Donald Davidson on Action, Mind and Value [1st ed.] 9789811572296, 9789811572302

Table of contents :
Front Matter ....Pages i-xiii
Objects and Events in Davidson (Jennifer Hornsby)....Pages 1-17
Davidson’s Meta-normative Naturalism and the Rationality Requirement (Robert H. Myers)....Pages 19-36
Acting Against Your Better Judgement (Rowland Stout)....Pages 37-53
A Davidsonian Theory of Evaluative Judgment (Ya-Ting Chang)....Pages 55-75
Davidson on Emotions and Values (Pascal Engel)....Pages 77-92
Pleasure Makes Pro-attitudes (Jeremias Koh, Neil Sinhababu)....Pages 93-105
On the Distinctive Value of Knowledge (Kok Yong Lee)....Pages 107-127
The Continuity of Davidson’s Thought: Non-reductionism Without Quietism (Claudine Verheggen)....Pages 129-144
Methodological Considerations in the Triangulation Argument (Wai Chun Leong)....Pages 145-169
Meaning, Evidence, and Objectivity (Olivia Sultanescu)....Pages 171-184
Davidson’s Triangulation Argument in the Logic of Actions (Syraya Chin-Mu Yang)....Pages 185-209

Citation preview

Logic in Asia: Studia Logica Library Series Editors: Fenrong Liu · Hiroakira Ono · Kamal Lodaya

Syraya Chin-Mu Yang Robert H. Myers Editors

Donald Davidson on Action, Mind and Value

Logic in Asia: Studia Logica Library Series Editors Fenrong Liu, Tsinghua University and University of Amsterdam, Beijing, China Hiroakira Ono, Japan Advanced Institute of Science and Technology (JAIST), Ishikawa, Japan Kamal Lodaya, Bengaluru, India Editorial Board Natasha Alechina, University of Nottingham, Nottingham, UK Toshiyasu Arai, Chiba University, Chiba Shi, Inage-ku, Japan Sergei Artemov, City University of New York, New York, NY, USA Mattias Baaz, Technical university of Vienna, Austria, Vietnam Lev Beklemishev, Institute of Russian Academy of Science, Russia Mihir Chakraborty, Jadavpur University, Kolkata, India Phan Minh Dung, Asian Institute of Technology, Thailand Amitabha Gupta, Indian Institute of Technology Bombay, Mumbai, India Christoph Harbsmeier, University of Oslo, Oslo, Norway Shier Ju, Sun Yat-sen University, Guangzhou, China Makoto Kanazawa, National Institute of Informatics, Tokyo, Japan Fangzhen Lin, Hong Kong University of Science and Technology, Hong Kong Jacek Malinowski, Polish Academy of Sciences, Warsaw, Poland Ram Ramanujam, Institute of Mathematical Sciences, Chennai, India Jeremy Seligman, University of Auckland, Auckland, New Zealand Kaile Su, Peking University and Griffith University, Peking, China Johan van Benthem, University of Amsterdam and Stanford University, The Netherlands Hans van Ditmarsch, Laboratoire Lorrain de Recherche en Informatique et ses Applications, France Dag Westerstahl, Stockholm University, Stockholm, Sweden Yue Yang, Singapore National University, Singapore Syraya Chin-Mu Yang, National Taiwan University, Taipei, China

Logic in Asia: Studia Logica Library This book series promotes the advance of scientific research within the field of logic in Asian countries. It strengthens the collaboration between researchers based in Asia with researchers across the international scientific community and offers a platform for presenting the results of their collaborations. One of the most prominent features of contemporary logic is its interdisciplinary character, combining mathematics, philosophy, modern computer science, and even the cognitive and social sciences. The aim of this book series is to provide a forum for current logic research, reflecting this trend in the field’s development. The series accepts books on any topic concerning logic in the broadest sense, i.e., books on contemporary formal logic, its applications and its relations to other disciplines. It accepts monographs and thematically coherent volumes addressing important developments in logic and presenting significant contributions to logical research. In addition, research works on the history of logical ideas, especially on the traditions in China and India, are welcome contributions. The scope of the book series includes but is not limited to the following: • • • •

Monographs written by researchers in Asian countries. Proceedings of conferences held in Asia, or edited by Asian researchers. Anthologies edited by researchers in Asia. Research works by scholars from other regions of the world, which fit the goal of “Logic in Asia”.

The series discourages the submission of manuscripts that contain reprints of previously published material and/or manuscripts that are less than 165 pages/ 90,000 words in length. Please also visit our webpage: http://tsinghualogic.net/logic-in-asia/background/

Relation with Studia Logica Library This series is part of the Studia Logica Library, and is also connected to the journal Studia Logica. This connection does not imply any dependence on the Editorial Office of Studia Logica in terms of editorial operations, though the series maintains cooperative ties to the journal. This book series is also a sister series to Trends in Logic and Outstanding Contributions to Logic. For inquiries and to submit proposals, authors can contact the editors-in-chief Fenrong Liu at [email protected] or Hiroakira Ono at [email protected].

More information about this series at http://www.springer.com/series/13080

Syraya Chin-Mu Yang Robert H. Myers •

Editors

Donald Davidson on Action, Mind and Value

123

Editors Syraya Chin-Mu Yang National Taiwan University Taipei, Taiwan

Robert H. Myers York University Toronto, ON, Canada

ISSN 2364-4613 ISSN 2364-4621 (electronic) Logic in Asia: Studia Logica Library ISBN 978-981-15-7229-6 ISBN 978-981-15-7230-2 (eBook) https://doi.org/10.1007/978-981-15-7230-2 © Springer Nature Singapore Pte Ltd. 2021 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

In the last twenty years of his life, Donald Davidson brought his causal theory of reasons and action and his truth-theoretic semantics for natural language together into a fully unified theory of mind and language, clarifying the roles that normativity, rationality, and objectivity play in his treatment of intentional action, evaluative judgement, and radical interpretation. At the core of this programme is his triangulation argument, which aims not merely at a satisfactory theory of language and thought but also at a full-fledged theory of knowledge by explaining why rational agents must have knowledge of the world, knowledge of other minds, and self-knowledge. By showing how communication among rational agents is possible, Davidson sheds new light on the nature of human knowledge and the objectivity of meaning and ultimately extends this grand programme to questions concerning values, emotions, and other related issues. The papers in this volume were originally written for (and most were presented at) the Third Taiwan Metaphysics Conference (TMC-2017), which was held in November 2017, at the National Taiwan University. As 2017 was the centenary anniversary of Davidson’s birth, the conference took as its theme the development of his philosophy over time and its continuing relevance to ongoing debates. This resulted in a wide range of innovative reflections, many of which focussed on the pivotal role that Davidson’s conception of agency plays in his later philosophy and its impact on his epistemology, his philosophy of language and mind, and his philosophy of values. The participants critically assessed central elements of Davidson’s programme, and offered reappraisals of his seminal contributions to, and his continuing influence on, the development of contemporary philosophy. The particular focus on agency revealed Davidson’s views to have been more dynamic and less reductive than hitherto acknowledged—pointing, as we shall see, towards important new possibilities not only in the theory of knowledge but also in the philosophy of action, mind, and value. Chapter One “Objects and Events in Davidson”, by Jennifer Hornsby, focusses on Davidson’s metaphysics, and specifically on the conception of events that Davidson introduced when he took the operation of causality always to consist in the relation cause obtaining between a pair of events. Hornsby argues that Davidson’s v

vi

Preface

conception of events as individual entities stands in the way of according objects a role in the causal world, and that it cannot allow for the understanding of general statements about events. But she goes on to suggest that, when Davidson turned to epistemological questions, he relinquished the ontology that guided his early work on action and came to take a much more relaxed view of “what there is”. Hornsby concludes by discussing the implication of this for Davidson’s physicalism, which he used his thesis of the Nomological Character of Causality to establish. In Chapter Two “Davidson’s Meta-normative Naturalism and the Rationality Requirement”, Robert H. Myers invokes Davidson’s causal account of normative properties to defend normative realism against a powerful objection. Michael Smith has famously argued that realists cannot explain why people act on their normative judgements as reliably as they do, and that constructivism (or perhaps constitutivism) is therefore to be preferred. Normative realists have often countered this argument by insisting that they can establish (1) that rationality typically requires people to act on their normative judgements and (2) that people typically are rational. Myers acknowledges that normative realists can establish (1) but argues that (2) poses problems both for non-naturalists and for reductive naturalists. He closes by suggesting that Davidson’s non-reductive (or “anomalous”) form of naturalism may give normative realists a better chance of success. In Chapter Three “Acting Against Your Better Judgement”, Rowland Stout defends a Davidsonian approach to weakness of will against some recent arguments by John McDowell and adapts the approach to meet other objections. Instead of treating one’s better judgement as a conditional judgement about what is desirable to do given available reasons, he proposes to treat it as an unconditional judgement about what is desirable to do from a rational perspective that one takes to be the right perspective to have. He argues that this makes sense of Aristotle’s claim that desire is for the good or the apparent good: judgements of desirability generally concern the apparent good, whereas judgements of desirability from rational perspectives that are judged to be the ones to have are judgements of the actual good. Weakness of will occurs when one’s actual rational perspective is not the one that one takes to be the one to have—i.e. when one’s judgement of the apparent good does not coincide with one’s judgement of the actual good. One makes two judgements—one from an adopted perspective that one judges to be the one to have and one from one’s actual perspective. Chapter Four “A Davidsonian Theory of Evaluative Judgment”, by Ya-Ting Chang, develops a Davidsonian theory of evaluative judgement. As she notes, because the Humean theory of motivation remains the dominant view, most philosophers endorse the idea that desire is necessary for motivation. As a result, they often feel compelled to endorse non-cognitivism in order to defend internalism, since it seems to them that evaluative judgements must either express or be desires if they are to motivate action. But then evaluative judgements cannot be true or false, for desires lack cognitive content. Chang aims to avoid this result by reconstructing a Davidsonian theory of evaluative judgement. She argues that, by following Davidson and taking evaluative judgements to be conative propositional attitudes, we can see how they can be both necessarily motivating and truth-apt.

Preface

vii

In Chapter Five “Davidson on Emotions and Values”, Pascal Engel discusses Davidson’s views about emotions and values. He demonstrates that Davidson held a causal version of the cognitive theory of emotions, which he associated with an objective conception of values. Engel confronts Davidson’s account of the correctness of emotions with the fitting-attitude view of the relation of emotions to values and argues that the two approaches can be reconciled if the correctness of an emotion is construed as a form of idealization. In Chapter Six “Pleasure Makes Pro-attitudes”, Jeremias Koh and Neil Sinhababu consider what it is to be a pro-attitude, the sort of mental state that can motivate one to act for a reason. They argue that pro-attitudes must do more than simply cause the right behaviour when combined with an appropriately linked belief. However, they also argue against requiring pro-attitudes to be accompanied by beliefs that the object is valuable, or treating pro-attitudes as perceptions of value. On their view, pro-attitudes must not only motivate action, but cause more pleasure when one thinks about attaining their objects than when one thinks about failing to attain them. Chapter Seven “On the Distinctive Value of Knowledge”, by Kok Yong Lee, defends the distinctive value thesis, i.e. the thesis that knowledge is more valuable than any of its proper parts. The first part of the paper considers three prominent arguments against the distinctive value thesis: the swamping argument, the generalized swamping argument, and the ad hoc argument. Lee maintains that they all rely on the mistaken idea that the distinctive value thesis is incompatible with what he calls epistemic veritism, the view that the value of knowledge is parasitic on the value of truth. In the second part of the paper, Lee argues that the distinctive value thesis is not only compatible with epistemic veritism but in fact follows from it. The crucial observation here is that knowledge does not contain or presuppose falsehood in the way mere justified true belief does. In Chapter Eight “The Continuity of Davidson’s Thought: Non-reductionism Without Quietism”, Claudine Verheggen discusses the continuity of Davidson’s thought. Her goal is to show that two commonly held conceptions of Davidson’s philosophy, one to the effect that there is a shift from radical interpretation to triangulation, in particular, a shift from reductionism to non-reductionism, the other to the effect that Davidson’s non-reductionism is incompatible with constructive philosophizing, are in fact misconceptions. Verheggen argues that, though Davidson’s account of meaning, being non-reductionist, does not spell out non-circular sufficient conditions for any particular utterances to have the meaning they have, and for someone to have a language and thoughts at all, it is nonetheless non-quietist, for it does give us necessary conditions for these phenomena to occur. These necessary conditions first emerge from Davidson’s reflections on radical interpretation and are further articulated in his writings on triangulation, which also demonstrate that radical interpretation is but an instance of triangulation, since they make clear that successful radical interpretation in effect requires triangulation. In Chapter Nine “Methodological Considerations in the Triangulation Argument”, Wai Chun Leong examines Davidson’s triangulation argument, which purports to show that an agent could not have propositional thoughts if it had never interacted

viii

Preface

with other agents in certain requisite ways. As Leong notes, not many philosophers accept this argument, and indeed no consensus has emerged on how exactly the argument goes. He thus sets out to reformulate it, paying special attention to methodological considerations bearing on the notion of relevant similarity, which he believes are often overlooked in the literature. He also proposes a tentative modification of the kind of interaction that triangulation requires. Chapter Ten “Meaning, Evidence, and Objectivity”, by Olivia Sultanescu, takes up the question of what, according to the conception of meaning offered by Davidson, makes expressions meaningful. She addresses this question by reflecting on Kathrin Glüer’s recent response to it, arguing that Glüer misconstrues both the evidence for meaning that the radical interpreter must rely on and the way in which the principle of charity must be deployed. The articulation of the correct construal of the evidence and the principle reveals the thoroughly non-reductionist aspect of Davidson’s conception of meaning. Sultanescu maintains that this aspect becomes even clearer in Davidson’s later work, through the articulation of the triangulation argument, and shows how this argument helps explain what makes expressions meaningful. She ends with a brief discussion of the question whether Davidson has the resources to account for the objectivity of meaning. In the final chapter, “Davidson’s Triangulation Argument in the Logic of Actions”, Syraya Chin-Mu Yang provides a formal machinery to complement the epistemic background in Davidson’s triangulation argument by presenting a structural modelling of the argument in a semantic framework of an epistemic logic of actions. Yang focusses on two aspects. Firstly, he suggests that to guarantee that the propositional content of what the speaker says in the ongoing communication is a justified true belief, the speaker’s utterance must be an assertion governed by the knowledge norm, assuming Davidson’s characterization of knowledge in terms of holistically justified true beliefs. Secondly, the author claims that ongoing communication succeeds only if what the speaker says becomes common knowledge. Accordingly, the required underlying system for the desired logic of actions will be a logic of common knowledge, in which the notion of common knowledge is characterized in terms of the knowledge account of assertion. A semantics appropriate for the logic of common knowledge is given. Yang further proposes that the transition from the speaker’s initial assertion to common knowledge can be formulated by virtue of individual agents’ learning and group learning, resulting from what the hearer learns from the speaker’s assertion and what the speaker learns from the hearer’s response, and the following interactions between them. Two act-operators for individual agents’ learning and group learning are introduced; the required semantic rules are stipulated. We owe thanks to all of the authors and referees of the manuscripts, to the members of the organizing committee, and to the participants in the joint conference, including all of the speakers, discussants, and attendees, and the staff of the Department of Philosophy, National Taiwan University. In particular, we would like to express our gratitude to Jennifer Hornsby, Claudine Verheggen, Pascal Engel, and Rowland Stout for their contributions to TMC3-2017 and to this volume. We also appreciate the help in the organization of the conference that we received from a large number of friends, especially Derek Baker, D.-M. Deng,

Preface

ix

Chung-I Lin, Rong-Lin Wang, and Chun-Ping Yen. We are deeply indebted to Fenrong Liu and Hiroakira Ono, the editors-in-chief of the book series Logic in Asia (LIAA), for their encouragement and long-term support of the TMCs and the preparation of this volume. Thanks also go to Leana Li, Editorial Director of Humanities and Social Sciences, for her contributions to this volume. Finally, and above all, we owe special thanks to Miss Wendy Huang. Without her generous financial support of the Wendy Huang Program and the Taiwan Metaphysics Colloquium series, this collection could only have materialized in some merely, perhaps even inaccessible, possible worlds. Taipei, Taiwan Toronto, Canada

Syraya Chin-Mu Yang Robert H. Myers

Contents

Objects and Events in Davidson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jennifer Hornsby Davidson’s Meta-normative Naturalism and the Rationality Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Robert H. Myers

1

19

Acting Against Your Better Judgement . . . . . . . . . . . . . . . . . . . . . . . . . Rowland Stout

37

A Davidsonian Theory of Evaluative Judgment . . . . . . . . . . . . . . . . . . . Ya-Ting Chang

55

Davidson on Emotions and Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pascal Engel

77

Pleasure Makes Pro-attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jeremias Koh and Neil Sinhababu

93

On the Distinctive Value of Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . 107 Kok Yong Lee The Continuity of Davidson’s Thought: Non-reductionism Without Quietism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Claudine Verheggen Methodological Considerations in the Triangulation Argument . . . . . . . 145 Wai Chun Leong Meaning, Evidence, and Objectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 Olivia Sultanescu Davidson’s Triangulation Argument in the Logic of Actions . . . . . . . . . 185 Syraya Chin-Mu Yang

xi

Contributors

Ya-Ting Chang Independent Scholar, Sydney, Australia Pascal Engel Ecole des hautes études en sciences sociales, Paris, France Jennifer Hornsby Birkbeck, University of London, London, UK Jeremias Koh Department of Philosophy, National University of Singapore, Singapore, Singapore Kok Yong Lee Department of Philosophy, National Chung Cheng University, Chiayi, Taiwan Wai Chun Leong Department of General Education, Macau University of Science and Technology, Taipa, Macau, China Robert H. Myers Department of Philosophy, York University, Toronto, ON, Canada Neil Sinhababu Department of Philosophy, National University of Singapore, Singapore, Singapore Rowland Stout School of Philosophy, University College Dublin, Dublin, Ireland Olivia Sultanescu Society of Fellows in the Liberal Arts, The University of Chicago, Chicago, IL, USA Claudine Verheggen Department of Philosophy, York University, Toronto, ON, Canada Syraya Chin-Mu Yang Department of Philosophy, National Taiwan University, Taipei, Taiwan

xiii

Objects and Events in Davidson Jennifer Hornsby

Abstract The focus of this paper is Davidson’s metaphysics, and specifically the conception of events that Davidson introduced when he took the operation of causality always to consist in the relation cause obtaining between a pair of events. I argue that Davidson’s conception of events as individual entities stood in the way of according objects a role in the causal world, and that it cannot allow for the understanding of general statements about events. But I suggest that when Davidson turned to epistemological questions, he relinquished the ontology that guided his early work on action and came to take a much more relaxed view of “what there is”. I touch on the implications of this for Davidson’s physicalism, which he used his thesis of the Nomological Character of Causality to establish. Keywords Davidson’s metaphysics · Events · Action · Davidson’s physicalism · Nomological character of causality

1 Introduction Davidson’s standing as among the greats of twentieth-century philosophy is owed not only to the quality of the work he published over more than forty years but also to its astonishing scope and breadth. Its breadth and systematic character is particularly 1 With the exception of two essays (Davidson 1967c, 2001e) and a book written close to the end of his life and published posthumously in 2005, references to Davidson’s essays herein are all to essays (three dozen of them) which are in one or other of three of Davidson’s anthologies (2001c, d, 2004) and are cited in accordance with the dates given therein. Davidson 2001c is a second edition which retains the Introduction to the first edition published in 1980 and contains as Appendices two essays (1985a, b) not found in the first edition. Davidson 2004 helpfully contains the contents of earlier anthologies, giving dates of the essays.

This chapter is in its final form and it is not submitted for publication anywhere else. J. Hornsby (B) Birkbeck, University of London, Malet Street, London WC1E 7HX, UK e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_1

1

2

J. Hornsby

evident in his paper ‘Three Varieties of Knowledge’. There he wrote: ‘We need an overall picture which not only accommodates all three modes of knowing, but makes sense of their relations to one another’ (p. 208 in 1991).1 Davidson’s overall picture was got from his unification of accounts of knowledge of one’s own mind, knowledge of the world and knowledge of others’ minds. His accounts of knowledge of these three kinds rest upon his answers to questions in philosophy of mind and language— about perception, about action, and, by way of his account of radical interpretation, about linguistic understanding. In his philosophy of mind, Davidson fended off physicalistic reductionism. His thesis of the nomological character of causality shielded him against it. In Davidson’s view, human beings’ interactions with the world outside their bodies and with one another are all of them constituted by causal relations between events. When the vocabulary of the mental—of belief and desire and so on—is in play, the constitutive ideal of rationality holds sway. But given causality’s nomological character, events that can be picked out in the everyday terms used when that ideal is in play are, Davidson said, describable also in the terms in which scientists’ laws are framed. Thus, the physical vocabulary of scientists has application to the mental events we participate in as rational beings, so that these are physical events. This is Davidson’s anomalous monism, of the early 1970s. Davidson’s anti-reductionism attracted me to his work when I first knew some of it, way back in my graduate student days. Of particular appeal to me was Davidson’s understanding of human thought, language and action as structured by rationality. But I have found myself less and less in agreement with Davidson as the years have gone by. In the present paper, I look for a new way to pinpoint the source of my disagreement. I begin by challenging Davidson’s event-causal metaphysics, according to which a phenomenon is causal only insofar as it is underpinned by a relation between events (see 1967a). Davidson’s event ontology is salient in his early writings in philosophy of action; and I shall pay most attention to those (Sects. 3 and 4). I shall say a little also about his account of perception (Sect. 5). This is with a view to showing that when Davidson reached his overall picture of our modes of knowing, he had come to appreciate that objects, not only events, stand to one another in relations in the causal world. I want to suggest that any overall evaluation of Davidson’s writings in philosophy of mind and epistemology should take account of a certain change of mind on Davidson’s part—a change between his early work on action and his later work in epistemology. But I won’t come to that until I’ve said a fair bit about Davidson’s philosophy of action.

2 Objects and Events I start with something about Davidson’s particularist ontology—about how he saw the two categories—of events and of objects.

Objects and Events in Davidson

3

A monist takes there to be things of only one basic kind in the universe. Davidson’s own monism, depending as it does on causality’s nomological character, takes the basic kind of things there are to be events. Small wonder, then, that in arguing for monism, Davidson should have defended his claim that events constitute a fundamental ontological category (p. 180 in 1969a)—a category of ‘unrepeatable particulars, or “concrete individuals”’ (p. 181 in 1970b). He defended this claim, thinking that we need to be persuaded of the existence of events, but could take for granted the existence of objects—of objects which ‘remain the same through changes’ (p. 311 in 1985b) (Occasionally Davidson speaks of things in the category of objects as ‘substances’ (e.g. pp. 170, 175, 178 in 1985b). But I shall follow him here in using the word ‘objects’ [even if ‘object’ can seem less than an ideal term for the category, given that persons then are objects].). There are philosophers who conceive of objects as four-dimensional entities— space–time worms, as their objects are sometimes called. Quine was one such philosopher, believing entities to be of a single kind, and Davidson quite explicitly disagreed with him about this, saying that objects and events belong in distinct categories (p. 310 in 1985b). Davidson had shown himself to be in agreement with Strawson, who said that events do not provide ‘a single, comprehensive and continuously usable framework of reference of the kind provided by physical objects’ (p. 173 in 1969a). And Davidson could hardly have dismissed such a framework of reference. He spoke of ‘the language of science as only a suburb of our common language’, saying that ‘science can add mightily to our linguistic and conceptual resources, but can’t subtract much’ (p. 305 in 1985b). Non-scientific language being replete with words for macrophysical objects, Davidson was bound to find an ontology of objects quite unproblematic. His claim that events comprise a fundamental category amounts to the claim that they are ‘on a par, ontologically, with familiar objects’ (p. 304 in 1985b). Although he set himself against Quine’s four-dimensionalism, Davidson thought of events and objects alike as spatiotemporal entities, items in both categories individuated by spatiotemporal location.2 What distinguished between entities in the two categories, he thought, was the different manners in which they relate to locations in space and time. As he put it: ‘It may be that events occur at a time in a place while objects occupy places at times’ (p. 301 in 1985b). When told that they occupy places, one thinks of objects as taking up space in the world. One might wonder then whether an event should be thought of, analogously, as taking up time in the world. Davidson appears to have thought so; for although he spoke of an event as something that occurs ‘at a time’, he also said that an ‘intentional action may take much time’ (p. 87 in 1978). So it might be thought that Davidson’s events must occupy intervals of time rather than merely moments, just as objects occupy volumes of space rather than merely points. So it might be thought. But the

2 Davidson

had earlier suggested a non-spatiotemporal criterion of the individuation of events, but he said ‘Quine has made clearer to me what was wrong with my original suggestion, and I hereby abandon it.’ (p. 309 in 1985b).

4

J. Hornsby

analogy relies upon supposing that time is “occupied” just as space is. And I shall come to a difficulty about supposing this in the next section. When Davidson first presented his theory of events, he said that it was ‘silent about processes, states and attributes if these are different from individual events’ (p. 210 in 1969a). He found a place for states within his event ontology when he said that ‘states and dispositions are not events, but the onslaught of a state or disposition is’ (p. 12 in 1963). But he never said anything about a place for processes in his metaphysics, despite the fact that he often spoke of processes himself. He spoke of ‘the all-important process of weighing reasons’ and ‘a process of reasoning’ (pp. 36, 41 in 1969b); of ‘electrical and chemical processes’, ‘neurological processes’, ‘mental processes’, and ‘cognitive processes’ (e.g. pp. 245, 246, 247 in 1973a); and again of ‘the process by which we understand utterances’ (p. 12 in 1984a), ‘the process of determining what is true’ (p. 70 in 1986), and ‘the causal process that connects beliefs with the world’ (p. 46 in 1988). With so much to be understood by reference to processes, one is bound to wonder whether an ontology of events alone can suffice for an understanding of causality. The reason why events have such a major part in Davidson’s philosophy is that he was so much concerned with questions about mind and action where his accounts of phenomena are all of them causal accounts. As such, they make use of a single notion of causation—‘the ordinary notion of cause’ as Davidson would have it (p. 9 in 1963, pp. 49, 51, 53 in 1971), or ‘the garden-variety of causation’ (p. 52 in 1971). According to this notion, wherever causation is found, there is a causal relation, expressed with the word ‘cause’, between one event and another (or between a state and an event, as it will be if one relies on the claim that ‘the onslaught of a state or disposition is an event’ [see above]). Now the word ‘cause’ (or an equivalent) standing for a relation between events isn’t present in every sentence that might be taken to invoke an idea of causation in some sense or other. For example, sentences on the pattern ‘p because q’ which provide causal explanations appear to have no equivalents saying that events are related by ‘cause’.3 Again, there are statements which do contain the word ‘cause’ but which make no mention of an event as a cause. Davidson for his part thought that wherever an object is spoken of as a cause, ‘the ellipsis of event causality’ is found, so that mention of the object just as such is eliminable. ‘We may expand the account of the cause to embrace an event’, he said (p. 53 in 1971): to say that the ball broke the window just is to say that its motion caused the breakage (p. 54 in 1971). Well, there doesn’t appear to be any ellipsis in ‘The ball broke the window’—not at least as ‘ellipsis’ is ordinarily understood (As it is ordinarily understood, there is ellipsis where words needed to complete a grammatical construction are omitted so as to avoid redundancy.). 3 In

Hornsby 1999, I argued that they could not be so understood. I pointed out that in ‘Actions, Reason and Causes’ (1963), Davidson set out to argue for the conclusion: ‘Rationalization is a species of causal explanation’ but came to summarize that conclusion saying ‘Reasons are causes’. I gave examples in which there is an instance of a rationalizing (thus causal) explanation by Davidson’s lights, but in which there is not so much as a candidate for a statement on the pattern of ‘r caused a’. However that may be, Davidson believed that wherever a causal explanation could be given, the explanation could be recorded in a statement saying that one thing had caused another.

Objects and Events in Davidson

5

Davidson spoke of ellipsis again when it came to human agency. ‘Although we say the agent caused the death of the victim, … this is an elliptical way of saying that some act of the agent caused the death of the victim’ (p. 49 in 1971). But can it not be asked ‘Who killed the victim? And would Davidson have said that ‘Who caused -----?’ is everywhere to be treated as elliptical for ‘Whose action caused -----?’ I don’t know. But however that may be, in Davidson’s view wherever an agent is cast in a causal role, really an event—specifically an action—must be taken to have played that role.

3 Events and Time Davidson didn’t think that actions are just any old events: as he said, actions ‘invite special modes of description and explanation’. Even so, actions have no distinctive ontological standing among events, according to Davidson. ‘What is special [about actions]’, he said, ‘turns out to be neither a matter of ontology nor, primarily at least, of grammar’ (p. 293 in 1985b). It ought to be possible then to obtain a general view of how Davidson thought of the nature and language of events by looking at what he had to say about actions.4 And I start here with an example he used of a particular action. When Davidson made his case that events are particulars, he illustrated his view of the logical form of action sentences, taking the sentence ‘Meyer climbed Kibo’. He said: On my proposal, ‘Meyer climbed Kibo’ is analysed as saying that there exists an event that is a climbing of Kibo by Meyer: in symbols, ‘(∃x) (Climbed (Kibo, Meyer, x))’. (p. 185 in 1970b).

In the symbolic version here, a 3-place predicate ‘climbed’ is invoked. But according to the analysis, the predicate ‘is a climbing of Kibo by Meyer’ applies to an event which exists. If the symbolic version is equivalent to the analysandum, then an event that is a climbing of Kibo by Meyer exists only if Meyer climbed Kibo. The use of the past tense is needed to speak of the existence of an event. Commitment to events’ existence appears to be commitment to entities of the past, to things which have a place in history, as it were. And it is understandable that Davidson should so conceive events. For he takes a verb of action to be a verb that says what someone did (p. 105 in 1967b). It is only to be expected that the past tense should be found in all of Davidson’s examples of action sentences. (‘Jones buttered the toast’, ‘Amunsden flew to the North Pole’, ‘Brutus stabbed Caesar’, etc.) 4 Davidson’s

focus was exclusively on human actions, where that which is done by an agent is intentionally done, and this will be my focus too. But I note that some of the questions I mean to raise about the causal notions required to understand human agency could equally be questions about non-human animal agency, some of them indeed questions about the agency of inanimate objects. Davidson recognized that there is teleological explanation (p.xvi in 2001a).

6

J. Hornsby

I don’t think we should be surprised that Davidson’s events should be confined to things that are over and done with. This may be a consequence of his category of events being, as he saw it, a category of concrete individuals, and his following Quine in taking existence to be expressed by the first-order existential quantifier. To speak of events as concrete is to inspire a view of them as fixed and unsusceptible to change—as unchanging, as we take the past to be. According to Davidson, ‘we say that one event is identical with another, quantify over events, and count events under sortals’ (in 1985b, 306). And unless someone has already done something, there is as yet nothing—no “whole”, individuable entity of such a sort as can be counted. The events whose existence Davidson defended in the course of rejecting Quine’s four-dimensionalism are then indeed matters of history. Call the conception of events in which they are a kind of existent, alongside objects, the “metaphysical” conception. A different conception is surely needed if events are to feature in any general account of agency. Presumably Davidson would not have denied that it is possible to say what will occur but has not yet occurred—your putting on of your shoes tomorrow morning, as it might be, or the Queen’s giving a speech to Parliament next Thursday. But if events are found here, they are things of the future, not events according to the metaphysical conception. It must be a question for Davidson how a sentence that seems to speak of agency but whose verb is not in the past tense is to be understood. Davidson for his part often used verbs in what might be thought of as a sort of generic tense, as if future and past tense sentences alike could be subsumed by it. This generic tense is found at the very start of his work on agency—in the famous example of his paper ‘Actions, Reasons and Causes’, where he says ‘I flip the switch. I turn on the light …’ (p. 4 in 1963). Here ‘flip’ and ‘turn on’ might seem to be simply a present tense. But if Davidson’s reader is to have been presented with an actual example, then she must think that Davidson turned on the light, that turn on the light was something he did. It might be thought that it can’t really matter that a reader is forced to think of the action as over, as past, in order to have the example in mind—that it makes no difference to what is involved in a person’s turning on of a light whether he has already turned the light on or will do so in the future. ‘The time at which any particular action occurs may be earlier than or later than the time at which it has a place in the totality of occurrences viewed from outside any particular temporal perspective’, it might be said. This response appeals to the idea of a time at which an action takes place. So it fits with Davidson’s speaking of events as occurring at a time. But we start to see a real problem for Davidson when we remember that, as he himself allowed, it can take time to do something. If action takes time, then it takes place over time. And Davidson could hardly disagree, given what he told us about action verbs. In keeping with his view that action explanation is teleological explanation (p. xvii in 1980a), Davidson told us that action verbs (verb phrases actually) are ‘definable in terms of some terminal stage, outcome or consequence’ (n. 2, p. 4 in 1963). When a verb phrase which is so definable is used to say what someone did, the relevant terminal stage, outcome or consequence is in place: the end or goal of she who did it—her telos—has been reached. But it is only when a person has reached her goal that she

Objects and Events in Davidson

7

has finished doing what she was doing in order to reach it. For example: If Jane walked to the library, so that walk to the library was what Jane did, then reaching the library was Jane’s end or goal, and her arrival at the library was the ‘terminal stage’ of her walk to the library, its ‘outcome’ her being at the library. What account might Davidson have given of someone who was in the process of doing something? How is ‘Jane was walking to the library’ to be understood? It might or might not be that Jane has arrived at the library when this is said. She might have encountered a cordon which prevented her from getting the whole way to the library; or she might have changed her mind and decided to go somewhere else. The fact that a person was doing something is no guarantee that there was ever such an event as is required for the truth of ‘She did it’. In an example made famous by philosophers, someone was crossing the street when she was knocked over, with the consequence that she didn’t cross the street.5 Equally, to switch to the present tense now, someone may be doing something for the sake of an end or goal, yet will not come to have done it. The point here is that Davidson’s action sentences are not merely past in tense, but also they assume the past perfect. ‘Climbed’ and ‘was climbing’, share a tense (both are past), but they differ from one another in the aspect they import: ‘climbed’ imports perfect aspect, ‘was climbing’ (likewise ‘is climbing’ and ‘will be climbing’) has imperfect aspect (I say that ‘climbed’, “imports” perfect aspect because it is only when she has done it [past perfect] that it can be said that she did it [preterite, i.e. simple past].). Use of the imperfect with an action verb conveys an idea of ongoing progress. If she is [/was/will be] walking to the library, then she is [/was/will be] advancing towards the library. Davidson for his part had nothing to say about action verbs as they take imperfect aspect.6 In taking action verbs to be verbs used to say what someone did he ruled it out that there should be anything to be said about someone who is or was in the process of doing something. One might wonder how it could be that Davidson came to give an account in which all thought of agency is thought of what is past, and from which the idea of progress is entirely absent. Well, this is surely owed to the fact that from the outset of his writing on action, the ex post facto stance of one who explains why someone has done something is in place. Davidson’s question was: ‘What is the relation between 5 Davidson

fails to mention that the Meyer of his example, the German geologist Hans Meyer who climbed the mountainous volcano Kibo in 1889, had attempted to climb Kibo two years earlier and had had to turn back when he found he lacked necessary equipment. In 1887, Meyer was climbing Kibo, but he did not climb Kibo in 1887. 6 In the course of criticizing Kenny’s account of action sentences, Davidson said ‘“Cass walked to the store” can’t be given as “Cass brought it about that Cass is at the store”, since this drops the idea of walking’ (p. 110 in 1967b). It could equally be said that it drops the idea of Cass’s having made progress towards the store (by walking there, as it is in this case). Kenny, for his part and unlike Davidson, did introduce an idea of making progress, and progress which might be halted. For Kenny’s “performance verbs” are Davidson’s action verbs; and for these, Kenny endorsed the schema ‘S was ϕ-ing not only if S ϕ-ed’ (p. 123 in Kenny 1963). And he provided examples: ‘A man may be walking to the Rose and Crown, and yet never walk there, perhaps because he was run over on the way’; and ‘One can start knitting a sweater and produce a scarf’ (p. 122).

8

J. Hornsby

a reason and an action when the reason explains the action by giving the agent’s reason for doing what he did?’ (p. 3 in 1963, my italics). Davidson explicitly introduced actions as entities at the start of ‘The Logical Form of Action Sentences’. Jones did it slowly, deliberately, in the bathroom, with a knife, at midnight. What he did was butter a piece of toast. We are too familiar with the language of action to notice at first an anomaly: the ‘it’ of ‘Jones did it slowly, deliberately, …’ seems to refer to some entity, presumably an action …. (p. 105 in 1967b).

Does the ‘it’ here really seem to refer to an entity? Davidson himself used a bare infinitive to say what it was that Jones did: ‘what he did was butter a piece of toast’. Did Davidson suppose that an unrepeatable particular or ‘concrete individual’ is referred to with the words ‘butter a piece of toast’? Butter a piece of toast is something Jones did and might do again tomorrow, and something which anyone might do at one or another time. If butter a piece of toast is to be called an action, then actions are not entities. ‘Jones buttered the toast’ predicates ‘buttered the toast’ of Jones, and it makes reference to no entities save for Jones and the toast. We saw that events stand alongside objects as a kind of particular in Davidson’s ontology. It could be that dissatisfaction with Quine’s four-dimensional objects led Davidson to think that there must be two different kinds of spatiotemporal entity. At any rate, when it came to treating action, Davidson sought paraphrases of ordinary language sentences in the predicate calculus, using quantification over events, with an existential quantifier signifying real existence (such as is conveyed in Quine’s slogan ‘To be is to be the value of a variable’).7 I suggest that this stood in the way of his recognizing that his action verbs are used to make predications—that something is predicated of Jones when it is said that he buttered the toast, and something is 7 Davidson’s paraphrases were at the service of providing a treatment of adverbs, many of these being predicates of events in Davidson’s view. But if it is allowed that adjectives modify predications made using nouns—as, say, ‘short order’ or ‘experienced’ can modify ‘cook’—then it can be allowed that adjectives modify predications made using verbs—as ‘slowly’ and ‘in the bathroom’ modify ‘buttered the toast’. I simplify here. Just as there are distinctions to be made between kinds of predication, so in turn there are distinctions to be between different modes of modification of predicates. Although this isn’t the place to delve into questions about meaning theories, I would note just two points. (1) In providing a free-standing account of his (past-tensed) action sentences, Davidson apparently contravenes his own principle that a natural language meaning theory should show how the meaning of a sentence depends on the meanings of its parts, and thereby show how knowledge of a finite number of features of a language suffices for the understanding of all the sentences of that language (See Davidson 1967c.). Adherence to that principle would require, for instance, that ‘climbed’ be seen as composed from ‘climb’ and ‘-ed’ (the past tense morpheme), so that ‘climb’ could be seen as recurring in ‘will climb’, ‘was climbing’, ‘is climbing’, etc., ‘-ed’ be seen as recurring in ‘buttered’, ‘stabbed’ etc., and ‘was — ing’ be found in ‘was climbing’ ‘was buttering’, ‘was stabbing’, etc. (2) Davidson’s assumption that action verbs are monadic fails to allow for what I spoke of as ‘acting on an object’. To allow for that, one would need to take the ‘buttered’ in ‘Jones buttered the toast’ to introduce a relation between Jones and the toast, sc. the relation in which a pair of objects stand if the first buttered the second.

