-justification, Fred might assert MJ. But, dogmatists would say, Fred’s justification for bDOG depends neither on his believing MJ, nor on his having evidence for MJ in the form of additional perceptual, introspective, memorial, or intuitional experiences (see Ginet 2014a, 296). Klein’s objection to foundationalism is based on the premise that if Fred doesn’t accede to Doris’s demand for a meta-justification, the result is a kind of arbitrariness

Huemer calls his view ‘phenomenal conservatism’, but it is essentially the same view Pryor defends. In the Klein-Ginet debate on infinitism, Ginet too defends dogmatic foundationalism (Ginet 2014a, 285f). 14 For a proper understanding of this claim, it is essential to bear in mind the difference between propositional and doxastic justification. The claim is not, using the example of Fred’s belief, that eDOG is sufficient for Fred’s belief to be justified. Rather, the claim is that eDOG is sufficient for Fred to have justification for believing that the dog is on the couch. If eDOG is internally defeated (perhaps because Fred knows he has a brain lesion causing frequent dog hallucinations), then according to dogmatic foundationalism, Fred has justification for believing the dog is on the couch while, at the same time, if Fred were to believe the dog is on the couch, his belief would not be justified. Here are two passages showing that, according to Huemer and Pryor, an experience as of p is sufficient for having defeasible justification for believing p. Huemer: “If it seems to S as if P, then S thereby has at least prima facie justification for believing that P” (2001, 99). Pryor: “My view is that whenever you have an experience as of p, you thereby have immediate prima facie justification for believing p” (Pryor 2000, 536). 13

244

M. Steup

that is inconsistent with Fred’s claim that, as a justified belief, bDOG is probably true. Now, while none of Klein’s arguments suggest that he endorses the doxastic reading of the principle of meta-justification, DMJ, his argument against foundationalism clearly indicates that he should accept the evidential version of the principle, EMJ. This principle delivers exactly what Klein demands: If bDOG is indeed justified, then Fred has evidence available to him—in the form of suitable experiences—to which, when Doris goes meta, he can appeal to justify his claim that eDOG is a reason for bDOG. According to dogmatic foundationalism, Fred’s justification for bDOG doesn’t require such evidence. If Klein is right and Fred’s belief isn’t justified unless Fred has available to him reasons for the meta-justification Doris demands, then dogmatic foundationalism is mistaken. So, Klein’s argument against foundationalism is really an argument against dogmatic foundationalism. And on this score, I find myself in complete agreement with him. I think Klein is right in insisting that first-order justification requires the possession of reasons that a subject can appeal to when, in a dialogue like the one between Fred and Doris, a meta-justification is requested. And I do not think Klein would want to insist that this requirement is doxastic—that first-order justification requires meta-justificatory beliefs. It seems to me, therefore, that Klein should reject dogmatic foundationalism, but not non-dogmatic foundationalism. Oddly, in responding to Ginet’s claim that a visual experience can by itself—i.e., autonomously—justify the corresponding visual belief (Ginet 2014a, 285f), Klein appears to concede as much, in effect endorsing dogmatic foundationalism. Listing various areas of agreements with Ginet, Klein says, “In a slightly extended use of ‘reason’ such things as perceptual states, memories, or understanding the meaning of an expression are reasons that can make a belief at least partially justified” (Klein 2014b, 292). Here, Klein agrees with Ginet that a visual experience can confer at least some degree of autonomous justification for the corresponding visual belief.15,16 My response to this concession is twofold. First, I think Klein is right 15

Epistemologists sometimes speak, as does Klein in the cited passage, of a belief as being partially justified. I have trouble discerning what ‘partial justification’ is supposed to mean. There is a truism about justification: it comes in degrees. Beliefs like “2 + 2 ¼ 4” or “I exist” have the highest degree: complete certainty. If we identify fully justified with having the highest degree of justification, then all beliefs less than certain are merely partially justified. That would be an odd outcome, particularly since knowledge, not requiring certainty, would require merely partial, not full, justification. Perhaps the idea is that any belief whose degree of justification falls short of being knowledgegrade is merely partially justified. But that, too, would be strange. In general terms, the problem is that if justification comes in degrees, every justified belief is fully justified to degree n. (Consider having a fever, which comes in degrees. It makes little sense to speak of a ‘partial fever’.) Thus, when Klein concedes that experiential states can make a belief at least ‘partially justified’, I take him to mean that such states can provide at least some degree of justification. 16 It might be objected that Klein means to accept merely the weaker claim that experiential states can be reasons, without conceding the stronger claim that they can provide autonomous justification (justification independent of any meta-justification). Alas, in that case, what Klein lists as an area of agreement would be primarily an area of disagreement. For on page 285 (of Ginet 2014a), the page to which Klein refers, Ginet defends foundationalism by insisting that states such as perceiving something or understanding the meaning of a proposition can justify autonomously. If Klein were to clarify his position by saying that he merely agrees that experiences can function as reasons, without

