Inquiry, Knowledge, and Understanding 0192896091, 9780192896094

Table of contents :
Title_Pages
Dedication
Preface
IntroductionInquiry_and_Epistemology
The_Aim_of_Inquiry
Knowledge
Conditions_on_Knowledge
Understanding
Epistemic_Value
Scepticism
Appendix_1
Appendix_2
Bibliography
Index

Citation preview

Inquiry, Knowledge, and Understanding

Inquiry, Knowledge, and Understanding CHRISTOPH KELP

1

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Christoph Kelp 2021 The moral rights of the author have been asserted First Edition published in 2021 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2020946999 ISBN 978–0–19–289609–4 DOI: 10.1093/oso/9780192896094.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

For Mona

Preface This book has benefited from discussions with and feedback from many colleagues and friends. First and foremost, I’d like to offer special thanks to Mona Simion with whom I have thought through the issues in this book more than with anyone else. In addition, I’d like to express my gratitude for helpful feedback to my colleagues at COGITO Epistemology Research Centre, Kristoffer Ahlstrom-Vij, Heather Battaly, Christoph Baumberger, Bob Beddor, Claus Beisbart, Sven Bernecker, Alexander Bird, Michael BlomeTillman, Cameron Boult, Elke Brendel, Fernando Broncano-Berrocal, Jessica Brown, Georg Brun, Herman Cappelen, Adam Carter, Matthew Chrisman, Annalisa Coliva, Juan Comesaña, Marian David, Henk DeRegt, Ezio DiNucci, Philip Ebert, Anna-Maria Eder, Miguel Egler, Catherine Elgin, Markus Eronen, Jeremy Fantl, Claire Field, Giada Fratantonio, Lizzie Fricker, Miranda Fricker, Jane Friedman, Mikkel Gerken, Sandy Goldberg, Alvin Goldman, Emma Gordon, Peter Graham, Patrick Greenough, John Greco, Stephen Grimm, Thomas Grundmann, Katherine Hawley, John Heil, Jaakko Hirvelä, Frank Hofmann, Joachim Horvath, Jonathan Jenkins Ichikawa, Matt Jope, Jepser Kallestrup, Klemens Kappel, Kareem Khalifa, Dirk Kindermann, Peter Klein, Stephan Krämer, Jennifer Lackey, Maria Lasonen-Aarnio, Clayton Littlejohn, Fed Luzzi, Jack Lyons, Aidan McGlynn, Matt McGrath, Conor McHugh, Robin McKenna, Alan Millar, Lisa Miracchi, Anne Meylan, Luca Moretti, Jennifer Nagel, Ram Neta, Nikolaj Nottelman, Erik Olsson, Carlotta Pavese, Nikolaj Jang Lee Linding Pederson, Andy Peet, Tommaso Piazza, Duncan Pritchard, Jim Pryor, Agustin Rayo, Wayne Riggs, Blake Roeber, Sven Rosenkranz, Felipe Santos, Susanna Schellenberg, Pedro Schmechtig, Susanna Siegel, Waldomiro Silva-Filho, Gerhard Schönrich, Martin Smith, Roy Sorensen, Matthias Steup, Ernest Sosa, Kurt Sylvan, Alessandra Tanesini, Raphael van Riel, Jonathan Vogel, Daniel Whiting, Tim



vii

Williamson, and Crispin Wright. I am sure that I have missed people who deserve to be acknowledged by name and hope they will accept my apologies. Elements of this book have been presented on a number of occasions, including Arché 20th Anniversary Conference, University of St Andrews (2019); Epistemology Workshop, University of Copenhagen (2019); Closure and Warrant Transmission Workshop, University of Edinburgh (2019); Fifth Conference on Analytic Philosophy, Salvador de Bahia (2018); Knowledge First Workshop, Cardiff University (2018); Metaphysics of Knowledge Workshop II, University of St Andrews (2017); Senior Seminar, University of Glasgow (2017); CSMN/ ConceptLab Visiting Speakers Seminar, University of Oslo (2017); Leuven-Bristol Workshop, University of Bristol (2016); New Trends in Epistemology, FINO/SIFA Graduate Conference in Mind, Language, and Science, University of Pavia (2017); Rutgers Epistemology Conference, Rutgers University (2017); Philosophy Colloquium, Washington University St. Louis (2017); Rational Capacities Conference, University of Luxembourg (2016); Epistemology Research Group, University of Edinburgh (2016); 2016 EEN Meeting, Paris (2016); Flash Epistemology Workshop, University of St Andrews (2016); Mathematical Collaboration, University of St. Andrews (2018); Wissen, Verstehen, Erklären, University of Dresden (2017); The Value of Understanding, UFBA, Salvador de Bahia (2014); EPSA 2013, University of Helsinki (2013); Understanding and Justification, University of Bern (2014); Explanatory Power II: Understanding through Modelling? Epistemology, Semantics, and Metaphysics of Inadequate Representation, University of Bochum (2013); Epistemic Justification and Reasons, University of Luxembourg (2013); Bled Philosophical Conferences, Bled (2013); Neue Perspektiven der Epistemischen Rechtfertigung, University of Dresden (2013); CCPEA. Academia Sinica, Taipei (2012); 2012 EEN Meeting, Universities of Bologna and Modena (2012); Epistemology Workshop, University of Groningen (2012); GAP8, University of Konstanz (2012); Philosophy Colloquium, University of Essen (2011). Thanks to the audiences for their very helpful comments on the material in this book.

viii



Finally, I’d like to thank two anonymous reviewers from Oxford University Press for their very helpful comments on the entire manuscript and Peter Momtchiloff and the entire Oxford University Press team for invaluable support with the publication process. Some of the material of this book draws on and further develops ideas that have been published in earlier papers. I’d like to thank the journals for granting me permission to use materials from Kelp, C. 2014. ‘Two for the Knowledge Goal of Inquiry’. American Philosophical Quarterly 51: 227–32; Kelp, C. 2015. ‘Understanding Phenomena’. Synthese 192: 3799–816; Kelp, C. 2017. ‘Knowledge First Virtue Epistemology’. In Knowledge First: Approaches in Epistemology and Mind, edited by A. Carter, E. Gordon, and B. Jarvis. Oxford: Oxford University Press; Kelp, C. 2018. ‘Inquiry, Knowledge and Understanding’. Synthese. doi. org/10.1007/s11229-018-1803-y; Kelp, C. 2019. ‘Inquiry and the Transmission of Knowledge’. Philosophy and Phenomenological Research 99: 298–310; and Kelp, C. 2020. ‘Theory of Inquiry’. Philosophy and Phenomenological Research. doi-org.ezproxy.lib.gla.ac. uk/10.1111/phpr.12719.

Introduction Inquiry and Epistemology

Let’s start with the central methodological idea of this book, which is to take inquiry (or finding out about things) as the starting point for epistemological theorizing. What this means in practice is that, in this book, I will be thinking a good deal about inquiry, and I will argue that thinking about inquiry allows us to develop an attractive way of doing epistemology. In a nutshell, the reason for this is that thinking about inquiry gives us a systematic way of developing answers to a range of central epistemological questions that are not only novel but also promising. Some of the most fundamental questions in epistemology concern (i) the nature of core epistemic phenomena as well as (ii) their value and (iii) the extent to which we possess them. My ambition here is to use the central methodological idea to develop new answers to all three of the above questions in a systematic way. The core epistemic phenomena that will take centre stage here are knowledge and understanding. However, my investigation will broach a variety of further relevant epistemic phenomena, most notably epistemic abilities, and epistemic sources such as deduction. In the remainder of this Introduction, I will sketch the view with a very broad brush. The ensuing chapters will then do the job of filling in the details.¹ One interesting property of inquiry is that it is a type of activity with an aim. When we are inquiring, we are trying to find out something, to settle something, to understand something, etc. A distinction that is of ¹ Recent years have witnessed a surge of interest in the debate between traditionalist and experimental philosophers. Since this book falls in the traditionalist camp, it is only fair to acknowledge the fact that a lot of work in experimental philosophy has been done that raises important concerns for traditionalism (e.g. Buckwalter and Stich 2014, Machery et al. 2004, Petrinovich and O’Neill 1996, Schwitzgebel and Cushman 2015, Swain et al. 2008, Weinberg 2016, Weinberg et al. 2001, Wright 2010). With this disclaimer in play, I hasten to add that I will leave discussion of experimental philosophy for another occasion. Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0001

2

, ,  

some importance for present purposes is that there are two different types of inquiry. First, one can inquire into specific questions such as the question of whether Boris Johnson will be prime minister of the UK in 2019, whether the Rolling Stones have ever played a gig in Ruanda, when the Battle of Hastings took place, who won the Wimbledon mixed doubles competition in 2018, and so on. One claim that I expect to enjoy widespread agreement is that an inquiry into a specific question aims at settling the question inquired into: Question-Settling Aim One’s inquiry into Q aims at settling Q; alternatively: inquiry into Q aims at properly closing Q for oneself in the affirmative/negative.

For instance, an inquiry into who won the Wimbledon mixed doubles competition in 2018 aims at settling who won the Wimbledon mixed doubles competition in 2018, an inquiry into when the Battle of Hastings took place aims at settling when the Battle of Hastings took place, and so on. Second, one can inquire into general phenomena such as the UK’s exit from the European Union, the rise of the Roman Empire, the origins of species, the death of JFK, and so on. Another claim that I expect to enjoy widespread agreement is that an inquiry into a general phenomenon aims at understanding the phenomenon inquired into: Understanding Aim One’s inquiry into phenomenon P aims at understanding P.

For instance, inquiry into the rise of the Roman Empire aims at understanding the rise of the Roman Empire, inquiry into the death of JFK aims at understanding the death of JFK, and so on. While I expect that some consensus on the aims of the two forms of inquiry is easy to generate, more substantive epistemological questions about them remain. In particular, there are a number of substantive accounts of both the aim of inquiry into specific questions and understanding on the market. For instance, some have unpacked the aim of



3

inquiry into specific questions in terms of true belief, others in terms of justified belief, and yet others in terms of knowledge about the answer. Likewise, there are a number of different accounts of understanding on the market: some analyse understanding in terms of knowledge, others in terms of true belief, and yet others in terms of justified belief. My own view is that the aim of inquiry of both types is to be understood in terms of knowledge. Inquiry into specific questions— and, in particular, into specific whether questions, which is what I’ll be focusing on here—aims at knowledge that p (not-p). Understanding and hence the aim of inquiry into general phenomena are to be understood in terms of systematic knowledge that p. Note that while knowledge is key to understanding, understanding requires more than just knowledge: the relevant knowledge must in addition be systematic. That is to say, the various pieces of knowledge must be hooked up in the right way. It won’t come as much of a surprise that knowledge plays a central role in the view I am developing here. Note, however, that this is not essential to the main project of the book, which is to explore the central methodological idea of taking inquiry as the starting point for epistemological theorizing. The reason for this is that it is entirely compatible with this methodological idea that some other epistemic phenomenon may have assumed the central place of knowledge instead. That knowledge is of central importance is a function of the methodological idea in conjunction with a set of arguments. Knowledge must earn its keep as something of key significance. That it does so is by no means a foregone conclusion in the present framework. In this way, the present approach differs from knowledge first epistemology, which has been on the rise in epistemology since Williamson’s seminal 2000 book. Now, I do not mean to deny that there are a wide range of affinities and similarities between the present approach and knowledge first epistemology. Many of them will become clear in due course. However, it is important to keep in mind that a key methodological idea of knowledge first epistemology is that it takes the distinction between knowledge and ignorance as the starting point for epistemological theorizing (Williamson 2010, 208). As a result, that knowledge is of key significance in epistemology is not optional on a knowledge first

4

, ,  

approach. This is why the present view cannot legitimately be placed in the knowledge first camp. A further idea that is of considerable importance for the purposes of this project is that the above aims are constitutive aims of inquiry into, respectively, specific questions and general phenomena. I will defend this idea in due course. For now, I will content myself with saying a few words about what it means to say that the above aims are constitutive aims of our two types of inquiry: they are essential to these activities. Anything that does not have settling a certain question as its aim will not count as an inquiry into a specific question. And anything that does not have understanding a certain phenomenon as its aim will not count as an inquiry into a general phenomenon. The reason this is so important is that it means that inquiry turns out to be a species of the broader genus of activities with constitutive aims. It is this feature of inquiry that provides the key to all the answers to the fundamental epistemological questions that I aim to address here. The perhaps most obvious case is that of epistemic value. I will argue that activities with constitutive aims constitute critical domains of value in which the constitutive aim corresponds to a for-its-own-sake value relative to this domain. This paves the way for an answer to a particularly hard question in the theory of epistemic norms and values, to wit, exactly which epistemic values are epistemically valuable for their own sake. It also allows us to make progress towards solutions of important value problems in epistemology which concern the relative value of some key epistemic properties. The perhaps most famous value problem in epistemology is the so-called Meno problem (Plato 1956), which asks us to explain why knowledge is more valuable than mere true belief. It will come as no surprise that my approach to various value problems in epistemology, including the Meno problem, will employ the idea that knowledge is a for-its-own-sake epistemic value. Since knowledge is the constitutive aim of inquiry into specific questions, it is valuable for its own sake relative the domain constituted by this activity. Since true belief isn’t, knowledge comes out as being more valuable than true belief. Again, I expect there to be agreement from the knowledge first camp, in that I expect knowledge firsters to be sympathetic to the idea that knowledge is a for-its-own-sake value in the epistemic domain. At the



5

same time, I also expect there to be important differences. In particular, on my view, understanding also comes out as a final value in the epistemic domain. The view I am proposing offers a pluralistic approach to epistemic value. In contrast, I would expect knowledge firsters to be more attracted to a monistic alternative, according to which knowledge is the only for-its-own-sake epistemic value. It may not be particularly surprising that the central methodological idea has potential to make headway when it comes to matters of epistemic value, especially once we are clear that inquiry is an activity with a constitutive aim. At the same time, it may be less straightforward that it should do the same when it comes to issues of the nature of epistemic phenomena. Perhaps one of the most ambitious aims of this book is to show that it does. More specifically, I will argue the following two points. First, there is a class of activities with constitutive aims that require what is sometimes called a network analysis (Strawson 1992), according to which the nature of the activity and its constitutive aim, as well as various other phenomena, are constitutively related to one another and can only be understood in terms of each other. Second, inquiry falls into this class of activities and, as a result, inquiry, knowledge, understanding, and belief are constitutively related to one another and can only be understood via a network analysis. My ambition is to develop this network analysis is some detail. If successful, this part of the project will lead to novel accounts of the nature of two key epistemic phenomena, knowledge and understanding. Both of these accounts will be non-reductive. In this respect, there are further affinities between my project and knowledge first epistemology, which agrees that knowledge does not admit of reductive analysis. Again, however, there are important differences between the two approaches. First, while my own preferred account of knowledge is indeed non-reductive, my project is in principle compatible with a reductive account of knowledge. In contrast, it is part and parcel of the knowledge first programme knowledge does not admit of reductive analysis. Second, the non-reductive account of the nature of knowledge my project will develop differs markedly from the standard knowledge first epistemological rival. According to the standard knowledge first account of knowledge, knowledge is essentially the most

6

, ,  

general factive mental state (Williamson 2000). In contrast, the central idea that my account builds on is that knowledge is, as a constitutive matter of fact, the aim of inquiry. Third, I offer not only a nonreductive account of knowledge but also of understanding, whereas we might expect knowledge first epistemology to offer a reductive analysis of understanding in terms of knowledge. The third central issue that this book will address concerns the extent of our knowledge. In other words, I want to address the problem of scepticism. According to the perhaps most powerful sceptical argument, the argument from ignorance, first, we don’t know that we are not radically deceived. Second, if we don’t know that we are not radically deceived, then we don’t know much of what we ordinarily take ourselves to know. From this it follows, third, that we don’t know much of what we ordinarily take ourselves to know. Standard responses to the argument from ignorance deny one of its two premises. Or, to be more precise, in the case of those who deny the second premise, what’s denied is the following transmission of knowledge principle that is widely believed to motivate it: Transmission If one knows that p, competently deduces q from p, thereupon comes to believe that q, then one knows that q.

Deniers of Transmission typically also hold that we know much of what we ordinarily take ourselves to know even though we don’t know that we are not radically deceived (e.g. Dretske 1970, Nozick 1981). The perhaps more popular alternative is to hold on to Transmission and to maintain, contrary to the first premise in the sceptical argument, that we do know that we are not radically deceived. In fact, the most popular version of this view holds that we have this knowledge in virtue of deduction from some ordinary proposition that we know, thanks to Transmission (e.g. Moore 1939, Pritchard 2005, Sosa 1999, Williamson 2000). The view I will develop differs from standard views in that it takes issue with both premises of the sceptical argument. I show that there is compelling independent reason to believe that Transmission fails, and that, as a result, the second premise of the sceptical argument is



7

unmotivated. While this means that the standard account of how we could know that we are not radically deceived is off the table, I argue that there is a better alternative available. Key to this alternative is the idea that many of us have the ability to recognize whether certain possibilities could easily obtain. It is through the exercise of this ability that we can acquire basic (i.e. non-inferential) knowledge of the denials of various sceptical hypotheses. In this way, the central methodological idea offers a novel approach to the sceptical problem, one that combines a robust anti-scepticism with Transmission failure. This completes my initial overview of the central methodological idea and how it aims to address the perhaps three most fundamental issues in epistemology. To repeat, so far I have worked with a very broad brush. This means that a lot of substantive work remains to be done in the chapters to come. This work will fill in the details of what I have outlined here. However, it will also go beyond the above sketch in that it will address a range of further important issues in epistemology, beyond the central themes of the nature, value, and extent of various epistemic phenomena. A couple of things before getting down to business: First, a disclaimer: the main ambition of this book is constructive. I aim to develop a novel way of approaching epistemological theorizing and to use this way to offer a new and systematic treatment of the most fundamental issues in epistemology. Since I want to put the constructive work centre stage, I will spend less time on critical discussion of rival views. I do not mean to say that I don’t discuss alternative views at all. Rather, I will restrict my discussion to the closest-in-spirit and most prominent competitors to the various accounts I will develop here. Given the breadth of the project, I hope this restriction can be excused. Second, here is a game plan for the book: Chapter 1 focuses on epistemologically substantive accounts of the aims of inquiry into specific questions. It mounts a detailed case that knowing that p/not-p is the aim of inquiry into whether p. To this end, I first develop two arguments that the knowledge aim of inquiry compares favourably with the main rivals in the literature, according to which the aim of inquiry is true belief or justified belief. Next, I show how these arguments can be generalized to other views about the aim of

8

, ,  

inquiry that might be conceived. Finally, I respond to a number of objections to the idea that knowledge is the aim of inquiry and the argument developed in support of it. Chapter 2 uses the idea that inquiry into specific questions is an activity with a constitutive aim and that knowledge is this constitutive aim to develop a non-reductive account of knowledge. The key idea is that certain activities with constitutive aims do not lend themselves to reductive analysis and instead afford a so-called network analysis in which each element cannot be properly understood without grasping the connections with other elements in the network. After addressing objections, I compare my account with Williamson’s non-reductive account of knowledge as a sui generis mental state and argue that the former is preferable to the latter. Chapter 3 focuses on conditions of knowledge. It argues that, for activities with constitutive aims featuring normative properties, it is not uncommon to find substantive conditions on both the means of attaining the aim and the environment. Moreover, I show that inquiry is a case in point. Since knowledge is the constitutive aim of inquiry, it follows that there are substantive constraints on the means of attaining knowledge and on the environment in which knowledge can be had. I go on to develop a more detailed account of these constraints in terms of abilities to know and show how these conditions can be used to solve the Gettier problem. Finally, I discuss a number of key objections and offer responses. Chapter 4 moves on to inquiry into general phenomena and develops a non-reductive systematic knowledge account of understanding. According to this view, roughly, maximal understanding of a phenomenon is maximally comprehensive and well-connected knowledge of it, degrees of understanding are a function of distances from maximal understanding, and understanding a phenomenon can be truly attributed when one surpasses a contextually determined threshold on degrees of understanding. I discuss a number of objections to this view and compare it to the most prominent rival views in the literature, according to which understanding is, in essence, knowledge of explanations. Again, I argue that the systematic knowledge account comes out on top.



9

Chapter 5 turns to issues relating to epistemic value. It argues that activities with constitutive aims constitute value domains in which the constitutive aims are domain-relative for-their-own-sake values. Applied to the case of the two forms of inquiry, we get the result that knowledge and understanding are valuable for their own sake in the domains constituted by these activities. In addition, I argue that the two forms of inquiry constitute the epistemic domain, thus shedding light on the boundaries and the structure of the epistemic domain. Finally, I show that the resulting view can solve a number of so-called value problems in epistemology, including the difficult tertiary value problem according to which knowledge must come out as more valuable than mere true belief as a matter of kind. Chapter 6 addresses the problem of scepticism. More specifically, it focuses on a particularly difficult sceptical argument, which proceeds from the plausible claims (i) that we don’t know that we are not radically deceived and (ii) that, if so, we don’t know much at all to the problematic sceptical conclusion that we don’t know much at all. I argue that there is reason to take issue with both premises of this argument. More specifically, I present a novel theoretical argument against the principle the knowledge transmits across competent deduction, which motivates the second premise. And I develop a new way of resisting the first premise. The key idea here is that we can have basic knowledge of the denials of sceptical hypotheses thanks to an ability to know that certain possibilities could not easily obtain. Having dealt with some objections, I compare the approach to scepticism developed here with its closest competitor, the sensitivity-based approach, and argue that there is reason to favour the former.

1 The Aim of Inquiry In the Introduction, I introduced some characterizations of the aims of inquiry into specific questions and general phenomena that are lightweight in the sense that they ought to be acceptable to everyone. However, especially in the case of inquiry into specific questions, there is also a lively debate on the issue in the epistemological literature. This chapter engages with this debate. It aims to show that knowledge is the aim of inquiry into specific questions.

1. The Aim of Inquiry into Specific Questions Let’s first recall the lightweight characterization of the aim of inquiry into specific questions in terms of question-settling. For present purposes, I will focus on inquiry into specific whether questions, i.e. questions of the form whether p. Accordingly, here is the lightweight characterization of the aim of inquiry that I will be working with in what follows: Question-Settling Aim One’s inquiry into whether p aims at settling the question whether p; alternatively: inquiry into whether p aims at properly closing for oneself the question whether p in the affirmative/negative.

And now I’d like to turn to the main views that are held in the epistemological literature on the aim of inquiry: Knowledge Aim One’s inquiry into whether p aims at coming to know that p/not-p.¹ ¹ Champions of Knowledge Aim include Williamson (2000) and Millar (2011). Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0002

   

11

True Belief Aim One’s inquiry into whether p aims at coming to truly believe that p/not-p.² Justified Belief Aim One’s inquiry into whether p aims at coming to justifiably believe that p/not-p.³

Crucially, these substantive epistemological views are not in competition with Question-Settling Aim. Rather, they can be seen as offering substantive accounts of what exactly it takes to settle a question. To repeat, my own view is that knowledge is the aim of inquiry. In what follows I will mount a defence of it. In order to achieve this, I will first argue that any account of the aim of inquiry that imposes a weaker requirement than knowledge is too weak. More specifically, I will first offer a couple of arguments that Knowledge Aim compares favourably to its two main competitors and then provide reason to think that the argument generalizes. After that, I will argue that any account that imposes a stronger requirement is too strong.

1.1 The Aim of Inquiry Requires Knowledge Argument 1: Commitment Release. Consider activities with aims (henceforth AAs for short). There are many examples. To take just one, consider running a certain marathon. My first argument for Knowledge Aim involves a principle about being released from commitments towards attaining the aim of AAs. To get a clearer view, I’d like to say a few words about having aims of attaining successes in AAs, being committed to attaining them, and the relation between the two. One may have the aim of attaining the success of an AA. One does so when one has the intention to attain the activity’s aim. For instance, you

² True Belief Aim has been defended, for instance, by Jonathan Kvanvig (2003) and Michael Lynch (2005). ³ Donald Davidson (2005), Richard Feldman (2002), Richard Rorty (1995), and Jay Rosenberg (2002) are among the most prominent advocates of Justified Belief Aim.

12 , ,   may have the aim of running the marathon. You do so when you have the intention of running it. In addition, one may be committed to attaining successes in AAs. The nature of the commitment depends on how it arises (or how it is sustained). The perhaps most familiar case is one in which one desires to attain success. For instance, you may desire to run the marathon in question. In this case, I will say that you have a practical commitment to running it. You may also have promised someone to run it. In this case, you have a moral commitment to running the marathon. Or you may have been contracted to run it, in which case your commitment will be contractual. Other forms of commitment are conceivable. What’s more, one may simultaneously have various different kinds of commitment towards attaining success in an AA. Besides desiring to run the marathon, you may have promised and signed a contract that you would do so. In this case, you are practically, morally, and contractually committed to running the marathon. In the standard case, one has the aim of attaining the success of an AA because of some set of commitments one has towards attaining it. For instance, in the standard case, you desire to run the marathon in question and thereupon form (and subsequently sustain) the intention to do so.⁴ Note, however, that being committed to attaining the success of an AA is one thing and having the aim of attaining it is another. Most importantly for present purposes, one can be committed to attaining the success of an AA even though one does not have the aim of attaining it.⁵ Suppose you once desired to run the marathon and thereupon formed the intention to

⁴ Note that while this is how things go in the standard case, it is not how they must go. You may promise someone to do something that you have no desire to do and form an intention to do that thing based on your promise. Here you have the aim of doing what you promised to do, but your aim is neither generated nor sustained by a desire to do it. Likewise, what sustains your aim may change. While initially you acquired an aim based on a desire, you later promised that you would do what you desire to do. Subsequently you lose the desire but keep the intention and hence the aim because of your promise. ⁵ One may wonder whether it is also possible to have an aim that one has no commitment towards. In my view, the answer is likely to be yes. It is conceivable that one forms an intention to do something based on a desire and subsequently loses the desire but keeps the intention. If the commitment generated by the desire is the only commitment one has towards attaining the relevant aim, one is now in a position in which one has the aim but no commitment towards attaining it. That said, for present purposes, this possibility is of little interest, which is why I am happy to set it aside here.

   

13

do so. Since then you have also promised that you would run it and have signed a contract that you would do so. As a result, you now have a practical, moral, and contractual commitment towards running it. Recently, however, you have lost any intention to do so, perhaps because your desire to run the marathon was replaced by an even greater desire not to run it. In this case, you no longer have the aim of running a marathon. However, it does not follow that you are no longer committed to running it. While, when you lost the desire, you also lost your practical commitment towards running the marathon, the promise you made and the contract you signed are still binding and so continue to commit you to running it. And if you end up not running the marathon, you do not live up to your commitments. In this way, you may be committed towards attaining success in an AA even though you don’t have the aim of doing so. There are a number of ways in which you may be released from your commitments towards attaining success in an AA. As I already indicated, you may lose interest and as a result lose your desire to run the marathon. This will release you from your practical commitment. And while the mere loss of desire will typically not release you from your moral and contractual commitments, the cooperation of the person you made the promise to and the other party to the contract might.⁶ That said, and most importantly for present purposes, one way in which one can be released from all commitments one may have towards attaining the aim of an AA is by actually pulling off the feat, i.e. by scoring the relevant success. This gives us the following first crucial thesis about AAs: Commitment Release If, at t, one attains the aim of a given AA, then, at t, one is released from all commitments towards attaining this aim.

⁶ In addition, one may have a moral commitment to fulfil other commitments, such as contractual ones, that one has taken on. In that case, one won’t be released from one’s moral commitment unless one has been released from one’s contractual commitment. The same may hold for other types of commitments.

14 , ,   For instance, when you have covered the distance, you will be released from your practical, moral, and contractual commitments to running the marathon. With these points in play, consider the following case: The Hire You are a geologist. I have hired you for two weeks to find out whether (D =) a certain mine that I am considering buying still has diamonds in it. Since I need to be in a remote location with no means of communication for the next two weeks, we agree to meet at the mine two weeks from now. You send your team of workers to the mine. After a day’s work, they bring you a sample of a deposit of stones they found. You run all available tests on the sample of the stones, all of which suggest that the stones are indeed diamonds. On the basis of this evidence, you come to believe that D. Since there is still a considerable amount of time before our meeting, you and your team pack your bags and get on the next flight home to spend time with your families. Meanwhile, I return unexpectedly early from my trip to the breaking news that the seller has placed a deposit of fake stones in the mine that are so cleverly crafted as to be indistinguishable from real ones by currently available tests. What’s more, I also learn that the fake stones were placed exactly where you found the deposit and that you have since left to see your family. You are currently back home with your family and entirely unaware of the news. The final twist in the story is that, unbeknownst to everyone, there actually is a deposit of real diamonds in a hidden corner of the mine.

Notice that your belief that there are diamonds in the mine was formed on the basis of excellent evidence, evidence that is strong enough to ensure that your belief that D is justified. In addition, your belief that D is true: there are diamonds in the mine, albeit in a hidden corner. That means that you have a justified true belief that D. According to two out of the three of the main contenders in the debate on the aim of inquiry—to wit, True Belief Aim and Justified Belief Aim—you have attained the aim of your inquiry into whether D. In conjunction with Commitment

   

15

Release, these views predict that you are released from all commitments towards settling the question whether D. Now here is the key question: is this prediction correct? In particular, are you released from your contractual commitment to settle the question whether D? No. As far as your contract is concerned, what you ought to be doing is to inspect the mine rather than taking time off to see your family. To see this, put yourself in my shoes (i.e. in the shoes of your employer). I know that what you found wasn’t a deposit of diamonds. While I might concede that you have an excuse for no longer working in the mine, I could rightly insist that you go back to work and fulfil your contract. Crucially, there is no need to negotiate a new contract with you. The old contract is still binding. None of this would make any sense if you had been released from your contractual commitment. Thus, the prediction that you are released from all commitments towards settling the question whether D, which both True Belief Aim and Justified Belief Aim generate in conjunction with Commitment Release, is mistaken.⁷ In contrast, Knowledge Aim does not make a mistaken prediction here. After all, the case is a standard Gettier case and so your belief falls short of knowledge. According to Knowledge Aim, then, you do not attain the aim of your inquiry into whether D. Knowledge Aim, therefore, does not predict that you are released from your contractual commitment to settling the question you were employed to settle. Knowledge Aim is thus preferable to the competition on this count. Argument 2: Progress. Let’s move on to the second argument. Often AAs admit of varying degrees of progress towards attaining their aims. Progress here is a function of times and distances from attaining the aims. One has made progress on a given AA between t1 and t2 if and only if, at t2, one is closer to attaining the aim of this AA than at t1. Suppose you are about to cross the finishing line of the marathon you are

⁷ Compare also: You have contracted a moving company to move all your furniture from your old address to the new one. So long as the company has moved only part of your furniture, they have not fulfilled the contract. As far as the contract is concerned, they ought to continue working on moving your furniture rather than call it a day early, say. You may also rightly take steps to get them back to work, without first having to negotiate a new contract. This is so even if they have an excellent excuse for having called it a day, say, because they justifiably believe that they have already moved all your furniture.

16 , ,   currently running. You have made progress compared to when you crossed the starting line because you are now closer to attaining its aim. It is not hard to see that this definition of progress yields the following thesis: Progress If, at t2, one has not attained the aim of a given AA and if one makes progress towards attaining its aim between t1 and t2, then one has not attained its aim at t1 either.

With Progress in play, let’s return to inquiry once more. In particular, consider the following case: The Insight The local prison at which you are a warder has received a tip that an inmate is in possession of illegal drugs (henceforth ‘is holding’ for short). You are charged with finding out whether this is correct. The first inmate you investigate has a history of drug dealing and abuse. You find a little bag of white powder under their bed. Moreover, upon questioning, the inmate credibly admits to being in possession of drugs. On the basis of this evidence, at t1, you come to believe that some inmate is indeed holding. At t2, however, you discover that the bag does actually not contain drugs and that the confession was false. Perhaps the inmate himself thought that someone else had planted drugs on him and was confessing to minimize the impending punishment. So, you go back and resume your inquiry. At the same time, some inmate is indeed holding. It is just that it isn’t the one you investigated.

Now here is what I take to be an overwhelmingly plausible claim about the Insight: you make progress on your inquiry into whether some inmate is holding between t1 and t2 (henceforth also ‘the crucial claim’). By way of support for the crucial claim, notice, first, that, at t2, you have discovered that a certain piece of evidence that, at t1, appears to support

   

17

the target proposition is misleading. Second, at t2, you can tick one person off the list of suspects you were not able to tick off at t1. The second reason why Knowledge Aim is preferable to the competition is that only Knowledge Aim is compatible with the crucial claim. Here is why: according to the crucial claim, you made progress on the inquiry into whether some inmate is holding between t1 and t2. By Progress, it follows that at t1 you did not attain the aim of your inquiry into this question. However, at t1 you have a justified true belief that M. According to both True Belief Aim and Justified Belief Aim, then, you did attain the aim of your inquiry into whether some inmate is holding at t1. The competition is thus unable to accommodate the crucial claim. In contrast, Knowledge Aim encounters no difficulties here. After all, at t1, your belief that some inmate is holding is gettiered and so does not qualify as knowledge. According to Knowledge Aim, you do not attain the aim of your inquiry into whether some inmate is holding at t1. As a result, Knowledge Aim is entirely compatible with the crucial claim. Generalization. I take these arguments to make a pretty compelling case that Knowledge Aim is preferable to the rival views that have been defended in the literature. While this constitutes considerable progress towards establishing that the aim of inquiry does indeed require knowledge, it might be thought that it doesn’t quite do the trick. After all, couldn’t there be further rivals that escape these arguments? Arguably, the answer to this question is no. For any rival to Knowledge Aim, R, that imposes a weaker requirement than knowledge, there will be some Gettier case in which the weaker requirement is satisfied but the agent doesn’t know the target proposition (P). What’s more, in any such case, there will be a fact (F) in virtue of which the agent fails to know. For instance, in the Hire, P = D = there are diamonds in the mine and F = the stones in the deposit you discovered are fakes. It is now easy to see how the above argument can be recovered for any R that imposes a weaker requirement on the aim of inquiry than knowledge. All that needs to be done is to construct the relevant Gettier case and update P and F in the above cases accordingly. The resulting cases will favour Knowledge Aim over any weaker R. And the same goes, mutatis mutandis, for the Insight.

18 , ,   By the same token, there is reason to think that the aim of inquiry requires knowledge.

1.2 The Aim of Inquiry Requires Nothing More than Knowledge Let’s move on to the second part of my argument for Knowledge Aim. To begin with, recall that I said that the aim of inquiry into whether p can be characterized in a lightweight fashion as aiming at properly closing for oneself the question whether p in the affirmative/negative. I’d now like to bring into play what I take to be a very attractive thesis about belief. Here goes: Belief One has a belief on whether p if and only if inquiry into the question whether p is closed for one in the affirmative/negative.

Does Belief offer an account of the nature of belief? Some have argued that the answer is yes.⁸ While I will get back to this issue in due course, note that for now it’s not necessary to embrace so strong a claim. All I need is that it is a truth about belief, whether or not it also captures its essence. This much, however, I will assume without further argument, at least for now. Since the aim of inquiry into whether p is to properly close inquiry into whether p for oneself, one will attain the aim of inquiry into whether p if and only if one does indeed properly close inquiry into whether p for oneself. In conjunction with Belief, we get the result that one attains the aim of inquiry into whether p if and only if one has a proper belief on whether p.⁹ By the same token, any belief on whether p whatsoever will ⁸ See Pamela Hieronymi’s (2009) insightful paper, which also makes a convincing case for this claim (cf. Friedman 2019, Harman 1986). ⁹ Note that the idea that the conditions for success in inquiry into whether p are the conditions for proper belief that p, in conjunction with the fact that the leading candidates in the debate on the aim of inquiry are knowledge, true belief, and justified belief, gives us reason to expect that there are three main candidates for proper belief, i.e. knowledge, truth, and justification. Note that this, in turn, tallies nicely with the fact that the leading candidates in

   

19

ipso facto be improper if and only if it does not attain the aim of inquiry into whether p. Crucially, it is implausible that there should be a requirement stronger than knowledge such that any belief that doesn’t satisfy it will ipso facto be improper. By way of evidence, let’s look at the perhaps most obvious candidates for stronger requirements. These are higherorder knowledge, understanding why, and certainty. There is excellent reason to think that it is not the case that any belief whatsoever that doesn’t satisfy these requirements will ipso facto be improper. To see this, consider a small child who believes that there is a dummy over there. Suppose this child has not acquired the concept of knowledge yet and so isn’t even in the ballpark for higher-order knowledge. Does it follow that her belief is ipso facto improper? Clearly not. Or consider your belief that Razorlight are playing in Manchester later this year. Suppose you don’t have a belief about why they are playing and so haven’t attained understanding why they are playing and suppose you are not certain that they are playing. Does it follow that your belief is ipso facto improper? Again, that doesn’t seem right. But, then, by the same token, there is reason to think that proper belief does not require higher-order knowledge, understanding why, or certainty. What about less obvious candidates? For instance, what about an account of proper belief that requires what we may call knowledge+, where one knows+ that p if and only if (i) one knows that p and (ii) one’s degree of justification for p is ever so slightly higher than what’s required for knowing that p? To assess the prospects of a knowledge+ account, note that there will be cases in which one knows that p but doesn’t know + that p. The key question now is whether, in such cases, one’s belief that p is ipso facto improper? It seems to me that the answer here is no. But even if you don’t share this intuition, note that knowledge+ is not a property that we use in our evaluation of beliefs, nor one that we might easily come to so use. The vast majority of us don’t even have the concept of knowledge+. As a result, the vast majority of us don’t even have

the related debate over the norm of belief are knowledge and truth. Moreover, it provides evidence that construing the aim of inquiry in terms of properly closing questions does not beg the question in favour of Knowledge Aim. After all, truth and justification remain live candidates for what it takes for a belief to be proper.

20 , ,   the conceptual resources to evaluate beliefs in terms of knowledge+. No surprise, then, that knowledge+ does not have much currency in our everyday lives. Even if the above intuition isn’t enough to establish that the knowledge+ account of proper belief is false, at least there is warrant for setting it aside on the grounds that it has no role to play in our evaluations of beliefs. And, of course, the same goes, mutatis mutandis, for other candidate accounts of proper belief such that most of us don’t have the concepts needed to deploy it. Finally, might there not be candidates for a general account of proper belief featuring a stronger requirement than knowledge other than the obvious ones and ones like knowledge+? For what it’s worth, I cannot think of any. But that may, of course, be my fault. The best I can do for now is to leave it to my critics to put their proposals on the table. I’d be more than happy to look into their prospects. I have argued that the aim of inquiry requires knowledge and that it requires nothing more than knowledge. In this way, then, there is reason to think that knowledge is indeed the aim of inquiry. This completes my positive case for Knowledge Aim.

2. Objections and Replies While I think that there is strong reason to accept Knowledge Aim, the literature also contains a number of arguments against it. In what follows, I will look at the most popular candidates. My aim is to show that all of them remain ultimately unsuccessful. Before getting down to the nitty-gritty, I’d like to point out that there is at least one important argument that I will not discuss in any detail. In a nutshell, the idea here is that knowledge cannot be the aim of inquiry because knowledge has no role to play in conducting inquiry (e.g. Kaplan 1985). Now, I take it that the best that champions of this argument can hope for is to pose a challenge for champions of Knowledge Aim: they need to explain what the role of knowledge in inquiry might be. And since the arguments for Knowledge Aim I just offered do precisely that, I will not address this argument in detail. With this disclaimer in play, let’s get down to business.

   

21

2.1 Against the Arguments for Knowledge Aim The first two objections I would like to consider are due to an anonymous referee. They aim to show that the arguments in Section 1 above for Knowledge Aim do not provide support for Knowledge Aim after all. Well-Conducted Inquiry. The key thought of the first objection is that champions of rival views might hold that, in the Hire and the Insight, you do attain success in inquiry but that your inquiry falls short in a different way. To see why one might think this, note first that, in both cases, the inquiries are complex activities that feature further nested inquiries. For instance, in the Hire, your inquiry into whether D is true proceeds by the nested inquiry of inspecting some diamonds, coming to believe that (R=) the diamonds you inspected are real, and inferring that D is true. Second, and crucially, the thought is that while the complex inquiry is ultimately successful, the inquiry is not well conducted because it relies on the results of unsuccessful nested inquiries. For instance, in the Hire, the inquiry proceeds via a sub-inquiry that delivers the false belief that R. In this way, it may be possible to resist the above argument, at least for champions of True Belief Aim. In response, it’s not so clear exactly how this is supposed to help mitigate the force of the arguments. Consider the Insight first. Here, you do indeed make a mistake. You acquire a false belief that the first inmate you investigated was holding. However, you find out about it and correct it. As a result, your overall complex inquiry may yet be not only successful but also well conducted in that it does not rely on the results of unsuccessful nested inquiries. What’s more, the key claim in this argument is that when you realize your mistake you make progress in your inquiry. Whether your complex inquiry is ultimately successful and/or well conducted in the relevant sense is orthogonal to this claim and hence to the success of the argument. While it will do for my purposes if one of my arguments is successful, it may be worth considering whether this objection can get traction against the first argument. Let’s return, then, to the Hire once more. At first glance, things may look promising here. After all, as we have already seen, your inquiry is once again a complex one. More importantly yet, it does rely on the results of an unsuccessful nested inquiry.

22 , ,   On reflection, however, it is once again unclear how this helps. After all, the fact remains that Commitment Release in conjunction with True Belief Aim delivers the result that you are released from your contractual commitment. And that still doesn’t seem right. What champions of True Belief Aim need to do in order to block the above argument is to resist Commitment Release. Simply countenancing a further normative property of inquiry won’t be enough. But might they not resist Commitment Release by rejecting it in favour of a principle that replaces all occurrences of ‘success’ and its cognates by ‘well-conductedness’ and its cognates? On the face of it, that would do the trick. After all, since your inquiry isn’t well conducted, the new principle doesn’t deliver the result that you are released from your contractual commitment. Unfortunately, there is reason to think that even this move won’t work, for two reasons. First, one might wonder why Commitment Release should be about well-conductedness rather than success. After all, what we are talking about is activities with an aim. It would be surprising that key principles governing activities with aims such as Commitment Release should be about well-conductedness rather than success. Second, we can tweak the case such that it is well conducted after all. To see this, just suppose that by accident a small number of real diamonds were dropped in the deposit alongside all the fakes and that, by yet another accident, the sample you examined contained all and only the real ones. Now even the belief that R from which you infer and come to believe the target proposition that D is true. As a result, your inquiry is well conducted in the relevant sense. However, that doesn’t change the fact that you are not released from your commitment and that I may take steps to get you back to work in the Hire. The Terminus of Inquiry. The second objection introduces the notion of the terminus of an activity, which is the point at which an agent can bring the activity to a rational close. The key thought now is that knowledge that one has attained its aim is the terminus of any activity with an aim. If this is correct, then the terminus of inquiry into p is going to be knowledge that p/not-p if the aim of inquiry is true belief, as per True Belief Aim. The question that arises is whether the data that I have

   

23

taken to support Knowledge Aim can’t really be explained by the idea that knowledge is the terminus of activities with aims and hence of inquiry. For instance, in the Hire, when you arrive at a true belief that D, according to True Belief Aim you attain success in your inquiry into whether D. However, if knowledge is the terminus of inquiry, then you do not attain the terminus of your inquiry into whether D. And perhaps this is what explains the problematic data. In response, first, there is reason to think that knowledge of success is not the terminus of any activity with an aim. To see why not, suppose that you are running a marathon. Unbeknownst to you, the finishing line has been wrongly placed at 25 miles. When passing the finishing line, you will bring your running activity to a rational close. Yet you don’t know that you have successfully run a marathon. After all, you simply haven’t covered the distance. It comes to light that you can bring an activity to a rational close without knowing that you have attained its aim. Here is a more plausible account of the terminus of any activity with an aim. First, belief that one has attained the aim normally closes the activity. When you think you have done it, you will normally stop doing what you think you have done. For instance, when running a marathon, you will normally stop running when you believe that you have run the distance. Second, rational belief that one has attained its aim brings the activity to a rational close. For instance, when running a marathon, you will rationally stop running when you rationally believe that you have run the distance. It is easy to see that this alternative view avoids the problems of the knowledge view. In the original marathon case, when you cross the finishing line, you acquire a rational belief that you have covered the distance. That’s why it’s rational to stop running.¹⁰ ¹⁰ Suppose my stomach is currently digesting the breakfast I ate earlier today. It is engaged in an activity with an aim, to wit, the digestion of the breakfast. Can this activity be brought to a rational close? If the answer is yes, then even the above account of the terminus of activities with aims in terms of rational belief is too strong. After all, neither I nor my stomach has beliefs on the matter. That said, this further development won’t be of much help for champions of the present strategy. After all, whatever it is that does feature in the correct account of the terminus of activities with aims cannot be success-entailing, for the reasons outlined above. It may also be worth noting that these considerations also help with a looming overintellectualization worry for champions of Knowledge Aim. In conjunction with the account of the terminus of activities with aims in terms of rational belief, Knowledge Aim entails that the

24 , ,   Second, once it is clear that rational belief rather than knowledge is the terminus of activities with aims, there is little reason to think that the data that I took to support Knowledge Aim can be explained in terms of the terminus of inquiry. The reason for this is that, to explain the trouble that champions of True Belief Aim encounter as a result of allowing that success in inquiry has been attained in cases like the Hire and the Insight, it had better be the case that the terminus of inquiry is not attained in these cases. Or to be more precise, it had better be the case that the terminus of inquiry is not attained at the crucial junctures at which, according to True Belief Aim, success is attained and which cause trouble for its champions. The problem is that rational beliefs are present at these junctures. As a result, there is excellent reason to think that the data that I took to support Knowledge Aim cannot be explained in terms of the terminus of inquiry. It may be useful to look at how this plays out in a particular case. To this end, let’s return to the Hire once more. Recall that what needs to be explained is the datum that you are not released from your contractual commitment to inquire into whether D. Since you have a true belief that D, champions of True Belief Aim cannot do so by holding that you didn’t succeed in inquiry. The question is whether they can do so by appealing to the distinction between success in inquiry and the terminus of inquiry. Now, the only way this distinction might do the trick is as follows: although you did attain success, you didn’t attain the terminus of your inquiry. Since your true belief falls short of knowledge, it is also easy to see that if the terminus is knowledge, then you haven’t attained the terminus of your inquiry into whether D and the envisaged strategy may work in the way intended. But once we are clear that the terminus of activities with aims is rational belief (rather than knowledge), it is also clear that you have attained the terminus of your inquiry into whether D. After all, your belief that D is rational. As a result, the prospects for terminus of inquiry is rational belief that one knows. If we want to acknowledge that unsophisticated agents who don’t have the concept of knowledge can reach the terminus of inquiry, we may legitimately worry that the present account is too demanding. My strategy for dealing with this is as follows: if we want an account of the terminus of activities with aims that applies to agents who don’t have the relevant beliefs or concepts, there is independent reason for thinking that it will have to be something weaker, as the case of my stomach digesting my breakfast clearly indicates.