Objects and Events in Davidson

9

predicated of Jones when he is said to be buttering the toast. Again, although there is now a different form of predication, something is predicated of Jones when Jones is said to be a cook (Compare ‘A cook is what [/something] Jones is’ with ‘Buttering the toast is what [/something] Jones is doing.).

4 Causality in Action I have meant to cast doubt on whether Davidson justified his treatment of actions as entities. But however that may be, if we are to understand ‘events’ as Davidson did, then we must take events to be such things, and the only such things, as stand one to another in relations of cause and effect. And it is the causal character of his theory of action which I shall explore now. In Davidson’s account of what it is for there to have been an action, ‘cause’ enters at two points. ‘Cause’ relates reasons and actions; and ‘cause’ relates bodily movements and events which they cause and in whose terms the movements can be re-described. Hence, Davidson’s claim that ‘All I ever do is move my body; the rest is up to nature’. In his picture, reasons cause bodily movements from which issue chains of events in accordance with the workings of nature, so that there are effects ‘the agent has wrought in the world beyond his skin’. ‘Mere movements of the body are all the actions that there are’, Davidson said (p. 55 in 1971).8 There appear to be two ways in which Davidson’s ‘mere movements’ might be understood. On the one hand, Davidson spoke of them as ‘actions we do not do by doing something else’ so that a mere movement might be thought to be a person’s moving her body—someone’s raising her arm, for example. On the other hand, mere movements take place in nature where they cause actions’ effects, so that each of them might be thought to be a body of a person’s moving—someone’s arm’s rising, for example. Well, Davidson came to think it wrong to suppose we have to choose between these two understandings.9 He wrote: ‘If my arm went up because I raised it nothing must be added to my arm going up to make it a case where I raised my arm. … If I raise my arm, then my raising my arm and my arm rising are one and the same event.’ (p. 103 in 1987a). This seems right: someone’s moving a part of herself is a case of self-movement: one’s moving (a part of) oneself and the movement of (that part of) oneself are not two different things. Consider what it is that one sees when one sees someone (say) raising her arm. One is not inclined to say that one sees both her doing something and her arm’s doing something. If it were supposed that really one saw only her body moving and that her moving her body was something different, 8 Notice

that the present tense of the ‘do’ and ‘cause’ of this account has to be understood as what I called generic, in line with ‘I flip the switch’, ‘I turn on the light’. 9 I say ‘came to think’ because in his earlier writings (both in 1963 and in 1971), Davidson gave every evidence of thinking that someone’s moving her body and her body’s moving are distinct events. I assumed that Davidson did think that (or perhaps I assumed that it was the right thing to think) when I wrote Hornsby 1980. Throughout that book, the many mistakes that ensue upon thinking it evolved. Suffice here to say that I have changed my mind about any number of things!

10

J. Hornsby

then it would seem to be supposed that her moving her body was invisible—as if a person could never be seen to be at work in the world. We can ask now what it is for someone to be moving an object which is not any part of herself. So suppose that she raised the glass held in her right hand by raising her right arm. In this case too, her raising the glass and the glass going up, thought of as things that were actually going on, will strike one as being one and the same. Like her arm, the glass went up because she raised it. One’s intuition, then, is that her raising the glass no more introduces an event additional to the glass’s going up than her making her arm go up introduces any additional event. If she is aware that she is holding the glass, as one may assume she is if she intentionally raises it, then it cannot really be, as Davidson would have it, that it is up to her whether her arm moves, but only up to nature whether the glass moves. There is a general question here about what it is to act on an object, whether to raise something, to stab something, to pick something up, to put something down, to rip something up, to chop something into pieces, to open something, to break something, to squash something ….10 Davidson thought that an event of an agent’s acting on an object is always to be distinguished from something’s thus happening to the object, so that action on an object everywhere consists in the relation cause obtaining between two events. He was explicit about this when he considered an example of his closing a door, saying that it is ‘an error to confuse what my action does cause—the closing of the door—with something utterly different—my action of closing the door’ (p. 56 in 1971). But when this is treated as an actual example, so that ‘the closing of the door’ applies to an event there was, we have to imagine a particular closing of a particular door, and we know which closing it was. It was the very closing of the door in which Davidson participated—the closing of the door by Davidson, as one might say. Is this really different from Davidson’s closing of the door? There is no doubt that we can differentiate in thought between a sort of thing someone might do at some time—put a book on the table, say—and a sort of terminal state that would ensue from a thing of that sort’s being done—the book’s being on the table. So when confronted with an example—Don’s putting the book on the table, say—we can find two different things to think about—something involving Don and something involving the book. But if we are to suppose that an action has actually taken place, so that there exists an event as Davidson meant that, then we must think that Don himself actually put the book on the very table he put it on. And then we shall be hard-pressed to distinguish between something in which Don was involved and some additional thing in which the book was involved. What we think of can just as well be described either as Don’s putting the book on that table or as the book’s being put on that table by Don. 10 If you want some further examples of verbs of acting on, you might note that it is an offence in English law for anyone to mutilate, kick, beat, nail or otherwise impale, stab, burn, stone, crush, drown, drag or asphyxiate any wild mammal with intent to inflict unnecessary suffering. Such examples may help to make it clear that for someone to have been acting on an object for a stretch of time, she need not have been in unmediated physical contact with the object throughout that stretch.

Objects and Events in Davidson

11

If the foregoing has something right, then it must be allowed that change does not require that there should be two events standing to one another in the relation of ‘cause’. Change can happen by virtue of one object’s acting on another.

5 Perception, Objects and Events It is possible to see a sort of likeness between Davidson’s account of action and what he had to say about perception in his paper ‘A Coherence Theory of Truth and Knowledge’ (1983). We’ve seen that Davidson took bodily movements to connect our reasons for acting with the world within which we act, so that bodily movements are sandwiched, as it were, between agents’ states of mind and their effects beyond agents’ bodies. When it comes to perception, Davidson took beliefs to be ‘states of people that [among other things] are caused by … events inside …the bodies of entertainers’ (p. 138 in 1983); and he said that ‘Sensations are what connect the world and our beliefs’ (p. 142 in 1983). Sensations, then, are sandwiched between states of the external world and their effects on perceivers’ states of mind. As in the case of action, so in the case of perception, objects feature in Davidson’s story only insofar as they participate in such events as are mentioned in statements recording perceivers’ mental states/events. Objects don’t feature in their own right, as it were: they were not assigned any causal role in Davidson’s account of perception in 1983.11 Still, Davidson appears to have changed his mind about how we should think about the perception of objects. When he introduced his idea of triangulation, as ‘a three-way relation among two speakers and a common world’ (p. xv in 2001d), he treated a case in which each of the two speakers is related to an object—a table in his example (pp. 117–8 in 1992). It could be that Davidson meant us to think of the two speakers as responsive merely to ‘features of reality’, rather than to be seeing the table as such. I’m not sure. But in any case I imagine that Davidson would have wanted to allow for a relation between a person and an object when she sees it. The perception of objects would seem to be fundamental in an explanation of our ability to have thoughts about objects, to act on them, and to have one or another sort of knowledge about them. So it might well be that Davidson would have wanted to adjust the account he gave of perception in 1983, an account which, in talking of beliefs as effects of ‘states of the external world’ apparently confined itself to perceiving that – –. However that may be—whatever view of perception Davidson may have settled on—he abandoned what had seemed to be the metaphysical basis of his causal theories, according to which objects and events are members of disjoint metaphysical categories. For in his later work, Davidson often talked of ‘objects and events’ in 11 He

might even have allowed a sense in which an object stands in relation to a person who sees it, a relation of acting on. If the words ‘act on’ don’t come as naturally here as they do when a person is, say, picking an object up, perhaps this is because there is always a physical distance between a perceived object and the sense organs by virtue of whose workings it can be perceived.

12

J. Hornsby

conjunction. He spoke, for instance, of ‘events and objects outside of the person’ (in 1987b, 18); and he said that ‘words and thoughts are, in the most basic cases, necessarily about the sorts of objects and events that commonly cause them’ (p. 45 in 1988). There are many further examples of Davidson’s talking of objects and events in a single breath—examples in which Davidson saw no need to distinguish between the parts that may be played by objects and the parts that may be played by events in the causal world. For just some of these examples, see p. 127 in 1980b; p. 162 in 1982b; p. 304 in 1985a; p. 114 in 1987b; pp. 20, 23, 27, 29 in 1987b; pp. 50, 52 in 1988; p. 202 in 1990a; p. 96 in 1991; p. 120 in 1992; p. 9 in 1995a; p. 47 in 1995b; p. 127 in 1995c; p. 86 in 1998; p. 137 in 2001b. Evidently Davidson failed to adhere to his claim that events alone play a causal role. When he turned to questions in epistemology, he cast himself far adrift from the metaphysical distinction he had made, between two different categories of spatiotemporal particulars. He came to appreciate that objects, not just events, must be mentioned in saying what transpires in the world of change that we inhabit, and that objects must be taken to possess causal powers.

6 Davidson’s Quasi-nominalism, and Causal Powers When he was engaged in metaphysics, Davidson said that he was “silent about processes, states and attributes if these are different from individual events” (see Sect. 2 above). Nonetheless, Davidson was happy to use terms for all such things. I pointed out how often he had things to say about processes. And under the head of ‘attributes’, he brought what he had spoken of as ‘traits’, as ‘properties’, as ‘dispositions’, and as ‘abilities’, all of which he attributed both to people and to other objects. Davidson was something of an ontological liberal.12 Unlike Quine, who was guided by maxims of scientific acceptability and ontological economy, Davidson ‘was not concerned to improve on [natural language] or to change it’ (p. 305 in 1985b). Yet Davidson followed Quine in a kind of nominalism inasmuch as he insisted that the semantic roles of predicates of all sorts was to be accounted for making use of the notion of satisfaction, satisfaction being a sort of converse of ‘true of’. Taking such a view of predication, and having defended events as members of the domain of a first-order quantifier, it is perhaps only to be expected that Davidson should not have shown much interest in the variety of things that are introduced when predications are made of attributes, whether these attributes be traits or dispositions or abilities. Attributes of all these sorts may be reckoned causal powers. And Davidson did make use of the idea of a causal power in his early work. He spoke of ‘solubility and of being a solvent as equally causal powers’, of ‘what it is an agent’s power to do’, 12 Or at least came to be an ontological liberal. In 2005, he wrote: ‘[1] If some (second-level) predicates are true of abstract objects, those objects must exist’. But he also wrote ‘[2] their existence does not explain the role of such predicates’ (p. 158 in 2005b). I take [1] to be an expression of ontological liberalism, in which Davidson can be contrasted with Quine, and [2] to be an expression of a kind of nominalism that he shared with Quine.

Objects and Events in Davidson

13

and of ‘the actualizing of a power’ (1973b). But he never brought the fact that agents have causal powers into his account of action. Nor therefore did he think of agents’ powers as ever actualized—as realized, or exercised, when they act. There is an explanation of why Davidson should have left the powers of human agents out of his philosophy of action. When reason explanations are provided, there is a species of causality that belongs with an idea of a person’s doing something which is not in view. Davidson’s argument that reasons are causes of actions started from supposing that someone had intentionally done something, and sought to uncover what is involved in saying why he did what he did. The fact of his possessing the power to do it is then taken as read: it does not belong in an explanation why he did it, even though he could not have done it unless he had possessed the power to do it—had had the relevant ability or capacity or skill. With the focus on the question why, it is easy to forget that the agent was playing a causal part—was exercising one or another agential power—so long as he was doing whatever it was. If instead of asking about reason explanations, we ask what it is for someone to act in this, that or the other way, then it comes to attention that agents have the power to do what they do, and that someone’s actually doing something is a matter of their exercising their powers. I think that the phenomenon of human agency will be reconceived if one abandons Davidson’s idea that causality is found only insofar as causal relations obtain between events. People can make such marks on the world as they intend to by virtue of being equipped with the powers of rational agents. And once it is granted that they possess those powers, there is no call to think that their acting has to be a matter merely of their moving their bodies, leaving it up to nature to provide for their doing anything else that they might want to. Davidson came to say that ‘ordinary explanations of action, perception, memory, and reasoning, as well as the attribution of thoughts, intentions, and desires are riddled with causal concepts’ (p. 96 in 1990b). I take it that where causal concepts are used in explanations and attributions, causal powers are seen to be at work, and that the causality then in the picture is not reducible to event–event relations. It seems then that Davidson came to allow a conception of causality in no way reliant on his metaphysical conception of events. Even so, he never revised his account of action.13

13 The

account he gave in ‘Agency’ (1971) is rehearsed in 1982a. And it is spelled out again in the essay ‘Aristotle’s Action’ (2001e), where Davidson persists both in his view that causality is (exclusively) a matter of events standing to one another in the relation cause, and in his view that if an agent does something to an object, there is an event that is a change in the agent and a different event that is a change in the object (here a patient). Davidson represents himself as sharing a view of agency with Aristotle. But this seems plainly wrong. Aristotle said: ‘A thing is capable of causing motion because it can do this, it is a mover because it actually does it. But it is on the movable that it is capable of acting. Hence there is a single actuality of both alike.’ (Physics, Book 3, Part 3.).

14

J. Hornsby

7 Nomologicality and Davidson’s Monism of Events Evidently, I have some fundamental disagreements with Davidson. So it is in the matter of his argument for anomalous monism. Nevertheless, I think that there is no need to dispute the idea which I take to underlie his monistic thesis. So I want to finish by returning in critical spirit to anomalous monism as Davidson first presented it, but to go on to suggest that something deserving the name of anomalous monism might be endorsed by one who accepts Davidson’s vision of ourselves as distinctively rational beings. Davidson’s anomalous monism is a species of physicalism, and, as such, a response to the mind–body problem. Since that problem has always seemed to concern substances (minds and bodies, if you like), it can be perplexing that a monistic thesis concerning events should be supposed to provide an answer to it. But it is intelligible that Davidson should have thought that a thesis about events was needed to avoid dualism when one takes into account his view that: [T]he causal nexus provides for events a comprehensive and continuously usable framework for the identification and description of events analogous in many ways to the space–time coordinate system for material objects (p. 180 in 1969a).

This can make it seem that insofar as the world is a causal world a place for events must be provided for, and objects can go along for the ride as it were. It’s as if we must suppose there are two classes of entities, one class responsible for the world’s being a changing world wherein there is psychophysical causation, the other class serving for the world’s being a spatial world in which material objects may exist at one or another time. Perhaps one is bound to suppose this if our inhabiting a causal world is to be taken to be a matter simply of events standing in causal relations one to another. But to take that to be so would be to deny that objects, including people, have genuinely causal powers. Davidson invoked his Principle of the Nomological Character of Causality in order to argue for identities between mental events (events described with the everyday vocabulary of the mental) and physical events (events described using the vocabulary in which laws are framed). It must be a question whether the events of the Nomologicality Principle can be conceived as Davidson’s metaphysics demands. Commitment to events’ existence, as genuinely concrete entities, I argued, could only be commitment to entities of the past. A quite different commitment is made when the Nomological Principle of Causality is put forth. For the Principle says that events related as cause and effect ‘have descriptions that instantiate a law’. And ‘lawlike statements’ are ‘general statements, supporting counterfactual and subjunctive claims, and supported by their instances’ (in 1970a, 217). Lawlike statements quantify over events whenever they might occur. So the events of the Nomologicality Principle cannot be the events treated by Davidson in his account of action. And if a quite different conception of events was adduced when Davidson concerned himself with agency from that which is in play when laws are stated, then he could hardly have established identities of mental events with physical events. Davidson then appears to be left without any actual argument for his physicalism.

Objects and Events in Davidson

15

I’ve suggested that the strictures on what causation requires found in Davidson’s early work were silently set aside in much of his work from the 1990s and on. It might then be thought that Davidson ought simply to have owned up to his new understanding of the role that, thanks to their powers, objects play in the causal world and given up on the whole idea of an anomalous monism. Yet I think it right to credit Davidson with an idea that underpinned his version of monism. ‘Nomological slack between the mental and the physical is essential as long as we conceive of man as a rational animal’, Davidson said in ‘Mental Events’ (p. 223 in 1970a). And later he said ‘the concepts of “folk psychology” cannot be incorporated into a coherent and comprehensive system of laws of the sort for which physics strives’ (pp. 24–5 in 1987b). Forgetting the difference between causation and causal explanation (cp. n. 3 above), Davidson’s idea was surely that such causal explanations as depend upon the use of concepts brought to bear when a being is treated as a rational animal cannot be served up by physical (natural scientific) theory; but the explanations for all that are of happenings in the spatiotemporal world, requiring no kind of dualism.14 Of course there may be some who will quarrel with this idea. But I think that Davidson gave it further substance when he turned to the work which gave rise to his overall picture of kinds of knowledge and their interdependence (2001d). Back in ‘Actions, Reasons and Causes’, Davidson said that ‘justifying and explaining an action so often go hand in hand’ (p. 8 in 1963). Rational explanation of action, then, is a special case of a causal explanation—not mere causal explanation, as one might put it. And when he said ‘Causality and rationality go hand in hand in important cases’ (p. 290 in 1977), he didn’t confine himself to the explanation of action. Later, when he came to speak of range of phenomena of the mind, which he spoke of as ‘riddled with causal concepts’ he came to say that ‘Ordinary explanations of action, perception, memory, and reasoning, as well as the attribution of thoughts, intentions, and desires, is riddled with causal concepts’ (p. 96 in 1990b). I take it that where causal concepts are used in explanations and attributions, causal powers are seen to be at work, and in the case of the phenomena of concern to Davidson when he wrote about subjectivity, intersubjectivity and objectivity, these are causal powers proper to rational beings, without whose attribution human knowledge of any kind cannot be understood.

14 I should acknowledge that Davidson always intended something more precise than this by ‘anoma-

lous monism’. In two essays printed in his 2005a, he defends anomalous monism against criticisms of his ‘Mental Events’ (1970a). He then relies on his conception of laws as not containing causal concepts (see p. 71 in 1997), on a distinction between strict and non-strict laws which he says is essential to the argument for anomalous monism, and he argues that the notion of supervenience he had intended has been misunderstood. So Davidson made a new case for his doctrine. I can’t attend to it here. I would only say that his conception of events as particulars, which I criticized for failing to allow that action ever is or was ongoing, is retained.

16

J. Hornsby

7.1 Coda I have attended to Davidson’s metaphysics, in part because I think it has been little attended to, and in part because it draws attention to a change of view on Davidson’s part about which he isn’t explicit, and thus which isn’t often recognized. I have drawn attention to the change of view, wanting to find in his determination to portray ourselves as rational animals something which guided Davidson’s work from start to finish and which many of us will applaud. So far as Davidson’s work on agency is concerned, I have been relentlessly critical. I think that my criticisms continue to have a point insofar as it is not Davidson alone to whom they might be directed. Taking his event ontology, as it shows up in his account of the logical form of action sentences, to be a solution to many problems, Davidson said: The solution I have proposed to these problems may at first seem ontologically excessive, but I believe it is now fairly widely accepted by logicians, philosophers, and those linguists who are interested in the semantics of natural languages. (p. 286 in 2001e).

I think that Davidson is correct in his belief, and that inside and outside philosophy a notion of event which conforms to his metaphysical conception is assumed.15 The conception of events that Davidson worked with in presenting an account of agency required him to think of actions one at a time, as it were, and all of them over and done. And this in turn requires a view of agency from outside the course of life, an exclusively third-personal view in which there is no experience of the passage of time. Davidson thus has no truck with the perspective of one who is leading a life, and who knows her own reasons for doing what she is doing. As it seems to me, it is no accident that when it comes to ‘Knowing One’s Own Mind’ (1987b), Davidson fails to include under the head of such self-knowledge, one’s knowing what one is oneself doing. (I find myself discontent with Davidson on the subject of self-knowledge more generally. But that surely is another story.)

References Aristotle. Physics. www.filosofia.unimi.it ›zucchi› NuoviFile› Barnes—Physics. Davidson, D. 1963. Actions, reasons and causes. In Davidson, Donald. 2001c (pp. 3–19). Davidson, D. (1967a). Causal relations. In Davidson, Donald. 2001c (pp. 149–162). Davidson, D. (1967b). The logical form of action sentences. In Davidson, Donald. 2001c (pp. 105– 122). Davidson, D. (1967c). Truth and meaning. Synthese, 17(3), 304–323. Davidson, D. (1969a). The individuation of events. In Davidson, Donald. 2001c (pp. 21–42). Davidson, D. (1969b). How is weakness of the will possible? In Davidson, Donald. 2001c (pp. 207– 227). 15 Myself

I cannot find it regrettable, as Davidson did, and as many now would, that ‘Aristotle was not in a position to discuss [questions about agency] with the relative clarity with which [they] can be raised in the context of today’s logic and semantics’ (p. 284 in 2001e).

Objects and Events in Davidson

17

Davidson, D. (1970a). Mental events. In Davidson, Donald. 2001c (pp. 207–227). Davidson, D. (1970b). Events as particulars. In Davidson, Donald. 2001c (pp. 181–187). Davidson, D. (1971). Agency. In Davidson, Donald. 2001c (pp. 43–61). Davidson, D. (1973a). The material mind. In Davidson, Donald. 2001c (pp. 245–259). Davidson, D. (1973b). Freedom to act. In Davidson, Donald. 2001c (pp. 63–81). Davidson, D. (1977). Hume’s cognitive theory of pride. In Davidson, Donald. 2001c (pp. 277–290). Davidson, D. (1978). Intending. In Davidson, Donald. 2001c (pp. 83–102). Davidson, D. (1980a). Introduction to first edition of Davidson, Donald. 2001c (pp. xv–xxi). Davidson, D. (1980b). A unified theory of thought, meaning, and action. In Davidson, Donald. 2004 (pp. 151–166). Davidson, D. (1982a). Rational animals. In Davidson, Donald. 2001d (pp. 95–105). Davidson, D. (1982b). Empirical content. In Davidson, Donald. 2001d (pp. 159–175). Davidson, D. (1983). A coherence theory of truth and knowledge. In Davidson, Donald. 2001d (pp. 137–152). Davidson, D. (1984a). First-person authority. In Davidson, Donald. 2001d (pp. 3–14). Davidson, D. (1985a). Adverbs of action. Appendix A in Davidson, Donald. 2001c (pp. 293–304). Davidson, D. (1985b). Reply to Quine on events. Appendix B. In Davidson, Donald. 2001c (pp. 305– 311). Davidson, D. (1986). The interpersonal comparison of value. In Davidson, Donald 2004 (pp. 59–74). Davidson, D. (1987a). Problems in the explanation of action. In Davidson, Donald. 2004 (pp. 101– 116). Davidson, D. (1987b). Knowing one’s own mind. In Davidson, Donald. 2001d (pp. 15–38). Davidson, D. (1988). The myth of the subjective. In Davidson, Donald. 2001d (pp. 39–52). Davidson, D. (1990a). Epistemology externalized. In Davidson, Donald. 2001d (pp. 193–204). Davidson, D. (1990b). Representation and interpretation. In Davidson, Donald. 2004 (pp. 87–99). Davidson, D. (1991). Three varieties of knowledge. In Davidson, Donald. 2001d (pp. 205–220). Davidson, D. (1992). The second person. In Davidson, Donald. 2001d (pp. 107–121). Davidson, D. (1995a). The problem of objectivity. In Davidson, Donald. 2004 (pp. 3–18). Davidson, D. (1995b). The objectivity of values. In Davidson, Donald. 2004 (pp. 39–57). Davidson, D. (1995c). Could there be a science of rationality? In Davidson, Donald. 2004 (pp. 117– 134). Davidson, D. (1997). Indeterminism and anti–realism. In Davidson, Donald. 2001d (pp. 69–84). Davidson, D. (1998). The irreducibility of the concept of the self. In Davidson, Donald. 2001d (pp. 85–91). Davidson, D. (2001a). Introduction to Davidson, Donald 2001d (pp. xiii–xviii). Davidson, D. (2001b). What thought requires. In Davidson, Donald. 2001d (pp. 135–149). Davidson, D. (2001c). Essays on actions and events (2nd edn.). Oxford: Clarendon Press. Davidson, D. (2001d). Subjective, intersubjective, objective. Oxford: Clarendon Press. Davidson, D. (2001e). Aristotle’s action. Reprinted in English (the original was in French) in Davidson, Donald 2004 (pp. 277–294). Davidson, D. (2004). Problems of rationality. Oxford: Clarendon Press. Davidson, D. (2005a). Language truth and history. Oxford: Clarendon Press. Davidson, D. (2005b). Truth & predication. Cambridge, MA: Harvard University Press. Hahn, L. E. (Ed.). (1999). The philosophy of Donald Davidson. IL: Open Court Publishing. Hornsby, J. (1980). Actions. London: Routledge and Kegan Paul. Hornsby, J. (1999). Anomalousness in action. In Hahn 1999 ed (pp. 623–635). Kenny, A. (1963). Action, emotion and will. London: Routledge & Kegan Paul.

Davidson’s Meta-normative Naturalism and the Rationality Requirement Robert H. Myers

Abstract According to Michael Smith, normative realists cannot explain why people act on their normative judgments as reliably as they do, so we should embrace constructivism (or perhaps constitutivism) instead. Normative realists have countered by insisting that they can establish (1) that rationality typically requires people to act on their normative judgments and (2) that people typically are rational. I acknowledge that normative realists can establish (1) but argue that (2) poses problems both for non-naturalists and for reductive naturalists. I close by suggesting that Davidson’s non-reductive (or “anomalous”) form of naturalism may give normative realists a better chance of success. Keywords Normative reasons · Practical rationality · Reductive naturalism · Non-reductive naturalism · Anomalous monism · Non-naturalism

1 Introductory Remarks Although Donald Davidson is best known for his account of motivating reasons, toward the end of his life he did write about normative reasons, arguing for a novel form of realism possessing two very attractive features. On the one hand, it is not just non-reductive but also non-revisionary, refusing to compromise in any way on the thought that the prescriptive authority of normative reasons is objective and extends to all possible agents. On the other hand, it still treats normative properties as causal properties and thus avoids many of the epistemological problems that bedevil realisms of the sort recently advanced by Derek Parfit, David Enoch, and Thomas Scanlon.1 1 See,

e.g., Parfit (2011, Chap. 32), Enoch (2011, Chap. 7), and Scanlon (2014, lecture 4).

This chapter is in its final form, and it is not submitted for publication anywhere else. R. H. Myers (B) Department of Philosophy, York University, 4700 Keele Street, Toronto, ON M3J 1P3, Canada e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_2

19

20

R. H. Myers

Revisionary naturalists and non-naturalists will of course push back against the possibility of such an anomalous naturalism, as we might call it, by insisting that strongly prescriptive properties could not possibly be determined by—because they would be “just too different” from—basic physical properties in the way that causal properties must be.2 They will insist that normative realists must choose: If their principal concern is to provide a causal epistemology for normative judgments, they will have to make their peace with revisionary naturalism; if their principal concern is to maintain the strong prescriptivity of normative truths, they will have to take their chances with non-naturalism. As I have argued elsewhere, I think this is too quick.3 Strongly prescriptive properties would undoubtedly be very different from basic physical properties, since they would be strongly prescriptive and basic physical properties would not, but this does not entail that the differences between these properties would be so great that strongly prescriptive properties could not possibly be determined by basic physical properties in the way that causal properties must be. Understanding how basic physical properties determine strongly prescriptive properties may be difficult, but this is not enough, by itself, to prove that they do not. Perhaps an explanation will be found; perhaps these relations are ultimately brute. In fact, my view, for what it is worth, is that this worry about anomalous naturalism is less serious either than the worry that revisionary naturalism gives up on normative reasons when it gives up on strong prescriptivity or than the worry that non-naturalism makes normative reasons “queer” by making knowledge of them sui generis.4 However, since this assessment of the dialectical situation is far from being common ground among normative realists, the question I would like to take up in this chapter is whether there is some other ground on which normative realists might be induced to choose Davidson’s anomalous naturalism over revisionary naturalism and non-naturalism. My answer to this question will be that such common ground can be found in the thought that an acceptable account of normative reasons must be able to meet what we might call the rationality requirement. Importantly, this requirement has two parts: An acceptable account of normative reasons must be able to make sense of the fact (1) that rationality typically requires people to form motivating states in accord with their normative judgments and (2) that people typically are rational. My claim will be that normative realists of all stripes can make good sense of (1), but that (2) raises very serious problems for them that only anomalous naturalism can overcome. This rationality requirement should be understood as the logical culmination of a sequence of progressively more plausible requirements—most notably, to my mind, Bernard Williams’s “explanatory” requirement, Christine Korsgaard’s “internalism” 2 The

“just too different” slogan is due to Enoch. See, e.g., his 2011, p. 4. Myers (2019). Davidson famously defends anomalous monism in the philosophy of mind in his 1970. 4 For the unfortunately named but still extraordinarily influential “argument from queerness,” see Mackie (1977, Chap. 1). 3 See

Davidson’s Meta-normative Naturalism and the Rationality …

21

requirement, and Michael Smith’s “practicality” requirement—that opponents of realism have articulated over the years in their attempts to find fault with that view. I shall briefly recap this (I trust familiar) history in Sect. 2 with a view to demonstrating that, although the earlier requirements can in every case be dismissed by realists as begging the question against them, the rationality requirement in which they culminate looks to be one that even realists may have to accept. Section 3 will consequently be given over to considering whether revisionary naturalists and non-naturalists can meet this rationality requirement, or whether, failing that, they could after all dismiss it as begging the question against them. As I indicated a moment ago, my conclusion will be that the first part of the rationality requirement causes them little trouble, but that the second part is significantly more difficult for them to deal with. Not surprisingly, this encourages a number of them to question whether it actually is a requirement on reasons that people should typically be rational. But that, as I shall argue, is an awkward position to sustain. Finally, in Sect. 4, I shall demonstrate that this is precisely where Davidson’s anomalous naturalism enjoys an advantage; by allowing that strongly prescriptive properties can be involved in causal explanations of normative judgments and motivating states, it obliges us to ask what it could be about people that makes them receptive to such influences. And since the Davidsonian answer is that people must have a systemic interest in getting normative matters right, it turns out that, for Davidson, the rationality of people is a requirement on the existence of reasons. It is only because people typically are rational that reasons can exist and have an influence on them.

2 Three Precursors of the Rationality Requirement Almost forty years ago, Bernard Williams proposed a requirement on reasons for action that still has many adherents. More influential than the requirement itself, however, have been the thoughts about explanation leading up to it: If something can be a reason [a normative reason] for action, then it could be someone’s reason [someone’s operative reason] for acting on a particular occasion, and it would then figure in an explanation of that action. (1980, 106)

This does not tell us how reasons for action should be expected to figure in explanations of people’s actions when people succeed in acting on their reasons.5 But Williams very quickly clarified this point by saying: …the agent should acquire the motivation [to act in the prescribed manner] because he comes to believe the reason statement, and…he should do the latter…because, in some way, he is considering the matter aright. (1980, 108–9)

5 Unless

I note otherwise, I will use the term "reason" throughout this paper to refer to normative reasons, not to operative or motivating reasons.

22

R. H. Myers

For anyone interested in the truth of normative realism, the next move was the real cause for concern. For Williams went on to explain what he took to be involved in considering matters aright by saying: … the…reasons statement…will have to be taken as roughly equivalent to, or at least entailing, the claim that if the agent rationally deliberated…he would come to be motivated to [act in the prescribed manner]. (1980, 109)

Williams’s account of rational deliberation was Humean in nature, in the sense that it aimed at a kind of desire satisfaction compatible with the possibility of relativism.6 So, we were left with something like this: Williams’s explanatory requirement: A has reason to Ø in circumstances C if and only if she would come to be motivated to Ø in C if she were empirically informed and deliberated rationally (in the Humean sense).

But how, more precisely, were Williams’s initial thoughts about explanation supposed to lend support to his claim that people’s reasons for action must be accessible to them via informed and rational deliberation? Normative realists deny that people’s reasons for action must be accessible to them via informed and rational deliberation. They think it is perfectly possible that an empirically informed person who is not also normatively informed might lack a rational way of working out the truth about her reasons for action. But since they do not think rational deliberation is the only way to acquire true beliefs, they do not think this inability to infer her way to the truth would rule out the possibility that this person could still succeed in recognizing her reasons and acting on them. Consider Gilbert Harman’s example, in which some boys are about to set fire to a cat. Why could it not happen that a person comes to believe she has reason to intervene simply because she suddenly appreciates that what the boys are doing to the cat is cruel? What prevents a reason for action from figuring so directly in the explanation of an action?7 Instead of squarely addressing this pivotal explanatory question, Williams tended at this point to go normative, assuming that a person could have a reason to perform some action only if informed and rational deliberation would make her such that action would have some appeal for her, inasmuch as she would then possess aims that would be furthered by that action. But of course this is not an assumption any normative realist would feel obliged to accept. It begs the question against them even more blatantly than Williams’s initial claims about explanation do. Harman himself takes a different line on this issue, allowing that direct explanations by reasons may be possible, but insisting that they are never actually as good as explanations couched in the Humean terms Williams prefers. This is a difficult case to make, however, since what counts as a good explanation of action is by 6 More

could of course be said about what, exactly, distinguishes Humean accounts of deliberation from others, but this should do for our purposes here. 7 For Harman’s example, see his 1977, Chap. 1. McDowell (1995) raises similar questions about the implications to be drawn from examples of this sort.

Davidson’s Meta-normative Naturalism and the Rationality …

23

no means a matter of agreement among the many different parties to these metanormative debates.8 If reasons for action have an important explanatory dimension, an explanatory dimension that creates serious difficulties for realists, it must surely be because there is some more widely accepted fact that stands in need of explanation. As we shall see in a moment, this is precisely the direction in which Smith develops Williams’s argument. Korsgaard develops it differently, however, so let us quickly look at her position before turning to Smith’s. Korsgaard rejects Williams’s Humean account of rational deliberation and in particular the relativism that Williams allows for, preferring to think of deliberation in Kantian terms, but her requirement on reasons is otherwise similar to his.9 And at first sight, the argument that she offers for her requirement seems to rest on an even stronger explanatory claim than his does, for her claim seems to be that people’s reasons for action must always exercise some influence over the motivations shaping their actions. As she says: An internalist theory is a theory according to which the knowledge (or the truth or the acceptance) of a moral judgment implies the existence of a motive (not necessarily overriding) for acting on the judgment. If I judge that some action is right, it is implied that I have, and acknowledge, some motive or reason for performing that action…On an externalist theory, by contrast, such a conjunction of moral comprehension and total unmotivatedness is perfectly possible: knowledge is one thing and motivation another. (1986, 8)

As she goes on to clarify, however, this is misleading in two respects. First, the distinction needs to be applied more broadly; it holds, not merely of moral judgments, but of normative judgments generally. Second, the distinction needs to be drawn more carefully, so even internalists can acknowledge the possibility of what Korsgaard calls (1986, 12) true irrationality. In saying that a normative judgment implies the existence of a motive to act accordingly, internalists do not mean to deny that someone could sincerely make such a judgment but utterly lack the corresponding motive. Their point is rather that this would necessarily reveal a measure of irrationality, either in the judgment or in the motive. Externalists, by contrast, deny that such cases must always involve a measure of irrationality.10 What rationality requires in the way either of normative judgments or of motivating states is, in their view, a contingent matter, dependent on further facts about people’s normative beliefs and the contexts in which they act. So, we end up with this: Korsgaard’s internalism requirement: “Practical-reason claims, if they are really to present us with reasons for action, must be capable of motivating rational agents.” (1986, 11)

Once again, however, it now looks like the argument for this requirement is more normative than explanatory. Korsgaard’s thought seems to be that reasons for action 8 As

we shall see, one important question dividing philosophers here is how the motivating states explaining people’s actions are themselves to be explained. 9 Once again, while more could be said about what, exactly, distinguishes Kantian accounts of deliberation from others, this should do for our purposes here. 10 Since realists, as I understand them, deny that normative truths are mere products of rational deliberation, they will count as externalists in this sense.

24

R. H. Myers

must be guaranteed to count in favor of actions and that the only way such a guarantee can be secured is by identifying people’s reasons for action with the conclusions they would reach if they were empirically informed and deliberating rationally about what to do. As we might put her point, externalists threaten to leave people without “reasons” to act on their reasons.11 Because they do not identify people’s reasons with the conclusions of their empirically informed and fully rational deliberations, they cannot guarantee that it is always rational for empirically informed people to act on their reasons; so, they need to provide an alternative account of how reasons for action count in favor of actions. But the trouble with this is that an alternative solution to this “normative problem” is ready to hand. Externalists can guarantee that reasons for action count in favor of actions by identifying people’s reasons for action with the conclusions they would reach if they were empirically informed and thinking veridically about what to do. If a normative notion is all that is needed, truth will surely serve just as well as rationality.12 Thus, we still have not been given a non-question-begging argument with which to confront normative realists. Korsgaard might try arguing (much as Harman did) that explanations of actions couched in terms of substantive reasons will never actually be as good as explanations of actions couched in the Kantian terms that she prefers. But then she too will run into the fact that normative realists understand action differently than she does. Williams’s hope that explanatory considerations will provide common ground over which normative realists and their opponents can fight is still a long way from being realized. Although his account of rational deliberation, like Williams’s, aims at a kind of desire satisfaction and in consequence is more coherentist than rigidly Kantian, Smith is largely in agreement with Korsgaard’s internalism requirement and indeed represents himself (1995, 109) as following in her footsteps, even to the point of sharing her refusal to embrace Williams’s relativism.13 He puts the requirement like this: The internalism requirement tells us that it is desirable for an agent to Ø in certain circumstances C, and so she has a reason to Ø in C, if and only if, if she were [both empirically informed and] fully rational, she would desire that she Øs in C. (1995, 112)

He thinks Korsgaard errs in understanding this requirement according to what he calls (1995, 110) the example model, rather than what he calls the advice model, but that is a somewhat tangential issue that need not detain us here. More important for our purposes is the fact that Smith offers an argument for the internalism requirement of the sort we have been seeking. His claim is that the truth of the internalism requirement best explains why people are so reliably motivated to act on their normative judgments. He agrees with Korsgaard that imperfectly rational people are not always motivated to act this way, but he maintains that even they typically are and that this fact is best explained by acknowledging that the 11 See,

e.g., Korsgaard (1996, Sect. 1.4) and Korsgaard (1997, Sect. 3). McDowell (1995) makes points of this broadly realist sort very forcefully. 13 Smith discusses his differences here with Williams in his 1995, Sect. 2. 12 Again,

Davidson’s Meta-normative Naturalism and the Rationality …

25

internalism requirement is true and that normative judgments are to be analyzed accordingly. Smith’s argument for this claim can be understood as having two steps. The first involves establishing that the reliability with which people are motivated to act on their normative judgments is best explained by supposing rationality requires them always to be so motivated and they are typically rational. Smith calls this the practicality requirement; we might put it as follows: Smith’s practicality requirement: (1) “If an agent judges that it is right for her to Ø in circumstances C, then either she is motivated to Ø in C or she is practically irrational.” (1994, 61) (2) People typically are practically rational.14

The second step involves establishing that the truth of this practicality requirement is best explained by supposing internalism to be correct and externalism incorrect, the strategy being to argue that, if one’s normative judgments are judgments about what one’s informed and coherent self would advise one to do, forming one’s motivating states in line with one’s normative judgments will always enhance coherence and so be something rationality requires one to do, whereas forming one’s motivating states in line with one’s judgments concerning independently obtaining reasons could sometimes fail to contribute anything to coherence at all.15 Once again, however, normative realists are understandably going to complain that this begs the question against them. If what needs explaining is the fact that people typically are motivated to act on their normative judgments, there is no reason why this should require us to establish that rationality always requires this of them. So long as we can establish that rationality typically requires this of them, that will do the trick. As is often noted, the stronger requirement might in any case prove impossible to meet, even given internalism. If I believe I am often mistaken about what my informed and coherent self would advise me to do, do I nonetheless enhance coherence by forming my motivating states in line with my normative judgments?16 Be that as it may, however, realists will certainly insist that Smith’s practicality requirement is too strong. But now consider the rationality requirement: The rationality requirement: Reasons for action must be such that (1) rationality typically requires people to form motivating states in line with their normative judgments, and (2) people typically are rational.