15

Benign Infinity

245

in accepting that visual experience can be a source of justification, for otherwise he’d be endorsing the highly problematic circle-of-belief thesis. Second, I do not think he should agree that a visual experience can be a source of at least some degree of autonomous justification. By ‘autonomous’ justification, Klein means the kind of justification for a visual belief that does not require reasons for thinking that the visual experience makes the visual belief probable. His response to Ginet is that the infinitist does not deny the possibility of autonomous justification, but instead insists that a subject has stronger justification for a visual belief if she has reasons for the truth-conduciveness of the corresponding visual experience. It seems to me this response gives up too much. It concedes that infinitism is correct only as an account of how to increase already-existing justification that is in no need of metajustification at all. This is an odd concession because it conflicts with the basic point that if bDOG is justified, then Fred must already possess whatever evidence is needed for meeting Doris’s demand for meta-justification. Perhaps Klein’s concession is motivated by the need to acknowledge that the human mind is finite, incapable of entertaining an infinite number of reasons. Accordingly, Klein says infinitism demands merely the availability of an infinite number of reasons (2014a, 279). But what does availability amount to? The challenge is obvious: if construed too stringently, infinitism succumbs to the finite mind objection; if construed too loosely, it’s difficult to see why meta-justificatory reasons are necessary at all. Klein sheds some light on the meaning of ‘available’ by saying that for reasons to be available at a given time t, they “must be appropriately ‘hooked up’ to S’s beliefs and other mental contents at t,” and he emphasizes that for a reason to be available in this sense, it needn’t be an occurrently held belief (2014a, 279). However, are the meta-reasons that Klein thinks justification requires—that is, the reasons Fred needs if he is to meet Doris’s demand for a meta-justification— available to Fred if he has no evidence at all for taking eDOG to be truth-conducive? I think not. For the needed reasons to be available to Fred when Doris goes meta, he must already be in possession of evidence for the truth-conduciveness of clear and vivid visual experiences. Suppose Klein agrees with this. In that case, he must concede that experiences can be reasons, but he need not concede that experiences can be autonomous reasons.

15.7

Infinitism, Inference, and the Creation Problem

Klein defends infinitism by arguing that non-inferential justification—a belief’s being basic—is impossible. Hence, Carl Ginet points out, infinitism as defended by Klein is the view that all justification is inferential (Ginet 2014a, 284). Klein conceding that they can function as autonomous reasons, then my response would be twofold: First, as an infinitist, that’s exactly what he ought to be saying. Second, the point in question would primarily be an area of disagreement between him and Ginet.

246

M. Steup

agrees (Klein 2014b, 294).17 However, I do not think that, qua being an infinitist, Klein is committed to the claim that non-inferential justification is impossible. Suppose Fred’s justification for bDOG does indeed come from a large set of experiential states that originate in perception, introspection, memory, and intuition. Let’s suppose this series is at least potentially infinite. Each time Fred attempts to give a reason in response to Doris’s recurring requests for justification, a new experiential state is formed and supplies Fred with a further reason. Suppose if Fred were a being with an infinite amount of time and patience then he could go on forever obliging Doris with further reasons. Would it follow that all justification is inferential? It doesn’t seem to me that it would. Consider again Fred’s belief at tbefore. If infinitism as just outlined is correct, then Fred’s justification at tbefore for bDOG comes from a finite set of experiences to which Fred, if he had an infinite amount of time, could add infinitely many further experiences.18 Why should we think that because of that possibility, Fred’s justification for bDOG at tbefore is inferential? From the fact that at tbefore Fred has reasons for the premises of (perhaps infinitely many) meta-justificatory arguments, it doesn’t follow that at tbefore he has silently rehearsed or verbally articulated any of these arguments. It’s one thing to have reasons for the premises of meta-justificatory arguments, and it’s another to articulate these arguments and infer their conclusions. While Fred offers some such arguments at tafter, he entertained none at tbefore. At that time, his justification for bDOG was not the result of inference. Hence, even if infinitism were true, it wouldn’t follow that all justification is inferential. It might be objected that, contrary to what I am claiming, non-dogmatic foundationalism makes all justification inferential. My response is that there is a difference between (a) possessing evidence for the premises of an inference and (b) making an inference. At tbefore, Fred already has all the evidence needed for the meta-justificatory inference that his justification for bDOG comes from a truthconducive source, namely eDOG. But at tbefore, Fred hasn’t yet made any such inference. Yet bDOG is justified at tbefore. Hence, at tbefore, Fred’s justification for bDOG is non-inferential. The objection, then, employs an odd understanding of ‘inferential’. As long as bDOG is justified without being inferred from at least one other belief, there is no basis for claiming its justification is inferential. In defending infinitism, Klein claims that all justification is inferential. Citing Jonathan Dancy, Ginet responds that if all justification is indeed inferential, then justification is impossible because inference does not create justification—it merely transfers it (see Ginet 2014a, 290; Dancy 1985, 50). Klein replies that justification

In a 2010 paper, Klein says, “In other words, infinitism holds that all propositional knowledge is inferential” (161). 18 At each of these times, the number of his experiences would still be finite. And, we might add, the number of Fred’s presently occurring experiential states might not swell excessively because with the passage of time, some previously held experiential states will no longer obtain. 17