   

25

explaining the relevant data by appealing to the notion of the terminus of inquiry are dim, on reflection.

2.2 The Aim of Inquiry Can’t Require Truth The first two objections took issue with my arguments for Knowledge Aim. They aimed to show that the arguments do not support Knowledge Aim. The remaining objections switch gear and opt for a more robust approach. They all aim to show that Knowledge Aim is false for one reason or another. The first objection I will consider ventures to show that the aim of inquiry cannot be thought to involve truth. It will come as no surprise that this line has been pursued mainly by champions of Justified Belief Aim. And while versions of it have been given by a number of different champions of this view, it is particularly clearly stated in the following passage by Davidson: [T]ruths do not come with a ‘mark’, like the date in the corner of some photograph, which distinguishes them from falsehoods. The best we can do is test, experiment, compare and keep an open mind. But no matter how long and well we and coming generations keep at it, we and they will be left with fallible beliefs. We know many things, and will learn more; what we will never know for certain is which of the things we believe are true. Since it is neither visible as a target nor recognizable when achieved, there is no point in calling truth a goal. (Davidson 2005, 6)

There are actually two arguments in this short passage. Both lay down a condition on aims. The first argument holds that aims must be visible as targets. The second, in contrast, claims that aims must be recognizable as having been achieved when they have been achieved. Crucially, the argument continues, truth doesn’t have either property and so cannot be an aim. A fortiori, it cannot be the aim of inquiry. While I could discuss the arguments individually, I won’t. One reason for this is that if I manage to successfully argue that these arguments fail,

26 , ,   its champions might just try to replace visibility or recognizability with a different condition. Rather, what I’ll do is focus on what I take to be the underlying structure of the argument. Note that both conditions first place an epistemic accessibility requirement on aims (visibility, recognizability) and then move on to the claim that truth does not satisfy this accessibility requirement. The Inaccessibility Argument 1. For something to be an aim, it must satisfy some epistemic accessibility requirement. 2. Truth does not satisfy this epistemic accessibility requirement. 3. Hence, truth cannot be an aim. This argument fails. The reason for this is that there no is way of unpacking the epistemic accessibility requirement that makes both (1) and (2) true. Let’s return to Davidson once more: the above passage suggests that he takes it that in order to satisfy the epistemic accessibility requirement, you must know for certain that you have attained the relevant success. This promises to make (2) true, at least on the plausible assumption that there is fairly little in general that can be known for certain. At the same time, (1) comes out clearly false. Completing a marathon can clearly be an aim. At the same time, it’s not possible to know for certain that you have done so when you have done so. After all, despite your best efforts to ensure otherwise, it may always be the case that something went wrong with the equipment used to measure the distance with the result that you stopped running prematurely and so failed to cover the distance. Note also that this doesn’t even start mentioning the more extravagant types of error possibility (Cartesian demons, etc.) that are standardly taken to show that it is very hard to know anything for certain. On the other hand, we can opt for a weaker account of the epistemic accessibility requirement in order to give (1) a better chance of coming out true. For instance, we might unpack the epistemic accessibility

   

27

requirement in terms of knowledge that you have attained the aim. To see why this promises to make (1) true, note that, in the marathon case, you can plausibly know that you have run the distance, for instance in virtue of knowing that if you have crossed the finishing line, you have run the distance and coming to know that you have crossed the finishing line. (I hasten to add that you will not come to know that you have run a marathon if your epistemic environment is not hospitable in that the equipment used to measure the distance was malfunctioning, Cartesian demons were meddling, etc. Crucially, however, the fact that you may fail to know (e.g. when the environment is not hospitable) does not entail that you don’t frequently come to know (e.g. when it is not).) In this way, the weaker account of the epistemic accessibility requirement gives us a better chance of making (1) come out true. Unfortunately, now there is excellent reason to think that (2) is false. After all, it is hard to deny that we can come to know the truth. Suppose I wonder whether Fermat’s last theorem is true. It is hard to deny that I can come to know that it is true, for instance, by asking my mathematics teacher. Again, this is not to say that I might also fail to acquire this knowledge, for instance, if I am in an epistemically unsuitable environment. But now that I am not, there is no in principle obstacle for me to come to know what I want to find out. As a result, the weaker account of the epistemic accessibility relation will make (2) come out false. In general, the trouble with the Inaccessibility Argument is that it is just not clear why it should be any harder to satisfy any given accessibility requirement for truth than for any other aim. That’s why ways of unpacking what it takes to recognize something that makes (1) true will make (2) false and vice versa.¹¹

¹¹ It might be thought that this is not entirely fair to champions of the inaccessibility argument. After all, when offering this argument, they typically have a specific account of truth in mind, to wit, one according to which truth is objective and mind-independent and may fail to be present even when we have all agreed on the answer to the questions, have the best possible justification for our answer, etc. But then again, many things are objective and mindindependent and may fail to be present even when we have agreement, the best possible justification, etc. Rocks, lakes, and trees are among the many points in case. Crucially, most of us do not have difficulties in recognizing a rock, a lake, or a tree when we see one. Or, to be more precise, we do not have any such difficulties unless the requirements for what it takes to do

28 , ,  

2.3 The Aim of Inquiry Can’t Be Knowledge The second objection to Knowledge Aim I’ll discuss here is due to Jonathan Kvanvig. Kvanvig’s thought is that knowledge can’t be required for success in inquiry because true belief is all that’s needed. He offers two arguments to this effect, one from the first person point of view, the other from the third person point of view. Let’s start with the former. Here is Kvanvig: My concern here is with attributing complex intentional states to small children and nonhuman animals that lack the conceptual resources to be accurately characterized by those states and yet . . . engage in inquiry. Given this concern, the concept of truth has advantages over those of knowledge, for knowledge is a more complex concept than truth: A cognitive being can have the concept of truth without that of knowledge, but not vice versa, because truth is a conceptual component of knowledge. Still, the concept of truth is not itself innate, leaving open so are unduly stringent. Objectivity and mind-independence are not the kinds of property that make recognition hard either. But perhaps even this isn’t the full story. Perhaps what champions of the inaccessibility argument have in mind is the following: if truth is objective and mind-independent, then, in order to recognize that it is present, we’d have to step outside our skin and compare the naked facts with the contents of our beliefs. It’s not possible to recognize that the truth is present when it is because it is not possible to do that. I do not mean to deny that if that were indeed required for recognizing a truth to be present, it would be impossible to do so. The trouble is, of course, that there is precious little reason to think that, in order to recognize truths, we’d have to do anything as fancy as that. Compare: in order to recognize that Sandy Goldberg is American, I do not need to see his certificate of citizenship. I can recognize that he is American from the way he sounds. No fancy methods are required here. Similarly, it is just not clear why recognizing truths requires fancy methods. For instance, it’s not clear that in order to recognize that it is true that there is milk in the fridge I need to do much over and above exercising my ability to recognize milk. Champions of the inaccessibility argument might concede the point but worry that I have moved away from an objective account of truth to a minimal one and that their argument was never meant to work on minimal accounts of truth in the first place. On reflection, this objection does not stand up to scrutiny. The above point was that you may be able to recognize Fs even though you don’t have the opportunity and perhaps are unable to get into direct cognitive contact with the features in virtue of which Fs are Fs. For instance, you might be able to recognize Sandy Goldberg as being American even though you don’t have the opportunity to get into direct cognitive contact the relevant properties in virtue of which he is American. If so, there is little reason to think that the fact that you are unable to get into direct cognitive contact with the naked facts will prevent you from recognizing truths.

   

29

the possibility that . . . one can engage in inquiry absent even the concept of truth. Even so, such inquiry . . . is teleological, directed at some goal. The aim of such inquiry, from within the intentional states of the cognizer, is not to find the claims that are true, for such a characterization requires possession of the concept of truth by the cognizer. From the perspective of the cognizer with respect to a particular proposition p, the goal in question is to ascertain whether p or not-p, not to ascertain whether or not p is true. Once we see the teleological nature of inquiry . . . in this way, the most accurate way to describe it from the outside is in terms of . . . truth, for the concept of truth allows us to generalize across all the particular instances of attempting to ascertain whether p or not-p, which is the accurate portrayal from the inside of the nature of inquiry. . . . It is in this way that the goal of inquiry . . . [is] to be identified with finding the truth. (Kvanvig 2003, 145–6)

Here is my reconstruction of the argument: The First Person Argument 1. From the first person perspective of an unsophisticated cognizer who does not have the concepts of truth and knowledge (= UC), it’s most accurate to characterize the aim of inquiry as ascertaining whether p or not-p. 2. If, from the first person perspective of a UC, it’s most accurate to characterize the aim of inquiry as ascertaining whether p or not-p, then, from the third person perspective, it’s most accurate to characterize the aim of inquiry in terms of truth rather than knowledge. 3. Hence, from the third person perspective, it’s most accurate to characterize the aim of inquiry in terms of truth rather than knowledge.

30 , ,   One obvious difficulty with this argument lies with premise 1. To see this, note that agents who do not have the concepts of truth and knowledge will not (or at any rate need not) have the concept of ascertaining either. As a result, from the perspective of unsophisticated inquirers, the aim of inquiry cannot plausibly be characterized as involving ascertaining. The only way in which the aim of inquiry can be thus characterized is from the outside. Note that even if this problem can be fixed and a more suitable characterization of the aim of inquiry from the point of view of a UC can be found, there is reason to think that the argument will not go through because premise 2 is no less problematic. To see why, note that the aim of inquiry can in any case not only involve the concept of truth. After all, to attain the aim of inquiry, it’s not enough that a certain proposition be true; the inquiring agent must also be related in the right way to this truth; more specifically, they must also believe it. But now note that unsophisticated inquirers may not have the concept of belief either. If so, there is excellent reason to think that the accurate portrayal of the aim of inquiry from the first person perspective of an unsophisticated agent, whatever it may be, is simply not a good guide to the correct characterization of the aim of inquiry from the third person perspective. Finally, we can of course agree with Kvanvig that the aim of inquiry is to ascertain whether p/not-p, from the third person perspective, that is. But, of course, that simply leaves open the question as to what it takes to ascertain whether p/not-p, i.e. whether that involves acquiring knowledge that p/not-p, a true belief that p/not-p, etc. Section 1 above has made a case that it involves coming to know. And, as we have now seen, the First Person Argument provides no compelling reason for thinking otherwise. Kvanvig’s first argument fails. But perhaps the second does better. Here is Kvanvig once more: Another way to assess the goal of inquirers from a third-person perspective is to consider what their behavior would be if they were given full information about their situation. For example, if we consider the situation of an inquirer whose appetite for information has

   

31

been sated by some found answer, but who does not know that answer to be correct, we could ask what such an inquirer would do if he or she was informed that the answer was not known to be correct. It is plausible to assume that inquiry would resume under such conditions, and this fact may be thought to lend credence to the view that knowledge is the goal of inquiry. (Kvanvig 2003, 47)

In a nutshell, then, the thought here is that looking at how inquirers would behave if they were given full information about their situation allows us to provide reason for thinking that some particular thesis about the aim of inquiry is correct. In particular, the fact that inquirers would resume inquiry upon being informed that their belief falls short of knowledge provides some reason in favour of Knowledge Aim. Of course, Kvanvig does not think that this argument for Knowledge Aim is ultimately successful, for two reasons. Here is one: The first is that we would get the same result if the person in question had not found the truth and were informed of such. So this argument for a connection between [the aim of inquiry]¹² and knowledge is no better than a similar argument connecting [the aim of inquiry] with discovered truth. (Kvanvig 2003, 147)

To see just why this response fails, let’s first consider two games involving basketball: in the first (call it three-pointer), the aim is to take shots until one sinks a three-pointer; in the second (call it mid-court), the aim is to take shots until one sinks a shot from the far side of the mid-court line. Note that the aim of mid-court is logically stronger than the aim of three-pointer. After all, making a mid-court shot entails making a threepointer but not vice versa.

¹² Kvanvig himself talks about curiosity here. Since my main interest here lies with the aim of inquiry, I have reformulated the argument accordingly. I don’t think that this does Kvanvig an injustice. After all, he takes himself to be arguing that the above argument that knowledge is the aim of inquiry fails.

32 , ,   With these points in play, suppose you are playing a game of midcourt. Of course, we would expect you to resume taking shots if someone informed you that were not on the far side of the mid-court line. At the same time, we’d also expect you to resume taking shots if someone informed you were not beyond the three-point line. After all, if you are not beyond the three-point line, you are not beyond the mid-court line either. Informing you that you weren’t beyond the former is effectively informing you that weren’t on the far side of the latter. No surprise, then, that you would resume taking shots. Suppose, next, that you are playing a game of three-pointer. Here, we’d, of course, also expect you to resume taking shots if you are informed that you are not beyond the three-point line. But now suppose that, in a game of three-pointer, you are informed that you are not beyond the mid-court line. Would you resume taking shots? Presumably not. Rather what you’d do is ask for clarification as to how this piece of information is relevant. Finally, suppose that some people are playing a game involving shooting basketball. You know that it’s either mid-court or three-pointer but you don’t know which. You are trying to arrive at an answer by looking at how players would behave if they are given various pieces of information. Suppose you discover that they resume taking shots upon being informed that they are not on the far side the mid-court line (rather than asking why this is relevant, say). In this case, you have reason for thinking that they are playing mid-court rather three-pointer. Crucially, the fact that they would equally resume taking shots were they informed that they were not beyond the three-point line does precious little to change this. After all, since not being beyond the three-point line entails not being beyond the mid-court line, that’s also entirely as it should be. The argument that they are playing mid-court rather than three-pointer stands. It should now be easy to see why Kvanvig’s first response to the argument he considers fails. Just as the aim of mid-court is logically stronger than the aim of three-pointer, what champions of Knowledge Aim take the aim of inquiry to be is logically stronger what champions of True Belief Aim take it to be. After all, knowledge entails true belief but not vice versa. Note next that what Kvanvig proposes on behalf of champions of Knowledge Aim is basically to arrive at an answer as to

   

33

whether the aim of inquiry is knowledge or true belief by looking at how inquirers would behave if they are given various pieces of information. He concedes that upon being given the information that inquirers don’t know, they would resume inquiry but claims that this provides no reason for Knowledge Aim over True Belief Aim because we’d get the same result if we give inquirers the information that their beliefs aren’t true. But, as the basketball cases make abundantly clear, this is not correct. On the contrary, the fact that inquirers would resume inquiry upon being informed that they lack knowledge (rather than asking why this is relevant, say) does provide reason to think that Knowledge Aim rather than True Belief Aim is correct. The argument Kvanvig considers thus stands. But perhaps Kvanvig’s second reason for thinking that the argument doesn’t work fares better: Here it is: [I]f we are to compare these two proposals [i.e. Knowledge Aim and True Belief Aim], we should inform our inquirers fully of their precise condition. Upon having their appetite for information sated, they are in one of three conditions: (i) they have not found the truth but think they have; (ii) they have found the truth but not in such a way that they have knowledge; and (iii) they have found the truth and know the information in question to be true. I submit that the only one of these three situations in which inquiry would resume would be condition (i). To inform someone that she had found the truth but had fallen short of knowledge would be met, I submit, with relative indifference. (Kvanvig 2003, 147–8)

In fact, it’s clear that Kvanvig takes the above considerations not only to indicate that the argument for Knowledge Aim he considers is unsuccessful but also to provide reason to think Knowledge Aim is false. After all, as we fill in the blanks for inquirers even further, in the sense that we provide information not only about whether their beliefs qualify as knowledge but also about whether they are true, it becomes clear that what really matters to whether inquirers would resume inquiry is not whether they know but whether they have the truth. Since, upon fuller

34 , ,   information, resumption goes with truth, at the end of the day, there is reason for thinking that True Belief Aim is correct. Here is one way in which champions of Knowledge Aim may respond to Kvanvig’s argument. If one is informed that p is true, one will typically come to know that p. As a result, even though one may not have known before, one knows now, in which case it’s entirely in line with Knowledge Aim that inquiry would not resume in this case.¹³ Kvanvig grants this point but insists that there is a kind of case that will do the job for him, which I will henceforth refer to as the Mixed Track Record case. Here it is: It is true that testimony often generates knowledge, but it is also possible that a testifier is someone with sufficient credibility to give one grounds for doubt that one has knowledge, but not with sufficient credibility to generate knowledge of that to which he or she testifies. Cases of this sort occur, for example, when the testimony is that of experts with a mixed track record. Their track record might be good enough to provide grounds for doubting that we are right when they disagree with us, but spotty enough that their word is insufficient for knowledge. When such a testifier says that one has found the truth but lacks knowledge, such testimony does not impart knowledge, but neither would it be, in ordinary cases, a cause for further investigation (when the motivation is merely that of curiosity). ‘Yeah, yeah, who cares?’ and ‘Whatever’ come to mind when imagining how the curious might respond to such testimony. (Kvanvig 2003, 148)

It’s not hard to see how Mixed Track Record cases may be thought to get Knowledge Aim into trouble. In these cases, you have a belief that p. You are also fully informed about your situation in the sense that you have been given not only the information that you don’t know that p but also that p is true. Finally, the case is set up such that you do not thereby come to know that p. Since here you would still not resume inquiry, there is

¹³ In fact, I am not convinced that this is the right account of what’s going on here, for reasons I’ll explain in due course.

   

35

reason to think that the aim of inquiry requires only true belief, not knowledge. Unsurprisingly, I think the Mixed Track Record cases do not serve to make a convincing case against Knowledge Aim. To drive this point home, I’d like to start by asking why, according to champions of True Belief Aim, it is appropriate to respond to the assertion that your belief is true but falls short of knowledge by saying ‘Yeah, yeah, who cares?’ or ‘Whatever’ (henceforth also the ‘Whatever’ response for short). The obvious answer is that the ‘Whatever’ response is appropriate because the assertion is simply irrelevant to you. You are interested in whether you have attained the aim of inquiry. According to True Belief Aim, you have done so when you have a formed true belief. But, then, when someone points out to you that your belief is true but falls short of knowledge, that’s irrelevant. And, of course, irrelevant assertions can appropriately be dismissed by the ‘Whatever’ response. In what follows, I will argue that this explanation of the ‘Whatever’ response is not best one on the market. Crucially, a better explanation enables champions of Knowledge Aim to accept Kvanvig’s key claim that, in Mixed Track Record cases, there is no cause for inquiry to resume. As a result, at the end of the day, the argument does not serve to make a compelling case against Knowledge Aim. First, consider a variation of the Mixed Track Record case in which two things happen: an agent with a mixed track record (henceforth M) first tells you that p is true and a little later a highly reliable agent (henceforth R) tells you that you don’t know that p. Is it appropriate for you to dismiss either R or M by a ‘Whatever’ response? Clearly, the answer here is no. And the same goes for another variation of the case in which you are talking sequentially to two agents with mixed track records, M1 and M2, such that M1 first tells you that p is true and a little later M2 tells you that you don’t know that p. Of course, if, in the original version of the Mixed Track Record case, the reason why it is appropriate for you to dismiss M’s assertion by a ‘Whatever’ response is that the assertion is simply irrelevant to you, then the same should go for at least one of the two assertions in these variations of the case. In particular, the assertions telling you that you don’t know should be also irrelevant to you and so should be equally

36 , ,   dismissible by a ‘Whatever’ response. Since that’s not what we find, there is reason to think that irrelevance is not what explains why the ‘Whatever’ response is appropriate in the Mixed Track Record case. But what, then, does explain it? To answer this question, consider yet another variation of the Mixed Track Record case in which the person who tells you that your belief that p is true but falls short of knowledge is R, the highly reliable agent. Recall that Kvanvig grants that the version of the case doesn’t pose a problem for Knowledge Aim because you come to know that p. While this may appear plausible at first glance, on reflection, I am not so sure that this is right. To see this, note that when someone tells you that a certain proposition is true but you don’t know it to be true, this is rather puzzling. Why? Here is one attractive answer. Normally, when someone tells you something, the point is to generate the corresponding belief in you, at the very least.¹⁴ However, it’s hard to see how R’s assertion could do that. After all, you couldn’t come to rationally believe its content.¹⁵ And R surely doesn’t expect you to come to believe something that you couldn’t rationally come to believe, simply because they say so. But then what is the point of R’s assertion? Nothing obvious springs to mind. Of course, given R’s outstanding credentials as a testifier, it may not be appropriate for you to dismiss their testimony by a ‘Whatever’ response. Perhaps you owe it to R to try and find out what they might take the point of their assertion to be. So, suppose you ask: ‘What are you trying to say?’ And suppose R were to respond: ‘Simply that p is true but you don’t know that p.’ At this stage, it does seem appropriate for you to dismiss R by a ‘Whatever’ response. What comes to light, then, is that the ‘Whatever’ response remains appropriate to assertions of p is true but you don’t know that p, no matter whether the speaker has a mixed track record or is highly reliable. We now have the resources to offer a better explanation of why the ‘Whatever’ response is appropriate in the Mixed Track Record case. It is not because the assertion (in particular, the information that you don’t ¹⁴ In fact, my own view is that the function of assertion is to generate knowledge. See (Kelp 2018a) for a defence. ¹⁵ This is a point that is often made by and associated with champions of knowledge-based accounts of justified belief. However, it turns out that it is independently defensible. For a compelling argument (that’s available even to internalists), see Douven (2009).

   

37

know) is irrelevant to you and so can appropriately be dismissed by a ‘Whatever’ response. Rather, it is that assertions of p is true but you don’t know that p cannot come to be rationally believed by the hearer. As a result, they don’t have a discernible point. This will allow us to explain why assertions of p is true but you don’t know it are puzzling no matter whether they are made by highly reliable agents or agents with mixed track records. It also serves to explain why they can appropriately be dismissed by a ‘Whatever’ response.¹⁶ Crucially, there is good reason to think that this explanation allows champions of Knowledge Aim to resist Kvanvig’s argument. To see why, let’s return once more to Kvanvig’s Mixed Track Record case. Even if it is correct that M is a good enough testifier to give you reason for doubt when M disagrees with you, it is far from clear that M’s disagreeing with you gives you reason for doubt on every single occasion. One type of case in which it arguably doesn’t is when you can appropriately dismiss their assertion by a ‘Whatever’ response. But, of course, if M’s assertion doesn’t give you reason for doubting that you know, then, even by the lights of Knowledge Aim, it furnishes no reason for thinking that you haven’t attained the aim of inquiry and so provides no cause for resuming inquiry either. The result that we get, then, is that the fact that, in the Mixed Track Record case, M’s assertion provides no cause for resuming inquiry is entirely compatible with Knowledge Aim. As a result, Kvanvig’s third person argument fails.

2.4 The Aim of Inquiry Can’t Require Justification The next prominent argument against Knowledge Aim ventures to show that ‘our epistemic aim’ cannot entail justification. While it may not be entirely straightforward that this targets the thesis that knowledge is the aim of inquiry, it is close enough to make the argument worth a ¹⁶ Again, in case of a highly reliable testifier such as R we may owe it to them to figure out what the point may be. As a result, we may not be able to dismiss their assertion by a ‘Whatever’ response straight away. But it’s not clear that we have the same obligation towards an agent with a mixed track record such as M. Given that we don’t, the ‘Whatever’ response will be appropriate straight away.

38 , ,   closer look. What’s more, in Chapter 5, I will develop an account of the epistemic domain according to which the domain is constituted by inquiry. As a result, knowledge (rather than true belief) is a for-itsown-sake value relative to this domain, because it is the aim of inquiry. And, as will become clear momentarily, that knowledge (rather than true belief) is a for-its-own-sake value in the epistemic domain is enough to run a version of the argument. Here it is, then, in a passage from Marian David: To say that believing p is justified or unjustified is to evaluate believing p, in some sense, as a good thing or as bad thing, as having some positive status or some negative status. The suggestion is that this type of evaluation, epistemic evaluation, is most naturally understood along broadly teleological lines, as evaluating beliefs relative to the standard, or goal, of believing truth and avoiding error . . . Although knowledge is certainly no less desirable than true belief, the knowledge-goal is at a disadvantage here because it does not fit into this picture in any helpful manner . . . In particular, any attempt to understand justification relative to the knowledge-goal would invert the explanatory direction and would make the whole approach circular and entirely unilluminating. After all, knowledge was supposed to be explained in terms of justification and not the other way round. (David 2001, 154)

Here is my reconstruction of the argument: The Anti-Justification Argument 1. If knowledge is the epistemic aim and epistemic justification is analysed teleologically as derivative from the epistemic aim, then epistemic justification is analysed as derivative from knowledge.¹⁷

¹⁷ Note that teleological analyses typically are thought to proceed from for-its-own-sake value. That’s why the claim that knowledge (rather than truth) is a for-its-own-sake value in the epistemic domain will do to get the argument off the ground.

   

39

2. If epistemic justification is analysed as derivative from knowledge, then knowledge cannot be analysed reductively in terms of epistemically justified true belief. 3. Knowledge can be analysed reductively in terms of epistemically justified true belief. 4. Either knowledge is not the epistemic aim or epistemic justification is not analysed teleologically as derivative from the epistemic goal. [1–3] 5. Epistemic justification is analysed teleologically as derivative from the epistemic goal. 6. Knowledge is not the epistemic goal. [4,5] How convincing is this argument? As a first observation, it is clearly valid. So, let’s grant this much. Instead, let’s take a closer look at the argument’s key premises, i.e. (1), (2), (3), and (5). Now, I take it that neither (1) nor (2) can plausibly be denied. (1) is equivalent to if epistemic justification is analysed teleologically as derivative from the epistemic aim, then: if knowledge is the epistemic aim, epistemic justification is analysed teleologically as derivative from knowledge, which is clearly analytic. Moreover, (2) is true in virtue of the fact that reductive analyses of the sort at issue here cannot be circular (more on this in Chapter 2). After all, if epistemic justification is analysed teleologically as derivative from knowledge, then epistemic justification is analysed in terms of knowledge. If so, any reductive analysis of knowledge in terms of epistemic justification is bound to fall foul of the non-circularity requirement on reductive analyses. In contrast, the prospects for denying (5) may look somewhat brighter at first glance. The perhaps most obvious way of avoiding a teleological account of epistemic justification is by going for a strongly internalist account instead. For instance, one might say that a belief is justified if and only if it is blameless, where blamelessness is not itself unpacked as derivative from the epistemic goal (Kvanvig 2014). One obvious problem with this line is that a version of the Anti-Justification Argument will go through so long as the reductive analysis of knowledge features some key condition that is analysed teleologically as derivative from the epistemic

40 , ,   aim, even if we don’t go on to identify this condition with the epistemic justification condition. And the trouble is that it is eminently plausible that knowledge does feature some such condition. In particular, it is eminently plausible that, in order to know, one must satisfy some form of reliability condition (cf. Goldman 1979, 2012). But, of course, reliability is most naturally analysed teleologically as derivative from the epistemic goal.¹⁸ As a result, there is excellent reason to think that (5) (or a version of (5) that could take its place in a suitably adapted version of the AntiJustification Argument) is independently plausible also. If (suitably adapted versions of) (1), (2), and (5) hold, the AntiJustification Argument (suitably adapted) gives us the following result: if knowledge is the epistemic aim, then knowledge does not admit of reductive analysis in terms of epistemically justified true belief. After all, what the Anti-Justification Argument shows is that (1), (2), (3), and (5) entail that knowledge is not the epistemic goal. So, if (1), (2), and (5) hold, then if knowledge is the epistemic goal, (3) must be false, i.e. knowledge does not admit of reductive analysis in terms of epistemically justified true belief. Now, David goes on to argue that since knowledge does admit of reductive analysis in terms of justified true belief, knowledge cannot be ¹⁸ Might it be possible to resist a teleological account of the reliability condition? Let’s suppose that teleologists and non-teleologists can agree that there is an epistemic condition on knowledge, C, such that one’s belief satisfies C if and only if one believes reliably. For instance, assuming that C is a normative condition (perhaps because all genuinely epistemic conditions are normative ones), what we are assuming is that teleologists and non-teleologists can agree that there is an epistemic norm, N, such that one’s belief complies with N if and only if it is reliable. What teleologists will hold is that the direction of analysis here proceeds from right to left. The normative condition is unpacked in terms of reliability, which, in turn, is understood teleologically as derivative from the epistemic goal. In contrast, non-teleologists will have to say that the direction of analysis is from left to right. The reliability condition is analysed in terms of a norm of belief and so in a non-teleological manner. (Note also that reliability can be analysed in the standard way in terms of its conduciveness to truth, so long as we avoid identifying truth with the epistemic goal.) While this is indeed one way of accommodating the reliability condition on knowledge whilst continuing to deny the relevant version of (5), I suspect that it will not turn out to be hugely popular. In fact, it is hard to deny that the teleological account of reliability is far more plausible than the non-teleological alternative. To see just one reason for this, consider the question of why we should think that there is an epistemic norm that corresponds to reliable belief. Teleologists can offer an attractive answer. Reliable means are generally good means to achieving ones aims. That’s why it makes sense to think that there is a reliability norm of belief. I just find it very difficult to see how a non-teleologist could come up with anything that comes even close to an equally plausible answer. Accordingly, there is reason to think that resisting the Anti-Justification argument in this manner is not the way forward.

   

41

the epistemic aim. Crucially, however, no one, David included, has ever provided a single argument that knowledge admits of reductive analysis in terms of justified true belief. All we have ever done on this front is to hope for the best.¹⁹ But, of course, in the absence of a viable argument that knowledge admits of reductive analysis in terms of justified true belief, the Anti-Justification Argument fails to make a compelling case against the claim that knowledge is the epistemic aim. As a result, the threat to Knowledge Aim that is posed by the Anti-Justification Argument can also be defused. Even if it establishes the conditional that if knowledge is the epistemic goal, then knowledge does not admit of reductive analysis in terms of epistemically justified belief, it simply provides no reason for detaching in the intended way. In fact, there is reason to think that the argument can now be turned on its head. Section 1 has developed a series of arguments that knowledge is the aim of inquiry. If, in addition, the epistemic domain is constituted by inquiry (as I will argue in Chapter 5), these arguments now provide excellent reason to believe that knowledge is the epistemic aim.²⁰ By the same token, there is reason to think that the way forward with the conditional that the Anti-Justification argument establishes is not by applying modus tollens, as David would have it, but rather modus ponens. If so, what the Anti-Justification Argument in conjunction with the arguments for Knowledge Aim really amounts to is an argument that knowledge does not admit of reductive analysis in terms of justified true belief.

3. Conclusion Inquiry is an activity with an aim. This chapter has focused on one specific form inquiry may take: inquiry into specific questions. I have made a case that attaining the aim of this form of inquiry requires ¹⁹ In fact, what has become clear in the wake of the Gettier literature is that we are holding out hope in the face of continued failure. ²⁰ Or, to be more precise, there is reason to believe that knowledge is the relevant epistemic aim, the one that justification is to be analysed as derivative from and the one for which true belief is a competitor.

42 , ,   knowledge. To this end, I have developed two arguments according to which Knowledge Aim is preferable to all of its weaker rivals in the literature and I have offered reason for thinking that these arguments generalize. I have also argued that the aim of inquiry into specific questions requires nothing more than knowledge. In other words, Knowledge Aim is true. I then looked into a number of prominent objections to Knowledge Aim and found them uncompelling. Finally, it is worth noting that one of these arguments effectively makes a case that if Knowledge Aim is true, knowledge does not admit of reductive analysis in terms of epistemically justified true belief. Of course, if knowledge does not admit of such reductive analysis, the question that immediately arises is what else we can say about the nature of knowledge. It is precisely this question that I will take up in Chapter 2.

2 Knowledge Chapter 1 took a closer look at the aim of inquiry into specific questions. In this chapter, I will turn to the question of what knowledge is. My aim is to use the result of my discussion in Chapter 1—in particular, the claim that knowledge is the aim of inquiry into specific questions—in order to shed light on this question.

1. A Network Analysis 1.1 Reductive and Network Analyses Let’s start by following Peter Strawson in distinguishing two models for philosophical analysis. Here is his characterization of the first model: The Dismantling Model [The Dismantling Model] represents it [i.e. philosophical analysis] as a kind of dismantling of a complex structure into simpler elements, a process which terminates only when you reach pieces which cannot be further dismantled. (Strawson 1992, 19)

Most importantly for present purposes, dismantling analyses allow us to understand complex phenomena by breaking them down into simpler elements which are independently understood. Crucially, for this to work as envisaged, dismantling analyses must satisfy the following condition: Priority The simpler elements in the analysans enjoy explanatory priority over the complex structure in the analysandum. Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0003

44 , ,   Any dismantling analysis that doesn’t satisfy Priority will be ill-suited to deliver an understanding of the complex phenomenon in terms of the simpler elements. A noteworthy consequence of Priority is that dismantling analyses must also be non-circular. After all, if a given analysis is circular, the simpler elements do not enjoy explanatory priority over the complex phenomenon. Here is a paradigm case of a dismantling analysis in epistemology: JTB One knows that p if and only if one justifiably and truly believes that p. JTB ventures to allow us to understand the nature of knowledge by dismantling knowledge into justification, truth, and belief, where all three enjoy explanatory priority over knowledge.¹ Unfortunately, now there is near universal consensus in the literature that JTB cannot be correct after all. There are cases in which one has a justified true belief that p but doesn’t know that p (Gettier 1963). The take-home lesson of these cases was widely believed to be that we need to fix JTB by identifying a further condition on knowledge, one that deals with the problematic cases. The trouble is that this has proven a surprisingly difficult task. For all known fixes, there are cases in which the proposed conditions for knowledge are present and knowledge is absent.² It is fair to say that the majority of epistemologists continue hold onto the idea that there exists a workable fix for JTB (henceforth also ‘the traditional analytic project’). However, as I already pointed out in Chapter 1, there is a non-negligible and rapidly growing number of dissenters who have abandoned this project. They think that we cannot define knowledge in terms of justified true belief and something else. As I already indicated, I take myself to be among these dissenters. One reason for turning my back on the traditional analytic project is the ¹ JTB might not count as a fully successful instance of the dismantling model of philosophical analysis as Strawson would have it. In fact, it won’t be, unless one is prepared to hold that justification, truth, and belief are not themselves in need of dismantling. ² For some overviews of the Gettier literature, see Lycan (2006), Slaght (1977), and Shope (1983).



45

modus ponens version of David’s argument from Chapter 1. Another is that we have better theoretical options. There are a range of attractive alternatives to the traditional analytic project on the market. This book develops the one I prefer in some detail. Once we abandon the traditional analytical project and its attempt to arrive at a workable analysis of knowledge within the dismantling model, the question that immediately arises is what exactly our alternatives are. Fortunately, the dismantling model is not the only game in town. In fact, it is once again Strawson who proposes an alternative. Here goes: The Network Model Let us imagine, instead, the model of an elaborate network, a system, of connected items, concepts, such that the function of each of item, each concept, could, from a philosophical point of view, be properly understood only by grasping its connections with the others, its place in the system. (Strawson 1992, 19)

One important difference between the network model and the dismantling model is that successful network analyses do not require Priority to be satisfied. In particular, elements in the network need not enjoy explanatory priority over the phenomenon to be analysed, or, at the very least, not all of them do. And, unsurprisingly, network analyses are also compatible with circularity. Once we have abandoned the dismantling model for our analysis of knowledge, we may still hope to offer an analysis that fits the network model. In what follows, I will develop just such a network analysis.

1.2 Activities with Constitutive Aims and Norms We do many things and pursue many aims. Often the relation between what we do and the aim we pursue is contingent. For instance, suppose you aim to get a new passport. To this end, you go to the embassy of your country, because you are an expat. country. In this case, going to the embassy is aimed at getting a new passport. The relation between

46 , ,   aim—getting a new passport—and activity—going to the embassy—is contingent here. Most importantly for present purposes, going to the embassy might not be aimed at getting a new passport. Suppose you work at the embassy. If so, you will more often than not go to the embassy without getting a new passport. What’s more, when you do so, going to the embassy is simply not aimed at getting a new passport. While the relation between activities and aims is often contingent, it can also be non-contingent. Most importantly for present purposes, there are activities that have constitutive aims as well as constitutive norms (henceforth ACANs for short). The reason I am interested in ACANs is that often enough ACANs lend themselves to network but not dismantling analysis, that inquiry is just such an ACAN, and that, as a result, knowledge does not admit of dismantling analysis. First things first—to get a better idea of what ACANs are, I will try to say a bit more about them. Let’s start with some examples. Target archery, trick-taking card games, snooker, and chess are examples of ACANs, among many others. The constitutive aim of target archery is to hit the target. Among the constitutive norms are norms specifying admissible means for attaining the aim, for instance that one must use a bow and arrow. The constitutive aim of whist, the famous trick-taking card game, is to score five points. Two of the constitutive norms are that it is played with a set of fifty-two cards and that it is played by four players. The constitutive aim of baseball is to score more runs that the other team. Two of the constitutive norms are that baseball is played by teams of nine and that one team bats while the other fields. The constitutive aim of chess is to checkmate one’s opponent. Among the constitutive norms are norms specifying the starting positions of the various types of pieces as well as a set of legal moves.³

³ Note that to say that an activity has a certain constitutive aim is not to say that an agent engaging in that activity has this aim too. For instance, you can play a game of chess and aim to lose. In this case, the aim of the game of chess, which is to checkmate your opponent, does not coincide with your aim, which is to be checkmated by your opponent. By the same token, to engage in an ACAN cannot require you to adopt the ACAN’s aim. What, then, does it take to engage in an ACAN? This is a hard question that I won’t be able to address here. For some proposed answers, see, e.g., Kelp and Simion (2018), Maitra (2011), and Williamson (2000).



47

With the examples in play, what does it mean to say that the norms and aims of a given ACAN are constitutive? The answer is that these norms and aims are essential to the ACAN. Anything that does not have these norms and aims will not be this ACAN. For instance, anything that does not have checkmating one’s opponent as its aim and is not governed by a norm that one must move the bishop diagonally will not be chess. Likewise, anything that does not have scoring more runs than the other team as its aim and is not governed by a norm requiring play with teams of nine will not be baseball, and so on. Before moving on, I’d like to look at a couple important properties of ACANs that set them apart from activities with contingent aims. First, every token of a given ACAN can be assessed in terms of these aims and norms. It is always prima facie appropriate to ask whether the ACAN’s constitutive aim was attained and whether its constitutive norms were respected. For instance, it is always prima facie appropriate to ask whether a certain chess player checkmated their opponent or whether they moved the bishop diagonally. Likewise, it is always prima facie appropriate to ask whether a certain baseball team scored more runs than the other team and whether they played with a team of nine players. Note that, in this way, ACANs differ from activities that have their aims contingently. It is not always prima facie appropriate to ask whether a certain aim was attained when this aim is contingent. For instance, suppose your spouse knows full well that you work at the embassy and that you just received a new passport. When you return home from a long day of work at the embassy, it will not be prima facie appropriate to ask whether you got a new passport. Second, when one doesn’t attain the constitutive aim of an ACAN one engages in, one must always remain unsuccessful in at least some way. If you take a shot in target archery that doesn’t hit the target, you must remain unsuccessful in some way. If you play a game of whist and do not score five points, you are bound to remain unsuccessful in some way. Again, ACANs differ from activities that have their aims contingently in this way. If one doesn’t attain an aim that an activity may have as a contingent matter of fact, one is not bound to remain unsuccessful in any way. For instance, if you return home from a

48 , ,   long day of work at the embassy, you may not remain unsuccessful in any way if you didn’t get a new passport.

1.3 ACANs, Dismantling, and Network Analysis What I want to argue is that there are ACANs that do not admit of dismantling analysis. In a nutshell, the reason for this is that any attempt at dismantling for these ACANs is spoiled by circularity. Even so, I will argue that a network analysis is still very much on the cards for these ACANs. In what follows, I will provide a few relevant examples. My aim here is to provide evidence not only that ACANs with this profile exist but also that they are not at all uncommon. In this way, this subsection lays important groundwork for the next subsection, which argues that inquiry is an ACAN with just this profile and develops a network analysis of knowledge in terms of its place in inquiry. Let’s move on to the examples then. My first case of an ACAN that lends itself to network but not dismantling analysis is chess. At first glance, this might seem surprising. After all, it might be thought that the prospects for a dismantling analysis of chess are quite bright, really. To see why, note first that we can offer necessary and sufficient conditions for attaining the constitutive aim of chess: one attains the constitutive aim of chess if and only if one checkmates one’s opponent. What’s even better is that we can also give necessary and sufficient conditions for the right-hand side of this biconditional: one checkmates one’s opponent if and only if one places one’s opponent’s king under inescapable threat of capture. Granted, it may initially appear as though the prospects for a successful dismantling analysis are looking up. On reflection, however, there is reason to think that this can’t be quite right. To see why not, note that the right-hand side of the last biconditional itself contains elements that are complex and so require dismantling. For instance, we might ask what a king is or what capturing amounts to. Of course, there are interesting things to be said here. For instance, to answer the question what a king is we might specify its starting point and the ways in which it moves. Crucially, however, the project of attempting to understand what a



49

king is without making reference to the game of chess is hopeless. For instance, we can’t understand what a king is without invoking the claim that it is a type of piece in chess. But, of course, chess is again not a simple phenomenon, and our dismantling analysis cannot stop here. We’ll need to offer an analysis of chess. The trouble is that there is no chance to offer a successful dismantling analysis of chess without including the claim that its constitutive aim is to checkmate the other player. But now we have gone around in a circle. Our analysis of checkmating forced us to make reference to the game of chess, and our analysis of chess forced us to make reference to checkmating. Priority is violated (because of a violation of non-circularity). What comes to light, then, is that although the prospects for a successful dismantling analysis of chess may initially have looked bright, closer scrutiny reveals that they are really quite dim, if not entirely hopeless. My second example is trick-taking card games. While there is no shortage of concrete examples to focus on, I’d like to take whist because it is particularly popular kind of trick-taking card game. Whist has the following constitutive aim: one wins a game of whist if and only if one scores five points. Champions of dismantling analyses might once again be heartened by the existence of necessary and sufficient conditions. That said, the right-hand side of our biconditional is once again complex and so requires further dismantling. Most poignantly, we might ask what point scoring amounts to. And, again, it might be thought that there is reason for optimism in that a dismantling analysis of point scoring can be given along the following lines: after all tricks have been played, the side which has won more tricks gets one point for each trick they have taken after the first six. Note that this still does not bring our dismantling project to an end. For instance, we may ask what exactly trick taking amounts to. In the spirit of a dismantling analysis, we may venture to offer an analysis of trick taking, for instance in terms of trick playing and a designated property. The trouble is that there is no way to proceed in our dismantling endeavour from here without making reference to the game of whist. To see this, note first that different trick-taking card games have different designated properties: in some games, it is the lowest-ranking card, while in others, including whist, it is the highestranking card that wins the trick. In order to proceed in our dismantling

50 , ,   project, we need to say what exactly it takes for a card to have the designated property. And this cannot be done without making reference to the game of whist.⁴ In this way, our project of offering a dismantling analysis of the constitutive aim of whist leads us back to the game of whist. Once again, we have gone around in a circle. Priority is once more violated. Again, the prospects for a dismantling analysis are dim. My last example is baseball. Recall that a team attains the constitutive aim of baseball if and only if it scores more runs than the other team. If we wanted to embark on a dismantling project, we will have to further analyse the right-hand side of this biconditional. For instance, we will have to say what it is to score a run. Here goes: a player scores a run if and only if they return safely to the home plate via the first, second, and third bases, touching the first base first, the second base second, the third base third, and the home plate last, before three outs are recorded and no obligations to reach bases safely on batted balls are violated. Now, there are plenty of questions to answer, including what a base is, what an out is, what the relevant obligations are, what it is to reach bases safely, and so on. The trouble is that it is again impossible to provide these answers without referring to the game of baseball. To take just one example, consider the question of what a base is. If you ask this question, the standard answer you will get is that bases are the most important elements in a baseball field, each one of which a player must touch to score a run in baseball. What comes to light is that it is not at all uncommon that ACANs do not admit of dismantling analysis because of a circularity issue. On the upside, there is little reason for thinking that we must now abandon the attempt to understand the activities, their norms, and aims altogether. After all, a network analysis is not committed to avoiding circularity and so might still be made to work. In fact, a network analysis seems particularly well suited here. To properly understand chess, whist, and

⁴ Perhaps this difficulty can be circumvented. It could be held, for instance, that the designated relation is always analysed in terms of the highest card. Even so, we are bound to run into the same kind of trouble once we try to understand what it takes for something to be the highest value card. The reason for this is that this turns on card ranking, which again differs from one game to the next. To understand what the highest card is we need to invoke the ranking rule of whist.



51

baseball we must understand their constitutive norms and aims. In the case of chess, this includes that the constitutive aim of chess is to checkmate the opponent, that a constitutive norm is that castling is a type of move in chess, and so on. In the case of whist, this includes that the constitutive aim is to score five points, that a constitutive norm is that the game is played with a pack of fifty-two cards, and so on. In the case of baseball, this includes the constitutive aim of scoring more runs than the other team, that it is played by two teams of nine, and so on. To properly understand these constitutive norms and aims, we must at the very least understand that they are constitutive of, respectively, chess, whist, and baseball. For instance, there is no way to properly understand what checkmating amounts to unless we understand that it has something to do with chess. Likewise, there is no way to properly understand what scoring more runs that the other team amounts to unless we understand that it has something to do with baseball, and so on. To properly understand chess, whist, baseball, and their constitutive aims and norms, we are thus bound to move around in a circle. But, again, this is entirely compatible with a network analysis. And, more importantly, the resulting analysis is no less illuminating for all that.

1.4 Inquiry, Knowledge, and Belief We have seen that there is a diverse range of ACANs which do not lend themselves to dismantling analysis. In all of these cases, the reason for this is that our dismantling endeavour ultimately leads us around in a circle. As a result, a key requirement for dismantling analyses is not met, to wit, Priority. Even so, I also argued that a network analysis is still very much on the cards. In what follows, I will develop a network analysis for knowledge. The key idea is that knowledge is defined in terms of its place in an ACAN that joins the ranks of ACANs which lend themselves to network but not dismantling analysis. What’s more, the reason why a dismantling analysis won’t work is that any dismantling endeavour is bound to lead us around in a circle, with the result that a key requirement on dismantling analyses, Priority, is not met.