This looks like a better bet. There is no obvious reason to doubt that this requirement could be met by internalists, and no obvious reason either to worry that it begs the question against normative realism. But there certainly is at least some reason to wonder whether it typically is the case that people enhance coherence by forming their motivating states in line with their judgments about independently existing reasons 14 Smith

does not explicitly include (2); but see below, especially note 21. Smith’s own statement of this argument, see his 1994, Chap. 5. 16 For further discussion of this important worry, see, e.g., Dreier (1996). 15 For

26

R. H. Myers

for action. So, we need to consider in more detail whether realists can after all meet this requirement, or whether, failing that, they can find grounds for complaining that it in fact still begs the question against them.

3 Realist Responses to the Rationality Requirement One way in which normative realists might try to meet this rationality requirement is by arguing that people’s desires typically are such that they typically do enhance coherence by forming their motivating states in line with their judgments about independently existing reasons for action. The most straightforward way for them to do this would be simply to argue that acting in accord with independently existing reasons for action is something people typically desire to do.17 But this line of response seems problematic in two closely related respects. First, on the assumption that each "typically" here is to be understood in something more than just a statistical sense, it is far from clear that a desire to act in accord with independently existing reasons for action is one that people typically have.18 Second, even if possession of some such desire were to be typical in the appropriate sense, it is far from clear that this would be enough to establish that people typically enhance coherence by acting in line with their judgments about how to satisfy that desire. One might have supposed this second "typically" could be secured only by supplying a suitable account of the particular kind of coherence that is to be enhanced in rational deliberation and not just a suitable account of the particular sorts of desires that people typically possess. Yet it would hardly be plausible if at this point realists were to maintain that we understand coherence in the context of discussions about rational deliberation as requiring that people’s motivating states typically be formed in line with their judgments about independently existing reasons for action. By contrast, it would by no means be implausible for Smith to maintain that we understand coherence (or, as he prefers to call it, systematic justification) in the context of discussions about rational deliberation as requiring that people’s motivating states typically be formed in line with their judgments about what their informed and coherent selves would advise them to do.19 The case would have to be made, of course, and one might wonder whether Smith has made it, but there is clearly something to the thought that people who too easily formed motivating states at odds with their judgments about what their informed and coherent selves would advise them to do could not possibly count as coherent deliberators themselves. If we are to understand the rationality of deliberation in terms of its success in achieving 17 For an example of a naturalist who takes this line, see Brink (1989, Chap. 3); for a non-naturalist,

see Enoch (2011, Chap. 9). we shall see, I believe people’s desires necessarily aim, as a system, to get such reasons right, but that is a different matter. 19 See Smith (1994, especially Sect. 3.5) for more on his understanding of the notion of systematic justification as it relates to strength of will. 18 As

Davidson’s Meta-normative Naturalism and the Rationality …

27

a kind of coherence among the deliberator’s various desires and motivating states, it seems only appropriate that the kind in question should somehow prioritize coherence between her motivating states and her judgments about what her informed and coherent self would advise her to do. Given that they are on comparatively weak ground here, normative realists do well to take a different line on the rationality requirement. Instead of accepting Smith’s broadly coherentist account of rationality, they should propose an account of rationality more in keeping with their realism. Indeed, instead of identifying people’s reasons with the results of their empirically informed and fully rational deliberations, as the internalists do, why not identify people’s empirically informed and fully rational deliberations as those deliberations that are ideally responsive to their independently existing reasons? In order to meet the rationality requirement, normative realists would then have to establish that people are ideally responsive to their independently existing reasons for action when they typically do form their motivating states in line with their judgments about matters of that sort. This would take some arguing, of course, but as their predicament would now seem to be structurally on a par with Smith’s, there is no reason to suppose they would find it harder to make their case than Smith would find it to make his. As an example of a normative realist who takes this externalist approach, consider how Derek Parfit defines practical rationality in terms of reason-responsiveness: We are rational if we want and do what we have most reason to want and do. (2006, 359)

In fact, at one point, when discussing the move from normative beliefs to motivating states, he even goes so far as to say: …if we were fully practically rational, we would always make that move, and without any further thought. (2006, 334; my emphasis)

This might suggest that he thinks normative realism can actually underwrite Smith’s stronger practicality requirement. Elsewhere, however, he says: When our desires depend on…normative beliefs, these desires are rational only when, and because, these beliefs are rational. (2001, 30; his emphasis)

At the end of the day, therefore, it is only a significantly weaker claim that Parfit is prepared to defend, a claim to the effect that people ideally responsive to reasons typically form motivating states in line with their normative judgments.20 But now the important point to notice here is that even if normative realists can defend this claim, as I think they probably can, this will not be enough, by itself, to meet our rationality requirement, and so will not be enough to fend off Smith’s explanatory challenge. For recall that the rationality requirement has two parts; a claim about rationality, to the effect that it typically requires people to form motivating states in line with their normative judgments, and a claim about people, to the effect 20 Revisionary naturalists could also define practicality rationality in terms of reason-responsiveness.

But it is not easy to think of one who clearly does.

28

R. H. Myers

that they typically act in accord with this requirement. It is only if both of these claims are true that we are in a good position to explain why people in fact act on their normative judgments as reliably as we are supposing they do; but no argument has been given yet for the second claim. And the worry is that, while normative realists may be able to establish the first claim by defining rationality as a kind of ideal responsiveness to independently existing reasons, this maneuver may make it very difficult for them to establish that the second claim is also true. To see the extent of the problem here, consider how our internalists might approach the issue. Williams and Smith would presumably argue that it is in the nature of desires and other motivating states to tend toward the sort of coherence in terms of which they analyze rationality.21 So, they would maintain, if promoting coherence of this sort always or at least typically requires people to form their motivating states in line with their normative judgments, as I am allowing it very well might, then people typically will form their motivating states in roughly that manner. Korsgaard would of course reject Williams and Smith’s assumption that motivation is so dependent on desire. She would insist on arguing instead that people’s aim, in acting, is literally to constitute themselves as agents and that the principles of self-constitution just are the principles of rationality.22 However, the general form of her argument would be very similar to Williams and Smith’s. Her claim would be that it is in the nature of intention and action that people tend to form their intentions to act in more or less the manner rationality calls for. This suggests that normative realists may need to provide an argument along these lines as well. For most of them, this would mean arguing that it is in the nature of action that people’s motivating states tend toward the sort of reason-responsiveness with which they identify rationality. And this could pose a problem, since most of them believe no such thing about action. In fact, I think it would be fair to say that most of them share Williams and Smith’s view that motivation requires desire and that desires tend toward nothing more than coherence. How then could they make the case that people typically are rational in their externalist sense? Once again, they could try arguing that acting in accord with independently existing reasons for action is something people typically have an overwhelming or at least coherence-determining desire to do. But that, as I maintained earlier, would be a very difficult case for them to make. Indeed, it is partly because they despair of making any such case that most normative realists insist on defining rationality in terms of reason-responsiveness and not in terms of coherence. Of course, even if it is true that most normative realists share Williams and Smith’s conviction that motivation requires desire and that desires tend toward nothing more than coherence, there notoriously are some normative realists who reject it and who hold instead that normative beliefs can constitute motivating states.23 And it may be 21 In Smith’s case, this has become clearer with his turn to constitutivism. See Smith (1994, Sect. 6.2),

but also Smith (2013, 2015). especially her 1997 and 1999, but of course this theme runs throughout her 2009 and the papers collected in her 2008. 23 Many take their lead from John McDowell. See his 1979 and 1996. 22 See

Davidson’s Meta-normative Naturalism and the Rationality …

29

thought that this puts them in a better position to explain why people typically are rational, since they can simply insist that it is in the nature of normative beliefs that they tend both to be reason-responsive and to move people to act. However, as Smith and others have long argued, it is not such a simple matter to deny the connection between motivation and desire. Motivating states are dispositions to adjust behavior in light of relevant beliefs, and it is doubtful that normative beliefs could ever constitute such dispositions.24 For the time being, therefore, I am going to proceed on the assumption that this line of argument is not open to realists. But if this line is not open to them, and the Humean line considered above is also closed, what options do realists have left? Some normative realists may conclude at this point that Smith’s explanatory challenge actually is guilty of begging the question against them. Without going so far as to deny that people frequently do form motivating states in line with their normative judgments, they may take exception to Smith’s assumption that this is typical in a sense that implicates some constitutive feature of action. Why not assume instead that this is a purely contingent matter to be explained in ordinary evolutionary and social–psychological terms? This need not be to deny that rationality calls for people to form motivating states in ways that optimize reason-responsiveness. Nor need it be to deny that forming motivating states in line with one’s normative judgments is typically what is optimal, where this is not a purely contingent matter but a consequence of what optimality amounts to in contexts of practical deliberation. The suggestion would simply be that such constitutive claims apply only to action qua rational and not to action per se. Consider again the example of Parfit. Although he never explicitly addresses Smith’s explanatory challenge, he does make some remarkable claims about Williams: Williams did not understand the concept of a reason. (2011, vol. 2, 435) [He] did not understand the distinctive concept of a non-psychological, purely normative reason. (2011, vol. 2, 435; his emphasis)

As evidence for this, he says: Williams thinks of reasons, not as facts that count in favour of having some desire or acting in some way, but as facts that might motivate us. (2011, vol. 2, 435; his emphasis) Williams’s claims about reasons, and about what people ought to do, are really psychological claims about how we might be motivated to act. (2011, vol. 2, 452)

But this is puzzling, for Williams’s claims about reasons were obviously not just claims about how we might be motivated to act; they were claims about how we would be motivated to act if we were empirically informed and deliberated rationally about what to do, and Williams obviously regarded such claims as normative. Parfit cannot possibly have missed this.25 So why is he making these remarkable claims? 24 See, 25 In

e.g., Smith (1987). For a dissenting voice, see Little (1997). fact, he acknowledges it elsewhere in the same discussion (p. 436).

30

R. H. Myers

What Parfit is objecting to here, I submit, is the fact that Williams holds normative questions hostage to psychological ones. After all, even if it is true that Williams usually was very careful to distinguish normative claims from psychological claims, his comments about explanation suggest that he still thought normative facts are beholden to psychological facts in an important sense. When he insisted that it must be possible for a person’s reasons to figure in correct explanations of her actions, he clearly had in mind that what makes this possible must be some systemic aim the person has in acting. He understood this aim in terms of Humean desire satisfaction; Korsgaard and Smith, as we have seen, understand it differently. But the important point to note here is that Parfit thinks such claims are all based on a grievous misunderstanding. He maintains that, as realists, we are committed to the existence of facts about what people have reason to do. And facts are facts; they are whatever they are, whether people are aiming to act on them or not.26 Now, at first sight, this can seem nothing more than common sense. However, I think further reflection reveals it to be deeply problematic. For how could there be facts about what people have reason to do if people were not capable of acting on them?27 But to act on normative facts is not simply to be caused to behave as one does by true normative beliefs that one possesses; getting normative matters right must be one’s systemic aim in acting, and one must typically have some success at doing just that.28 So Williams, Korsgaard and Smith are quite right to insist that reasons for action are in some sense hostage to constitutive features of action. Where all three go wrong, however, is in their assumption that this counts in favor of internalism and against normative realism. After all, if all we know is that people’s aim in acting is to get normative matters right, this leaves it open whether reasons are independently existing facts or facts about what people’s empirically informed and fully rational deliberations would lead them to conclude. But what this does mean, of course, is that it will not after all be possible for normative realists to escape the challenge of explaining how it could be that people’s systemic aim in acting is to act on independently existing reasons for action. And that requires them to show either that desires aim to accord with such facts or that beliefs alone can motivate action. So, let us now return to this issue and consider whether their prospects here really are as bleak as I’ve been suggesting.

4 Anomalous Naturalism Now at this point, many realists will feel obliged to find some way of allowing that normative beliefs are capable of constituting motivating states. Their hope would then be to establish that people’s aim, in forming normative beliefs, is to get their

26 Parfit

(2017, Chap. 12) provides a very powerful statement of this view. Myers (2012, especially the Postscript) for further development of this point. 28 This last point deserves more argument than I can give it here. 27 See

Davidson’s Meta-normative Naturalism and the Rationality …

31

independently existing reasons for action right so they might do what they have independently existing reasons to do. As I noted earlier, however, I find it hard to see how normative beliefs actually could be capable of literally constituting motivating states. But I think it is quite plausible to suppose that they typically influence motivating states in ways that realists can exploit to their advantage. So, let us consider how an argument of this sort might go. A good place to start is with Thomas Scanlon’s idea that motivating states are typically “judgment-sensitive” attitudes.29 Although Scanlon himself evinces some ambivalence on this point, I take his thought to be that motivating states, while strictly speaking counting as desires, modally separable from one’s normative beliefs, are nonetheless typically conditioned by those beliefs in significant ways, with the result that a desire to Ø is typically not a relatively simple disposition to do whatever one believes will increase one’s chances of getting what one wants but instead a more complicated disposition to do whatever one believes will increase one’s chances of satisfying the independently existing reasons one believes one has for Ø-ing in the first place. What needs to be added to this are grounds for holding that motivating states are typically judgment-sensitive not merely in a statistical sense but also in a sense that implicates some constitutive feature of action. It is not enough to establish that people’s normative beliefs regularly influence the contents of their desires; it also matters why they enjoy this influence. What we need, more specifically, are grounds for holding that people’s desires are typically conditioned by their beliefs about independently existing reasons for action because desires always aim, as a system, to get such matters right. Individual desires may on occasion fall very far short of this goal; considered as a system, however, this must always be their aim.30 And this, I suggest, is where anomalous naturalism comes into its own. For if normative properties are to be understood as being causal properties, it is going to be on the grounds that they have a role to play in causal explanations of why people come to have the particular normative beliefs and the particular motivating desires that they do. So, questions will arise about how explanations of this sort might go and about what must be true of people for them to work. And if Davidson’s views about content determination are sound, the answer is going to be that such explanations are possible only because people have systemic interests in believing what is true and desiring what is right. Davidson’s deceptively simple idea here is that the contents of people’s propositional attitudes must in the first instance be fixed by their typical causes, but that in the first instance there would be no saying which causes are the typical ones if people were not triangulating on them. It is only by actively participating in the back and forth that triangulation involves that people can isolate the properties that are the typical causes of their thoughts and utterances, for what makes certain causes typical in the sense that is relevant here is precisely the fact that they emerge as salient in contexts where questions about content are addressed. And since it is very difficult 29 Scanlon 30 I

develops this idea in his 1998, Chap. 1. See also his 2014, especially lecture 3. develop this point further in Myers (2012). Hieronymi (2009) also takes Scanlon to task here.

32

R. H. Myers

to see why people would participate in the back and forth that triangulation involves if they did not share systemic interests in believing what is true and desiring what is right, it would seem to follow that possession of these two systemic interests is a constitutive feature of agency and hence that our rationality requirement could be met without having to deny that normative realism is true.31 Of course, many details would have to be filled in, and many objections would have to be answered, before we would be in a position to decide how compelling this line of argument might finally be. My point here is simply that Davidson’s anomalous naturalism, unlike either non-naturalism or revisionary forms of naturalism, does at least put realists in a position from which they can hope to meet the rationality requirement. Unlike non-naturalists, Davidson believes normative properties are causal properties, so he is committed to regarding agents as having whatever properties are required to be receptive to such causes. Unlike revisionary naturalists, Davidson believes normative properties are strongly prescriptive, so the properties he is committed to regarding agents as having must succeed in making them sensitive to such prescriptions. If the triangulation argument is sound, the requisite sensitivities will be systemic and involve aiming to get one’s normative beliefs and motivating states right. And if this is what people systemically aim at, presumably they typically will succeed in forming motivating states in line with their normative beliefs. Revisionary naturalists will respond to this by doubling down on their revisions. Having already forsaken strong claims about the prescriptive authority of normative reasons, they are not going to go out of their way to defend strong claims about the interest agents take in acting on such reasons. They will be inclined to agree with those anti-realists, such as Williams and Smith, who hold that motivating states aim at nothing more than a kind of internal coherence. Some revisionary naturalists will also follow Williams and Smith in identifying rationality with coherence promotion, strengthening the link between agency and rationality by weakening the link between rationality and reasons. Other revisionary naturalists will break with Williams and Smith by identifying rationality with reason-responsiveness, understood of course in realist terms, strengthening the link between reasons and rationality by weakening the link between rationality and agency. Both approaches make it very difficult to understand how agents could ever act for reasons, since that, as I insisted earlier, requires that they somehow are aiming to bring their behavior into line with their reasons. Now it might be objected that revisionary naturalists could escape this predicament by linking reasons to desire satisfaction. If, say, my better-informed and clearerthinking self would want me to avoid certain sorts of situations, could that not be both a natural property of those situations and a reason for me to avoid them? And is it not plausible to hold that agents have it as their aim to act on such considerations?32 But surely any naturalism along these lines would be unacceptably revisionary in 31 For more on this argument and its place in Davidson’s philosophy, see Myers and Verheggen (2016). 32 See Railton (1986) for a version of revisionary naturalism that proceeds along something like these lines.

Davidson’s Meta-normative Naturalism and the Rationality …

33

other important respects. It might prove capable of meeting the rationality requirement that has been our focus here. But it would do so at the cost of making reasons for action deeply relativistic and relative precisely to the motivating states over which they are expected to have authority. If some maneuver along these lines is the best that revisionary naturalists can do by way of a strategy for meeting the rationality requirement, I think normative realists have no choice but to place their bets either with some non-revisionary form of naturalism or with non-naturalism. But what can non-naturalism offer? As the case of Parfit shows, non-naturalists may also be tempted to take a revisionary line here, insisting that normative properties, being strongly prescriptive properties, necessarily make claims on the motivating states agents should have, but denying that agents, by their very nature, take an interest in conforming their behavior to such claims. However, most non-naturalists try to take a more conciliatory line, allowing that agents share a systemic interest in getting normative matters right. The problem is that their non-naturalism makes it very difficult for them to argue for this claim. Much as knowledge of normative properties becomes sui generis, so too our interest in them becomes a mystery. To the best of my knowledge, Scanlon never says whether he thinks people’s motivating states are typically judgment-sensitive in a merely statistical sense, or whether he sees this as marking a constitutive feature of agency.33 But one might certainly wonder why he could not make the constitutive claim and argue for it through an analysis of what agency necessarily involves. If agents are creatures who act for reasons, and acting for reasons is possible only if one is aiming to get normative matters right, why could Scanlon not simply assert that every agent must possess this aim? Why should we suppose that Davidson’s anomalous naturalism puts us in a better position to meet the rationality requirement than Scanlon’s non-naturalism does? The problem with this, however, is that agents could aim to get normative matters right without aiming to get independently existing reasons right. Their aim might instead be simply to act as their empirically informed and ideally coherent selves would advise, in which case they would presumably not be reliably motivated to act on their judgments about independent reasons. If this is the best Scanlon can do in his efforts to meet the rationality requirement, therefore, he seems destined to fall short. He seems unlikely to be able to explain why his account of reasons and his account of agency are linked to one another in the very close and very secure way that the rationality requirement demands. Of course, in response to this, Scanlon might protest that no explanation of this link is really necessary. Much as he thinks we are free to build our normative epistemology on the method of reflective equilibrium whether or not we can adduce any independent reasons for maintaining that our pre-theoretical judgments about normative matters are on the right track, so too he might say we are free to assume the existence of a very close and very secure link between the nature of normative reasons for action

33 This

is what leaves him open to objections such as the ones Hieronymi offers in her 2009.

34

R. H. Myers

and the systemic aims agents have in acting even if we have no explanation of that link ready to hand. But this is a lot to be asked to take on faith.34 By comparison, there is much to be gained if Davidson’s argument can be shown to work. For people will be motivated to engage in the back and forth that triangulation requires only if the normative matters they are aiming to get right concern independently existing reasons for action, and triangulation will succeed in fixing contents for their normative beliefs and motivating states only if independently existing reasons for action are capable of exercising some causal influence on them. Thus, Davidson’s argument explains why there must be the very close and very secure link that the rationality requirement demands between his account of reasons and his account of agency. Moreover, it does so in a way that simultaneously promises to explain why Scanlon might after all be correct in thinking that a viable normative epistemology can be based on the method of reflective equilibrium. People’s systemic interest in and pre-theoretical grasp of normative matters are not offered as mere articles of faith; they are grounded in a powerful and persuasive account of how the contents of their attitudes are fixed. But of course this argument has no chance of working unless strongly prescriptive properties can also be causal, and that is a possibility that very few philosophers of any stripe are prepared to countenance these days. As I have argued elsewhere, however, it is not always clear what the worries here are. In saying that normative properties play a part in causal explanations of people’s normative judgments and motivating states, Davidson certainly did not mean to be suggesting there are laws guaranteeing that well-placed and properly trained observers always form the normative judgments and motivating states that they should. For Davidson, causal properties operate at levels of explanation where strict laws are not available. They mark regularities that, while counterfactual supporting, still tolerate various exceptions. So in saying that normative properties play a part in causal explanations, Davidson was in a way simply noting how intelligible and predictable it can be that people should form normative judgments and motivating states correctly.35 He did also acknowledge, however, that such causal explanations committed him to a controversial claim—that normative properties can be involved in causal explanations of people’s normative judgments and motivating states only because they vary with and are determined by basic physical properties in the right way.36 And, these days, very few philosophers of any stripe are prepared to allow that normative properties could vary with and be determined by basic physical properties in the right way. As I mentioned at the outset, I take a different view of this matter myself. I grant that normative properties cannot be (analytically or otherwise) reduced to basic physical properties, but I do not see how that prevents them from varying with 34 Scanlon

discusses normative epistemology and the method of reflective equilibrium in his 2014, especially lecture 4. 35 See Davidson (1967, 1970, 1973), Myers and Verheggen (2016, especially Chap. 7), and Myers (2019). 36 See especially Davidson (1970). Davidson (1995) extends the argument from the philosophy of mind to metaethics.

Davidson’s Meta-normative Naturalism and the Rationality …

35

and being determined by such properties in the way causal properties must be. This clearly raises some deep worries in which anomalous naturalism could founder, and clearly I have not done anything to quiet those worries here. But revisionary naturalism and non-naturalism are also plagued by deep worries, and as I said earlier it is not obvious that they are not worse. In any case, my goal here has not been to establish that anomalous naturalism is more plausible overall than revisionary naturalism and non-naturalism. My claim is simply that, while Williams’s explanatory requirement, Korsgaard’s internalism requirement, and Smith’s practicality requirement are all guilty of begging questions against realism, the rationality requirement is one realists need to meet and anomalous naturalists are better able to meet it than revisionary naturalists or non-naturalists. Revisionary naturalists have a hard time even accepting that the rationality requirement must be met; non-naturalists are often committed to meeting it but have a hard time following through. By contrast, anomalous naturalists, because they allow that normative properties are both strongly prescriptive and causal, are in a position to do much better. For they can argue that action is possible only because agents have it as their systemic aim to get independently existing normative matters right.

References Brink, D. (1989). Moral realism and the foundation of ethics. Cambridge and New York: Cambridge University Press. Davidson, D. (1967). Causal relations. Reprinted in his Essays on actions and events. Oxford: Clarendon Press, 1980. Davidson, D. (1970). Mental events. Reprinted in his Essays on actions and events. Oxford: Clarendon Press, 1980. Davidson, D. (1973). The material mind. Reprinted in his Essays on actions and events. Oxford: Clarendon Press, 1980. Davidson, D. (1995). The objectivity of values. Reprinted in his Problems of rationality. Oxford: Clarendon Press, 2004. Dreier, J. (1996). Review of The moral problem. Mind, 105, 363–367. Enoch, D. (2011). Taking morality seriously. Oxford: Oxford University Press. Harman, G. (1977). The nature of morality. New York: Oxford University Press. Hieronymi, P. (2009). The will as reason. Philosophical Perspectives, 23, 201–220. Korsgaard, C. (1986). Skepticism about practical reason. Journal of Philosophy, 83, 5–25. Korsgaard, C. (1996). The sources of normativity. Cambridge: Cambridge University Press. Korsgaard, C. (1997). The normativity of instrumental reason. Reprinted in her 2008. Korsgaard, C. (1999). Self-constitution in the ethics of Plato and Kant. Reprinted in her 2008. Korsgaard, C. (2008). The constitution of agency. Oxford: Oxford University Press. Korsgaard, C. (2009). Self-constitution. Oxford: Oxford University Press. Little, M. (1997). Virtue as knowledge: Objections from the philosophy of mind. Nous, 31, 59–79. Mackie, J. L. (1977). Ethics. Harmondsworth: Penguin Books. McDowell, J. (1979). Virtue and reason. Reprinted in his Mind, value, and reality. Cambridge, MA: Harvard University Press, 1998. McDowell, J. (1995). Might there be external reasons? Reprinted in hisMind, value, and reality. Cambridge, MA: Harvard University Press, 1998.

36

R. H. Myers

McDowell, J. (1996). Two kinds of naturalism. Reprinted in his Mind, value, and reality. Cambridge, MA: Harvard University Press, 1998. Myers, R. (2012). Desires and normative truths: A holist’s response to the sceptics. Mind, 121, 375–406. Myers, R. (2019). Davidson’s meta-normative naturalism. Journal of the History of Analytical Philosophy, 7, 48–58. Myers, R., & Verheggen, C. (2016). Davidson’s triangulation argument: A philosophical inquiry. New York: Routledge. Parfit, D. (2001). Rationality and reasons. In D. Egonsson (Ed.), Exploring practical philosophy: From action to value. Burlington, VT: Ashgate. Parfit, D. (2006). Normativity. In R. Shafer-Landau (Ed.), Oxford studies in metaethics (Vol. 7). Oxford: Oxford University Press. Parfit, D. (2011). On what matters (Vols. 1 and 2). Oxford: Oxford University Press. Parfit, D. (2017). On what matters (Vol. 3). Oxford: Oxford University Press. Railton, P. (1986). Moral realism. Philosophical Review, 95, 163–207. Scanlon, T. M. (1998). What we owe to each other. Cambridge, MA: Harvard University Press. Scanlon, T. M. (2014). Being realistic about reasons. Oxford: Oxford University Press. Smith, M. (1987). The Humean theory of motivation. Mind, 97, 36–61. Smith, M. (1994). The moral problem. Oxford: Blackwell. Smith, M. (1995). Internal reasons. Philosophy and Phenomenological Research, 55, 109–131. Smith, M. (2013). A constitutivist theory of reasons: Its promise and parts. LEAP: Law. Ethics, and Philosophy, 1, 9–30. Smith, M. (2015). The magic of constitutivism. American Philosophical Quarterly, 52, 187–200. Williams, B. (1980). Internal and external reasons. Reprinted in his Moral luck. Cambridge: Cambridge University Press, 1981.

Acting Against Your Better Judgement Rowland Stout

Abstract I defend a Davidsonian approach to weakness of will against some recent arguments by John McDowell and adapt the approach to meet other objections. Instead of treating one’s better judgement as a conditional judgement about what is desirable to do given available reasons, it is proposed to treat it as an unconditional judgement about what is desirable to do from a rational perspective that one takes to be the right perspective to have. This makes sense of Aristotle’s claim that desire is for the good or the apparent good: judgements of desirability generally concern the apparent good, whereas judgements of desirability from rational perspectives that are judged to be the ones to have are judgements of the actual good. Weakness of will occurs when one’s actual rational perspective is not the one that one takes to be the one to have—i.e. when one’s judgement of the apparent good does not coincide with one’s judgement of the actual good. One makes two judgements—one from an adopted perspective that one judges to be the one to have and one from one’s actual perspective. Keywords Weakness of will · Practical reasoning · Motivating judgement · Intention · Perspective · Anscombe · McDowell · Davidson

1 Davidson (1980, 101). I will preserve Davidson’s talk of ‘judging’ and ‘judgement’ in this discussion, even though such talk gives the misleading impression that intending is taken to be an action of some sort. I mean nothing more by these terms than would be expressed by talking about thinking. Thinking that φ-ing is desirable does not require any conscious or unconscious mental action and may be understood in completely dispositional terms.

This chapter is in its final form and it is not submitted for publication anywhere else. R. Stout (B) School of Philosophy, University College Dublin, Newman Building Belfield, Dublin, Ireland e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_3

37

38

R. Stout

1 Introduction Because he identifies intention with judgement, Davidson appears to face the problem of weakness of will. For Davidson, intending to φ is judging (all-out) that φ-ing is desirable.1 This judgement is the consequence of practical reasoning—a process that leads one to form judgements about what to do in response to reasons. Suppose that weak-willed action is intentional action that goes against one’s better judgement. In virtue of being intentional, such action is caused by reasons through a process of practical reasoning. But the weak-willed agent’s better judgement too is the result of practical reasoning and recommends not doing that action. So, we appear to have practical reasoning simultaneously resulting in opposing recommendations—that the action is and that it is not desirable. The one that determines the agent’s actual action we might call a ‘motivating judgement’. The other is the agent’s ‘better judgement’. The fact that practical reasoning may result in opposing judgements would seem to undermine its supposed status as a rational faculty of any sort let alone one that determines, all by itself, how you act. Davidson’s approach shares with Kant’s what is sometimes called ‘motivational internalism’. Kant describes the Will as the power to act according to the idea of rules (Groundwork 36, Critique of Practical Reason 32). Acting intentionally is exercising this power; we might say that it is exercising sensitivity to rationality, where we should think of rationality rather abstractly as a system of reasons. Davidson (1980, 7–8) describes actions done with intentions as actions caused by reasons. Now, although he does not talk about powers and sensitivity, he is certainly committed to the idea that motivation is internal to practical judgement; the judgements that issue from practical reasoning can motivate you to act all by themselves. There is no need for any endorsement of such judgements, through a further act of will, to get you to form an intention and then to act.2 This is by contrast with the approach to agency that goes back at least to Hobbes. The definition of the will, given commonly by the Schools, that it is a rational appetite, is not good. For if it were, then could there be no voluntary act against reason. For a voluntary act is that which proceedeth from the will, and no other. But if instead of a rational appetite, we shall say an appetite resulting from a precedent deliberation, then the definition is the same that I have given here. Will, therefore, is the last appetite in deliberating. (Leviathan, chapter VI, paragraph 55)

In the Hobbesian model, we can separate the willing from the practical reasoning. As a result, it provides a very straightforward—indeed literal—rendering of weakness of will. A weak-willed agent may work out what is the best thing to do but then fail to act on it due to a failure in this last step—the step where willing would turn the practical judgement into action. In this approach, better judgements may or may not motivate one, but if they do not that is not because there is an opposing motivating judgement. What opposes one’s better judgement is just a recalcitrant will. It is difficult not to think of the final stage in the process described by Hobbes as arbitrary—an arbitrary will—given that it is not sensitive to reason. The space 2 See

for example Davidson (1980, 87).

Acting Against Your Better Judgement

39

that this model opens up for weakness of will can start to look like an unbridgeable chasm. Also, Hobbes’s idea that there is a last appetite in deliberating misconstrues the nature of deliberation and action. Practical reasoning does not end at the onset of acting; while one is in the process of acting one continues to reason practically about what one is doing—about how to be doing it.3 Since practical reasoning does not end in advance of acting, there is no location in the causal process where we can fit in a non-rational willing. Davidson’s proposed solution to the problem of weakness of will is to make a sharp distinction between motivating practical judgements and better judgements. The motivating judgement that φ-ing is desirable (the judgement that Davidson identifies with the intention to φ) is an unconditional judgement. It is based on reasons, but those reasons do not figure in the content of the judgement. By contrast, the better judgement that φ-ing is desirable is a conditional judgement; it is the judgement that φ-ing is desirable all things considered. And Davidson spells out what ‘all things considered’ amounts to by describing this judgement as that φ-ing is desirable on the grounds that r, where r includes all considerations that seem relevant to the agent.4 When suffering from weakness of will, an agent simultaneously judges on the basis of all the relevant evidence that φ-ing is desirable (motivating judgement) and judges that on the basis of all the relevant evidence φ-ing is not desirable (better judgement). The failure to adapt one’s judgement (based on all the available reasons) of what is desirable to a judgement about what is desirable given all one’s available reasons exemplifies a kind of irrationality; but it is a psychological possibility. While generally acknowledged to be ingenious, this solution has left many people puzzled about the nature of the proposed practical reasoning that leads to non-motivating better judgements. In this chapter, I will consider several issues concerning Davidson’s account that show that it cannot be quite right as it stands. As a result of thinking about how to adapt and develop the account in the light of these issues, I will propose a different model of what a better judgement is. It is judging from a rational perspective that one takes to be the one to have that φ-ing is desirable. In weakness of will, one simultaneously judges from the perspective one takes to be the one to have that φ-ing is desirable and judges from one’s own perspective that φ-ing is not desirable. According to this idea, one’s better judgement involves a second-order judgement about how one should judge but does not involve a conditional judgement in the way Davidson suggests. In order to make sense of this idea, I need to spell out what it is to think something from a perspective that is not one’s own.

3 This

is a point highlighted by Thompson (2008, part 2), and which he attributes to Anscombe (1957), arguing at the same time that it is obscured by Davidson’s insistence that actions are events that are describable with the perfective rather than continuous aspect. 4 Davidson (1980, 38–41).

40

R. Stout

2 Intention But before considering how to improve on Davidson’s conception of a better judgement and how we reason towards it, I will consider McDowell’s (2010) suggestion that the problem here is with Davidson’s conception of motivating judgement instead. As I said, Davidson (1980, 101) identifies an intention to φ with a judgement emerging from practical reasoning that φ-ing is desirable. This identification is challenged by McDowell (2010), who, while agreeing with Davidson that practical reasoning takes us all the way to intention and action, denies that practical reasoning yields a normative judgement about what is desirable to do. He argues instead that the conclusion of practical reasoning is a factual judgement—that φ-ing is what we are actually doing. I suggest that McDowell expresses an important insight about practical reasoning here—one that will help us make something of Davidson’s conception of weakness of will—but that he draws too strong a conclusion from this insight. The insight is one that McDowell attributes to Anscombe (1957), namely that practical reasoning may function by taking into account a single desirable characteristic of some piece of behaviour and need not take into account all the possible desirable characteristics of all the alternative actions. Suppose someone suggests that we go to the fair—that it will be fun. I might think that that is a good idea and so decide to go. Here, I base my action of going to the fair on the fact that it will be fun without having to establish first that this is the best way to have some fun or that there are no other actions with desirable characteristics that outweigh this one. McDowell makes the point, using Anscombe’s example of eating some pig’s tripes because it is good for the health. Someone who takes some pig’s tripes to get some vitamin X into him may or may not have considered other things that bear on the question whether to do that, perhaps that the pig’s tripes will probably taste disgusting. Surely, silence about this does not make Anscombe’s syllogism somehow incomplete, as Davidson implies. (2010, 425)

Practical reasoning does not work by optimizing some value function across all alternatives, nor should it. We would never get anything done. As McDowell puts it, it is wrong to assume that practical reasoning yields ‘overall verdicts’ of desirability. The point is even clearer in Buridan’s ass cases. Suppose I can pick just one can of baked beans from a supermarket shelf. There may be nothing to recommend one can of the same brand of beans over any other can. But I do not hang about looking for a reason to prefer one to another. I note that this one is a can of beans, and given that I want to pick a can of beans, I go ahead and pick it. I had a reason for picking that can of beans even though I did not judge it to be better than any other can. The reason I had for picking this can could have been transformed into a reason for picking another can. But we have to be a bit careful in working out what this shows. I think it would be a mistake to conclude that the same system of reasons that my behaviour is sensitive to recommends both that I pick this can and that I pick that can. This is what Anthony Kenny (1975) seems to conclude based on what he describes as the logic of satisfactoriness. He starts off making the observation about practical reasoning that I have been attributing to Anscombe and McDowell—namely that it

Acting Against Your Better Judgement

41

can recommend sufficient means to an end without having to recommend the one and only or best means to an end. But he goes on to say: ‘If someone has expressed his current wants adequately in a goal-fiat, then there is something inconsistent in his refusing to welcome a fiat which is derivable from it in the logic of satisfactoriness’. (1975, 91). The trouble with thinking that there is a logical system underlying practical reasoning and that it recommends picking this can and it recommends picking that can is that intentional action cannot be considered to be an exercise of sensitivity to such a logical system. Sensitivity to such a system would require us to pick this can and to pick that can, which is impossible in the case under consideration. We are trying to articulate a notion of practical reasoning where practical reasoning takes us all the way through to intention and action. Such practical reasoning could not recommend both of two equally good but jointly incompatible courses of action. But now, we seem to have a dilemma. We are rejecting a conception of practical reasoning as determining the most desirable course of action, and we are rejecting a conception of practical reasoning as recommending all equally desirable courses of action. What is left? I think there is just one way out. It is to say that the system of reasons that my behaviour is sensitive to recommends picking this can and it does not recommend picking that can, while another system of reasons that my behaviour might have been sensitive to recommends picking that can and not this can. There might be multiple equally good ways to reason practically corresponding to equally good systems of reasons that such reasoning is sensitive to.5 These different ways of reasoning practically are associated with different rational perspectives. From one rational perspective the thing to do is to pick this can of beans; from another it is to pick that can of beans. So there are many conceptions of practical rationality, and the one that an agent is sensitive to may be arbitrary in some way. Picking this can of beans is the thing to do according to the conception of practical rationality that I am sensitive to, but there is nothing recommending that conception of practical rationality over another one that would have recommended something quite different. That the thing to do is to pick this can of beans is relative to a particular conception of practical rationality. It does not follow that we should be relativists about the recommendations of practical rationality across the board. As I will show later, we may avoid such relativism by noting that conceptions of practical rationality may lead to self-referring recommendations. I may be sensitive to a conception of practical rationality that recommends being sensitive to a different conception of practical rationality. This opens up the possibility of doing something intentionally, which is recommended by a conception of practical rationality that I am sensitive to, while acknowledging that ultimately this is not the thing to do as it is not recommended by the conception of rationality that is the one to be sensitive to. McDowell takes Davidson to be claiming that practical reasoning leads to an overall judgement of desirability. But this is unfair. As McDowell notes, Davidson 5I

think this fits what Price (2008, Chap. 1) is claiming in a useful discussion of Kenny’s logic of satisfactoriness when he says that it is not best thought of as a logic as such.