15

Benign Infinity

247

emerges “when the set of propositions that are appropriately adduced as reasons expands” (Klein 2014b, 294). The thought is that justification begins emerging when the number of [rational] inferences has reached a critical mass, and that its degree increases the more of these inferences are piled onto each other. Here I agree with Ginet and Dancy. I can’t find any plausibility in the claim that justification emerges from inferences alone, provided there are enough of them. If all justification were indeed inferential, then it would be unclear where Fred’s justification for bDOG comes from at tbefore, because at that time he hasn’t yet made any inferences that bear on the justification of his belief that the dog is on the couch. Can Klein remain an infinitist but allow for the possibility of non-inferential justification? I have argued that he can. As I see it, his primary commitment is to the principle of evidential meta-justification. When Doris demands from Fred a reason for thinking that eDOG is a reason for bDOG, she is asking a legitimate question, and it would be irrational of Fred to say he can’t or won’t answer it. But, as I have argued above, advocating the principle of evidential meta-justification is compatible with bDOG’s being basic or non-inferentially justified. What it is not compatible with is dogmatic foundationalism, according to which eDOG alone justifies bDOG. Now how, according to non-dogmatic foundationalism, is Fred’s justification for bDOG created? In a nutshell, it is created by reasons for taking eDOG to be a reliable or trustworthy belief source—put differently, by reasons for believing that basing bDOG on eDOG is a truth-conducive way of forming a belief. This, it seems to me, is the correct answer to the question of how justification is created.19 It seems to me as well that this answer is compatible with infinitism. What matters is the content of the reasons, not how many there are. If we are to have justification for our beliefs, our reasons must include reasons for the truth-conduciveness of our belief sources. Reasons of this kind create justification for our beliefs. Whether their number is finite or infinite doesn’t matter here.

15.8

Infinity and the Argumentative Regress

Since Klein thinks neither foundationalism nor coherentism is a viable solution to the regress problem, he recommends infinitism as the only workable solution. However, exactly which type of regress—doxastic, evidential, or argumentative— is supposed to be infinite? Clearly not the first. Klein would readily agree that finite

19

For a defense of this view, see Steup (2004) and Steup (forthcoming). A controversy similar to that over dogmatic vs. non-dogmatic theories of justification is the dispute between reductionism and anti-reductionism in the epistemology of testimony. The reductionist position is that testimony fails to have justificatory power unless it is backed up with evidence of trustworthiness. Reductionism, then, can be seen as the view that testimonial justification requires meta-justification. For a defense of reductionism, see Fricker (1994, 1995). For a reductionism-friendly hybrid view, see Lackey (2006, 2008). For a defense of the anti-reductionist view, see Sanford and Henderson (2006).

248

M. Steup

minds cannot have an infinite number of beliefs. When people infer one belief from one or more further beliefs, the set of beliefs this regress consists of is finite, not infinite. Suppose non-dogmatic foundationalism is correct. Fred’s justification for bDOG comes from eDOG and additional experiential states that are Fred’s reasons for taking eDOG to be truth-conducive. Is it plausible to claim that Fred’s justification for bDOG comes from an infinite set of such states? I think what applies to beliefs applies to experiences as well. At a given time, subjects don’t have an infinite set of experiential states. But when, prompted by Doris’s request for a meta-justification, Fred starts contemplating what might justify him in believing that eDOG is a reason for bDOG, then, Klein might argue, a new experiential state might arise each time Fred moves along a further step in the regress of giving reasons. The number of such potential experiential states might in principle be infinite. I will explore this possibility further below. Now consider the argumentative regress. In practice, it cannot be infinite. Sooner or later, other pressing matters, like getting a good night’s sleep, will put a stop to Fred’s attempts to oblige Doris. Klein might reply that if we have access to a potentially infinite number of meta-justificatory reasons, then if Fred and Doris were beings with an infinite amount of time, they could carry on their dialogue forever. And this seems to follow from two premises to which Klein appeals and which I myself find rather plausible: P1: Each time Fred supplies Doris with a further reason, it is legitimate for Doris to ask Fred what it is that justifies him in asserting this reason. P2: Each time Doris asks Fred to justify the reason he just offered, it would be irrational for Fred—that is, inconsistent with his starting point that bDOG is justified—to reply that he doesn’t have a reason.

If P1 and P2 are true, then the conclusion that the argumentative regress is indeed infinite, at least in principle, seems inescapable. Of course, in practice, the activity of reason-giving is bound to come to an end eventually. But that’s consistent with the main point: if P1 and P2 are correct, then the argumentative regress could go on forever, at least in principle. Ginet demurs. He says it’s one thing to claim that there is an at least in-principle available infinite series of reasons, and it’s another thing to give an example of such a regress or an algorithm for generating one (see Ginet 2014a, 290).20 Ginet’s request for an example or an algorithm is fair. In the next section, I’ll attempt to spell out what’s going on when, after Doris has gone meta, the dialogue between her and Fred continues for some distance.

Ginet writes, “To make it plausible that there actually occur justifications having the endlessly ramifying structure that infinitism says all justifications must have, the infinitist must provide representative examples of particular such structures possessed by cognitively normal human subjects, examples about which it would be credible that cases essentially like them actually occur” (Ginet 2014a, 290). 20

15

Benign Infinity

15.9

249

The Introspective Track

Let us review what ensues when Doris goes meta. Fred claims his reason for bDOG is eDOG. Doris makes the meta-move by asking Fred why he thinks eDOG is a reason for bDOG, or, put differently, why he thinks that eDOG makes it likely that bDOG is true. Fred replies that eDOG has the following property: it is a clear and vivid visual experience. In effect, he offers a TIP Argument: A1 (1) bDOG originates in eDOG: a clear and vivid visual experience. (2) Visual experiences of this type are reliable. ∴ bDOG is probably true. Doris can now choose between three different tracks of moving the regress forward: She can ask Fred for either a reason in support of (1), a reason in support of (2), or, finally, a reason in support of why Fred thinks that (3) follows from (1) and (2). I will discuss the first two of these tracks.21 If Doris asks Fred to justify premise (1) of A1, she starts the introspective track. Fred will respond by offering a second TIP argument: AIN1 (1) I am, in a way that’s clear and vivid, introspectively aware that eDOG is a clear and vivid visual experience.22 (2) Introspective experiences of this type are reliable. ∴ It’s highly likely that PREMISE (1) of A1 is true.23 Suppose Doris next asks Fred to justify premise (1) of AIN1. Since Fred knows this premise through introspection, he will offer an argument that repeats the reasoning of AIN1: AIN2 (1) My reason for (1) of AIN1 is clear and vivid introspection. (2) Introspective experiences of this type are reliable. ∴ PREMISE (1) of AIN1 is probably true. Suppose Doris carries on and says to Fred, “You just presented me with a new argument: AIN2. This argument has two premises, each of which requires a reason. 21