52 , ,   Let’s look at the central ACAN then. Unsurprisingly, it is inquiry into whether p. Recall that the aim of inquiry into whether p could be characterized in a lightweight fashion as properly closing the question whether p in the affirmative/negative for oneself (alternatively, as settling that question). Crucially, inquiry is not only an activity with an aim but also an ACAN. The aim of settling questions is not only the aim of inquiry but also its constitutive aim. Any activity that does not have this aim is ipso facto not the activity of inquiry.⁵ What about the constitutive norms of inquiry? One prima facie plausible candidate is that in inquiry one must form or work toward positioning oneself to form beliefs via epistemic abilities.⁶ That’s why going to the local brainwashing service and having a certain belief installed will not count as inquiring into the corresponding question. More importantly for present purposes, I want to suggest that there are constitutive norms specifying kinds of move in inquiry, just as in chess there are constitutive norms specifying kinds of move in chess. In particular, one important kind of move is to close inquiry in the affirmative or negative. Now, the suggestion is that inquiry is an ACAN that joins the ranks of chess, whist, and baseball in the sense that it lends itself to network but not dismantling analysis. What does this network look like? Here is a rough idea: the network relates knowledge, belief, and inquiry in the sense that knowledge is the constitutive aim of inquiry and belief is the result of the kind of move in inquiry that closes it in the affirmative or negative.⁷

⁵ It’s worth keeping in mind here that activity aim and agent aim can come apart (see n.3). Accordingly, it is entirely compatible with the claim that inquiry constitutively aims at questionsettling that a given inquirer does not have question-settling as her aim. ⁶ I will return to epistemic abilities in Chapter 3. For detailed discussion, see Kelp (2016, 2017, 2018). While I think this is roughly right, alternatives are conceivable. Evidentialists, for instance, might say that inquiry essentially involves gathering evidence (and that’s why making use of the brainwashing service doesn’t qualify as inquiring). ⁷ While I will describe enough of the network for my purpose of illuminating the nature of knowledge in terms of its place in it, I will not describe it in its entirety. Most importantly, suspension of judgement will have a place in the network (as the result of another type of move). However, since suspension of judgement is of little relevance for understanding the nature of knowledge, I will set it (as well as any other phenomena that may belong in the network) aside here.



53

With the rough idea in play, let’s look at some of the details. Recall that I said that inquiry into whether p aims at settling the question whether p; or, alternatively, it aims at properly closing the question whether p for oneself in the affirmative/negative. The way in which knowledge enters the picture, I suggest, is that it specifies what it takes to attain the constitutive aim of inquiry. In particular, I want to suggest the following necessary and sufficient conditions for the question-settling aim of inquiry: Question-Settling Inquiry into the question whether p is properly closed for one in the affirmative/negative if and only if one knows that p/not-p.

I said that belief is the result of a kind of move in inquiry and, more specifically, that it is the result of the type of move that one makes when one closes inquiry into whether p in the affirmative or negative. More specifically, I want to suggest that belief is the state in which inquiry is closed for one either in the affirmative or in the negative: Belief One believes that p/not-p if and only if inquiry into whether p is closed for one in the affirmative/negative.

With these points in play, the network is up and running. Note that, just as in the chess case earlier, we find necessary and sufficient conditions for some of the relevant phenomena. And, again, this raises the question whether we might not hope for relevant dismantling analyses after all. Crucially, just as in the cases of chess, whist, and baseball, the answer is no. To see this, note that the right-hand side of Question-Settling is itself complex and would therefore require further dismantling. Moreover, we can say interesting things about it too, including that knowledge requires some epistemic condition(s) (e.g. justification or reliability) and belief. However, first, we have already seen in the discussion of the AntiJustification Argument from Chapter 1 that if (i) knowledge is the aim of inquiry and (ii) there is an epistemic condition that is analysed teleologically, then the epistemic condition will have to be analysed in terms

54 , ,   of knowledge.⁸ Since there is excellent reason to think that both (i) and (ii) hold, the prospects for a dismantling analysis of knowledge are dim. After all, a dismantling analysis will lead us from knowledge to the teleological epistemic condition on knowledge and from there back to knowledge. As a result, the analysis is going to be circular and so bound to violate Priority. Moreover, second, by Belief, the necessary and sufficient conditions for belief feature inquiry. But, of course, inquiry is again not a simple phenomenon, and our dismantling analysis cannot stop here. We’ll need to offer an analysis of inquiry. The trouble is that there is no chance of offering a successful dismantling analysis of inquiry without including the claim that its constitutive aim is to settle the question at hand. As a result, we are once again tied up in a circle. Our analysis is thus bound to violate Priority on yet another count, and so the prospects for a successful dismantling analysis are dim. Of course, what this suggests in the first instance is that we should not expect a dismantling analysis of what it is to settle a question. Note, however, that the reason why settling questions doesn’t admit of dismantling analysis will also serve to show that we should not expect a dismantling analysis of knowledge either. This is very obvious in the case of the epistemic condition. If knowledge features an epistemic condition and the epistemic condition is to be analysed in terms of knowledge, there will be no successful dismantling of knowledge in terms of the epistemic condition. Regarding the belief condition, any further dismantling of belief will feature inquiry, and any further dismantling of inquiry will lead us back to knowledge. The attempt at dismantling knowledge will at this point be circular and so violate Priority as well. Just as in the cases of chess, whist, and baseball, then, there is reason to think that a dismantling analysis is not forthcoming. Yet, again, there is an upside. There is little reason for thinking that we must now abandon the attempt to understand the nature of knowledge altogether. After all, a network analysis is not committed to Priority and so might still be made to work. In fact, a network analysis seems particularly well suited here. To properly understand inquiry, we must ⁸ See also Chapter 3 for more on justification.



55

understand its constitutive norms and aims, including that the constitutive aim of inquiry is to settle the question at hand, a constitutive norm is that closing questions in the affirmative/negative is a kind of move in inquiry, etc. To properly understand these constitutive norms and aims, we must at the very least understand that they are constitutive of inquiry. For instance, there is no way to properly understand what settling a question amounts to unless we understand that it has something to do with inquiry. To properly understand the activity of inquiry and its constitutive aims and norms, we are thus bound to move around in a circle here. But, again, this is entirely compatible with a network analysis. And, more importantly, the resulting analysis is no less illuminating for all that. At this stage, the network analysis should have come into view with sufficient clarity. Even so, I have said fairly little about why one should buy it in the first place. That’s the task I’ll take up next.

1.5 Evidence for the Network Analysis For the vast majority of work on the nature of knowledge in the literature, the part on why we should accept it focuses mainly (if not entirely) on Gettier cases. This is no surprise, given that the project of fixing JTB is effectively the project to coming to grips with Gettier cases. For those who expect me to explain how my account handles various Gettier cases, I have bad news: you’ll have to wait until Chapter 3. In fact, I will not look into any predictions about cases here. What I want to focus on instead is a number of theoretical benefits that the network analysis brings in its train, both for the theory of knowledge and for the theory of belief. Here goes. First, the network analysis allows us to understand the nature of knowledge even once we agree that a dismantling analysis is not forthcoming. This means that we will be able to reap the benefits of a nonreductive epistemology like the one this book develops whilst not being at a complete loss when it comes to the important task of understanding the nature of knowledge.

56 , ,   Second, the network analysis places the ACAN of inquiry and hence the phenomena of inquiry, knowledge, and belief which it connects into a broader class of ACANs (alongside chess, whist, and baseball) that does not admit of dismantling analysis because of a circularity issue. In this way, the proposal promises to deliver an important diagnostic insight as to why we shouldn’t expect knowledge to admit of dismantling analysis in the first place. Third, the following highly attractive thesis about belief drops right out of the network analysis, to wit, that belief is constitutively governed by a knowledge norm.⁹ Here is how. Recall first that, by QuestionSettling and Belief, knowledge that p (not-p) is proper belief that p (not-p). As a result, belief is governed by a knowledge norm. And since, on the present network analysis, the relations between inquiry, knowledge, and belief hold as a matter of constitutive fact, this knowledge norm is a constitutive norm of belief.¹⁰ Fourth, the network analysis allows us to avoid some important downsides that the thesis that belief is constitutively governed by a knowledge norm might be thought to have. For instance, it might be thought that it forecloses a dismantling analysis of knowledge in terms of belief. While this is, of course, true, it is no longer worrying. After all, the network analysis also explains why we shouldn’t expect a dismantling analysis of knowledge to be viable in the first place. It might also be thought that the thesis about belief commits us to the implausible claim that belief can be fully analysed in terms of knowledge or, at the very least, that knowledge enjoys explanatory priority over belief. Again, given the network analysis, this is incorrect. This is because both are analysed by their place in the network. Since the network places them on an explanatory par, knowledge does not enjoy priority over belief on this front. Since the network also features inquiry, belief cannot be fully analysed in terms of knowledge. ⁹ While I do take this to be a particularly neat consequence of the account, I won’t stop to defend it here. Fortunately, this won’t be necessary either since what I take to be a compelling case for this claim has already been made in the literature (e.g. Adler 2002, Huemer 2007, McHugh 2011, Simion, Kelp, and Ghijsen 2016, Williamson 2000). ¹⁰ See Simion, Kelp, and Ghijsen (2016) for a more detailed account of what it means to say that belief is governed by a knowledge norm (in terms of Geach-style attributive goodness Geach 1956).



57

2. Objections and Replies 2.1 That’s Not Inquiry Suppose you walk down the street and enter a red house. In this case, you may form a belief that the house you entered is red and you may come to know this. At the same time, it may well be that you form your belief automatically. In that case, however, you form your belief without ever inquiring. But how can this be if beliefs are supposed to be the result of moves in inquiry? I’d like to sketch two responses to this worry. The first maintains that cases of automatic belief formation are cases of inquiry. To see why this might be plausible, suppose that, a little later, someone asks the question whether the house you have entered is red. Given that you still believe that it is red, you return a positive answer straight away, i.e. without inquiring further (Case 1). Now contrast Case 1 with a case in which you never formed a belief about the colour of the house in the first place. When asked whether the house you have entered is red, you will not return a positive answer straight away. Rather, you inquire further (Case 2). Why is it that, in Case 1 you don’t inquire further, while in Case 2 you do? An attractive answer to this question is that in Case 1, but not in Case 2, you made a move in inquiry. But since making a move in inquiry requires inquiry, there is reason to believe that you did inquire in our toy case of automatic belief formation, i.e. Case 1. Second, even those who still want to resist counting cases of automatic belief formation as cases of inquiry will be hard pressed to deny that the problematic cases of automatic belief formation and cases of what they would count as genuine inquiry are species of a broader genus of activity which we may call ‘finding things out’.¹¹ In the above case, for instance, ¹¹ Couldn’t we just say that the activity is belief formation or acquiring knowledge? No. To think of the activity as acquiring knowledge trivializes Knowledge Aim. After all, True Belief Aim and Justified Belief Aim are no longer serious competitors. To see that we can’t construe the activity as belief formation, note that not every belief-forming process has settling questions as its aim. Some are aimed at making us feel better about ourselves, some at producing false beliefs that increase the probability of survival, etc. Since ACANs have their constitutive aims essentially, it is hard to see how belief formation could still be an ACAN that has the constitutive aim of settling questions.

58 , ,   even if you don’t count it as inquiring, you are still engaged in the activity of finding out whether the colour of the house is red. Crucially, inquiry and finding things out share their constitutive aim and a range of key constitutive norms. Just like inquiry, finding things out has properly closing questions as a constitutive aim. If what you are doing does not aim at properly closing questions, then what you are doing is not finding things out. The two activities have the same constitutive aim. What’s more, finding things out and inquiry share key constitutive norms. Recall that an important constitutive norm of inquiry is that in inquiry one must form or work toward positioning oneself to form beliefs via epistemic abilities.¹² That’s why going to the local brainwashing service and having a certain belief installed will not count as inquiring into the corresponding question. Note that the same goes, once again, for finding things out. This is evidenced by the fact that being brainwashed into a certain belief counts no more as engaging in the activity of finding things out than it counts as inquiring. Another way of responding to the above worry, then, is to grant that the problematic cases of automatic belief formation do not count as cases of inquiry but to insist that they do still count as cases of a broader type of activity. This type of activity, finding things out, can accommodate the problematic cases of automatic belief formation. At the same time, finding things out shares the constitutive aim and key constitutive norms with inquiry. As a result, just like inquiry, the broader type of activity is an ACAN which has knowledge as its constitutive aim and closing questions in the affirmative/negative as a central kind of move which results in belief. This will allow me to get the network analysis off the ground, even if it turns out that I was wrong to think that it is inquiry. While it is possible to respond to the worry by retreating to a different ACAN such as ‘finding things out’, my own preference is for the broad account of inquiry, which allows that cases of automatic belief formation are cases of inquiry. Unsurprisingly, then, this is the view I will continue to be working with in what follows. ¹² I will return to epistemic abilities in Chapter 3. For detailed discussion, see Kelp (2016, 2017, 2018).



59

2.2 Belief Doesn’t Close Inquiry According to the proposed view, belief closes inquiry. Given that this is so, doesn’t it follow that once one has formed a belief, one must stop inquiring, and if one continues to inquire, one cannot believe? And isn’t that implausible? Isn’t it entirely possible to inquire into what one already believes? While a positive answer to the last two of these questions may look plausible at first glance, things are not entirely straightforward on reflection. This is because when one has a belief and continues to inquire, one will often inquire into questions that are subtly different from the one about which one holds a belief already. Suppose, for instance, that you have a belief that a certain defendant is guilty and continue to inquire. How plausible is it that you are still trying to find out whether the defendant is guilty? Not very plausible. After all, you already believe that they did it. In keeping with that, there are questions in the vicinity that you may still pursue, for instance, whether there is further evidence that they are guilty. Accordingly, one response to the above worry is that what’s really going on when one continues to inquire when one already has a belief is that one inquires into a (perhaps subtly) different question from the one about which one holds a belief already (Friedman 2019). Still, there is a lingering worry that this response isn’t fully satisfactory at the end of the day. To see why not, suppose you have formed a belief that p and are offered a handsome sum for continuing to inquire into whether p. Isn’t that something you could still do to collect the money? But, of course, if you can, it can’t be that belief closes inquiry. Let’s suppose this is right, then. If so, the claim that belief closes inquiry needs to be qualified. To see how, compare: you are playing a game of chess. You have captured all your opponent’s pieces except for the king. At the same time, you still have two bishops and a king left. Since your opponent knows you to be a skilled player, they resign and you win the game. Now suppose that someone offers you a handsome sum to continue to play until the bitter end. While you do continue to play after your opponent has resigned in this case, there is a clear sense in which the game is over once your opponent has resigned. Here is one way of saying more about this sense: the aim of winning the game can no

60 , ,   longer generate reason for you to continue playing. Rather, any reason to continue playing will now need to be game-external (e.g. here, to get the money). Unsurprisingly, I want to suggest that the same goes, mutatis mutandis, for inquiry and belief. Even if we want to allow that you can continue to inquire into whether p after you have formed a belief that p/not-p, there is a clear sense in which inquiry into whether p is closed for you once you form a belief that p, say. One way of saying more about this sense is that the aim of settling the question whether p can no longer generate reason for you to continue inquiring into whether p: to do so, you will now need an external reason, such as the prospects of a reward or an interest in a (perhaps subtly) different question. Note that these two responses will also work for seemingly more difficult cases. Suppose that you believe that the hob is off. Now you are seeking to confirm that it is. Isn’t this a case in which you believe that the hob is off and yet continue to inquire into the question whether it is off?¹³ No, for the reason just given. Once you have a belief that the hob is off, the aim of settling the corresponding question can no longer give you a reason to inquire into this question. What you need is an external reason such as the reward of the additional confidence (perhaps given the stakes). What’s more, it’s plausible that what you are really inquiring into is a subtly different question, e.g. whether you can confirm that the hob is off or whether you can be sure that the hob is off.

2.3 Knowledge Is Not the Constitutive Aim of Inquiry All right, one might say, inquiry into whether p is an ACAN and it constitutively aims at settling the question whether p. But even so, it’s not very plausible that inquiry constitutively aims at knowledge. To see why not, note that there could be communities of inquiring agents whose belief-forming processes are insufficiently reliable to generate knowledge in them. Members of these communities surely engage in inquiry. And even if their inquiries may be thought to aim at settling questions, they ¹³ Thanks to an anonymous referee for pressing me on this.



61

cannot plausibly be thought to aim at knowledge. Rather, they must be aiming at something weaker than knowledge. Similarly, there could be communities of inquiring agents who are much more reliable than we are. Members of these communities also surely engage in inquiry. And even if their inquiries may be thought to aim at settling questions, they cannot plausibly be thought to aim at knowledge either. Rather, they must be aiming at something stronger than knowledge. If this is right, inquiry aims only contingently at knowledge. But then it can’t be that knowledge is the constitutive aim of inquiry and the above account of the nature of knowledge collapses.¹⁴ As a first observation, note that this objection assumes that, in the possible communities under consideration, it is the aim of inquiry that changes, while a number of other things remain fixed, including that the agents engage in inquiry and what is required for knowledge. Accordingly, one way of responding to the objection is to argue that it is really something else that’s changing. In fact, I think that there is more than one option that can be made to work here. In what follows, I’ll look at three of them in turn. One way of responding to the objection is by holding that the requirements for knowledge are different across the different communities. More specifically, in the community of agents that are less reliable than we are, knowledge requires a lower degree of reliability, whereas, in the more reliable community, it requires a higher degree. The result is that, in all three communities, agents engage in inquiry and their inquiries aim at knowledge. It might be thought that this response smacks of ad hocness. After all, why should we ever think that the requirements of knowledge vary in just this way? Fortunately, there is a promising answer even to this question. There is a family of views in epistemology, perhaps most prominently associated with Edward Craig’s influential work (Craig 1990), according to which the content of the concept of knowledge is ¹⁴ Thanks to Mona Simion for bringing this to my attention. Essentially the same worry could be raised with inquiry in different domains. For instance, it might be thought that the aims of scientific and legal inquiry are more stringent that the aim of ordinary inquiry. It is easy to see that the responses I offer below work just as well here. Thanks to Miguel Egler for raising this issue.

62 , ,   fixed by its function. Now, Craig takes it that the function of the concept of knowledge is to flag good informants. However, there is an alternative proposal, which I (e.g. Kelp 2011) and others (e.g. Kappel 2010, Rysiew 2012) have defended elsewhere, according to which the function of the concept of knowledge is to flag when a question is properly closed for one. What this proposal effectively amounts to is that the function of the concept of knowledge is to flag when the aim of inquiry has been reached. Given that, according to this view, the function of the concept determines its content, if the requirements for reaching the aim of inquiry vary, the requirements for knowledge will vary as well. On this view, then, we’d expect the requirements for knowledge to vary between the three communities in just the way the response to the above objection would have it. Here is another way of responding to the objection. The key idea here is to explain the objection in terms of the distinction between activity aims and agent aims. To see how it works, let’s return to chess: suppose a grandmaster is playing a schoolchild. No time control rules are operative. At the same time, it is a longstanding grandmaster ethos that schoolchildren must be defeated within forty-five seconds; otherwise playing them is just too boring. On the other hand, it is schoolchildren’s ethos to last for at least one minute when playing grandmasters. This is not to say that winning wouldn’t be nice, but rather to have some aim that is more realistically attainable. Finally, suppose our grandmaster checkmates the schoolchild as expected but needs ninety seconds to do so. While she has attained success in chess, she has failed to honour grandmaster ethos. On the other hand, our schoolchild has not attained success in chess but has done well when it comes to schoolchildren’s ethos. Note that what is happening here is that there are certain normative practices built around the game of chess; or, to be more precise, around a certain type of match. Crucially, these practices are extrinsic to the game in that they do not affect its constitutive norms and aims. Note also that it makes sense to have the extrinsic normative practices. After all, it makes sense to have aims that are both challenging and attainable. Unsurprisingly, the key idea of the second response is that the same is going on, mutatis mutandis, in the different communities. Inquiry constitutively aims at knowledge. And the agents in all three communities



63

engage in inquiry. However, agents in the less and more reliable communities also have some additional normative practices built around the activity of inquiry. In the case of the less reliable community, positive normative status is assigned to certain failures in inquiry. Since it makes sense to have aims that are both challenging and attainable and since they cannot attain the aim of inquiry, this makes sense. In the case of the more reliable community, negative normative status is assigned to some successes in inquiry. Again, since it makes sense to have aims that are both challenging and attainable and since attaining the aim of inquiry is not challenging at all, this makes sense also. The last response I will consider here takes seriously the idea that the aim of the activity of inquiry changes. To see how it works, let’s return to chess once more. Note first that ‘chess’ does not pick out a single game. Rather it picks out a family of games.¹⁵ For some of these family members we have names. Blitz chess is one example, Chess960 another. For many others, we don’t. Some of them we do play; others we don’t (though we might have played at least some of them). Crucially, each family member has its own constitutive aims and norms, which distinguish it from the others. For instance, one key rule of Blitz chess which distinguishes it from more familiar members of the chess family is a time control rule. Roughly, one has only ten minutes to win the game. The time control rule also affects the aim of Blitz chess, which is not only to checkmate the opponent but to do so within the allocated time. Note that the specific time control rule also distinguishes Blitz chess from other members of the chess family with different time control rules, such as Bullet chess or Lightning chess. Finally, note that whichever member of the game in the chess family one is playing, one can correctly be said to be playing chess. With these points in play, let’s return to inquiry. The nub of the present response is, again, that the same is going on, mutatis mutandis, in the case of inquiry. ‘Inquiry’ does not pick out a single activity but a family of activities.¹⁶ Each family member has its own constitutive aims and norms, which distinguish it from the others. According to this ¹⁵ How this range is fixed is a tricky question that I do not attempt to settle here. ¹⁶ Again, I won’t offer a detailed account of how the range is fixed.

64 , ,   response, then, communities of agents whose inquiries do not constitutively aim at knowledge engage in different activities from those in which we engage. For instance, the community of more reliable agents engages in an activity that has a stronger constitutive aim than knowledge. In fact, that’s just what distinguishes this activity from the one we engage in. At the same time, all of these activities are members of the family of inquiry. As a result, agents in all communities can correctly be said to be inquiring.¹⁷

2.4 Activities Don’t Have Aims One very fundamental objection one might have is that activities simply aren’t the kinds of thing that can have aims. Only agents have aims; activities don’t. Or, at the very least, talk of aims is at best metaphorical when it comes to activities. Given how central the idea that activities have aims is for the present view, doesn’t that undermine the entire project?¹⁸ One might wonder just how damaging this objection is. In particular, one might wonder whether we can’t analyse talk of aims of activities in terms of talk of aims of agents? A straightforward proposal would be that to say that an activity has an aim means that there is an agent who has adopted this aim and intends to bring it about by engaging in the activity. For instance, to say that the aim of chess is to checkmate one’s opponent means that there is some player who has adopted the aim of checkmating their opponent and intends to bring it about by playing chess. Unfortunately, there is excellent reason to think that this won’t work. Consider chess. It is possible to play chess without adopting the aim of

¹⁷ Note that this response leaves open the possibility that true belief is the aim of inquiry of some members of the family of inquiry, or at least minimally reliable true belief. (Thanks to an anonymous referee for pointing this out.) One advantage of this consequence is that it serves to make even clearer how the present view differs from knowledge first epistemological alternatives. After all, it allows that true belief is the constitutive aim of at least some members of the family of inquiry. At the same time, it’s important to be clear that this is not much of a concession to champions of True Belief Aim. After all, even if there are such members of the family of inquiry, they are not members we engage in, nor might we easily have done so. ¹⁸ Thanks to an anonymous referee for pressing me on this.



65

checkmating one’s opponent. Suppose you are playing a vindictive dictator who will have your head if you beat them. It only makes sense for you to play to lose. In fact, it could even be that both players in a game of chess are playing to lose. However, that does not change the fact that the aim of chess—and indeed the aim of the game of chess that you are playing—is to checkmate the opponent. The straightforward proposal fails. In fact, once we are clear that the straightforward proposal fails, we may start wondering whether talk of aims of activities really is metaphorical after all. To see why, note that, at least for token activities, that a certain activity has a certain aim explains why agents adopt this aim, rather than the other way around. For instance, when you are playing a game of chess, what explains why you adopt the aim of checkmating your opponent is that checkmating your opponent is the aim of chess, rather than vice versa. But this kind of explanatory work does not really sit well with the idea that aim talk is at best metaphorical when it comes to activities. Still, a residual uneasiness about aim talk for activities may remain. Perhaps the following proposal may resolve this uneasiness. We can analyse constitutive aims of activities in terms of constitutive norms. More specifically, the thought is that a constitutive aim of an activity boils down to a constitutive norm that specifies a success condition for the activity under consideration. For instance, on this proposal, to say that the constitutive aim of chess is to checkmate the opponent is to say that there is a constitutive norm of chess according to which one wins a game of chess if and only if one checkmates one’s opponent.¹⁹ If this proposal is correct, we can in principle do away with talk of aims of activities and replace it with talk of norms and success conditions.²⁰

¹⁹ To keep things simple, I will bracket the phenomena of forfeiting and resigning here. ²⁰ One may wonder just how much is gained by replacing aim talk with success talk. Shouldn’t anyone who was uneasy with the thought that activities have aims also be uneasy with the thought that activities have success conditions? Might we not say that just as aim talk only applies to agents, the same goes for success talk? No, in the case of success talk it is clear that it applies beyond agents. Consider my digestive system. We may be uneasy about saying that my digestive system aims to digest my breakfast. But it would be odd to be uneasy about saying that my digestive system successfully digested my breakfast.

66 , ,  

2.5 Dismantling in Ramsey-Lewis Style? Another worry one might press against my view is whether it isn’t best understood as a kind of dismantling analysis after all. In particular, the idea is that the best way of interpreting what the account amounts to is as a functional analysis à la Ramsey (1990) and Lewis (1970). Once we have delineated the network, we go on to replace the key terms we are interested in—in the present case, knowledge, belief, and inquiry—by variables and then quantify over these variables. What we end up with is a so-called Ramsey-sentence, which offers a reductive analysis of knowledge, belief, and inquiry all at the same time, as it were. And isn’t that tantamount to a form of dismantling of these phenomena?²¹ Perhaps. But if it is, I want to suggest that what the worry highlights is that Strawson’s way of putting things simply isn’t maximally perspicuous. In particular, I think that there are actually two important distinctions between philosophical analyses that Strawson’s distinction between the dismantling and network model by itself doesn’t take into account, at least not explicitly. The first is between atomistic and holistic analyses. Atomistic analyses aim to illuminate complex phenomena by breaking them down into simpler ones. Holistic analyses, in contrast, aim to shed light on phenomena by tracing relations to other phenomena. The second distinction is between reductive and non-reductive analyses. On the standard view, reductive analyses allow us to transform any statement about the thing to be reduced into a statement about the things it is reduced to (Carnap 1967). For instance, JTB qualifies as a reductive analysis in this sense. It allows us to transform any statement about knowledge into a statement about justified true belief. Of course, to count as genuinely reductive, the analysis must satisfy Priority and thus be noncircular. That’s why JTB qualifies as reductive, whereas an analysis of knowledge according to which one knows that p if and only if one knows What’s more success talk also does explanatory work when it comes to activities. Suppose I had the aim of running a marathon by running 10 miles. My aim is unattainable. What explains this unattainability is the fact that running a marathon has a success condition that cannot be attained by running 10 miles. Here success conditions that apply to activities do crucial explanatory work. ²¹ Thanks to Timothy Williamson for pressing this worry on me.



67

that p or 2 + 2 = 4 doesn’t. In contrast, and most importantly for present purposes, non-reductive analyses eschew the commitment to noncircularity. Here is one way to (partially) map these two distinctions onto Strawson’s distinction: any analysis that is reductive is a dismantling analysis. If that’s right, then, of course, Ramsey-Lewis functional analyses will come out as dismantling. After all, they are reductive in nature. And if the account I proposed is indeed best understood as a Ramsey-Lewis functional analysis, it falls in the dismantling camp also. Now, I am not sure that this mapping best captures what Strawson had in mind. That said, my purpose here is neither to push a particular reading of Strawson nor to force a certain label for my account. I do believe, however, that Strawson was onto something that’s useful for my (epistemological) purposes. And I also think that once we keep the two distinctions apart, we can get a clear view of what it is. JTB and the bulk of the attempts to fix it that have been proposed in the wake of Gettier are analyses that are both atomistic and reductive. Once we abandon this type of approach and ask what, if anything, we can still say about the nature of knowledge, the key move I am proposing is to abandon atomism in favour of holism. Rather than trying to understand the nature of knowledge by breaking it down into simpler elements, the proposal is to shed light on it by tracing relations to other phenomena. That’s how the account aims to say something illuminating about the nature of knowledge after having parted ways with the traditional analytical project. And this key move is entirely unaffected by the above worry, even if it sticks and the account comes out as dismantling. Suppose, then, that I am right not only in that the correct analysis of knowledge is holistic rather than atomistic but also in that it takes the shape I suggest. Even so, one point that the above worry forcefully makes is that holistic analyses need not be non-reductive but can still be reductive. As a result, even if the worry can be allayed, what we are left with is a question that is of interest in its own right, to wit, whether the account is best understood as a reductive or a non-reductive account. It might be thought that the answer to this question is straightforward: go reductive. After all, isn’t that clearly the preferable option? I don’t mean to deny that it would be neat to have a reductive analysis of

68 , ,   knowledge. At the same time, I think that there is some reason for caution. To begin with, Ramsey-Lewis functional analyses are always threatened by the so-called permutation problem, i.e. that the relevant Ramsey-sentence is realized in ever so many different ways (e.g. Smith 1994). Note that the present account is particularly likely to succumb to the permutation problem, at least in its present shape. The reason for this is that the number of statements at issue in the account to be ‘ramsified’ is so very small. As a result, the chance of there being a large number of things that realize it seems particularly high. Here is another problem with opting for the Ramsey-Lewis route. Recall that what we are trying to do is understand the nature of knowledge. The trouble is that the Ramsey-Lewis route isn’t particularly well suited for this purpose. To see why not, consider the following: you are at the luggage carousel at the local airport. Someone took your suitcase. You ask who. ‘The person with the pink and green underwear’ is not a particularly illuminating answer to this question (even if true), at least not unless you have some idea of who the person with the pink and green underwear is. Likewise, ‘Knowledge is the denotation of x in there is an x, there is a y, there is a z . . . such that . . . ’ is not a particularly illuminating answer to the question as to what knowledge is, at least not unless you have some idea of what y, z . . . are. But, of course, to answer the questions as to what y is, you need to have some idea of what x, z . . . are, to answer the question of what z is, you need to have some idea of what x, y . . . are, and so on. Not only are we moving around in a circle, but we also appear to be doing so blindly. In this way, a second important downside of the Ramsey-Lewis route is that it just doesn’t deliver the kind of illumination that would allow us to make progress on understanding the nature of knowledge.

3. The Competition I have argued that certain ACANs afford network analyses in that they cannot be properly understood without understanding their constitutive norms and aims, and the constitutive norms and aims cannot be properly understood without understanding the activity they are constitutive



69

norms and aims of. I have also made a case that inquiry is an ACAN. Since knowledge is the aim of inquiry, this means that we should not hope for a dismantling analysis of knowledge but opt for a network analysis instead. In keeping with this, it may be possible to specify necessary and sufficient conditions for knowledge. Crucially, we must not expect these to be noncircular. What we get, then, is a non-reductive account of knowledge. Of course, the above is not the only non-reductive account of knowledge on the market. Arguably the most prominent alternative is due to Timothy Williamson. He claims that knowledge is a mental state, or, to be more precise, that it is the most general factive mental state, and that this sheds some light on the nature of knowledge (Williamson 2000; cf. Nagel 2013, Miracchi 2015). At this stage, one might wonder which of the two accounts is preferable. As a first observation, note that the two accounts are in principle compatible. It could be that knowledge is the constitutive aim of inquiry (henceforth the inquiry view) and that it is the most general factive mental state (henceforth the mental state view). Perhaps, like quantic entities which have both a particle and a wave nature, knowledge has both a normative and a psychological nature. That said, anyone who accepts at least one of the two views will have to make a choice between the following three incompatible views: the first is to accept both the inquiry view and the mental state view; the second is to accept only the mental state view; and the third is to accept only the inquiry view. In what follows I will provide reason to believe that the last option is the one to go for. In a nutshell, this is because the mental state view cannot be correct because, pace Williamson, knowledge simply is not a mental state. Note also that, as a result, Williamson is genuine competition. After all, he endorses the mental state view and so is firmly committed to either the first or the second option.

3.1 Preliminaries Let’s first take a look at what exactly Williamson means when he says that knowledge is a mental state. It might be thought that this thesis is not particularly controversial. After all, it is widely agreed that

70 , ,   knowledge is a species of belief and that belief is a type of mental state. If so, of course, knowledge is a mental state as well. Obviously, however, this can’t be what Williamson has in mind if he wants this thesis to shed light on the nature of knowledge. And, indeed, it isn’t. What is widely accepted is that being in some mental state is necessary for knowing that p. In addition, Williamson claims that there is a mental state such that being in it is sufficient for knowing that p. In other words: Mental State There is a mental state such that being in it is necessary and sufficient for knowing that p.

This puts knowledge on a par with other paradigmatic mental states such as belief, desire, hope, fear, and intention. It is also what gets Williamson in the ballpark of a non-reductive account of knowledge. And, of course, it is also much more controversial. Before I get down to the nitty-gritty, I’d like to briefly say a few words about my general strategy. In order to mount a case against Williamson, I will offer an alternative to the thesis that knowledge is a mental state, which I’ll argue is better than what Williamson has to offer. In a nutshell, here is my alternative: knowledge is an epistemic state only. And here is a more careful statement of what it amounts to. One key thesis is that knowledge is an epistemic state. The thought is that knowledge is an epistemic state in the same sense in which, according to Williamson, knowledge is a mental state: Epistemic State There is an epistemic state such that being in it is both necessary and sufficient for knowing that p.

This puts knowledge on a par with other paradigm epistemic states such as justified belief and irrational belief. I take it that Williamson does not deny Epistemic State. Rather, what I’d expect him to maintain is that both Mental State and Epistemic State are true—that is, that knowledge is both an epistemic state and a mental state. This is where the second key



71

thesis of my alternative comes in: we will do well to refrain from adopting the thesis that knowledge is a mental state alongside the thesis that it is an epistemic state. That’s what I mean when I say that knowledge is an epistemic state only. With these preliminaries in play, let’s move on to the arguments.

3.2 Causal Explanations of Action According to one prominent objection to Mental State, knowledge cannot be a mental state because knowledge isn’t apt to feature in causal explanations of actions in the right way. The nub of the argument is that causation is local in the sense that causal efficacy supervenes on narrow features of the cause (here: the agent). Since knowledge is factive, it isn’t a narrow feature of the agent and isn’t apt to feature in causal explanations of actions. But any bona fide mental state must be apt to so feature. Hence, knowledge cannot be a bona fide mental state. By way of response to this objection Williamson argues that there is reason to believe that causal explanations of actions need not be restricted to narrow features of agents. On the contrary, jettisoning causal explanations of actions in terms of non-narrow (or broad) features of agents may come at a significant explanatory loss. The crucial point is that the probability of continued action, given knowledge, is higher than the probability of continued action, given true belief. The reason for this, in turn, is that knowledge is more robust (in the sense that it is less likely to be lost) than true belief. Since the probability of continued action, given continued belief, is much higher than the probability of continued action, given lack of belief, the probability of continued action, given knowledge, is higher than the probability of continued action, given true belief. As a result, continued action is better explained in terms of knowledge than in terms of true belief. In order to illustrate this point, Williamson considers the case of a burglar who is ransacking a certain house all night in search of a diamond. He considers two explanations of this action, one in terms of the burglar’s knowledge that the diamond is in the house, the other in terms of the burglar’s true belief that it is there. The burglar’s belief that

72 , ,   the diamond is in the house is more likely to persist throughout the night, given that he knows it is there, than given that he has a true belief. After all, true belief that the diamond is in the house might be essentially based on inference from a false premise belief, say, that the diamond is in the drawer. Once the false premise belief is discovered to be such, the probability that the burglar will abandon his belief that the diamond is in the house will increase significantly. In contrast, knowledge cannot be essentially based on inference from a false premise. Given that this is so, it cannot be the case that the burglar discovers that his premise belief is false. The burglar’s continued belief that the diamond is in the house is less likely to be lost in the way described, given that he knows, than given that he merely believes truly. Of course, the probability that the burglar continues ransacking the house all night in search of a diamond, given that he continues to believe that the diamond is there, is much higher than the probability of so ransacking it, given that he has abandoned his belief way before the end of the night. As a result, the probability of the burglar ransacking the house all night, given that he knows that the diamond is there, is higher than the probability of him ransacking the house all night, given that he truly believes that the diamond is there. The burglar’s action is better explained in terms of knowledge (Williamson 2000, 62–4). To be clear, I do not mean to take issue with Williamson’s response to the classical objection. On the contrary, I will grant him that knowledge does raise the probability of action in the way Williamson takes it to and that, as a result, action is better explained in terms of knowledge than in terms of true belief. But why mention it at all, then? The answer is that thinking more carefully about Williamson’s response to the classical objection eventually does provide reason against Mental State. Or, to be more precise, it does so in the sense that it provides reason not to accept Mental State in addition to Epistemic State.²² Here’s why. ²² In this respect, my argument differs from the (perhaps seemingly similar) ones in Magnus and Cohen (2003) and Molyneux (2007) both of which attempt to show that explanations of actions in terms of broad conditions—and, in particular, knowledge—aren’t better than in terms of narrow conditions—and, in particular, some form of belief. Unlike them, I do not mean to take a stance on whether explanations of actions in terms of broad conditions such as knowledge are ever better than explanations of actions in terms of narrow conditions here. Rather, my point specifically concerns the debate over Mental State. In particular, what I aim to show is that



73

As a first observation, note that not only mental states can contribute to the explanation of action. Properties of mental states may do so as well. To return to Williamson’s case of the burglar, his ransacking of the house all night in search of a diamond will be better explained by a dogmatic or by an appropriate desire to find the diamond than by a desire to find it. Similarly, his action will be better explained by a dogmatic or a well-evidenced belief that the diamond is in the house than by a belief. The reason for this is in essence the same as the reason why explanations in terms of knowledge are better than explanations in terms of belief. Dogmatic and appropriate desires and dogmatic and well-evidenced beliefs are more robust (in the sense that they are less likely to be lost) than desires and beliefs. As a result, the probability of continued action conditional on dogmatic or well-evidenced belief is higher than the probability of continued action conditional on belief (and similarly for desire). However, dogmatic or well-evidenced beliefs are not mental states in their own right. Rather, dogmaticity and well-evidencedness are properties of such mental states as belief and desire. With these points in play, let’s suppose we are trying to decide whether to augment Epistemic State by Mental State. To this end, let’s ask the following question: what is more plausible? That knowledge better explains action than belief because knowledge is a mental state in its own right or that knowledge better explains action than belief because qualifying as knowledge is a property of belief that better explains action. Before answering this question, let us broaden our perspective a bit. More specifically, let’s consider a series of explanations of actions such as Williamson’s burglar ransacking the house all night in search of a diamond: (i) in terms of knowledge; (ii) in terms of justified true belief; (iii) in terms of true belief; (iv) in terms of belief; (v) in terms of absence of belief. Note that Williamson’s argument serves to show that (i) is better than (ii), (ii) is better than (iii), (iii) is better than (iv), and (iv) is better than (v). After all, thanks to differences in robustness, the probability of action conditional on knowledge is higher than that conditional on justified true belief, which, in turn, is higher than the probability of even if Williamson is right about broad conditions and causal explanations of action, there is still reason not to adopt Mental State in addition to Epistemic State.

74 , ,   action conditional on true belief, and so on. If Williamson is right and knowledge is a mental state, there are two junctures at which there is a mental difference in kind between our explanations of action: at the first juncture, between the explanation in terms of knowledge and that in terms of justified true belief, and at the last juncture, between the explanation in terms of belief and absence of belief. Now, here is a question: what would indicate a difference in kind at any particular juncture in the series? One plausible hypothesis is that one such indication would be that there is a large difference in the quality of explanation at that juncture, in the sense that the probability of action conditional on one state is much higher than that conditional on the other. Note that this hypothesis is confirmed by the juncture that uncontroversially instantiates a difference in kind, that is, that between belief (iv) and absence of belief (v). The probability of action conditional on belief is much higher than the probability of action conditional on absence of belief. In Williamson’s case, the probability of the burglar ransacking the house all night in search of a diamond conditional on his believing that the diamond is there is much higher than the probability of this happening conditional on his not believing that the diamond is there. What about the juncture at which, according to Williamson, there is another difference in kind, to wit, between (i) and (ii)? It is not hard to see that we do not find a similarly significant difference in the conditional probabilities. Cases of justified true belief that fall short of knowledge are Gettier cases. However, Gettier cases are rare. The probability of being in a case of justified true belief without knowledge is generally extremely low. If the probability of being in such a case is generally extremely low, since an agent with a justified true belief has knowledge if and only if he is not in such a case, then the probability that an agent will act conditional on knowing will generally be only marginally higher than the probability of his acting conditional on having a justified true belief. It comes to light that while explanations of action in terms of belief are significantly better than explanations of action in terms of absence of belief, the same is not true in the case of explanations of action in terms of knowledge vs explanations of actions in terms of justified true belief. Add to this that there are other differences (between explanations (ii)



75

and (iii) as well as (iii) and (iv)) that, whilst not being mental differences, affect the quality of explanation of action. The picture we get is one in which we have a significant difference in the quality of explanation between (iv) and (v) and then a more or less gradual increase in the quality of explanation between (i) and (iv). With this picture in play, let’s return to our initial scenario in which we are trying to decide whether or not to add Mental State to Epistemic State, and to our question as to what is more plausible: that knowledge better explains action than belief because knowledge is a mental state in its own right or that knowledge better explains action than belief because qualifying as knowledge is a property of belief that better explains action. In view of the available evidence, it is hard to deny that the answer is that it is the latter. It is hard to deny that the more or less gradual increase in quality of explanation we find here fits much more nicely with an explanation in terms of properties of belief that add to the robustness of the belief all the way from (i) to (iv) than with an explanation in terms of such properties between (ii) and (iv) and an explanation in terms of a different kind of mental state between (i) and (ii).²³ However, this answer is, of course, not available to champions of Mental State, who are firmly committed to the former answer. In contrast, those who accept that knowledge is an epistemic state only can easily give the more plausible answer. According to them, knowledge, justified true belief, and true belief are all epistemic states, that is, instantiations of epistemic properties by mental states. Since they reject Mental State, they may hold that there is no difference in the kind of mental state invoked in explanations (i) and (ii). At the same time, their view is entirely compatible with the orthodox view in the philosophy of mind, which entails that there is a difference in kind of mental state at issue in explanations (iv) and (v). They can, of course, also allow that

²³ But is it really correct that there is a more or less gradual increase in the quality of explanation between (i) and (iv)? Isn’t it more plausible that there is another significant difference between an explanation in terms of belief and true belief? I must confess that I am not convinced about this. That said, there is reason to think that I could in principle grant Williamson as much. After all, given just how rare Gettier cases are, it is quite plausible that the difference in quality of explanation is higher between (ii) and (iii) as well as (iii) and (iv) than between (i) and (ii). If so, it remains hard to see why exactly there should be a difference in kind just where Williamson needs it, i.e. between (i) and (ii).

76 , ,   properties of mental states, including epistemic ones, can affect the quality of explanations of action. Given that epistemic states are instantiations of epistemic properties by mental states, this means that they can allow that epistemic states might affect the quality of explanations of action. Finally, it may well be that different epistemic states affect the quality of explanations of action in different ways. In fact, given that different epistemic states are associated with different degrees of robustness of the underlying mental states, this is to be expected. As a result, facts such as that the probability of successful action conditional on knowledge turns out to be higher than conditional on justified and/or true belief do not constitute a problem for those holding that knowledge is an epistemic state only. It comes to light that those who accept that knowledge is an epistemic state only can give the more plausible answer to our question about causal explanations of action, whereas those who follow Williamson in accepting that it is also a mental state are committed to the less plausible alternative. In this way, we have one important reason for favouring the former view.

3.3 Theoretical Virtues Rejecting Mental State will also allow us to secure at least two theoretical virtues in our theories of mental and epistemic states. First, our theory of mental states will be more parsimonious if it does not countenance knowledge as a mental state in its own right, alongside other mental states such as belief and desire. Second, knowledge would be somewhat queer among epistemic states. After all, for other epistemic states (with the possible exception of seeing that), there has never been much doubt that they are not mental states in their own right, that is, not even according to Williamson. Rather, they are instantiations of epistemic properties by mental states. As a result, steering clear of Mental State will allow us to attain a more uniform theory of epistemic states according to which all epistemic states are instantiations of epistemic properties by mental states. Resisting Mental State promises to deliver a theory of mental and epistemic states that is more parsimonious and uniform



77

than one that embraces Mental State. There is thus further reason to favour the thesis that knowledge is an epistemic state only.

4. Conclusion This chapter has developed a novel account of knowledge. In contrast to traditionalist approaches, which have attempted to dismantle knowledge into simpler elements, including justified and/or true belief, this account offered a network analysis of knowledge. It aims to understand the nature of knowledge in terms of its place in a network featuring inquiry and belief. I have argued that this network analysis brings a number of theoretical benefits with it, including a viable account of knowledge, and a diagnosis of why we shouldn’t expect knowledge to admit of dismantling analysis. In addition, it made the attractive thesis that belief is constitutively governed by a knowledge norm available without according knowledge explanatory priority over belief. I responded to a number of objections to the view and took a closer look at the most prominent competing view(s) championed by Williamson. Chapter 3 will focus on conditions on knowledge and the Gettier problem.

3 Conditions on Knowledge Chapter 2 offered a novel account of knowledge which aims to allow us to understand the nature of knowledge in terms of its place in a network of related phenomena, including inquiry and belief. In this chapter, I will look at a set of epistemic conditions that are widely believed to be necessary for knowledge—namely epistemic justification and being in a suitable epistemic environment—and show how they can be fitted into my preferred account. What’s more, I will take up the key issue in the traditional debate on the nature of knowledge, to wit, the Gettier problem, and show how the aforementioned conditions on knowledge can help us deal with it. In a nutshell, on my view, a belief is knowledge only if it is produced and perhaps sustained by an ability to know in a suitable environment. And a belief is justified if and only if it is produced and perhaps sustained by an ability to know, whether or not the environment is also suitable. Gettier cases are among the cases in which the agent believes via an ability to know in an unsuitable environment. That’s why the beliefs of agents in Gettier cases are justified but fall short of knowledge. The remainder of this chapter will develop this idea in considerable detail. There are two main reasons for going into this amount of detail. First, I want my account of conditions on knowledge to be continuous with the general story about ACANs that is key to the account of knowledge. Second, the account of justified belief that I defend will do work in my discussion of the problem of scepticism in Chapter 6. What’s more, the details of the account matter. So, it’s important to get them into clear view.

Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0004

  

79

1. Knowledge and Ability Recall that on the view developed in Chapter 2, inquiry into whether p aims at properly closing the question whether p for oneself and that one does this if and only if one knows that p (not-p). For now, let’s focus on proper question closing. Note that the aim of inquiry features a normative property. To attain the aim of inquiry into whether p, the question whether p must be properly closed for one. One interesting fact about ACANs that feature normative properties in their aims is that they often impose further substantive constraints on the means of attaining the ACAN’s constitutive aim and on the environment. Consider, for instance, fairly winning an archery competition. To win an archery competition fairly it won’t be enough that you score more points than your opponents. There are constraints on the means for attaining the aim. For instance, you may not do so with the help of performance-enhancing drugs or equipment (perhaps AR glasses that tell you exactly whether you are aiming correctly). Even if performanceenhancing drugs or equipment help you win the competition, you will not thereby have won fairly. In fact, it is impossible for you to win fairly in this way. Similarly, there are constraints on the environment. For instance, your competition must not be hampered by performancediminishing drugs (while you aren’t). Even if you end up winning, if your competition has been hampered by performance-diminishing drugs (while you aren’t), you will not thereby have won fairly. In fact, again, it’s impossible for you to win fairly in this situation. Since of the aim of inquiry features a normative property, the question arises whether there are substantive constraints on the means of attaining its aim and on the environment here as well. There is reason to think that the answer is yes. To see this, note that certain means are clearly unfit for properly closing a certain question for one. Think, for instance, of going by the results of the toss of a fair coin, which will give you the true answer with a probability of 0.5. If you close a question in this way, you will not have done so properly. In fact, just as the use of performance-enhancing drugs will make it impossible for you to win a race fairly, so the use of a fair coin toss as a way of closing a question will make it impossible for you to close the question properly. And, likewise,

80 , ,   there are environments that make it impossible to properly close certain questions for one. The perhaps best-known examples here are various forms of radical sceptical scenarios in which virtually nothing is as it seems. If you are subject to radical deception and close certain questions such as whether you have hands, you will not have done so properly. In fact, just as being in a competition in which your rivals are hampered by performance-diminishing drugs (while you aren’t) will make it impossible for you to win fairly, so being in a radical sceptical scenario will make it impossible for you to close certain questions properly. Given that one knows if and only if one attains the aim of inquiry, these considerations provide excellent reason to believe that knowledge features substantive constraints on the means of attaining knowledge and the environment. Now, this will not be particularly surprising to anyone with even a passing interest in epistemology. After all, it is widely agreed that knowledge requires justification and a suitable environment. And it is only natural to think of the justification condition as a substantive constraint on the means of attaining knowledge and to think of the condition requiring a suitable environment as a substantive constraint on the environment. In this way, what the present proposal predicts is borne out by what is widely regarded as independently plausible. And, of course, that’s good news. The question remains whether more can be said about these substantive constraints. Can they be made more precise? Are they completely independent or are they related? In what follows, I will develop answers to these questions. Again, my strategy is to use theoretical machinery from the general theory of ACANs. In particular, my answers will appeal to accounts of ability to attain the aims of ACANs, their exercise, and competent ACAN moves. Let’s start by taking a closer look at these accounts.

1.1 ACAN Abilities Practitioners of ACANs may have the ability to attain the aim of a given ACAN. Consider, for instance, a simple version of archery—call it ARCH—in which shots are taken at a disc-shaped target with no further

  

81

structure from a certain distance. The constitutive aim of ARCH is to hit the target. Now, you, the reigning world champion in ARCH, have the ability to hit the target, while I, a blind man, don’t. Let’s take a closer look at these abilities (henceforth ‘ACAN abilities’). ACAN abilities involve ways of move production. Your ability to hit the target involves a way of shooting. More than one way of move production may be available to any one agent. For instance, you may shoot with your right hand or with your left hand. Of course, not all ways of move production will qualify as ACAN abilities. While you may have the ability to hit the target in ARCH when shooting with your right hand, you may not have this ability when shooting with your left. ACAN abilities are relative to conditions. You may have the ability to hit the target in ARCH when shooting with your left when sufficiently concentrated, sober, not being shoved while releasing the arrow, when shooting in normal winds, when there are no jokesters destroying the target at the last second, and so on. The conditions relative to which one may have an ACAN ability may differ across different ways of move production. While you may have the ability to hit the target in ARCH when shooting with your left hand but only when you are completely sober, you may have this ability when shooting with your right even after a couple of beers. ACAN abilities involve dispositions to attain the relevant aims. Dispositions are associated with (i) trigger (T) and (ii) manifestation (M) conditions. In the case of ACAN abilities, the manifestations are, of course, attainments of the ACAN’s aim. In addition, the trigger conditions are uses of ways of move production. Furthermore, dispositions are (iii) relative to conditions (C). In the case of ACAN abilities these conditions are the ones to which the relevant ability is relative. To possess an ability to attain the aim of a given ACAN, S, relative to conditions C, a practitioner must have a way of move production, W, such that they are disposed to attain S’s aim (M) when using W (T) in C (C). For instance, to have the ability to hit the target in ARCH relative to C you must have a way of shooting such that you are disposed to hit the target when using it in C. Finally, (iv) dispositions are widely regarded as being associated with counterfactual conditionals of the form ‘were T to obtain in C, the possessor of the disposition would exhibit M’. This means

82 , ,   that practitioners of ACANs who have an ability to attain success in a given ACAN, S, relative to conditions C will be such that were they to use the way of move production at issue in this ability (T) in C (C), they would attain S’s success (M). Those who have the ability to hit the target in ARCH relative to C will be such that were they to use the way of shooting at issue in this ability in C, they would hit the target.¹ A way of move production that qualifies as an ability to attain the aim of one ACAN may or may not so qualify for another ACAN. Consider another ACAN, ARCH’. ARCH’ is just like ARCH except that the target has a different shape from that of the target in ARCH. Instead of being disc-shaped, it is star-shaped. It may be that, in a certain set of conditions, C, a way of move production—shooting with your left hand, say— disposes you to hit the target in ARCH but not ARCH’. Your left-handed way of move production qualifies as an ability to hit the target in C relative to ARCH but not ARCH’. You are not that good an archer when shooting with your left hand. At the same time, it may also happen that a(nother) way of move production qualifies as an ability to hit the target in C for both ARCH and ARCH’. It may be that, in C, shooting with your right hand disposes you to hit the target not only in ARCH but also in ARCH’. Your right-handed way of move production qualifies as an ability to hit the target in C relative to ARCH and ARCH’. You are that good an archer when shooting with your right hand. For that reason, a way of move production may also qualify as an ability to attain the aim for a range of ACANs. A way of move production qualifies as an ability to attain the aims of a given range, R, of ACANs only if it disposes one to attain the relevant aim when used for each ACAN in R. Your right-handed way of move production may constitute

¹ Properties (i), (ii), and (iv) above are among what Jennifer McKitrick (2003, 157) calls ‘markers of dispositionality’, that is, properties of dispositions that are widely agreed upon among contributors to the debate on dispositions. For what I take to be convincing motivations of property (iii), see, e.g., Mumford (1998) and Sosa (2015). Finally, it might be better to construe dispositions as being associated with a probability of manifestation conditional on triggering in suitable conditions, rather than with the counterfactual conditionals at issue in (iv) above. One advantage that this kind of account of dispositions has for the application to abilities is that it can easily make sense of the fallibility of abilities, even in favourable conditions. After all, the relevant probability of manifestation may but need not be 1, even when the triggering conditions obtain in favourable conditions. See, e.g., Healey (1991) and Suarez (2007) for more on probabilistic approaches to dispositions.

  

83

an ability to hit the target across ARCH, ARCH’, ARCH” (squareshaped), etc. Your left-handed way of move production may constitute an ability to hit the target across ARCH, ARCH”, but not ARCH’, etc. With these points in play, I’d like to propose the following account of ACAN abilities: ACAN Ability One has an ability to attain the aim for a range, R, of ACANs and relative to conditions, C, if and only if one has a way of move production, W, such that, for any S∈R, using W in C disposes one to attain the aim of S.²

1.2 Exercises of ACAN Abilities I want to suggest that exercises of ACAN abilities are uses of ways of move production involved in ACAN abilities. Or, more precisely: ACAN Exercise One exercises an ability, A, to attain the aim for a range, R, of ACANs and relative to conditions, C, if and only if one has A and produces a move via the way of move production at issue in A.

It is important to note that placing the agent in conditions relative to which they do not have an ACAN ability can have different effects on an agent’s performances. Some such conditions will result in preventing an agent from using their way of move production. For instance, being too drunk, distracted, nervous, shoved while releasing the arrow, and so on will prevent you from using the way of shooting that qualifies as an

² See Kelp (2016, 2018) for a more detailed account of ACAN Abilities (as well as ACAN Exercise and Competent ACAN Move below). Note also that the full account of ACAN abilities additionally features what I call a groundedness condition on abilities, which is a generalization of Millikan’s (2000, ch. 4) idea that genuine abilities are distinguished from mere behavioural dispositions by being relative to conditions in which they were acquired via learning or natural selection. That said, the grounding condition does no substantive theoretical work here. Accordingly, for present purposes at least, it can safely be set aside.

84 , ,   ACAN ability relative to some (albeit different) conditions. I will henceforth refer to conditions that, when not satisfied, prevent the agent from using their way of move production as conditions of shape (SH).³ According to ACAN Exercise, then, exercising an ACAN ability requires that SH be satisfied. In contrast, other such conditions do not prevent the agent from using their way of move production when not satisfied and so allow them to exercise their ability anyway. Suppose, for instance, that you fire a shot that would have hit the target had it not been for a jokester who destroys the target when the arrow is about to hit it. Even though your shot misses the target, you do get to produce a move via the way of shooting that qualifies as an ACAN ability relative to some (albeit different) conditions. I will henceforth refer to conditions that, when not satisfied, do not prevent the agent from using their way of move production as environmental conditions (EN). According to ACAN Exercise exercising an ACAN ability does not require that EN be satisfied.

1.3 Competent ACAN Moves Competent moves in an ACAN require the exercise of an ACAN ability. When producing a shot in ARCH, your shot will be competent only if it is produced by an ability to hit the target.

³ Note that, as a result, unsuitable SH may and often do explain incompetent performance in an indirect way. What directly explains incompetence is that no way of move production that qualifies as an ability was used. For instance, suppose you are given a drug with the result that you are not in suitable SH for any ability to hit the target in ARCH you may have. That you are drugged explains why you don’t shoot in a way that qualifies as an ability to hit the target (perhaps you mess up your aim), which in turns explains why your shot is not competent. Similarly, suppose a certain official aiming to marry a couple is too drunk, with the result that they are not in suitable SH for their ability to marry couples. That the official is too drunk explains why they don’t perform the ceremony in a way that qualifies as an ability to marry couples (perhaps they mess up the words), which indirectly explains why their performance was not competent. Finally, suppose that I just received a letter about my application to do graduate studies and want to find out whether I was admitted into my favourite programme. If I am too nervous, this may explain why I read in a way that does not qualify as an ability to know (perhaps I am reading the words wrong), which indirectly explains why the formation of my belief was not competent.

  

85

However, a competent move requires more than the exercise of an ACAN ability. To see this, let’s return to the case in which you are the reigning world champion of ARCH. Suppose that you are currently engaging in ARCHX in which the target changes its position discontinuously, randomly, and rapidly. Let’s assume, as is plausible anyway, that you do not have the ability to hit the target in ARCHX. You have no way of shooting that disposes you to attain the aim of ARCHX, no matter what conditions we may place you in. Suppose you take a shot using a way of move production that disposes you to hit the target in a range R of ACANs and relative to conditions C. Here you exercise your ACAN ability to hit targets in range R and relative to C. However, that does not make your shot competent. The ability you exercise is the wrong ability for the ACAN you are engaging in. (An even clearer example may be the following: suppose you have a way W of producing lay-ups in basketball that qualifies as an ability to score relative to some C. Currently you are standing at the three-point line and have two seconds to score a basket to win the game. Suppose you produce a shot via W, which, of course, doesn’t even get close to the basket. On the present view, you exercise an ability to score. However, your shot is not competent. Again, the reason for this is that you are exercising the wrong ability to score here.) For a move to be competent, then, it must be a move in an ACAN, S, that is within the range R for which your way of move production qualifies as an ability. Contrast the situation described above with one in which you engage in an ACAN that, we may assume, is within the range, R, of your ACAN ability, but in which a jokester prevents the shot from being successful. Here you not only exercise an ability to hit the target, but your shot is also competent. There is thus reason to believe that, in order to produce a competent move in a given ACAN, the ACAN must be within the range (of ACANs) of the ACAN ability exercised. The above considerations thus motivate the following account of competent moves: Competent ACAN Moves A move in a given ACAN, S, is competent if and only if it is produced by an exercise of an ACAN ability to attain the aim for a range, R, of ACANs and relative to conditions, C, such that S∈R.

86 , ,  

1.4 Substantive Conditions on Proper Question Closing Recall that we saw that there is reason to think that there are substantive constraints on both the means and the environment for attaining the aim of inquiry. And given that knowledge is the aim of inquiry, there is also reason to think that there are substantive constraints on the means and the environment for attaining knowledge. The good news was that this is independently plausible, since knowledge is widely believed to feature both a justification condition and a condition requiring a suitable environment. However, the question remained as to whether more can be said about these constraints. With the accounts of ACAN abilities, their exercises, and competent ACAN moves in play, I am now in a position to answer this question. I want to unpack both substantive constraints in terms of the notion of an ACAN ability. In this way, then, there is a relation between these two conditions. More specifically, I want to suggest that the constraint on the means for attaining the aim of inquiry/knowledge is unpacked in terms of a competence condition and the constraint on the environment in terms of the satisfaction of the EN of the abilities that produce the belief. In other words: Conditions on Proper Question Closing The question whether p is properly closed for oneself only if (Constraint on Means =) one’s belief that p/not-p is competently produced and (Constraint on Environment =) the EN of the abilities by means of which one’s belief that p/not-p is produced are satisfied. Conditions on Knowledge One knows that p only if (Constraint on Means =) one’s belief that p is competently produced and (Constraint on Environment =) the EN of the ability by means of which one’s belief that p is produced are satisfied.

To see exactly what Constraints on Means and Environment amount to, here is what the relevant instances of ACAN Ability, ACAN Exercise, and Competent ACAN Moves that serve to precisify these constraints come to:

  

87

Ability to Know One has an ability to know propositions in a range, R, and relative to conditions, C, if and only if one has a way of belief formation, W, such that, for any p∈R, using W in C disposes one to form knowledgeable beliefs that p. Exercises of Abilities to Know One exercises an ability, A, to know propositions in range R and relative to conditions C if and only if one has A and forms a belief via the way of belief formation at issue in A. Competent Belief One competently believes that p if and only if one’s belief that p is produced by an exercise of an ability to know propositions in range R and relative to conditions C such that p∈R.

The results we get are the following alternative statements of Conditions on Proper Question Closing and Knowledge: Conditions on Proper Question Closing One has properly closed the question whether p for oneself only if (C1 =) one’s belief that p/not-p is produced by an exercise of an ability to know propositions in range R and (C2 =) relative to conditions C such that p∈R, and the EN of the ability to know that produced one’s belief that p/not-p are satisfied. Conditions on Knowledge One knows that p only if (K1 =) one’s belief that p is produced by an exercise of an ability to know propositions in range R and relative to conditions C such that p∈R, and (K2 =) the EN of the ability to know that produced one’s belief that p are satisfied.

1.5 Why These Conditions? Now, even if we are willing to go along with the idea that for ACANs with constitutive aims that feature normative properties there may be

88 , ,   substantive constraints on both the means of attaining these aims and the environment, and even if we are willing to grant that inquiry is an ACAN with a constitutive aim featuring a normative property, why should we think that this specific way of precisifying these constraints is correct? While I don’t have anything conclusive to offer in response to this question, there are some considerations that provide relevant evidence. The first is that the above precisification accommodates the motivations for the substantive constraints on the means and environment for attaining the aim of inquiry. For instance, we can now explain why you cannot properly close a question via mere guesswork. Mere guesswork simply couldn’t be an ability to know. This, in turn, is because mere guesswork simply doesn’t dispose one to come to know propositions in any range. But, of course, if mere guesswork couldn’t be an ability to know, beliefs acquired by mere guesswork simply couldn’t be competent. Given C1 of Proper Question Closing, it follows that it’s impossible to properly close a question via mere guesswork. Similarly, we can explain why it’s impossible to properly close certain questions when one is radically deceived. Even if the beliefs one forms are competent, the EN of the ability to know that produced these beliefs could not be satisfied. This, in turn, is because the disposition to know at issue in this ability simply could not extend to the EN of radical deception. By C2 of Conditions on Proper Question Closing, it follows that it’s impossible to properly close certain questions when one is radically deceived. The second reason for thinking that the precisification is correct is that it serves to address one of the central problems in the literature on the nature of knowledge, i.e. the Gettier problem. To see how, consider first the following two famous Gettier cases: Stopped Clock Having come down the stairs, you look at the grandfather clock in the hallway, see that it reads 8.22 and on that basis comes to believe that it is 8.22. The clock has an outstanding track record of functioning properly and you have no reason to think that it is currently not accurate. Your belief is true. It is in fact 8.22. Unbeknownst to you, however, the clock has stopped exactly twelve hours ago.

  

89

Fake Barns You are driving through the countryside and are currently looking out of the window of your car. You see what appears to be a barn in the field and form a perceptual belief that there is a barn in the field. Unbeknownst to you, you are looking at one of the few real barns in an area peppered with barn façades that are so cleverly constructed as to be indistinguishable from real barns from your position on the road.

In Stopped Clock, you don’t know that it is 8.22 and in Fake Barns, you don’t know that there is a barn in the field. One of the central difficulties for traditional reductive analyses of knowledge is to explain the absence of knowledge in these cases. Fortunately, the present proposal can succeed where traditional reductive analyses have failed. It allows us explain the absence of knowledge in Gettier cases in terms of the violation of some substantive conditions on knowledge. More specifically, the reason why agents in Gettier cases lack knowledge is that the EN for the ability to know that produces the relevant beliefs in Gettier cases are not satisfied. To see in more detail why this is so, let’s return to our two toy cases and look at how things play out here. In Stopped Clock, the clock you are taking a reading from is stopped. As a result, you are not in conditions C relative to which you have the ability to acquire knowledge about the time. After all, your way of belief formation does not dispose you to form knowledgeable beliefs about the time in conditions in which the clock you are taking a reading from is stopped. At the same time, the conditions you find yourself in do not prevent you from forming your belief in the way that constitutes an ability to know relative to some (but different) conditions. This means that you are in unsuitable EN for the ability that produced your belief about the time. By K2 of Conditions on Knowledge, your belief that it’s 8.22 falls short of knowledge. Similarly, in Fake Barns, the fact that you are in a part of the country in which fake barns predominate means that you are not in conditions C relative to which you have the ability to recognize barns. At the same time, the conditions you find yourself in do not prevent you from

90 , ,   forming your belief in the way that constitutes an ability to know relative to some (but different) conditions. This means that you are in unsuitable EN for your ability to recognize barns. By K2 of Conditions on Knowledge, your belief that there is a barn in the field falls short of knowledge. Let’s move on to the third reason, then. To see this one, recall that we found it only natural to identify the constraint on the means with the independently plausible justification condition on knowledge. What the above precisification of the constraint on means does, then, is to deliver the following precise account of the justification condition on knowledge: Justified Belief One’s belief that p is justified if and only if it is formed by an exercise of an ability to know propositions in range R and relative to conditions C such that p∈R.⁴

In a nutshell, the third reason for thinking that the above precisification of the constraints on the aim of inquiry is correct is that (modulo the natural identification of the constraint on means with the justification condition on knowledge) it delivers Justified Belief, which is independently attractive. In fact, I have defended Justified Belief at length in a recent book (Kelp 2018). There I argue that the account offers promising solutions to a wide range of central problems in the theory of justification, including the new evil demon problem, the clairvoyant problem, the generality problem, the new Gettier problem and the lottery paradox. Now, I could not hope to even begin to rehearse my case for Justified Belief, not least because it spans the entirety of the earlier book. That’s why I will have to content myself with (i) observing that the proposed precisification effectively allows me to incorporate the relevant material into the present project and (ii) referring the interested reader to the book for more detailed arguments.

⁴ For other accounts of justified belief in terms of abilities to know, see Miracchi (2015), Schellenberg (2018), Silva (2017). For other accounts of justified belief in terms of knowledge, see, e.g., Bird (2007a), Jenkins Ichikawa (2014), Littlejohn (2013), Millar (2010), Reynolds (2013), Simion (2019), Sutton (2007), and Williamson (2010).

  

91

2. Objections 2.1 Redundancy Even if the above account of the absence of knowledge in Gettier cases works, one may wonder whether it is even required on the present view. To see why one might think it isn’t, note that, when engaging in ACANs, we don’t always succeed. On the contrary, failure may very well be frequent. When we do fail, we may ask why we fail. Suppose you take on a shot in ARCH and miss. You may very well wonder why your shot was unsuccessful. There are two generic sources of failure in ACANs: the problem lies either with the agent or with the environment. By way of illustration, let’s return to your unsuccessful archery shot. Typical sources of failure are that your aim is off, your hands are too shaky, and so on. Here failure is sourced in the agent, and, more specifically, in your way of shooting, which falls short. However, it may also be that there is nothing wrong with the agent. Rather, the problem lies with the environment. In the archery case, this may be, for instance, when your shot would have been successful had it not been for a gust of wind that brought it off its path to the target or a rogue bird which picked it up in mid-flight. Now, since inquiry is an ACAN, the same goes for inquiry. We may fail to attain the aim of a given inquiry due to problems with the agent or with the environment. For instance, your belief about whether a certain defendant is guilty may fail to attain the aim of inquiry because of bias or wishful thinking. In this case, the problem lies with you, and, more specifically, your way of belief formation, which falls short. Alternatively, the problem may lie with the environment. In the above case, it may be that unbeknownst to everyone, the perpetrator has an identical twin who has mistakenly been arrested and is now standing trial. When you form a belief that the defendant is guilty based on the excellent evidence for their guilt, there may be nothing wrong with you. Rather, your failure is sourced in a problem with the environment. Since, on the present proposal, knowledge is the aim of inquiry, the above story about explaining failures in ACANs allows us to explain instances of absence of knowledge. In particular, we can now explain the

92 , ,   absence of knowledge in Gettier cases. Gettier cases are very plausibly cases of failure to know that are sourced in problems with the environments. By way of illustration, recall that, in Stopped Clock, you don’t know that it is 8.22 and in Fake Barns, you don’t know that there is a barn in the field. Why not? As we have already seen, there doesn’t seem to be any issue with you here. Your way of proceeding is just fine. You are perfectly competent at reading clocks and at recognizing barns. Rather, the problem lies with the environment. The fact that the clock you are taking a reading from is stopped explains why you don’t know that it’s 8.22. Similarly, the fact that there are too many fake barns in the area explains why you don’t know that there is a barn in the field. What comes to light is that the absence of knowledge in Gettier cases can be explained in a perfectly well-motivated way. In fact, it’s but an instance of a familiar form of explanation of failure in terms of unsuitable environments. But if we have a perfectly viable explanation of why agents in Gettier cases lack knowledge at hand, it looks as though we simply do not need the conditions on knowledge developed above to do the trick. In response, note first that I would not at all be bothered if it turned out that we simply do not need substantive conditions on knowledge to offer a satisfactory account of Gettier cases. If anything, this would be grist for my mill. That said, there is reason to think that we do need substantive conditions on knowledge after all. This is because there is an important difference between standard cases of failure in ACANs, on the one hand, and Gettier cases, on the other. To see this, let’s return to the case in which your shot in ARCH fails due to a gust of wind. While the gust of wind may explain past failures and while it may be a reliable means of predicting future failures, note that it does not exclude success. After all, a second gust of wind may bring your shot back on target, with the result that it ends up being successful. Gettier cases are different in that unsuitable environmental conditions not only explain past failure and serve to reliably predict future failures, but also exclude success. When you form a belief that it’s 8.22 by taking a reading from a stopped clock, it’s impossible for you to come to know that it is 8.22 on the basis of this reading, even if the reading is perfectly competent. And if you form a belief that there is a barn in the field by looking at a field peppered

  

93

with indistinguishable fakes, it’s impossible for you to come to know that there is a barn in the field by looking, even if your belief is perfectly competent. Given that this is so, a proper treatment of Gettier cases must not only explain why knowledge happens to be absent in Gettier cases but also account for why knowledge couldn’t be present in Gettier cases. And, of course, this strongly suggests that the absence of knowledge in Gettier cases must be explained in terms of the violation of some substantive condition on knowledge.

2.2 Queerness One can attain the aim of an ACAN via a move that is not competent. Consider a case in which you take a shot in ARCH. Suppose you are completely drunk when taking the shot. As a result, your shot is not competent. But now suppose that, by an incredible stroke of luck, your shot finds the target anyway. Your shot attains the aim of ARCH without being competent. Moreover, attaining the aim of a given ACAN does not require suitable conditions for one’s ability either, as the two gusts of wind case we looked at earlier clearly indicates. Why think that these observations cause trouble for the proposed view? To answer this question, note that on the present view inquiry differs from other familiar ACANs in just this respect. To attain the aim of inquiry, your belief must be competent and the environment must be suitable. As a result, one may now worry that the present view turns inquiry into an oddball amongst ACANs. Now, it might be thought that all we need by way of response is to recall that, for ACANs with aims that feature normative properties, there may be substantive constraints on both means and environments for attaining the aim. This was clearly shown by the case of fairly winning an archery competition. While I think this line of thought goes a long way towards a viable response here, there is a residual worry that may be worth addressing. To see this, note that while the fair winning case demonstrates the existence of some substantive constraints, it does not demonstrate the existence of competence constraints in particular. After

94 , ,   all, it’s easy to see that you can win an archery competition fairly without competence. Despite the fact that you are drunk beyond comprehension, all of your shots find the target by complete luck and none of your opponent’s does. You win the competition fairly but not by means of competence. Perhaps, then, inquiry is an oddball among ACANs after all just because it imposes an unheard-of competence requirement on the attainment of its aim. To repeat, I do believe that the fact that there are substantive constraints on means and environment for attaining the aims of some ACANs with aims featuring normative properties goes a long way towards addressing the issue we are facing here. After all, once we see that there are some such substantive constraints, it’s a short way to acknowledging that there may also be different ways in which these substantive constraints manifest themselves. But once we have granted this, the worry that the competence requirement on attaining the aim of inquiry makes inquiry into a problematic oddball amongst ACANs is deprived of much, if not all of its force. Even so, to eliminate any residual worry that may remain, it will be useful to show that inquiry isn’t the only ACAN featuring the sort of competence constraint we find in Conditions on Knowledge. One of my favourite examples is the activity of joining consenting adults in lawful marriage (henceforth LM). LM is an ACAN, the aim of which is to effect the lawful marriage of the target couple. Moves are tokens of ceremonies of a certain type—call it c. The aim of LM is a token of c that effects the lawful marriage of the target couple. (The definitions of LM ability, exercise thereof, and competent move as expected.) Crucially, there is reason to believe that one joins partners in lawful marriage only if one competently conducts a token of c. To see this, suppose you, the person performing the ceremony, fail to competently conduct a token of c. This may be for two reasons. You may fail to conduct a token of c altogether, say because you are too drunk and so do not manage to say enough of the relevant words. Or else the ACAN may not be in the range of c, say because one of the partners is underage and so a different kind of ceremony is needed. In either case, the couple will not end up lawfully married. Since joining the couple in lawful marriage is the success condition of LM, this means that the aim of LM will not be

  

95

attained. If so, there is reason to believe that there is a competence condition on attaining the aim of LM. What’s more, there is also reason to believe that one joins partners in lawful marriage only if the EN of the ability exercised are satisfied. To see this, suppose that EN are not satisfied. For instance, one key EN for the ability to join partners in lawful marriage is that any caveat lodged against the marriage has been discharged. So, suppose that this EN is not in place. There remains an undischarged caveat against the marriage. If so, again, the couple will not end up lawfully married, even if you have competently conducted a token of c. Since joining the couple in lawful marriage is the aim of LM, this means that the aim of LM will not be attained. As a result, there is also reason to believe that attaining the aim of LM requires that the EN of the ability exercised are satisfied.⁵ ⁵ In a recent paper, Jaakko Hirvelä (2019) argues that success and competence can always come apart. He directly addresses the above lawful marriage case, which he claims is a case in point. Hirvelä gives three reasons for this. First, he takes it that having a disposition to fail to conduct tokens of c excludes the competence to join couples in marriage. Second, he considers the rejoinder that the disposition at issue in the competence is to effect lawful marriage when producing a token of c. In response, he claims that this commits me to an unduly fine-grained way of individuating ways of move production that isn’t in line with how I characterize ways of move production for other ACANs, such as archery, where I appeal to something more coarsegrained such as shooting with one’s left or right hand. Third, he considers a case in which the law is about to be changed and has already changed in many nearby worlds. In this case, he claims, an unwitting official doesn’t have the disposition to lawfully marry a couple by conducting c any longer. At the same time, given that the old law remains in force until the end of the ceremony, the couple ends up lawfully married. None of these reasons is convincing. The first claim is simply false. Having a disposition to fail to produce moves in a given ACAN by means of ability does not exclude the ability to attain the aim in that ACAN. Suppose I have vowed never to produce moves in a given ACAN via my only ability. Instead I either refrain from practising altogether or have taken up practising in a way that falls short of an ability. Does this mean that I lost my ability or that if I were to break my vow once and produced a move via this ability, the move would no longer be competent? Clearly not. The second reason doesn’t fare much better. Contrary to what Hirvelä says, I think that the ways of move production at issue in ACAN abilities will turn out to be quite fine-grained. In the archery case, for instance, the way of shooting at issue in an ability to hit the target in archery must be much more fine-grained than what I laxly labelled ‘shooting with one’s right hand’. There are many ways of shooting with one’s right hand that do not qualify as an ability, and anything that does must be quite fine-grained. Finally, the third claim rests on an implausible modal condition on dispositions/abilities, one on which the failure in nearby worlds in different conditions bears on the presence of the disposition/ability in the actual world. However, it is widely agreed that abilities may be fragile and that if there is a modal condition on dispositions/abilities at all, then the modal condition needs to hold the conditions of the actual world fixed. Once we are clear on these points, it is easy to see that Hirvelä’s variation of the marriage case is not one of success without competence either.

96 , ,   Try-outs for team sports are another example. The constitutive aim of try-outs is to make the team. Whether you make the team will depend on how you perform on a variety of relevant tasks. To take a simple example, consider a try-out for a team that will compete in basketball free-throw competitions. Suppose you are the team coach. Suppose (i) you are trying out a player who is drunk and takes shots, many of which find the hoop, by bouncing them randomly off the floor. Or suppose (ii) the player you are trying out refuses to take shots from the free-throw line and will only take shots, many of which go in, from right under the basket. Finally, suppose (iii) one of the players produces many shots that go in but only because they have a helper with a wind machine in the wings. You will not select any of them. By the same token, there is reason to believe that only shots that are both competent and taken in suitable environmental conditions are successes in your try-out (only those count towards a prospective draft). We thus have another ACAN featuring the sort of competence constraint we find in Conditions on Knowledge.

2.3 Back on the Road to a Dismantling Analysis But don’t the substantive conditions on knowledge put us back in the ballpark of a traditional dismantling analysis? What we now have is a view according to which knowledge features substantive conditions such as justification and a suitable environment. What’s more, one of these conditions serves to explain the absence of knowledge in Gettier cases. So, why not take this to be a traditional dismantling analysis of knowledge after all? Fortunately, it is easy to see why these conditions do not lend themselves to a dismantling analysis. After all, both substantive conditions on knowledge are themselves unpacked in terms of knowledge and so are bound to violate the non-circularity condition on dismantling analyses. It may also be worth noting that there is nothing queer about this. The same holds for ACANs in general, including ACANs such that attainment of the ACAN’s aim requires competent ACAN moves. This is because, in general, an ability to attain the aim of a given ACAN is defined in terms of the ACAN’s aim. For instance, the ability to hit the

  

97

target in ARCH is defined in terms of hits. Accordingly, the prospects for a dismantling analysis of the aim of an ACAN in terms of a competent ACAN move are very dim indeed. Note also that this doesn’t change once we consider other ACANs in which attainment of the ACAN’s aim requires competence and suitable environmental conditions. The ability to join consenting adults in lawful marriage is defined in terms of jointing consenting adults in lawful marriage. A reductive analysis of the aim of this ACAN in terms of a competence is no more promising here than in the archery case. And, of course, the same goes mutatis mutandis, for inquiry.⁶ At the same time, note that the fact that the relevant ACAN abilities are defined in terms of the ACAN’s aim does not mean that we cannot offer adequate explanations of why certain moves were guaranteed to remain unsuccessful in terms of the failure of, for example, the competence condition on attaining the aim or in terms of a condition requiring suitable environmental conditions for that matter. For instance, consider a version of the marriage case in which there is a caveat lodged against the marriage. One question we may ask is why a couple didn’t end up lawfully married. Of course, as we have seen above (Section 2.1) we don’t need to invoke substantive conditions on attaining the ACAN’s aim to explain this. However, we may also observe that, given that the caveat was lodged against the marriage, the couple couldn’t have ended up lawfully married. And this can indeed be adequately explained in terms of one of the substantive conditions on lawful marriage. Since, once again, the same goes, mutatis mutandis, for inquiry, we need not worry that the fact that ACAN abilities are defined in terms of ACAN aims will render any explanation of failure that ultimately must appeal to ACAN abilities inadequate.

3. Conclusion This chapter focused on conditions on knowledge. I argued that, for ACANs with aims featuring normative properties, it is not uncommon ⁶ Note also that this is just another way of putting one of the key insights from the AntiJustification Arguments from Chapter 1.

98 , ,   to find substantive constraints on both the means of attaining the aim and the environment and that inquiry is a case in point. I then developed in more detail an account of what these conditions might look like, which appeals to the notion of an ACAN ability and provided reason for accepting this account. The result is a further development of the network analysis from Chapter 2 which allows us to accommodate widely agreed conditions on knowledge (justification and a suitable environment) and to explain the absence of knowledge in a range of cases, including the notorious Gettier cases. Again, I looked into a number of objections and provided responses. In contrast with previous chapters, I did not discuss competing views. The main reason for this is that I do so at considerable length in the book that develops my preferred account of justified belief. Rather than rehearsing these arguments here, it seemed preferable to save some time and refer the reader to the earlier work instead (Kelp 2018). This completes my discussion of the nature of knowledge. Chapter 4 will focus on inquiry into general phenomena and understanding.

4 Understanding Chapter 3 focused on conditions on knowledge and the Gettier problem. In this chapter, I will move on from inquiry into specific questions to inquiry into general phenomena. More specifically, I will take a closer look at the understanding aim of inquiry into general phenomena and develop an account of understanding that unpacks it in more detail. Again, as will become clear in due course, this account will also take the form of a non-reductive network analysis.

1. The Aim of Inquiry into General Phenomena When I introduced the discussion of the aim of inquiry into general phenomena, I offered the following lightweight characterization: Understanding Aim One’s inquiry into general phenomenon P aims at understanding P.

First things first, inquiry into general phenomena is an ACAN, just like inquiry into specific questions. The aim of understanding is a constitutive aim of inquiry into general phenomena. Second, recall that in the parallel debate about the aim of inquiry into specific questions, we found a number of substantive epistemological theses about the aim of inquiry. These substantive theses could be viewed as unpacking a lightweight characterization of the aim of inquiry into specific questions in terms of question-settling. Now, it is arguable that less ink has been spilled on substantive epistemological theses about the aim of inquiry into general phenomena. That said, there is a lively debate on the nature of understanding. And, of course, given Understanding

Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0005

100 , ,   Aim, these views of understanding will come with commitments about the aim of inquiry into general phenomena. To the extent that Understanding Aim is lightweight enough to enjoy widespread consensus, it will provide a condition of adequacy for more detailed accounts of understanding of general phenomena. Any adequate account of understanding of general phenomena will have to accommodate the thesis that understanding is the aim of inquiry into general phenomena. While Understanding Aim may provide an important checkpoint for an independently conducted debate about the nature of understanding, it will come as no surprise that my own approach accords Understanding Aim a key role in theorizing about understanding, rather than marginalizing it to one condition of adequacy any account of understanding must satisfy. More specifically, I will approach the question of what understanding is by thinking in more depth about the aim of inquiry into general phenomena. To be clear, my aim is not to end up with a reductive analysis of understanding. Rather, in what follows, I will work towards a network analysis connecting inquiry into general phenomena and understanding general phenomena in much the same way as the network analysis from Chapter 2 connected inquiry into specific whether questions and knowledge.

1.1 Phenomena I will start my analysis by thinking about inquiry into general phenomena. To begin with, I’d like to offer a brief sketch of what I take phenomena to be. Metaphysically, phenomena form a rather heterogeneous bunch: persons (Barack Obama), objects (Planet Earth), events (the Big Bang), processes (the rise of the Roman Empire), and instantiations of properties and relations (the independence of mass and acceleration) all qualify. And, of course, this list remains incomplete. That said, there are a couple of features that phenomena share. First, phenomena are real in a relevant sense. Obama and Planet Earth are real. The Big Bang is real, as are the rise of the Roman Empire and the independence of mass and acceleration. This is why they may count as phenomena. In contrast, Obama’s twin sister and the twenty-eighth



101

planet of our solar system are not real, and neither are Creatio ex nihilo, the rise of the Finnish Empire in the twentieth century, and the dependence of mass and acceleration. That’s why they do not count as phenomena. Second, phenomena come with structure. More specifically, they have internal structure or they are essentially part of a broader structure. Planet Earth has internal structure: it consists of an inner core, an outer core, a mantle, and so on. The number 1 may not have internal structure. However, it is essentially part of a broader structure, i.e. the natural number structure. Doesn’t the metaphysical heterogeneity of phenomena render a uniform account of the aim of inquiry into them problematic? Shouldn’t we offer separate accounts for different kinds of phenomena? Fortunately, there is reason to think that this won’t be necessary. After all, for every phenomenon, no matter what its metaphysical nature might be, there is a set of true propositions that describes it. Structures help regiment the set of true propositions describing a phenomenon. It is true propositions about the structural relations between its elements and true propositions about intrinsic properties of the structure of the phenomenon that matter. Let’s call the totality of true propositions describing a phenomenon its ‘full account’. It is the full account of a general phenomenon that is the target of an inquiry into it. While phenomena are metaphysically heterogeneous, each phenomenon has a full account. As a result, a uniform account of the aim of inquiry into general phenomena is still very much a live option.¹ ¹ An anonymous referee asked me whether phenomena can be analysed in terms of questions. Perhaps. For instance, it might be that a phenomenon can be analysed in terms of its full account and that the full account in turn, can be analysed in terms of a set of whether questions. Alternatively, it might be that a phenomenon, X, can be analysed in terms of a question like ‘What’s the whole truth about X?’ Again, the answer to this question will lead us to the full account of X. So, we might also be able to analyse a phenomenon in terms of its full account and the full account, in turn, in terms of a question about the whole truth about the phenomenon. All of this is in principle compatible with the view I develop here. One potential obstacle for an analysis of phenomena in terms of questions lies with the claim that phenomena have structure. Consider the view that analyses phenomena in terms of a set of whether questions. Which propositions go into this set and which ones don’t? If the analysis of phenomena ends with the relevant set of whether questions, it is hard to see how one could answer this question. Now, perhaps all this shows is that an analysis in terms of a question like ‘What’s the whole truth about . . . ?’ is more promising. One issue with this proposal is that it appears to be so very inclusive. On the face of it, the whole truth about the Big Bang, for instance, includes propositions such as that it is your favourite topic of conversation, that some

102 , ,  

1.2 Just More Specific Questions? Given that this is so, it is tempting to think of inquiry into a general phenomenon as a series of inquiries into specific questions. After all, each true proposition corresponds to the true answer to a whether question, i.e. the target of an inquiry into a specific question. It is also tempting to think of the aim of inquiry into a general phenomenon as properly closing all of the specific whether questions, the true answers to which correspond to the full account of the general phenomenon. If this is correct, then there is nothing special about inquiry into general phenomena. It is just a collection of inquiries into specific questions. But, of course, if inquiry into general phenomena is just a collection of inquiries into specific questions, then the aim of inquiry into a general phenomenon coincides with the attainment of all the aims of the inquiries into specific questions in this collection. In other words, the aim of inquiry into a general phenomenon consists in properly closing all the specific questions that correspond to the true answers in its full account. And given that knowledge is the aim of inquiry into specific questions, the result that we get is that the aim of inquiry into general phenomena is knowledge of its full account. What’s more, given Understanding Aim, it looks as though we get a straightforward account of understanding of general phenomena in terms of knowledge. I agree that this is an attractively simple account of the aim of inquiry into and understanding of general phenomena. And, in fact, there is more to be said in its favour. After all, it fits the way in which we typically inquire into general phenomena and the way in which we typically aim to acquire understanding about general phenomena. To see this, consider an example. You want to understand the rise of the Roman Empire people don’t believe in it, that I was thinking about it on Easter Monday 2019, as well as disjunctions with infinitely many disjuncts, including one of the propositions just mentioned. But it would be a stretch to say that those propositions are part of our ordinary understanding of the phenomenon of the Big Bang. Given that this is so, there is reason to think that we need further restrictions on what counts as a phenomenon. The claim about structure provides just such a restriction, and a promising one at that. Perhaps one can get the right restriction out of the relevant ‘What the whole truth about . . . ?’ questions. If so, my account is compatible with an analysis of phenomena in terms of questions (though if the correct restriction differs from my proposal about structure, that proposal will have to be revised accordingly). If not, there is reason to think that the attempt to analyse phenomena in terms of questions is in trouble.



103

and launch an inquiry into this phenomenon. How are you going to proceed? You might buy a relevant history book or find an expert on the topic. What happens as you read through the book or as the expert tells you more about the rise of the Roman Empire? The answer is that you close more and more questions about the phenomenon and, if all goes well, properly so. In this way, that inquiry into a general phenomenon can be understood as the activity of settling a series of questions about it with the ultimate aim of properly closing all of them fits the way in which inquiry into general phenomena typically proceeds. I do not mean to deny that the simple account has considerable attractions. At the same time, the truth of the simple account threatens to be disastrous through and through, at least for present purposes. To see this, recall that my aim is to offer a network analysis which connects inquiry into general phenomena and understanding general phenomena in the same non-reductive way as the network analysis in Chapter 2 connected inquiry into specific questions and knowledge. The reason the simple account threatens to be disastrous is that it is most naturally understood as offering a reductive analysis of inquiry into general phenomena in terms inquiry into specific questions and, as a result, a reductive analysis of understanding in terms of knowledge.² How, if at all, can this disaster be averted? In a nutshell, the answer is that, perhaps somewhat surprisingly, there is reason to think that the simple account is mistaken. To be clear, I do not mean to say that the aim of inquiry into general phenomena doesn’t feature knowledge of its full account. On the contrary, I take the above considerations in favour of the simple account to show that it does.³ Rather, the thought is that the simple account is too weak. The aim of inquiry into a general phenomenon features more than just knowledge of its full account. To see why, consider the following case:

² Note that this is a view that knowledge firsters might be sympathetic to. ³ In addition, the claim that the aim of inquiry into general a phenomenon requires knowledge of its full account is independently plausible. In fact, this claim can be established by arguments parallel to the ones that show that the aim of inquiry into specific is knowledge (see Appendix 1 for details).

104 , ,   The Scatterbrain You want to find out about phenomenon P. To achieve this, you offer a reward for anyone who gives you information about P. You subsequently receive all the information about P and come to know the full account of P. However, each proposition you know you know on the basis of testimony from a different person. Many of the pieces of information you receive are in different languages (all of which you speak).

Have you attained the aim of your inquiry into P? If all that’s required of you is that you know the full account of P, then the answer is yes. Crucially, however, there is reason to think that this cannot be quite right. To see why, note that you believe each proposition of the full account of P separately, as it were, on the basis of the testimony of a different testifier. As a result, you can, of course, answer any question about P. Crucially, many of your answers are bound to remain unsatisfactory in an important respect. Suppose that I ask you whether q is true, and you offer the correct answer. Suppose, next, that I ask you why you would think that. In your current predicament, the only answer you can give is that a certain testifier told you so. Somewhat irked by this, you go on to consider a number of further things that you know about P and ask yourself why you would think that they are true. Unfortunately, again and again, the only thing you can come up with is that some person or other told you so. Recall that you wanted to find out about P. In other words, you had a desire to find out about P. Now, here is the crucial question: in your predicament, is your desire satisfied? Note that if knowledge of the full account of P is the aim of inquiry into P, then the answer to this question has to be yes. After all, you do know the full account of P. At the same time, it seems plausible that the answer does not have to be yes. The very fact that you are unable to justify your various beliefs about P beyond pointing to the fact that you were told so by some testifier may rightly leave you dissatisfied. You may rightly feel that the job is not done yet. If so, of course, knowledge of the full account cannot be all there is to the aim of inquiry into general phenomena. Something is missing. But what? Here is a promising thought: although you are in possession of all the information about P, you haven’t connected the dots yet. That is to say,



105

you know everything there is to know about P but your underlying beliefs are not connected in the right way. For instance, say that q is a fact about P and that q is true because of r. Since you know everything there is to know about P, you know that q, you know that r, and you know that q because r. However, you believe all of these propositions on the basis of (and only of) the say-so of some testifier. That’s why, when asked why you would think that q is true, all you can do is refer to the testifier. In particular, your belief that q is true is not based on your beliefs that r is true and that r causes q, say. Accordingly, my proposal is that, in addition to knowing the full account of a phenomenon, the aim of inquiry into general phenomena also requires the knowledge to be maximally well connected in the following sense: an agent’s knowledge of the full account of some phenomenon, P, is maximally well connected when the basing relations that obtain between the agent’s beliefs about P reflect the agent’s knowledge about the explanatory and support relations that obtain between the members of the full account of P. For instance, in the above case, when you know the full account of P, you will know that q, that r, and that r caused q. In order to have maximally well-connected knowledge, you must additionally believe q on the basis (at least in part) of the fact that r is true and that r caused q. With these points in play, here is my account of the aim of inquiry into general phenomena: Maximally Systematic Knowledge Aim One’s inquiry into phenomenon P aims at having maximally comprehensive and well-connected knowledge of the full account of P (henceforth also ‘maximally systematic knowledge of P’ for short).

This is how inquiry into a given general phenomenon requires more than knowledge of the full account of that phenomenon: one’s knowledge must also be connected in the right way.⁴

⁴ Note that this view of the aim of inquiry is compatible with the idea that phenomena can be analysed in terms of questions (see n.1). After all, it may be that the phenomenon is nothing but a set of questions, but to attain the aim of inquiry into it we must do more than simply know all the answers (i.e. connect the dots in the right way).