42

R. Stout

distinguishes all-out judgements of desirability from conditional all things considered judgements of desirability. The all-out judgement is based on all the reasons one has but is not about those reasons. On the other hand the conditional judgement is that a certain action is desirable given all these reasons, and so is partly about one’s reasons. It is the all-out judgement that Davidson identifies with an intention. But McDowell, in describing Davidson’s position, switches between talking about allout judgements of desirability and talking about judgements of all-out desirability.6 In the latter the content of the judgement rather than the judgement itself is qualified as being all-out. Such a judgement could fairly be described as a verdict of overall desirability. But the former could not. An all-out judgement of desirability is just the judgement that some course of action is desirable, not that it is most desirable. It is important to be clear that when Davidson talks about judging an action to be desirable he does not mean judging that it has some desirable characteristic (even if the judgement is just based on noting some desirable characteristic). There is a sense of ‘desirable’ in which we could judge that picking each can of beans is desirable. But if this was what was meant then judging that picking this can of beans is desirable could not be identified with the intention to pick it. If some course of action is desirable in the sense that Davidson must be after then it is the thing to do. Two incompatible courses of action cannot both be judged as desirable in this sense. But it does not follow that ‘desirable’ means ‘most desirable’. I might judge that picking this can of beans is the thing to do without committing myself to any claim about its superiority to picking that other can of beans. Of course I am, at least implicitly, judging that picking that other can of beans is not the thing to do, but this rejection of the desirability of alternative courses of action need not itself be based on reasons. I may have a reason to pick one can without having a reason not to pick another can. So, I think a charitable reading of Davidson here would line him up with Anscombe on the nature of practical reasoning. But this is compatible with his not accepting another aspect of practical reasoning that McDowell attributes to Anscombe— namely that the conclusion of practical reasoning is a factual not a normative judgement. McDowell argues that the recognition that practical reasoning does not yield an overall verdict of desirability supports the idea that practical reasoning does not yield any sort of judgement of desirability, but rather yields a factual judgement about what one is doing. According to McDowell, the conclusion of practical reasoning—a judgement that expresses my intention—is not that picking this can of beans is the thing to do or that it is desirable. It is that I am picking this can of beans. It is a judgement that may be distinguished from the sort of judgement that a detective following me around the supermarket might make about me not by its content but by the sort of knowledge I have of it. It is not by observation that I (unlike the detective) know I am picking this can of beans. I hope to have shown that the consideration about the structure of practical reasoning, that it may take into account a single desirable characteristic and reach a conclusion without making comparisons with alternative courses of action with 6 Contrast

McDowell (2010, 419) with McDowell (2010, 423).

Acting Against Your Better Judgement

43

their own desirable characteristics, does not require us to say that the conclusions of practical reasoning are factual as opposed to normative judgements. Davidson’s proposal that these conclusions are all-out judgements of desirability stands up to this argument from McDowell.

3 Non-motivating Judgements of Desirability Davidson makes a general point against any account (and this would have included McDowell’s) that attempts to explain weakness of will just by considering the practical reasoning that leads to motivations. Such accounts deny that practical reasoning leading to motivation is in direct conflict with practical reasoning towards better judgements by denying that practical reasoning that leads to motivation results in judgements of desirability. For McDowell it results instead in judgements about what one is actually doing and for others it might not results in judgements of any sort. But Davidson points out that conflicts in deliberation are not restricted to conflicts between motivating judgements and better judgements. We also have conflicts between various non-motivating judgements in the process of practical deliberation. Davidson calls these moral conflicts (1980, 33). We need an account of why such conflicts do not leave the agent making straightforwardly contradictory judgements, and this account can then be carried over to why conflicts between motivating and non-motivating judgements do not involve straightforward contradictions. If weakness of will is an instance of a wider problem that includes moral conflict then our solution to it should carry over to the wider problem. So perhaps we should look to an account of non-motivating practical rationality rather than to an account of motivating practical rationality to address the problem of weakness of will. Consider a simple bit of deliberation about whether or not to switch off the TV and go to bed. Suppose that I see the force of each of two different sets of considerations that lead to opposite recommendations about how to act. From one point of view it makes sense to switch off the TV and go to bed as I have to get up early in the morning and have a busy day ahead of me. From another point of view I should keep watching the TV in order to see what happens in the end and the show has only got another half hour to go. Both arguments weigh with me, but only one of them leads to my intention and action. What does it mean for them to weigh with me or for me to see the force of these arguments if they do not actually motivate me? Davidson’s answer is that I can see the force of an argument for the desirability of some action when I endorse a prima facie judgement that that action is desirable.7 Prima facie judgements have a conditional structure. So I judge that switching off the TV is desirable given that I need my sleep and simultaneously judge that continuing to watch the TV is desirable given that doing so will enable me to see what happens in the end. Such judgements do not contradict one another. 7 Davidson

(1980, 35 ff.).

44

R. Stout

Davidson then develops a model for how we arbitrate between these different judgements. According to Davidson, the agent puts both sets of considerations into a bigger syllogism in which all the considerations may be weighed up. When all the considerations that are available to the agent are included in the prima facie judgement the agent has made a judgement that switching off the TV is desirable, say, given all the available reasons. And this counts as their better judgement. Weakness of will occurs when the agent fails to judge unconditionally that switching off the TV is desirable despite judging that it is desirable given all the available reasons. But this account of non-motivating better judgements is unsatisfactory as it stands for several reasons. The first issue is that Davidson explains weakness of will in terms of a discrepancy between motivating judgement and better judgement, and not necessarily as a failure in the motivating judgement. This means that a failure in one’s better judgement should count for Davidson as a case of weakness of will even when the motivating judgement is perfectly responsive to reasons. But it is no failure not to act according to a mistaken prima facie all things considered judgement. For example, I might make a mistake about what are the available reasons or make a mistake as to what follows from them. Clearly this sort of mistake should not count as weakness of will even though it falls under Davidson’s definition. One way to respond would be to require that in cases of weakness of will one’s better judgements are not mistaken. But ruling out weakness of will for all cases where one’s better judgement is incorrect seems unmotivated and also not in step with Davidson’s conception. Suppose I weigh up the available reasons for and against switching off the TV and going to bed and quite wrongly conclude that they favour my switching off the TV—perhaps it is very important for me to finish watching that particular programme though I do not take this into account properly. I form the incorrect prima facie judgement that switching off the TV is desirable all things considered. But also suppose that through laziness I fail to act according to this incorrect judgement. This should still count as weakness of will. A better response—one which I will develop later—is to add another normative layer to Davidson’s conception of a better judgement. Instead of taking an agent’s better judgement to be that φ-ing is desirable given all the available reasons we should perhaps take it to be that φ-ing should be judged to be desirable. On this conception of a better judgement it is not the conditionality of the desirability that is distinguishing it from a motivating judgement of all-out desirability. It is the fact that it is a second-order judgement. Failing to act on the second-order judgement is failing to form the corresponding first-order judgement. And failing to form the corresponding first-order judgement (and then acting) would still count as a failure of will even when that second-order judgement is incorrect. The second issue concerns the relationship between the practical reasoning that leads to motivating judgements and the practical reasoning that leads to better judgements. We do not reason towards a better judgement and then simply infer a motivating judgement. This would make weakness of will weirdly irrational.8 Why would

8 This

is Tenenbaum’s (1999) criticism of Davidson’s account.

Acting Against Your Better Judgement

45

one not be able to make the simple inference required? Although Davidson’s introduction of the ‘Principle of Continence’ might suggest that he took it to be a principle of rationality that one infers one’s motivating judgement from one’s better judgements, this cannot really be what he had in mind. So what is the relationship between the two kinds of reasoning? McDowell (2010) takes them to be quite separate kinds of reasoning. He takes better judgement reasoning not to be practical reasoning at all, but just a species of theoretical reasoning. But this would make it mysterious why we should care about what results from such reasoning; it would be merely theoretical and thus irrelevant to practical decision-making. Both kinds of reasoning involve sensitivity to practical reasons, which means that they must really both be practical reasoning.9 It might seem from the way Davidson describes the cases that he too takes practical reasoning to work in different ways in the two cases, as if in better judgement reasoning one is inferring from reasons whereas in motivating judgements the reasons are causing one’s judgement by weighing with one psychologically. The idea would be that psychological effects might affect this weighing process but would be absent from the more logical inferring process. But practical reasoning can only be made sense of as the exercising of sensitivity to the inferences that should be made from reasons. There is no other way that reasons may weigh with one. Davidson is certainly not contrasting a rational process with a merely causal process. So we must assume that the reasoning leading to better judgements is the same sort of reasoning that leads to motivating judgements, but that somehow they may lead to different judgements in the same situation. Perhaps we might think of the distinction as being that between online and offline thinking. Because reasoning towards a motivating judgement commits one to action, various psychological factors, e.g. anxiety or laziness—may interfere which would not have any force when the reasoning is offline and not leading to action directly. The challenge for this view, however, is to make sense of the offline better judgement reasoning. Since, in the normal course of events, this reasoning would not merely lead to a better judgement, but would lead to a motivating judgement too, it is not clear why it should count as offline and be free from the psychological factors that distort motivating judgements. We might take this offline reasoning that leads to non-motivating practical judgements to be reflective or second-order reasoning. One derives a motivating judgement of desirability and then reflects on what the reasons are that one was employing. But by itself this is not going to explain why the reflective judgement should ever diverge from the motivating one. Also it does not explain how we might have conflicting motivating judgements. As we saw, such moral conflict is a standard feature of moral deliberation. Tenenbaum makes a suggestion that might help here. He claims that ‘different desires might present an object as good from different perspectives’. (Tenenbaum 1999, 899) ‘A desire provides the agent with a direct and primary cognition’. (1999, 901) One of these will be a motivating judgement. But these cognitions may then be reflected upon and evaluated in a better judgement. ‘One can think of the akratic 9 See

Stout (2019) where I respond in more detail to this suggestion of McDowell.

46

R. Stout

agent as someone who has a desire whose presentation of an object wins over his reflective understanding of the good’. (1999, 906). This is not quite right yet. An agent’s better judgement does not have to coincide with any desire they have. But we can skip this talk of desire altogether. An agent can adopt multiple perspectives and make judgements of desirability from those perspectives.10 The one that will motivate them will be a judgement of desirability from their own perspective. They can reflect on these judgements and make secondorder judgements about them, including the judgement that the judgement they should form is that φ-ing is desirable. This is the idea that emerged from consideration of the previous issue too. There is one further criticism of Davidson’s account, which also applies to Tenenbaum’s. It is the most commonly heard criticism, but also fairly easily dealt with. Sometimes when we act against our better judgement, our better judgement is itself an all-out judgement of desirability, with nothing conditional or second-order about it.11 I just think I should switch off the TV and go to bed, but do not act on this thought. It is incredibly implausible to suppose that all non-motivating judgements of desirability are second-order or conditional. For one thing it makes a nonsense of the idea of practical reasoning to reduce it to reasoning about reasons. As Sebastian Rödl observes: “As a reason for doing something is something from which one may reason practically to a conclusion, this describes practical reasoning as reasoning from a reason for doing something to the conclusion that there is a reason to do it.” (2007, 18). The solution from Davidson’s point of view should be to acknowledge that better judgements of desirability have two components. The agent judges unconditionally from some perspective that φ-ing is desirable and at the same time judges that that is the perspective to have and therefore the judgement they should make. In cases of weakness of will the agent simultaneously judges that φ-ing is not desirable. They make this judgement from the perspective they themselves actually have rather than from the perspective they think they should have. As it stands Davidson’s account of weakness of will, in common with most of its competitors, does not attribute a genuine contradiction in the thinking of the weakwilled agent (or the morally conflicted agent). On the one hand, this is a good thing, as the weak-willed agent is merely irrational, not incoherent. But without such a contradiction it is not clear where the real conflict between better judgement and motivating judgement lies. My alternative suggestion is that the contents of the two judgements are genuinely contradictory but that because one is from the agent’s own perspective and the other is from an adopted perspective that the agent thinks is the one to have, this contradiction does not involve psychological incoherence. At the heart of the matter is the dynamic nature of practical rationality. Practical rationality is not merely a way of making recommendations about what to do; once 10 Remember a rational perspective in this discussion should be thought of as associated with a way of reasoning practically. 11 See for example Bratman (1979) who insists that better judgements need not be conditional in form.

Acting Against Your Better Judgement

47

it reaches a certain level of sophistication, it also makes recommendations about how to think. What this means is that a particular conception of practical rationality may make recommendations about what conception of practical rationality to be sensitive to. So in being sensitive to such a conception of rationality one changes the conception of rationality one is sensitive to and so moves from thinking that the thing to do is the one that is recommended by the original conception of practical rationality to thinking that the thing to do is the one that is recommended by the recommended conception of practical rationality. This makes sense of Aristotle’s repeated claim that desire is for the good or the apparent good.12 If we assume that sensitivity to practical rationality is a dynamic process, involving different requirements as it progresses through different stages in which the conception of practical rationality that the agent is sensitive to itself evolves, then exercising this sensitivity must involve both drawing interim conclusions and developing ultimate, or at least more considered, conclusions. And these conclusions, both deriving from practical rationality, may conflict. We might think of the apparent good as what a conception of practical rationality recommends at a particular stage, and the good as what it recommends ultimately, after it has finished adapting to its own recommendations for itself. An interim judgement that something is the thing to do is a judgement about the apparent good, and while making such a judgement one might simultaneously doubt that this is actually or ultimately the thing to do. A claim about the apparent good is always made within the scope of a particular rational perspective, whereas a claim about what is actually good, even if it is rooted in some conception of practical rationality, transcends that conception. Understanding the good and apparent good in this way, it is clear that although we are always sensitive to the apparent good we are not always sensitive to the actual good. The tripe eater or the shopper who chooses a can of beans is sensitive to what is apparently good inasmuch as they are responding to what is recommended directly by a certain conception of practical rationality. While such sensitivity involves the capacity to alter their thought about what to do in the light of countervailing reasons, it does not require any commitment to this being the thing to do ultimately. There is no attempt by either agent to work out whether their reasoning is better than alternative ways to reason. And indeed we cannot say from outside of the scope of their rational perspective that eating tripe or choosing that can of beans is actually or ultimately the thing to do. (Nor can we say that it is not actually the thing to do.) Being sensitive to the good (what is ultimately the thing to do) is only possible when the conception of practical rationality that one is sensitive to is advanced enough to make recommendations about itself. Before that stage—e.g. for nonhuman animals and infants—rational sensitivity can only be for the apparent good. Weak-willed agents are one step up from this, as they have the capacity to be sensitive to the actual good, not just the apparent good. But when they are weak-willed they fail to be sensitive to the good they know about and are only responding to the apparent good. 12 See

Nicomachean Ethics 1113a23-31 and De Anima 433a26-29. I make no claim that this is the right way to understand Aristotle.

48

R. Stout

In cases of weakness of will, one fails to adapt the way one is thinking about what to do to the way one is thinking is the way to think about what to do. One may adopt the rational perspective that one thinks is the one to have and judge from that adopted perspective that switching off the TV is the thing to do. But one’s own perspective still works with the conception of practical rationality that one simultaneously thinks is not the one to have, and from that perspective one thinks that watching some more TV is the thing to do.

4 Adopting a Perspective You Do not Actually Have So my suggestion to improve Davidson’s conception of non-motivating judgements of desirability is to say that they are all-out (unconditional) judgements from perspectives that are not the agent’s own perspective. They have exactly the same form as motivating judgements of desirability. What distinguishes these latter judgements is that they are from the agent’s own perspective. The agent’s better judgement about what to do has two components: the all-out unconditional judgement (from an adopted perspective) about what to do and the judgement that this is the way to think about what to do. So, it consists in thinking from some perspective that φ-ing is desirable and also thinking that that perspective is itself desirable—that that is the right way to think. In exhibiting weakness of will in continuing to watch the TV, one simultaneously judges from one’s own perspective that watching more TV is desirable and judges from a perspective that one judges to be the perspective to have that watching more TV is not desirable. This suggestion preserves Davidson’s conception of a motivating judgement, but changes his conception of a non-motivating better judgement. Instead of being a conditional prima facie judgement that all things considered switching off the TV is desirable, it is a second-order but unconditional judgement that, according to the desirable way to think about this, switching off the TV is desirable. Adopting what one takes to be the desirable way to think about this, one judges from that perspective unconditionally that switching off the TV is desirable. This suggestion raises the urgent question of what it is to think something from an adopted perspective that is not one’s own. There is no question that we do adopt perspectives that are not our own; it is one of our great skills. But the question is what this amounts to and also whether it makes sense to talk about making judgements from such adopted perspectives. Furthermore, we need to understand what it is for a perspective to count as one’s own. Adopting perspectives is involved in play-acting, in imagination, in sympathetic conversations, in considering arguments, and in giving advice, just to pick a few instances. When we do adopt a perspective that is not our own this is with relation to a particular practice or mode of behaviour that has its own rules or norms, for example, a game or an argument or a conversation. In normal life, we are constantly going into and out of such practices. In adopting a perspective we do not simply change the way we behave to a way in which we do what the conception of practical rationality associated with that perspective recommends. We do not lose our way of

Acting Against Your Better Judgement

49

behaving, but instead we find a space in it for this sensitivity to other conceptions of practical rationality. So, adopting a different perspective is not the same thing as making that perspective one’s own temporarily. But nor is it the same as merely pretending to have that perspective. That would make sympathy into insincerity. Consider the example of play-acting. Play-acting behaviour is sensitive to a conception of practical rationality associated with the role one is playing. If I am playing ‘mud pies’ in a sandpit I may conclude that the thing to do is to cook this pie and then eat it. From the adopted perspective of that role I think that this is the thing to do. My behaviour generally is sensitive to a conception of practical rationality that recommends that in this situation the thing to do is to play ‘mud pies’. And the rule of the game is to play-act doing what is the thing to do from the rational perspective of cooking and eating pies. So the thing to do is to play-act cooking and eating a pie. But of course, I do not really think that eating this mud pie is the thing to do. From my own perspective I think eating a mud pie is the wrong thing to do. Given that I do not think from my own perspective that eating this mud pie is the thing to do, I do not intend to eat it. Thinking that eating this pie is the thing to do from the perspective of the role, and accepting that I am in a context where that role is called for—i.e. the game-playing context—means that I intend to play-act eating the pie. For the sake of the game I am adopting a conception of practical rationality that is not my own; if it were my own then my actual behaviour, not just my play-acting, would be sensitive to it. More generally we might try out the following theory of what it is to adopt a rational perspective. It involves exercising two skills simultaneously: the ability to determine recommendations issuing from a certain conception of practical rationality and the ability to transform these recommendations into recommendations for our actual behaviour. Subject, S, adopts a perspective, P, for the sake of a practice, M, if and only if

1. there is a function implicit in the practice, M, that transforms one set of recommendations about what is to be done (the perspectival recommendations) into another set of recommendations about what is to be done (recommendations for actual behaviour); 2. S’s behaviour is sensitive to a conception of practical rationality that includes the recommendations of P transformed by M. M is what links the perspectival recommendations to actual behaviour. It includes entry and exit rules for the practice as well as rules for transforming perspectival recommendations into actual behaviour. Knowing M is knowing how to engage in that sort of practice. You know when to start and stop playing, and, when playing, you know to smack your lips with the mud pie presented to your mouth and say ‘yum yum’, rather than actually to eat anything, when the recommendation of your adopted perspective is that you eat the pie. In order to be able to play the game properly you must know this and also know what the recommendations of P are. For example,

50

R. Stout

you should only eat a pie that has been cooked, and that you should work on the assumption that this (pointing to a pie-shaped bit of mud) is a pie. We can now make sense of the idea that when playing the game you think (or judge) that this is a pie. On the one hand, it is clear that you are not working on the assumption that this is a pie in your actual behaviour—otherwise you might actually eat it. So, you do not actually think this is a pie. But within the perspective of the game and before these perspectival recommendations are transformed by M into actual behaviour, you are working on the assumption that this is a pie. Within the game you accept the following reasoning: because this is a pie and it has not yet been cooked, it should be placed in this oven before being eaten. You think from the perspective of the game that it should be eaten, and then, you think from your own perspective that the thing to do is to raise it to your lips and go ‘Yum yum’. Something similar goes on in imagination, where there is no visible behaviour manifested. I may take on some alternative conception of practical rationality with respect to my imagined thoughts. Concluding from this adopted perspective that the thing to do is to destroy my enemy I may imagine accordingly. Again, it is not that I think from my own perspective that the thing to do is to destroy my enemy; otherwise I would be acting accordingly. I think from the imagined perspective that the thing to do is to destroy my enemy, but I do not actually think—i.e. from my own perspective—that this is what to do. Fantasy is a circumscribed realm of behaviour isolated from our normal active life, in which we can exercise sensitivity to adopted structures of practical rationality. This is an important skill as it enables us to try out different schemes of practical rationality as part of the never-ending task of selfconsciousness. By experimenting in this way, we come to see how we might improve the way we think about what to do. Working for an employer is another relatively circumscribed area of behaviour. Your work place may be a context where you can take on sensitivity to a conception of practical rationality that is not what your behaviour generally is sensitive to. When your boss tells you that your role is to identify your goals with the goals of the company you may well adopt the company’s conception of practical rationality for the sake of your work behaviour. You think that the way to behave at work is to do what is the thing to do when you are at work from the adopted perspective of a good company person. Your actual behaviour is by and large the same as that recommended by the company’s conception of practical rationality. But when your work behaviour stops being isolatable from your goals in life, you then have to choose to continue in alienation or to step out of the circumscribed role and stop adopting that perspective, perhaps by taking sick leave, becoming a whistle-blower or by resigning.13 Another example would be that of the love cheat, who may try to divide their love life into separate zones. Perhaps she thinks that the way to behave with respect to Giles is to do whatever is the thing to do from the perspective of a person who loves 13 Contrast a company ‘Yes-Man’ who loses themselves in the company way of behaving and Melville’s Bartelby who refuses to adopt a perspective other than his own at work, replying ‘I would prefer not to’, when given instructions by his employer. Bartleby’s almost noble refusal to ‘play the game’ extends into a self-destructive rigidity of mind leading to his eventually refusing food that someone else thinks he should eat and starving to death in prison.

Acting Against Your Better Judgement

51

only Giles and the way to behave with respect to Jim is to do what is the thing to do from the perspective of a person who loves only Jim. She thinks, adopting the first of these perspectives, that she should have a relationship only with Giles, and she thinks, adopting the second perspective, that she should have a relationship only with Jim. But neither of these is how she thinks from her own perspective. They are just ways of thinking that she adopts when she is with either man. The difficulty arises for her if the way to behave with respect to Giles cannot be kept separated from the way to behave with respect to Jim. A deeper problem perhaps, even if this difficulty does not arise, is that she is never really herself with either man. Thinking something from an adopted perspective does not mean you really believe it. You are not disposed to act on the assumption unless the thought is from your own perspective. Your own perspective is the one that your actual behaviour is sensitive to. But if the adopted perspective is one that connects you with someone in a conversation, then thinking something from that perspective does mean you may assert it and defend it. Indeed this defence may itself be from your own perspective if you think that the adopted perspective is the one to have. So despite thinking from your own perspective (i.e. employing the structure of practical rationality that you are in fact sensitive to in your behaviour) that φ-ing is the thing to do, you may assert the opposite, employing the structure of practical rationality that you think is the one to be sensitive to. That is the claim you may be committed to defend. This means that there is an important difference between actually thinking that φ-ing is desirable and thinking that φ-ing is actually desirable. Actually thinking that φ-ing is desirable is thinking this from one’s own perspective—the perspective that one’s actual behaviour is sensitive to. Thinking that it is actually desirable is thinking that from the desirable perspective to be thinking from φ-ing would be thought to be desirable. So, asserting and defending claims are practices that one may engage in by adopting perspectives that are not one’s own. The same goes for the practice of giving advice. When asked what to do by someone who is looking to buy some tripe for the sake of their health, I may suggest that she goes to a particular butchers shop. I do not actually think that eating or buying tripe is the thing to do—I may even disapprove of eating meat—but I can adopt the perspective that I take to be the one my interlocutor should have for the sake of giving her some advice. I may even advise myself by adopting the perspective I take to be the one I should have. If the weak-willed TV watcher exercises sensitivity to the adopted perspective she takes to be the one she should have, this may result in her telling herself that she should switch off the TV. She may assert this and in so doing commit herself to justifying the claim if required to. She will not assert that the thing to do is to keep watching TV, as she is not committed to any justification of this claim. The conception of practical rationality she is actually committed to justifies the assertion that switching off the TV is the thing to do. But even though her assertions and her advice are sensitive to this conception of practical rationality, her behaviour generally is not. She does not actually intend to switch off the TV. Adopting the perspective she takes to be the one to have she concludes that switching off the TV is the thing to do, but this is not in fact her perspective, and so she does not follow her own advice.

52

R. Stout

Why she does not follow her own advice is another matter—a matter for psychological examination. Perhaps she is too tired and comfortable to change the way she thinks about what to do. So, the second-order goal—to think that the thing to do is to switch off the TV—fails to be achieved, and the agent fails to form the first-order intention to switch off the TV. She thinks that the way to think is that the thing to do is to switch off the TV, but she does not think that the thing to do is to switch off the TV.14 This means that she thinks she is wrong to think that the thing to do is to watch more TV. And this is why weakness of will seems so paradoxical. If you are wrong to think that p and the thing to think is in fact not-p, then it follows that it is not the case that p. Presumably, the weak-willed agent knows this. She actually thinks that watching some more TV is the thing to do, while thinking that it is not actually or ultimately the thing for her to be doing. And she might think this even though she knows full well that if something else is actually or ultimately the thing to do, then that something else is the thing to do. So, she thinks that p (switching off the TV is ultimately the thing to do), knows that if p then q (switching off the TV is the thing to do) but does not think that q. So, the weak-willed agent fails to apply Modus Ponens in her thinking. This may seem like an implausible degree of irrationality. But note that her failure to apply Modus Ponens in her thinking is not a failure to see how she ought to apply Modus Ponens. She knows exactly what she should think; her failure is to think what she thinks she should think. She may be working on the assumption that watching more TV is the thing to do in the way she thinks and acts. This is manifested by her continuing to watch the TV. But at the same time, she may be working on the assumption that this is not really the thing to do. This is manifested in the pattern of her assertions and also in her eventually changing the way she thinks (or at least trying to change it) so that she changes her mind and stops watching TV. Before she succeeds in forming the intention to switch off the TV, she thinks that she should form that intention. At that time, she thinks both that she should not switch off the TV and that she should change the way she thinks. She thinks she is wrong to think that she should continue to watch the TV. Weakness of will involves doing something while thinking you are wrong to be doing it. On the Davidsonian conception of action I am defending, this is to think something is desirable while thinking you are wrong to think this. Somehow this seems more paradoxical. How could you fail to make such an obvious inference in your thinking? But the point is that the sensitivity to rationality of your thinking may encounter obstacles like anxiety or tiredness that the sensitivity to rationality of your assertions does not encounter. The sensitivity to rationality of your thinking may encounter the same obstacles that the sensitivity to rationality of your action 14 Sarah Buss takes a similar line, though her focus on ‘all things considered’ judgements is misplaced in my view. ‘On my account, then, the characteristic perversity of weak-willed action lies not in a conflict between the weak-willed agent’s all-things-considered evaluative judgement and her intention, but in a conflict between her actual all-things-considered evaluative judgement and the all-things-considered evaluative judgement, she believes she would have reached if, when she acted, she had accepted the authority of her own reason”. (Buss 1997, 36).

Acting Against Your Better Judgement

53

does. This is because what you think determines what you do. Really thinking that something is the thing to do is harder than thinking that it is really the thing to do.

References Anscombe, G. E. M. (1957). Intention. Cambridge MA: Harvard University Press. Bratman, M. (1979). Practical reasoning and weakness of Will. Nous, 13(2), 153–171. Buss, S. (1997). Weakness of Will. Pacific Philosophical Quarterly, 78(1), 13–44. Davidson, D. (1980). Essays on actions and events. Oxford: Oxford University Press. Kenny, A. (1975). Will, Freedom and Power. Oxford: Blackwell. McDowell, J. (2010). What is the content of an intention in action. Ratio, 23, 415–432. Price, A. (2008). Contextuality in practical reason. Oxford: Oxford University Press. Rödl, S. (2007). Self-consciousness. Cambridge MA: Harvard University Press. Stout, R. (2019). Practical reasoning and practical knowledge. Canadian Journal of Philosophy, 49(4), 564–579. Tenenbaum, S. (1999). The judgment of a weak will. Philosophy and Phenomenological Research, 59(4), 875–911. Thompson, M. (2008). Life and action. Cambridge MA: Harvard University Press.

A Davidsonian Theory of Evaluative Judgment Ya-Ting Chang

Abstract It is acknowledged that the debate between internalism and externalism is closely related to the debate between cognitivism and non-cognitivism regarding evaluative judgment. As the Humean theory of motivation is the dominant view, most philosophers endorse the idea that desire is necessary for motivation. It follows that the internalist usually appeals to non-cognitivism to argue for internalism, because if evaluative judgments express/are desires, then it seems that evaluative judgments must motivate. But then evaluative judgments cannot be true or false, for desires lack cognitive contents. Such a view seems to be counterintuitive, as we tend to think of evaluative judgments as having cognitive contents. The aim of this paper is to settle this issue by reconstructing a Davidsonian theory of evaluative judgment. I will argue that, following Davidson, by taking evaluative attitudes as conative propositional attitudes, we will be able to see how evaluative judgments can be both necessarily motivating and truth-apt. Keywords Donald Davidson · Evaluative judgment · Motivation · Cognitivism · Internalism

1 Introduction The discussion over judgment internalism is not only about motivation, but also intertwined with the meaning or semantics of evaluative judgments. Arguments for This chapter is in its final form and it is not submitted for publication anywhere else. This paper was finished while the author was a visiting scholar at the Department of Philosophy and Research Centre for Agency, Values, and Ethics (CAVE), Macquarie University, from August 2018 to August 2019 and funded by the Ministry of Science and Technology, Taiwan, under the Postdoctoral Research Abroad Program (MOST: 107-2917-I-564-002-A1). Y.-T. Chang (B) Independent Scholar, Sydney, Australia e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_4

55

56

Y.-T. Chang

judgment internalism often appeal to the action-guiding character of evaluation. However, within the Humean framework, only desire has a direct impact on human action. That is to say, only desire issues motivation. It follows that if judgment internalism holds, then evaluative judgment must have a necessary connection with desire. From the Humean perspective, belief and desire are distinct mental entities: belief is a cognitive attitude inert to motivation, whereas desire is a conative attitude lacking cognitive content. It follows that evaluative judgments do not have cognitive content. Without cognitive content, evaluative judgments cannot have truth values. Whether evaluative judgments have truth values or not is the debate between cognitivism and non-cognitivism. If evaluative judgments are cognitive, then they are truth-apt; if evaluative judgments are conative, then they are not truth-apt. This is a different issue from whether evaluative judgments must motivate. But as I just mentioned, these two issues are intertwined, thus one cannot hope to deal with either independently. The problem here is that both judgment internalism and cognitivism are backed by certain strong intuitions about the essential features of evaluative judgment. On the one hand, it is undeniable that our evaluation guides our action in a way such that our evaluative judgments regularly result in subsequent actions. On the other hand, it is undeniable that evaluative judgments have cognitive contents and can be true or false. What we encounter is a predicament of explaining both features of evaluative judgment within the Humean framework. Call this the Humean compatibility issue regarding evaluative judgment. Sticking to the motivational feature of evaluative judgment, most Humeans choose to renounce the view that evaluative judgments have cognitive contents. Donald Davidson is an exception.1 He argues that there is a way to account for both features of evaluative judgment if we re-examine the fundamental relation between evaluation and language. Specifically, Davidson points out that evaluative judgments must have truth values in order to play a role in practical reasoning. But Davidson never directly deals with the Humean compatibility issue. In this paper, I aim to reconstruct a Davidsonian theory of evaluative judgment in virtue of resources found in Davidson’s works and demonstrate how it can solve the Humean compatibility issue. In order to carry out the Davidsonian theory of evaluative judgment I have in mind within the Humean framework, I will begin with an investigation of judgment internalism in section II. In section III, I reconstruct Davidson’s diagnosis of the fundamental connection between evaluation and language. Then in section IV, I propose a Davidsonian theory of evaluative judgment and show how it accommodates both judgment internalism and cognitivism with certain necessary supplementary explications. I conclude with a brief review of my argument in section V.

1 Michael

(1994).

Smith is probably the most prominent exception in contemporary philosophy. See Smith

A Davidsonian Theory of Evaluative Judgment

57

2 Evaluative Judgment Internalism There are two significant questions regarding evaluative judgment. Firstly, there is the question about the connection between evaluative judgment and motivation. This question leads to the debate between judgment internalism and judgment externalism. Hereafter, I sometimes call these two positions “internalism” and “externalism” for simplicity if the context is clear enough. The simplest form of judgment internalism can be captured by the intuition shown as follows: Simple Internalism: If a person judges that φ-ing is valuable, then, necessarily, he is somewhat motivated to φ. Judgment externalism is better characterized as the denial of internalism, saying that there is no such necessary connection between evaluative judgment and motivation. Secondly, there is the question about the content and the objectivity of evaluative judgments. This question leads to the debate between evaluative cognitivism and non-cognitivism. Roughly, cognitivism maintains that evaluative judgments, or their representations, have objective truth values, because they are cognitive. Again, let us put it into the simplest form: Simple Cognitivism: Evaluative judgments have objective truth values. Non-cognitivism is the opposing view of cognitivism, maintaining that evaluative judgments are non-cognitive or conative and thus cannot have truth values. A satisfactory theory of evaluative judgment is required to answer both questions in a consistent manner. Furthermore, it would be better if one’s answer to one question provides resources for supporting one’s view on the other. Since my aim is to solve the Humean compatibility issue and most Humeans are judgment internalists, I begin by scrutinizing simple internalism and examining common ways to argue for it. At the end of this section, a predicament of internalism will emerge. A popular approach to judgment internalism is to appeal to a form of noncognitivism—expressivism, which claims that public evaluative judgments function to express conative attitudes such as desires, approval or disapproval, or pro/con attitudes.2 For example, an agent’s utterances of his evaluative judgments are said to express his conative attitudes. If expressivism holds, it seems that internalism follows: if whenever an agent judges that travelling around the world is wonderful he is expressing his desire to travel around the world, then he seems to thus be motivated to travel around the world. Unfortunately, there are critical defects of expressivism. What is worse, according to Richard Joyce, even if expressivism holds, internalism does not follow.3 Let me elaborate on them respectively. The theoretical results of expressivism are obvious. If making an evaluative judgment is expressing a desire, then evaluative judgments have no truth value, because desires are not truth-apt. Although in simple cases, the idea of expressivism allows 2 Leading

ideas of expressivism in the twentieth century can be found in Ayer (1946), Blackburn (1984), and Gibbard (1990). 3 Joyce (2002).

58

Y.-T. Chang

us to understand what an agent means or what his mental attitude is, it is widely acknowledged that it fails to explain the meaning of evaluative judgments when they are embedded in larger or complex sentences. This is usually called the embedding problem, originated from Frege and Geach.4 And it turns out to be an intractable problem for expressivism. It is wildly accepted that, in addition to assertion, evaluative sentences may occur in conditionals, conjunctions, disjunctions, or other indirect discourses. But it is hard to see how, if an evaluative judgment expresses a person’s attitude, it contributes to the meaning of a complex sentence as a whole. Consider the following argument employing modus ponens.5 1. It is wonderful to travel around the world. 2. If it is wonderful to travel around the world, then it is wonderful to stay in different countries. 3. Therefore, it is wonderful to stay in different countries. The very first thing everyone would agree on is that this is a valid argument. However, it is hard to see how expressivism can show validity without talking about truth values. Since (1) is an assertion, according to expressivism, it expresses a certain attitude towards travelling around the world. But (2) is a conditional. To accept a conditional is not to assert anything, thus (2) does not express any attitude. It follows that the term “wonderful” has different meanings in (1) and in (2). And it becomes unclear how the non-cognitivist can show the argument to be valid. Some proponents of expressivism have been trying to develop a semantics for evaluative judgments on the basis of non-cognitive attitudes.6 Nevertheless, even if the expressivist succeeds in establishing the required semantics, it does not entail the truth of internalism. According to Richard Joyce, the truth of expressivism does not guarantee the truth of judgment internalism.7 The reason is easy to grasp. One can express an attitude with an utterance without having the requisite attitude. You can say “I’m sorry” to express your apology without feeling sorry. You can also utter “travelling around the world is wonderful” to express a pro-attitude towards travelling around the world, while having no such pro-attitude in your mind. Linguistic expressions or conventions do not represent a relation holding between a speaker and his mental attitudes. Still, an insincere apology, in which the speaker does not have the requisite attitude for apology, is an apology. Similarly, an expression of an evaluative judgment still expresses a favourable or a pro-attitude, even though the speaker may not actually have such an attitude. If expressivism is true, what it informs us is merely that public evaluative judgments function to express conative attitudes. It does not follow that a speaker uttering an evaluative judgment actually has a corresponding pro-attitude, and thus does not entail the occurrence of motivation. 4 Geach

develops this objection with an idea inspired by Frege. See Geach (1965). also Sinnott-Armstrong (1993: 209–302). 6 See Gibbard (1990), Blackburn (1993), and Schroeder (2008) for the expressivist semantics projects. 7 Joyce (2002: 337–339). See also Ridge (2006). 5 See

A Davidsonian Theory of Evaluative Judgment

59

A natural move for internalism is to appeal to the mentalistic approach to noncognitivism, in which linguistic expressions play no role. Such an approach is to identify evaluative judgments with conative mental states or attitudes: Mentalistic non-cognitivism: evaluative judgments are conative attitudes such as desires. The mentalistic construal of non-cognitivism logically entails judgment internalism. Obviously, if evaluative judgments are desires and desires necessarily motivate, then evaluative judgments necessarily motivate. However, as Joyce points out, if this is what non-cognitivism claims, then the entailment between non-cognitivism and internalism turns out to hold trivially and the truth of non-cognitivism ceases to offer substantial justification to internalism. Furthermore, Joyce asserts that “taking the argument in this mentalistic direction makes the whole issue an empirical one—a very unwelcome result”.8 It is unwelcome because such an approach attempts to settle the issue in terms of scientific investigations of what counts as “making an evaluative judgment”, what counts as “having a conative attitude”, and whether they could be empirically identified with each other. We thus reach the conclusion that non-cognitivism cannot support judgment internalism. Nevertheless, given the Humean theory of motivation, we are still certain that evaluative judgment must have something to do with desire if internalism is to be true. Recall that simple internalism regarding evaluative judgment asserts that there is a certain internal connection between evaluative judgment and motivation.9 In folk psychology, there are two distinct kinds of mental states or attitudes, cognitive attitudes such as beliefs and non-cognitive attitudes such as desires. The Humean tells us that belief alone does not motivate. This implies that to motivate, the occurrence of desire is necessary. If one is motivated to do something, one must desire to do it, whether one has a belief that doing so is good or not. Such a Humean theory of motivation asserts that there is a conceptual link between desire and motivation. And according to judgment internalism, there is a conceptual link between evaluative judgment and motivation. For judgment internalism to hold within the Humean framework, there must be a conceptual link between evaluative judgment and desire. As we have precluded the possibility that making an evaluative judgment is identical with having a desire, the conceptual connection between evaluative judgment and desire must exist in a different way. At the end of this paper, we will see how a 8 Joyce

(2002: 343).