I’ll trust readers can easily apply the results of my discussion to the regress that ensues when Doris asks Fred to justify his assumption that the premises of A1 support the conclusion of A1. 22 I’m simplifying a bit. As far as premise (1) of the initial TIP argument, A1, is concerned, Doris can focus on two different questions. She might ask Fred how he knows bDOG is based on (originates in) eDOG, and how he knows that eDOG is a clear and vivid visual experience. I discuss what happens if she asks the second of these questions. The structure of unfolding series of arguments will be the same for the regress that would ensue if Doris were instead to ask the first of these questions. 23 I’m relying here on an additional premise: if a belief originates in a (highly) reliable source, it’s highly likely that the belief is true. Thanks to Cherie Braden for pointing that out.

250

M. Steup

Please give me your reason for the first premise of AIN2.” Well, what is Fred’s reason for claiming he knows through introspection that his knowledge of what his visual experience is like originates in introspection? There is only one thing Fred can say, namely, “I know this through introspection.” Hence, to defend premise (1) of AIN2, he will offer this argument: AIN3 (1) My reason for (1) of AIN2 is clear and vivid introspection. (2) Introspective experiences of this type are reliable. ∴ PREMISE (1) of AIN2 is probably true. AIN2 and AIN3 are identical except for one difference: the first premise of AIN2 refers to the first premise of AIN1, and the first premise of AIN3 refers to the first premise of AIN2. If Doris asks Fred to justify the first premise of AIN3, Fred will repeat the previously given argument, and if Doris ask Fred to justify the first premise of this further argument, he will again repeat the previously given argument, and so forth. Two things are now perfectly clear. First, this pattern is in principle infinite. Second, if Doris and Fred are reasonable, they will recognize that further pursuit of this regress is pointless: it will not deepen their understanding of why Fred is justified in believing, or how he knows, that eDOG, which is a clear and vivid visual experience. The basic point is that he has introspective awareness of what his visual experience is like, and that he is introspectively aware of his introspective awareness. Therefore, although the introspective track of the regress could in principle carry on endlessly, it ceases to produce further justificational juice, which amounts to a kind of termination.24

15.10

The Nature of Introspective Awareness

It might be objected that the introspective track reveals the following: if Fred’s justification for bDOG requires that he have reasons supporting the premises of the initial TIP argument, A1, then the following vicious regress ensues. It must be that Fred: (1) has clear and vivid introspective awareness that eDOG is a clear and vivid visual experience, (2) has clear and vivid introspective awareness that he has clear and vivid introspective awareness that eDOG is a clear and vivid visual experience,

24

Doris might ask Fred why he takes himself to know that clear and vivid introspective experiences are truth-conducive. His reply should be that he remembers that they are. For the sake of illustration, he might say, for example, that he remembers not having been in situations in which he thought he was seeing a dog and in which it turned out that he was not seeing a dog but smelling one (or hearing one, or feeling one with his hands). Put differently, memory tells us that we are highly reliable in identifying a specific perceptual experience as visual, auditory, tactile, or olfactory.

15

Benign Infinity

251

and so forth. In short, for Fred to have justification for PREMISE (1) of A1, he must have an infinite series of layers of introspective meta-awareness. As a finite mind, Fred is not capable of such a feat. Hence, the principle of evidential metajustification has the consequence that Fred’s belief about the nature of eDOG cannot be justified. There is a straightforward response to this worry. Introspective awareness is, as Roderick Chisholm would have put it, self-presenting: it comes with introspective awareness of itself. To know I’m introspecting, I don’t need to meta-introspect. For example, when I have a headache, I know I do because I can feel it. Feeling my headache has awareness built into it. Likewise, when Fred is introspectively aware that he is having a visual experience of the dog—as opposed to an auditory, tactile, or olfactory one—that introspective awareness is self-presenting. Fred does not need an additional introspection to know he is introspectively aware of what kind of an experience eDOG is.25 Let’s review. When Doris asks Fred how he knows eDOG is a clear and vivid visual experience, Fred replies that he knows this through introspection. When Doris asks him how he knows that he knows this through introspection, he will say, “I’m introspectively aware that I’m introspectively aware of it.” Doris might now say, “And how do you know that?” To answer this question, Fred need only repeat his previous answer.

15.11

The Memorial Track

Now suppose Doris next asks Fred to justify the second premise of A1, his initial meta-justificatory argument: (2) Clear and vivid visual experiences are reliable. Fred replies that he remembers that visual experiences of this type nearly always produce true beliefs, and thus he offers the following argument: AME1 (1) I have a clear and vivid memory that clear and vivid visual experiences are reliable.26 (2) Clear and vivid memories are reliable. ∴ It’s highly likely true that my clear and vivid visual experiences are true.27

25

My point is not that introspection is infallible. It is that awareness of what one introspects does not require a regress of meta-introspections. Thanks to Cherie Braden for suggesting I address this point. 26 If Doris were to challenge this premise, a regress along the introspective track would ensue. Its structure would replicate the regress discussed in the previous section. 27 I’m relying again on the additional premise that origination in a reliable source generates likelihood of truth.