106 , ,   Might it be that the aim of inquiry into general phenomena requires even more than maximally systematic knowledge of it? Again, that seems implausible. In fact, the considerations from the discussion of inquiry into specific questions carry right over. Let’s first look at the obvious candidates: higher-order knowledge, understanding why, and certainty. First, there is no reason to think that agents who do not operate a concept of knowledge could ipso facto not attain the aim of inquiry into any phenomenon. But then higher-order knowledge cannot plausibly be what is required here. Second, note that the full account of a given phenomenon will include some propositions that aren’t answers to why questions. If so, understanding why cannot be the attitude that is required for reaching the aim of inquiry into general phenomena (more on this in Section 4.2 below). Finally, third, it’s far from clear that someone who has maximally systematic knowledge of some phenomenon but isn’t absolutely certain about every proposition is ipso facto disqualified from attaining the aim of inquiry into this phenomenon. So, certainty isn’t a plausible candidate either. What about less obvious candidates such as knowledge+? Again, it’s implausible that this should be required for reaching the aim of inquiry into general phenomena if only because most of us don’t have the relevant concepts (e.g. most of us don’t have the concept of knowledge +) and so don’t have the conceptual resources to evaluate beliefs in terms of these concepts. Again, this is not to say that there might not be other candidates that I am overlooking. But, then, the best I can do is still to leave it to the opposition to put their proposals on the table.

2. A Network Analysis 2.1 Inquiry, Understanding, and Views Of course, given Understanding Aim, these results about the aim of inquiry into general phenomena have implications for the debate about understanding phenomena. After all, according to Understanding Aim, the aim of inquiry into general phenomena is understanding.



107

Accordingly, it might be thought that understanding a general phenomenon requires maximally systematic knowledge about it. While this is perhaps the most straightforward way of connecting the results concerning the aim of inquiry into general phenomena with Understanding Aim, there is reason to think that it can’t be quite right. After all, it is plausible that understanding of general phenomena is frequently within reach even for finite human beings, whereas maximally systematic knowledge of them isn’t. A more promising way of effecting the connection is to identify the ultimate aim of inquiry with the highest degree of understanding: Understanding Aim* One’s inquiry into general phenomenon P ultimately aims at maximal understanding of P.

But, of course, if Understanding Aim* holds and if the ultimate aim of inquiry into general phenomena is identified with maximally systematic knowledge of P, then we get the following account of maximal understanding: Maximal Understanding One has maximal understanding of a phenomenon, P, if and only if one has maximally systematic knowledge of P.

Maximal Understanding states necessary and sufficient conditions for maximal understanding. The question that immediately arises is whether Maximal Understanding serves as part of a dismantling analysis of understanding. Unsurprisingly, I want to say that the answer is once again no. Let me explain. To begin with, it will be useful to make a slight detour. Recall that I pointed out that we human beings are finite and, as a result, rarely, if ever, in the ballpark for maximal understanding of general phenomena. At the same time, we frequently attribute understanding of general phenomena to each other. It would be nice if we could make sense of this practice. Moreover, it would be nice if we could make sense of this

108 , ,   practice such that some of these attributions can come out true. How can this be done? The answer I’d like to propose combines an account of degrees of understanding with a contextualist semantics for outright attributions of understanding. Given that Maximal Understanding gives us the maximum degree of understanding, it will come as no surprise that my account of degrees of understanding measures degrees of understanding in terms of approximations to the maximum degree: Degrees of Understanding Degree of understanding of P is a function of distance from maximally systematic knowledge of P: the closer one approximates maximally systematic knowledge of P, the higher one’s degree of understanding of P.

Regarding the contextualist semantics for outright attributions of understanding, the key idea of my proposal is that an attribution of outright understanding of P is true just in case one approximates maximally systematic knowledge of P closely enough. How close is close enough? As is commonly the case with gradable expressions, the answer is that this depends on contextual factors. In other words, context determines a threshold of distance from maximally systematic knowledge that one must surpass for outright understanding to be truly attributable to one: Outright Understanding, Generic ‘S understands P’ is true in context c if and only if S approximates maximally systematic knowledge of P closely enough to surpass a threshold on degrees of understanding determined by c.

Many questions remain about both Degrees of Understanding and Outright Understanding. I will return to at least some of them in a moment. For now, I’d like to focus on the features of Degrees of Understanding that support the idea of a network analysis. How do we measure degrees of understanding of P, when one doesn’t have maximal understanding? Whatever the full answer, I want to suggest that we will



109

have to have a look at what I will call one’s view of P. One way of homing in on the property of a view of P is that one’s view of P stands to understanding of P as one’s belief on whether p stands to knowledge whether p. Belief is the result of a kind of move in inquiry into the specific questions. Similarly, a view is the result of a kind of move in inquiry into the general phenomena. More specifically, one’s view of P is a function of the set questions about P that are closed for one and the connections one has effected between them. To assess how closely one approximates maximally systematic knowledge of P, we must look at one’s view of P and ask how closely it approximates maximally systematic knowledge of P. The answer to this question will give us one’s degree of understanding. With this point in play, it is easy to see why the prospects for a dismantling analysis of understanding of general phenomena are no brighter than in the case of knowledge. Recall that, in the case of knowledge, a dismantling analysis is out of the question because it would have to proceed from inquiry to question-settling to knowledge to belief and back to inquiry. Likewise, what is becoming clear now is that the same goes, mutatis mutandis, for understanding. Here a dismantling analysis would have to proceed from inquiry to understanding to systematic knowledge to views and back to inquiry. In both cases, any attempt at a dismantling analysis will be spoiled by a violation of Priority due to circularity.

2.2 More on Degrees and Outright Understanding With the network up and running, let’s return to Degrees of Understanding and Outright Understanding and look at some of the most obvious questions that arise about them. I’ll start with Degrees of Understanding. I said that to measure degrees of understanding of P we must look at one’s view of P and ask how closely one’s view of P approximates fully systematic knowledge of P. While this is, of course, a start, it leaves open many questions of detail. For instance, one concerns whether how closely one’s view approximates maximally systematic knowledge of a

110 , ,   phenomenon is a function only of what one knows about P and how well one’s knowledge is connected; or whether parts of one’s view that fall short of knowledge can make a difference also. More specifically, one may wonder whether false or unjustified beliefs about P can negatively affect one’s degree of understanding of P. And one may also ask whether justified beliefs or at least justified true beliefs that fall short of knowledge can positively affect one’s degree of understanding of P. While all of the above are fascinating questions, I will not stop to address them here. Rather, I will content myself with one further clarificatory remark: to keep things simple, I will work on the assumption that one’s degree of understanding of P is a function only of what one knows about P and how well one’s knowledge is connected. This simplifying assumption should not raise too many eyebrows. After all, a lot of criticism that knowledge-based accounts of understanding encounter is that they are too strong: understanding doesn’t require knowledge. I will address the most prominent objections in this camp in due course. For the present, note that, if anything, I am making my life harder by adopting the simplifying assumption. If the simplifying assumption is dropped and we allow that justified true beliefs or perhaps even justified false beliefs can get one closer to maximally systematic knowledge, we gain additional resources for dealing with some of the relevant objections. I don’t believe that I need those resources here. However, it is worth keeping in mind that they are in store if needed. Here is another question one may ask about Degrees of Understanding. It is widely agreed that degrees of understanding can themselves be measured along at least two dimensions, i.e. breadth and depth. How can Degrees of Understanding make sense of this? In response, it may be tempting to identify degree of breadth of understanding with degree of comprehensiveness of knowledge and degree of depth with degree of well-connectedness. However, it is likely that things are more complicated, especially when it comes to depth. Here things like knowledge of detail will matter as well, which means that it is likely that degree depth of understanding is going to be a function of degree of both comprehensiveness and well-connectedness of knowledge. Regarding Outright Understanding, the perhaps most obvious question that remains is how exactly the threshold is fixed at a given context.



111

This is a vexed question. Fortunately, it is not of central importance for present purposes. That said, here is my preferred answer: context supplies a task concerning the phenomenon such that one is close enough to maximally systematic knowledge of a given phenomenon just in case one would successfully perform this task, if one were to have the relevant skills needed to do this. In other words: Outright Understanding, Specific ‘S understands P’ is true in context c if and only if S approximates maximally systematic knowledge of P closely enough to be such that S would (be sufficiently likely to) successfully perform any task concerning P determined by c, if, in addition, S were to have the skills needed to do so and to exercise them in suitably favourable conditions.⁵

While I am optimistic about the prospects of this account, I don’t want to hang too much on it. Fortunately, for present purposes, I don’t need to. After all, my ambition here is not to provide a fully worked-out account

⁵ It might be thought that the subjunctive conditional here is problematic. After all, as Robert Shope (1978) has argued at length, biconditionals featuring a subjunctive are prone to what he calls ‘the conditional fallacy’. In the simplest case, the biconditional is of the form ‘p
(q □! r)’. In one version of the fallacy (V1), p is true and q is not, but if q were true, q’s being true (and/or r’s being true) would lead to p’s no longer being true. In another version of the fallacy (V2), p is true and q is not, but if q were to be true, this would lead to r’s no longer being true. Fortunately, there is reason to think that Outright Understanding, Specific does not fall prey to the conditional fallacy. Concerning V1, note that we can add further conditions to the antecedent of the subjunctive such that they entail that the left-hand side of the biconditional is true. A biconditional of the form ‘p
((p & q) □! r)’ will evidently be safe from V1. In the case of Outright Understanding, Specific, we can require that the context remains the same and S approximates maximally systematic knowledge to the same degree and in the same way. Since whether or not an agent surpasses the threshold depends only on the context and the degree and way of approximation to maximally systematic knowledge, adding the above conditions to the antecedent guarantees that if the antecedent of the subjunctive is true, the left-hand side of Outright Understanding, Specific is true also. Concerning V2, note that if the corresponding conditional is necessarily true, again there is no need to worry: ‘p
((p & q) □! r)’ will evidently be safe from V2 when ‘p
□((p & q) ! r)’ is true also. In Outright Understanding, Specific, the addition of the proviso that conditions be suitably favourable arguably ensures that the corresponding conditional is necessarily true. That said, it is in principle possible to avoid stating Outright Understanding, Specific in terms of a subjunctive conditional. One alternative strategy appeals to the notion of an epistemic duplicate and claims that ‘S understands P’ is true in a context just in case some epistemic duplicate of S would successfully perform the tasks determined by context. (Thanks to Catherine Elgin for pointing this out to me.) I decided to opt for the subjunctive version because it strikes me as most intuitive and more elegant certainly than the ‘epistemic duplicate’ version.

112 , ,   of understanding but rather to say enough for the epistemological purposes of the present project.

3. Objections With my account of understanding in play, I’d now like to look at some objections to it. To home in on where the action will be, note that my account is knowledge-based in the sense that it takes understanding to require knowledge. Note also that this is not a lightweight aspect of my account but rather a substantive commitment. At the same time, there is no shortage of objections to knowledge-based accounts of understanding in the literature. In what follows, I will look at the most prominent ones.

3.1 Gettiered Understanding The first objection to knowledge-based accounts of understanding is due to Kvanvig (2003), who argues that understanding, but not knowledge, is compatible with (a certain form of) gettierization. To bring this point home, Kvanvig invites us to consider the following case: Comanche Consider, say, someone’s historical understanding of the Comanche dominance of the southern plains of North America from the late seventeenth until the late nineteenth centuries. Suppose that if you asked this person any question about this matter, she would answer correctly. Assume further that the person is answering from stored information; she is not guessing or making up answers, but is honestly averring what she confidently believes the truth to be. Such an ability is surely constitutive of understanding, and the experience of query and answer, if sustained for a long enough period of time, would generate convincing evidence that the person in question understood the phenomenon of Comanche dominance of the southern plains. (Kvanvig 2003, 197–8)



113

Crucially, Kvanvig points out that while the history aficionado’s beliefs will normally also qualify as knowledge, they need not. The case can be set up as a Gettier case and so the history aficionado’s beliefs are only luckily true. Here’s how Kvanvig ventures to achieve this: For example, most history books might have been mistaken, with only the correct ones being the sources of the understanding in question and with no basis in the subject for preferring the sources consulted over those ignored. (Kvanvig 2003, 198)

The thought here is that the case is relevantly analogous to Fake Barns. (Recall that, in Fake Barns, the agent acquires a true belief that they are facing a barn whilst driving alongside a field that otherwise contains only cleverly constructed fakes.) For that reason, the history aficionado understands Comanche dominance, even though the beliefs that constitute their understanding are gettiered and hence do not qualify as knowledge. One problem I have with this case is that it is actually not clear that Fake Barns is the correct model for it. After all, in Fake Barns, the agent (i) comes to truly believe one proposition, to wit, that they are looking at a barn. Moreover, (ii) they might easily have believed the same proposition, but (iii) that belief would have been false. In contrast, in the Comanche case, the history aficionado (i) acquires a body of interconnected true beliefs about Comanche dominance. What might easily have happened is (ii) that they might have come by a different body of interconnected beliefs. Finally, (iii) while some members of the body of interconnected beliefs the history aficionado might have arrived at, including some central ones, would have been false, it is far from clear that all of them would have been false (or even unknown).⁶ Here is a more adequate model for the Comanche case: someone selects a particular school for their daughter on the ground that it is housed in a nicer building than all the other schools in the neighbourhood. Suppose, furthermore, the school selected happens to be the only ⁶ In fact, I am not the only one who has this concern. See Khalifa (2013b) for a similar complaint.

114 , ,   school in the neighbourhood that teaches evolutionary theory instead of creationism. Note that this case parallels the Comanche case closely, much more closely than the Fake Barns case. After all, just as in the Comanche case, in this case, (i) the daughter acquires a body of interconnected true beliefs about a certain phenomenon, here the origin of species. What’s more, (ii) she might easily have come by a different body of interconnected beliefs about it, that is, if she had been sent to a different school. Finally, (iii) while some members of the body of interconnected beliefs, including central ones, would have been false, it is far from clear that all of them would have been false (or even unknown). It comes to light that the Comanche case parallels the school case much more closely than the Fake Barn case. Now, the crucial question is whether, in the school case, the daughter’s beliefs about the origin of species qualify as knowledge. In my view, the answer here is yes. Surely, the daughter can come to know things she learns at school about the origin of species, even if different schools she might easily have attended instead would have taught her falsehoods. If so, Kvanvig has failed to make a compelling case that, in the Comanche case, the beliefs about Comanche dominance the history aficionado arrives at fall short of knowledge. On the contrary, if anything, there is reason to think, pace Kvanvig, that the history aficionado’s beliefs about Comanche dominance do qualify as knowledge, even if different books they might easily have read instead would have led them to believe falsehoods. As a result, Kvanvig’s case fails to make a compelling case against knowledge-based accounts of understanding. That said, it may be worth noting that the present view has the resources to provide yet another account of the Comanche case, one that can concede that many of the history aficionado’s beliefs in the Comanche case fall short of knowledge. Recall that, according to Outright Understanding, attributions of understanding are given a contextualist semantics. Roughly, one understands a phenomenon just in case one knows enough about it to surpass a contextually determined threshold. Notice next that even if, in the Comanche case, some of the history aficionado’s beliefs about Comanche dominance are gettiered, others may qualify as knowledge, including (i) what the relevant theory about Comanche dominance says and (ii) that it is a highly plausible theory. Now, this knowledge might be enough in the sense required by Outright



115

Understanding if the contextually determined threshold is sufficiently low. Consider, for instance, a context in which all we are looking for is someone who will correctly answer a number of questions about Comanche dominance. In such a context, it would seem that someone who, like the history aficionado, knows what the correct theory about Comanche dominance says and knows it to be highly plausible, will know enough to meet our needs.⁷ Given that this is so, it is also plausible that the threshold operative in this context should be low enough to allow us to truly attribute to this agent understanding of the phenomenon. But now notice that the context just described fits the context that Kvanvig sets up like a glove. After all, what Kvanvig takes to be the crucial evidence that the history aficionado understands Comanche dominance is that the history aficionado is such that ‘if you asked [her] any question about this matter, she would answer correctly’ (Kvanvig 2003, 197). In consequence, the present view may be able to secure the result that, in the relevant context, understanding of Comanche dominance can truly be attributed to the history aficionado even if many of their beliefs about the issue are gettiered.

3.2 Models, Idealizations, and Thought Experiments in Science A second set of objections to knowledge-based accounts venture to show that, as opposed to knowledge, understanding is not factive. That is to say, one can come to understand something even though some of the beliefs that constitute one’s understanding are false. Elgin presents a number of arguments along these lines, one of which concerns the role of idealizations, models, and thought experiments in science. She considers a variety of examples including the ideal gas law: The ideal gas law, for example, accounts for the behaviour of gases by characterizing the behaviour of a gas composed of dimensionless, ⁷ We might have to add that this person is not aware of any alternative theories that they believe to be equally or almost equally plausible. Otherwise, they might not be prepared to answer in accordance with what the theory says. Notice, however, that the history aficionado also satisfies this additional requirement.

116 , ,   spherical molecules that are not subject to friction and exhibit no intermolecular attraction. There is no such gas. Indeed, there could be no such gas. Nonetheless, scientists purport to understand the behaviour of actual gases by reference to the ideal gas law. (Elgin 2009, 326)

As the case of the ideal gas law illustrates, idealizations, simplified models, and thought experiments play a central role in scientific theorizing and constitute a crucial part of our understanding of the subject matter. Our understanding of the behaviour of gases rests crucially on the ideal gas law. At the same time, these idealizations, models, and thought experiments do not describe any part of the natural world and so are not literally true. The ideal gas law is a case in point: the natural world features no gases with the properties the ideal gas law assumes gases to have. It is, therefore, not literally true. It goes without saying that if idealizations, models, and thought experiments are not literally true, then no one can know them to be literally true. At the same time, it is simply not part of the present view that a proposition or theory can contribute to our understanding of various scientific phenomena only if it is known to be literally true. Instead, by the lights of the present view, it will so contribute if it allows us to make progress in the direction of maximally systematic knowledge of the phenomena. And there is excellent reason to think that idealizations, models, and thought experiments do this much for us. After all, we can still know what (theories including) these idealizations, models, and thought experiments say. For instance, we can know what a theory including the ideal gas law says. We can also know that actual scientific entities approximate the idealizations, etc. in various respects and to various degrees. For instance, we can know that actual gases approximate ‘ideal gases’ in various respects. Finally, we can know the limitations of idealizations, etc. Since this knowledge advances us in the direction of maximally systematic knowledge of the relevant phenomena—in the case of the ideal gas law, of the physics of gases—by the present view, it constitutes part of our understanding of them. It thus comes to light that while the fact that idealizations, etc. are not true of any actual scientific entities may mean that they cannot be part of our understanding of the



117

relevant scientific phenomena in the most straightforward way, it does not follow that they cannot constitute part of our relevant understanding at all.

3.3 Scientific Progress Another argument against knowledge-based accounts of understanding, also due to Elgin (2009, 325–6), proceeds along the following lines: scientific understanding progresses from the crude to the sophisticated. For instance, the first seed of a great evolutionary biologist’s understanding of the evolution of humankind may consist in a belief that human beings evolved from apes. Throughout their education their beliefs about the evolution of humankind became refined. The crude belief was replaced by a belief that humans and apes had a common hominid ancestor. It also became embedded in a web of beliefs about evolution that make them the expert they are now. On a social level, our contemporary understanding of astronomy developed from Ptolemy’s theory of the movement of the planets, via Kepler, Newton, and relativity theory to string theory. Elgin claims that developments like these may manifest and often do manifest genuine cognitive progress, genuine progress in understanding. Throughout the various stages of the development, our understanding of the subject matter—e.g. of the origin of species or astronomy—is embodied by the theory of the subject matter we accept at that stage. Crucially, however, not only the crude beginnings of scientific theorizing on a subject matter, but also even the best theories science has produced thus far are all false. In consequence, on any account on which understanding is factive, the sophisticated are on par with the unsophisticated: neither has achieved understanding. It is not hard to see that Elgin’s objection will work only on the assumption that, according to knowledge-based accounts of understanding, there can be progress in scientific understanding of a phenomenon as we move from one theory to another only if at least the successor theory is (known to be) true. Again, however, the present view is simply not committed to this assumption. On the present view, progress in

118 , ,   scientific understanding of a phenomenon will occur if we make progress towards maximally systematic knowledge of the phenomenon in question. Moving from one false theory to another may do the trick. In fact, Elgin’s own case of moving to a theory including the ideal gas law plausibly constitutes a case in point here. Even if we grant that, just like its predecessor, this theory is false, it may nonetheless serve to increase our knowledge about the physics of gases, as we have just seen. Again, just because the contribution to progress is not made in the most straightforward way, it does not follow that no contribution to progress is made at all.⁸

3.4 Understanding via Incompatible Theories The last objection I will discuss here is due to Zagzebski, who argues that incompatible theories can give their champions understanding of some phenomena. Here is Zagzebski: More than one alternative theory may give understanding of the same subject matter. This makes sense if we think of a theory as a representation of reality, where alternative representations can be better or worse, more or less accurate. But more than one may be equally good, equally accurate. This form of understanding does not presuppose knowledge or even true belief, and if we assume that two competing representations of the same part of reality cannot both constitute knowledge, it cannot be a form of knowledge. (Zagzebski 2001, 244)

Just as Elgin before, Zagzebski presupposes that, according to knowledge-based accounts, champions of a theory can attain understanding of the underlying phenomenon only if they know the theory to be true. As I have already pointed out, the present view is not ⁸ It may be worth noting that, in an insightful paper, Alexander Bird (2007b) forcefully argues that scientific progress must be understood in terms of accumulation of knowledge. If successful, Bird’s argument turns the tables on those who think that knowledge-based accounts of understanding have difficulties in explaining progress in science.



119

committed to this kind of claim. In fact, it is not hard to see that the present view can allow that more than one alternative theory can deliver an equal degree of understanding. After all, it is possible for adherents of two (or more) distinct theories to be equidistant from maximally systematic knowledge of a certain phenomenon.⁹ In that case, the present view predicts that they have the same degree of understanding of the phenomenon.

4. The Competition I’d like to start my inquiry into how my view of understanding compares with the closest and most prominent competition by looking at the Received View of understanding: The Received View One understands why p if and only if one knows why p (i.e. for some q, one knows that q explains why p).¹⁰

Obviously, the Received View is quite different from the view I have outlined above. In fact, it might be thought that it is so different as to not be a competitor at all. To see why one might think this, consider a widely recognized distinction between two types of understanding: objectual understanding and explanatory understanding. The two kinds of understanding differ in that they have different objects. Objectual understanding takes objects (or phenomena) as objects. Explanatory understanding takes propositions as objects. More specifically, propositions that qualify as correct answers to why questions. In other words, explanatory understanding is understanding why. With these two types of understanding in play, it is easy to see why one might think that the Received View isn’t a competitor for my preferred ⁹ Furthermore, two agents may be equidistant from maximally systematic knowledge of a phenomenon because they are approximating such knowledge via different routes (e.g. breadth vs depth). ¹⁰ Advocates of the received view include Hempel (1965), Lipton (2004), Kitcher (1989), and Salmon (1989).

120 , ,   view. The Received View is a view about explanatory understanding and my view is a view about objectual understanding. Crucially, however, there is a further thesis that promises to bring the two views into competition. The key idea of this view is that explanatory understanding is all that matters. Objectual understanding isn’t a particularly interesting philosophical phenomenon. Kareem Khalifa characterizes the view nicely in the following passage: Explanationism Nothing of philosophical importance is lost if all instances of objectual understanding are treated as instances of explanatory understanding. (Khalifa 2017, 85)

Let us call the view that combines the Received View and Explanationism ‘Classical Explanationism’. It is easy to see that Classical Explanationism is in competition with my own view. According to Classical Explanationism, knowledge of explanations is all that matters philosophically for objectual understanding. The same does not hold for my own view. There are a number of straightforward problems with both components of Classical Explanationism. In what follows, I will look at some of these as well as at a more sophisticated development of them, due to Khalifa, which is a more serious competitor for my own view.

4.1 Explanatory Understanding The Received View. The Received View identifies explanatory understanding with knowledge of explanations. One of the core problems for this view is that it is arguably too strong. For, as Peter Lipton argues in a paper aptly entitled ‘Understanding without Explanation’ (Lipton 2009), it is possible to acquire understanding of some phenomenon without acquiring an explanation of it. In what follows, I will briefly outline Lipton’s argument. Lipton first distinguishes between explanations, on the one hand, and the cognitive benefits of explanations, on the other. He then identifies



121

understanding a phenomenon with having the cognitive benefits of an explanation of it and distinguishes between four types of cognitive benefits an explanation may offer: knowledge of (i) causal information, (ii) necessity, (iii) possibility, and (iv) unification (Lipton 2009, 43–4). In order to show that one can have understanding without explanations, he argues that one can come to enjoy these benefits of explanation without at the same time coming to have an explanation of the relevant phenomenon. More specifically, he adduces four cases, in each of which the agent comes by one of the above-mentioned types of benefits without at the same time acquiring an explanation of the phenomenon at issue. For present purposes, I would like to focus on one of Lipton’s cases, which involves knowledge of necessity without explanation. Lipton considers an agent who comes to know that it is impossible for gravitational acceleration to be dependent on mass by appreciating Galileo’s famous thought experiment in which a heavy and a light mass are tied together. Intuitively, the agent comes to understand why gravitational acceleration is independent of mass here.¹¹ At the same time, the agent does not acquire knowledge of an explanation of why this must be so. The reason Lipton gives for this is that an explanation would have to offer a ‘direct answer’ to the question why acceleration is independent of mass, which the thought experiment does not provide. More specifically, Lipton suggests that ‘to explain by showing necessity requires a kind of constructive argument that not all proofs or thought experiments supply.’¹² It is not hard to see that it is precisely constructive arguments that will enable their possessors to give direct answers to the relevant questions that are required by explanation. Crucially, however, arguments by reductio ad absurdum do not qualify as constructive in the relevant sense. Since Galileo’s thought experiment offers at best a reductio ad

¹¹ Though see Khalifa (2017, ch. 5) for a competing account according to which Galileo’s thought experiment establishes not the independence of mass and acceleration but the independence of our concepts of mass and acceleration. ¹² Lipton (2009, 47). I take it that Lipton’s theses (i) that explanations must offer direct answers to the relevant why questions and (ii) that explanations by showing necessity require constructive arguments constitute constraints that any satisfactory account of explanation (by showing necessity) will have to satisfy. As a result, his argument will go through on any viable account of explanation.

122 , ,   absurdum of the proposition that mass and acceleration are not independent, it does not serve to explain why they are independent.¹³ The upshot of Lipton’s argument for the Received View is obvious. Given that one can have understanding why without knowing an explanation of what one understands, the Received View is too demanding. More specifically, it seems plausible that the problem the view faces here is one of over-intellectualization. The intellectual demands placed on understanding—knowledge of an explanation—are unrealistically high. Less sophisticated cognitive achievements can qualify as understanding. It is worth noting that my own account promises to avoid this problem. According to Degrees of Understanding, as one acquires more systematic knowledge about a phenomenon, one advances one’s understanding of it. Crucially, that is just what happens when one learns about Galileo’s thought experiment. One acquires systematic knowledge about the relation between gravitational acceleration and mass, thereby advancing one’s understanding of this phenomenon. So, my account predicts correctly that one’s understanding increases even in cases like Lipton’s Galileo case. What about attributions of outright understanding? Again, there is excellent reason to think that Outright Understanding can steer clear of the over-intellectualisation problem. Recall that here the thought is that whether outright understanding can truly be attributed to one depends on whether one knows enough to be able to successfully complete a contextually determined task. Crucially, the contextually determined tasks may be very easy. In certain contexts, for instance, when talking about small children, knowing enough to be able to answer a set of fairly easy multiple-choice questions may do the trick. That’s why there is every reason to think that Outright Understanding can deliver right results when it comes to attributions of outright understanding in Lipton’s cases also.

¹³ Lipton (2009, 47–8). Note, first, that Lipton argues that the situation is analogous in the case of mathematical proofs, which also allows for a distinction between explanatory and nonexplanatory proofs. Here, too, Lipton is attracted by the idea that proofs by reductio are not explanatory. Second, those who are not convinced by this particular case may recall that Lipton offers three other cases of understanding without explanation. To those who remain unmoved by all of them, I’d say that the onus is on them to show why.



123

Khalifa’s Explanation-Knowledge-Science Model. A more sophisticated account of understanding in terms knowledge of explanations has recently been defended by Kareem Khalifa. His ‘ExplanationKnowledge-Science model’ (EKS model) comprises the following two theses: EKS1 S1 understands why p better than S2 if and only if (A) Ceteris paribus, S1 grasps p’s explanatory nexus more completely than S2; or (B) Ceteris paribus, S1’s grasp of p’s explanatory nexus bears greater resemblance to scientific knowledge than S2’s. EKS2 S has minimal understanding of why p if and only if for some q, S believes that q explains why p, and q explains why p is approximately true. (Khalifa 2017, 14) Khalifa offers a careful explanation of various of the relevant terms in EKS. For reasons that will become clear in due course, I will focus on two: ‘scientific knowledge’ and ‘explanatory nexus’. Let’s start with ‘scientific knowledge’. Khalifa offers the following account of scientific knowledge why p: Scientific Knowledge Why S has scientific knowledge that q explains why p if and only if the safety of S’s belief that q explains why p is because of her scientific explanatory evaluation.

Crucially, scientific explanatory evaluation features consideration of potential explanations, comparing them and forming a belief based on the result of the comparison. The other term I want to look at is ‘explanatory nexus’. The reason for this is that it provides an important aid to seeing just how kindred in spirit the EKS-model is to my own account. Khalifa initially defines

124 , ,   ‘explanatory nexus’ as ‘the set of correct explanations of p as well as the relations between those explanations’ (Khalifa 2017, 6), adding that by ‘explanation’ he means what many others would call ‘explanatory information’ such that the following counts as an alternative to the initial definition: ‘the nexus of p is the totality of explanatory information about p’ (Khalifa 2017, 6 n.2). It’s difficult not to notice the similarities between what Khalifa calls an explanatory nexus and between what I call the full account of a phenomenon. The similarity between the two accounts comes out even more clearly once we consider Khalifa’s accounts of maximal (or ‘ideal’) understanding and his account of outright attributions of understanding. Here goes: Very roughly, ideal understanding is maximally scientific knowledge of a complete explanatory nexus. (Khalifa 2017, 15)

‘S understands why p’ is true in context C if and only if S has minimal understanding and S approximates ideal understanding of why p closely enough in C. (Khalifa 2017, 5)

This already looks quite close to the account I sketched above (in Section 2) and defended in earlier work (e.g. Kelp 2015, 2017), especially once we are clear what Khalifa means by ‘explanatory nexus’. Now that we are clear on the similarities, let’s take a closer look at the differences between Khalifa and me. Here is Khalifa once more: I’ve learned much about how to think about degrees of understanding from Kelp (2015). One small difference: Kelp takes maximum/ideal understanding as his starting point and gleans comparative and outright conceptions of understanding from there. These are only methodological points. I briefly cover the substantive differences between our views in Chapter 4. (Khalifa 2017, 4 n.1, my emphasis)



125

Crucially, the differences that Khalifa identifies as ‘the substantive ones’ between us (the ones discussed in his Chapter 4), concern the issue of Explanationism, a more sophisticated version of which Khalifa endorses but which I reject. Let’s take a closer look at these differences, then.

4.2 Explanationism Khalifa wants to combine his EKS-model with a sophisticated version of Explanationism. Why not go for the standard version instead? One important reason here, which Khalifa also notes, is that we seem to be able to get understanding in virtue of acquiring non-explanatory knowledge of classifications, discoveries and, so on (Khalifa 2017). Given that this is so, we will do well to avoid the combination of the EKS-model with standard Explanationism. After all, by EKS2, the EKS-model countenances a minimal threshold on understanding according to which there can be no understanding without belief in approximately correct explanations. If we can get understanding from classifications, discoveries, and so on, it would appear that we can get understanding without even belief in correct explanations. That’s why something more sophisticated is needed. Khalifa’s idea is to go for a view he calls Quasi-Explanationism. Here goes: Quasi-Explanationism Nothing of philosophical importance is lost if all instances of objectual understanding are treated as instances of: (a) explanatory understanding, or (b) being on the right track to having explanatory understanding. (Khalifa 2017, 85) Moverover, Khalifa unpacks what it is to be on the right track to having explanatory understanding in terms of a property of proto-understanding. While understanding is about explanations, proto-understanding is about explanatory roles, which describe components of explanations or, if they

126 , ,   don’t, at least feature in scientific knowledge of a correct explanation (Khalifa 2017, 86–7). I do not mean to deny that Khalifa’s move to Quasi-Explanationism constitutes progress in that it helps him avoid some of the problems that Explanationism encounters. At the same time, I think that even QuasiExplanationism remains unsatisfactory at the end of the day. In a nutshell, the reason for this is that there are phenomena that we can achieve significant degrees of understanding about, enough to make outright attributions of understanding come out true, without acquiring any knowledge or even beliefs about correct explanations or explanatory roles for that matter. What’s more, any such knowledge does not affect our understanding of these phenomena in any way. Here is one example: the layout of my house. Consider me, the owner of the house. I know that there is a bathroom on the right as you enter on the ground floor, a kitchen to the left, the living room straight ahead, that there are three bedrooms on the first floor, two on top of the ground floor bathroom and kitchen, the master bedroom on top of the living room, that the master bedroom has an en suite bathroom, that the main bathroom is between the master bedroom and the two other bedrooms, that there is an attic but no basement, and so on. It is hard to deny that I have a better understanding of the layout of my house than my little daughter, who doesn’t know that there is an attic. And my daughter’s understanding, in turn, is better than my friend’s, who has only ever been on the ground floor and knows nothing about the layout of the first floor or the attic. Finally, my friend’s understanding is better than my postman’s, who has never entered the house at all. What’s more, it is even plausible that outright attributions of understanding to agents who know what I know about the layout of my house come out true in many contexts. For instance, ‘Chris understands the layout of their house’ comes out true in many contexts. Here is another example: the meaning of ‘drake’. I know that ‘drake’ means male duck. This knowledge goes a long way towards understanding the meaning of ‘drake’. In particular, it is plausible that, in virtue of knowing that ‘drake’ means male duck, I understand the meaning of ‘drake’ better than someone who only knows that ‘drake’ denotes some sort of male animal, without knowing which one. And, again, outright



127

attributions of understanding to agents who know that ‘drake’ means male duck come out true in many contexts. ‘Chris understands the meaning of “drake”’ comes out true in many contexts.¹⁴ Crucially, knowledge or even true beliefs about explanations do not enter the picture at all, and neither does knowledge of explanatory roles. I do not need beliefs about the correct answers to questions about why there is a kitchen to the left, why there are three bedrooms on the first floor, why there is a kitchen at all, etc. to attain a significant degree of understanding of the layout of my house, nor do I need this for outright attributions of understanding to come out true in many contexts. On the contrary, beliefs about correct explanations or explanatory roles seem to be irrelevant to understanding here. All of the above agents may or may not know why there is a kitchen to the left, etc. However, this additional knowledge does not seem to affect their understanding of the layout of my house in any way (though, of course, it may affect their understanding of different phenomena such as the quality of the design of the house). And the same holds, mutatis mutandis, for my understanding of the meaning of ‘drake’. If knowledge or even true beliefs about correct explanations or explanatory roles are irrelevant to our understanding of certain phenomena, then even Quasi-Explanationism is in dire straits. A question that remains is why Quasi-Explanationism runs into trouble. One promising ¹⁴ Isn’t this case unfair to Khalifa? After all, Khalifa is quite clear that he takes understanding of word meanings, concepts, and the like to be something quite different from understanding of empirical phenomena and that his account is an account of understanding of empirical phenomena only. Three points by way of response: first, even if successful, this move doesn’t allow Khalifa to sidestep the problem raised by the case of the layout of my house. And that’s all that I really need to cause trouble for Khalifa is one case. Second, Khalifa says that his view ‘should be compatible with whatever turns out to be the best account of these kinds of non-explanatory understanding.’ (2017, 3) I take the ‘should’ here to express a desideratum for his account of understanding. As I will argue below in more detail, there is reason to think that Khalifa’s account doesn’t satisfy this desideratum. Here is the key idea: there is a view (mine), which may very well be the best, that accounts for at least a certain kind of non-explanatory understanding in terms of objectual understanding. However, this is incompatible with the part of Khalifa’s view that nothing of philosophical interest is lost if we eliminate objectual understanding. Third, suppose Khalifa can offer a convincing motivation for bracketing understanding of word meanings, concepts, and the like. Even so, the fact remains that my account does not have to. In fact, it can offer a unified account of understanding of word meanings, concepts, and the like, on the one hand, and empirical phenomena, on the other. As a result, my account compares favourably with Khalifa’s in any case.

128 , ,   answer is that the phenomena understood here—the layout of A’s house and the meaning of ‘drake’—do not have explanatory structure. Rather, what we have here is a case of understanding of phenomena with a different kind of structure. But, of course, any attempt to analyse understanding of phenomena that simply do not have explanatory structure in terms of knowledge or even mere belief about correct explanations or explanatory roles is doomed to failure. That’s why any attempt to rescue Explanationism is ultimately bound to remain unsatisfactory.¹⁵ In contrast, it is easy to see that my account can do better. After all, even though these phenomena don’t have explanatory structure, they have some structure. This structure is described by the full account of these phenomena. Accordingly, we can come to understand them by acquiring systematic knowledge of the relevant full accounts: the more closely we approximate maximally systematic knowledge, the better our understanding of these phenomena. One might worry that this does Khalifa an injustice.¹⁶ To see why, recall, first, that Khalifa’s claim is that nothing of philosophical importance is lost if all instances of objectual understanding are treated in terms of explanatory understanding or being on track to having it. Crucially, second, in cases of objectual understanding without explanatory understanding/being on track to it, we can replace talk of understanding in terms of talk of knowledge without loss. And that’s true of both of the above cases. When we say that I understand the layout of my house, we might as well say that I know the layout of my house. And when we say that I understand the meaning of ‘drake’, we might as well say that I know the meaning of ‘drake’. Nothing of philosophical substance will be lost here. Note that there will still be differences. For instance, on my view utterances of ‘My daughter understands the layout of my house better ¹⁵ Here is yet another difference between Khalifa and me. Khalifa’s project seems to qualify as a dismantling analysis which ventures to unpack understanding in terms of knowledge and knowledge in terms of safe because competent belief. In contrast, abandoning the dismantling approach in favour of a network alternative is at the very heart of my own project. At the same time, Khalifa could embed his view in the kind of non-reductive framework that I favour. That’s why I will set these differences aside here. ¹⁶ Thanks to an anonymous referee for pressing me on this. The remainder of this subsection engages with their very helpful comments.



129

than the postman’ comes out true. Moreover, ‘Chris understands the meaning of “drake”’ comes out true in the vast majority of contexts. In contrast, on Khalifa’s view, while ‘My daughter knows the layout of my house better than the postman’ and ‘Chris knows the meaning of “drake”’ come out true, ‘My daughter understands the layout of my house better than the postman’ and ‘Chris understands the meaning of “drake”’ don’t. And one might wonder whether these differences don’t favour my view over Khalifa’s. After all, isn’t it intuitive that ‘My daughter understands the layout of my house better than the postman’ and ‘Chris understands the meaning of “drake”’ come out true, at least in the right contexts? I expect that Khalifa would deny that these differences are of genuine philosophical significance. At best, the above considerations indicate that Khalifa’s view doesn’t respect how competent speakers of English use the word ‘understanding’. However, what Khalifa is offering is an account of understanding, not a semantics for ‘understanding’. Moreover, there is little reason to think that having a view that respects how competent English speakers use the word ‘understand’ is of philosophical importance. So, the above considerations aren’t all that damning for Khalifa. I agree. In fact, as a stand-alone argument, they don’t carry much weight at all against Khalifa. It is worth keeping in mind, however, that I am not seeking to give a stand-alone argument, nor do I need to. This is because what we are looking at is a comparison between two rival accounts of understanding, Khalifa’s and my own. And in this particular setting, the above considerations can still do some work. After all, given that the two accounts perform equally well on all other fronts and this is the only point of difference, they do provide reason to favour my account. More importantly yet, there are further and more substantive differences. To get them into sharp relief, note first that we may inquire into phenomena with non-explanatory structure. For instance, a potential buyer may inquire into the layout of my house and a non-native speaker may inquire into the meaning of ‘drake’. What’s more, we may make progress and eventually succeed in our inquiries into phenomena with non-explanatory structure. For instance, upon seeing a floor plan of my house, the potential buyer may make progress in their inquiry into its

130 , ,   layout, and when the non-native speaker finds out that ‘drake’ means male duck, they may well have succeeded in their inquiry into the meaning of ‘drake’. Now, here is the key question I want to ask: what is the aim of inquiry into phenomena with non-explanatory structure such as the layout of my house and the meaning of ‘drake’? It will not come as much of a surprise that my answer is: understanding of the target phenomena. On my view, then, inquiry into phenomena with non-explanatory structure has the same aim as inquiry into any other general phenomenon. In this way, my account offers an attractively unified answer to this question. Note also that my view can easily accommodate the possibility of progress and success in inquiries into phenomena with non-explanatory structure. Having had a look at the floor plan, the potential buyer approximates maximally systematic knowledge about the layout of my house more closely than before. As a result, they understand the layout better than before, thus making progress in their inquiry. Similarly, upon learning that ‘drake’ means male duck, the non-native speaker approximates maximally systematic knowledge about the meaning of ‘drake’ more closely than before. In fact, they approximate it closely enough that outright attributions of understanding of the meaning of ‘drake’ come out true in the vast majority of contexts, including the one they found themselves in when embarking on their inquiry. If so, their inquiry is successful. In contrast, the question about the aim of inquiry into general phenomena with non-explanatory structure causes trouble for Khalifa. Since the target phenomena of these inquiries don’t have explanatory structure, according to Quasi-Explanationism, they simply cannot be understood. As a result, Khalifa is faced with the following trilemma: (i) He agrees that inquiry into phenomena with non-explanatory structure aims at understanding these phenomena. In that case, Khalifa cannot accommodate the possibility of attaining the aim of inquiry into these phenomena, nor of making progress, which is a considerable cost for his view. (ii) Inquiry into general phenomena never aims at understanding and always only aims at knowledge (or else true belief, justified belief, etc.) In that case, he will run into trouble with cases like the Scatterbrain (Section 1.2), which suggest otherwise. (iii) Inquiry into phenomena with



131

non-explanatory structure does not aim at understanding, whereas inquiry into phenomena with explanatory structure does. While this option avoids the problems of the two earlier ones, it puts Khalifa at an important disadvantage vis-à-vis my own view. After all, while my view can offer a unified account of the aim of inquiry into general phenomena, Khalifa will have to settle for something that is considerably more gerrymandered. Crucially, this disadvantage is of genuine philosophical significance, not one that can be relegated to the semantics of ‘understanding’. Before moving on, I want to consider a reason that Khalifa offers for thinking that his view is preferable to the kind of view that I favour. In a nutshell, the worry is that the kind of view that I favour makes understanding too easy to come by. Any bit of knowledge about a phenomenon, no matter how irrelevant, will get you closer to maximally systematic knowledge about it. On my view, it would appear that, by the same token, it is bound to constitute an advance in understanding of phenomenon. However, that doesn’t seem to be right. To take just one example, knowing that the kitchen is my favourite room doesn’t seem to advance anyone’s understanding of the layout of my house. In response, recall that, on my view, phenomena have structure and that structures help regiment their full accounts. In particular, it is true propositions about intrinsic properties of the structure that matter. Given that this is so, not every bit of knowledge about a phenomenon will advance one’s understanding of it. In particular, what Khalifa identifies as irrelevant knowledge won’t do the trick. In the case of the layout of my house, knowing that the kitchen is my favourite room won’t. After all, it’s not a true proposition about intrinsic properties of the structure of the phenomenon understood.

5. Conclusion This chapter has developed a novel account of the nature of understanding. Like the account of the nature of knowledge, the account offers a network analysis rather than attempting to dismantle the phenomenon into simpler elements. What’s more, one key thesis of my account is that

132 , ,   understanding is the constitutive aim of inquiry into general phenomena. Besides this, I have offered a set of necessary and sufficient conditions for maximal understanding and degrees of understanding, as well as for attributions of outright understanding, all of them, of course, firmly within the non-reductive framework of the network analysis. I showed that the arguments that inquiry into general phenomena requires systematic knowledge can be converted into arguments that understanding requires systematic knowledge also. In this way, the inquiry-based framework served to motivate yet another substantive view in epistemology, this time about understanding. The remainder of the chapter addressed a number of objections to knowledge-based accounts of understanding and showed that the view compares favourably with the main rival view in the literature. Chapter 5 will leave questions about the nature of epistemic states behind and turn to questions about their value.

5 Epistemic Value Chapter 4 offered a novel account of understanding. Like the account of knowledge from Chapter 2, it is non-reductive and analyses understanding in terms of its place in a network connecting inquiry into general phenomena, understanding, and views. While there is certainly more to be said about this network and the nature of its various nodes, I will leave these issues for another occasion. Instead, I will turn to questions about the value of knowledge and understanding.