9 What we have been concerned with is the connection between evaluative judgment and motivation.

Analytically, there are two possibilities. Either the connection holds between judging evaluatively and motivation or it holds between evaluation and motivation. The first sort of internalism is usually called judgment internalism, as we have seen, which states the necessary connection between genuine evaluative expressions, thoughts, or convictions and motivation. It is the sort of view about the nature of evaluative thought or language. The second sort of internalism has the name of existence internalism, as Stephen Darwall labels it, which makes a metaphysical claim about the necessary connection between evaluative facts or truths and motivation. See Darwall (1992, 1995). Although what Darwall talks about are normativity and ethical or normative judgments, the characterization applies to evaluative judgment as well.

60

Y.-T. Chang

Davidsonian understanding of the relation between evaluative judgment and desire can make room for the compatibility between judgment internalism and cognitivism. It seems that we had better go back to a fundamental question about what it is to make an evaluative judgment. To know this, it is necessary to know what it is to make a judgment. From the orthodox point of view, “to judge”, “hold a judgment”, and “make a judgment” refer to a mental act,10 namely the act of affirming or denying an object’s having a certain property. One should bear in mind that, in current discourse, to put it this way has no metaphysical implication of the independent existence of “properties”. What it is supposed to mean is the mental act of affirming that a particular object falls under a given concept. Or in Davidson’s words, to make a judgment is to apply a concept, to affirm that an object or an event or a situation has a certain characteristic.11 Following this line of thought, we can distinguish between factual or descriptive judgment and evaluative or normative judgment. Descriptive judgments are mental acts of affirming that such and such objects have certain descriptive properties. In contrast, evaluative judgments are mental acts of affirming that such and such objects have certain evaluative properties. No wonder a number of philosophers tend to hold the cognitivist view which identifies making judgments as believing. If to judge is to believe, there is no reason to deny that making an evaluative judgment is being in the cognitive state of believing. And it seems to follow that the difference between descriptive judgments and evaluative judgments lies in the properties being judged. However, the concept of judgment must be different from the concept of belief; otherwise, there would be no such debate between cognitivism and non-cognitivism. The core idea of cognitivism is that evaluative judgments are apt to have objective truth values. This is different from the realist claim about the existence of evaluative properties. A judgment’s being true does not and need not entail the existence of a certain abstract entity. Hence, it is not necessary to characterize making an evaluative judgment as believing that a certain evaluative property exists. In addition, according to the internalist intuition, an essential difference between descriptive judgment and evaluative judgment is that the latter is action-guiding but the former is not. Even the externalist must admit that there is a regular connection between evaluative judgment and its subsequent action.12 Taking these two considerations together, a proper characterization of evaluative judgment, it seems to me, has to capture the asymmetry between descriptive judgment and evaluative judgment on the one hand and the similarity between them on the other hand. The asymmetry is that one has a direct influence on action, while the other does not. The similarity is that both have cognitive content such that we can talk about their truth and falsity. We will see how this is possible with resources found in Davidson’s theory of evaluative judgment.

10 See

introduction of Björnsson et al. (2015) and Svavarsdóttir (2006). (1995a, b: 9). 12 For an argument for this point, see Stroud (2003). 11 Davidson

A Davidsonian Theory of Evaluative Judgment

61

3 Language and Evaluation: Davidson’s Diagnosis If we begin with the orthodox view that making a judgment is mentally affirming that something has a certain property, then we will naturally accept the idea that making an evaluative judgment is mentally affirming that something has a certain evaluative property. For Davidson, to ask what it is to mentally affirm that something has a certain evaluative property is to ask what the relation between evaluation and language is. And the question about the relation between evaluation and language is the question about how evaluative attitudes are expressed in language. Evaluations or evaluative attitudes are attitudes such as desiring, wishing, wanting, taking to be right, holding to be good, hating, disliking, taking to be wrong, holding to be bad,…, and so on, whatever attitude an agent has that can be represented as valuing something.13 In ordinary discourse, our evaluative attitudes are represented in terms of explicitly evaluative terms. If you have a positive attitude towards the brunch you just finished, you might well express this attitude in virtue of an explicitly evaluative term to form a sentence that expresses your evaluative judgment such as “The brunch is delicious!” or “The brunch tastes good!”. No wonder it is common that the study of value theory has focused on explicitly evaluative terms. It appears that we can understand evaluative judgment in terms of the study of the meaning of “good”, “bad”, or any term explicitly made for expressing evaluation. Contrary to this line of thought, Davidson makes two claims, one negative and the other positive.14 I shall discuss the negative claim in this section and leave the positive claim to the next section: The negative claim: We can learn relatively little about ethics and values generally by concentrating on explicitly evaluative language. The positive claim: There is a connection between evaluative attitudes and meaning, a connection that is fundamental to understanding both. Davidson offers two arguments to support the negative claim, a minor argument, and a major one. The minor argument shows that we can express our evaluative attitudes without the use of explicitly evaluative terms. Let me reconstruct the argument in terms of the following example. Suppose I am shopping for a dress with my friend who shares my taste. When I say “This dress is very Korean”, my friend would immediately understand that I have a positive attitude towards that dress. I don’t have to say anything such as “It is good that this dress is very Korean”. Moreover, the sentence “This dress is very Korean” could express more accurately my attitude than the sentence “This dress is good” could do. Explicitly evaluative terms are not necessary for expressing our evaluations. What are necessary are our evaluative attitudes towards the sentences being evaluated.

13 Davidson 14 Davidson

(1984), 20. (1984: 19).

62

Y.-T. Chang

The fact is that, although we often express our evaluations by using evaluative terms, we can, and we actually do, express our evaluative attitudes without them. Evaluation therefore cannot be a speech act as expressivism suggests. According to Davidson, the underlying reason is that there is no linguistic convention governing our use of a word and the meaning of the word. We are thus unable to understand evaluative judgments in terms of explicitly evaluative terms, as the use of such terms has no necessary connection with the agent’s evaluative attitudes. One might have an evaluative attitude without using an explicitly evaluative term to express it. And one might use an explicitly evaluative term to express an evaluative attitude that one does not have. Evaluative attitudes are not speech acts, so the relation between evaluation and language does not lie in the attitude that a speaker has towards his audience. This is not to deny that expressivism might be true in saying that public expressions of evaluative judgments can function to express evaluative attitudes. The point is that the function of linguistic expressions of evaluative judgments is one thing, while the nature of evaluative judgment is another. As Davidson maintains, the real issue is how to characterize evaluative judgments in terms of the speaker’s evaluative attitudes towards his sentences. This leads us to Davidson’s major argument for the negative claim. This is an argument contending that focusing on evaluative terms represents an incorrect asymmetry between descriptive sentences and evaluative sentences, and therefore leads to inadequate views on evaluative sentences. Assuming that there is a clear asymmetry between belief and evaluation. As Davidson sees it, traditional approaches to understanding evaluative sentences often appeal to their structural similarity with descriptive sentences. When a person judges that winter is cold, it is natural to characterize him as believing that winter is cold or holding it as true that winter is cold. It is therefore natural to think of the person “embracing”, “accepting”, or “holding” descriptive sentences of the following form:15 (B)Winter is cold. Similarly, given that our evaluations are towards states of affairs or actions or events, it is natural to think of a person who judges that the coldness of winter is valuable “embracing”, “accepting”, or “holding” evaluative sentences of the following form: (E)It would be good if winter is cold. This has a form similar to descriptive sentences that express the cognition of certain states of affairs. As Davidson puts it, following this line of thought, the relation between evaluation and language is to be understood as this: to make an evaluative judgment or to value a certain proposition is to embrace or accept an evaluative sentence, that is, to hold it as true that “It would be good if winter is cold”. This is in 15 See

Davidson (1984: 24) for his use of these terms.

A Davidsonian Theory of Evaluative Judgment

63

contrast with the relation between belief and language: to make a descriptive judgment or to believe a certain proposition is to embrace or accept a descriptive sentence, that is, to hold it as true that “Winter is cold”.16 Such an approach takes “embracing an evaluative sentence” as the key relation between evaluation and language, in contrast with taking “embracing a descriptive sentence” as the key relation between belief and language. From Davidson’s point of view, taking “embracing an evaluative sentence” as the basic relation between evaluation and language heavily relies on the study of explicitly evaluative sentences, i.e. “sentences that ‘say’ that something is desirable or good or obligatory”.17 In this way, the asymmetry between evaluation and belief is characterized in terms of the content of such sentences. To believe a proposition is to hold it as true that something has a certain descriptive property. In contrast, to value a proposition is to hold it as true that something has a certain evaluative property. But there is something troublesome in this approach. Below is Davidson’s diagnosis: Such an approach encourages one of two distortions: either … [evaluation] is seen as a special form of belief, a belief that certain states or events have a moral or other kind of value, or are obligatory, etc.; or evaluative sentences are thought to lack cognitive content. The first of these views makes evaluation too much like cognition; the second bifurcates language in an unacceptable way by leaving the semantics of evaluative sentences unrelated to the semantics of sentences with a truth value.18

This quoted passage shows why neither cognitivism nor non-cognitivism is satisfactory for understanding evaluative judgment. Just like sentence (B) says that something has a certain descriptive property, sentence (E) says that some state of affairs is good. If we contrast evaluative judgments with descriptive judgments by studying the difference between embracing explicitly evaluative sentences and embracing descriptive sentences, it is easy for us to fall into the debate between cognitivism and non-cognitivism. The reason is obvious. What is it to embrace a sentence? To embrace a sentence is to hold it as true. Cognitivism is the view that to value a proposition is to hold an evaluative sentence as true. Non-cognitivism, as the denial of cognitivism, is the view that to value a proposition is not to embrace an evaluative sentence. Rather, to value a proposition is to express a pro/con attitude, emotion, or approval/disapproval— that is, to perform a certain speech act. Opting for either position seems to leave something out. On the one hand, to value something is quite different from holding something as true. The cognitivist approach fails to capture the non-cognitive aspect of evaluation. On the other hand, it is hard to deny that evaluative sentences have cognitive contents. The non-cognitivist approach fails to offer a proper semantics for evaluative sentences. As Davidson observes, the study of the language of morality or value by and large focuses on explicitly evaluative language: through the investigation of explicitly evaluative terms, it is said we can know about evaluative properties and thus evaluative 16 For

further discussions, see also Bricke (2012). (1984: 25). 18 Davidson (1984: 25–6). 17 Davidson

64

Y.-T. Chang

facts or truths. However, Davidson’s arguments have shown that such a way gives little hope of seeing the genuine relation between evaluation and language. Davidson’s diagnosis also sheds light on why the formulation of non-cognitivism usually appeals to “expressions” of evaluative judgments, while most of the time cognitivism can simply be formulated as a view about evaluative judgments identified with beliefs.19 The underlying difference lies in a fundamental disagreement on the key relation between evaluation and language. For Davidson, what matter to evaluation are our evaluative attitudes towards the propositions being valued rather than our uses of evaluative words. The asymmetry between belief and evaluation thus is better characterized in a way different from the distinction between “embracing a descriptive sentence” and “embracing an evaluative sentence”. According to Davidson, we can characterize the asymmetry in another equivalent way that is promising in preserving both the cognitive and the conative aspects of evaluative judgment. How could this be done?

4 A Davidsonian Theory of Evaluative Judgment We have seen why it is inadequate to take “embracing a sentence” as the fundamental relation between evaluation and language. Either it leads to the cognitivist approach that focuses the study of value on explicitly evaluative terms, or it leads to the noncognitivist approach that takes values as expressions of emotions or conative attitudes. In this section, I reconstruct Davidson’s arguments for his positive claim that the key relation between evaluation and language lies in the connection between evaluative attitudes and language or meaning. We will then be able to distinguish between two distinct issues: the issue about understanding a person’s evaluative attitude and the issue about the truth condition of an evaluative judgment. According to Davidson, the asymmetry between evaluation and belief can be characterized in another way—a speaker’s having different mental attitudes towards sentences or propositions. This is to suggest taking evaluative attitudes as conative propositional attitudes, in contrast with beliefs as cognitive propositional attitudes.20 Instead of the contrast between embracing a descriptive sentence such as (B) and embracing an evaluative sentence such as (E), Davidson suggests that we characterize the asymmetry between evaluation and belief as: (B*)Believing “winter is cold” to be true. and. (E*) Wanting or desiring “winter is cold” to be true.

19 For the cognitivist view that regards moral judgments as expressions of beliefs, see, for example, Smith (1994), and Brink (1997) For the cognitivist discussion without the use of expression, see Svavarsdóttir (1999), and Hills (2015). 20 See Davidson (1984, 1985).

A Davidsonian Theory of Evaluative Judgment

65

Davidson argues that we could think of a person who puts a positive value on the coldness of winter as “wanting or desiring the sentence ‘winter is cold’ to be true”, just like we could think of a person who believes in the coldness of winter as “believing the sentence ‘winter is cold’ to be true”. Taking such a contrast as basic, the relation between evaluation and language can be investigated in terms of the study of the relation between propositional attitudes and meaning. Therefore, we are able to understand people’s evaluative attitudes in the same way as we understand their beliefs. At this point, we need a briefing of Davidson’s theory of interpretation to proceed. Here, I summarize Davidson’s theory of interpretation for propositional attitudes and their contents. In Davidson’s view, the possibility of interpretation of a speaker’s propositional attitudes and their contents depends on two presumptions. Firstly, the interpreter and the speaker share the same common ground of living. Secondly, the interpreter must adopt the principle of charity when processing interpretation—that is, to interpret others, we have to fit our own propositions and attitudes to their words and attitudes “in such a way as to render their speech and behaviour intelligible”.21 In addition, since the contents of a speaker’s propositional attitudes are highly interconnected in the way that any propositional content depends on its consistency with indefinitely many other propositional contents, we cannot attribute to a speaker an individual propositional attitude without attributing many other propositional attitudes. And the only way this is possible is to attribute to him the propositional attitudes that we have on our own. Desires, beliefs, and meaning (the contents of propositional attitudes) are holistic in the same way—in the way that they are largely coherent and consistent with each other.22 This is just an outline of how to understand evaluative attitudes as propositional attitudes. My present concern is to defend Davidson’s thesis that it is better that we take evaluative attitudes as propositional attitudes, for this will enable us to characterize evaluative judgment more accurately. Another strong reason for taking evaluative attitudes as propositional attitudes in parallel with cognitive attitudes can be grasped via the passage below.23 The same sentences are the objects of both belief and desire: this reinforces the claim that the interpretation of the evaluative attitudes proceeds along the same general lines as the interpretation of the cognitive attitudes.24

When we interpret other people, we try to make them intelligible by interpreting not only their beliefs but also their evaluative attitudes. The fact that we can have different kinds of attitudes towards the very same (descriptive) sentence ensures that the relation between evaluative attitudes and sentences is as basic as the relation between beliefs and sentences. Successful interpretation relies on seeing the logical relations among cognitive propositional attitudes such as beliefs, conative propositional attitudes such as desires and the meanings of the propositional contents. This 21 Davidson

(1984: 35). also Myers (2004, 2013). 23 For this point, see Davidson (1984, 1986, 1980). 24 Davidson (1986: 71). 22 See

66

Y.-T. Chang

unified theory of desires, beliefs, and meanings not only allows us to understand others’ thoughts but also makes it possible for us to understand others’ actions.25 According to Davidson, once the basic connection between evaluative attitudes and descriptive language is established, we can interpret explicitly evaluative judgments in virtue of interpreting a person’s evaluative attitudes. And the simplest way is to “identify desiring a sentence to be true with judging that it would be desirable if it were true”.26 Let me make it clear. The very question I am interested in dealing with in this paper is what it is to make an evaluative judgment. From the orthodox perspective, making an evaluative judgment is mentally affirming that a certain object falls under an evaluative concept. Usually, we characterize such a mental act in this way—a person holds a sentence with an explicitly evaluative predicate as true. What Davidson suggests is that there is another conceptually equivalent way to characterize the mental act—that is, a person has an evaluative attitude towards a non-evaluative sentence. Call this the equivalence thesis: The Equivalence Thesis: Evaluation or evaluative judgment can be characterized in two equivalent ways: desiring a sentence to be true or judging that it would be desirable if a sentence were true. Note that these two ways of characterizing evaluative judgment are equivalent because they both represent the asymmetry between belief and evaluation successfully. Nevertheless, Davidson further argues that, from the radical interpreter’s perspective, the attitude of desiring is conceptually more basic than the attitude of holding true. It is more basic in the sense that “if we know enough about a person’s desires, we can work out what he believes, while the reverse does not hold”.27 Here, desire must not be considered as simply non-relative, but it should be considered as a form of simple preference—a relation among an agent and two alternatives, one of which is preferable or desired more strongly than the other by the agent. The possibility that measurable desires can be reduced to simple preferences ensures that preference is more basic than simple desire. Our choices of one course of action rather than another or our preferences for one state of affairs over another are determined by two things: the values we set on possible outcomes and our beliefs about the likelihood of those outcomes. The interpreter can work out an agent’s preferences in terms of identifying the extensional or causal relation between the agent and his preferences, that is, in terms of knowing what causes the agent to prefer a certain state of affairs over another. And once the interpreter knows the agent’s preference, he can work out what the agent believes. Davidson illuminates this point with an example of mountain climbing.28 Suppose you want to see into the next valley. Upon your choosing to act, there are two possible choices and three possible outcomes. If you climb the easier mountain, K1, you might 25 For

more discussions on evaluative attitudes, see also Preyer (2011). (1984: 37). 27 Davidson (1984: 26). Davidson argues for this point in Davidson (1980) as well, see especially 158–159. 28 Davidson (1984: 27). 26 Davidson

A Davidsonian Theory of Evaluative Judgment

67

or might not see the view you expect. If you climb the harder mountain, K2, you are certain to see the view you want. The best outcome is that you climb K1 and see the view, and the worst outcome is that you climb K1 and get disappointed. Climbing K2 is somewhere in between. According to Davidson, given your value on the outcomes, what you choose reflects what you believe. Since you rank the outcomes in the way specified above, your belief in the probability of succeeding in seeing the view on K1 will determine your choice. If you believe it is very likely to succeed, you will climb K1. If you think it is very likely to fail, you will climb K2. So your desire to see the view, together with the background of the situation, will let us know what you believe. This is the sense in which belief is embedded in desire, so Davidson argues. When Davidson makes the point that desire is in this sense more fundamental than belief, what he is concerned with is still the condition that makes interpretation possible, or, to put it in another way, the condition that makes understanding human action possible. He is not concerned with defending judgment internalism. In all his writings, he seems to take judgment internalism for granted. Fortunately, it seems to me that we have enough resources for constructing an argument for judgment internalism now. Still, this is an argument for understanding what it is to make an evaluative judgment. Here is the argument. Making an evaluative judgment can be characterized as either believing or desiring, as these two ways are equivalent in characterizing evaluative judgment. However, given that desire is conceptually prior to belief, the possibility of characterizing evaluative judgment as believing an evaluative sentence to be true depends on the possibility of characterizing evaluative judgment as desiring a descriptive sentence to be true. This is because, as Davidson insists, with proper methodology, we can derive belief and meaning from desire. As I see it, this turns out to imply that, for any evaluative judgment, we can find a description that characterizes it as a person’s having an evaluative attitude towards a descriptive sentence. But it is not true that, for any evaluative judgment, we can find a description that characterizes it as a person’s holding an evaluative sentence as true. Therefore, the understanding of evaluative judgment entails the understanding of evaluative attitude. The talk of understanding is an ordinary way of talking about concepts. So what the above suggests is that the concept of evaluative judgment entails the concept of evaluative attitude. In this way, the concept of evaluative judgment entails the concept of motivation. Judgment internalism is thus justified within the Humean framework. We now turn to the second issue about evaluative judgment—whether it has truth value or not. For this, Davidson answers in the affirmative. Call this the truth condition thesis: The truth condition thesis: Evaluative judgments have truth values. Davidson offers two arguments for the truth condition thesis, one from the perspective on the nature of evaluative judgment and the other from the concern about practical reasoning. Davidson’s first argument stems from the idea shown in the following passage.

68

Y.-T. Chang Creatures anything like us, creatures capable of thought and action, have concepts of the evaluative properties, and employ these concepts in making judgments.… …. For a creature to have a concept, it is not sufficient that it reacts differentially to things that fall under the concept: it requires that the creature be able to classify things it believes fall under the concept while aware that it may be making a mistake. To apply a concept is to make a judgment, and judgments may be true or false according to the creature’s own understanding of the concept.29

The first thing is that making a judgment is applying a concept. If one accepts this, one must accept that a judgment is either true or false, as the object being judged either falls under the concept employed or does not fall under it, whatever kind of judgment it is. This is uncontroversial for descriptive judgment. For Davidson, the point applies to evaluative judgment as well. If evaluative concepts are concepts at all, they must contribute to meanings when applied, and things can thus be said to fall under an evaluative concept or not. Following this line of thought, it is not surprising that Davidson also holds that evaluative judgments have truth conditions.30 And it is not surprising that the issue about whether evaluative judgments have truth values, whereas being an issue about the objectivity of values, is not a metaphysical issue about the existence of values. For Davidson, of course there are evaluative properties. To put it in his slogan, “values are rooted to things”.31 Nevertheless, Davidson insists that when we talk about values, what concerns us is not the locations of properties, but the objectivity of evaluative judgment. To ask the ontological question about what kind of entity value is, where it resides, or what entity makes an evaluative judgment true, is to ask the question about the source of values. But there is another way to talk about the objectivity of values, that is, to ask how we get to know the contents of evaluative judgments. To know the contents of evaluative judgments is to know how the contents are determined. This is the epistemological problem about the nature of evaluative judgment. According to Davidson, there is a distinct problem conceptually prior to the epistemological problem, one that asks what makes it possible for us to form judgments or to have thoughts. And the answer to this problem, hopefully, can show us the answer to the epistemological problem.32 What makes it possible for us to form a judgment, in Davidson’s view, is our possession of the concept of error. And our possession of the concept of error depends on our having the concept of objective truth. This is, basically, what the above quotation manifests. Having thought or having concepts requires us to be aware that we might be making mistakes, mistakes of our classification of what we believe to fall under a certain concept by our own lights. It follows that judgments, descriptive or evaluative, must have truth values. The truth value of a judgment is fixed by the facts and the content of the judgment. And in the sense that they have fixed truth values (that is, true, false, or no value), evaluative judgments are as objective as factual 29 Davidson

(2000: 54–55). (2000: 56). 31 Davidson (1995a, b: 52). 32 Davidson argues for this point in Davidson (1995a, b). 30 Davidson

A Davidsonian Theory of Evaluative Judgment

69

judgments. The truth of an evaluation is independent of how we make the evaluative judgment, just as the truth of a belief is independent of whether we have the belief or not.33 The beliefs and values we share and the common ground among us make us understand the shared criteria of what good and bad are, or what right and wrong are. Values are objective because we share a way of life, so Davidson argues.34 A subtle point needs to be addressed here. To classify things that I believe to fall under an evaluative concept is not to determine the content of an evaluative judgment on my own. Classifying things in terms of the criteria I have, as I understand the criteria, is appealing to a certain standard independent of me. This is different from my arbitrary assignment of something in any particular case. We do not live alone; we communicate with others. We understand the content of a particular judgment by interpreting others’ attitudes, both cognitive and evaluative, in a holistic way. Davidson maintains that, in learning to understand others’ utterances, we refer to things or objects common to us. More specifically, it is because we share the common ground of experience that we can understand others’ expressions. Given this, it is natural to think that we understand the expressions of evaluations in the same way as we understand the expressions of descriptive properties. We grasp evaluative concepts in the same way as we grasp descriptive concepts. And what makes our judgments true, what makes us judge that an object has a certain property, is the fact that there is a “systematic relationship between the attitude-causing properties of things and events, and the attitudes they cause”.35 Just as shining objects tend to cause us to believe that those objects are shining, honourable objects tend to cause us to honour those objects or, in other words, tend to cause us to have an evaluative attitude called honour towards those objects. It is the shared criteria or the interpersonal—that is, objective—standards that determine where the truth lies.36 A common misgiving is that Davidson’s view seems to commit to the guise of the good (GG) thesis, which is dubious and usually proved to be wrong by counterexamples.37 Here, I do not intend to defend or argue against the GG thesis. What I would like to do is to show that the Davidsonian view does not commit to the thesis 33 Here’s Davidson’s suggestion on objectivity: “A judgment is objective if it is true or false, or possibly neither, but its truth value (true, false or neither) is fixed: its truth value is independent of who makes the judgment, and of the society or period in which the thinker lives. The truth value of a judgment depends on just two things: the facts, and the contents of the judgment, the proposition being judged”. Davidson (1995a, b: 42). 34 This is fundamentally different from the view that values are objective because they are what we agree on. See Davidson (1995a, b: 51). 35 Davidson (1995a, b: 47), see also Davidson (2000, 1986). 36 Davidson (1995a, b: 50). One might have the worry that Davidson’s view commits to notorious subjectivism, and thus, as is well known, easily leads to relativism. But that is not true. Davidson thinks that making a judgment entails allowing the gap between what is judged to be good and what is actually good, and thus, values are objective in the sense that evaluative judgments have objective truth values. Actually, Davidson’s argument for evaluative judgment can be taken as a challenge to moral relativism. For further discussion about how it goes and the relativist replies, see (Gowans 2016) and (Prinz 2007). 37 For counter-examples, see (Stocker 1979).

70

Y.-T. Chang

and thus is not vulnerable to its criticisms. This clarification is important because one can defend a Davidsonian internalism without defending the GG thesis.38 In general, the GG thesis is regarded as saying something similar to Kant’s formulation of the old schools in Critique of Practical Reason: “We desire only what we conceive to be good; we avoid only what we conceive to be bad”.39 The formulation is considered as the minimal thesis because it states only that we desire what is believed by us to be good, not what is actually good. David Velleman famously characterizes Davidson’s view as rooted to the GG thesis. According to Velleman, Davidson’s approach is to “incorporate the valence of desire into its content, by describing desire, not as a favourable attitude towards the representation of some outcome, but rather as an attitude towards a favourable representation of the outcome”. On this understanding, “[d]esires are conceived as value judgments with intrinsic justificatory forces, so that the desire motivating an agent can be identified with the reason guiding him”.40 According to Velleman, Davidson’s identification of pro-attitudes with evaluative judgments is a way to provide justificatory force for desires. Thus when an agent does something according to his desire, he is rationally justified. And he is rationally justified because his desires are aiming at the good. Such an interpretation is partly influenced by Davidson’s theory of intentional action, in which desire can play a role in constituting reasons for action. I will not go into the issue of reasons for action here. Here, the concern is whether having an evaluative attitude or making an evaluative judgment involves pursuing the good. Robert Dunn raises a similar criticism: “Where Davidson groups together all the practical states that may enter into the explanation of action as desires with evaluative contents, he makes a mistake in categorizing desires (havings of goals) as values”.41 Dunn thinks that Davidson credits evaluative contents to pro-attitudes, and thus, the contents can serve as justification of actions. Under this interpretation, Davidson identifies desires with values. However, there is textual evidence against Velleman’s and Dunn’s interpretations. Davidson is aware of the possibility that one has a pro-attitude towards a certain action and yet does not regard it as worth performing. For this point, he offers an impressive example: “a man may all his life have a yen, say, to drink a can of paint, without ever, even at the moment he yields, believing it would be worth doing”.42 To clarify, it is important to notice the distinction between evaluative attitudes and evaluative judgments. Desires or pro-attitudes, according to Davidson, are a kind of attitude distinct from cognitive attitudes. But whatever an attitude is, it must have a descriptive content. It is natural that propositional attitudes have propositional contents. If you have a desire to drink a can of paint, then your desire has a descriptive content “I drink 38 Sergio Tenebaum thinks that Davidson commits to the GG thesis in his account of akrasia. See Tenenbaum (2007). 39 Kant (1997), Ak5: 58–9. 40 See Velleman (2000: 103–104). 41 See Dunn (2006: 101). 42 Davidson (1963: 4).

A Davidsonian Theory of Evaluative Judgment

71

a can of paint”. To put this desire into words, the desire and its content constitute an evaluative judgment “it is desirable that I drink a can of paint”. This judgment is evaluative because, as mentioned above, it can be characterized in terms of an evaluative attitude with propositional content—desiring “I drink a can of pain” to be true, and not because the content of the desire is evaluative. Many philosophers have confused this distinction. Once we keep the distinction in mind, we can realize why, in this example, the man has a pro-attitude towards drinking a can of paint, but does not regard the action of drinking the paint as valuable. He might actually think it a bad action. Indeed, what one desires to do need not be what we ordinary conceive to be good or valuable. In our everyday life, at times we have an inclination to perform some action we judge to be value-neutral. The feeling of obligation to follow the traffic rule at an intersection at midnight is not necessarily aiming at the good. Pro-attitudes are psychological inclinations or tendencies towards things that we observe; they do not aim at the (conceived) good.43 The second argument supporting the truth condition thesis relies on the validity of practical reasoning. Davidson observes that when we reason about what to do, we must combine factual judgments with our evaluations to reach a conclusion. In a simple piece of practical reasoning, we reason from a desire to eat something sweet and a belief that the chocolate is sweet to a conclusion that we intend to eat chocolate. To put desire and belief into expressions that can be used to reason to a conclusion that guides an agent’s action, Davidson maintains that the expressions of desires must be evaluative in form.44 An evaluative judgment can preserve an agent’s pro-attitude and meanwhile serves as a premise. As Davidson suggests, “if practical reasoning can be shown to be (in some cases) valid, then premises and conclusion must have truth value”, because “validity is defined as a truth preserving mode of reasoning”.45 I will come back to elaborate on this point later. Davidson’s insistence on the truth condition thesis leads several philosophers to read him as a cognitivist in the traditional sense. For instance, both Sergio Tenenbaum and Paul Hurley specifically put evaluative belief instead of evaluative judgment in their discussions of Davidson’s theory of akrasia.46 They take it that, by evaluative judgments, Davidson means beliefs involving evaluative properties. Usually, those who endorse the idea that evaluations are beliefs do not think evaluative judgments can settle the practical question. For instance, Nomy Arpaly maintains that “an agent’s 43 For Davidson, pro-attitudes include: “desires, wantings, urges, promptings, and a great variety of moral view, aesthetic principles, economic prejudices, social conventions, and public and private goals and values in so far as these can be interpreted as attitudes of an agent directed towards actions of a certain kind. The word ‘attitude’ does yeoman service here, for it must cover not only permanent character traits that show themselves in a lifetime of behaviour, like love of children or a taste for loud company, but also the most passing fancy that prompts a unique action, like a sudden desire to touch a woman’s elbow”. Davidson (1963: 4). 44 See Davidson (1978: 86). The same point is made implicitly in Davidson (1970) as well. 45 Davidson (2000: 55). 46 See Tenenbaum (2007, 2010), and Hurley (2002). Both Tenenbaum and Hurley find the apparent inconsistency between Davidson’s claim of the truth-aptness of evaluative judgment and his insistence of internalism. This drives them to develop their own approaches to solve the problem.

72

Y.-T. Chang

best judgment is just another belief, and for something to conflict with one’s best judgment is nothing more dramatic than ordinary inconsistency between beliefs, or between beliefs and desires”.47 Dunn argues that evaluative judgments about action are purely cognitive or theoretical.48 In this view, an evaluative judgment does not play any special role in determining subsequent action. One’s judging that having organic food is good cannot settle or even influence the answer to one’s practical question about what to eat for dinner. Hence, for these philosophers, there is no problem, either logical or psychological, if an agent’s action is contrary to the evaluative judgment he holds. Obviously, this cannot be Davidson’s view. Davidson cannot take evaluative judgments as identical with beliefs, since beliefs are inert to motivate. What he says is rather that evaluative judgment entails motivation and that evaluative judgments have truth values. And this point can be well addressed in the context of practical reasoning, although Davidson did not make the point explicitly. Here, I try to elaborate the point on his behalf. A piece of practical reasoning aims to answer the practical question, the question about what to do. To feature in a piece of reasoning, my desires and beliefs that I take as premises must be expressed in propositions, with which valid inferences can be made. According to Davidson, propositional expressions of desires must be evaluative in form. That is to say, the propositional expressions of my desires are characterizations of my evaluative judgments. To serve as a premise, it must be a proposition or a sentence. And it is the same for serving as a conclusion. Propositional expressions of beliefs are descriptive judgments that can be characterized in terms of holding a proposition as true. Similarly, propositional expressions of desires are evaluative judgments that can be characterized as holding an evaluative proposition as true. Therefore, we can formulate the following piece of reasoning without difficulty: Major Premise (Desire) hold “It is desirable to eat something sweet” as true. Minor Premise (Belief) hold “Chocolate bar is sweet” as true. Conclusion hold “It is desirable to eat chocolate” as true. What are included in the quotation marks are the propositional expressions of attitudes an agent has, and they are the genuine premises and conclusion with which an inference is made. The “hold it as true” attitude perfectly shows what it is like for the agent to process the piece of practical reasoning. He must hold those propositions as true in order to solve his practical issue. If he cannot be characterized as holding those propositions as true, it is hard to see in what sense this piece of practical reasoning is his. It is now clear that when we talk about desires and beliefs in the discourse of action, we are in effect talking about judgments rather than mental attitudes that can be loosely taken to have no content at all. Recall that we have two equivalent ways to characterize a judgment, to characterize it as holding a proposition as true or to characterize it as believing/desiring a proposition to be true. My elaboration can be supported by Davidson’s claim that an all-out or unconditional evaluative judgment 47 See

Arpaly (2000: 512). (1987).

48 Dunn

A Davidsonian Theory of Evaluative Judgment

73

is an intention.49 Although a propositional expression of an evaluative attitude is different from an evaluative attitude as a propositional attitude, they are equivalent characterizations of the very same mental act. For the propositional expression of an evaluative attitude to serve as a premise of practical reasoning, it must have a truth value. Otherwise, we cannot talk about the validity of practical reasoning. When you are considering whether to play an online game tonight or to keep working on your dissertation, you have a desire the expression of which is evaluative in a form such that it represents your desire on the one hand and functions well as a premise on the other hand. From this, we can at most infer that evaluative judgments have truth values. We cannot deduce judgment internalism. However, as I previously argued, if judging and having a certain propositional attitude are just different ways to characterize the very same mental act, then it entails judgment internalism. If every mental act that can be called evaluative judgment can be characterized in terms of evaluative attitude, there is a conceptual connection between making an evaluative judgment and desire where desire is broadly construed as a wide range of evaluative attitudes.

5 Conclusion The Davidsonian theory of evaluative judgment I have reconstructed in this paper shows us a promising way to solve the compatibility issue between judgment internalism and cognitivism within the Humean framework. First of all, I pointed out that, following Davidson’s point of view, the Humean compatibility issue lies in a misunderstanding of the relation between language and evaluation. The fundamental relation between language and evaluation does not lie in expressions of evaluation. Nor does it lie in explicit evaluative words. Rather, it lies in evaluative attitudes. And this is the crux of understanding the nature of evaluative judgment. Secondly, I argued that sticking to the point that to judge is to mentally affirm that an object falls under a certain concept, a judgment is better conceived as a mental event. And according to Davidson, there are two equivalent ways to characterize an evaluative judgment in terms of a person’s attitude towards a sentence—to take it as “embracing or believing an evaluative sentence” or to take it as “having an evaluative attitude towards a descriptive proposition”. To understand an evaluative judgment thus is to understand a person’s attitude of believing or his having an evaluative attitude. Thirdly, given that desire is conceptually more basic than belief, to understand an evaluative judgment ultimately is to understand an agent’s evaluative attitude. This conceptual dependency reveals the conceptual connection between evaluative judgment and desire by taking evaluative attitudes as propositional attitudes that can be characterized as “desiring a descriptive sentence to be true”. On the one hand, that is why making an evaluative judgment must motivate. On the other hand, it also manifests how evaluative judgments can have truth values. 49 Davidson

(1978: 99).

74

Y.-T. Chang

I hope this Davidsonian theory of evaluative judgment can shed some light on the discussion over evaluative judgment. If we can find a way to account for both the cognitive and the conative features of evaluative judgments, why do we stick to views that can only account for one?

References Arpaly, N. (2000). On acting rationally against one’s better judgment. Ethics, 110, 488–513. Ayer, A. J. (1946). Language, truth, and logic (2nd ed.). London: Gollancz. Björnsson, G., Strandberg, C., Olinder, R. F., Eriksson, J., & Björklund, F. (Eds.). (2015). Motivational internalism. Oxford: Oxford University Press. Blackburn, S. (1984). Spreading the word. Oxford: Clarendon Press. Blackburn, S. (1993). Essays in quasi-realism. New York: Oxford University Press. Bricke, J. (2012). Hume and Davidson: Passion, evaluation, and truth. In I. Kasavin (Ed.), David Hume and contemporary philosophy (pp. 141–156). Cambridge Scholars Publishing. Brink, D. O. (1997). Moral motivation. Ethics, 108, 4–32. Darwall, S. (1992). Internalism and agency. Philosophical Perspectives, 6, 155–174. Darwall, S. (1995). The Brisitsh moralists: Inventing internalism. In The British moralist and the internal “Ought”: 1640–1740 (pp. 1–22). Cambridge: Cambridge University Press. Davidson, D. (1963). Actions, reasons, and causes. In Essays on actions and events (pp. 3–21). Oxford: Clarendon Press. Davidson, D. (1970). How is weakness of the will possible? In Essays on actions and events (pp. 21–42). Oxford: Clarendon Press. Davidson, D. (1978). Intending. In Essays on actions and events (pp. 83–102). Oxford: Clarendon Press. Davidson, D. (1980). A unified theory of thought, meaning, and action. In Problems of rationality (pp. 151–166). Oxford: Clarendon Press. Davidson, D. (1984). Expressing evaluations. In Problems of rationality (pp. 19–37). Oxford: Clarendon Press. Davidson, D. (1985). Replies to essays I-IX. In B. Vermazen & E. Hamilton (Eds.), Essays on Donald Davidson: Actions and events (pp. 195–229). Oxford: Clarendon Press. Davidson, D. (1986). The interpersonal comparison of values. In Problems of rationality (pp. 59–74). Oxford: Clarendon Press. Davidson, D. 1995. The objectivity of values. In Problems of rationality (pp. 39–57). Oxford: Clarendon Press. Davidson, D. (1995). The problem of objectivity. In Problems of rationality (pp. 3–18). Oxford: Clarendon Press. Davidson, D. (2000). Appendix: Objectivity and practical reason. In Problems of rationality (pp. 52– 57). Oxford: Clarendon Press. Dunn, R. (1987). The possibility of weakness of will. Indianapolis: Hackett. Dunn, R. (2006). Values and the reflective point of view: On expressivism, self-knowledge and agency. London: Routledge. Geach, P. T. (1965). Assertion. Philosophical Review, 74, 449–465. Gibbard, A. (1990). Wise choises, apt feelings. Oxford: Clarendon Press. Gowans, C. 2016. Moral relativism. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy. Hills, A. (2015). Cognitivism about moral judgement. In R. Shafer-Landau (Ed.), Oxford studies in metaethics (pp. 1–25). Oxford: Oxford University Press. Hurley, P. (2002). A Davidsonian reconciliation of internalism, objectivity, and the belief-desire theory. The Journal of Ethics, 6, 1–20. Joyce, R. (2002). Expressivism and motivation internalism. Analysis, 62, 336–344.