252

M. Steup

Carrying on, Doris asks Fred to justify the second premise of AME1. That is, she asks Fred to explain how he knows that clear and vivid memorial experiences are reliable. Fred answers that he remembers that they are, and he proceeds to give the following argument: AME2 (1) I have a clear and vivid memory that clear and vivid memories are reliable. (2) Clear and vivid memories are reliable. ∴ It’s highly likely true that clear and vivid memories are reliable. As Doris did when she and Fred were on the introspective track, she might now say to him, “You just presented me with a new argument: AME2. This argument has two premises, each of which requires a reason. Please give me your reason for PREMISE (2) of this argument.” Here is Fred’s response: AME3 (1) I have a clear and vivid memory that clear and vivid memories are reliable. (2) Clear and vivid memories are reliable. ∴ It’s highly likely true that clear and vivid memories are reliable. AME2 and AME3 are two tokens of the same type. Fred’s giving AME2 is one speech act, his giving AME3 is another. But in giving AME3, Fred isn’t saying anything new. Although Doris and Fred could in principle go on forever, nothing of value will henceforth be achieved. What is of epistemic value has already been captured: Fred knows through memory that clear and vivid visual experiences are reliable, and he knows through memory that clear and vivid memorial experiences are reliable.28 If Doris is reasonable, she will abstain from asking Fred again to justify PREMISE (2) of AME3, will consider the matter closed, and will buy him a drink.29

15.12

The Five Stages of the Justificational Regress

One of Ginet’s objections to Klein’s infinitism concerns the content of the allegedly infinite regress. Ginet asks for an illustration of what’s going on when Doris and Fred continue traversing the allegedly infinite regress for at least some distance. In the previous two sections, I have taken Ginet’s request seriously and examined what

28

Some epistemologists would strenuously object that it’s irrational to attempt to justify the reliability of memory by using memory itself. For example, Fumerton (1995, 177) says, “You cannot use perception to justify the reliability of perception! You cannot use memory to justify the reliability of memory! You cannot use induction to justify the reliability of induction! Such attempts . . . involve blatant, indeed pathetic, circularity.” (See also Alston 1993.) I’m not persuaded by the objection that using our faculties to establish the reliability of our faculties involves vicious circularity (see Steup 2013). 29 If Doris were to ask Fred to justify why he thinks A1’s conclusion follows from A1’s premises, she would trigger a similar regress along an intuitional track.

15

Benign Infinity

253

ensues when Doris and Fred carry on for quite a while. A noteworthy outcome presented itself: paradoxically, although the argumentative regress is in principle infinite, there is nevertheless a clear sense in which it terminates in its final stage. Let’s review the overall structure of the argumentative regress. STAGE 1: Fred asserts that the dog ate the cookie. When Doris asks him to justify this belief, he mentions three further propositions: the dog is on the couch, the dog looks guilty, and Fred’s cookie is gone. This is the doxastic stage. STAGE 2: When Doris asks him to justify his belief that the dog is on the couch, Fred might say what is clearly a proper response, “I can see it.” Since he is an experiential foundationalist, he expresses this point by saying that he has a visual experience as of a dog on the couch. This the experiential grounding stage. STAGE 3: When Doris goes meta and asks Fred why he takes eDOG to be a reason for bDOG, he offers an initial TIP argument: eDOG is a clear and vivid visual experience, and such experiences are truth-conducive. This the initial meta stage. STAGE 4: Next, the regress branches into two tracks. Initiating the introspective track, Doris asks Fred to justify his claim that eDOG is a clear and vivid visual experience. In response, Fred offers two further TIP arguments: AIN1 and AIN2. Initiating the memorial regress, Doris asks Fred to justify his claim that clear and vivid memorial experiences are truth-conducive. In response, Fred offers another pair of TIP arguments: AME1 and AME2. This is the supplementary meta stage. STAGE 5: Doris points out that there is an algorithm for generating an infinite number of additional TIP arguments. She might say, “Fred, you just used four supplementary TIP arguments with two premises each. In principle, you owe me eight more arguments to justify these premises. These arguments will in turn have premises in need of justification, and so forth ad infinitum.” This is the stage of vacuous infinity. If Doris is reasonable, she should add, “On the other hand, I agree with you that these arguments will only repeat the main justification you already offered. I accept, therefore, that you have provided me with a satisfying defense of your claim that eDOG is a reason for bDOG, and ultimately with a complete justification of your assertion that the dog ate your cookie.” The justificational regress, then, is both finite and infinite. It is finite because, when it comes to revealing the flow of genuine justificational juice, STAGES 2–4 do all the work. Further TIP argumentation fails to be productive, as it does not identify any additional source of justification. So, STAGE 5 is superfluous. The regress now deteriorates into vacuous infinity. This type of infinity is benign and does not pose a problem for the claim that Fred is justified in believing that the dog is on the couch.