1. A Normative Framework 1.1 Normativity in Epistemology It is common practice in contemporary philosophy to distinguish a variety of normative domains. The best-known examples are the moral, the practical, and the aesthetic domains. Normative domains feature distinctive values and norms. For instance, desire satisfaction is widely agreed to be a value in the practical domain, and that one ought to do what satisfies one’s desires is widely agreed to be a norm in the practical domain. In contrast, beauty is widely agreed to be a value in the aesthetic domain, and a norm may be that one ought to do what promotes beauty. Moral philosophy and aesthetics are normative disciplines. They concern, respectively, the moral and the aesthetic domains. When we are doing moral philosophy/aesthetics, we will have to answer a number of questions about the moral/aesthetic domain. Among these questions are: (i) What kind of things are the unmoved movers in the respective domains? Is it values, as teleologists would have it, or it is norms, as deontologists would maintain? (ii) How many unmoved movers are there in the domain? Is there only one, as monists would hold, or are Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0006

134 , ,   there more, as pluralists would have us believe? (iii) What exactly are the unmoved movers in the domain? Different moral/aesthetic theories provide different answers to these questions. By way of illustration, consider classical utilitarianism and classical Kantianism in moral philosophy. Classical utilitarianism takes happiness to be the only central moral value. Its central norm holds, roughly, that an action is morally permissible if and only if it causes the greatest happiness for the greatest number. It is easy to see that classical utilitarianism takes values to be the unmoved movers and so is teleological (re i). After all, the central norm is analysed in in terms of the central moral value. In this way, values enjoy explanatory priority over norms. What’s more, classical utilitarianism countenances only one central moral value, to wit, happiness. As a result, it is monistic (re ii). Finally, it identifies happiness as the only central moral value (re iii). Classical Kantianism takes the Kantian Moral Law to be the only central moral norm and the good will as its central value. In contrast with classical utilitarianism, classical Kantianism, at least on a standard reading, assigns explanatory priority to the norm, i.e. the Kantian Moral Law. The central value is understood in terms of the norm as follows: ‘a good will is a will whose decisions are wholly determined by . . . the Moral Law’ (Johnson and Cureton 2019, §2). Here the central moral value is analysed in terms of the central moral norm. On classical Kantianism, then, norms are the unmoved movers. The view is deontological (re i). Like classical utilitarianism, there is only one central moral norm, the Kantian Moral Law, which means that the view is monistic (re ii). Finally, it identifies the Kantian Moral Law as the only central moral norm (re iii). Epistemology is a normative discipline, just like moral philosophy and aesthetics. Epistemology is concerned with the epistemic domain, which is also a normative domain. Accordingly, when doing epistemology, we need to answer a number of questions about the epistemic domain, including (i) whether the unmoved movers in the epistemic domain are values or whether they are norms; (ii) whether there is only one kind of unmoved mover in the epistemic domain or whether there is a plurality; and (iii) what exactly the unmoved movers in the epistemic

 

135

domain are. Different epistemologies provide different answers to these questions. One central aim of this section is to provide answers to precisely these questions. In a nutshell, on the view I will develop momentarily, values have explanatory priority over norms (re i) and there are at least two central values in the epistemic domain (re ii), to wit, knowledge and understanding (re iii). It is easy enough to just say these things. It is quite a bit harder to argue for them. In fact, there is reason to think that the task of arguing for specific answers to the above questions is even harder for the epistemic domain than it is for the moral, practical, and aesthetic domains. To see why, note first that, in the case of the moral, practical, and aesthetic domains, we have a reasonable enough pre-theoretic grasp of their boundaries and how they differ from one another. For instance, we have at least some pre-theoretic grasp of the difference between the moral value of an action and its practical value. We manifest this grasp, for instance, when we acknowledge that stealing may well be good for the thief (practically good) while, at the same time, being immoral (not morally good). Our pre-theoretic grasp of the boundaries of these domains is of considerable importance for the project of answering questions about the moral and practical domains that we are interested in. At the very least, they provide an important checkpoint for our theorizing about these domains. For instance, it explains why a moral theory featuring a norm that takes stealing to be generally morally permissible will not enjoy much popularity. The reason why it will be harder to answer the same questions about the epistemic domain than about the moral and practical domains is that we do not have a similar pre-theoretic grasp of the epistemic domain. For instance, we don’t have a pre-theoretic grasp of the difference between epistemic and practical values. Consider the question whether acting on a belief in high-stakes cases is epistemically or practically bad. It’s not clear that we have a good pre-theoretic handle on this question. And, quite plausibly, one important reason for this is the absence of a pretheoretical grasp of the epistemic domain.¹ ¹ Thanks to an anonymous referee for suggesting this.

136 , ,   Another central aim of this section is to shed some light on the boundaries and the structure of the epistemic domain. In a nutshell, the answer is that ACANs in general constitute normative domains. The ACANs’ constitutive aims are central values in these domains. Norms are to be explained in terms of these values. And, most importantly, the epistemic domain is a domain constituted by an ACAN, to wit, by inquiry.

1.2 Critical Domains To get there I need to first introduce the idea of critical domains (Sosa 2007). Critical domains are normative domains that are associated with goods or values for that domain. Any critical domain has at least one central value. Central values are fundamental to the domain in the sense that it cannot be explained fully in terms of other values in that domain. Another way of putting this point is that central values are valuable for their own sake, relative to the domain. The qualification is important, because values in critical domains may not enjoy any of the more familiar kinds of value, such as moral, prudential, or aesthetic value. Crucially, central values organize the evaluations in that domain in the sense that all other domain-specific values and norms can be explained fully in terms of the central values. Critical domains are thus teleologically structured. One example Sosa uses to illustrate these rather abstract ideas is the domain of coffee. The thought here is that in this domain, a central value is ‘liquid coffee that is delicious and aromatic’ (2007: 73, henceforth good coffee for short). From the point of view of the domain of coffee, this value cannot be explained fully in terms of other values. (In keeping with that, it may be possible to explain it in terms of other values external to the domain, e.g. the value of pleasant experiences.) Crucially, this value organizes our evaluations in that domain in the sense that the values of other relevant entities, as well as norms in the domain, are explained in terms of it. For instance, we may ask questions about good beans, baristas, and so on. What it takes for a bean to be a good bean and for a barista to be a good barista will be explained in terms of the central

 

137

value of good coffee. In the present case, the value is instrumental value. Good beans and good baristas are good insofar as they are conducive to (i.e. effective means to) good coffee. And the same goes, mutatis mutandis, for norms in the domain of coffee. How one ought to roast beans, brew coffee, and so on is explained in terms of the central value of good coffee, too. The reason I am mentioning all this is that I want to suggest that ACANs constitute critical domains. More specifically, I want to put forward the following eminently plausible idea: ACANs’ constitutive aims correspond to central values in the domains they constitute. Consider, for instance, chess. The constitutive aim of chess is to checkmate one’s opponent. As a result, checkmating one’s opponent is a central value in the domain constituted by chess. Given that this is so, we can ask our three questions about domains constituted by ACANs: (i) Are the unmoved movers values or norms? (ii) Is there only one value or are there more? (iii) What exactly are the unmoved movers? Since the domains constituted by ACANs are critical domains and since critical domains are teleologically structured, (i) receives a straightforward answer. The unmoved movers are values; in other words, the domains constituted by ACANs are teleologically structured. As I will argue in due course, there is reason to think that this is independently plausible. A more interesting question is (ii). If constitutive aims of ACANs correspond to central values in the domains they constitute, does this mean that domains constituted by ACANs are monistic? There is reason to think that at least some such domains are indeed monistic. Chess is a plausible example. Checkmating is not only a central value in the domain of chess, but, in fact, it is the only one. That said, it is possible for the domains constituted by ACANs to feature more than one central value. A noteworthy example is snooker. The constitutive aim of snooker is to score more points than one’s opponent. This gives us one central value of snooker. However, the central value corresponding to the constitutive aim cannot be the only central value of snooker. To see why, note that to understand what the constitutive aim of snooker amounts to, we require further information on how points are scored. We need to know, for instance, that potting a red ball is worth one point, that potting a black

138 , ,   ball is worth seven points, and so on. Crucially, the value attaching to potting balls of various colours cannot be explained fully in terms of the value of scoring more points than one’s opponent. And, conversely, the value attaching to scoring more points than one’s opponent cannot be explained fully in terms of the value of potting balls of various colours. This means that snooker must have more than one central value. Here is what’s going on in snooker, on a more theoretical level. To begin with, note that snooker is an ACAN that nests further ACANs. Consider potting, safety, and snookering. These activities are nested in snooker in that, to play snooker, you may at least sometimes engage in any of them and you have to engage in at least some of them. Crucially, potting, safety, and snookering are all ACANs themselves. That is to say, they have their own constitutive aims and norms. For instance, potting constitutively aims at legally pocketing an object ball. Now, the key reason why snooker has more than one central value is that it is an ACAN such that the constitutive aim is defined in terms of a set of (perhaps suitably related) successes of nested ACANs. In other words, it is an ACAN that computes progress towards and attainment of its constitutive aims at least partly by scoring successes in nested ACANs. In what follows, I will also refer to ACANs that fit this bill as score-based ACANs. Score-based ACANs have more than one central value: one is given by the ACAN’s constitutive aim, another by any of the constitutive aims of nested ACANs, successes in which are used for scoring. It is easy enough to see why score-based ACANs must have more than one central value. It is not possible to fully explain the value that is given by its constitutive aim in terms of the value of the successes used for scoring. After all, even after we have agreed on the latter values, the question of what it takes to attain the value given by the ACAN’s constitutive aim remains open. And, vice versa, it’s also not possible to fully explain the value given by the successes used for scoring in terms of the value of the constitutive aim. Even if we agreed on these latter values, the question of how the former are scored remains open also.² ² It may be worth noting that not all ACANs are score-based ACANs in this sense. Chess, for example, isn’t. Even though there may be nested ACANs such as various openings and gambits, success in chess is not scored in terms of whether or not a certain opening, gambit, etc. was successful.

 

139

What comes to light, then, is that the question of whether the domains constituted by ACANs are monistic or pluralistic remains open. Some of them, such as chess, plausibly are monistic, whilst others, score-based ACANs, are pluralistic. Which way our answer to question (ii) will go, then, may vary from one ACAN to the next. And, of course, the same holds for question (iii).

1.3 The Epistemic Domain Recall that while we had a reasonable pre-theoretic grasp of the boundaries of some normative domains, we do not have much of a grip on the epistemic domain. This makes theorizing about the epistemic domain particularly difficult. In particular, we do not have the same kind of checkpoint that constrains theorizing about other domains for which we do have a reasonable pre-theoretic grasp. In the absence of a reasonable pre-theoretic grasp of the epistemic domain, I want to make the following suggestion: the epistemic domain is constituted by an ACAN, to wit, by inquiry. I do not have a direct argument for this suggestion. That said, I will, in a moment, provide reason to believe that the suggestion allows us to give answers to the key questions about the normative domains that are not only motivated but also plausible. In this way, the suggestion enjoys at least some indirect support: it makes the difficult issue of how to delineate the epistemic domain tractable and it provides plausible answers to central questions about it. My suggestion is that the epistemic domain is constituted by inquiry. But now recall that I distinguished between two forms of inquiry: inquiry into specific questions and inquiry into general phenomena. Given that this is so, we may wonder exactly how the epistemic domain is constituted by inquiry. To answer this, I want to make my suggestion a little more precise: the epistemic domain is constituted by inquiry into general phenomena, which is a score-based ACAN. Recall that score-based ACANs compute progress towards and attainment of their constitutive aims, at least partly by scoring successes in nested ACANs. In the case of inquiry into general phenomena the relevant nested ACAN is inquiry

140 , ,   into specific questions. To see that inquiry into specific questions is nested in inquiry into general phenomena, note that to engage in inquiry into general phenomena, you may engage in inquiry into specific questions; in fact, it is plausible that, in this case, you will have to do so. To see that inquiry into general phenomena is a score-based ACAN, note that attaining and progress towards attaining the aim of inquiry into general phenomena, i.e. understanding, are computed at least in part in terms of success in inquiries into specific questions, i.e. knowledge. But now recall that there is reason to think that a score-based ACAN must have more than one central value. This is because it is not possible to fully explain the value that is given by its constitutive aim in terms of the value of the successes used for scoring. Note that this is independently plausible for the domain constituted by inquiry into general phenomena. The value of knowledge cannot be fully explained in terms of the value of understanding. Even if we are clear that understanding is the constitutive aim of inquiry into general phenomena and so a central value in the domain constituted by it, the question of how progress is scored remains open. And, vice versa, even if we are clear that knowledge is how progress is scored, the question of whether the constitutive aim is attained and the corresponding central value is realized remains open. With these points in play, we can now answer our three key questions about the epistemic domain. Since the epistemic domain is constituted by an ACAN and since ACANs are teleologically structured, so is the epistemic domain (re i). Since the epistemic domain is constituted by inquiry into general phenomena, which is a score-based ACAN, and since score-based ACANs are pluralistically structured, so is the epistemic domain (re ii). Finally, since inquiry into general phenomena nests inquiry into specific questions and since attaining and progress towards attaining the aim of inquiry into general phenomena is computed at least in part in terms of success in inquiries into specific questions, both understanding and knowledge are central values in the epistemic domain. In the absence of a reasonable pre-theoretic grasp of the epistemic domain, theorizing about it is difficult and may even seem altogether intractable. The account of the epistemic domain I have developed in this section brings it into clearer view in that it sheds light on its boundaries

 

141

and structure and provides motivated and plausible answers about a number of central questions we’ll have to address when theorizing about normative domains. With the account in play, I will now turn to some value problems in epistemology. I will argue that this account of the epistemic domain paves the way towards attractive solutions to even the hardest incarnations of these problems.

2. Value Problems 2.1 Knowledge One of the key tasks of epistemology is to explain the distinctive value of knowledge. While it is widely recognized that it won’t be hard to explain why knowledge has some value, it’s believed to be difficult to explain why knowledge is distinctively valuable. The reason for this, in turn, is that, in order to successfully do so, one must solve a range of socalled value problems, that is, roughly, problems of explaining why knowledge is more valuable than belief that enjoys certain other properties. It is widely believed that knowledge is more valuable than mere true belief. Unfortunately, on reflection, it’s not entirely clear exactly why this should be the case. This was noted already by Plato (1956) in the Meno, where Socrates challenges his interlocutor to explain why exactly we should think that knowledge is more valuable than mere true belief. After all, a mere true belief regarding the right way to Larissa would seem to be just as useful as knowledge of the way to Larissa: both will get us where we want to go. But surely knowledge will be distinctively valuable only if our impression is right, and it is more valuable at least than mere true belief. Accordingly, any satisfactory account of the distinctive value of knowledge must explain how this can be. This gives us: The Primary Value Problem Any satisfactory account of the value of knowledge must explain why knowledge is in some respect more valuable than mere true belief.

142 , ,   Some have claimed that simply meeting the Primary Value Problem won’t be enough to give a satisfactory account of the value of knowledge. Jonathan Kvanvig, for one, argues that more is needed: suppose that there is some logically weaker property than knowledge, P, which may but need not be truth, such that knowledge is not more valuable than belief that has P. In that case, it would be wrong to think that knowledge is distinctively valuable. After all, we have no reason to care about knowledge rather than belief that has P. In view of these considerations, Kvanvig favours the following constraint: The Secondary Value Problem Any satisfactory account of the value of knowledge must explain why knowledge is in some respect more valuable than belief that falls short of knowledge. (Kvanvig 2003, xii–xiii)

Duncan Pritchard (Pritchard, Millar, and Haddock 2010) ups the stakes even further; according to him, in order to account for the distinctive value of knowledge, one must explain why knowledge is more valuable than the corresponding belief that falls short of knowledge not just as a matter of degree, but also as a matter of kind. This gives us: The Tertiary Value Problem Any satisfactory account of the value of knowledge must explain why knowledge is in some respect more valuable than belief that falls short of knowledge not just as a matter of degree but as a matter of kind. (Pritchard, Millar, and Haddock 2010, 8)

Of course, if we can solve the Tertiary Value Problem, then we will automatically also have solved the Primary and the Secondary Value Problems. But what, then, does it take to solve the Tertiary Value Problem? What does it take to show that knowledge is more valuable than belief that falls short of knowledge not just as a matter of degree but as a matter of kind? Pritchard’s (2010, 8) answer is that we will have to show that knowledge is valuable for its own sake.

 

143

The trouble, according to Pritchard, is that the prospects of solving the Tertiary Value Problem are rather dim. He takes it that there is only one candidate that might deliver the goods on the market, which is associated with virtue epistemology. In a nutshell, according to virtue epistemology, knowledge is a cognitive achievement. Since achievements in general are valuable for their own sake, so is knowledge (e.g. Greco 2010). According to Pritchard, the trouble is that virtue epistemology is mistaken: knowledge isn’t a cognitive achievement. As a result, the only contender on the market that stood any chance of solving the Tertiary Value Problem in the first place turns out to fail. That’s not exactly great news. In fact, Pritchard himself takes his argument to motivate a form of scepticism about the distinctive value of knowledge. Contrary to what appears to be widely believed, knowledge is not distinctively valuable. One may wonder whether the Tertiary Value Problem really places an adequate demand on accounts of the distinctive value for knowledge. I will get back to this question in due course. For now, I will simply grant Pritchard that it does. The reason for this is that, as I will argue in a moment, my account can solve even this version of the value problem. If the demands of this problem are unrealistically high, there is all the more reason for thinking that my account does all it needs to in order to adequately account for the value of knowledge. In addition, I’d already like to flag that my account is not virtue-epistemological in the sense that it does not rely on the thesis that knowledge is a cognitive achievement.³ As a result, the account offers an alternative for those who think that explaining the distinctive value of knowledge requires solving the Tertiary Value Problem. Virtue epistemology is no longer the only game in town. In fact, it is not hard to guess what my solution to the Tertiary Value Problem looks like: since knowledge is a central value in the epistemic domain, it is valuable for its own sake, relative to the epistemic domain. It is easy to see that knowledge enjoys a different kind of value from belief that falls short of knowledge. After all, belief that falls short of knowledge does not enjoy the status of a central value in the epistemic domain. It follows that knowledge has a kind of value, one that belief that falls short ³ That said, it is compatible with it (see Appendix 2).

144 , ,   of knowledge does not enjoy: knowledge is a central epistemic value; it is valuable for its own sake, relative to the epistemic domain. The Tertiary Value Problem is thus solved.

2.2 Understanding With this account of the epistemic value of knowledge in play, I’d now like to move on to understanding. Of course, one key task in epistemology is to explain the distinctive value of knowledge. It is hard to deny, however, that another no less significant task is to explain the distinctive value of understanding. And it might be thought that, just as in the case of knowledge, in order to achieve this, we will have to solve some set of value problems. Recall that, in the case of knowledge, this approach was initially motivated by the intuition that knowledge is more valuable than mere true belief. Likewise, I am confident that many would agree that understanding a phenomenon is more valuable than merely having some knowledge about it. Accordingly, there is reason to think that any satisfactory account of the distinctive value of understanding must explain why this is so. In other words, it must solve the following value problem: The First Value Problem Any satisfactory account of the value of understanding must explain why understanding a phenomenon is more valuable than having some knowledge about it.

And just as the Primary Value Problem could be adapted, so we can adapt the Secondary Value Problem: The Second Value Problem Any satisfactory account of the value of understanding must explain why understanding a phenomenon is more valuable than that which falls short of understanding.

 

145

And finally, we may, of course, follow Pritchard in holding that in order to explain the distinctive value of understanding, we must show that understanding has a different kind of value from that which falls short of it: The Third Value Problem Any satisfactory account of the value of understanding must explain why understanding is in some respect more valuable than that which falls short of it not just as a matter of degree but as a matter of kind.

Again, solving the Third Value Problem is the hardest challenge and will suffice to solve the other two. Unfortunately, unlike in the case of the Tertiary Value Problem, it is difficult to see how the present account could do so. After all, on my view, understanding is a central epistemic value, i.e. something that has final value relative to the epistemic domain. The trouble is that knowledge is also a central epistemic value, even if it falls short of understanding. As a result, it is hard to see how the present view could have the resources to solve the Third Value Problem. Recall that I shelved the question of whether the Tertiary Value Problem really places an adequate demand on accounts of the distinctive value of knowledge. I would now like to return to this question. Or, to be more precise, the question I am really interested in is whether the Third Value Problem places an adequate demand on accounts of the distinctive value of understanding. As I am about to argue, the answer to this question is no. To begin with, let’s consider Pritchard’s original motivations for the Tertiary Value Problem. Here goes: [I]f one regards knowledge as being more valuable than that which falls short of knowledge merely as a matter of degree rather than kind, then this has the effect of putting knowledge on a kind of continuum of value with regard to the epistemic, albeit further up the continuum than anything that falls short of knowledge. The problem with this ‘continuum’ account of the value of knowledge, however, is that it fails to explain why the long history of epistemological discussion has focused specifically on the stage in this continuum of value that

146 , ,   knowledge marks rather than some other stage (such as a stage just before the one marked out by knowledge, or just after). Accordingly, it seems that accounting for our intuitions about the value of knowledge requires us to offer an explanation of why knowledge has not just a greater degree but also a different kind of value than whatever falls short of knowledge. (Pritchard, Millar, and Haddock 2010, 7–8).

What motivates the Tertiary Value Problem is a worry about accounts according to which the value of knowledge is on a continuum with the value of belief that isn’t knowledge. Any such account would not do proper justice to the distinctive value of knowledge. In addition, Pritchard also claims that the only way in which we can avoid placing knowledge on such a continuum is by showing that knowledge enjoys a different kind of value. That’s why he thinks that any adequate account of the distinctive value of knowledge must solve the Tertiary Value Problem. Now, I think that Pritchard’s point that any adequate account of the value of knowledge must take knowledge off a value continuum with belief that falls short of knowledge is well taken. What’s more, the same goes, mutatis mutandis, for understanding. However, I part ways with Pritchard when he contends that the only way in which this can be done is by showing that knowledge/understanding enjoys a different kind of value. Here is why. On the present view, understanding and knowledge are both central epistemic values, that is to say, they are both valuable for their own sake, relative to the epistemic domain. However, they are different central epistemic values; neither one can be fully explained in terms of the other. As a result, crucially, it is not as if understanding simply realizes more of the same value that knowledge also realizes. In this way, on the present view, the value of understanding is not on a continuum with the value of knowledge. Since knowledge and understanding are both central epistemic values, they are values of the same kind. At the same time, since they are different central epistemic values, they are not on a value continuum. What comes to light, then, is this: Pritchard may well be right in thinking that an adequate account of the value of knowledge/understanding must

 

147

take knowledge/understanding off a value continuum with what falls short of knowledge/understanding. Even so, he is wrong to think that the only way to achieve this is by showing that knowledge/understanding enjoys a different kind of value than that which falls short of knowledge/ understanding. By the same token, the Tertiary/Third Value Problem is thus too demanding. The question remains what a more adequate demand on accounts of the distinctive value of knowledge/understanding might look like. One attractive suggestion is simply to hold that the account must take knowledge/understanding off a value continuum with what falls short of knowledge. Alternatively, it is also plausible that any account on which knowledge/understanding comes out as being valuable for its own sake, if only relative to the epistemic domain, will meet any adequate demand on accounts of the distinctive value of knowledge/understanding. Fortunately, it is easy enough to see that the present account not only meets the revised challenge but also the condition that’s sufficient for meeting any adequate demand on accounts of the distinctive value of knowledge/understanding. After all, on the present view, understanding is a central epistemic value and so does enjoy for-its-own-sake value, relative to the epistemic domain.

3. Objections In what follows, I will consider a series of objections to my account of the distinctive value of knowledge and understanding. While a lot of my discussion will be focused specifically on the value of knowledge, it is easy to see that most (if not all) of the objections could be raised, mutatis mutandis, for understanding. Fortunately, my responses will work, again mutatis mutandis, here as well.

3.1 No Domain-Relative For-Its-Own-Sake Value One might worry that my proposed solution to the Tertiary Value Problem doesn’t really work because there is no such thing as domain-relative

148 , ,   for-its-own-sake value. Rather, if something has for-its-own-sake value, then it has it simpliciter. If so and if Pritchard is right that in order to solve the Tertiary Value Problem knowledge needs to have for-its-own-sake value, then the proposed account does not solve this version of the value problem. Now, I already argued that the Tertiary Value Problem places too strong a demand on accounts of the value of knowledge to be adequate. What’s more, I also showed that the demand is out of kilter with the motivations Pritchard adduces, which are more plausible. The claim that the right solution to the Tertiary Value Problem requires that knowledge must have for-its-own-sake value exacerbates the problem, especially if we do not countenance domain-relative for-its-own-sake value. Here is why. While I would strongly resist the claim that there are no domainrelative for-its-own-sake values, let’s assume that it is true. Even so, it remains plausible that there are what I called central values in domains, i.e. values that organize the evaluations in that domain in the sense that all other domain-specific values and norms can be explained fully in terms of them, while they cannot be fully explained in terms of the other values. For instance, even if we don’t want to say that checkmating one’s opponent has for-its-own-sake value because it does not have for-itsown-sake value simpliciter, it is plausible that it is a central value in the domain of chess. But given that this is so, it is possible to solve the Tertiary Value Problem without showing that knowledge has for-its-own-sake value. In fact, the above account does precisely that. Even if there is no such thing as for-its-own-sake value relative to the epistemic domain, knowledge remains a central value in the epistemic domain. At the same time, belief that falls short of knowledge is not a central value. In this way, knowledge enjoys a different kind of value from that of belief that falls short of knowledge and so the account solves the Tertiary Value Problem. Since it does so without showing that knowledge has for-its-own-sake value, it comes to light that solving the Tertiary Value Problem does not require showing that knowledge has for-its-own-sake value.⁴ ⁴ It may also be worth repeating that, even if there is no such thing as domain-relative forits-own-sake value, the account preserves the original motivations for the Tertiary Value

 

149

3.2 Central Values vs Fundamental Goods But even if this is right, reservations about whether this solution to the Tertiary Value Problem actually works may remain. To see why, note that Pritchard himself countenances something in the vicinity of what I call a central value. He calls it a fundamental good in that domain. A fundamental good in a domain here is something that at least sometimes has domain-relative value for its own sake (Pritchard, Millar, and Haddock 2010, 11–12). Now, it might be thought that what I call a central value just is what Pritchard calls a fundamental good. But if that’s right, then there is excellent reason to think that showing that knowledge is a central value in this sense is not enough to solve the Tertiary Value Problem. After all, on the standard interpretation of all of the above versions of the value problem, what needs to be done in order to solve any one of them is to show that knowledge always has the required form of excess value. But if all we show is that knowledge is a central value, where that is taken to be tantamount to what Pritchard calls a fundamental good, all we have shown is that knowledge at least sometimes enjoys a different kind of value from that of beliefs that fall short of knowledge. In that case, however, we won’t have solved the Tertiary Value Problem. The key to my response to this worry is that there is an important difference between what I call a central value and what Pritchard calls a fundamental good after all. Pritchard’s fundamental goods at least sometimes enjoy domain-relative for-its-own-sake value. Of course, to make the point that my account doesn’t solve the Tertiary Value Problem stick, Pritchard will need the additional claim that fundamental goods do not (and more specifically that knowledge does not) always enjoy this form of value. So, suppose they don’t. The trouble is that it now gets pretty hard to see how they could still play the role that central values are supposed to play, i.e. organize the evaluations in the domains to which they are central. This is easiest to see for domains with only one central Problem, i.e. taking knowledge off a value continuum with belief that falls short of knowledge. What’s more, it also vindicates the special attention that knowledge has received among epistemologists. After all, it makes perfect sense for epistemologists to pay special attention to central values in the epistemic domain.

150 , ,   value. Take chess, for instance. I already mentioned that it is pretty uncontroversial that checkmating is the only central value in the domain of chess. Now suppose that checkmating is only sometimes valuable for its own sake in the domain of chess. Since nothing else has for-its-ownsake value in chess, tokens of checkmating that are not valuable for their own sake will simply not enjoy any chess value. But, in that case, just how does checkmating organize other evaluations in chess? For instance, it seems quite plausible to say that a good chess move is one that makes checkmating the opponent more likely. But that won’t work on the present proposal. After all, if the case at hand is one in which checkmating the opponent has no chess value, there will be no chess value to making checkmating more likely. There will be no good moves in such a game (nor bad ones for that matter). Alternatively, it would seem quite plausible to say that a good chess player is one who tends to checkmate opponents (perhaps of a certain quality) often enough. Again, it’s hard to see how this could work on the present proposal. After all, it may be that our player tends to do so only in matches in which checkmating has no chess value. In that case, checkmating can’t confer any value on them, and so they wouldn’t count as a good chess player. I wouldn’t be surprised if many would find these consequences undesirable enough to dismiss the idea that it is Pritchard’s fundamental goods for a given domain that organize the evaluations in that domain. That said, at the very least, if it were Pritchard’s fundamental goods for a domain that do the job here, our story about how this works will have to be considerably more complicated than we may initially have thought, which is already a minus. What’s more, in order to flesh out the more complex theory, we will need some story about when tokens of a fundamental good are indeed valuable for their own sake in that domain and when they aren’t. Now, for starters, it is pretty hard to even begin to tell this story. Suppose checkmating is only sometimes valuable for its own sake relative to the domain of chess. When is a token checkmating valuable for its own sake in chess and when isn’t it? These questions strike me as pretty tough to tackle. What’s worse, we must make sure that, whatever our answer might be, it does not compromise the role of fundamental goods as organizers of the evaluations of the relevant domains. This means that we will do well not to invoke other values such as moral or prudential ones. Here is why. Suppose that whether a

 

151

token of a fundamental good is valuable for its own sake relative to the domain at hand turns on facts about some domain-external type of value such as moral or prudential value. In that case, it now appears that what organizes the evaluations in the domain is really the relevant domainexternal type of value, not the fundamental goods in that domain. If this isn’t immediately obvious, consider the case of chess once more. If what determines whether a token checkmating is valuable for its own sake in the domain of chess is, say, facts about moral value, then the question whether a move is a good move is ultimately determined by these very facts, as is the question whether a player is a good player. After all, it is the moral facts that ultimately determine whether the move is good rather than having no value at all. And it is the moral facts that determine whether the match at hand contributes to the player being a good player. What I take these considerations to suggest is that we will do well to construe the values that organize the evaluations in a given domain as always enjoying for-its-own sake value, relative to that domain. This will allow us to avoid not only the undesirable consequence we are otherwise settled with but also the complications that we are bound to encounter. Perhaps most importantly, it will enable us to retain our straightforward understanding of what makes for, for example, a good move in chess and a good chess player. Given the role that central values are meant to play, then, there is excellent reason for thinking that they differ from what Pritchard calls fundamental goods in a domain. In particular, there is reason to think that central values are always valuable for their own sake, relative to their proper domains. What this means for the case of knowledge is that knowledge is always valuable for its own sake, relative to the epistemic domain, whereas belief that falls short of knowledge is not. By the same token, the worry that the proposed account does not solve the Tertiary Value Problem because it establishes the relevant excess value only for some items of knowledge can be laid to rest.

3.3 A Superficial Solution? One might worry that even if the present account solves the value problem, it doesn’t do so in anything but a superficial sense. On the present account, knowledge is a central epistemic value because it is the

152 , ,   aim of inquiry into specific questions. But even if that solves the value problem, one might think that it misses what’s at its very heart, to wit, the question why this should be. Why should it be knowledge that we pursue in inquiry rather than, say, true belief or something else?⁵ By way of response, let’s first consider chess. Suppose that there are (perhaps alien) theorists of chess who have discovered that checkmating has distinctive value in chess and have wondered why this might be. Suppose, further, that they recently agreed that the distinctive value of checkmating is explained by the fact that checkmating is the constitutive aim of chess and, as a result, checkmating is a central value in the domain constituted by chess. Suppose, finally, that a sceptic raises the worry that this doesn’t get to the heart of the issue: the important question, they suggest, is why it is checkmating that we pursue when playing chess rather than, say, perpetual chess or something else. The answer to this worry is that we pursue checkmating in chess rather than perpetual chess or something else because checkmating is the constitutive aim of chess. That is to say, if the question that is at the heart of the value problem for chess is why we pursue checkmating in chess rather than perpetual chess or something else, the fact that checkmating is the constitutive aim of chess provides the answer. By the same token, the worry that the account of the value of checkmating under consideration does not get to the heart of the value problem can be laid to rest. And the same goes, mutatis mutandis, for the worry that my account of the value of knowledge does not get to the heart of the value problem. But perhaps the question was meant to be slightly different. Perhaps the question at the heart of the value problem was not why we pursue knowledge in inquiry rather than true belief or something else but rather why knowledge is the aim of inquiry rather than true belief or something else. In response, let’s turn to chess once more. In particular, I’d like to first consider the question why checkmating is the aim of chess rather than perpetual chess or something else. This question simply doesn’t have a very satisfactory answer, not beyond the observation that if it had a ⁵ Thanks to an anonymous referee for raising this worry.

 

153

different aim, it would have been a different game. And, again, the same holds, mutatis mutandis, for inquiry. This leaves the question why we are playing chess rather than some other game. I think it is impossible to answer this question fully. This is because there is no good answer to why we are playing chess rather than a game that is just like chess except that its constitutive aim is to capture the king. At the same time, I don’t mean to deny that it is possible to answer this question partially. Chess is enjoyable to play, much more so than many alternatives we might have played instead but don’t. And that explains why we are playing chess rather than these alternatives. But note that in giving this answer we have now left the domain of chess. What explains why we play chess rather than some alternative turns on chessexternal, practical factors, to wit, enjoyability. Of course, there is much more to be said about what exactly makes chess enjoyable to play, and more so than some alternatives we don’t play but might have played instead. My best guess is that part of the story about what makes chess enjoyable to play is that it achieves a good balance between a variety of factors, including some that pull in opposite directions such as playability and challengingness. And what makes chess more enjoyable to play than some alternatives is that it achieves a better such balance. Crucially, since we have now left the domain of chess, these details are not for chess theorists to work out and so can safely be set aside by chess theorists. Unsurprisingly, once again, I want to say that the same goes, mutatis mutandis, for inquiry. The question why we engage in inquiry rather than in a similar activity with a slightly different constitutive aim is impossible to answer fully. For instance, it’s impossible to answer the question why we engage in inquiry rather than an activity that is just like inquiry except that it aims at knowledge+.⁶ At the same time, I don’t mean to deny that it is possible to answer this question partially. Inquiry is useful to engage in, much more so than many alternatives we might have engaged in instead but don’t. And that explains why we are

⁶ Knowledge+ made an appearance in Chapter 1. Recall that one knows+ that p if and only if (i) one knows that p and (ii) one’s degree of justification for p is ever so slightly higher than what’s required for knowing that p.

154 , ,   engaging in inquiry rather than of these alternatives. But note that in giving this answer we have now left the domain of epistemology. What explains why we engage in inquiry turns on epistemology-external, practical factors, to wit, usefulness. Again, there is much more to be said about what exactly makes inquiry a useful activity to engage in, and more so than some alternatives we might have engaged in instead. My best guess is that inquiry achieves a good balance between a variety of factors, including some that pull in opposite directions, such as attainability and quality (Carter, Jarvis, and Rubin 2015). And what makes it more useful to engage in than some alternatives (including an activity that constitutively aims at true belief) is that it achieves a better such balance. Crucially, since we have now left the domain of epistemology, these details are not for epistemologists to work out and so can safely be set aside here.

3.4 The Epistemic Value of Justified and True Belief Let’s suppose that this story works in the way envisaged. If so, we have a solution to the value problem, including in its difficult tertiary incarnation, and, by the same token, a promising account of the distinctive value of knowledge. While this would, of course, constitute a significant achievement, it cannot be the whole story about epistemic value. After all, while no one would deny that knowledge has epistemic value, it is also immensely plausible that justified and true belief have epistemic value also. At the very least, it is plausible that justified and true belief are epistemically better than, respectively, unjustified and false belief. And here is where another difficulty for the present account arises. How can it be that justified and true belief are epistemically better than, respectively, unjustified and false belief? Note that traditionalists have little trouble in answering this question. After all, they take true belief but not false belief to be the relevant central value in the epistemic domain. If, in addition, they go for a teleological account of justification according to which, roughly, justification is essentially truth-conducive, they also have an attractive account of why justified belief is better than unjustified belief ready to hand. After all, justified belief is a good means to the central

 

155

epistemic value of true belief whereas unjustified belief isn’t. Hence, justified belief enjoys instrumental epistemic value over and above unjustified belief. But this story is evidently not available to those who think that knowledge is the relevant final value in the epistemic domain. How, then, can they account for the epistemic value of justified and true belief? On the upside, the question about justified belief is fairly easy to answer. After all, champions of the present view are free to adopt a teleological account of justification as well. Of course, they will do well to construe justification not as essentially truth-conducive but rather as essentially knowledge-conducive. Once they have done so, however, the account of the value of justified belief can proceed as expected: justified belief is epistemically better than unjustified belief because justification is conducive to the central epistemic value of knowledge. In fact, we have already seen in Chapter 3 that my own account of justification falls into just this camp. Unsurprisingly, then, I would not consider the move from an account of justification according to which it is essentially truth-conducive to one according to which it is essentially knowledge-conducive as a cost. On the contrary, in my view, there is independent reason to make it. Accounting for the epistemic value of true belief, on the other hand, is considerably harder. After all, the idea that true belief is essentially knowledge-conducive in the way that justification is does not carry much promise. What should a champion of the above account say about the value of true belief? Here is one response: true belief simply does not have distinctively epistemic value and so is not better than false belief. While this response may seem crazy at first, I believe that, on reflection, it is not without merit. To see this, suppose that you believe a certain proposition in a way that is random in the sense that the chances of arriving at a true belief in this way are fifty-fifty. Suppose, on this particular occasion, you happen to hit upon a truth. What’s good about your belief? Of course, now that you have arrived at a true belief you may well be more likely to succeed in your projects than had your belief been false. But that just means your belief enjoys additional practical value in virtue of being true. What we are after, however, is distinctively epistemic value. Accordingly, the

156 , ,   question we need to ask is what’s good about your belief from a distinctively epistemic point of view. If you are already convinced that true belief is a central epistemic value, the answer is obvious. What’s good about your belief is that it enjoys for-its-own-sake epistemic value. But once you have given up the idea that true belief is a central epistemic value, it is just not clear why we should think that what you arrived at enjoys value that’s distinctively epistemic at all. What’s more, the impression of craziness mentioned above may very well be the result of (what we are assuming to be) an erroneous theory of epistemic value according to which true belief is a central epistemic value. But since mistaken views frequently give rise to mistaken impressions, we need not attach much evidential weight to the impression of craziness either. Even if, on reflection, this response is better than it may initially have appeared, I wouldn’t be surprised if many considered adopting it a cost for the present view. After all, at the very least, the view is revisionist in a way that many may find objectionable. The question, then, is whether there is a story to be told that accords positive epistemic value to true belief over and above false belief without taking this value to be forits-own-sake value, relative to the epistemic domain. Fortunately, the answer to this question is yes. This leads me to my second response. Let’s start by asking whether true belief could be a good means to knowledge. It may be thought that the answer has to be no. After all, a true belief either already is knowledge or it isn’t. As a result, true belief cannot be instrumental to knowledge. There surely is an important sense in which this thought is correct. More specifically, I agree that a true belief at a given time cannot be a good means to knowledge at that time. However, in keeping with that it may well be that present true belief is a good means to future knowledge in a way in which false belief isn’t an equally good means to future knowledge. If so, true belief does have instrumental epistemic value over and above false belief. To see what I am getting at here, let’s return once more to Williamson’s burglar case from Chapter 2. More specifically, consider the following variation of it: two burglars have entered our two houses to steal our diamonds. Both have beliefs about where the diamonds are: in the safes under our beds, say. Crucially, only the belief of the burglar at your house is true. While my diamonds are usually in the safe, this very night, I took them out to wear them at a

 

157

party. At present, the burglars open the safes. Shortly after that, the one at your house comes to know where the diamonds are. In contrast, the one at my house is still in the dark. In fact, he may never discover the location of the diamonds. What this case forcefully illustrates, then, is that present true belief makes future knowledge more likely than present false belief. In this way, present true belief is a better means to future knowledge than present false belief. On this line, it turns out that true belief does enjoy instrumental epistemic value over and above false belief because present true belief is a better means to future knowledge than present false belief.⁷

4. Conclusion This chapter has used the framework developed in the previous chapters to offer a novel account of the epistemic domain as a domain that is constituted by inquiry. What’s more, the account delivers new solutions to various value problems in epistemology, including (the plausible motivations of) the relevant versions of the controversial Tertiary Value Problem. More specifically, I have argued that ACANs constitute critical domains, which have their constitutive aims as central values, i.e. things that are valuable for their own sake, relative to the domains they constitute. Since both inquiry into specific questions and inquiry into general phenomena are ACANs and since they have, respectively, knowledge and understanding as their constitutive aims, we get the result that knowledge and understanding are valuable for their own sake, relative to the domains constituted by these ACANs. The epistemic domain is identified with the domain constituted by inquiry into general phenomena in which both knowledge and understanding are final values. As such, they have a kind of value that anything falling short of knowledge/ understanding does not have.

⁷ Note also that true belief may have other kinds of extrinsic epistemic value as well. Since knowledge entails true belief, perhaps they inherit value from knowledge in a way in which false beliefs don’t. Thanks to Mona Simion for drawing my attention to this option.

158 , ,   One thing that may be worth noting is that, while I did respond to a number of objections to my view, unlike in earlier chapters I did not consider how this view compares with the competition. The reason for this is that I am not convinced that there is much in the way of competition to consider. After all, it is entirely compatible with the present account of the value of knowledge and understanding that the two are valuable in many other ways as well. In fact, I’d be happy to countenance a vast variety of ways in which knowledge and understanding are valuable. If any of them works, that’s great. It means that knowledge and/or understanding are even more valuable/valuable in other respects than I have outlined here. (In fact, I believe that they are and have explored the prospects of some relevant avenues concerning the value of knowledge elsewhere; see Kelp and Simion 2017, Simion and Kelp 2016). For that reason, there is no need to try to establish that my view compares favourably with alternative accounts of the value of knowledge and/or understanding.

6 Scepticism Chapter 5 offered a novel account of epistemic value. In this final chapter, I will turn to the last central epistemological question, which concerns the extent of our knowledge. My aim is to show that, contrary to what the sceptic might have us think, we know a lot about the world, perhaps even more than we may have thought. To begin with, I will introduce the sceptical problem.

1. The Argument from Ignorance Consider the following famous sceptical argument: The Argument from Ignorance AI1. You don’t know that (~SH =) sceptical hypothesis SH (e.g. that you are not a handless brain in a vat (BIV)) is false. AI2. If you don’t know that ~SH (e.g. that you are not a handless BIV), then you don’t know a corresponding ordinary empirical proposition O (e.g. that you have hands) to be true. AI3. Hence, you don’t know that O (e.g. that you have hands). Both premises of the Argument from Ignorance are highly plausible. AI1 is intuitively compelling. You don’t know that you are not a handless BIV. How could you? After all, everything would seem to you exactly as it would, were you to be a normal handed person. AI2 is motivated by the following attractive principle:

Inquiry, Knowledge, and Understanding. Christoph Kelp, Oxford University Press (2021). © Christoph Kelp. DOI: 10.1093/oso/9780192896094.003.0007

160 , ,   Transmission of Knowledge If one knows that p, one competently deduces q from p, and one comes to believe that q based on one’s deduction, then one thereby comes to know that q.¹

But, of course, in conjunction, AI1 and AI2 commit us to the highly implausible sceptical claim, AI3, that you don’t even know that you have hands. The Argument from Ignorance is perhaps the strongest sceptical argument on the market. If we want to avoid its sceptical conclusion, we will have to deny at least one of its premises. In fact, as I will argue in the remainder of this chapter, there is reason to resist both of the argument’s premises. Or, to be more precise, we will do well to deny both AI1 and Transmission of Knowledge, which motivates AI2. I will start my discussion with Transmission of Knowledge.

2. The Case against Transmission of Knowledge Many think that competent deduction is a way of extending one’s knowledge. What’s more, they think that Transmission of Knowledge captures this thought at least roughly. Others disagree. They argue that cases like the following by Fred Dretske (1970) cause trouble for Transmission of Knowledge: Zebra You are at the zoo. Currently you are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since you can tell a zebra from the way it looks, you come to know the

¹ Williamson (2000), Hawthorne (2004). This principle also sometimes goes by the name of Closure. In contrast with the debate on knowledge, the literature on warrant distinguishes sharply between closure principles and transmission principles. Since the above would clearly be categorized as a transmission principle there, I decided to go with the label Transmission of Knowledge.



161

proposition that Z = it is a zebra. From this you competently deduce the proposition that ~CDM = it is not a cleverly disguised mule and you thereupon come to believe ~CDM.

What foes of Transmission of Knowledge typically point out at this stage is that, intuitively, your belief that ~CDM does not qualify as knowledge. Given that this intuition is correct, Transmission of Knowledge fails. After all, as the description of the case clearly indicates, you satisfy all three conditions in the antecedent of the principle. However, since you don’t know what you come to believe, you do not satisfy the consequent. Even if it is intuitive that you do not come to know that ~CDM in Zebra, Transmission of Knowledge is also highly intuitive (Williamson 2000). This raises the question whether we should accept the counterexamples and reject Transmission of Knowledge or else hold on to Transmission of Knowledge and embrace that agents in cases like Zebra do acquire knowledge. One key aim of this chapter is to argue that we should indeed reject Transmission of Knowledge. Crucially, the argument I will offer is theoretical in the sense that it relies on a number of independently plausible premises to establish the failure of Transmission of Knowledge, rather than just on the intuition of absence of knowledge in cases like Zebra. What’s more, the argument also ushers the way towards carving out a certain type of case in which Transmission of Knowledge fails. If Transmission of Knowledge fails, these failures should be borne out by the correct account of knowledge. The second aim of this chapter is to argue that the account of knowledge developed in Chapters 2 and 3 can do just that. To be more precise, I will show that this kind of account has the resources to accommodate not only that Transmission of Knowledge fails in the relevant type of case, but also that all remaining instances of Transmission of Knowledge continue to hold. In this way, my preferred account of knowledge not only bears out failures of Transmission of Knowledge but also supports a suitably restricted version of the principle.

162 , ,  

2.1 The Argument This section offers an argument against Transmission of Knowledge. In order to achieve this, I will once more focus on inquiry and, more specifically, on inquiry into specific whether questions. Recall the lightweight characterization of the aim of inquiry: Question-Settling Aim One’s inquiry into whether p aims at settling the question whether p.

Another thesis that I take to be no less plausible is that question-begging excludes question-settling. This gives us: No Settling If one’s inquiry into whether p proceeds via a question-begging argument (henceforth is question-begging for short), then it doesn’t settle the question whether p.

Now consider the following two cases involving inquiry into whether ~CDM: Inquiry 1 You are at the zoo. Currently you are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since you can tell a zebra from the way it looks, you come to know that Z. You now want to find out whether ~CDM. To settle this question, you take a hair sample from the animal and perform a DNA analysis. The result of the analysis is that the animal is indeed a zebra. From this, you competently deduce and thereupon come to believe that ~CDM. Inquiry 2 You are at the zoo. Currently you are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since you can tell a zebra from the way it looks, you come to know that Z. You now want to find out whether ~CDM. To settle this question,



163

you exploit the entailment from Z to ~CDM to competently deduce and thereupon come to believe that ~CDM.