A Davidsonian Theory of Evaluative Judgment

75

Kant, I. (1997). Critique of practical reason. New York: Cambridge University Press. Myers, R. (2004). Finding value in Davidson. Canadian Journal of Philosophy, 34, 107–136. Myers, R. (2013). Interpretation and value. In E. Lepore & K. Ludwig (Eds.), A companion to Donald Davidson (pp. 314–327). Oxford: Willy-Blackwell. Preyer, G. 2011. Evaluative attitudes. In J. Malpas (Ed.), Dialogues with Davidson: Acting, interpreting, understanding (pp. 325–342). Cambridge, MA: The MIT Press. Prinz, J. J. (2007). The emotional construction of morals. New York: Oxford University Press. Ridge, M. (2006). Sincerity and expressivism. Philosophical Studies, 131, 487–510. Schroeder, M. (2008). Being for. Oxford: Oxford University Press. Sinnott-Armstrong, W. (1993). Some problems for Gibbard’s norm-expressivism. Philosophical Studies, 69, 297–313. Smith, M. (1994). The moral problem. Oxford: Blackwell Publishers. Stocker, M. (1979). Desiring the bad: An essay in moral psychology. The Journal of Philosophy, 76, 738–753. Stroud, S. (2003). Weakness of will and practical judgment. In S. Stroud & C. Tappolet (Eds.), Weakness of will and practical irrationality (pp. 121–146). Oxford: Clarendon Press. Svavarsdóttir, S. (1999). Moral cognitivism and motivation. The Philosophical Review, 108, 161– 219. Svavarsdóttir, S. (2006). Ho do moral judgments motivate? In J. Dreier (Ed.), Contemporary debates in moral theory (pp. 163–180). Wiley-Blackwell. Tenenbaum, S. (2007). Appearances of the good: An essay on the nature of practical reason. Cambridge: Cambridge University Press. Tenenbaum, S. (2010). Akrasia and irrationality. In T. O’Connor & C. Sandis (Eds.),A companion to the philosophy of action (pp. 274–281). Wiley-Blackwell. Velleman, J. D. (2000). The guise of the good. In The possibility of practical reason (pp. 92–122). Oxford: Clarendon Press.

Davidson on Emotions and Values Pascal Engel

Abstract Although Davidson has never written specifically on emotions, we can reconstruct his views from his writings on attitudes and values. He defends a causal version of the cognitive theory of emotions which he associates to an objective conception of values. I confront his account of the correctness of emotions with the fitting attitude view of the relation of emotions and value. They can be reconciled if the correctness of an emotion is construed as a form of idealization. Keywords Davidson · Emotions · Justification · Values · Fitting attitudes · Correctness

1 Introduction Donald Davidson is a philosopher. By this I mean that his views were meant to be systematic and to be related not only to his own views and to other topics on which he wrote, but also to all important issues within philosophy as a whole, even when he did not spell out his positions explicitly. This makes him unique among other analytic philosophers, at a time when over-specialization is more or less the rule. This was the case with his views on ethics and meta-ethics, on which he did not write very much, but which we can reconstruct (Davidson 2004; Myers 2004, 2013; Engel 2017). This is also the case with his views on emotions. In this essay, I shall first recall the background of contemporary theories of emotion. Then I shall try to reconstruct Davidson’s analysis of emotions, and to discuss how it relates to his conception of

This chapter is in its final form and it is not submitted for publication anywhere else. P. Engel (B) Ecole des hautes études en sciences sociales, 54 Bd Raspail, 7006 Paris, France e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_5

77

78

P. Engel

value. Then I shall, more speculatively, try to spell out to what extent his views can answer some of the problems that have been raised by recent theorists of emotions. I shall suggest that Davidson’s views could have been compatible with a specific conception of the fitting analysis of the relation between emotions and values.

2 The Background of Contemporary Views on Emotions There have been, within contemporary analytic philosophy, quite a number of theories of emotions. One major feature of these views, since the 1950s, has been a divorce between the theories inspired by Hume and those inspired by Husserl and Meinong, which in many ways persists today, although the gap between the “analytic” conceptions and the “phenomenological” ones tends now to narrow down. At the time when Davidson wrote on these issues, there was a divide, in the English speaking world, between a British tradition led by Ryle (1949) and Kenny (1963) on the one hand, and philosophers like Chisholm (1969) and Findlay (1963) on the other, who were Meinong’s heirs. There was, nevertheless, during the 60s and 70s, a small tradition of writing on emotions within mainstream analytic philosophy, with the work of Thalberg (1977), Pitcher (1965), Lyons (1980) and Wilson (1975) among others, with which Davidson was familiar. Later, during the 80s and 90s, work on emotions started within cognitive science, which was to breed the contemporary boom on emotion research during the last thirty years. Davidson did not contribute to this cognitive science literature, to which, I suspect, he was in many ways hostile. His main references within psychology were the kind of psychology based on probability and decision theory that he himself pioneered in Stanford during the 50s (Davidson 1957; Tversky and Slovic 1982), and psychoanalysis, in which he had become more and more interested, in particular when he discussed these issues with Cavell (1993). It would be wrong, however, to suppose that Davidson approached these topics as from the outside. On the contrary, from the very start of his philosophical career, he thought about desire, emotions, passions, and the role of rationality in mental life. His dissertation on Plato’s Philebus, the main philosophical dialogue on pleasure and desire, bears witness of his early and deep interest in these issues. His now classical essays on action, events and adverbs, and his analysis of Hume’s conception of the passions (1976) were in large part the product of a dialogue with Ryle, Kenny and Chisholm. And his later essays, from 1980 to 2003, on the paradoxes of irrationality and on the divided self, are directly the product of his thinking on these issues, as is his late essay on Spinoza (1999). The problem of the nature of emotions is much behind his writings on values, which have been recently the focus of much attention (Myers 2004, 2012, 2013). So although the secondary literature on Davidson’s views on emotions is rather small (Green 2013), the issue is in many ways quite central to his views on mind, action, and values. Before looking at Davidson’s views, it will be important to give some background of recent views on emotions. There are three main strands:

Davidson on Emotions and Values

79

First, on most views, emotions are mental states or episodes involving the following features: 1. 2. 3. 4. 5. 6. 7.

a sensory experience associated to an intentional component a kind of appraisal, or some evaluative mode based on an affect physiological changes, facial expressions and other behavioral effects characteristic feelings and phenomenal qualia cognitive and attentional processes an action-tendency or some other kind of motivational component. a neuronal basis.

As everyone knows, these features are neither exhaustive nor inclusive, and taken together they do not give necessary and sufficient conditions for emotions: some emotions do not issue in specific behaviors or in characteristic feelings, not all conceptions of emotions entail a cognitive component (behaviorist theories reject it), some emotions give rise to no action. But overall, many writers agree on this pattern of features, although each conception emphasizes some features and downplays some others, and it is reasonable to expect that any serious theory of emotion should account for these features or explain why some of them do not play a genuine role. Second, theories of emotion also diverge on the nature of emotions as mental phenomena. Are these episodes, dispositions, feelings, sentiments, passions, moods? Behaviorists view them as pieces of behavior producing feelings (James 1889). Some theorists see them mostly as conative states (Fridja 1986; Clore and Huntsinger 2007). Cognitivists see them as based on beliefs (Solomon 1984; Nussbaum 2001). To what extent do they depend on their neurological basis (Ledoux 1996)? To what extent are they socially and culturally constructed? What are the differences between basic and non-basic emotions? Are there kinds of emotions, such as those which are purely affective or those which are sometimes called “epistemic”? Can emotions combine, and give rise to mixed emotions? Indeed not all theories agree on these issues. Third, an important chapter of the philosophy of emotions concerns what we may call the epistemology of emotions: can emotions be justified? If so, in what sense? Can they be based on reasons? In what sense can they be rational? Can they be in some sense true? If they contain an evaluative element in what sense can they involve some normative appraisal and stand in some cognitive relation to values? Since at least Kenny (1963), who took up some medieval terminology (see also Chisholm 1969; Teroni 2007), it has been customary to distinguish the material object of an emotion (say the particular dog I am now afraid of) from its formal object (the typical kind of thing which is the object of fear, the fearable). An emotion is correct if it is appropriate to its formal object. It is also usual to talk of the cognitive basis of the emotion as the epistemic source from which the emotion flows (say my perception of a salivating dog). This source or basis is distinct from the object of the emotion (the dog) and from its content (that this dog is dangerous). This framework seems to presuppose that emotions can have a propositional content which is truthevaluable and that they can be in some sense judgements, hence to presuppose what is called a “cognitivist” conception of emotion. But the idea that emotions involve

80

P. Engel

valuing need neither rest on any cognitivist construal of emotions nor on the idea that they involve relations to propositional truths about values. The valuing can be constructed as being reduced to the affect itself, and values may be understood as mere responses to emotions, sentiments and other affective attitudes. Anti-realist conceptions of value may rest on such a response dependent conception. There is here a classical Euthyphronic dilemma: are values emotional responses or are they objective entities which emotional reckonings register or signal? In large part, the literature on the epistemology of emotions deals with the opposition between realist views about values, according to which emotions are based on some form of cognition (judgmental or perceptual) of values, and anti-realist views according to which values are the product or the projection of affective attitudes and emotions. Some realists take the cognitive basis to be a form of perception (Scheler 1913; Mulligan 1998; Tappolet 2016), anti-realists see values as merely response dependent, while “buck passing” views of values take them to be based on fitting attitudes (Brentano 1889; Scanlon 1998).1 So there are three major problems for all theories of emotions. The first is to understand on what kind of mental states—experiences, perceptions, beliefs, attitudes—the evaluations which emotions involve are based, and how they relate to their content. Let us call this the basis problem. The second is to examine whether these evaluations are relations to objective values or not, and in what sense. Let us call it the value problem. The third is to explain in what sense emotions can be said to be rational or irrational. This is the rationality problem.

3 Davidson’s Causal Theory of Emotions Now on this sketchy map, where would Davidson stand? He has an answer to the three problems, although he does not articulate these in a systematic way. He seems to be close to the cognitivist view, according to which emotions involve beliefs and judgments or are based on judgments. He is also an objectivist and a realist about values (Davidson 1984, 1995; Myers 2004; Myers and Verheggen 2016). But, unlike perceptual realists about emotions and fitting attitudes analyses, he does not base his realism about values on a theory of emotions. So what kind of relationship is there for him between emotions and values? What kind of rationality or irrationality is he prepared to give to emotions? Davidson’s essay on Hume’s theory of pride (1976) suggests an answer to the basis problem, although the article presents itself as a commentary on Hume’s doctrines. Hume, Davidson tells us, says that the object of an emotion is a proposition: Hume’s account of pride is best suited to what may be called propositional pride—pride described by sentences like, ‘She was proud that she had been elected president.’ Hume more

1 I take the “attitudinal view” of Deonna and Teroni (2012) to be intermediary between a perceptual

and a fittingness view.

Davidson on Emotions and Values

81

often speaks of being proud of something—a son, a house, an ability, an accomplishment— but it is clear from his analysis that cases of being proud of something (or taking pride in something, or being proud to do something) reduce to, or are based on, propositional pride. If Hume’s theory is to cope with the other indirect passions, a propositional form must be found for each of them. (Davidson 1980: 277–278)

Davidson seems to endorse Hume’s cognitivism about emotion: emotions are based on belief and judgment. But as Green (2013: 507–508) points out, there are two versions of this view: a robust or strong one, according to which an emotion is not only based on a belief or judgment, but consists in a belief (thus to fear that the dog is dangerous is to believe that it is dangerous), and a modest one, according to which the emotion at least involves a belief or a proposition, without being just a belief. On the modest view there is a belief on which the emotion is based, but the belief does not exhaust the nature of the emotion. Some kind of attitude, presumably affective, must also occur about the proposition which is the content of the belief. But then what is the relation between the belief and the emotion, for instance between anger and the belief which accompanies it? It is, according to Hume as read by Davidson, a causal relation: the belief that I have been treated unjustly is the cause of my being angry. To be proud of one’s having a beautiful house is to believe that one has a beautiful house, which causes the pride.2 But the causal relation between the belief and the emotion has to be channeled through an attitude, the attitude of approval or esteem, on the part of the subject of the emotion, of others who can share this emotion: ... the basic structure of pride and its etiology as Hume saw them is clear: the cause consists, first, of a belief, concerning oneself, that one has a certain trait; and second of an attitude of approbation or esteem for anyone who has that trait. Davidson (1980: 284, Davidson 1980: 284, Green 2013: 509)

Although Davidson’s article is only one in his work commenting on Hume’s views, it seems clear that he endorses such a cognitivist conception, which is perfectly consonant with his view of reasons as causes (Davidson 1963); the reason why the man is proud is that he has a belief which causes his pride, together with the attitude of approval. The causal theory of emotions, according to Davidson, does not merely say that an emotion is a belief. Rather it says that an emotion is a mental attitude which is caused in a certain way by a belief. Interestingly, in his essay on Spinoza’s theory of affects, Davidson ascribes the same causal theory of emotions to Spinoza, on the example of self-esteem: 2 “Hume’s

theory, more or less as he gives it: the cause of pride is a conjunction of the idea of a house, say, and a quality (beauty). The quality causes the separate and pleasant passion, which under the right conditions causes (by association) the similar pleasant passion of pride. The passion of pride itself always causes the idea of self to appear, and this idea must be related (causally, by association) to the idea of the object (the house) on which the quality is placed. In short, ‘That cause, which excites the passion [pride], is related to the object [self], which nature has attributed to the passion; the sensation, which the cause separately produces, is related to the sensation of the passion: From this double relation of ideas and impressions, the passion is derived’ (Davidson 1976,1980: 286).

82

P. Engel “What Spinoza calls “self-esteem” provides a better characterization of pride, since pride can, after all, be born of a correct estimate of one’s worth. Self-esteem, Spinoza says, is joy arising from thinking of our power of acting or of our actions ([Ethics] IIIP55S). Neither Hume nor Spinoza allows a separate judgment that having a certain strength is estimable; rather, to be caused pleasure (approbation) by the belief that one has a certain ability or strength is to value that ability or strength (Davidson 1993: 310–11).

Davidson’s causal and cognitive theory of emotions suggests prima facie answers to each of the main problems to a theory of emotion. First, because emotions involve beliefs and propositional attitudes, they are intentional mental states. But what kind of states are they? Because they have an intentional content, they cannot be merely behavioral dispositions. But since emotions are based on beliefs, and since beliefs are, at least in part, dispositions, emotions are, at least in part, dispositions. To be proud of having a beautiful house is to be disposed to answer queries about one’s house, and to act accordingly. An emotion, though, is not merely dispositional: there are indeed emotional dispositions, such as irascibility or cowardice, but emotions are most of the time occurrent episodes. In this sense they are not moods or feelings, although they can be associated to feelings. It is, as we can see from the quote just given about Spinoza’s causal theory, a certain kind of affective state, involving a valuing, which is caused by a belief. Second, Davidson’s causal theory of emotions suggests an answer to the value problem. The valuing is itself a certain kind of attitude. The emotion consists in this causal structure. That still does not tell us whether the valuing involves a relation to a separate entity, a value, or whether the value just consists in the valuing. Third, the causal theory gives us the basic element for Davidson’s answer to the rationality problem: since emotions are caused by beliefs, they can, through the beliefs, be reasons and have reasons. In this sense too they are, as intentional states, susceptible of being rational. An emotion is rational if the subject which has it can see it as rational, and if it can also be interpreted as such by others. Here the constraints are just those which weigh on actions and beliefs in Davidson’s conception of interpretation. A subject has authority over which emotions he has (one is a better judge of one’s anger that others) but he is not for that infallible (one can hate someone without realizing it) (Green 2013: 514). The fact that emotions are rational at least in the sense that they are interpretable and liable to the standards of rationality which are necessary for interpretation does not mean that all emotions are rational. Indeed they can be irrational, as cases of akrasia, self-deception and other episodes amply show. Davidson has often dealt with this issue, which is actually central for his causal conception of reasons: if an agent believes himself to have the best reasons to do something, why does he not do it? If an agent has good evidence for a belief that p, how can he come up with the belief that not p? In these cases of irrational behavior and belief, as Davidson notes, rationality and causality fall apart: a reason ought to be a rational cause, but with akrasia and self-deception the best reasons fail to be the right causes (Davidson 1982). Emotions also play an important role here, but not as rationalizers of action. They play a role in understanding why some actions, such as those made out of weakness of the will, or some beliefs, such as those acquired through wishful thinking or self-deception, can fail to be rational.

Davidson on Emotions and Values

83

However, these answers to the three problems are only prima facie, because we still have to understand, within the causal structure in which an emotion consists, in what sense an emotion can involve a belief, without being itself a belief, how this causal structure can give rise to a valuing, and how it can respond to reasons.

4 Davidson on Emotions and Objective Values Although Davidson’s remarks on emotions are often unsystematic, there is at least one issue of which he gives an explicit treatment: valuing and the problem of the reality of values. This, I shall try to show, gives us an answer to the main issues that are raised by a theory of emotion from Davidson’s perspective. Let us start with the value problem. It can be formulated in the following way: what is it for an emotion to be correct? In what sense can we talk of the truth of an emotion? The answer seems to be that it must in some sense be true to the facts that the emotion is a reaction to. It must also in some sense be justified, in the sense that there must be a reason, or some reasons for the emotion, which are themselves good reasons, and such that the emotion is appropriate. But how can we ensure the conditions of correctness of appropriateness of emotions? Some emotions are inappropriate or incorrect because their objects do not exist. Thus I may be afraid of a perfectly inoffensive spider or be angry at you for no reason. Davidson was familiar with this problem since his dissertation on the Philebus (1949) where Plato argues that some pleasures might be false. If we formulate this problem in terms of his causal theory, the question for Davidson is twofold: first if an emotion is caused by a belief, in what sense is it based, in the sense of justified, by the belief? Second, can it be justified if the belief is false? The answer to the latter question is easy: if the belief from which the emotion originates—say the belief that this dog is attacking me causing my fear of the dog—is false—the dog rushes to me affectionately—is my emotion inappropriate? Maybe it’s not: rushing dogs can be dangerous. Some emotions, however, are factive: thus to be disgusted or to be horrified is to be disgusted or horrified at something which is there (Gordon 1987). The answer to the first question is more complex. One’s pride in having a beautiful house is caused by the belief that one has a beautiful house. But even if the belief is true, does that make the emotion of pride correct and justified? To answer that question it is not enough to consider the belief upon which the emotion originates, but also whether the emotion is associated to a valuing which is objectively correct. But how does the emotion, which is a subjective feeling, related to the value? How can the emotion, so to say, track the value? There are two kinds of answer to this question to which Davidson is not attracted. The first is the Humean one. One could expect that Davidson would follow Hume, on the basis of his cognitive theory of emotions: the emotion is not related to a value, because there are no such things as values, although there are valuings, attitudes of ascribing values to things or states of affairs. Humeans take these attitudes to be based essentially on desires and motivations. If there are values, these are mere projections

84

P. Engel

out of our attitudes of appraisal. This entails that judgments about value are neither true nor false, and that our emotions cannot be objectively correct. But Davidson does not follow this Humean path. On the contrary he takes values to be objective, and thinks of himself as a realist about values. Our feelings and emotions are objectively correct because our judgments are (Davidson 1984, 1995). Anti-realist theories of value cannot account for this correctness. The other kind of answer is the perceptual realist one: emotions track values because they are perceptions of values, understood as some kind of real entity external to the sensible world, in which the emotion puts us into some relation. According to the terminology of formal objects of emotion presented above, the objects of emotions are evaluative properties or values. From this is natural to think that an emotion puts us in some sort of contact, perceptual or experiential, with a value, as the formal object of the emotion, taken as a real object out in the world, and that the emotion is correct if the perception on which it is based fits the formal object.3 The perceptual theory does not take the content of the emotion to be a propositional object. It takes it as the object of an experience, which may not involve any concept of judgment. And it takes values to be a certain kind of real entity, autonomous from agents, to which their perceptual experience makes them sensitive. Davidson rejects both the Humean and the perceptual views. He rejects an antirealist theory of values and takes these to be real and objective, but not in an ontological sense. The objectivity of values in not a matter of placing, within the natural world or in some Platonic realm, a certain kind of entity. It is a matter of being related, in a certain way, to an objective world of events: Objectivity depends not on the location of an attributed property, or its supposed conceptual tie to human sensibilities; it depends on there being a systematic relationship between the attitude-causing properties of things and events, and the attitudes they cause. What makes our judgments of the “descriptive” properties of things true or false is the fact that the same properties tend to cause the same beliefs in different observers, and when observers differ, we assume there is an explanation. This is not just a platitude, it’s a tautology, one whose truth is ensured by how we interpret people’s beliefs. My thesis is that the same holds for moral values. (Davidson 1994, 2004: 46)

His argument is not ontological, but epistemological: once we understand clearly how we can ascribe evaluative attitudes to people on the basis of their evaluative judgments, we shall be able to conclude that these attitudes are bound to track objective values. Davidson invites us, as he does in many other contexts, to start from the necessary features of interpretation. The familiar claims are the following4 : 1. Evidential basis: the task of interpretation is to ascribe to agents propositional attitudes, such as beliefs, desires, intentions and preferences, which have certain contents. Interpretation has to start from publicly observable features of agents and of their environment and must rest on an evidential basis. 3 This

is indeed a very rough and inaccurate presentation of the perceptual view. There are actually a number of views of this sort. See Mulligan (1998; Tappolet 2016) for the stronger versions. 4 Here I follow the very clear presentations by R. H. Myers in Myers and Verheggen (2016) (see also Lillehammer 2007). For a more detailed examination of the argument see Engel (2017).

Davidson on Emotions and Values

85

2. Holism: the contents of someone’s attitudes necessarily depend on the contents of many other attitudes. 3. Charity: given that the contents of attitudes are necessarily interconnected, one must presuppose that there is at least a minimal coherence between these contents, and ascription of coherent sets of content cannot be made unless the interpreter presupposes that the agent shares a large amount of true beliefs with him. 4. If agents are to be interpretable, they not only must share attitudes and contents which are largely similar to ours, but also largely correct. Let us call this the argument from interpretation. Any reader of Davidson will recognize the affinity between this argument and the one which he uses to refute radical skepticism: since interpretation presupposes a massive degree of agreement on beliefs which are largely correct, these beliefs have to be about an objective world (Davidson 1981). Davidson applies this reasoning to desires, then to values, expanding (iv) into. 5. If agents are to be interpretable, they share values which are largely similar to ours, correct, and objective. If the argument from interpretation to objective values is supposed to work, it must not simply bear on particular desires and other attitudes, because the holistic pattern that these might display does not guarantee that agents converge on the same desires and values. The problem here is very close to the problem known in utility theory as the problem of interpersonal comparisons of utilities (Davidson 1986). The desires in question must not be transitory, such as the particular desire for an ice cream at a particular time. They must be long standing and, as Davidson says “enlightened”: they must actually be beliefs about what is good in general, hence normative desires, about what is desirable: To what extent do these considerations apply to the evaluative attitudes? It is possible, I think, to show that the justified attribution of values to someone else provides a basis for judgments of comparisons of value, what is called the interpersonal comparison of values. But the comparability of values does not in itself imply agreed-on standards, much less that we can legitimately treat value judgments as true or false. Now I want to go on to suggest that we should expect enlightened values—the reasons we would have for valuing and acting if we had all the (non-evaluative) facts straight—to converge; we should expect people who are enlightened and fully understand one another to agree on their basic values. An appreciation of what makes for such convergence or agreement also shows that value judgments are true or false in much the way our factual judgments are. (1994, 2004:49)

The values on which we must expect a convergence must not be simply basic values, such as the value of basic human needs such as food, sex and safety, but also enlightened values, such as justice, equality or freedom. But how can we be sure that there is a convergence on such values? Davidson requires that such convergence can be reached only when people, within a community, understand each other. But do the minimal conditions of interpretation, together with the principles of charity and of rationality which accompany them, suffice to make agents “believer[s] of truth, and lover[s] of the good” (Davidson 1969; 1980: 222)? There must also be “shared criteria”:

86

P. Engel [I]f I am right, disputes over values (as in the case of other disputes) can be genuine only when there are shared criteria in the light of which there is an answer to the question who is right... When we find a difference inexplicable, that is, not due to ignorance or confusion, the difference is not genuine... The importance of a background of shared beliefs and values is that such a background allows us to make sense of the idea of a common standard of right and wrong, true and false. (Davidson 1995, 2004, 50–51).

But how do we know these criteria, and how they are shared? As Lillehammer (2007: 214–5) has remarked, the holistic strategy does not guarantee that “there is a uniquely fixed and determinate set of particular features of the world the positive or negative evaluation of which all agents must share if they understand each other and are otherwise well informed about the (non-evaluative) facts”. When considering this difficulty, Davidson comes back to the requirement of charity and good interpretation: when we find uninterpretable differences, this is a sign of bad interpretation (ibid). So when we ask how interpretation can actually converge on shared values, we are told that it must converge.5 Although Davidson’s argument from interpretation is an argument about the existence of objective practical values, he nowhere proposes a similar argument for the objectivity of epistemic values. Nevertheless, he sometimes gives hints at what he would say on this issue. There is indeed a wide debate about the nature of epistemic values and norms, but it need not concern us here. Let us only take it for granted that truth, rationality, justification and knowledge are plausible candidates.6 Our question is: could Davidson give a parallel argument about epistemic values and norms (whatever they are)? Such an argument would try to justify the objective status of values such as truth, knowledge, rationality and justification, through our judgments about these values, and it should show that we can converge on these values and norms. But such an argument makes no sense for Davidson: the truth of our beliefs, their rationality and their justifications, are not values or norms, which we could posit independently and which could be objects judgments. They are presuppositions or principles of interpretation. Rationality and charity (the sharing of true and coherent beliefs) are objective, in the sense that without these normative principles, we could not understand others or ascribe to them beliefs. So there is no need of an independent argument to this effect. But here again it is not clear that Davidson’s view yields a sufficiently robust, or sufficiently realist, conception of these values and norms. The fact that rationality and charity are what makes interpretation possible does no tell us how we can converge on these as values, or why knowledge can be a stronger value than true belief. And if the convergence is not guaranteed, we cannot say that

5 Myers (2012, 2013) (see also Myers and Veregghen 2016) gives a defense of the holistic strategy, but

concedes that Davidson’s realism about values cannot be stronger than the form of value objectivism of the kind for which contractualists like Scanlon (2014) can reach. 6 When he examines the suggestion that truth might be an epistemic norm, Davidson answers clearly that truth is neither a norm nor a goal (Davidson 1998). He is right: truth is not a value or a norm (Engel 2000). Epistemic norms and values are never self-standing, they are norms and values with respect to belief and knowledge: believing the truth and knowing the truth are at least prima facie values and norms, just as justification. This is disputable, but this question need not concern us here.

Davidson on Emotions and Values

87

emotions, and the attitudes on which they are based, are correct or not, and whether they track the right values.7

5 Fitting Attitudes and Idealization A striking feature of the argument from interpretation to the objectivity of values is that, unlike the Humean or the perceptual model, it does not talk of emotions and of their relation to values. It talks about desires and their relation to judgments about values. This is a consequence of the revised cognitive model that Davidson subscribes to: values are not entities with which we could be in direct contact through certain kinds of experiences, feelings or emotions. For Davidson our relation to values is necessarily indirect, and goes through our interpretation of others and our sharing various attitudes and reactions with them. In this respect Davidson keeps an important feature of the cognitive theory which he ascribed to Hume: an emotion is caused by a belief and by the attitude of approval of those who share the emotion. This intersubjective character is essential, and makes for the objectivity of the shared values. But as we saw, it fails to give us an account of the correctness of emotions. Davidson could have proposed an alternative analysis of the relation between emotions and values which could account for the correctness of emotions. This analysis is what is now called the fitting attitude analysis of value, according to which values consist in a certain relation between attitudes which are fitting in response to these values. On such a view, values are neither the expressions of our attitudes nor independent realities which could be perceived. Evaluative concepts have to be explained in terms of fitting or appropriate emotions. On such views, values are response dependent, as they are for Humeans, but they are neither subjective nor projections out of our attitudes. They are based on our judgements about the correctness, or the reasons that one has to have these attitudes.8 Davidson does not explicitly discuss such neo-Brentanian views, but he was certainly familiar with these from his reading of Kenny’s Action, emotion and will (1963) and from the work of Chisholm, whose views on action he discussed intensively.9 The fitting attitude analysis starts from emotions. It does not say that they can be true or false, but that they are fitting or not, and this fittingness is itself an objective matter. In the terminology presented above, emotions have a formal object, which is a value property. Thus the formal object of fear is the fearable, the formal object of love is the lovable, the formal object of admiration is the admirable. But 7 For

an argument to the effect that Davidson must adopt a stronger notion of normativity in the epistemic domain, see Engel (2008). 8 There are actually a number of versions. For presentations see (Chisholm 1969; Mulligan 1998; Danielsson and Olson 2007; Tappolet 2016, Chap. 3). It is sometimes associated to the «buck passing» account of values: the buck is passed to reasons. Scanlon (1998, 2014), Skorupski (2010) are the main contemporary defenders of such views. 9 See Brentano (1889), Chisholm (1969), and Davidson ‘s essays in reply to Chisholm in Davidson (1980).

88

P. Engel

how can the view be made something other than a tautology? One can fear objects which are not fearable (little innocuous spiders) or which do not exist (monsters). How can the view yield fittingness to objective values? Davidson does not talk of the fitness of attitudes and emotions, but he often characterizes the causal link between action, belief and reason in terms of the notion of appropriateness. An action is caused “in an appropriate way” for a given reason if the action fits the reason (the cases where it does not fit are cases of “deviant causal” chains”). Interestingly most of Davidson’s examples of deviant causal chains are cases where an emotion intervenes in the causal sequence leading from intentions and reasons for an action to the action. Thus the climber who is so nervous that he releases the rope holding his partner out of the desire to free himself from his weight and his belief that he could do this by releasing the rope is under the causal influence of an emotion. Famously Davidson tells us that he despairs of spelling out “the way in which attitudes cause actions if they are to rationalize the action” (Davidson 1980: 79). He might say the same thing about the fitness of emotion to value. But that does not bring his view closer to the fittingness account. Davidson actually does not accept the idea that values are the formal objects of the emotions, for two reasons. The first can be recovered from his dispute with Chisholm and Kenny. When he deals with the formal object of actions, Kenny (1963) tells us that one encounters the problem of “variable polyadicity” of action verbs: how can they have a single formal object, given that actions are relative to all sorts of circumstances, such as when, how, where, by whom the action was done? Davidson’s answer to this problem in “The logical form of action sentences” (1967) is well known: he proposes to add to action predicates argument places for events, and to construe action sentences as quantifying over them. Thus he breaks down the very notion of a formal object into a core property (expressed by the action verb), events and the properties of these events. In accepting that the events which make up an action are real entities in the world, Davidson rejected the view that actions could have a formal object. One can presume that Davidson would have rejected in the same way the notion of a formal object of emotions, although his analysis still involves a commitment to properties. But his later writings, as I have tried to show, do not entail any commitment to value properties. The second reason has to do with his rejection of the foundationalist model of justification. This model is much present in the perceptual account of emotions as experiences of value. Crude versions of this account say that emotions are direct perceptions of evaluative properties. But these crude versions encounter many difficulties, such as these: the phenomenology of emotions is not the same as the phenomenology of perceptions, there can be mixed and conflicting emotions, and emotions seem to be liable to rational assessment in a way in which perceptions are not (Tappolet 2016, Chap. 1). But even if one does not adopt a perceptual model, and if one accepts a version of the fitting attitude analysis, one has to accept at least the principle that when one has an emotional response to a perceived object or event, then it thereby seems to you that that object or event possesses some evaluative property. The justification of the emotion, its being correct, supervenes on the content of the mental states

Davidson on Emotions and Values

89

on which it is based, that is its cognitive basis.10 This is a form of prima facie and immediate justification: it seems to you, from your perceiving of this salivating dog, that it is dangerous, and thereby you experience fear of the dog.11 The perception justifies the evaluative judgment in which consists the emotion. It is important here to distinguish this relation of justification between the cognitive basis and the evaluative judgment from the causal relation which the cognitive theory postulates between the perception and the judgment. When Davidson adheres, as we saw above, to a modest version of the cognitive theory, he does not take the relation to be one of justification. However the idea that emotional justification comes from the awareness of a cognitive basis has been strongly criticized: not only can one be prima facie justified and wrong, but also there can be large differences in the ways cognitive bases and emotions represent objects and properties, so cognitive bases are not sufficient to justify emotional responses.12 The relationship between emotional experience and evaluative beliefs need be neither direct nor foundational in the sense suggested by the simple perceptual model or in the sense of the model of prima facie justification. It can be holistic, and such that the emotional experience and its relation to values is further confirmed by related beliefs.13 The fit between the perceived situation, the emotion and the issued value judgments may be more a matter of coherence than a matter of perceptual basis, and the correctness of emotions need not be based on some mysterious capacity of grasping the values within the emotional experiences. As de Sousa (2005) argues, the appearance of tautology of the fitting attitude analysis (the formal object of love is the lovable, of fear the fearable) cannot be dispelled “because the attainment of success for emotions—the actual fit between the object or target of the emotion and its formal object—depends on a vast holistic network of factors that transcend my actual responses”. Clearly Davidson has more sympathy with the idea that the justification of emotions, and their correctness, are more a matter of holistic relations between experiences, beliefs and facts than a matter of immediate justification. As we saw, Davidson, as any Quinean, does not like the notion of intentional object of attitudes or emotions. He dislikes any idea that there could be objects “present to the mind” (Davidson 1989). When he spells out his account of the relation between emotions and desire states to values, his line consists in explaining them through causal relations between facts, speakers, and their attitudes. But, as we saw, the main difficulty which his account has to face is: what kinds of facts can secure the fit between emotional attitude and value? Facts about human nature? Biological facts? Social facts? And how can we go from these facts to commitments to values?

10 See

Deonna and Teroni (2012), Goldie (2000), Mulligan (1998). Basically this view is close to Brentano’s view that emotions “manifest themselves to be correct” (Chisholm 1969). 11 This is indeed just the reverse of the James-Lange theory. 12 For this kind of criticism of the prima facie view, sometimes called “emotional dogmatism”, see Brogaard and Chudnoff (2016), Echeverri (2019). 13 See in particular Brady (2014), Roesser and Todd (2014).

90

P. Engel

So, Davidson could not be a fitting attitude theorist of the relation between emotions and values. However, there is a version of this theory which might be understood as a development of a Davidsonian idea. Our attitudes, such as belief, desire, hope, or regret, but also our emotions, such as love, hate and disgust, are all associated to various presuppositions about the kinds of judgements, inferential relations, causal profiles, and strategies that they involve. In particular emotions are all the more complex that they are tied to complex presuppositions (think for instance of mixed emotions). These presuppositions are in many ways contingent and can vary, depending on social, historical, or cultural factors. They are in various ways tied to our nature, and can depend on all kinds of circumstances in individuals: for instance some people are afraid of tiny spiders, many believe or desire weird things. The holistic structure of the justification of emotions is in large part the product of this diversity and of the diversity of these factors. These contingencies and links, however, do not affect the essence or nature of the attitudes. Each attitude has an ideal profile, one which it ought to have. It is particularly the case for belief: it is an attitude which is associated, implicitly or not, in a believer, to what he takes himself to believe, to what he considers that he ought to believe, which is constrained by the rational norms of interpretation. Davidson said in “Mental events” (1980: 223) that the domain of the mental is governed by “a constitutive ideal of rationality”. An interpreter has to start from this ideal profile, and the subjects of emotions and other attitudes could not be interpreted if they did not aim at this ideal profile, even when they are far from instantiating it in their actions and beliefs. The correctness of an emotion is the fit between the causal profile and the ideal attitudinal profile. The value involved is what ideally the emotion would be an approval of. The same idea can be found in some partisans of the fitting attitude view of emotions and values. Thus J. Findlay, an anglophone philosopher who worked in the Brentanian tradition, talks in reference to Bishop Butler’s notion of moral sense, of Butlerian attitudes, those which are idealised to an impartial point of view: one abstracts away from personal biases and pretends to take policies which are reasonable and turns one’s back on a certain range of facts (Findlay 1954). We can understand Davidson’s conception of emotion and values in the same way: emotions are correct, and track objective values, when we put them within the right set of relations. The attitudes that are correct are not those that are made so by a certain range of natural facts, but those that one ideally would reach if one turned one’s back on those facts, and tried to adopt an idealized point of view. This form of idealization is nothing different from the objective standpoint on values and norms, which Davidson meant to be reachable from his interpretation argument.

6 Conclusion There are many more aspects of Davidson’s views on emotion that I have not discussed in this article: in particular his account of how emotions play a role in the processes which lead to irrational behavior and belief. His conception of emotions

Davidson on Emotions and Values

91

and of their justification is, within the contemporary literature of emotions, very original. He stands in between the two main strands of analysis which still dominate the field today: the Humean and the Brentanian one. To the first he owes a causal theory of emotions. With respect to the second, he comes close to a fitting attitude view. But his holism and his externalism about the mind put him on a distinctive path.

References Brentano, F. (1889). Vom Ursprung sittlicher Erkenntnis, Leipzig. (The origin of our knowledge of right and wrong. London: Routledge.: Engl R.Chisholm & E.S Chneewind, Trans.). Brogaard, B., & Chudnoff, E. (2016). Against emotional dogmatism. Philosophical Issues, 26, 59–76. Brady, M. (2014) Emotional insight. Oxford: Oxford University Press. Cavell, M. (1993). The psychoanalytic mind, from Freud to philosophy. Cambridge, Mass.: Harvard University Press. Chisholm, R. (1969). Brentano’s theory of correct and incorrect emotions. Revue Internationale de Philosophie, 20, 78(4), 395–415. Clore, G., & Huntsinger, R. (2007). How emotions inform judgment and regulate thought. Trends in Cognitive Sciences, 11(9), 393–399. Davidson, D. (1949). Plato’s Philebus. New York: Garland Publishing Co. 1990. Davidson, D. (1957). (with P. Suppes and S. Siegel) Decision making: An experimental approach. Stanford, Calif.: Stanford University Press (New edition. Westport, Conn.: Greenwood Press. 1977). Davidson, D. (1963). Actions, reason and causes. In Davidson 1980. Davidson, D. (1967). How is weakness of the will possible? In Davidson 1980. Davidson, D. (1976). Hume’s cognitive theory of pride. In Davidson 1980. Davidson, D. (1980). Essays on actions and events. Oxford: Oxford University Press. Davidson, D. (1981). A coherence theory of belief and knowledge. In Davidson 2001. Davidson, D. (1982). Paradoxes of irrationality. In Davidson 2004. Davidson, D. (1984). Expressing evaluations. In Davidson 2004. Davidson, D. (1989). What is present to the mind. In Davidson 2001. Davidson, D. (1995). The objectivity of values. In Davidson 2004. Davidson, D. (1999). Spinoza’s causal theory of the affects. In Y. Yovel (Ed.), Desire and affect: Spinoza as psychologist. New York: Libble Room Press. Davidson, D. (2001). Subjective, objective, intersubjective. Oxford: Oxford University Press. Davidson, D. (2004). Problems of rationality. Oxford: Oxford University Press. Danielsson, S., & Olson, J. (2007). Brentano and the buck-passers. Mind, 116, 511–522. Deonna, J., & Teroni, F. (2012). The emotions. London: Routledge. De Sousa, R. (1986). The rationality of emotions. Cambridge, Mass: MIT Press. De Sousa, R. (2005). Emotional truth. Proceedings of the Aristotelian society, 76(76), 265–275. Echeverri, S. (2019). Emotional justification. Philosophy and Phenomenal Research, 98(3), 541– 566. Engel, P. (2000). Is truth a norm? In P. Kotatko, P. Pagin & G. Segal (Eds.), Interpreting Davidson (pp. 37–51). Stanford, Calif.: CSLI. Engel, P. (2008). Davidson on epistemic norms. In C. Amoretti et N. Vassalo (Eds.), Knowledge, language and interpretation: On the philosophy of Donald Davidson (pp. 123–146). Munich: Ontos Verlag.