15.13

Benign Infinity

Here is an analogy. Consider the following equation: 1 þ 2 þ 3 þ ... þ n ¼

n2 n þ : 2 2

254

M. Steup

Its truth is not obvious. Now consider the following large square in which each natural number is represented with a little square. It illustrates the above equation for n ¼ 4.30

1þ2þ3þ4¼

42 4 þ ¼ 10: 2 2

If one looks at the square, one sees that the little squares representing the natural numbers on the left side of the equation make up one half of the large square plus n2 little squares, in this case one half of four little squares. As one comprehends this, one intuits that the equation works for any value of n where n is a natural number. A nagging doubter might suggest drawing a square for n ¼ 5 to see whether we get the same result, and then a square for n ¼ 6, just to make sure, and so forth. Obviously, nothing will be accomplished by carrying on. Two conclusions follow. First, the initial intuition triggered by apprehending the diagram is sufficient for understanding why the equation is true. Considering additional squares is pointless. It will not deepen our understanding. Second, since the equation is true for any n, it has an infinite number of instances. Such infinity is benign. The fact that there is an infinite number of instances, and the fact that as finite minds we could not comprehend the actual propositions stating the equations for extremely large values of n, does not limit at all our understanding of the equation’s truth. The regress of reasons for Fred is similarly benign. The infinity involved in no way threatens to undermine Fred’s justification. Therefore, there is no epistemic harm in terminating the regress at STAGE 4.

15.14

Does the Argumentative Regress Increase Justification?

I have argued that we must distinguish between two claims. The first is that the argumentative regress is in principle infinite. The second is that all justification is inferential. I have also argued that the former claim does not entail the latter. Klein, James Brown presented this example at his recent Reinhardt Lecture, “Pure and Applied: Mathematics and Ethics,” at CU Boulder on October 6, 2017.

30

15

Benign Infinity

255

though, thinks it does. Hence he needs a response to Ginet’s objection that inference cannot create, but can only transfer, justification (Ginet 2014a, 290f). Ginet’s point is that infinitism makes it impossible to understand where justification comes from. Klein responds by saying that justification emerges as the argumentative regress expands (Klein 2014a, 294). The more reasons one gives, the more one’s justification increases. There are three problems with this claim. First, as I argued above, this response leaves unexplained the case in which bDOG is already justified at tbefore. Second, it makes the justification of our beliefs dependent on the activity of reasongiving—a consequence I find utterly implausible. Third, when Fred has completed the supplementary meta stage and moves on to STAGE 5, the argumentative regress deteriorates into pointlessness, and there is no additional value to be added by continuing. I don’t find it plausible, therefore, that the longer the regress lasts, the more justification is generated. In fact, I find it doubtful that by responding to Doris’s continued demands for reasons, Fred increases his justification for bDOG to any degree. According to nondogmatic foundationalism, Fred’s belief is justified at tbefore because of: (i) eDOG, (ii) Fred’s clear and vivid introspective awareness that eDOG is a clear and vivid visual experience, (iii) memorial awareness that clear and vivid introspective and visual experiences are truth-conducive, and (iv) memorial awareness that clear and vivid memories are truth-conducive. These are reasons Fred already has before his dialogue with Doris commences. Obliging Doris’s requests to justify bDOG, Fred does not generate new reasons for his belief but merely articulates and makes public the reasons he had to begin with. I’m inclined to think, therefore, that the argumentative regress that ensues between Doris and Fred does not increase Fred’s justification for bDOG at all. It merely reveals where Fred’s justification comes from.

15.15

Foundationalist Infinitism

I conclude by summing up where I agree and where I disagree with Klein’s infinitism. First and most fundamentally, I agree with Klein that justification requires meta-justification. Second, I agree with Klein that the argumentative regress is in principle infinite. Third, I agree with Klein that the infinity involved is benign. However, and here my disagreement with Klein begins, the infinity in question is benign not because, as Klein suggests, it is merely an infinity of potential reasons. Rather, it is benign because an argumentative regress that infinite beings might carry on forever ceases at STAGE 5 to produce any new reasons. It merely repeats previously stated reasons. The potential infinity of the argumentative regress does not, therefore, require an infinite number of reasons. Second, I disagree with Klein’s claim that the longer the regress continues, the more justification is produced. I claim instead that if bDOG is justified to begin with, then Fred must have, prior to his dialogue with Doris, all the reasons needed for complete meta-justification. Due to these reasons, Fred’s belief is justified, and indeed justified to a degree that, in typical cases, is sufficient for knowledge.

256

M. Steup

Fred’s belief does not gain additional justification when, prompted by Doris, he uses these reasons to provide meta-justification for bDOG. Third, I disagree with Klein’s claim that justification is created through inference. Fred’s belief was already completely justified before Doris and Fred began their dialogue. At that time, bDOG was justified non-inferentially: not by any inferences, but instead by Fred’s initial visual experience, eDOG, and the various introspective and memorial states providing him with the reasons for the reliability of eDOG. I do not, therefore, see a conflict between foundationalism and infinitism. However, infinitism isn’t compatible with just any form of foundationalism. Infinitism results from the principle of meta-justification. Dogmatic foundationalism rejects this principle. Hence, I strongly suspect Klein rejects dogmatism.31 If he does, I find myself again in agreement with him. If dogmatism is true, there can be justification without meta-justification. If there can be justification without meta-justification, there will be a point when Fred has no answer at all when Doris asks him, “What justifies what you just said?” Klein would claim that this introduces an unacceptable kind of arbitrariness. On that, I couldn’t agree more with him.32

References Alston, W. (1989). Epistemic justification. Essays in the theory of knowledge. Ithaca: Cornell University Press. Alston, W. (1993). The reliability of sense perception. Ithaca: Cornell University Press. Audi, R. (1993). The structure of justification. New York: Cambridge University Press. BonJour, L. (1985). The structure of empirical knowledge. Cambridge: Harvard University Press. Dancy, J. (1985). Introduction to contemporary epistemology. Oxford: Blackwell. Davidson, D. (1983). The coherence theory of truth and knowledge. In D. Henrich (Ed.), Kant oder Hegel (pp. 423–438). Stuttgart: Klett Cotta. Feldman, R. (2003). Epistemology. Upper Saddle River: Pearson. Fricker, E. (1994). Against gullibility. In B. K. Matilal & A. Chakrabarti (Eds.), Knowing from words (pp. 125–161). Dordrecht: Kluwer.