There is a clear difference between Inquiry 1 and 2. While there is nothing wrong with your inquiry in Inquiry 1, in Inquiry 2, your inquiry is problematic. What explains this difference? By far the best answer that I can think of is that, in Inquiry 2, as opposed to 1, your inquiry is question-begging (more on this below in Section 2.2). Reflection on the difference between Inquiry 1 and 2 thus motivates the following thesis: Question-Begging In Inquiry 2, your inquiry into whether ~CDM is question-begging.²

Now, one might wonder what all of these rather unexceptional points have to do with Transmission of Knowledge. To see the answer, recall that there is a more substantive characterization of the aim of inquiry in terms of knowledge: Knowledge Aim Inquiry into whether p aims at knowledge that p/not-p.³

It is easy to see that from Question-Settling Aim, Question-Begging, No Settling, and Knowledge Sufficiency, it follows that, in Inquiry 2, your inquiry into ~CDM does not lead you to knowledge that ~CDM. But now recall how your inquiry in Inquiry 2 proceeds: from Z, which you know, you competently deduce and thereupon come to believe that ~CDM. If Transmission of Knowledge holds, your inquiry leads you to

² It may be worth noting that even champions of Transmission of Knowledge and the related transmission of warrant principle (see also Section 2.2 below) acknowledge this point (Pryor 2000, Markie 2005, Pritchard 2007). ³ It may be worth noting that, for the purposes of this argument, all that I need is the sufficiency direction of Knowledge Aim. Since the competition in the literature advocates weaker aims of inquiry, they would all agree on the sufficiency direction of Knowledge Aim. By the same token, the relevant premise of my argument should not raise too many eyebrows.

164 , ,   knowledge that ~CDM. Hence, Transmission of Knowledge fails.⁴ Crucially, however, this failure is not just supported by an intuition that you do not know that ~CDM. Rather, we have an independent theoretical argument for this which exploits a couple of platitudinous principles relating the aim of inquiry, question-begging and questionsettling, an independently plausible principle relating knowledge and the aim of inquiry, and a highly attractive explanation of a clear difference between Inquiry 1 and 2 in terms of question-begging.

2.2 Diagnosis Reflection on the structure of question-settling serves to further support this result. To see how, note first that there is a lively and related debate in the literature on whether the following principle holds: Transmission of Warrant If one has warrant for p and one competently deduces q from p, then one thereby has warrant for q.

One interesting fact that this debate has unearthed is that the division on Transmission of Warrant is driven by a disagreement about the structure of warrant (Pryor 2004). In particular, we can distinguish between three ⁴ Here is a more detailed statement of the argument: 1. Transmission of Knowledge holds. [Assumption for RAA] 2. In Inquiry 2, you know that Z, competently deduce ~CDM from Z, and thereupon come to believe that ~CDM. [Description of Inquiry 2] 3. In Inquiry 2, you come to know that ~CDM. [1,2] 4. In Inquiry 2, your inquiry into whether ~CDM is question-begging. [Question-Begging] 5. In Inquiry 2, your inquiry into whether ~CDM doesn’t settle the question whether ~CDM. [4, No Settling] 6. In Inquiry 2, your inquiry into whether ~CDM does not attain its aim. [5, QuestionSettling Aim] 7. In Inquiry 2, your inquiry into ~CDM doesn’t lead you to knowledge that ~CDM, i.e. you don’t come to know that ~CDM. [6, Knowledge Aim] 8. ⊥ [3,7]. 9. Transmission of Knowledge does not hold. [1,8]



165

relevant views: Conservatism (Wright 2004, 2007, White 2006), Moderatism (Coliva 2012, 2015) and Liberalism (Pryor 2000, 2004, Davis 2009). In the case of perceptual warrants these views disagree about the status of propositions like ~CDM for acquiring warrant for propositions like Z by perceptual means. Conservatives claim that you need an antecedent warrant for propositions like ~CDM. Moderates claim that, while such propositions need not be antecedently warranted, they must be antecedently assumed in some sense to be specified. In contrast, liberals disagree with both: they deny that propositions like ~CDM need be antecedently warranted or assumed. They do grant, however, that they have the potential to defeat one’s warrant for propositions like Z. However, they will do so only in special circumstances, such as when one has reason to think that ~CDM is true. Here is how this division bears on the question of Transmission of Warrant. Conservatism and Moderatism license failures of Transmission of Warrant. In Zebra, for instance, they hold that ~CDM must already be in place (as either antecedently warranted or assumed) for perceptual means to generate a warrant for Z. As a result, when you reason from Z to ~CDM, you effectively fall foul of circular reasoning. Since circular reasoning cannot confer warrant on its conclusion, warrant fails to transmit from Z to ~CDM. In contrast, Liberalism does not require ~CDM to be in place in any way for you to acquire a warrant for Z by perceptual means. As a result, the inference from Z to ~CDM will not come out as circular either and is perfectly suited to transmit warrant to the conclusion. Let’s now consider the structure of inquiry. We can easily enough imagine the following analogues to Conservatism, Moderatism, and Liberalism about the structure of warrant, which I will label I-Conservatism, I-Moderatism, and I-Liberalism respectively. IConservatism claims that in order to settle the question whether Z by perceptual means, for instance, you need to already have settled the question whether ~CDM. I-Moderatism claims that that’s too much. All that’s needed is that ~CDM is already true. I-Liberalism takes not even this to be required. You can settle the question whether Z by perceptual means whether or not ~CDM has already been settled or whether or not it is already true.

166 , ,   The crucial point is that it is highly plausible that I-Moderatism is true of the structure of inquiry. Certain facts must already be in place for you to be able to settle the question whether Z by perceptual means. For instance, when you are in a part of the world that is predominantly populated with cleverly disguised mules, when the animal before you is a cleverly disguised mule, etc., you cannot settle the question whether Z just by looking. That’s why I-Moderatism is true rather than ILiberalism. At the same time, in order to settle the question whether Z by looking, you don’t have to already have settled the question whether the relevant facts are indeed in place. You do not need to already have settled the question whether you are in cleverly disguised mule county, whether ~CDM, etc. That would be much too demanding. In fact, if such a requirement were in place, it is hard to see how we could ever settle any question at all. That’s why I-Moderatism is true rather I-Conservatism. Given that Moderatism licenses transmission of warrant failures, it should now come as no surprise that question-settling is not transmissive either. In fact, the reason for this is parallel in both cases. Recall that, according to Moderatism, perceptual means will give you a warrant for Z only if ~CDM is assumed. That’s why when you deduce ~CDM from Z, you fall foul of circular reasoning. Warrant fails to transmit across your deduction. Similarly, settling the question whether Z by looking requires that you are not in disguised mule county, that ~CDM, etc. When, having settled Z by looking, you move on to settle the question whether ~CDM by deduction from it being a zebra, you fail in a similar way: you fall foul of begging the question. The property of questionsettling fails to transmit across the deduction as well. But, of course, given that knowledge is sufficient for question-settling, this means that in cases in which the property of question-settling fails to transmit across competent deduction, you don’t come to know the conclusion of the deduction. Given that, in some such cases (e.g. Inquiry 2), it is independently plausible that you know the premise and that you believe the conclusion based on competent deduction from the premise, this means that Transmission of Knowledge is bound to fail. In this way, the argument against Transmission of Knowledge receives further support by reflecting on the structure of question-settling.



167

2.3 The Case Before moving on, I’d like to point out that we also have what it takes to specify a certain type of case in which Transmission of Knowledge fails due to question-begging. To see this, note that the reason why, in Inquiry 2, your inquiry into whether ~CDM begs the question generalizes. If I-Moderatism is true, settling the question whether p in way Wp will typically (if not invariably) require that a certain set of facts ΔWp is already in place. More specifically, I want to suggest that ΔWp is the set of facts that obtains if and only if what we may call the ‘false positive conditions’ of Wp do not obtain. By the ‘false positive conditions’ I mean the conditions under which W produces a belief on whether p that falls short of knowledge. Anyone who then ventures to settle any further question whether q by deducing, say, q from p, where q corresponds to some i ∈ ΔWp, will fall foul of question-begging. If, in addition, one knows that p by having settled the question whether p in Wp and has come to believe q by competent deduction from p, we will have a case of Transmission of Knowledge failure due to question-begging. We thus have a type of case in which Transmission of Knowledge fails due to question-begging. It is this type of case (henceforth ‘the Case’) that I will focus on in the remainder of this chapter.⁵ With the argument against Transmission of Knowledge in play, I will now return to my preferred account of knowledge. My aim here is to show that this account has the resources to accommodate not only that Transmission of Knowledge fails in the relevant type of case, but also that all remaining instances of Transmission of Knowledge continue to hold. In order to achieve this, I will first argue that, given certain assumptions, my account of knowledge validates an unrestricted version of Transmission of Knowledge. This establishes that the account can license instances of Transmission of Knowledge, including the

⁵ There may be other kinds of case in which Transmission of Knowledge fails, perhaps even due to question-begging. While this may mean that the account of Transmission of Knowledge failure is not complete, it does not diminish its significance. After all, understanding only one such kind of case means important progress on the issue.

168 , ,   target ones. Then I will show how the assumptions that commit me to Transmission of Knowledge can be replaced in such a way that Transmission of Knowledge fails just in the Case. In this way, my account of knowledge supports just the kind of restricted version of Transmission of Knowledge we are after.

2.4 The Validation of Transmission of Knowledge First things first, recall the transmission principle as well as the substantive conditions on knowledge that I developed in Chapter 3: Transmission of Knowledge If (C1 =) one knows that p, (C2 =) one competently deduces q from p, and (C3 =) one comes to believe that q based on one’s deduction, then (C4 =) one thereby comes to know that q. Conditions on Knowledge One knows that p only if (K1 =) one’s belief that p competent, and (K2 =) the EN of the ability to know that produced one’s belief that p are satisfied. Competent Belief One’s belief that p is competent if and only if it is formed by an exercise of an ability to know propositions in range R and relative to conditions C such that p∈R.

Recall also that abilities are intrinsically connected to success by the relevant dispositions. The strength of the success connection may vary in a number of ways. For instance, there may be variation in the probability of success conditional on the exercise of ability in C. The thought here is that the higher the probability of success given the exercise of ability in C, the stronger the ability (relative to C). Most importantly for present purposes, the strongest abilities (relative to C) feature sure-fire dispositions, i.e. dispositions such that triggering (the exercise of ability) in suitable conditions (C) guarantees manifestation (success). As a result, for abilities to know featuring sure-fire dispositions, competent belief in C entails knowledge.



169

Consider, next, the following bridge principles between Transmission of Knowledge and Conditions on Knowledge: Bridge 1 (C2) and (C3) hold if and only if one believes that q via the exercise of a deductive epistemic ability (DEAq). Bridge 2 (C1) identifies a member of CDEAq. Now consider the following assumptions: Assumption 1 The disposition at issue in DEAq is a sure-fire disposition. Assumption 2. There are no other members of CDEAq besides (C1). Assumption 3. The range of DEAq includes all propositions q one may competently deduce from p. My account of knowledge, Bridge 1 and 2, and Assumptions 1–3 jointly entail (and thus validate) Transmission of Knowledge. Here’s why. To begin with, by Bridge 1, Assumption 3, and Competent Belief, if (C2) and (C3) hold, then one will competently believe that q. Next, by Bridge 2 and Assumption 2, (C1) specifies the only members of CDEAq. Hence, if (C1) holds, the set of conditions to which DEAq is relative must be satisfied. Hence, if (C1)–(C3) hold, then one must also competently believe that q in CDEAq. But now recall that for any ability featuring a sure-fire disposition, competent belief in C entails knowledge. Since, by Assumption 1, the disposition at issue in DEA is a sure-fire disposition, it follows that if (C1)–(C3) hold, then one must not only competently believe that q in CDEAq, but one is also guaranteed knowledge that q, i.e. (C4) holds. Given my preferred account of knowledge, Bridge 1 and 2, and Assumption 1–3, we get the result that if (C1)–(C3) hold, then (C4) must hold also. Transmission of Knowledge is thus validated.

170 , ,  

2.5 Accommodating Transmission of Knowledge Failure Next, I will show how to accommodate failures of Transmission of Knowledge in and just in the Case. Of course, in order to deny Transmission of Knowledge, we must block the validation argument from the last subsection. That means that we will have to resist at least one of the bridge principles or assumptions the argument exploits. At the same time, wherever we may place our resistance, we had better made sure that we do not thereby close the door to explaining that the remaining instances of Transmission of Knowledge do hold. It is easy to see that this means that the bridge principles are off limits. And the same goes for Assumption 1. After all, it is needed to ensure that the satisfaction of (C1)–(C3) guarantees the satisfaction of (C4) even for uncontentious instances of Transmission of Knowledge. This leaves Assumption 2 and 3. As I am about to argue, both options can be made to work. The first and most important step is to connect ΔWp with Conditions of Knowledge. Since, on my view of knowledge, one has settled the question whether p if and only if one knows that p/not-p, it is immensely plausible that the way of settling the question whether p referred to in the Case, just is an epistemic ability (EA) that has p in its range at issue in Conditions on Knowledge. In other words, it is immensely plausible that Wp = EAp. Moreover, it is also immensely plausible that the set of facts that I-Moderatism requires to be in place for the question whether p to be settled will just be the set of conditions to which the epistemic ability is relative. In other words, it is also immensely plausible that ΔWp = CEAp. Given that this is so, we can describe the Case in terms of the conceptual resources of Conditions of Knowledge: (i) one knows that p via the exercise of an epistemic ability, EAp, that requires CEAp to be in place, (ii) one has come to believe that q via the exercise of a deductive epistemic ability, DEA, and (iii) q corresponds to some i ∈ CEAp. With the Case so described, here’s what we need to do to accommodate the relevant failures of Transmission of Knowledge: we need to disallow instances of Transmission of Knowledge for any q such that q corresponds to some i ∈ CEAp. It is not hard to see that there are two ways of achieving this, each corresponding to the denial of one of Assumptions 2 and 3. First, pace Assumption 2, we may maintain that,



171

in addition to one knowing the premise of one’s deduction, a further member of CEAp is that q not correspond to any i ∈ CEAp. Alternatively, pace Assumption 3, we may hold that the range of DEAq includes only propositions q such that q does not correspond to any i ∈ CEAp, alongside the requirement that one may competently deduce q from p, of course. It is easy to see that (i) either option will accommodate the failure of Transmission of Knowledge in the Case, whilst (ii) continuing to allow instances of Transmission of Knowledge to hold in all other cases. Does it matter which alternative we opt for? Yes. There is a crucial difference between the two approaches. To see this, let’s consider once again the first option, which denies Assumption 2 and holds on to Assumption 3. On this approach, while the belief in the conclusion of the deduction falls short of knowledge, it continues to be justified, at least on my preferred account of justification, i.e. Justified Belief (see Chapter 3). The reason for this is that, by Bridge 1, one’s belief that q is produced by the exercise of DEAq and, by Assumption 3, q is within the range of this ability. It follows from Competent Belief that one’s belief in q is competent and from Justified Belief that it is justified. In contrast, on the alternative approach, the conclusion belief qualifies as neither knowledge nor justified. Recall that here we restrict the range of DEAq to include only propositions q such that one may competently deduce q from p and q does not correspond to any i ∈ CEAp. For any q that corresponds to some i ∈ CEAp, then, while, by Bridge 1, one’s belief that q may well be produced by the exercise of DEAq, it will not be in the range of this ability. By Competent Belief, it is not competent and, by Justified Belief, it is not justified. Finally, it strikes me as highly plausible that, in the Case, the deductive inference fails to transmit not only knowledge, but also justified belief. For instance, when, in Inquiry 2, you come to believe ~CDM by competent deduction from Z, the question-begging character of your inquiry prevents you not only from coming to know ~CDM, but also from coming to believe it justifiably. As a result, I take the second option, which places further restrictions on the range of DEAq, to be preferable. This completes the account of Transmission of Knowledge failure I favour.

172 , ,  

2.6 Back to AI2 What remains to be seen is that this account of transmission failure is relevant to our task of dealing with the Argument from Ignorance. Fortunately, it is easy enough to see that it is. Sceptical hypotheses are paradigm cases of false positive conditions for our nearly all of our ways of belief formation. As a result, the falsity of sceptical hypotheses is among the conditions that need to be in place for perceptual ways of forming beliefs to settle questions (alternatively, for our epistemic abilities to give us knowledge). On the above account of transmission failure, knowledge will fail to transmit across deductions from ordinary empirical propositions to the denials of sceptical hypotheses. For instance, knowledge will fail to transmit across the deduction from you have hands to you are not a handless BIV. And, of course, if knowledge fails to transmit from ordinary empirical propositions to the denials of sceptical hypotheses, the key motivation that Transmission of Knowledge provided for AI2 will be lost.

3. The Case against AI1 Given that Transmission of Knowledge fails, including in the sorts of cases at issue in the Argument from Ignorance, it may seem that we now have a way to block the Argument from Ignorance. In particular, we can now deny the argument’s second premise, AI2, which is motivated by Transmission of Knowledge. Since denial of AI2 is independently motivated (by the above argument against Transmission of Knowledge), so is the resulting way of blocking the Argument from Ignorance. What’s more, given that Transmission of Knowledge seemed intuitively plausible, we even have an account of why we were so taken by the Argument of Ignorance. The reason is that it features an intuitively plausible but ultimately false principle. So far, so good. Of course, the mere fact that Transmission of Knowledge is false and, as a result, the Argument from Ignorance does not go through tells us precious little about the status of AI1 and AI3. Do you or don’t you know that you are not a handless BIV? Do you or don’t you know that you have



173

hands? While these questions do indeed remain open, it might be thought that there is no real need to address them in any detail at this stage. After all, as the Argument from Ignorance forcefully indicates, they are intuitively plausible. So, shouldn’t we just hold on to our intuitions here and maintain that AI1 and AI3 are both true, i.e. that you do know that you have hands and that you don’t know that you are not a handless BIV? What’s more, one might think that an inquiry-based approach only serves to confirm this intuition. After all, it might seem that while the question whether you have hands is settled, the question whether you are not a handless BIV isn’t, and in fact couldn’t be. In fact, doesn’t holding on to both AI1 and AI3 give us the happiest of solutions to the problem of scepticism? The sceptic is right about something: you don’t know that you are not a handless BIV. However, the non-sceptic is also right about something: you do know that you have hands. What goes wrong is that Transmission of Knowledge fails. What’s more, once we see that it fails, we can explain why we find the Argument from Ignorance so plausible in the way just sketched. Isn’t this all we ever wanted from a solution to scepticism?

3.1 Reasons for Resisting AI1 Unfortunately, the answer to these questions is no. One important reason for this is that you will be committed to abominable conjunctions such as: Abomination I know that I have hands, but I don’t know that I am not a handless BIV. (DeRose 1995, 28)

What’s more, the following doesn’t sound any better: Abomination* I have settled the question whether I have hands but the question whether I am not a handless BIV remains entirely open.

174 , ,   No one in their right mind would want to be committed to something abominable. Accordingly, views that do generate such commitments thereby suffer a strike against them. The bad news for the present proposal is that it generates a commitment to abominable conjunctions, including Abomination and Abomination*. Whatever else may speak in favour of the present proposal, then, this fact speaks against it. Here is a related problem with the present proposal. Suppose that knowledge is the norm of assertion. If so, by the lights of the present proposal, you may permissibly assert that you have hands and that this entails that you are not a handless BIV. However, you cannot permissibly assert that you are not a handless BIV. To see just why this is bad, suppose someone asked a champion of the present proposal whether they have hands. They would of course say yes. And when asked whether this entails that they are not handless BIVs, they would again say yes. But, of course, when asked whether they are not handless BIVs, they couldn’t say yes. Rather, they’d have to say something like ‘I can’t say.’ But that is just absurd. As Hawthorne aptly put it, in this case you are just like ‘Lewis Carroll’s Tortoise, that familiar object of ridicule who was perfectly willing to accept the premises of a modus ponens argument but was unwilling to accept the conclusion’ (Hawthorne 2005, 32).

3.2 Theoretical Underpinnings I take it that these problems indicate that we cannot respond to the sceptical argument simply by denying Transmission of Knowledge which motivates AI2. Rather, we must also deny AI1. That is to say, we must hold that you do know that you are not a handless BIV, that the question whether you are not a handless BIV is settled. Now, what I take to be the central task of any view that denies AI1 is to explain exactly how it is that you could possibly know that you are not a handless BIV—alternatively, exactly how it is that you could possibly settle the question whether you are not a handless BIV—given that everything would seem just the same if you were. Recall that, in Chapter 3, I argued that knowledge is acquired via abilities to know. This account constrains the central task of explaining exactly how it is



175

that you could possibly know that you are not a BIV. The explanation will need to invoke an ability to know. Unfortunately, it does not help with discharging this task. We cannot hope to explain how you could possibly know that you are not a handless BIV by simply pointing out that you do so via an ability to know. What we need, at the very least, is to identify a specific ability to know that can deliver knowledge that you are not a handless BIV and to offer an account of how this ability does the job. It is this task that I will take on in the remainder of this subsection. There is a popular account available on the market. According to this view, championed most prominently by G. E. Moore (e.g. 1939), you know that you are not a handless BIV on the basis of competent deduction from your knowledge that you have hands. On Moore’s view, then, the relevant ability to know is a deductive epistemic ability and the way in which this ability delivers knowledge that you are not a handless BIV is just the normal way in which exercise of this ability delivers knowledge. The trouble with the Moorean account is, of course, that it requires (relevant instances of) Transmission of Knowledge and so is unavailable to those who want to adopt the above way of denying Transmission of Knowledge. So, it looks as though foes of Transmission of Knowledge face a particularly difficult challenge here. To get my own preferred account of how you can know that you are not a handless BIV in view, I’d first like to focus on a slightly different type of case, which does not involve sceptical hypotheses. Suppose, for instance, someone were to put to us the hypothesis (TOS =) that there is a teapot orbiting the Sun somewhere between the Earth and Mars or (SM =) that the universe was created by a supernatural creator closely resembling a serving of spaghetti with meatballs. As a first observation, I want to insist that we know that these hypotheses are false. At the same time, it doesn’t look as though we will be able to produce a conclusive argument against these hypotheses. At the very least, many of us won’t be able to do this. So then, how can we know these facts? To arrive at what I take to be a promising answer to this question, I’d like to ask how a sane person might react to being presented with such a hypothesis. Here is what I take to be the sane response: ‘Get real! That’s just crazy!’ What’s going on here? My suggestion is that we have an epistemic ability that enables us to recognize that certain possibilities could

176 , ,   not easily enough obtain. It is via an exercise of this ability that we come to know that the possibilities described in TOS and SM could not easily enough obtain (and hence that ~TOS and that ~SM are both true). This suggestion offers an attractive explanation of the sane response. Upon consideration of the possibility described in TOS and SM, you come to know that these possibilities could not easily enough obtain by an exercise of your epistemic ability to recognize possibilities that could not easily enough obtain. ‘Get real! That’s just crazy!’ is an expression of this very knowledge. The fact that my suggestion offers an attractive explanation of the sane response provides some evidence that it is on the right track. There is further support for my suggestion. Consider the following two important properties of beliefs produced by recognitional abilities (henceforth recognitional beliefs). First, recognitional beliefs are cognitively spontaneous; they do not appear to be based on further evidence. For instance, when you form a perceptual recognitional belief that there is a chair before you, you do not appear to base this belief on further evidence such as the appearance as of a chair. Second, in case of recognitional beliefs, we are often unable to articulate how the recognitional ability works in any detail. For instance, someone may have the ability to recognize pieces by Beethoven without being able to articulate how they do so (other than by making the rather uninformative observation that it sounds like Beethoven) (Millar 2010). But now note that, our beliefs that the possibilities described in TOS and SM could not easily enough obtain plausibly satisfy these two key properties of recognitional beliefs. First, when we acquire these beliefs, we do not appear to base them on further evidence. Rather, they are cognitively spontaneous. Second, we may not be able to articulate how we recognize the relevant possibilities as such (other than by making the rather uninformative observation that they just look crazy). The fact that our beliefs fit the profile of recognitional beliefs further confirms my suggestion.⁶

⁶ Just how this ability might operate is a fascinating question. It may not be the job for philosophers to answer it, at least not fully, and I will not attempt to do so here. Instead, I will content myself with the observation that, even if we do not have an account of how the ability operates, the above still provides some evidence that we do indeed have this ability.



177

If I am right, we have an epistemic ability that enables us to recognize whether certain possibilities could not easily enough obtain. It will come as no surprise that it is just this ability that I want to suggest enables us to come to know the denials of sceptical hypotheses. Here is how. In many situations, the above account for the possibilities at issue in TOS and SM can be adapted for sceptical possibilities. For instance, suppose that someone asks you whether you have any siblings. Suppose you respond that you have two and I interject by asking how you can be so sure, after all you might be a BIV. In this case the sane response is once again something along the lines of ‘Get real! That’s just crazy!’⁷ Alternatively, consider a court case in which the solicitor attempts to undermine the prosecution’s case against the defendant by invoking sceptical possibilities. Again, we’d expect a response very much along the lines of ‘Get real! That’s just crazy!’ (except, of course, that it is likely that it will be phrased slightly differently). Just as in the TOS and SM cases, what’s doing the work here is your epistemic ability to recognize that certain possibilities could not easily obtain, which offers an attractive explanation of the sane response along the following lines: upon consideration of the possibility that you might be a BIV, you come to know that this possibility could not easily enough obtain by an exercise of your epistemic ability to recognize possibilities that could not easily enough obtain. Again, ‘Get real! That’s just crazy!’ is an expression of this very knowledge.⁸ In this way, even champions of the present approach to transmission failure can deny AI1, that is to say, they can allow that we know the denials of sceptical hypothesis. Crucially, they are known not via competent deduction from ordinary empirical propositions but rather via the exercise of a separate recognitional ability.⁹

⁷ Note also that the sane response is not, for instance ‘Well, I have siblings, so the BIV possibility doesn’t obtain.’ As a result, Moorean responses to the sceptical argument do not fit well with the common-sense reaction here. ⁸ It is also easy to see that our beliefs here fit the profile of recognitional beliefs just as well as in the TOS and SM cases, with the result that my proposal is further confirmed. ⁹ It may also be worth noting that invoking this ability in my response to the sceptic is not ad hoc. After all, the evidence for the existence of the ability is already provided by the TOS and SM cases. As far as the response to scepticism is concerned, then, the argument invokes an ability that there is independent reason to think exists.

178 , ,  

3.3 Diagnosis Finally, I want to iron out a last wrinkle in the proposed treatment of the sceptical problem. While it is hard to deny that there are indeed many situations in which nearly everyone would offer some version or other of the sane response when presented with sceptical possibilities, there are also situations in which most of us wouldn’t. The clearest examples are certain academic and educational settings. Here, many of us will not only fail to give the sane response, but we will even be very much taken by the sceptical possibilities and may be tempted to agree that we cannot know the denials of various sceptical hypotheses. Accordingly, any treatment of the sceptical problem according to which we can know the denials of sceptical hypotheses will not be fully adequate until it has explained why we tend to be so taken by AI1, at least in the relevant settings. And that’s what I want to do now. To begin with, note that many recognitional abilities have what I will call illusion conditions, that is, conditions under which exercises of these abilities make things seem to us in a way in which they are not. One classical example is the bent stick illusion. Here exercises of relevant perceptual recognitional abilities make it seem to us that a stick partly immersed in water is bent, whereas in fact it is straight. Note that this illusion only occurs in special settings, i.e. when the stick is immersed in water. In the vast majority of situations, we are not prone to this kind of illusion. Rather, we have no problem recognizing whether a certain stick is straight. Likewise, I want to suggest that, when considering the problem of scepticism in certain settings, most notably academic and educational ones, our ability to recognize whether certain possibilities could easily enough obtain falls prey to an illusion: it makes it seem to us as though sceptical possibilities could indeed easily enough obtain even though, in fact, they could not. And, of course, that’s what explains why we are so taken by them. What do we need to do in order to see through a given illusion? This varies from one case to the next. In case of the bent stick illusion, the answer is particularly easy: we just take the stick out of the water to verify that it is not bent. What about the sceptical illusion? My suggestion is that what will help reveal that the sceptical possibilities such as the BIV



179

possibility could not easily enough obtain is to consider them alongside possibilities such as the one that a teapot is orbiting the sun and that the universe was created by a flying spaghetti monster. Once we do so, we can see that these possibilities are no more outlandish than the familiar sceptical ones. As a result, we can now see the sceptical possibilities for what they really are: possibilities that could not easily enough obtain.¹⁰

4. Objections In what follows, I will consider some objections to my treatment of the sceptical problem, focusing mainly on the denial of Transmission of Knowledge. I hasten to add that there are others which I will turn to only in Section 5. There I will discuss so-called sensitivity-based approaches to scepticism, perhaps the closest rival of the present approach. More specifically, I will take a look at a number of objections that have been raised against sensitivity-based approaches to scepticism and show that my view is preferable because it does not fall prey to them.

4.1 Denying Bridge 1 One might wonder whether there isn’t another way of responding to the validation argument, one which allows us to hang on to

¹⁰ I assume that not everyone will agree that they can now see that sceptical possibilities could not easily enough obtain. It may even be that they see the above ‘crazy’ possibilities as ones that could easily enough obtain. But even this can be explained. Some illusions are harder to see through, and some ways of revealing them won’t work equally well for everyone. Thus, consider the grey strawberries illusion. Here we have a picture of some strawberries, which look red even though they are not: the picture does not contain any red colour. One way of revealing this illusion is by adding a grey strip at the bottom of the picture and another grey strip that comes up from that strip to the depicted strawberries. However, this way of revealing the illusion doesn’t work for everyone, just as may be the case with the sceptical illusion. In fact, just as in the sceptical illusion, it may be that some see the above ‘crazy’ possibilities as ones that could easily obtain, in the grey strawberries illusion, some subjects will simply see the grey strip as red as well.

180 , ,   Transmission of Knowledge after all. The key idea here is to not only place the aforementioned restrictions on the range of DEAq, but also to abandon Bridge 1 in favour of the following: Bridge 1* (C2) and (C3) hold if and only if one believes that q via the exercise of a DEAq such that q is in the range of this ability. The result that we get is that, in the Case, the agent simply does not competently deduce the conclusion from the premise. In this way, we can have an unrestricted version of Transmission of Knowledge, whilst also securing the result that, in the Case, the deductive inference does not transmit knowledge. And isn’t that the best of all worlds? I agree that this would be a very nice outcome. But now note that it is very plausible that whether you satisfy (C2) and (C3), i.e. whether you competently deduce one proposition from another, is a formal issue in the sense that it doesn’t (or at least needn’t) depend on what the semantic contents of your deduction are. For instance, suppose that you replace the propositions you want to employ in a deduction by variables and ask me to do the deduction for you. I should still be able to do so competently. But Bridge 1* threatens to violate the formality of competent deduction and clearly does so when combined with the further restrictions on the range of DEAq mentioned earlier. To see this, just suppose that Z has been replaced by A and ~CDM by B. On the present proposal, I cannot competently deduce B from A and A ! B simply because of how A and B have been introduced. That’s why I think that, on reflection, we will do well not to replace Bridge 1 by Bridge 1* and stick to the restriction strategy developed above.

4.2 Lost Knowledge The second objection also aims to show that there is a viable transmission principle for knowledge after all. To begin with, recall that I assumed the following version of the principle:



181

Transmission of Knowledge If one knows that p, one competently deduces q from p, and one comes to believe that q based on one’s deduction, then one thereby comes to know that q.

Now, while this principle comes close enough to the right principle for many purposes, in the final analysis, it arguably affords the following refinement: Transmission of Knowledge* If one knows that p, one competently deduces q from p, one comes to believe that q based on one’s deduction, and one retains one’s knowledge that p throughout, then one thereby comes to know that q.

Even if the above argument works against Transmission of Knowledge, it is not clear that it will also work for Transmission of Knowledge*. More specifically, it might be thought that once you go through the deduction and base your belief on the conclusion of it, you will have lost your knowledge of the premise. One question that arises at this stage is just how it is that knowledge of the premise is lost when one goes through the deduction. While I think that this question is at least not trivial to answer, I will set it aside here. The reason for this is that we will do well not to go down this route in any case. To see why, note that many, myself included, think that knowledge has normative significance, in the sense that it features in various norms, such as assertion, practical reasoning, and action.¹¹ The trouble is that if, in the Case, knowledge of the premise is lost as one goes through the deduction, then one is henceforth bound to violate all of these knowledge norms. To take just one example, suppose you lose your knowledge that Z as a result of competently deducing and thereupon coming to believe that ~CDM. The result is that all of your assertions of Z will now violate the knowledge norm of assertion and so are bound to be improper. ¹¹ See, e.g., DeRose (2002), Fantl and McGrath (2009), Hawthorne (2005), Hawthorne and Stanley (2008), Kelp (2018a), Simion (2016), Turri (2016a), Unger (1975), and Williamson (2000). I hasten to add that I do not subscribe to the knowledge norm of action.

182 , ,   What’s more, they are bound to be improper in a way in which they would not have been, had you not deduced and thereupon come to believe that ~CDM. But that just doesn’t seem to be right. Suppose that you have moved on from the zebra pen and encounter another visitor who asks you what animal is in the pen at the end of the path. If, in response, you assert that Z, the propriety of your assertion simply does not appear to hinge on whether you have come to believe that ~CDM based on a competent deduction from Z.¹² As a result, we will do well not to try to rescue an unrestricted transmission principle for knowledge by holding that when, in the Case, you go through the deduction and come to believe the conclusion based on it, you lose your knowledge of the premise along the way.

4.3 Question-Begging Doesn’t question-begging require an audience, someone against whom one is begging the question? And, doesn’t it require a context of disputation in which a question is begged? Moreover, don’t facts about question-begging vary across contexts of disputation, with the result that something that begs the question in some such context won’t do so in another? If the answer to these questions is yes, one might start to wonder whether the seemingly plausible Question-Begging is true after all. To see this, suppose first you are having a dispute with a radical climate change denier. It might well be that, on the present proposal, you are begging the question against your opponent because they would deny some of the premises of your case for the existence of climate change. Suppose, next, that you are having a dispute with a climate change agnostic. It might well be that the very same case that begged the question against the climate change denier does not beg the question against the agnostic. If so, it would now seem that question-begging varies with context of disputation. What’s more, it would also appear plausible that coming ¹² It is not hard to imagine similar cases in which Z is used as a premise in practical reasoning or acted upon.



183

to know that climate change is happening, say, is entirely compatible with the case for climate change begging the question against a diehard climate change denier. If so, begging the question (say against the diehard denier) is entirely compatible with attaining the aim of inquiry into it. This means that Question-Begging fails and the argument against Transmission of Knowledge does not go through. My inclination is to say that this view of question-begging is mistaken: question-begging is an epistemic phenomenon rather than a dialectical one. However, rather than driving this point home, I will try to provide some reason to think that my argument against Transmission of Knowledge is acceptable even to those whose preferences lie with a dialectical account of question-begging. To begin with, I’d like to introduce the following variations of the above Inquiry cases. The main difference between the cases is that they feature two people, you and I, rather than just you. Importantly, the thought is that I am your closest possible epistemic and psychological duplicate. That is to say, aside from necessary differences that arise from us being different people, I know, justifiably believe, believe, etc. just what you know, justifiably believe, believe, etc. Now consider: Inquiry 1* You and I are at the zoo. Currently we are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since we can tell a zebra from the way it looks, we both come to know that Z. We now want to find out whether ~CDM. You take the lead. To settle this question, you take a hair sample from the animal and perform a DNA analysis. The result of the analysis is that the animal is indeed a zebra. From this, you competently deduce and thereupon come to believe that ~CDM and present the corresponding argument to me. Inquiry 2* You and I are at the zoo. Currently we are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since we can tell a zebra from the way it looks, we both come to know

184 , ,   that Z. We now want to find out whether ~CDM. You take the lead. To settle this question, you exploit the entailment from Z to ~CDM to competently deduce and thereupon come to believe that ~CDM and present the corresponding argument to me.

Just as in Inquiry 1 and 2 earlier, there is a clear difference between Inquiry 1* and 2*. In Inquiry 1*, it would not be right for me to remain unmoved by your argument, whereas, in Inquiry 2*, it is perfectly appropriate for me to remain thus unmoved. But now recall that I am your closest possible epistemic and doxastic duplicate. If so and if, in Inquiry 2*, it is perfectly appropriate for me to remain unmoved by the argument, then you too should remain unmoved by it. Finally, any argument for a proposition, p, that you should remain unmoved by is also one that does not properly close the question whether p for you. As a result, your inquiry in Inquiry 2*, which exploits just this argument, does not properly close the question whether ~CDM for you. And, of course, this is all I need to get my argument against Transmission of Knowledge off the ground.

5. The Competition Of course, there is no shortage of responses to the sceptical problem in general and the Argument from Ignorance in particular in the literature. In line with the general policy of this book, I will look at how my account compares with the closest competitor. I take this to be sensitivity-based approaches to scepticism, which, like my own approach, deny Transmission of Knowledge. In what follows, I will compare the two approaches and provide some reason for favouring mine.

5.1 Sensitivity Very roughly, according to sensitivity accounts of knowledge, one knows that p if and only if one sensitively believes that p (Dretske 1970, Nozick



185

1981). And, again very roughly, one’s belief that p is sensitive if and only if at the closest worlds at which p false, one does not believe p. It’s easy to see that Transmission of Knowledge will fail on sensitivity accounts. Just consider Zebra once more. Your belief that Z is sensitive: at the closest worlds at which Z is false, the enclosure is empty or you are standing in front of a different enclosure in which case you are looking at some other kind of animal instead. In that case, you do not believe that Z. Since your belief that Z is sensitive, sensitivity accounts will predict that you know Z. Now suppose you competently deduce and thereupon come to believe that ~CDM. Is that belief sensitive as well? No. After all, at the closest worlds at which ~CDM false, you are looking at a cleverly disguised mule, in which case you still believe ~CDM (based on your competent deduction from Z). Since your belief that ~CDM isn’t sensitive, sensitivity accounts will predict that you don’t know ~CDM. Transmission of Knowledge fails.¹³

5.2 Problems for Sensitivity Abominable Conjunctions and Conversational Patterns. First, sensitivitybased approaches to transmission failure are committed to abominable conjunctions and problematic conversational patterns. Here is why. Unlike my preferred approach, sensitivity-based approaches to transmission failure do not leave open the question of what should be said about AI1 and AI3. Rather, sensitivity-based approaches are firmly committed to holding that AI1 is true and AI3 is false. If this isn’t immediately obvious, note that your belief that you are not a handless BIV is not sensitive: if you were a handless BIV, you’d still believe that you aren’t. Hence AI1 comes out true. At the same time, your belief that you have hands is sensitive: if you didn’t have hands, say, because you lost them in an accident, you would be painfully aware of this fact and so you

¹³ Not all versions of sensitivity accounts are committed to the failure of Transmission of Knowledge (e.g. Roush 2005). However, since my aim here is to compare my account of transmission failure with its most prominent rival, these views can safely be set aside, at least for present purposes.

186 , ,   wouldn’t believe that you have hands. Therefore, you know that you have hands. AI3 comes out false. In fact, on sensitivity-based approaches to transmission failure, it is precisely the fact that claims like AI1 come out true and claims like AI3 come out false that explains why Transmission of Knowledge fails to begin with. As a result, the truth of AI1 and the falsity of AI3 are not accidental features of sensitivity-based approaches to transmission failure but rather lie at their very heart. But given that claims like AI1 will come out true and claims like AI3 false on sensitivity-based approaches to transmission failure, it is easy to see that sensitivity-based approaches will be committed to abominable conjunctions and problematic conversational patterns. For instance, they will be committed to Abomination, i.e. ‘I know that I have hands, but I don’t know that I am not a handless BIV.’ And given the knowledge norm of assertion, they will also require you to say something like ‘I can’t say’ when asked whether you are a handless BIV after having affirmed that you have hands and that this entails that you are not a handless BIV. The Crazy Argument from Ignorance. To bring out yet another downside of sensitivity’s treatment of scepticism, consider next the following case: Milk You have just put a half-full carton of milk back in the fridge. Since you have been looking at the fridge ever since and nobody has opened the fridge in the meantime, you know that there is milk in the fridge.

Now consider a variation of the Argument from Ignorance: The Crazy Argument from Ignorance CAI1. You don’t know that (~IMF =) it’s not the case that invisible milk fairies have entered the fridge just before you closed it and by now have drunk all the milk that was in the carton. CAI2. If you don’t know that ~IMF, then you don’t know that there is milk in the fridge. CAI3. You do know that there is milk in the fridge.



187

It is easy to see that advocates of sensitivity will have to treat the Crazy Argument from Ignorance in the same way as the Argument from Ignorance. That is to say, just as they maintain that AI2 is false and AI1 and 3 are true, here too they will have to hold that CAI2 is false, while CAI1 and 3 are true. After all, your belief that there is milk in the fridge is sensitive, but your belief that the milk hasn’t been drunk by invisible milk fairies isn’t.¹⁴ Crucially, however, there is an important intuitive difference between the two cases. While AI1 is indeed intuitively plausible, the same does not go for CAI1. On the contrary, it is intuitively highly plausible that you know that ~IMF is false. The Alternative Deduction. To see the last problem for sensitivitybased approaches to transmission failure that I will discuss here, consider the following case by John Hawthorne (2005): Inquiry 3 You are at the zoo. Currently you are standing in front of the zebra enclosure and see a black-and-white striped equine creature inside. Since you can tell a zebra from the way it looks, you come to know that Z. You now want to find out whether ~CDM. In order to achieve this, you first exploit the equivalence between Z and Z & ~CDM to competently deduce Z & ~CDM and then apply conjunction elimination to arrive at the belief that ~CDM.

It’s easy to see that the above argument that knowledge fails to transmit across competent deduction in Inquiry 2 serves to establish the same result for Inquiry 3.¹⁵ It’s also easy to see that sensitivity-based accounts of knowledge can secure this result. After all, your belief that ~CDM is no more sensitive here than in the earlier Inquiry 2. Hence, you don’t know ~CDM here either. So far, so good.

¹⁴ If this isn’t immediately obvious, consider: were there to be no milk in the fridge, say, because you have not yet put it back, you wouldn’t believe there is. In contrast, had the milk been drunk by invisible milk fairies, you’d still believe that it hadn’t. ¹⁵ This is unsurprising once it is noted that the only difference between Inquiry 2 and Inquiry 3 is that the competent deduction of ~CDM from Z proceeds via a different route and the above argument simply doesn’t depend on the specific route of deduction.

188 , ,   Of course, if knowledge fails to transmit in Inquiry 3, we’d expect it to do so at some specific point in your deduction, i.e. either when you deduce Z & ~CDM from Z or when you deduce ~CDM from Z & ~CDM. The question is where exactly this point is. Say you are convinced that Transmission of Knowledge fails in cases like Inquiry 2 in which your competent deduction exploits the entailment from Z to ~CDM. If so, you will want to say that, in Inquiry 3, knowledge fails to transmit at the first step of the deduction, which exploits the equivalence between Z and Z & ~CDM. There are at least three reasons for this. First, it is independently plausible that if you cannot come to know that ~CDM via competent deduction from known Z, then the same goes for Z & ~CDM. Second (and relatedly), ~CDM is logically weaker than Z & ~CDM. As a result, if knowledge doesn’t transmit across competent deduction to ~CDM, then it would be at least surprising if it did transmit to the stronger Z & ~CDM. Third, the thesis that knowledge is closed under competent conjunction elimination is extremely plausible. Already the stronger principle that if one knows a conjunction, then one knows each conjunct is nearly universally accepted. Rejecting this principle will just be too big a bullet to bite. The problem for sensitivity theorists is that they are committed to holding that the deduction fails to transmit at the conjunction elimination step, rather than at the step that exploits the equivalence. The reason for this is that if your belief that Z is sensitive, then so is your belief that Z & ~CDM arrived at via competent deduction. After all, the closest worlds at which Z & ~CDM is false are just the worlds at which Z is false. According to sensitivity accounts, then, if you know Z and come to believe that Z & ~CDM via competent deduction from Z, then you also know that Z & ~CDM. Sensitivity accounts predict that knowledge does transmit across competent deductions that exploit the equivalence between Z and Z & ~CDM. So, the problem must lie with the conjunction elimination step. And that’s the wrong result.

5.3 Can We Do Better? Can my approach do better? In what follows, I will argue that it can. Since we have already seen that my preferred approach to transmission



189

failure can avoid the problem posed by abominable conjunctions and problematic conversational patterns, I will not return to these issues here. What’s more, I take it that the discussion of how to resist AI1 lays out a clear path to how the present proposal will deal with the Crazy Argument from Ignorance: the epistemic ability to recognize that certain possibilities could not easily obtain allows you to know not only that you are not a handless BIV but also that ~IMF. This means that there is only one problem remaining, to wit, the problem of the alternative deduction. Since there is no obvious solution to this problem on the table yet, I will take a closer look at this one. As a first observation, note that my own approach does not entail anything about which of the two steps in Inquiry 3 is the problematic one. It only tells us that knowledge does not transmit from Z to ~CDM and hence that at least one step must be problematic. On the upside, this means that my account is at least compatible with the right result here. At the same time, it would be nice to be able to do more than this, i.e. to secure the correct result rather than merely be compatible with it. Fortunately, this can be done also. Here is how. First, my own approach only specifies a set of propositions such that knowledge doesn’t transmit across competent deduction to members of this set. While this is too coarse-grained for specifying which step in a multistep deduction is the problematic one, it is all we need when we are dealing with one-step deductions. After all, if the deduction features only one step, then it’s clear exactly where transmission failure occurs. For instance, in Inquiry 2, the deduction proceeds in a single step, which exploits the entailment from Z to ~CDM. Here, it’s clear that this is where transmission failure occurs. While it would be nice to have a fully general account of just where knowledge fails to transmit in multistep deductions, I must confess that I don’t have one. However, I have the following proposal for a partial account: Transmission Failure If knowledge fails to transmit across a competent deduction that exploits a certain logical relationship between the members of a certain set of propositions, then it also fails to transmit across any competent

190 , ,   deduction that exploits an a priori equivalent logical relationship between the members of this set.

For instance, if knowledge fails to transmit across any competent deduction that exploits the entailment from p to q, then it also fails to transmit across any competent deduction that exploits a logical relationship a priori equivalent to the entailment from p to q. This much seems plausible enough. Now, we have already seen that knowledge fails to transmit across competent deduction from Z to ~CDM in Inquiry 2. We have also seen that, in Inquiry 2, the deduction is a one-step deduction, which exploits the entailment from Z to ~CDM. This means that the deduction fails at the step that exploits the entailment. Crucially, the entailment from Z to ~CDM is a priori equivalent to the equivalence between Z and Z & ~CDM, i.e. (Z ! ~CDM)  (Z  (Z & ~CDM)). By Transmission Failure, it follows that knowledge fails to transmit across any competent deduction that exploits the equivalence between Z and Z & ~CDM. In this way, the present approach can secure the desired result.