92

P. Engel

Engel, P. (2017). Davidson on objective values and reasons. Argumenta, 5. https://www.argume nta.org/article/davidson-objectivity-values-reasons-special-issue/. Engel, P. (2018). Review of Myers and Verheggen. Notre Dame reviews in philosophy. https://ndpr. nd.edu/news/donald-davidsons-triangulation-argument-a-philosophical-inquiry/. Findlay, J. N. (1954). The justification of attitudes. Mind. LXIII, 250, 145–161. Findlay, J. N. (1963). Meinong’s theory of objects and values. Oxford: Oxford University Press. Fridja, N. (1986). The emotions. Cambridge: Cambridge University Press. Gordon, R. (1987). The structure of emotions. Cambridge: Cambridge University Press. Goldie, P. (2000). The emotions: A philosophical exploration. Oxford: Clarendon Press. Green, M. (2013). The rationality of emotions. In E. Le Pore, & K. Ludwig (Eds.), The Blackwell companion to Donald Davidson (pp. 506–518). Oxford: Blackwell. James, W. (1889). The principles of psychology. Revised ed. New York: Dover Publications. 1950. Kenny, A. (1963). Action, emotion and will. London: Routledge (second ed 2003). Le Doux, J. (1996). The emotional brain. New York: Wendfield and Nicolson. Lyons, W. (1980). Emotion. Cambridge: Cambridge University Press. Mulligan, K. (1998). From appropriate emotions to values. The Monist, 81, 161–188. Myers, R. H. (2004). Finding value in Davidson. Canadian Journal of Philosophy, 34, 107–136. Myers, R. H. (2012). Desires and normative truths: A holist’s response to the sceptics. Mind, 121(482), 375–406. Myers, R. H. (2013). Interpretation and value. In E. Lepore & K. Ludwig, (Eds.), A companion to Donald Davidson. New York: Wiley-Blackwell. Myers, R., & Verheggen, C. (2016). Donald Davidson’s Triangulation Argument: A philosophical inquiry. New York: Routledge. Nussbaum, M. (2001). Upheavals of thought: The intelligence of emotions. Cambridge: Cambridge University Press. Pitcher, G. (1965). Emotion. Mind, 74(July), 326–346. Ryle, G. (1949). The concept of mind. London: Hutchinson. Roesser, S., & Todd, C. (Eds.). (2014). Emotion and value. Oxford: Oxford University Press. Scheler, M. (1913). Formalism in ethics and non-formal ethics of values (1913) (M. S. Frings & R. L. Funk, Trans.). Evanston: Northwestern University Press. Scanlon, T. (1998). What we owe to each other. Cambridge, Mass.: Harvard University Press. Scanlon, T. (2014). On being realistic about reasons. Oxford: Oxford University Press. Solomon, R. C. (1984). The passions: The myth and nature of human emotions. New York: Doubleday. Skorupski, J. (2010). The domain of reasons. Oxford: Oxford Universty Press. Tappolet, C. (2016). Emotions, values and agency. Oxford: Oxford University Press. Teroni, F. (2007). Emotions and formal objects. Dialectica, 61(3), 395–415. Tversky A., & Slovic P. (Eds.). (1982). Judgment under uncertainty: Heuristics and biases. Cambridge: Cambridge University Press. Thalberg, I. (1977). Perception, emotion and action: A component approach. New Haven, Conn.: Yale University Press. Wilson, J. N. (1975). Emotion and object. Cambridge: Cambridge University Press.

Pleasure Makes Pro-attitudes Jeremias Koh and Neil Sinhababu

Abstract Donald Davidson famously argues that when a person acts for a reason, we can characterize that person as having some sort of pro-attitude toward actions of a certain kind and believing that their action is of that kind. If acting for a reason must be caused by pro-attitudes, then the right account of pro-attitudes will help us correctly identify cases of action for reason. In this paper, we examine what it takes for a mental state to be a pro-attitude. First, we argue that having any mental state that systematically causes behavior when combined with an appropriately connected belief count as a pro-attitude is overly permissive. We do so by presenting an example of someone whose complex system of nervous tics mimics the causal structure of a pro-attitude, but where the resulting behavior does not count as action. Next, we argue that requiring beliefs about value or treating pro-attitudes as perceptions of value is overly restrictive. Such views fail to account for akrasia or attribute unprecedented functional properties to perceptual states. Finally, we propose an account of proattitudes that avoids these pitfalls: In addition to causing behavior, pro-attitudes cause more pleasure when one thinks about attaining their object than when one thinks about failing to attain their objects. Keywords Desire · Pro-attitudes · Motivation · Akrasia · Action · Pleasure

1 Davidson, Pro-attitudes, and the Explanation of Action Donald Davidson’s classic “Actions, Reasons, and Causes” (1963) argues that when a person acts for a reason, we can characterize that person as “(a) having some This chapter is in its final form, and it is not submitted for publication anywhere else. J. Koh (B) · N. Sinhababu Department of Philosophy, National University of Singapore, 3 Arts Link, Block AS3, #05-22, Singapore 117570, Singapore e-mail: [email protected] N. Sinhababu e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_6

93

94

J. Koh and N. Sinhababu

sort of pro-attitude toward actions of a certain kind and (b) believing (or knowing, perceiving, noticing, remembering) that their action is of that kind” (685). Davidson treats this combination of a pro-attitude and a belief as causally sufficient for acting for a reason. Explanations of action along these lines have many historical and presentday proponents, including Hume (1739/2000), Arpaly and Schroeder (2013), and Sinhababu (2017). In this paper, we examine what it takes for a mental state to be a pro-attitude. If action for a reason must be caused by pro-attitudes, the right account of proattitudes will help us correctly identify cases of action for a reason. If all mental states that interact systematically with beliefs to cause actions are counted as pro-attitudes, it will be too easy for a mental state to count as a pro-attitude. Various types of nonintentional behavior will then qualify as action for a reason. The opposite problem arises if we require that agents believe that their actions are valuable or perceive value in their actions, as people sometimes do things that they regard as valueless or bad, and human psychology may not even allow such representational states to motivate action. So views requiring these types of beliefs or perceptions will treat various actions that are done for reasons as if they were not done for reasons. We suggest a new solution that we hope will avoid undercounting or overcounting actions done for reasons. In addition to causing action in combination with a belief, the pro-attitude must include a disposition to feel more pleasure in thinking of attaining one’s intrinsic end than in thinking of not attaining it. As desire has this feature, our view is friendly to desire-belief accounts of acting for a reason. Section 2 explains why a pro-attitude cannot merely be any state that systematically causes behavior when combined with an appropriately connected belief. Section 3 explains why requiring beliefs about value for pro-attitudes or treating pro-attitudes as perceptions of value are excessively restrictive. Section 4 presents our solution: In addition to causing behavior, pro-attitudes cause more pleasure when one thinks about attaining their object than about failing to attain their objects.

2 Causing Behavior in Combination with a Belief Is Not Enough Systematically causing behavior in combination with a belief is not enough for a mental state to be considered a pro-attitude. We will demonstrate this with a simple counterexample and then distinguish the point of the counterexample from superficially similar ones that have attracted considerable interest. Consider the following case: Nellie sometimes believes that touching her nose would make everyone mad. When she thinks of making everyone mad, she gets nervous. When she gets nervous, she involuntarily performs the last behavior she thought of, which in this case is touching her nose. These psychological dispositions are stable and reliably cause Nellie to touch her nose and do other things that she

Pleasure Makes Pro-attitudes

95

thinks would make everyone mad—even though she regards her actions and their effects only with anxiety, and draws no satisfaction from them. If all it takes for a mental state to be a pro-attitude is to cause behavior of some kind in combination with a belief that a behavior is of that kind, Nellie’s dispositions to get nervous and touch her nose will constitute a pro-attitude toward making everyone mad, and perhaps also an instrumental pro-attitude toward touching her nose. This is because her dispositions systematically cause her to touch her nose, when combined with her belief that touching her nose would be an action that makes everyone mad. But Nellie clearly does not have a pro-attitude toward making everyone mad. Her nervousness gives the case additional force, as she seems to have a negative attitude toward making everyone mad, rather than a pro-attitude. As her case shows, systematically causing behavior when combined with a belief is not enough to make a mental state into a pro-attitude. Cases like Nellie’s are similar to cases of deviant causal chains, though the issues they address are different. Consider this classic example of a deviant causal chain from Davidson (1973): “A climber might want to rid himself of the weight and danger of holding another man on a rope, and he might know that by loosening his hold on the rope he could rid himself of the weight and danger. This belief and want might so unnerve him as to cause him to loosen his hold, and yet it might be the case that he never chose to loosen his hold, nor did he do it intentionally” (79). As Davidson writes, the problem of deviant causal chains arises because “not just any causal connection between rationalizing attitudes and a wanted effect suffices to guarantee that producing the wanted effect was intentional. The causal chain must follow the right sort of route” (78). The similarity between the cases is that Nellie’s touching of her nose does not qualify as intentional, and the causal chain in her case does not follow the right sort of route. However, the issues raised by these examples are different. Cases of deviant causal chains involve mental states already agreed to be pro-attitudes and beliefs, with the question being when their causing a bodily movement counts as intentional action or action for a reason. This paper addresses the question of what makes something a pro-attitude in the first place. So, typical solutions to the problem of causal deviance will not work here. They answer a question that arises only after our question has already been answered. Nellie’s case also resembles that of Quinn’s (1993) radioman, though here as well the issues at stake are fundamentally different. Quinn characterizes radioman as being in a strange functional state that disposes him to turn on radios that he sees to be turned off. In Quinn’s example, this disposition to turn on radios is all there is to radioman’s functional state. He has no further purpose in turning them on, such as hearing music or getting news. Quinn acknowledges that radioman has an odd proattitude toward turning on radios, but denies that this state gives radioman a reason for action. Radioman’s case thus concerns what qualifies as a reason for action. There is room for classifying radioman as having a pro-attitude toward turning radios on, but still not as acting for a reason. (Exactly what to say about radioman is somewhat unclear, because Quinn does not go into much detail about the nature of his mental state. We hope to have provided a clearer example by being more explicit about the nature of Nellie’s psychological dispositions.) One might naturally understand

96

J. Koh and N. Sinhababu

Quinn as presenting a counterexample to the Davidsonian account of acting for a reason. Radioman combines a pro-attitude with a belief about his action, but does not thereby act for a reason, making him a counterexample to the sufficiency of Davidson’s account of acting for a reason. Nellie, however, clearly does not do anything that even counts as intentional action. Nellie’s case points out how important it is for Davidsonian views to have a general account of what pro-attitudes are. The case is not a direct counterexample to any Davidsonian account of the nature of pro-attitudes, because Davidson does not seem to offer such an account. He just tells us what pro-attitudes are by giving examples: “desires, wantings, urges, promptings, and a great variety of moral views, aesthetic principles, economic prejudices, social conventions, and public and private goals and values in so far as these can be interpreted as attitudes of an agent directed toward actions of a certain kind” (686). While this is a helpful list of examples, there is no general rule for what makes something a pro-attitude here. While the last clause might seem to suggest a general account, it does not narrow things down enough to be helpful. If you feel surprised that you managed to lift something heavy, you have an attitude directed toward actions of a certain kind. But this would not seem to be a pro-attitude of the sort Davidson’s picture requires. At some level, believing that your action is of a certain kind is itself an attitude of an agent directed toward actions of a certain kind. After all, belief is a propositional attitude. But Davidson did not mean to treat belief about action as a pro-attitude. He treats beliefs as the things that pro-attitudes interact with, but never tells us that beliefs can be pro-attitudes. All this underscores the point that we need a clearer account of what makes a mental state a pro-attitude. A further remark of Davidson’s might seem to gesture at a general account of pro-attitudes. He writes, “It is not unnatural, in fact, to treat wanting as a genus including all pro-attitudes as species. When we do this and when we know some action is intentional, it is empty to add that the agent wanted to do it.” Here he seems to treat pro-attitudes in general as types of wanting. This might appear to allow him an easy answer to Nellie’s case. Nellie involuntarily touching her nose would not constitute a pro-attitude toward making everyone mad, because she does not want to touch her nose (i.e., she is not touching her nose deliberately). And indeed, it seems reasonable to treat wanting as a general category covering all pro-attitudes. But in our view, this response is still unsatisfactory, because it fails to tell us which essential feature of wanting Nellie’s nervous tendencies lack. The mental state that drives her behavior does not count as wanting. But why not? A fully satisfactory theory of pro-attitudes, or of wanting, will have to make this explicit. Until we have such a theory, a fundamental component of the Davidsonian picture still stands in need of clarification. Over half a century has passed since the publication of “Actions, Reasons, and Causes”, and there have been a number of theoretical advances on topics Davidson was addressing. One that we will assume in our discussion involves the structure of the pro-attitudes and beliefs that interact to motivate action. As we have seen, Davidson invokes pro-attitudes that pick out the kinds of actions that are to be performed and beliefs with content that the actions are of that kind. Contemporary decision

Pleasure Makes Pro-attitudes

97

theorists typically invoke motivational states that favor a particular state of affairs and beliefs with the content that the actions to be performed raise the probability of that state of affairs. For example, Jeffrey (1983) sees desirability as applying first to consequences—the states of affairs created by action—and only derivatively to the acts themselves. We see the decision-theoretic approach as superior. As it is important to our solution that pro-attitudes pick out states of affairs rather than actions, we will provide an independent argument for the decision-theoretic approach before proceeding. As we will argue, invoking pro-attitudes that pick out states of affairs rather than actions provides a unified account of their role in both motivation and instrumental reasoning. Instrumental reasoning is a process by which pro-attitudes create other proattitudes, when combined with means-end beliefs. Such reasoning can operate on attitudes that do not refer to actions. If you desire that your child be happy, the content of your desire does not pick out any of your actions. So your desire does not fit Davidson’s characterization of a pro-attitude. But it generates instrumental desires when combined with means-end beliefs. If you believe that having friends would make your child happy, you can form the instrumental desire that your child have friends. You can form this instrumental desire even if you do not believe that there is any action you can perform to help your child have friends. Instrumental desire formation thus combines desires for states of affairs with means-end beliefs about what would make those states of affairs more probable. The same structure of reasoning motivates action if the means are your action. If you believe that buying your child a present would make your child happy, you will be motivated to buy a present. So the decision-theoretic account unifies motivational reasoning and instrumental reasoning. It is unclear how the structure of reasoning Davidson suggests could generalize to instrumental reasoning. Davidson requires pro-attitudes to make reference to action, which neither the premises nor the conclusions of instrumental reasoning have to do. So that pro-attitudes have the right shape to figure in both instrumental reasoning and motivation, we will depart from Davidson and treat them as picking out states of affairs rather than actions. By itself, this modification to Davidson’s view does not solve the problem that Nellie’s case presents. Jeffrey and his followers themselves require some explanation of why the system of dispositions that leads Nellie to engage in behaviors that make everyone mad does not constitute a pro-attitude toward the state of affairs where everyone is mad. It interacts with means-end beliefs in just the way pro-attitudes are supposed to, on the decision-theoretic view. But the behavior that results clearly is not intentional action. So the problem we aim to solve is afflicts contemporary decision theorists as well as Davidson. Having explained the nature of the problem under discussion, we will now consider some more recent theories about what constitutes a pro-attitude.

98

J. Koh and N. Sinhababu

3 Requiring Beliefs About Value or Treating Pro-attitudes as Perceptions is Too Much Two recently popular views of pro-attitudes add elements beyond a role in causing action. One of these views requires belief that the action being done is good. Another of these requires pro-attitudes to themselves be perceptions of value. These views have the opposite problem of the view we considered above. They require too much for a mental state to be a pro-attitude and fail to correctly identify some cases where we do in fact act for reasons. One initially plausible way of dealing with problem cases like Nellie’s is to require belief that the object of the pro-attitude is good. Stampe (1986) and Quinn (1993) have suggested views along these lines. With this requirement, one could say that since Nellie does not believe that making everyone mad or touching her nose is good, her dispositions do not constitute a pro-attitude toward making everyone mad or touching her nose. This gets the intuitively right result that she is not acting for a reason. Requiring such a belief might seem like an unproblematic requirement to place on intentional action, because we believe that most of the things we pursue when we act for reasons are in fact good. Stampe in fact raises a case like Nellie’s, though he does not explore its consequences for pro-attitude constitution as we have. While distinguishing desire from belief, he notes that a belief does not usually “cause behavior that brings about its own satisfaction, though there are exceptions to that – like the thought, when serving in tennis, that one is going to double-fault. But we take these to be anomalies” (156). We can see how his requirement would similarly deal with his case. The belief that one will double fault may disturb one’s composure so that one does in fact double fault, but even so it does not constitute a pro-attitude. However, insisting on this requirement means saying that it is impossible for people to pursue or even desire things that they believe to be bad. Temptation to engage in behavior that one believes to be bad, in some sense or another, is an ordinary part of human life. Requiring belief that the action is good makes various types of akrasia impossible. While Nellie clearly did not believe that scratching her nose was a good thing to do, one can act while believing that one’s actions are immoral or irrational. Cases of procrastination make this clear. One can procrastinate even while occurrently believing that one is wasting time on something pointless and that one should get work done instead. Davidson himself recognizes this issue, writing that “pro attitudes must not be taken for convictions, however temporary, that every action of a certain kind ought to be performed, is worth performing, or is, all things considered, desirable.” He then provides a counterexample to the necessity of beliefs about value for pro-attitudes, writing that “a man may all his life have a yen, say, to drink a can of paint, without ever, even at the moment he yields, believing it would be worth doing” (686). We will say more about akrasia in the next section of this paper, when discussing our proposed solution. For now, it suffices to say that this way of dealing with problem cases like Nellie’s will not do, because it prevents akratic action from counting as action done for a reason.

Pleasure Makes Pro-attitudes

99

Another way one might try to address problem cases like Nellie’s is to treat proattitudes as perceptions of value (or goodness). Proponents of this strategy include Wallace (1999), Oddie (2005), Tenenbaum (2007), and Schafer (2013). On this view, Nellie’s nervous dispositions would not amount to a pro-attitude toward touching her nose or making everyone mad, because her pro-attitude is not itself a perceptual state that recognizes touching her nose or making everyone mad as valuable. While the previous solution added a further belief as a necessary condition to prevent motivational states like Nellie’s from constituting pro-attitudes, the present solution is to treat the pro-attitudes themselves as being perceptions of value. One attractive feature of perceptualism is that it accounts for the phenomenology of desire. Desires do more than cause us to act. They also give us particular feelings as we consider the objects of desire. These feelings often represent the desired things as valuable in various ways corresponding to our desires. For instance, food looks good to hungry people. When the hungry people had enough to eat and their desire for food goes away, the food does not look as appealing to them. Desire thus shapes our perception of its object, making us see its object as valuable. In accounting for the phenomenology of desire, perceptualism represents an advance over theories that treat it as purely motivational and neglect its other psychological effects. Some philosophers argue that it is impossible for pro-attitudes to have the representational nature of belief or perception, as perceptualism requires. Smith (1994) argues that treating an attitude as both motivational and representational in this way is incoherent, because these attitudes have inconsistent directions of fit. On his view, representational attitudes go out of existence when we discover that reality does not match their content—in other words, that reality is not as they represent it. In these cases, we recognize that the representations are false and abandon them. Meanwhile, motivational attitudes go out of existence when we discover that reality does match their content—in other words, that reality is the way we are motivated to bring it about. After all, motivation to bring about some state of affairs would be pointless if the state of affairs already obtains. So to put things together, any attitude that was both representational and motivational could not exist. If we thought reality did not match the content, its representational character would make it cease to exist, and if we thought reality did match the content, its motivational character would make it cease to exist. So no attitude can be both motivational and representational. If there were such an attitude, it would go out of existence whether or not we thought reality matched its content. Little (1997) and Price (1989) raise an important objection to Smith’s argument. Suppose a mental state represents a particular state of affairs as good. In keeping with the way representational states work, it will persist until someone discovers that the state of affairs is not good. The same mental state might motivate the person to bring about the state of affairs. In keeping with the way motivational states work, it would cease to exist when the state of affairs was already achieved. Such a state would not go out of existence no matter what in the way Smith suggests. It would stay in existence as long as one had evidence that the state of affairs represented was good, but was not yet achieved. So it seems to be conceptually possible for a pro-attitude to be a representational state without driving itself out of existence. We will grant

100

J. Koh and N. Sinhababu

Little and Price’s point and make a more empirical and psychological argument that human pro-attitudes are not representational states. As we see it, the problem for perceptualists is that contemporary Humean theories like those of Arpaly and Schroeder (2013) and Sinhababu (2009, 2017) provide a more simple and unified explanation of the phenomenology of desire than perceptualism does. These contemporary Humean theories assign a causal role to desire not only in motivating action, but also in generating a hedonically charged phenomenology in virtue of which we regard the objects of desire as valuable. The key difference between perceptualists and Humeans is that perceptualists see proattitudes like desire as constituted by a perceptual phenomenology, while Humeans see pro-attitudes as causing a perceptual phenomenology while not being perceptual states themselves. While this is a subtle difference, the advantages of the Humean approach are apparent when we see how sharply the perceptualist view of desire differs from the perceptual states that perceptualists analogize it to. Our visual, auditory, tactile, and olfactory perceptual experiences do not have any immediate motivational effects. They are merely ways we experience the world and are not immediate causes of intentional action. Experience is not, in general, a source of motivation. Experiences can cause belief—our sensory experiences often cause us to believe that the world is the way our sensations represent it—but they do not motivate action. In treating desires as experiences that motivate action, perceptualists invent a new type of perception that is unlike the other perceptual states familiar from psychology. If inventing a new type of perceptual state was the only way to accommodate the phenomenological and behavioral data, such inventions would serve an important purpose. But as Baker (2014) argues, the Humean theory provides us with an alternative that does not require positing perceptual states that are unlike any perceptual states we have encountered before. We can simply use desire as our one-and-only source of motivation instead of inventing new types of motivationally potent perceptual states that are unlike the ordinary perceptual states they are supposed to be analogous to. In avoiding such ad hoc inventions, the Humean theory helps us maintain the simpler psychological theory. Some versions of perceptualism, including Tenenbaum’s (2007) neo-scholastic view, treat desire as perceptual but locate motivational force entirely within beliefs caused by the perceptual states that are desires. As the perceptual states are perceptions of value, the beliefs that they generate will be beliefs about value. Tenenbaum thinks beliefs about value can motivate us to promote what we believe to be valuable. Such views combine the two ideas discussed in this section. They give perception a role in producing belief and treat beliefs about value as necessary for pro-attitudes simply by claiming that the beliefs are the pro-attitudes. These views share the problems of the sort of perceptualism that ascribes motivational force to perceptual states. Ordinary beliefs do not motivate action by themselves. They need to be combined with desires or other pro-attitudes in order to motivate us. Beliefs about how to get food motivate us only when combined with a desire for food, and beliefs about how to get sleep motivate us only when combined with a desire to sleep.

Pleasure Makes Pro-attitudes

101

As an analogy to the question of what makes a mental state a pro-attitude, consider the question of what makes an animal a koala. Even those who had not studied the development of koalas would have good reason to reject the view that a koala is a mouse that grew very large by eating eucalyptus leaves. The problems with this answer are analogous to the problems with perceptualist accounts of pro-attitudes that treat them as beliefs or perceptions that gained motivational force because of their evaluative contents. Animals do not, in general, grow to many times their usual size because they ate unusual foods, and mental states do not, in general, acquire sharply different new functional properties (like being able to motivate action) because they had unusual contents. Instead, what mental state types like belief, desire, and perception do can be characterized in a general way across all possible contents, just as the effects of food on an animal can be characterized in a general way across many possible foods. If one wants to argue that a koala is a mouse that grew very large from eating eucalyptus leaves, one should point to examples of animals that grow very large because they ate unusual foods. Similarly, perceptualists who claim that pro-attitudes are motivationally potent beliefs or perceptions should point to other cases where belief or perception gains radically new functional properties in virtue of having unusual contents. That will make it more empirically plausible that distinctive contents can bestow distinctive functional properties on mental states. So far, perceptualists have not done this, perhaps because there are no such cases. What happens if we accept the perceptualist account of pro-attitudes as motivationally potent beliefs or perceptions, when human beings do not have motivationally potent beliefs or perceptions? Then the perceptualist account combined with the empirical facts will entail that there are no pro-attitudes. The consequence for Davidson’s account of acting for a reason is that no human being ever acts for a reason. If one requires too much for a mental state to be a pro-attitude, one will fail to correctly classify cases of action for a reason. Defining pro-attitudes as mental state types that humans do not actually have generates an extreme version of this problem, where no human has what it takes to act for a reason.

4 Pleasure in Thinking of the Object Makes Pro-attitudes Now we present our view, on which pro-attitudes not only motivate action, but also cause us to experience pleasure upon having certain types of thoughts of their objects. This view fits nicely with what we know about the most fundamental pro-attitude in human psychology, desire. Desire causes us to experience more pleasure when we come to believe that we are more likely to attain its object, or when we have vivid mental representations of its object. So our view will classify cases of desire-driven behavior as action for a reason. We agree with Davidson that pro-attitudes must combine with means-end beliefs to motivate action. Desire can do this. But this motivational capacity alone is insufficient for generating intentional action, as the case of Nellie shows. An additional ingredient is needed, and we will argue that pleasure is suitable for playing this role.

102

J. Koh and N. Sinhababu

Why is pleasure a good candidate for being the missing ingredient that turns mere behavior into action for a reason and turns a cause of bodily movement into a proattitude? It is because being pleased by something is a simple way of seeing it as good. Pleasure, we might say, puts the “pro” in pro-attitude. This point is potential common ground between Humeans and perceptualists, as both invoke some kind of experience of value in constituting pro-attitudes. The difference between perceptualists and Humeans, as previously mentioned, is that perceptualists treat the experience of pleasure as itself motivational while Humeans see desire as separately generating both motivation and pleasant experience. The problem with perceptualism is that it takes a mental state type generally observed not to generate motivation and argues that it does so in one specific instance. Humeans avoid this problem by locating motivation in a mental state that systematically generates it—desire, which they do not regard as a perceptual state. On the Humean view, desire is a dispositional state that produces motivation when combined with an appropriate means-end belief and produces a hedonically charged phenomenology when combined with representations of its object. It is not itself a perceptual state itself, but merely a cause of perception-like hedonically charged experiences. Schroeder (2004) and Sinhababu (2017) emphasize desire’s ability to cause pleasure and displeasure when we mentally represent its object in various ways. While their formulations are slightly different from each other, they all treat desire as causing experiences of pleasure when we mentally represent its object in one way or another. Schroeder defends the strong claim that pleasure itself is in part constituted by representations of increasing desire satisfaction. To be pleased is (at least) to represent a net increase in desire satisfaction relative to expectation; to be displeased is to represent a net decrease in desire satisfaction relative to expectation. Intensity of pleasure or displeasure represents degree of change in desire satisfaction relative to expectations. (94)

He provides neuroscientific evidence that activation in the brain’s reward center (where desire contents are encoded) causes activation in important pleasure centers as well as the motor cortex. This puts desire in position both to cause pleasure and motivate action. Sinhababu argues that desire causes pleasure when we have various sorts of thoughts of its object. Desire that E combined with increasing subjective probability of E or vivid sensory or imaginative representation of E causes pleasure roughly proportional to the desire’s strength times the increase in probability or the vividness of the representation. (With decreasing subjective probability of E or vivid sensory or imaginative representation of not-E, it likewise causes displeasure.) (28)

These phenomena can be described more colloquially. Learning that you are going to get what you desire is a happy surprise. Daydreams where you vividly imagine what you desire are happy daydreams. Like Schroeder, Sinhababu suggests ways to quantify the amounts of pleasure generated by interactions between desires and our representations of their objects.

Pleasure Makes Pro-attitudes

103

On both of these views, Nellie’s touching of her nose does not suggest that she has a pro-attitude toward touching her nose or making everyone mad. The psychological dispositions that cause her to touch her nose do not cause pleasure when she mentally represents touching her nose or making everyone mad. Since they cause the opposite of pleasure—displeasure—they make it the case that she has the opposite of a proattitude—perhaps an aversion or a con-attitude. Neither Schroeder nor Sinhababu accept motivational hedonism, according to which the only object of our motivational states is pleasure. The relations they posit between pleasure and pro-attitudes take a different form, as the formulations above suggest. As far as their views are concerned, we can be motivated to bring about any state of affairs at all, including ones in which we have less pleasure or more pain. Their claim is that while you might desire anything at all, thinking of whatever it is that you desire will please you. Both of these views require that a disposition to feel pleasure be psychologically integrated into pro-attitudes. It is not enough that one simply have pleasure sometimes (or every time) when one thinks of what is disposed to bring about. So if someone were to spray a pleasant aroma into the air whenever Nellie thought about making everyone mad or touching her nose, that would not make her behavioral dispositions into pro-attitudes. In such a case, the pleasure would come from a source other than the mental state disposing one to perform the behavior. It will be hard to give a full account of what such psychological integration amounts to until we have a better general theory of mental dispositions (and dispositions in general). Treating the disposition to cause pleasure as a necessary condition for being a proattitude allows us to act for reasons even when we believe that what we are doing is bad. Such actions are often sweet, with pleasant representations of the object. For example, if I stay in bed even though that means being late for an important work meeting, I do something that I judge that I should not do. But staying in bed is nice and comfortable, and my already being in bed at that point of time gives me a pleasant sensory representation of continuing to stay in bed. The same can be said of smoking a cigarette, or having too much alcohol. While we might have some displeasure when we mentally represent the bad states of affairs our actions eventually bring about, representations of the lesser goods to be produced by action cause us pleasure, and their ability to do so is integrated into the motivational states that drive us. In these cases, even though we might feel displeasure afterward, our representations of the objects of akratic actions at that point in time are pleasant, or more pleasant as opposed to their negation (i.e., not having a cigarette, going home, and not having another drink). As previously discussed, cases of actions we believe are bad cause trouble for views on which beliefs about the value of the action are necessary for a mental state to be a pro-attitude. The present account avoids these problems. Potential objections might involve actions we perform with displeasure. One may feel displeasure as one performs an onerous task that one promised to do, concedes defeat in a political campaign, or apologizes for a wrong action. As these examples show, it is certainly possible for our intentional actions to cause displeasure in us as we perform them. But these cases are not counterexamples to the relation between

104

J. Koh and N. Sinhababu

pleasure and pro-attitudes that Schroeder and Sinhababu suggest, because their theories treat states of affairs and not actions as the proper objects of pro-attitudes. What they say we will regard with pleasure or avoid displeasure in thinking of is the state of affairs that we hope to achieve, not the action itself. In this, their accounts rightly depart from Davidson and join contemporary decision theorists as we discussed in Sect. 2. In each case above, the state of affairs that one avoids is likely to be viewed with displeasure. The sorts of people who will do onerous tasks to fulfill promises would be displeased to recognize that they have broken their promises. Someone who concedes defeat in a political campaign may feel displeasure in thinking of being ridiculed for refusing to acknowledge obvious defeat. Leaving the wronged party aggrieved is the sort of thing that someone apologizing regards with displeasure. Relations between desire and pleasure not only play a role in constituting proattitudes, but in helping us introspect their presence. As Ashwell (2013) argues, introspection of pro-attitudes is quite different from the introspection of evaluative beliefs. Instead, “wanting things makes you see them in a certain light, and this is how you introspectively know what you want” (255). To know whether an attitude is a pro-attitude, one simply has to think about its object. If thoughts of the attitude’s object are pleasing, then that attitude is most probably a pro-attitude. Desire’s ability to cause pleasure thus is helpful not only with the metaphysics of pro-attitudes, but with the epistemology of pro-attitudes as well. There are a number of reasons why philosophers of Davidson’s era might not have considered invoking dispositional relations to pleasure as a necessary condition for pro-attitudes. One is that characterizing pro-attitudes in terms of the actions they motivate goes poorly with such a solution. We have addressed this by treating proattitudes as aiming at the states they produce rather than the actions they motivate. Another is that behaviorism still exerted considerable influence over the philosophy of mind and action in Davidson’s time. This made empirically minded philosophers less inclined to invoke internal mental experiences in giving accounts of action. As the influence of behaviorism recedes, we can do so with pleasure.

References Arpaly, N., & Schroeder, T. (2013). In praise of desire. Oxford: Oxford University Press. Ashwell, L. (2013). Deep, dark … or transparent? Knowing our desires. Philosophical Studies, 165(1), 245–256. Baker, D. (2014). The abductive case for Humeanism over quasi-perceptual theories of desire. Journal of Ethics & Social Philosophy, 8(2), 1–29. Davidson, D. (1963). Actions, reasons, and causes. Journal of Philosophy, 60(23), 685–700. Davidson, D. (1973). Freedom to act. In T. Honderich (Ed.), Essays on freedom of action (pp. 137– 156). London: Routledge and Kegan Paul. Hume, D. (1739/2000). A treatise of human nature. Oxford: Oxford University Press. Jeffrey, R. (1983). The logic of decision. Chicago, IL: University of Chicago Press. Little, M. (1997). Virtue as knowledge: Objections from the philosophy of mind. Noûs, 31(1), 59–79. Oddie, G. (2005). Value, reality, and desire. Oxford: Oxford University Press.

Pleasure Makes Pro-attitudes

105

Price, H. (1989). Defending desire-as-belief. Mind, 98(389), 119–127. Quinn, W. (1993). Morality and action. Cambridge: Cambridge University Press. Schafer, K. (2013). Perception and the rational force of desire. Journal of Philosophy, 60(5), 258– 281. Schroeder, T. (2004). Three faces of desire. Oxford: Oxford University Press. Sinhababu, N. (2009). The humean theory of motivation reformulated and defended. Philosophical Review, 118(4), 465–500. Sinhababu, N. (2017). Humean nature. Oxford: Oxford University Press. Smith, M. (1994). The moral problem. Oxford: Blackwell. Stampe, D. (1986). Defining desire. In J. Marks (Ed.), The ways of desire (pp. 149–173). Chicago: Precedent. Tenenbaum, S. (2007). Appearances of the good: An essay on the nature of practical reason. Cambridge: Cambridge University Press. Wallace, R. J. (1999). Three conceptions of rational agency. Ethical Theory and Moral Practice, 2(3), 217–242.

On the Distinctive Value of Knowledge Kok Yong Lee

Abstract Intuitively, knowledge is distinctively valuable, i.e., knowledge is more valuable than any of its proper parts. Call it “the distinctive value thesis.” Recently, the distinctive value thesis has been forcefully challenged by three arguments, which I call “the swamping argument,” “the generalized swamping argument,” and “the ad hoc argument,” respectively. These three arguments rely on what I will call “epistemic veritism,” the view that the distinctive value of knowledge is parasitic on the value of truth. Against these arguments, I argue that not only is epistemic veritism compatible with the distinctive value thesis, the former in effect implies the latter. My argument depends on the idea that justification enough for knowledge does not contain or presuppose certain falsehoods in a way that mere justification does. Keywords Epistemic veritism · Distinctive value thesis · Swamping argument · Ad hoc argument · Value of knowledge

1 Introduction The discussion of the value of knowledge begins with the Meno problem, the problem of explaining why knowledge is more valuable than mere true belief. According to the standard analysis, knowledge implies true belief. The Meno problem thus indicates that knowledge is more valuable than some of its proper part.1 One may further ask the question whether or not knowledge is more valuable than any of its proper parts

1A

is a proper part of B if and only if A is a part of B but not the other way around.

This chapter is in its final form and it is not submitted for publication anywhere else. K. Y. Lee (B) Department of Philosophy, National Chung Cheng University, 168, Sec. 1, University Rd. Minhsiung, Chiayi 62102, Taiwan e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_7

107

108

K. Y. Lee

(or whether knowledge is “distinctively valuable” for short). Call this “the value problem.”2 Intuitively, knowledge is distinctively valuable. Call this “the distinctive value thesis.” Despite its initial plausibility, the distinctive value thesis has been forcefully challenged recently. Central to the recent anti-distinctive value thesis movement, as I see it, is the view that the distinctive value of knowledge is parasitic on the value of truth. Call this view “epistemic veritism.”3 Epistemic veritism is a natural corollary of two widespread views on knowledge, epistemic inquiry, and epistemic appraisal. First, it is widely held that the condition that turns one’s true belief into knowledge is exclusively truth-conducive. This view has been labeled “intellectualism.”4 Many have found intellectualism, at least initially, “highly intuitive” (DeRose 2009, 24) or even “clearly true” (Fantl and McGrath 2009, 28).5 Intellectualism implies that knowledge is a kind of true belief formed in a certain truth-conducive manner. Hence, the distinctive value of knowledge must depend on the value of truth. Second, many have regarded truth as the goal of epistemic inquiry and/or the norm of epistemic appraisal. For one thing, on the standard view, the goal of epistemic inquiry is obtaining truth and avoiding error. “We must know the truth; and we must avoid error,” William James says, “these are our first and great commandments as would-be knowers” (James 1992, 469; italics omitted). Similarly, Laurence BonJour claims that what “makes us cognitive beings at all is our capacity for belief, and the goal of our distinctively cognitive endeavors: truth” (BonJour 1985, 83; original italics).6 For another thing, it is widely accepted that epistemic evaluation 2 Notice

that “the distinctive value (of knowledge)” and “the value problem” are my own terms of art. Some philosophers might use these terms differently. For instance, Duncan Pritchard takes “the distinctive value of knowledge” to also require that knowledge possesses a different kind of value in comparison with its proper parts (Pritchard et al. 2010, 7–8). Moreover, I think it is fair to say that Pritchard’s assumption—that the difference in value between knowledge and its proper parts is of different kinds—is controversial. To avoid controversies, I will not adopt Pritchard’s assumption here. 3 Here, the value of truth includes both truth values and truth-conduciveness. This allows us to distinguish between the following cases: Case 1. One believes that p and that q1, …, qn, which are all true, but the q’s are logically irrelevant to p. Case 2. One believes that p and that q1, …, qn which jointly entail p. All of the q’s are true. While the number of true beliefs involved in Case 1 and Case 2 may be the same, one’s beliefs that q1 , …, qn in Case 2 are truth-conducive regarding p in a way that one’s beliefs that q1 , …, qn in Case 1 are not. Hence, according to epistemic veritism, one’s beliefs in Case 2 are more valuable than one’s beliefs in Case 1. I want to thank an anonymous referee for raising this point. 4 The term “intellectualism” is from Stanley (2005). The present formulation of intellectualism is from DeRose (2009). Fantl and McGrath (2009) call it “purism.” 5 For complaints about intellectualism, cf. Stanley (2005) and McGrath (2010). For a recent defense of intellectualism, cf. Lee (2020). 6 For more on the traditional view of epistemic goal, cf. David (2001). For an objection to the traditional view, cf. Riggs (2003).