Klein (2014c) explores options for finding rapprochement between infinitism and foundationalism. It is not clear to me, however, that the rapprochement he seeks is the kind of compatibility I have defended in this chapter. I am puzzled, to begin with, by Klein’s definition of foundationalism. He says that according to foundationalism, basic beliefs are non-arbitrarily held “even in the absence of reasons for them” (Klein 2014c, 117). Neither dogmatic nor non-dogmatic foundationalists should agree. The key point is the following: a basic belief is justified by mental states other than beliefs, namely by perceptual, introspective, memorial, or intellectual (intuitional) experiences or seemings. Basic beliefs, then, are never held in the absence of a reason. Does that mean they are always non-arbitrarily held? Dogmatists claim that an experience of p can, all by itself, without any need of any additional reasons, justify the belief that p. Non-dogmatic foundationalists claim that without meta-justification, a single experience is insufficient to remove arbitrariness. Since infinitism rests on the principle that justification requires meta-justification, I think the prospects for rapprochement are limited: infinitism does not conflict with non-dogmatic foundationalism but is impossible to reconcile with dogmatism. 32 I wish to thank Cherie Braden for valuable feedback and a fantastic editing job. 31

15

Benign Infinity

257

Fricker, E. (1995). Telling and trusting: Reductionism and anti-reductionism in the epistemology of testimony. Mind, 104(414), 393–411. Fumerton, R. (1995). Metaepistemology and Skepticism. Lanham: Rowman & Littlefield. Ginet, C. (2014a). Infinitism is not the solution to the regress problem. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 283–291). Oxford: Wiley. Ginet, C. (2014b). Reply to Klein. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 295–297). Oxford: Wiley. Huemer, M. (2001). Skepticism and the veil of perception. Lanham: Rowman & Littlefield. Huemer, M. (This volume). Finite minds. Klein, P. D. (1999). Human knowledge and the infinite regress of reasons. Philosophical Perspectives, 13(Epistemology), 297–325. Klein, P. D. (2010). Self-profiles: Peter Klein. In J. Dancy, E. Sosa, & M. Steup (Eds.), A companion to epistemology (2nd ed., pp. 156–163). Oxford: Wiley-Blackwell. Klein, P. D. (2014a). Infinitism is the solution to the regress problem. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 274–283). Oxford: Wiley. Klein, P. D. (2014b). Reply to Ginet. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (2nd ed., pp. 291–295). Oxford: Wiley. Klein, P. D. (2014c). Reasons, reasoning and knowledge: a proposed rapprochement between infinitism and foundationalism. In P. Klein & J. Turri (Eds.), Ad infinitum: New essays on epistemological infinitism (pp. 105–124). New York: Oxford University Press. Lackey, J. (2006). It takes two to tango: Beyond reductionism and non-reductionism in the epistemology of testimony. In J. Lackey & E. Sosa (Eds.), The epistemology of testimony (pp. 160–189). New York: Oxford University Press. Lackey, J. (2008). Learning from words. New York: Oxford University Press. Pollock, J. L. (1986). Contemporary theories of knowledge. Totowa: Rowman & Littlefield. Pryor, J. (2000). The skeptic and the dogmatist. Noûs, 34(4), 517–549. Pryor, J. (2014). There is immediate justification. In M. Steup, J. Turri, & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 202–222). Oxford: Wiley. Sanford, G., & Henderson, D. (2006). Monitoring and anti-reductionism in the epistemology of testimony. Philosophy and Phenomenological Research, 72(3), 600–617. Steup, M. (1996). An introduction to contemporary epistemology. Upper Saddle River: Prentice Hall. Steup, M. (2004). Internalist reliabilism. Philosophical Issues, 14(Epistemology), 403–425. Steup, M. (2013). Is epistemic circularity bad? Res Philosophica, 90(2), 215–235. Steup, M. (Forthcoming). Destructive defeat and justificational force: The dialectic of dogmatism, conservatism, and meta-evidentialism. Synthese. Tucker, C. (Ed.). (2013). Seemings and justification. New York: Oxford University Press.