6. Conclusion This chapter has taken a close look at scepticism. More specifically, I focused on what is perhaps the most difficult sceptical argument, the Argument from Ignorance. This argument features two highly plausible premises: that we don’t know that certain sceptical hypotheses are false and that, if so, we don’t know many ordinary propositions we take ourselves to know. If sound, it will establish that we know much less than we take ourselves to know. In fact, we know very little at all. I have argued that, whilst attractive at first glance, both premises of the Argument from Ignorance are problematic. More specifically, the transmission principle that underlies the second premise is false. In fact, the distinctively inquiry-based approach to epistemological theorizing helped to bring this out. What’s more, contrary to the first premise, we can know the denials of sceptical hypotheses. Crucially, the standard view of how we can know them is mistaken. It’s not that we can know



191

them by competent deduction from ordinary propositions we know. Rather, we can have recognitional knowledge of them through the exercise of an ability to come to know that certain possibilities could not easily obtain. Once again, I finished the chapter by responding to a range of objections and showing that my own view compares favourably with its closest competitor, the sensitivity-based approach to scepticism.

APPENDIX 1

In Chapter 4, I explored the claim that the aim of inquiry into general phenomena is understanding and developed a knowledge-based account of understanding. While we saw some reason for thinking that the aim of inquiry into general phenomena requires knowledge, I mentioned that more detailed arguments for this claim can be given. More specifically, I indicated that the arguments that inquiry into specific questions requires knowledge from Chapter 1 can be adapted to bring home the point about general phenomena also. In what follows, I show how this can be done.

1. Commitment Release It’s easy enough to see that one can run a similar argument via Commitment Release for understanding. Recall: Commitment Release If, at t, one attains the aim of a given activity with an aim, then, at t, one is released from all commitments towards attaining this aim. And now consider: The Hire 2 You are a geologist. I have hired you for two weeks to find out about (P =) the value of a certain mine I am considering buying. Since I need to be in a remote location with no means of communication for the next two weeks, we agree to meet at the mine two weeks from now. Thanks to a contact in the government you are able to access a classified and full report on the value of the mine, which includes detailed information about the fake diamonds, the real diamonds, and so on. On the basis of this report, you come to believe the full account of P. Since there is still a considerable amount of time before our meeting, you and your team pack your bags and get on the next flight home to spend time with your families. Meanwhile, I return unexpectedly early from my trip to the breaking news that the report is a fabrication of a corrupt government official who wanted to climb the career ladder. You are currently back home with your family and entirely unaware of the news. The final twist in the story is that, unbeknownst to everyone, despite being a fabrication, the full account of P the report gives is entirely true. To see how the argument will go, note first that you justifiably and truly believe in the full account of P. If we analyse the aim of inquiry into general phenomena in terms of true belief or justified belief, Commitment Release entails that you are released from your contractual commitment to find out about P. However, again, that’s the wrong result. Your contractual commitment is still binding, as is evidenced by the fact that I can take steps to get you back to work

194   without negotiating a new contract. In contrast, if we analyse the aim of inquiry into general phenomena in terms of knowledge, we will do better. Since your beliefs about P are all gettiered, they fall short of knowledge. According to a knowledgebased account of the aim of inquiry into general phenomena, it does not follow that you are released from your contractual commitment. In this way, there is further reason to favour it.

2. Progress Finally, let’s look at the argument from Progress. Recall: Progress If, at t2, one has not attained the aim of a given activity with an aim and if one makes progress towards attaining its aim between t1 and t2, then one has not attained its aim at t1 either. And now consider: The Insight 2 You are a geologist. I have hired you for two weeks to find out about (P =) the value of a certain mine I am considering buying. Since I need to be in a remote location with no means of communication for the next two weeks, we agree to meet at the mine two weeks from now. Thanks to a contact in the government, at t1, you are able to access a classified and full report on the value of the mine, which includes detailed information about the fake diamonds, the real diamonds, and so on. On the basis of this report, you come to believe the full account of P. At t2, however, you discover that the report is a fabrication of a corrupt agent who wanted to climb the career ladder. So, you start again. At t3, you have discovered the fake diamonds and the real diamonds in the mine. Now here is what I take to be an overwhelmingly plausible claim about the Insight 2: you make progress on your inquiry into P between t2 and t3 (henceforth also ‘the crucial claim 2’). By way of support for the crucial claim 2 notice, first, that, at t2, you have discovered that a certain account of P, at t1, is a mere fabrication and so you have to start from scratch. But that also means that, at t3, once you have discovered the fake diamonds and the real ones, you have made progress. The second reason why it is better to analyse the aim of inquiry into general phenomena in terms of knowledge than in terms of true belief or justified belief is that only the knowledge-based account is compatible with the crucial claim 2. Here is why: according to the crucial claim 2, you made progress on the inquiry into P between t2 and t3. By Progress, it follows that at t1 you did not attain success in your inquiry into P. However, again, we may think of the case as one in which you justifiably and truly believed the full account of P. According to any account of the aim of inquiry into general phenomena that analyses it in terms of true belief or justified belief, then, you did attain the aim of inquiry into P at t1. As a result, these accounts are unable to accommodate the crucial claim 2. In contrast, a knowledge-based

 

195

account encounters no difficulties here. After all, at t1, your beliefs about P are gettiered and so do not qualify as knowledge. According to a knowledge-based account of the aim of inquiry into general phenomena, you do not attain the aim of your inquiry into P at t1. As a result, a knowledge-based account is entirely compatible with the crucial claim. Given the plausibility of this claim, this means that there is yet further reason to think that the aim of inquiry into general phenomena requires knowledge.

APPENDIX 2

Chapter 3 develops a view about conditions on knowledge which accords abilities an important role. This is by no means a new idea. In fact, it is at the very heart of virtue epistemology which is one of the most flourishing research programmes in contemporary epistemology.¹ In previous work, I have defended versions of virtue epistemology myself, including a distinctively knowledge first version of virtue epistemology. In this appendix, I will say a few words about how the views of knowledge and justified belief that I developed in Chapters 2 and 3 relate to virtue epistemology in general and my earlier work on knowledge first virtue epistemology in particular. More specifically, I will argue that, while the view is compatible with a virtue epistemology, I now think that it is best not thought of as such.

1. Virtue Epistemology Key to virtue epistemology are (i) a theory of the normativity of performances; (ii) the claim that beliefs are performances with an epistemic aim; and (iii) the identification of key epistemological properties with normative properties of beliefs as performances. In what follows, I will briefly explain all three elements, starting with the theory of performance normativity. According to this theory, we can always ask whether a performance is successful, i.e. whether it attains its aim; whether it is competent, i.e. whether it is produced by an ability to attain its aim; and whether it is apt, i.e. whether it is successful because competent. Accordingly, performances can be assessed along three key dimensions: success, competence, and aptness. When virtue epistemologists say that beliefs are performances with an epistemic aim, they typically hold that the aim of belief is truth. Given that this is so, beliefs can be assessed along the aforementioned three dimensions, as successful (true), competent (produced by an ability to form true beliefs), and apt (true, because competent). Finally, virtue epistemologists typically identify key epistemological properties with normative properties of beliefs as performances. In particular, they typically identify justified belief with competent belief and knowledge with apt belief. The distinctively knowledge first version of virtue epistemology I developed in earlier work (e.g. Kelp 2016, 2017, 2018b) took all of (i)–(iii) on board. The most

¹ For recent virtue epistemologies, see, e.g., Broncano-Berrocal (2017), Carter (2016), Greco (2010), Kelp (2018b), Miracchi (2015), Pritchard, Millar, and Haddock (2010), Riggs (2002), Sosa (2015), Turri (2016b), and Zagzebski (1996). Recent collections on virtue epistemology include Battaly (2018), Fernandez Vargas (2016), Greco and Turri (2012), and Kelp and Greco (2020).

 

197

important difference from standard virtue epistemology is that knowledge takes the place of truth as the aim of belief. This difference may appear puzzling. After all, if knowledge is the relevant kind of epistemic success, how can knowledge be identified with apt belief, as (iii) would have it? To get the answer into clearer view, let’s first consider my preferred account of aptness: Apt ACAN Move An ACAN move is apt if and only if it is (i) successful, (ii) competent, and (iii) the EN of the ability exercised are satisfied. Next, recall the conditions on knowledge from Chapter 3 that I also defended in the earlier work: Conditions on Knowledge One knows that p only if (K1 =) one’s belief that p is produced by an exercise of an ability to know propositions in range R and relative to conditions C such that p∈R, and (K2 =) the EN of the ability to know that produced one’s belief that p are satisfied. It is easy to see that, given Apt ACAN Move and Conditions on Knowledge, a belief is successful if and only if it is apt. In this way, the virtue-epistemological identification of knowledge with apt belief turns out to be entirely compatible with the knowledge first claim that belief is a kind of performance that aims not at truth but at knowledge. It may also be worth noting that, as a result, the views of knowledge and justified belief developed in Chapters 2 and 3 can be understood along virtue-epistemological lines. More specifically the substantive constraints on means and environment on the aim of inquiry from Chapter 3 are tantamount to a virtue-epistemological aptness constraint on the aim of inquiry. Now, a couple of potential hiccups remain. First, one might still wonder whether the identification of knowledge and apt belief isn’t incompatible with my claim that belief is a state from Chapter 2. After all, isn’t this claim incompatible with something that the identification entails, i.e. that belief is a performance? By way of response, recall the possibility of dual natures which we already encountered in Chapter 2. This possibility may be realized here too. Like quantic entities which have both a particle and a wave nature, belief may have both a stative and a performative nature. As a result, although it is part of my view that belief is a state, the view remains compatible with virtue epistemology. Second, one may also wonder whether the identification of knowledge and apt belief doesn’t lead us back to the kind of dismantling analysis that I claimed to have abandoned. Fortunately, the answer to this question is no. The claim that apt belief is knowledge cannot be developed into a dismantling analysis of knowledge. This is because aptness is unpacked in terms of knowledge and so any attempt at a dismantling analysis is bound to fall foul of the non-circularity requirement on dismantling analyses. In this way, the identification of apt belief and knowledge that the present account delivers leaves the question about the nature of knowledge wide open. The network analysis can still provide the answer.

198  

2. Back-Pedalling There is reason to think that the views of justified belief and knowledge from Chapters 2 and 3 are entirely compatible with my earlier work on knowledge first virtue epistemology. Even so, I am now less convinced that interpreting the present view as a virtue epistemology, never mind a knowledge first virtue epistemology, is the right thing to do. Let me explain. Knowledge First. Recall that I already mentioned that there are important differences between my account and knowledge first epistemology. Most importantly, knowledge first epistemology takes the distinction between knowledge and ignorance as the starting point for epistemological theorizing. As a result, that knowledge is of key significance in epistemology is not optional on a knowledge first approach. In contrast, my account takes inquiry as the starting point for epistemological theorizing. That knowledge is of central importance is a function of the methodological idea in conjunction with a set of arguments. Knowledge must earn its keep as something of key significance. That it does so is by no means a foregone conclusion in the present framework. While I do not mean to deny that, on the present interpretation, my view has a number of features that are congenial to knowledge first epistemology, in view of these differences, I think that it would be a mistake to think of the view as knowledge first epistemological proper. And this goes for the view I defend here, as well as the view I defended in earlier work (e.g. Kelp 2016, 2017, 2018b). Beliefs as Performances. What’s more, there is reason for thinking that the view is not best understood as a version of virtue epistemology either. This is because there is reason not to take at least two of the three key elements of virtue epistemology—i.e. (i) the theory of the normativity of performances and (ii) the claim that beliefs are performances with an epistemic aim—on board. Let’s look at (ii) first. I said that my view is compatible with the virtue-epistemological idea that belief is a kind of performance. After all, while on my view belief is a state, it might be that belief has both a state and a performance nature. The question remains, however, whether it is a good idea to embrace the view that belief has a dual nature. One reason against is that it is less simple than the alternative view according to which belief is a state only. That said, the view that belief is a performance has met with considerable resistance on independent grounds. More specifically, it has been argued that belief is a kind of state and not kind of a performance.² And, of course, if this objection is on target, there is reason for thinking that we will do well not to take the key virtueepistemological claim that belief is a kind of performance with an aim on board.³ ² The most prominent version of this objection is due to Matthew Chrisman (2012, 2017). In a nutshell, Chrisman uses a typology of English verb phrases from Kenny (1963), Mourelatos (1978), and Vendler (1957) to argue that knowledge attributions are statives and that, as a result, knowledge attributions are about states and not performances. ³ One might wonder whether it is really optional on my view that belief is a kind of performance. After all, I do take justified belief to be competent belief. And doesn’t that commit me to the claim that beliefs are performances? No. Consider sitting. Sitting is a state. In keeping

 

199

The Centrality of Aptness. Another worry is with the theory of performance normativity that virtue epistemology endorses. To get this worry into clearer view, note first that, on this theory of performance normativity, aptness is a central normative category. In fact, there is reason to think that aptness is the most central normative category. This is because aptness is the fully desirable status of performances and any performance that falls short of aptness is a performance that falls short qua performance (cf. Sosa 2015). (I will hereafter refer to this thesis as ‘the centrality of aptness’.) Note just how important the centrality of aptness is for virtue epistemology. It makes a significant contribution to motivating the identification of apt belief and knowledge and is key to the virtue-epistemological solution to the Meno problem. If aptness is the fully desirable status of performances, it makes sense to identify apt belief and knowledge. After all, it gives us the attractive result that knowledge is the fully desirable status of belief, which goes a long way towards explaining the centrality of knowledge in epistemology, among other things. What’s more, if a performance that falls short of aptness is a performance that falls short qua performance, then a belief that falls short of knowledge is a belief that falls short qua belief. No surprise, then, that a belief that qualifies as knowledge is more valuable than a belief that falls short of knowledge. With these points in play, let’s consider the motivations for the centrality of aptness. The thought is that in standard cases of human performance there appears to be a genuine normative difference between producing a merely successful performance and an apt one. Consider a case in which you are taking a shot in ARCH. It makes sense for it to matter to you whether your shot is merely successful or actually apt. Here is perhaps the easiest way to see this. Suppose you are about to take a shot that’s very important to you, say, because it will decide whether or not you make the archery team that it has been your dream to join. You produce a competent shot, it hits the target, and you are on the team. At the same time, you had a helper in the wings who acted on the arrow to make it hit the target. Without their help, your shot would have missed. While your shot is successful and competent, it was not apt. Later on, you discover all of this. The key observation about this case is that it matters.

with that, one may or may not have the ability to sit and one may or may not have the ability to sit straight, for hours on end, and so on. These abilities are relative to conditions of shape and environment, and so on. Moreover, states can feature in ACANs. You may compete in a competition to see who can sit longest. Now certain abilities to sit, i.e. abilities to sit for prolonged periods of time, become relevant. We can ask whether you won the competition, whether you sat competently (i.e. as a result of the exercise of an ability to sit rather than, say, paralysis), and so on. Finally, states can feature in ACANs with normative success conditions. You may compete in a try-out for the sitting competition team. Now successes require you not only to sit still but to do so competently in suitable conditions (see Chapter 3). Abilities have become essential to success. None of this changes the fact that sitting is a state. Likewise, the fact that one may believe a proposition because of the exercise of a certain ability, that we can ask whether one’s belief is successful (i.e. qualifies as knowledge) or whether it was competent, and that success in inquiry requires competent belief in suitable environmental conditions are entirely compatible with the claim that belief is a state, as is, of course, the proposed identification of competent belief with justified belief.

200   It makes sense for us to feel that you have fallen short, and perhaps even that you don’t really deserve to be on the team. These considerations suggest that aptness has a normative significance that goes beyond that of success. By the same token, aptness seems to deserve its place as a (and perhaps the) central normative category in the theory of the normativity of performances. The reason I have started to harbour doubts about the centrality of aptness is that there are cases in which aptness does not seem to have additional normative significance. Consider the case of a frog which has the reliable ability to catch flies with its tongue. Crucially, we must take this ability to be triggered automatically rather than by intention. Luckily, this seems plausible enough anyway. Unfortunately for the frog, the environment has recently been taken over by a new species of fly that is considerably faster than the kind of fly that used inhabit it. In fact, at its current speed, the frog’s tongue won’t be fast enough to catch enough flies of the new fast kind. Luckily, there also is an unusual kind of gas in the environment which slows the flies down. Just now, the frog rolls out its tongue to catch a fly. Even if our frog succeeds, as we may, of course, assume it does, it will not succeed aptly. After all, unsuitable environmental conditions will prevent this from happening. Now here is the key question. Does this matter? Does it make sense for us to feel that the frog has fallen short and that he does not deserve, say, to no longer go hungry, etc.? In my view, the answer to these questions is a clear no. But if so, we may start to wonder whether the property of aptness captures anything of distinctive normative significance and, as a result, whether it really deserves its place as a central normative category in the theory of the normativity of performances. After all, there are cases in which whether or not a performance is apt simply does not matter/does not correspond to a shortcoming of the agent. It might be thought that so far all we have is evidence that leaves it open whether aptness is a central normative category. Some cases suggest that it is; others don’t. Crucially, however, there is an alternative account of the cases that seem to support the centrality of aptness, which appeals only to success and not to aptness. To see this, recall first that the standard motivation for the centrality of aptness comes from standard cases of human performance. Of course, standard cases of human performance are cases of intentional performance. For instance, in the above archery case, you intentionally produce a shot that turns out to be successful. But now note that you cannot perform without performing in a certain way. What’s more, when it comes to intentional performances, you cannot perform intentionally without performing intentionally in a certain way. For instance, you cannot intentionally take a shot in ARCH without intentionally doing so in a certain way (e.g. releasing the arrow thus). As a result, when intentional performances have an aim, their aim consists not only in bringing about the success of the relevant performance type, but also in doing so in the intended way. For intentional performances with an aim, then, we have two dimensions of success: the first is whether the performance attains its aim and the second is whether the performance attains its aim in the intended way. This gives us the resources to offer an alternative account of the kinds of cases that are typically adduced to motivate the centrality of aptness, such as the above archery case in which your shot is off target but successful anyway because of the intervention

 

201

of a helper in the wings. The performance attains its first aim: your shot finds the target. However, it does not attain its second aim: it does not find the target in the intended way. In this way, we have an attractive alternative explanation of the shortcomings of agents in cases of successful but inapt intentional performance. Crucially, this alternative explanation invokes only success and does not appeal to aptness at all.⁴ Where does this leave us? At first glance, standard cases of intentional human performances appear to suggest that there is a genuine normative difference between merely successful and apt performances which would justify the centrality of aptness. On reflection, however, there is reason to think that appearances are misleading. Once we move away from intentional performances and focus on automatic ones, it becomes hard to deny that aptness is not of central normative significance. What’s more, there is an attractive alternative account of cases motivating the centrality of aptness that doesn’t appeal to aptness. But, of course, if the centrality of aptness is false, virtue epistemology runs into even more trouble. After all, it now looks as though it is based on a flawed normative framework. And that, I take is a serious drawback for the view. In contrast, my own view is in no way committed to the centrality of aptness. On my view, knowledge is first and foremost the aim of inquiry. If there is any normative property of belief as performance that explains the normative significance of knowledge, it is success. And since the normative significance of success is not in question, the problem that virtue epistemologists encounter by endorsing a theory of performance normativity featuring the centrality of aptness can be laid to rest as well.

3. Conclusion In sum, the view of knowledge and justified belief I developed in Chapters 2 and 3 can be seen as a version of virtue epistemology, perhaps even the distinctively knowledge first variety thereof that I defended in earlier work (e.g. Kelp 2016, 2017, 2018b). That said, it transpires that there is reason not to do so. There are significant differences between my own view and both knowledge first epistemology and virtue epistemology. My view differs from knowledge first epistemology in that it doesn’t take the distinction between knowledge and ignorance as a starting point for epistemological theorizing, with the result that knowledge must earn its keep as

⁴ Note that it is compatible with this that there are cases in which aptness matters. In particular, the cases I have in mind are cases in which a performance is successful if and only if it is apt. Two things are worth bearing in mind here. First, aptness plausibly matters here only because success matters: it is not as if aptness has normative significance over and above success. Second, it’s not clear that standard virtue epistemology can allow this way in which aptness matters. After all, standard virtue epistemology typically doesn’t countenance the possibility of a performance that is successful if and only if apt. In fact, as we have seen earlier, some standard virtue epistemologists resist the thought vehemently.

202   something significant in epistemology. And it differs from virtue epistemology in that it is neither committed to the claim that belief is a kind of performance nor to virtue epistemology’s key theory of performance normativity, which features the centrality of aptness. In view of these considerations, I think that it might be wisest to curb my earlier enthusiasm for knowledge first virtue epistemology and to reconsider my allegiances to both knowledge first epistemology and to virtue epistemology.

Bibliography Adler, J. 2002. Belief ’s Own Ethics. Cambridge, MA: MIT Press. Battaly, H., ed. 2018. The Routledge Handbook of Virtue Epistemology. London and New York: Routledge. Bird, A. 2007a. ‘Justified Judging.’ Philosophy and Phenomenological Research 74: 81–110. Bird, A. 2007b. ‘What Is Scientific Progress?’ Noûs 41: 64–89. Broncano-Berrocal, F. 2017. ‘A Robust Enough Virtue Epistemology.’ Synthese 194: 2147–74. Buckwalter, W., and Stich, S. 2014. ‘Gender and Philosophical Intuition.’ In Experimental Philosophy, ii, edited by J. Knobe and S. Nichols. New York: Oxford University Press 307–46. Carnap, R. 1967. The Logical Structure of the World and Pseudoproblems in Philosophy. Berkeley, CA: University of California Press. Carter, A. 2016. ‘Robust Virtue Epistemology as Anti-Luck Epistemology: A New Solution.’ Pacific Philosophical Quarterly 97: 140–55. Carter, A., Jarvis, B., and Rubin, K. 2015. ‘Varieties of Cognitive Achievement.’ Philosophical Studies 172: 1603–23. Chrisman, M. 2012. ‘The Normative Evaluation of Belief and the Aspectual Classification of Belief and Knowledge Attributions.’ Journal of Philosophy 109: 588–612. Chrisman, M. 2017. “Performance Normativity and Here-and-Now Doxastic Agency. Synthese. doi.org/10.1007/s11229-017-1641-3. Coliva, A. 2012. ‘Varieties of Failure (of Warrant Transmission: What Else?!).’ Synthese 189: 335–54. Coliva, A. 2015. Extended Rationality. A Hinge Epistemology. Basingstoke: Palgrave Macmillan. Craig, E. 1990. Knowledge and the State of Nature. An Essay in Conceptual Synthesis. Oxford: Clarendon Press. David, M. 2001. ‘Truth as the Epistemic Goal.’ In Knowledge, Truth, and Duty, edited by M. Steup. Oxford: Oxford University Press 151–169. Davidson, D. 2005. Truth, Language and History. Oxford: Clarendon Press. Davis, M. 2009. ‘Two Purposes of Arguing and Two Epistemic Projects.’ In Minds, Ethics, and Conditionals: Themes from the Philosophy of Frank Jackson, edited by I. Ravenscroft. Oxford: Oxford University Press. DeRose, K. 1995. ‘Solving the Skeptical Problem.’ Philosophical Review 104: 1–52. DeRose, K. 2002. ‘Assertion, Knowledge, and Context.’ Philosophical Review 111: 167–203.

204  Douven, I. 2009. ‘Assertion, Moore and Bayes.’ Philosophical Studies 144: 361–75. Dretske, F. 1970. ‘Epistemic Operators.’ Journal of Philosophy 67: 1007–23. Elgin, C. 2009. ‘Is Understanding Factive?’ In Epistemic Value, edited by A. Haddock, A. Millar, and D. Pritchard. Oxford: Oxford University Press 322–330. Fantl, J., and McGrath, M.2009. Knowledge in an Uncertain World. Oxford: Oxford University Press. Feldman, R. 2002. ‘Epistemological Duties.’ In The Oxford Handbook of Epistemology, edited by P. Moser. New York: Oxford University Press 362–84. Fernandez Vargas, M., ed. 2016. Performance Epistemology: Foundations and Applications. Oxford: Oxford University Press. Friedman, J. 2019. ‘Inquiry and Belief.’ Noûs 53: 296–315. Geach, P. 1956. Good and Evil. Analysis 17: 32–42. Gettier, E. 1963. ‘Is Justified True Belief Knowledge?’ Analysis 23: 121–3. Goldman, A. 1979. ‘What Is Justified Belief?’ In Justification and Knowledge, edited by G. Pappas. Dordrecht: Reidel 1–23. Goldman, A. 2012. Reliabilism and Contemporary Epistemology. New York: Oxford University Press p.40. Greco, J. 2010. Achieving Knowledge. Cambridge: Cambridge University Press. Greco, J., and Turri, J., eds. 2012. Virtue Epistemology: Contemporary Readings. Cambridge, MA: MIT Press. Harman, G. 1986. Change in View. Cambridge, MA: MIT Press. Hawthorne, J., and Stanley, J. 2008. ‘Knowledge and Action.’ The Journal of Philosophy 105: 571–90. Hawthorne, J. 2005. The Case for Closure. In Steup, M. & Sosa, E. eds. Contemporary Debates in Epistemology. Blackwell, 26-43. Healey, R. 1991. The Philosophy of Quantum Mechanics: An Interactive Interpretation. Cambridge: Cambridge University Press. Hempel, C. G. 1965. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. New York: Free Press. Hieronymi, P. 2009. ‘Two Kinds of Agency.’ In Mental Actions, edited by L. O’Brien and M. Soteriou, 138–62. Oxford: Oxford University Press. Hirvelä, J. 2019. ‘Knowing without Having the Competence to Do So.’ Thought: A Journal of Philosophy 8: 110–18. Huemer, M. 2007. ‘Moore’s Paradox and the Norm of Belief.’ In Themes from G. E. Moore: New Essays in Epistemology and Ethics, edited by S. Nuccetelli and G. Seay. Oxford: Clarendon Press 142–57. Jenkins Ichikawa, J. 2014. ‘Justification Is Potential Knowledge.’ Canadian Journal of Philosophy 44: 184–206. Johnson, R. and Cureton, A. 2019. Kant’s Moral Philosophy, The Stanford Encyclopedia of Philosophy. Edward N. Zalta (ed.), URL = . Johnson, R., and Cureton, A. 2019. ‘Kant’s Moral Philosopy.’ In The Stanford Encyclopedia of Philosophy, edited by E. Zalta, https://plato.stanford.edu/archives/ spr2019/entries/kant-moral/, accessed 12 September 2020.



205

Kaplan, M. 1985. ‘It’s Not What You Know That Counts.’ Journal of Philosophy 82: 350–63. Kappel, K. 2010. ‘On Saying That Someone Knows: Themes from Craig.’ In Social Epistemology, edited by A. Haddock, A. Millar, and D. Pritchard. Oxford: Oxford University Press 69–88. Kelp, C. 2011. ‘What’s the Point of “knowledge” Anyway?’ Episteme 8: 53–66. Kelp, C. 2016. ‘Justified Belief: Knowledge First-Style.’ Philosophy and Phenomenological Research 93: 79–100. Kelp, C. 2017. ‘Knowledge First Virtue Epistemology.’ In Knowledge First: Approaches in Epistemology and Mind, edited by A. Carter, E. Gordon, and B. Jarvis. Oxford: Oxford University Press 223–45. Kelp, C. 2018a. ‘Assertion: A Function First Account.’ Noûs 52: 411–42. Kelp, C. 2018b. Good Thinking. A Knowledge First Virtue Epistemology. London and New York: Routledge. Kelp, C., and Greco, J., eds. 2020. Virtue-Theoretic Epistemology: New Methods and Approaches. Cambridge: Cambridge University Press. Kelp, C., and Simion, M.2017. ‘Commodious Knowledge.’ Synthese 194: 1487–502. Kelp, C., and Simion, M. 2018. ‘The C Account of Assertion: A Negative Result.’ Synthese 197: 125–37. doi.org/10.1007/s11229-018-1760-5. Kenny, A. 1963. Action, Emotion, and Will. London: Routledge and Kegan Paul. Khalifa, K. 2013b. ‘Understanding, Grasping, and Luck.’ Episteme 10: 1–17. Khalifa, K. 2017. Understanding, Explanation and Scientific Knowledge. Cambridge: Cambridge University Press. Kitcher, P. 1989. ‘Explanatory Unification and the Causal Structure of the World.’ In Scientific Explanation, edited by P. Kitcher and W. Salmon. Minneapolis, MN: University of Minnesota Press 410–505. Kvanvig, J. 2003. The Value of Knowledge and the Pursuit of Understanding. Cambridge: Cambridge University Press. Kvanvig, J. 2014. ‘Truth Is Not the Primary Epistemic Goal.’ In Contemporary Debates in Epistemology, ii, edited by M. Steup, J. Turri, and E. Sosa. Malden/ MA: Wiley-Blackwell 352–62. Lewis, D. 1970. ‘Anselm and Actuality.’ Noûs 4: 175–88. Lipton, P. 2004. Inference to the Best Explanation. London and New York: Routledge. Lipton, P. 2009. ‘Understanding without Explanation.’ In Scientific Understanding: Philosophical Perspectives, edited by H. De Regt, S. Leonelli, and K. Eigner. Pittsburgh, PA: University of Pittsburgh Press. Littlejohn, C. 2013. ‘The Russellian Retreat.’ Proceedings of the Aristotelian Society 113: 293–320. Lynch, M. 2005. True To Life. Why Truth Matters. Cambridge, MA: MIT Press. Lycan, W. (2006). On the Gettier problem problem. In Stephen Hetherington (ed.), Epistemology Futures. Oxford University Press. pp. 148–168. Machery, E., Mallon, R., Nichols, S., and Stich, S. 2004. ‘Semantics, Cross-Cultural Style.’ Cognition 92: B1–B12. McHugh, C. 2011. ‘What Do We Aim at When We Believe?’ Dialectica 65: 369–92.

206  McKitrick, J. 2003. ‘A Case for Extrinsic Dispositions.’ Australasian Journal of Philosophy 81: 155–74. Magnus, P., and Cohen, J. 2003. ‘Williamson on Knowledge and Psychological Explanation.’ Philosophical Studies 116: 37–52. Maitra, I. 2011. Assertion, Norms and Games. In Assertion: New Philosophical Essays, edited by J. Brown and H. Cappelen. Oxford: Oxford University Press 277–296. Markie, P. 2005. ‘Easy Knowledge.’ Philosophy and Phenomenological Research 70: 406–16. Millar, A. 2010. ‘Perceptual Knowledge and Recognitional Abilities.’ In The Nature and Value of Knowledge, edited by D. Pritchard, A. Millar, and A. Haddock. Oxford: Oxford University Press 120–143. Millar, A. 2011. ‘Why Knowledge Matters.’ Proceedings of the Aristotelian Society, Suppl. 85: 63–81. Millikan, R. 2000. On Clear and Confused Ideas. Cambridge: Cambridge University Press. Miracchi, L. 2015. ‘Competence to Know.’ Philosophical Studies 172: 29–56. Molyneux, B. 2007. ‘Primeness, Internalism and Explanatory Generality.’ Philosophical Studies 135: 255–77. Moore, G. E. 1939. ‘Proof of an External World.’ In Philosophical Papers. London: Allen & Unwin, 1959. Mourelatos, A. 1978. ‘Events, Processes, and States.’ Linguistics and Philosophy 2: 415–34. Mumford, S. 1998. Dispositions. Oxford: Oxford University Press. Nagel, J. 2013. ‘Knowledge as a Mental State.’ In Oxford Studies in Epistemology, iv, edited by T. Szabó Gendler and J. Hawthorne. Oxford: Oxford University Press 275–310. Nozick, R. 1981. Philosophical Explanations. Oxford: Oxford University Press. Petrinovich, L., & O’Neill, P. (1996). Influence of Wording and Framing Effects on Moral Intuitions. Ethology and Sociobiology, 17, 145–171. Plato. 1956. ‘Meno.’ In Protagoras and Meno, edited by W. K. C. Guthrie. London: Penguin 115–158. Pritchard, D. 2005. Epistemic Luck. Oxford: Oxford University Press. Pritchard, D. 2007. ‘How to Be a Neo-Moorean.’ In Internalism and Externalism in Semantics and Epistemology, edited by S. Goldberg. Oxford: Oxford University Press 68–99. Pritchard, D., Millar, A., and Haddock, A. 2010. The Nature and Value of Knowledge. Oxford: Oxford University Press. Pryor, J. 2000. ‘The Skeptic and the Dogmatist.’ Noûs 34: 517–49. Pryor, J. 2004. ‘What’s Wrong with Moore’s Argument?’ Philosophical Issues 14: 349–78. Ramsey, F. 1990. F. P. Ramsey: Philosophical Papers. Cambridge: Cambridge University Press. Reynolds, S. 2013. ‘Justification as the Appearance of Knowledge.’ Philosophical Studies 163: 367–83. Riggs, W. 2002. ‘Reliability and the Value of Knowledge.’ Philosophy and Phenomenological Research 64: 79–96.



207

Rorty, R. 1995. ‘Is Truth a Goal of Enquiry? Davidson vs. Wright.’ The Philosophical Quarterly 45: 281–300. Rosenberg, J. 2002. Thinking about Knowing. Oxford: Clarendon Press. Roush, S. 2005. Tracking Truth: Knowledge, Evidence and Science. Oxford: Oxford University Press. Rysiew, P. 2012. ‘Epistemic Scorekeeping.’ In Knowledge Ascriptions, edited by J. Brown and M. Gerken. Oxford: Oxford University Press 270–294. Salmon, W. 1989. ‘Four Decades of Scientific Explanation.’ In Minnesota Studies in the Philosophy of Science, xiii, edited by P. Kitcher and W. Salmon. Minneapolis, MN: University of Minnesota Press 3–199. Schellenberg, S. 2018. The Unity of Perception: Content, Consciousness, Evidence. Oxford: Oxford University Press. Schwitzgebel, E., and, Cushman, F. 2015. ‘Philosophers’ Biased Judgments Persist despite Training, Expertise and Reflection.’ Cognition 141: 127–37. Shope, R. 1978. ‘The Conditional Fallacy in Contemporary Philosophy.’ The Journal of Philosophy 75: 397–413. Shope, R. (1983). The Analysis of Knowing: A Decade of Research. Princeton: New Jersey: Princeton University Press. Silva, P. 2017. ‘Knowing How to Put Knowledge First in the Theory of Justification.’ Episteme 14: 393–412. Simion, M. 2016. ‘Assertion: Knowledge Is Enough.’ Synthese 193: 3041–56. Simion, M. 2018. ‘No Epistemic Norm for Action.’ American Philosophical Quarterly 55: 231–8. Simion, M. 2019. ‘Knowledge-First Functionalism.’ Philosophical Issues 29: 254–67. Simion, M., and Kelp, C.2016. ‘The Tertiary Value Problem and the Superiority of Knowledge.’ American Philosophical Quarterly 53: 397–409. Simion, M., Kelp, C., and Ghijsen, H.2016. ‘Norms of Belief.’ Philosophical Issues 26: 375–92. Slaght, R. 1977. ‘Is Justified True Belief Knowledge?: A Selective Critical Survey of Recent Work’, Philosophy Research Archives 3: 1–135. Smith, M. 1994. The Moral Problem. Oxford: Blackwell. Sosa, E. 1999. ‘How to Defeat Opposition to Moore.’ In Philosophical Perspectives, edited by J. Tomberlin. Oxford: Blackwell 137–149. Sosa, E. 2007. A Virtue Epistemology. Apt Belief and Reflective Knowledge, i. Oxford: Oxford University Press. Sosa, E. 2015. Judgment and Agency. Oxford: Oxford University Press. Stanley, J. 2005. Knowledge and Practical Interest. Oxford: Oxford University Press. Strawson, P. 1992. Analysis and Metaphysics: An Introduction to Philosophy. Oxford: Oxford University Press. Suarez, M. 2007. ‘Quantum Propensities.’ Studies in History and Philosophy of Modern Physics 38: 418–38. Sutton, J. 2007. Without Justification. Cambridge, MA.: MIT Press. Swain, S., Alexander, J., and Weinberg, J. 2008, ‘The Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp.’ Philosophy and Phenomenological Research 76: 138–55.

208  Turri, J. 2016a. Knowledge and the Norm of Assertion: An Essay in Philosophical Science. Cambridge: Open Book Publishers. Turri, J. 2016b. ‘Knowledge as Achievement, More or Less.’ In Performance Epistemology, edited by M. Fernandez. Oxford: Oxford University Press 124–134. Unger, P. 1975. Ignorance: A Case for Scepticism. Oxford: Clarendon. Vendler, Z. 1957. ‘Verbs and Times.’ Philosophical Review 66: 143–60. Weinberg, J. 2016. ‘Going Positive by Going Negative.’ In A Companion to Experimental Philosophy, edited by J. Sytsma and W. Buckwalter. Malden, MA, and Oxford: John Wiley & Sons 71–86. Weinberg, J., Nichols, S., and Stich, S. 2001. ‘Normativity and Epistemic Intuitions.’ Philosophical Topics 29: 429–60. White, R. 2006. ‘Problems for Dogmatism.’ Philosophical Studies 131: 525–57. Williamson, T. 2000. Knowledge and Its Limits. Oxford: Oxford University Press. Williamson, T. 2010. ‘Knowledge First Epistemology.’ In The Routledge Companion to Epistemology, edited by S. Bernecker and D. Pritchard, 208–18. New York: Routledge. Wright, C. 2004. ‘Warrant for Nothing (and Foundations for Free)?’ Proceedings of the Aristotelian Society, Suppl. 78: 167–212. Wright, C. 2007. ‘The Perils of Dogmatism.’ In Themes from G. E. Moore. New Essays in Epistemology and Ethics, edited by S. Nuccetelli. Oxford: Oxford University Press 25–48. Wright, J. 2010. ‘On Intuitional Stability: The Clear, the Strong, and the Paradigmatic.’ Cognition 115: 491–503. 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. 2001. ‘Recovering Understanding.’ In Knowledge, Truth, and Duty. Essays on Epistemic Justification, Responsibility, and Virtue, edited by M. Steup. Cambridge: Cambridge University Press 235–252.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages. abilities 1, 7–9, 27, 52, 58, 78, 80–90, 93–98, 112, 168–172, 174–178, 180, 189, 191, 196–198, 200 abominable conjunctions 173–174, 185–186, 189 activities with constitutive aims and norms (ACANs) 46–48, 51–52, 56–58, 60, 69, 79–86, 88, 91, 93–99, 136, 138–140, 197 ACAN Ability, see Principles, ACAN Ability ACAN Exercise, see Principles, ACAN Exercise Apt ACAN Move, see Principles, Apt ACAN Move Competent ACAN Moves, see Principles, Competent ACAN Moves score-based 138–140 aims 1–33, 35, 37–43, 45–55, 57–66, 69, 72, 79, 80–86, 88, 90–91, 93–107, 130–132, 135–138, 140, 152–153, 159, 161–164, 167, 183, 185, 193–198, 200–201 constitutive 4–5, 8–9, 46, 48, 51–52, 55, 57–58, 61, 63–65, 87, 99, 136–140, 152, 157 analysis atomistic 66–67 dismantling 43–46, 48–56, 66–67, 69, 77, 96–97, 107, 109, 128, 197 functional 66–67 holistic 66–67 network 5, 8, 45, 48, 50–51, 54–56, 58, 69, 77, 98–100, 103, 108, 131–132, 197 non-reductive 5, 8, 66–67, 69–70 reductive 5, 6, 8, 39–42, 55, 66–67, 69, 89, 97, 99–100, 103, 128, 132–133

Anti-Justification Argument 38–41, 53 Argument from Ignorance 159–160, 172–173, 184, 187, 190 belief 3–5, 14–15, 17–25, 30–36, 38–41, 44, 51–60, 64, 66, 70–78, 84, 86–93, 105, 109, 113, 117, 123, 125, 128, 135, 141, 142–143, 146, 148, 151–152, 154–157, 161, 167–169, 171–172, 176, 181, 185, 187–188, 193–194, 196–199, 201–202 Competent Belief, see Principles, Competent Belief true 3–4, 7, 9, 14, 17–18, 22–24, 28, 30, 32–33, 35, 38–42, 44, 64, 66, 71–77, 113, 118, 130, 141, 144, 152, 154–157, 193–194 True Belief Aim, see Principles, True Belief Aim Carnap 66 Carter 154, 196 Cases Fake Barns 89, 92, 113 Inquiry 1 162–164, 183–184 Inquiry 1* 183–184 Inquiry 2 162–164, 166–167, 171, 183–184, 187–190 Inquiry 2* 183–184 Inquiry 3 187–189 Mixed Track Record 34–37 Stopped Clock 88, 89, 92 the burglar case 71, 72, 73, 74, 156 The Case 160, 167–168, 170–172, 180–182 the Comanche case 113, 114 The Hire 14, 17, 21–24

210  Cases (cont.) The Hire 2 193 The Insight 16–17, 21, 24, 194 The Insight 2 194 The Scatterbrain 104, 130 Zebra 160–161, 165, 185 certainty 19, 106 Coliva 165 Conservatism 165, 166 Craig 61, 62 David 38, 40, 41, 45 Davidson 11, 25, 26 Davis 165 deduction 1, 6, 9, 160–164, 166–169, 171–172, 175, 177, 180–185, 187–191 dispositions 81–83, 88, 95, 168–169 domains 4–5, 9, 38, 41, 61, 133–137, 139–141, 143–157 critical 136–137 epistemic 9, 38, 134–136, 139–140, 143, 147–148, 151 Dretske 6, 161, 184 Elgin 111, 115–118 Explanationism 120, 125–128, 130 Classical Explanationism 120 Quasi-Explanationism 126–127 Explanation-Knowledge-Science Model 123, 125 First Person Argument 29 Friedman 18, 59 Gettier 8, 15, 17, 41, 44, 55, 67, 74–75, 77–78, 88–93, 96, 98–99, 113 Greco 143, 196 Inaccessibility Argument 26 inquiry 1–11, 14–18, 20–25, 28–35, 37, 41, 43, 46, 48, 52–54, 56–64, 66, 69, 77–80, 86, 88, 90–91, 93–94, 97–107, 109, 119, 129–130, 132–133, 136, 139–140, 152–154, 157, 162–167, 171, 173, 183–184, 190, 193–194, 197–199, 201

the cases Inquiry 1 1*, 2, 2*, 3*, see Cases, Inquiry 1 1*, 2, 2*, 3 terminus of 22–25 well-conducted 21–22 JTB 44, 55, 66–67 justification 3, 7, 18–19, 27, 36–41, 44, 53–54, 70, 78, 80, 86, 90, 96, 98, 130, 153–155, 171, 193–194, 196–198, 201 of belief 3, 7, 18, 36, 41, 70, 78, 90, 98, 130, 154, 155, 171, 193–194, 196–198, 201 Justified Belief, see Principles, Justified Belief Justified Belief Aim, see Principles, Justified Belief Aim Kaplan 20 Khalifa 113, 120–121, 123–131 knowledge 1, 3–11, 15, 17–20, 22–24, 27–46, 48, 51–58, 60–62, 64, 66–78, 80, 86, 88–93, 96–100, 102–128, 130–133, 135, 140–149, 151–161, 163–164, 166–172, 174–177, 180–182, 184, 186–191, 193–199, 201–202 as an epistemic state, see Principles, Epistemic State as a mental state, see Principles, Mental State Conditions on Knowledge, see Principles, Conditions on Knowledge higher-order 19, 106 Knowledge Aim, see Principles, Knowledge Aim knowledge+ 19, 20, 106, 153 knowledge first epistemology 3–6, 64, 196–198, 201–202 Kvanvig 11, 28–37, 39, 112–115, 142 Lewis 66–68, Liberalism 165–166 Lipton 119–122 Meno 4, 141, 199 Millar 10, 90, 142, 146, 149, 176, 196

 Moderatism 165–167, 170 Moore 6, 175 Nozick 6, 184 Plato 4, 141 Principles ACAN Ability 83, 86 ACAN Exercise 83–84, 86 Apt ACAN Move 197 Commitment Release 11, 13, 15, 22, 193 Competent ACAN Moves 84–86 Competent Belief 171 Conditions on Knowledge 86–87, 89–90, 94, 96, 168–170, 197 Conditions on Proper Question Closing 86–88 Degrees of Understanding 108–110, 122 Epistemic State 70, 72–73, 75 Exercises of Abilities to Know 87 Justified Belief 14–15, 17, 57, 90, 171 Justified Belief Aim 11, 14–15, 17, 25, 57 Knowledge Aim 10–11, 15, 17–18, 20–21, 23–25, 28, 31–37, 41–42, 57, 105, 163–164 Maximal Understanding 107–108 Mental State 70–73, 75–77 No Settling 162–164 Outright Understanding 108–111, 114, 122 Priority 43–45, 49–51, 54, 66, 109 Progress 15–17, 117, 194 Question Begging 163–164, 182–183 Question Settling 2, 10–11, 53, 56, 162–164 Question Settling Aim 2, 10–11, 162–164 Understanding Aim 2, 99–100, 102, 106–107 Understanding Aim* 107 The Received View 119–120, 122 Transmission of Knowledge 6, 7, 160–175, 179–181, 183–186, 188–190 Transmission Failure 189–190

211

Transmission of Warrant 164, 165 True Belief Aim 11, 14–15, 17, 21–24, 32–35, 57, 64 Pritchard 6, 142–143, 145–146, 148–151, 163, 196 Pryor 163, 164, 165 Ramsey 66–68 reliability 40, 53, 61 scepticism 6–7, 9, 78, 143, 173, 177–179, 184, 186, 190–191 sensitivity 9, 179, 184–188, 191 Smith 68 Sosa 6, 82, 136, 196, 199 Strawson 5, 43–45, 66–67 Transmission of Knowledge, see Principles, Transmission of Knowledge Transmission Failure, see Principles, Transmission Failure Transmission of Warrant, see Principles, Transmission of Warrant understanding 1–6, 8–9, 19, 44, 52, 55, 68, 98–103, 106–110, 112–133, 135, 140, 144–147, 151, 157–158, 167, 193 Degrees of Understanding, see Principles, Degrees of Understanding Maximal Understanding, see Principles, Maximal Understanding Outright Understanding, see Principles, Outright Understanding Understanding Aim, see Principles, Understanding Aim Understanding Aim*, see Principles, Understanding Aim* The Received View, see Principles, The Received View values 1, 4–5, 7, 9, 38, 50, 132–138, 140–152, 154–159, 193–194 central 134–138, 140, 143, 148, 149–152, 154, 157

212  values (cont.) epistemic 4–5, 145–147, 151, 154–156 for-its-own-sake 4, 38, 147–150, 156 of knowledge 133, 141, 143–147, 155, 158 of understanding 144, 146 value problems 4, 9, 141–149, 151, 157 The First Value Problem 144 The Primary Value Problem 141–142, 144 The Second Value Problem 144 The Secondary Value Problem 142, 144

The Tertiary Value Problem 142–143, 145–149, 151 The Third Value Problem 145 virtue epistemology 143, 196–199, 201, 202 White 165 Williamson 3, 6, 8, 10, 46, 56, 66, 69, 70–77, 90, 156, 160–161, 181 Wright 1, 165 Zagzebski 118, 196