Index

A Accuracy, 37, 58, 69, 194, 195 Alternatives to defeasibility, 3, 48–52, 165 Anscombe, G.E.M., 90, 93–96, 98 Availability of propositions, 184–186

B Bayesian approaches to defeaters, 8, 233 Bayesian epistemology, 214, 233 Belief revision, 6, 68, 103, 139, 237–239 Benign regress, 236, 237, 247–255 Brueckner, A., 149, 150, 152, 153, 155, 161, 164–166

C Cartesian skepticism, 6, 146, 148 Causal basing, 89 Cognitive limitations, 133, 141, 184 Complexity of propositions, 19, 101, 141, 181, 182, 184 Confirmation relation, 42, 45–48, 52 Counter-closure, 81, 156, 161–163, 166 Counterevidence, 4, 72–79, 82, 84, 86, 87

D Defeasibility, 1, 3–5, 26, 38, 49–51, 53, 60, 62, 65, 76, 80–83, 135–139, 160 Defeasible reasoning, 41, 42, 44, 45, 48 Degrees of justification, 77, 78, 194, 244 Deploying reasons, 4, 89, 93, 101, 183

E Enhanced knowledge, 4, 16, 22 Epistemic basing, 89 Epistemic closure, 5, 7, 146 Epistemic competence, 15, 17, 22 Epistemic defeat, 78, 79, 84, 85, 214 Epistemic humility, 130–143 Epistemic information, 8, 20, 21, 41, 49, 50, 53 Epistemic limitations, 132–135, 139–141, 143 Epistemic primitives, 2, 5 Epistemic responsibility, 9, 159, 207–210 Epistemic virtue, 3, 41, 132, 141 Epistemology, 1, 2, 4, 5, 8, 10, 21, 26, 36, 40, 59–61, 66, 76, 141, 146, 159, 162, 166, 200, 213, 214, 247 Epistemology of the infinite, 190, 191, 193 Etiology, 4, 31, 92, 93 Evidential defeat, 8, 233 Evidential underdetermination, 146 Externalist epistemology, 4, 9, 200

F Falsity-makers, 34 Finite-mind objection, 10, 171, 172, 185, 186, 190, 192, 245 Finite minds, 10, 171–186, 189, 236, 241, 245, 247, 251, 254 Formal epistemology, 10, 47, 213–233 Foundationalism, 8, 9, 26, 174, 198–207, 211, 236, 237, 239–245, 247 Foundationalist infinitism, 11, 255, 256 Foundationalist justification, 198–211 Fragmentation, 40, 43

© Springer Nature Switzerland AG 2019 B. Fitelson et al. (eds.), Themes from Klein, Synthese Library 404, https://doi.org/10.1007/978-3-030-04522-7

259

260 H Holism in epistemology, 41, 42 Hyperintensionality, 32, 34, 38

I Inference chains, 26, 38, 236 Infinitism, 1, 8, 10, 26, 147, 171–173, 175, 178, 186, 189–191, 198, 236–238, 243, 245–248, 252, 255, 256 Information gaps, 58, 61, 65 Intellectual humility, 77, 133 Internalist reliabilism, 4, 92, 105, 207 Internal reliability, 3

J Justification, 16, 26, 40, 59, 74, 92, 113, 130, 175, 190, 198, 216, 236 Justification defeat, 3, 33, 214 Justification requirement, 53, 62

K Kant, I., 97, 112, 118, 119, 125 Klein, P.D., 1–10, 16, 25–30, 32, 33, 35–37, 50, 60, 65, 74, 80–82, 90, 92, 97, 99, 101, 106, 116, 128, 130–132, 134–136, 139, 142, 146–166, 171–175, 177, 179, 180, 184, 186, 189–191, 198–210, 216, 236, 238–248, 254, 255 Knowledge as achievement, 40 Knowledge conditions, 3, 4, 8, 29, 62, 84, 86, 149, 159 Knowledge defeaters, 1, 3, 8, 50, 79, 81–83, 135, 136, 138 Knowledge-first, 5, 40, 44–48, 51, 52, 54, 78–80, 85 Knowledge from falsehoods, 25, 68, 80, 81 Knowledge in degrees, 16, 18–20, 53, 78, 79, 87, 227, 244, 255 Knowledge of reliability, 3, 17, 19–22, 59, 60, 63, 64 Knowledge without justification, 1, 6, 11, 17, 29, 43, 50, 65, 92, 239, 241–243, 256

L Lasonen-Aarnio, M., 79, 84, 85 Lehrer, K., 60, 83, 90–92, 96, 101–103, 105, 136, 159, 203

Index Metajustification, 204 Moderate skepticism, 7, 130 Mooreanism, 154–159, 165, 166 Moral agency, 117, 123, 124, 128

N Non-dogmatic foundationalism, 244, 246–248, 255 Normative epistemology, 43, 44, 47, 51, 193

P Practical rationality, 19, 114, 123 Privations, 94–96, 98–101, 104–106 Probabilistic justification, 10, 62, 191, 194 Psychological defeat, 78, 79, 87 Pyrrhonian skepticism, 16, 18, 127, 128, 147 Pyrrhonism, 117, 126–128, 146, 147

R Rational agency, 116–118, 124, 127 Reasoning agent, 5 Reasons for reliability, 17, 19, 22, 59, 60, 62–64, 205, 256 Reflective knowledge, 4, 15–22 Regress problem, 199, 200, 202, 204, 207, 210, 236, 247 Reliabilist foundationalism, 200, 203, 204, 206, 207 Responses to skepticism, 6, 113, 131, 147, 155, 166

S Skeptical argument, 5, 6, 111, 117, 126, 128, 130, 147 Skeptical hypothesis (SH), 6, 7, 111–117, 119, 120, 122–125, 127, 130, 131, 146, 149, 150, 152, 155–158 Skepticism, 1, 5–7, 11, 18, 64, 90, 111, 146, 198, 233, 241, 242 Structure of justification, 1, 8, 10, 198, 199

T Transcendental arguments, 6, 111–128 Types of defeaters, 49, 219

U Useful falsehoods, 1, 3, 4, 25, 80, 81 M Maximally accurate, 58, 59, 61–63, 65 Meta-epistemology, 49, 200, 211