On the Distinctive Value of Knowledge

109

is truth-conducive by nature. Epistemic evaluation, according to William Alston, “is undertaken from what we might call the ‘epistemic point of view’. That point of view is defined by the aim at maximizing truth and minimizing falsity in a large body of beliefs” (Alston 1985, 59). Similarly, Ruth Weintraub claims that “epistemic appraisal is made relative to one particular desideratum: maximal truth- and minimal falsity-content in our beliefs” (Weintraub 1997, 1; original italics). If what makes knowledge epistemically valuable is its fulfillment of such epistemic goals and/or norms, the distinctive value of knowledge must depend on the value of truth. In one way or another, philosophers have argued that the distinctive value thesis is in conflict with epistemic veritism. Arguments along this line, however, are incorrect. Not only is the distinctive value thesis compatible with epistemic veritism, the latter indeed implies the former, or so I shall argue. In the first part of this paper, I will examine three arguments that conclude, based on epistemic veritism, that having a reliably formed true belief is no more valuable than having a mere true belief (Sect. 2), that having a justified true belief is no more valuable than having a mere true belief (Sect. 2), and that having a piece of knowledge is no more valuable than having a mere justified true belief (Sect. 3), respectively. I find all of them wanting. In the second part, I will argue that knowledge being distinctively valuable is given rise to by epistemic veritism (Sect. 5). My argument rests on the view that justification enough for knowledge does not essentially contain or presuppose certain falsehoods in a way that mere justification does (Sect. 4). In the end, I will consider and reply to two possible objections (Sect. 6).

2 The (Generalized) Swamping Argument Some philosophers have argued that the value problem poses a serious problem for simple reliabilism, the view that knowledge is true belief formed by a reliable beliefforming process (or “a reliably formed true belief” for short). The argument runs as follows. Reliability is a truth-conducive notion; the degree of reliability of a belief-forming process depends on its capacity to produce true beliefs as opposed to false ones. Hence, arguably, the value of (reliable) belief-forming processes is parasitic on the value of true beliefs, meaning that the value of a (reliable) belief-forming process (a) is entirely conferred by and (b) does not exceed the value of true beliefs. For instance, the value of a state of likely obtaining X is parasitic on the value of X, other things being equal (Kvanvig 2003, 45). However, if all we value are true beliefs, having a reliably formed true belief is no more valuable than having a mere true belief. Consider Linda Zagzebski’s famous analogy (Zagzebski 2003, 13). A reliable coffee machine is valuable precisely because it is reliable in producing good-taste coffee. If all we value is good-taste coffee, the value of having a cup of good-taste coffee does not get a boost from being produced by a reliable coffee machine. Likewise, if all we value is truth, the value

110

K. Y. Lee

of having a true belief does not receive a boost from being reliably formed. Call this “the swamping argument” (cf. Swinburne 1999; Zagzebski 1996). The swamping argument can be reconstructed as follows: (S1) The distinctive value of reliable belief-forming processes is parasitic on the value of true beliefs. (S2) If the distinctive value of reliable belief-forming processes is parasitic on the value of true beliefs, the value of having a reliably formed true belief does not exceed the value of having a mere true belief.

Therefore, (S3) the value of having a reliably formed true belief does not exceed the value of having a mere true belief (cf. Kvanvig 2003, 63ff.).

Two remarks are called for. First, the swamping argument is characterized in terms of the notion of parasitic value. Another formulation of the argument appeals to the notion of instrumental value. For instance, Pritchard argues that the swamping problem arises from the idea that a reliable belief-forming process is instrumentally valuable relative to true belief (Pritchard et al. 2010, 9–10). For simplicity’s sake, I will only deal with the aforementioned formulation. The main conclusions of this paper will still hold even if we characterize the problem in terms of instrumental value. Second, S1–S3 is originally proposed to argue against simple reliabilism.7 If simple reliabilism is correct, S3 indicates that knowledge is not distinctively valuable. But knowledge is distinctively valuable. Therefore, simple reliabilism is incorrect. However, one can turn the table on the distinctive value thesis. If simple reliabilism is correct, S1 leads to epistemic veritism. Moreover, if simple reliabilism is correct, S3 indicates that knowledge is not distinctively valuable. Epistemic veritism, therefore, implies the denial of the distinctive value thesis. Notice that the swamping argument poses a serious problem for the distinctive value thesis only if simple reliabilism is correct. Not everyone is a simple reliabilist, though. While many may agree that knowledge is a kind of reliably formed true beliefs, perhaps few are willing to claim that reliably formed true belief is sufficient for knowledge. A non-simple reliabilist may agree that knowledge requires an additional anti-Gettier condition, and this condition may thus leave room for nonsimple reliabilism to account for the distinctive value of knowledge.8 At any rate, simple reliabilism is controversial. The argument S1–S3 is anything but a conclusive objection to the distinctive value thesis. For this reason, some might want to strengthen the swamping argument so that it depends on a more plausible account of knowledge. Kvanvig (2003) has pursued this line of thought (also see Pritchard et al. 2010). More precisely, Kvanvig argues 7 Jones

uses a similar argument to argue against the view that (a) “[t]he end of belief-formation is to gain true beliefs” and (b) that “[j]ustification is a means to attaining the end of true belief” (he calls it “epistemic instrumentalism”) (Jones 1997, 424). Epistemic instrumentalism is more general than reliabilism. 8 Goldman and Olsson (2009, 22–23) mention, but do not develop, this view.

On the Distinctive Value of Knowledge

111

that, given that knowledge is a kind of justified true belief, a similar argument can be constructed to show that epistemic veritism is in conflict with the distinctive value thesis. Crucial to Kvanvig’s argument is the claim that the distinctive value of justification is parasitic on the value of truth. According to Kvanvig, justification, if valuable at all, must be either extrinsically or intrinsically valuable. If justification is extrinsically valuable, its value must depend on the value of some other thing X. The natural candidate for X is truth since justification is an indication of truth (in Kvanvig’s term “a mark of truth” (2003, 52)). By contrast, if justification is intrinsically valuable, its value must consist in satisfying the goal of the game of epistemic inquiry (or the game of belief). “The claim that there is some important [category] that is important in its own right,” Kvanvig writes, “apart from the goal of the game (and apart from the goal of playing the game, for that matter), is preposterous” (2003, 54). The goal of the game of epistemic inquiry, for Kvanvig, is obtaining truth and avoiding error. Hence, regardless of whether justification is extrinsically or intrinsically valuable, its distinctive value is parasitic on the value of truth.9 We can construct a parallel argument showing that the value of having a justified true belief does not exceed the value of having a mere true belief: (G1) The distinctive value of justification is parasitic on the value of true beliefs. (G2) If the distinctive value of justification is parasitic on the value of true beliefs, the value of having a justified true belief does not exceed the value of having a mere true belief.

Therefore, (G3) the value of having a justified true belief does not exceed the value of having a mere true belief.

Call this “the generalized swamping argument.” The generalized swamping argument poses a more serious problem for the distinctive value thesis than the swamping argument does since, arguably, knowledge is (a kind of) justified true belief.10 Both the swamping argument and the generalized swamping argument indicate that epistemic veritism is in conflict with the distinctive value thesis. On the one hand, the idea underlying S1 and G1 is epistemic veritism. On the other hand, both S3 and G3 imply that knowledge is no more valuable than some of its proper parts. Neither the swamping argument nor the generalized swamping argument is sound, however. Consider S1 and G1. Both premises rest on epistemic veritism. I think epistemic veritism is correct, but the goal of this paper is not to defend the view. The goal, rather, is to show that epistemic veritism and the distinctive value thesis are 9 The

idea that the value of justification is parasitic on the value of truth is also supported by intellectualism. On intellectualism, justification—i.e., the condition that turns one’s true belief into knowledge—consists solely of truth-conducive factors. 10 The generalized swamping argument will be more powerful if “justification” is understood as a placeholder for whatever condition that turns true belief into knowledge.

112

K. Y. Lee

compatible and, even stronger, to show that the former implies the latter. Hence, I will not challenge S1 and G1. Both S2 and G2, by contrast, are implausible, and their problems have the same root cause. Let X and Y stand for different kinds of things and x and y for their instances, respectively. The principle of evaluation underlying S2 and G2 can be specified as follows: (PE) If the distinctive value of X is parasitic on the value of Y, then the value of having an y and an x does not exceed the value of having an y without an x.

While initially plausible, PE is problematic. On PE, the value of possessing a certain instance somehow supervenes on the value of its kind, but this has mistakenly confounded values of different sorts. In fact, the value of possessing an instance may be independent of the value of its kind. Consider the case of the Golden Goose. Arguably, the distinctive value of the Golden Goose is parasitic on the value of golden eggs—the Golden Goose is more valuable than the normal goose precisely because the former’s capacity of laying golden eggs. Still, it does not follow that the value of having a golden egg laid by one’s Golden Goose does not exceed the value of merely having a golden egg. The former state presupposes the possession of a Golden Goose while the latter does not. This makes all the difference. Even if all we value are golden eggs, being in the state of having a golden egg and a Golden Goose is still better than being in a state of merely having a golden egg since being in the former (but not the latter) state is more prolific in golden eggs, other things being equal. For instance, in the long run, the state of having a golden egg and a Golden Goose will arguably lead to the possession of more golden eggs than the state of having a golden egg. Viewed in this light, Zagzebski’s analogy is misleading at best. Let us call a coffee machine that can reliably produce good-taste coffee “a reliable coffee machine” and good-taste coffee produced by an unreliable coffee machine “mere good-taste coffee.” Suppose that the distinctive value of reliable coffee machines is parasitic on the value of good-taste coffee. Still, it does not follow that the value of having a cup of good-taste coffee produced by one’s reliable coffee machine does not exceed the value of having a cup of mere good-taste coffee. The former presupposes the possession of a reliable coffee machine while the latter does not, and this makes all the difference. Even if all we value is good-taste coffee, being in the state of having a cup of good-taste coffee and a reliable coffee machine is better than being in the state of having a cup of mere good-taste coffee, since the former (but not the latter) is prolific in good-tast coffee, other things being equal. For instance, in the long run, the state of having a cup of good-taste coffee and a reliable coffee machine will arguably generate more good-taste coffee than the state of merely having a cup of good-taste coffee does. Likewise, for belief-forming processes. Even if the distinctive value of reliable belief-forming processes is parasitic on the value of true beliefs, it does not follow that the value of having a true belief formed by one’s reliable belief-forming process does not exceed the value of having a mere true belief. The former presupposes the possession of a reliable belief-forming process while the latter does not. This makes all the difference. Even if all we value are true beliefs, being in a state of having a true

On the Distinctive Value of Knowledge

113

belief and a reliable belief-forming process is better than being in the state of having a mere true belief, since the former (but not the latter) is prolific in true beliefs. For instance, in the long run, the state of having a true belief and a reliable belief-forming process will arguably generate more true beliefs than the state of having a mere true belief does.11 I contend that the same can be said of justification. That is, even if the distinctive value of justification is parasitic on the value of true beliefs, it does not follow that the value of having a justified true belief does not exceed the value of having a mere true belief. The former presupposes the possession of justification while the latter does not. This makes all the difference. Even if all we value are true beliefs, being in a state of having a true belief and some justification for it is better than being in the state of having a mere true belief since the former (but not the latter) is prolific in true beliefs in some sense (one of the main tasks of this paper is to articulate why and how having a true belief and justification (enough for knowledge) essentially gives rise to more true beliefs than merely having a true belief (Sect. 4)). This gives us a very promising way to reply to the (generalized) swamping problem. But before we get to that, there is another anti-distinctive-value-thesis argument that is worth addressing.

3 The Ad Hoc Argument The Meno problem focuses on the difference in value between knowledge and true belief. The contemporary development of the value problem has pushed the discussion further by focusing on the difference in value between knowledge and mere justified true belief. On the orthodox view, knowledge consists of justified true belief plus a (fourth) condition that handles Gettier cases (cf. Lehrer 2000, 20ff.). Call the last condition “the G-condition.” Accordingly, knowledge is G-justified true belief (hereafter, “knowledge” and “G-justified true belief” will be used interchangeably). Mere justified true belief , in the present context, means justified true belief suffering from Gettier cases. If knowledge is distinctively valuable, G-justified true belief is more valuable than mere justified true belief. In an influential book, however, Kvanvig (2003) argues that having a G-justified true belief is no more valuable than having a mere justified true belief. According to Kvanvig, Gettier cases manifest that unexpected, unaware epistemic fortuities render justified true belief falling short of knowledge. This kind of epistemic fortuities “wears its disvalue on its sleeve” (2003, 115), and the function of the Gcondition is to identify and eliminate such fortuities. Kvanvig further argues that the G-condition has to be gerrymandered and ad hoc in order to cope with a great variety of Gettier cases (2003, 117ff.). However, being gerrymandered and ad hoc, 11 This

point has been developed by Goldman and Olsson (2009).

114

K. Y. Lee

the G-condition is of little value. But if the G-condition is not valuable, having a G-justified true belief is no more valuable than having a mere justified true belief. Call this “the ad hoc argument.” To facilitate the following discussion, we may reconstruct the ad hoc argument as follows: (K1) The G-condition, being gerrymandered and ad hoc, is of little value. (K2) If the G-condition is of little value, the value of having a G-justified true belief does not exceed the value of having a mere justified true belief.

Therefore, (K3) the value of having a G-justified true belief does not exceed the value of having a mere justified true belief.

The ad hoc argument also depends on epistemic veritism. As noted, Kvanvig argues that the value of justification is parasitic on the value of truth. This implies that the value of knowledge, which is a kind of justified true belief, is also parasitic on the value of truth. Hence, when Kvanvig claims that the G-condition is of little value, he means that the G-condition is of little value with respect to truth (or truthconduciveness). The question of whether or not knowledge is more valuable than mere justified true belief, hence, is tantamount to the question of whether or not having a G-justified true belief is more valuable than having a mere justified true belief from the point of view of truth. Unlike the (generalized) swamping argument, the ad hoc argument does not rely on PE. Like the (generalized) swamping argument, however, the ad hoc argument also relies on a problematic principle of evaluation, i.e., K2. But, first, let us also note that K1 faces its own problems. Kvanvig’s argument for K1, roughly, is as follows. The function of the G-condition is to identify “any recognizably disvaluable kind of luck, accidentality, or fortuitousness,” and the G-condition gains its value by eliminating such kind of disvaluable fortuity (2003, 115). However, being gerrymandered and ad hoc, the G-condition fails to identify and eliminate the kind of fortuities that is recognizably valuable. For instance, when discussing Peter Klein’s distinction between false-dependent defeaters and non-false-dependent defeaters, Kvanvig writes that “[t]he distinction between these kinds of defeaters tracks no intuitive difference in value, leaving us with an account of the nature of knowledge incapable of helping to explain the value of knowledge” (2003, 130). It is not true that the G-condition is gerrymandered and ad hoc, at least not necessarily. In Sect. 4, I will offer an account of the G-condition that is both theoretically motivated and simple. For now, let us suppose, for the sake of argument, that the G-condition is indeed gerrymandered and ad hoc. Still, K1 seems dubious as it is misguided to think that the correct account of the G-condition must track certain intuitive differences in value. While an account of the G-condition may be gerrymandered and ad hoc, the epistemic fortuities characterized by such condition can still be valuable. For an object (event) can be identified by various conditions, and just

On the Distinctive Value of Knowledge

115

because an object (event) can be identified by a gerrymandered and ad hoc condition it does not follow that the object (event) itself is not valuable. Suppose that I put the first edition of the King James Bible in the second closet of my working desk. The Bible is valuable but can be identified by a gerrymandered and ad hoc condition, namely, the thing that is in the second closet of my working desk. Some of Kvanvig’s remarks suggest that by saying that the gerrymandered and ad hoc G-condition is of little value, he means that the G-condition as such is of little value in explaining the distinctive value of knowledge. For instance, Kvanvig discusses and agrees with Timothy Williamson’s point that there is “[an] inverse proportionality … between the complexity and ad hoc character of an account of knowledge and the usefulness of such an account in an explanation of the value of knowledge” (2003, 116). I have qualms about the idea that there is a tension between the ad hoc character of the G-condition and the usefulness of the account of the G-condition vis-à-vis the value problem. It is not clear that no ad hoc G-condition is able to capture epistemically valuable status. But even if we grant this idea, the ad hoc argument still faces a problem. That is, when K1 is understood in this way, K2 becomes implausible. The reason is that, in general, whether or not the G-condition is useful in characterizing the distinctive value of G-justified true belief should have no direct bearing on whether or not having a G-justified true belief is more valuable than having a mere justified true belief. An analogy is helpful here. Suppose that I construct a very complicated proof for all theorems of a logic system. Theorems are logically more valuable than non-theorems (let us assume), but my proof is of no use in explaining that (since it is too complicated). Yet it does not follow that having a theorem is not logically more valuable than having a non-theorem. Let us move on to K2. Kvanvig’s argument for K2, roughly, is that since the Gcondition is the only difference between G-justified true belief and mere justified true belief, G-justified true belief would be more valuable than mere justified true belief only if the G-condition is valuable (cf. Kvanvig 2003, 115). However, Kvanvig does not explain how the difference in value between Gjustified true belief and mere justified true belief depends on the value of the G-condition per se. One possible explanation is to resort to the following: The Summation Account. The value of an object (event) as a whole is the sum of the values of its components.

Accordingly, the value of mere justified true belief is equivalent to the value of Gjustified true belief minus the value of the G-condition.12 The summation account thus gives rise to K2. The summation account, however, does not hold universally. The value of the whole does not always equal the sum of the values of its parts. In some cases, the value of the whole is greater than the values of its parts, while in other cases, lesser. For instance, the value of compassionate pain, presumably, is greater than the sum

12 This

is how DePaul understandings Kvanvig’s argument (cf. DePaul 2009, 120).

116

K. Y. Lee

of the values of compassion and pain; the value of undeserved pleasure, on the other hand, is lesser than the sum of the values of undesert and pleasure (Hurka 1998). More importantly, analyzing the concept of knowledge shows that the G-condition is not conceptually independent of justified true belief. To illustrate, consider the following analysis of knowledge: The No False Grounds Theory. S knows that p if and only if (i) S believes justifiably and truly that p and (ii) S’s justification for p consists only of true beliefs.13

We can see from this analysis that the G-condition, i.e., (ii), is not a condition independent of justification. Rather, the G-condition modifies (refines or qualifies) S’s justification. Admittedly, the no false grounds theory still suffers from counterexamples (cf. Feldman 2003, 31–32). But this should not debar us from drawing a useful moral from this example. That is, in general, knowledge is a kind of true belief whose justification is modified by the G-condition. That the G-condition modifies S’s justification is a prima facie reason to reject the summation account since, in many cases, the value of A as modified by B cannot be simply identified with the sum of the values of A and B. For instance, the value of chocolate milk cannot be simply identified with the value of the sum of the value of chocolate and the value of milk, whatever they are. Put more generally, if B modifies A, then, in many cases, A as modified by B is a basic unit of evaluation. A more plausible approach is to reject the summation account and argue instead that the value of knowledge is determined organically by the values of its parts.14 That is, the value of knowledge depends on the values of its parts evaluated as a whole. This account, however, does not (at least not necessarily) support K2: If the value of knowledge depends on the values of its parts evaluated as a whole, the difference in value between knowledge and mere justified true belief may not depend on the value of the G-condition per se. In particular, G-justification may be more valuable than mere justification even if the G-condition per se is of little value.

4 The Distinctive Feature of G-Justification We have seen three arguments indicating that epistemic veritism is incompatible with the distinctive value thesis. All of them have relied on a certain problematic principle of evaluation. The (generalized) swamping argument presupposes PE, while the ad hoc argument, K2. The correct principle of evaluation, as I will argue, shows that, even if the value of justification is parasitic on the value of truth, the value of having a justified true belief may still exceed the value of having a mere true belief. Having a G-justified true belief may likewise be more valuable than having a mere justified true belief. 13 This

is Feldman’s (2003, 31) formulation of Clark’s (1963) account. A’s value is essentially organic if A’s value is not the sum of its parts (Hurka 1998). Hurka has dealt with two possible interpretations of the organic unity of value, but I will not address this nuance here.

14 Roughly,

On the Distinctive Value of Knowledge

117

My argument runs like this. I will first point out that the main difference between Gjustification and mere justification is that the latter essentially contains or presupposes certain falsehoods in a way that the former does not. Accordingly, the state of having a G-justified true belief is more valuable than the state of having a mere justification true belief, given that all we value is truth. I will then extend this account to deal with the difference in value between knowledge and mere true belief. Having a G-justified true belief is more valuable than having a mere true belief inasmuch the same way as having a G-justified true belief is more valuable than having a mere justified true belief. Let us explore the distinctive feature of G-justification. A defining feature of Gjustified true beliefs is that they do not suffer from Gettier cases. Analyzing Gettier cases hence is crucial to illuminating the distinctive feature of G-justification. Gettier cases come in different shapes and sizes, but the core structure seems to be that the subject possesses certain epistemically defective ground in forming or sustaining her true belief (cf. Fogelin 1994). According to a prominent diagnosis, if one merely believes truly and justifiably that p, then one’s (mere) justification for p involves or depends on a certain falsehood. For instance, in one of the earliest replies, Michael Clark points out that Gettier’s original counterexamples show that “a proposition which is in fact true is believed on grounds which are in fact false” (Clark 1963, 46). Gilbert Harman expresses a similar thought, when he advances the principle that “[r]easoning that essentially involves false conclusions, intermediate or final, cannot give one knowledge” (Harman 1973, 47). Recently, Richard Feldman proposes that G-justification for p “does not essentially depend on any falsehood” (Feldman 2003, 37). Feldman talks about essentially depending on falsehoods, but he does not say much about the condition. William Lycan (and also Harman) helpfully interprets essentiality involved as indispensability. On Lycan, G-justification “must not essentially rest on a false assumption; any false assumption on which it does rest must be dispensable” (Lycan 2006, 157, original italics). Consider one of Gettier’s original counterexamples: Smith. Smith and Jones had applied for the same job. Smith believed justifiably that Jones got the job, and that Jones had ten coins in his pocket. Smith then inferred, from his justified beliefs, and came to believe that the man who got the job had ten coins in his pocket. However, unbeknown to Smith, he was the person who got the job, and he too had ten coins in his pocket. (cf. Gettier 1963)

Smith’s belief that the man who got the job had ten coins in his pocket is both true and justified. Yet, intuitively, he did not know. Why? A natural reply is that Smith’s belief is defective in such a way that its justification contains, or depends on, the falsehood that Jones got the job. More importantly, Smith indicates that this falsehood is essential for Smith to form or sustain his belief that the man who got the job had ten coins in his pocket since, if he were to realize its falsity, he would have discarded the belief. I contend that the same holds in the following case15 : 15 Cases

like Barn are different from Gettier’s original cases in that their subject’s justification for p is typically good enough for knowledge, and that certain strange environmental factors debar

118

K. Y. Lee

Barn. A man was in a farm and saw a barn. He thereby formed a belief that he was facing a barn. However, unbeknown to him, he was actually surrounded by numerous fake barns. (cf. Goldman 1976, 772–773)

Intuitively, the man’s justified true belief that he was facing a barn was not knowledge since he did not realize that he was surrounded by fake barns.16 However, cases like Barn are different from cases like Smith in that the former do not seem to indicate that the subject’s justification contains, or depends on, a certain falsehood. Some might want to conclude that the subject’s justification is not defective in any epistemically significant sense. But this conclusion is too hasty, for it seems to me that, in Barn, the subject’s justification for the proposition He was facing a barn is still defective in the sense that it presupposes the falsehood that he is in a normal farm where fake barns are extremely rare (cf. Sosa 1986). Again, this falsehood is essential for the subject to form or sustain his belief that he was facing a barn since, if he were to realize its falsity, he would have discarded the belief. In sum, I contend that reflecting on the (aforementioned) Gettier cases indicates that mere justification essentially contains or presupposes some falsehoods in a way that G-justification does not, or, equivalently, G-justification essentially contains or presupposes some truths in a way that mere justification does not.17 Some remarks are called for. First, the present account distinguishes justification from presupposition. I do not mean to claim that there is a clear-cut distinction between one’s justification and the presupposition of one’s justification. There are a variety of vague cases where it is hard to tell whether a certain proposition is one’s justification or its presupposition. That justification and presupposition are not mutually exclusive, however, will not affect the discussion below. Second, to say that knowledge requires that one’s justification does not essentially contain or presuppose a certain proposition p does not imply that animals or young children, who are not capable of grasping p, do not have knowledge. I do not want to get into the debate about whether animals and young children are able to know (and if they do, whether they are able to grasp their justification). More generally, I want to remain neutral in the debate between the internalist and externalist accounts of justification. In what follows, I will assume a rather weak condition for what counts as being contained or presupposed by one’s justification. On this account, the subject’s justification counts as containing or presupposing p so long as the subject acts (or has a disposition to act) as if p is the case. This does not give us a full-blooded her justified true belief from being knowledge. In Pritchard’s terminology, cases like Barn (but not cases like Smith) contain certain “environmental epistemic luck” (cf. Pritchard et al. 2010). For related cases, which Lycan (2006) calls the “unpossessed-defeater” type of cases, see Harman (1973, 142–143). However, it is not uncommon for cases like Barn to be considered as a kind of Gettier cases (cf., e.g., Goldman and McGrath 2015). 16 Some reject the alleged intuitions regarding cases like Barn (cf. Gendler and Hawthorne 2005; Lycan 1977). However, even if one is ready to accept that the subject does know in cases like Barn, the following discussion of the value of G-justification will not be affected. 17 The equivalence is guaranteed by the intuitive idea that G-justification, but not mere justification, must contain or presuppose some truths.

On the Distinctive Value of Knowledge

119

account of what it takes for a proposition being contained or presupposed by one’s justification, but it does give us what we need for the present purposes. As shown above, the subject in a Gettier case does act (or has a disposition to act) as if a certain false proposition holds. In this sense, animals and young children can also possess knowledge so long as they do not act (or have a disposition to act) as if a certain false proposition holds. Third, the present account shows that Kvanvig is wrong in claiming that the G-condition is gerrymandered and ad hoc. For one thing, the different ways that G-justification and mere justification are related to truths (or falsehoods) clearly cut an important epistemic joint. For another, this distinction is theoretically motivated as it is manifested by our intuitive judgments regarding Gettier cases. Fourth, some might further complain that the present account of G-justification is problematic because it offers a criterion for distinguishing G-justification from mere justification. More precisely, an account of G-justification might be said to be in conflict with the popular view that the concept of knowledge is unanalyzable (cf. Williamson 2000). The question of whether knowledge is analyzable lies beyond the scope of this paper. However, notice that the present account, strictly speaking, is compatible with the idea that knowledge is unanalyzable or that there is no non-circular analysis of knowledge. On the present account, G-justification differs from mere justification in that the latter essentially contains or presupposes a certain falsehood in a way that the former does not. But I do not provide a full characterization of the notion of essentiality in play. The present account, hence, is compatible with the possibilities that a full-blooded account of essentiality involves the notion of knowledge, or that the analysis of knowledge based on the present account is circular.18 For even if the concept of knowledge is indispensable in characterizing the notion of essentiality in play, it will not affect the fact that mere justification essentially contains or presupposes a certain falsehood in a way that G-justification does not. With some crucial points clarified and initial worries replied, let us suppose that G-justification contains or presupposes q1 , …, qn . Call it that G-justification consists of q1 , …, qn . By the same token, mere justification consists of q1 , …, qm , where some qi belongs to mere justification but not to G-justification (and vice versa). If mere justification essentially consists of certain falsehoods in a way that G-justification does not, then some qi of mere justification, which do not belong to G-justification, must be false. Similarly, if G-justification essentially consists of certain truths in a way that mere justification does not, then some qi of mere justification, which do not belong to G-justification, must be true. This explains why mere justification, but not G-justification, is inherently defective. We are now in a position to show that epistemic veritism implies the distinctive value thesis. In the next section, I will first show that epistemic veritism implies that 18 An interesting note: Williamson, a prominent proponent of the unanalyzability of knowledge, also writes that “a true belief essentially based on false beliefs does not constitute knowledge” (Williamson 2000, 78). If I am right, Williamson’s remark will not be inconsistent, for the unanalyzability of knowledge is compatible with the idea that G-justification does not essentially contain or presuppose of a certain falsehood.

120

K. Y. Lee

knowledge is more valuable than mere justified true belief. I will then show that knowledge is more valuable than mere true belief, given epistemic veritism.

5 The Distinctive Value of Knowledge When evaluating the value of having a justified true belief, it seems plausible to think that its value depends on the values of its components in a certain way. More precisely, the value of having a justified true belief is determined by the following principle: VJTB The value of having a justified true belief that p is equal to the value of having a true belief that p and its justification q1 , …, ql (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole.

Presumably, the principle of evaluation in play is more complex than the summation principle. But I do not have a full-blown account of the principle of evaluation to offer. In the absence of a better option, I will work with a somehow intuitive idea of the value of an object (event) evaluated as a whole. As far as I can tell, the following argument is compatible with any plausible principle of evaluation. Since G-justified true belief and mere justified true belief are special cases of justified true belief, the following are special cases of VJTB: VJTBG The value of having a G-justified true belief that p is equal to the value of having the true belief that p and its G-justification q1 , …, qn (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole. VJTBM The value of having a mere justified true belief that p is equal to the value of having the true belief that p and its mere justification q1 , …, qm (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole.

According to VJTBG and VJTBM , the question whether or not having a piece of knowledge is more valuable than having a mere justified true belief is tantamount to the question whether or not the value of having the true belief that p and its Gjustification q1 , …, qn (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole exceeds the value of having the true belief that p and its mere justification q1 , …, qm (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole. When it comes to evaluating the difference in value between having a piece of knowledge and having a mere justified true belief, we may further ignore the value of having that particular true belief. This move is not implausible. Consider a case where S1 believes that p based on some good evidence while S2 believes that p based on a lucky guess. Intuitively, S1 ’s belief is epistemically more valuable than S2 ’s. More importantly, the truth-value of p has no impact on our intuitions—our intuitions regarding the difference in value between S1 ’s and S2 ’s beliefs will not be affected by learning the truth-value of p. The reason is that mere justified true belief and G-justified true belief are different kinds of true belief—G-justified true belief

On the Distinctive Value of Knowledge

121

is a kind of true belief with G-justification, while mere justified true belief, mere justification. If having a G-justified true belief is more valuable than having a mere justified true belief, the difference in value must not depend on the G-justified true belief being true. Rather, the difference in the value must consist in the difference between G-justification and mere justification. It follows that, when comparing the value of having a G-justified true belief with the value of having a mere justified true belief, we may simply focus on the difference in value between having G-justification and having mere justification. In other words, whether or not having a G-justified true belief is more valuable than having a mere justified true belief depends on whether or not having G-justification is more valuable than having mere justification. For the sake of argument, suppose that epistemic veritism holds. It follows that the difference in value between having G-justification and having mere justification should be evaluated from the point of view of truth. Obviously, given the distinctive feature of G-justification, having G-justification is more valuable than having mere justification, from the point of view of truth, since G-justification essentially consists of some truths in a way that mere justification does not. In other words, epistemic veritism implies that having a piece of knowledge is more valuable than having a mere justified true belief. We are now in a position to show that having knowledge is indeed more valuable than having mere true belief. Here “mere true belief” is understood as denoting true beliefs that lack justification whatsoever. This does not prevent the subject from seeing herself as having justification. For instance, a subject merely believes truly that p if she wildly guesses that p and gets lucky. The subject’s belief in p lacks justification even though she may regard herself as justified in believing that p.19 In this light, we may regard mere true beliefs as a kind of “degenerated” justified true beliefs, i.e., justified true beliefs with null justification. That is, when one merely believes truly that p, we may take it that one’s justification for p consists of no proposition. By VJTB, the value of having a mere true belief is determined by the following: VMTB The value of having a mere true belief that p is equal to the value of having a true belief that p and its justification q1 , …, qo evaluated as a whole (while {q1 , …, qo } is an empty set).

According to VJTBG and VMTB, the question of whether or not having a piece of knowledge is more valuable than having a mere true belief amounts to the question of whether or not the value of having the true belief that p and its G-justification q1 , …, qn (where for any i with 1 ≤ i ≤ l, qi = p) evaluated as a whole exceeds the value of having the true belief that p and its justification q1 , …, qo evaluated as a whole (while {q1 , …, qo } is an empty set). Likewise, the value of p being true plays no role in determining the difference in value between having a piece of knowledge and having a mere true belief. Therefore, when comparing the value of having a Gjustified true belief and the value of having a mere true belief, we may simply focus 19 To clarify, believing that p truly based on misleading (but persuasive) information does not count

as merely believing truly that p. Misleading justification is still justification.

122

K. Y. Lee

on the difference in value between having G-justification and having no justification at all. Given epistemic veritism, it is obvious that having G-justification is better than having no justification at all since G-justification essentially consists of some truths (empty justification, of course, consists of no truth). That is, epistemic veritism implies that having a piece of knowledge is more valuable than having a mere true belief.

6 Possible Objections and Replies The alleged arguments against the distinctive value thesis based on epistemic veritism all fall short. On the contrary, the former, together with some plausible premises, implies the latter. The crux of my account is the idea that G-justification does not essentially consist of falsehoods in a way that mere justification does. I have clarified this account to avoid some initial confusions or worries. In this section, I will further consider and reply to two important objections to this thesis.

6.1 Kvanvig’s Objection Kvanvig has discussed a similar account of knowledge according to which knowledge requires “insulation from error.” On this view, bodies of evidence of knowledge do not “contain, presuppose, or confirm falsehoods” (Kvanvig 2003, 122). Kvanvig argues that the insulation-from-error account is of no use in solving the value problem since it is not even a correct account of knowledge. He offers two reasons. Firstly, the preface paradox and the lottery paradox show that our bodies of evidence nearly always support some falsehoods. In the preface paradox, an author justifiably believes the content of the book, although the book, presumably, contains some falsehoods. The author’s body of evidence for the content of the book thus supports some falsehoods. Similarly, in the lottery paradox, our body of evidence for each ticket being a loser also supports the false proposition that no ticket will win. These two paradoxes indicate that knowledge does not require insulation from error since, nearly always, our evidence also supports falsehoods (Kvanvig 2003, 122–123). Kvanvig’s second reason concerns statistical beliefs. A piece of statistical knowledge depends on statistical samples (i.e., statistical evidence), but some statistical samples may confirm falsehoods instead. Strictly, the preface paradox, the lottery paradox, and statistical belief do not necessarily undermine the insulation-from-error account. At best, they show that a body of evidence may contain, presuppose, or confirm falsehoods. It does not follow, however, that the body of evidence is enough for knowledge. For instance, suppose that, in the preface paradox, the author’s evidence for the content of her

On the Distinctive Value of Knowledge

123

book supports certain falsehoods. But this is not a counterexample to the insulationfrom-error account since, in this case, the author does not know all the contents of her book in the first place. Nevertheless, Kvanvig is right that a body of evidence enough for knowledge may contain, presuppose, or confirm falsehoods. Consider: Joe. Joe, the librarian, discovered that a book on the shelf was missing. He recalled that John had taken a book with him when he left a while ago and that the cover of the book in his hand looked like the missing one. Joe also recalled that no one was in the library at that time. So, Joe concluded, correctly, that John has the book. However, unbeknown to him, a janitor was on the other side of the library when John took the book.

Joe’s belief that John has the book is both true and justified. And it seems that Joe knows, too. Hence, Joe’s G-justification contains (or presupposes) the falsehood that no one was on the other side of the library when John took the book. However, the present account does not deny that G-justification may contain or presuppose falsehoods. In fact, the present account does not even imply that Gjustification does not essentially contain or presuppose certain falsehoods. Rather, the present account is a weaker contention that G-justification does not essentially contain or presuppose certain falsehoods in the way that mere justification does. In Joe, Joe’s G-justification does not essentially contain a falsehood in a way that mere justification does. To elaborate, suppose that we modify Joe such that Joe is merely justified in believing truly that John has a book. For instance, suppose that Joe believes that John has a book because he believes falsely that the man he saw in the library is John (rather the man Joe did see was John’s twin brother Jill, and John happens to have the book in question, etc.). Call this case Joe*. In Joe*, Joe’s belief that John has the book essentially depends on a certain falsehood (i.e., the man Joe saw in the library was John) in the following way: In Joe*, if Joe were to learn the falsity that leads him to have mere justified true belief, namely, if Joe were to learn that the man he saw in the library was not John, then Joe would have discarded his belief that John has the book. By contrast, the same does not hold in Joe. In Joe, if Joe has were to learn the falsehood contained in his justification, namely, if Joe were to realize that someone was on the other side of the library, he would still not have discarded his belief that John has the book. In other words, in Joe, Joe’s belief that John has the book does not essentially depend on a certain falsehood in a way that mere justified true belief does. The present account, thus, does not predict that Joe lacks knowledge in Joe. As a result, Kvanvig’s objections against the insulation-from-error account do not undermine the present account.

6.2 Knowledge from Falsehoods I have argued that knowledge may contain or presuppose falsehoods only if those falsehoods are not essential. Now, some might argue that my view is incompatible

124

K. Y. Lee

with the view of knowledge advanced by Warfield (2005) and others.20 According to Warfield, there are cases of knowledge from falsehood, according to which one may acquire inferential knowledge based on of false relevant premises (2005, 405). The following is one of Warfield’s examples: Meeting. I have a 7 pm meeting and extreme confidence in the accuracy of my fancy watch. Having lost track of the time and wanting to arrive on time for the meeting, I look carefully at my watch. I reason: ‘It is exactly 2:58 pm; therefore I am not late for my 7 pm meeting’. Again I know my conclusion, but as it happens it’s exactly 2:56 pm, not 2:58 pm. (Warfield 2005, 408)21

Meeting describes how the subject can come to know that she is not late for her 7 pm meeting by inferring from a false premise, i.e., it is exactly 2:58 pm. Warfield further contends that the false premise is essential for the subject’s possession of knowledge. Following Warfield, let us call knowledge that essentially depends on falsehoods (or false premises) ‘knowledge from falsehood’. Some might want to argue that the possibility of knowledge from falsehood undermines the present account of knowledge (or G-justification). The idea is that if Warfield’s diagnosis of cases like Meeting is correct, G-justification might essentially contain or presuppose certain falsehoods. Now, I do not think the present account is necessarily incompatible with the possibility of knowledge from falsehood; acute readers may recognize that this objection is similarly misguided as the objection mentioned in the previous section. The idea of that objection is that the present account of knowledge (or G-justification) requires that knowledge (or G-justification) does not essentially contain or presuppose certain falsehoods. But, strictly, the latter is not required by the account of knowledge (or G-justification) defended above. Instead, according to the view defended above, knowledge (or G-justification) does not essentially contain or presuppose certain falsehoods in a way that mere justified true belief (or mere justification) does. As a result, the present account is entirely compatible with the existence of knowledge from falsehood. To elaborate, suppose that we modify Meeting such that the subject merely justified in believing truly that she is not late for the meeting. For instance, suppose that the subject learn that he is not late for the meeting by looking at a watch, which has stopped working and happens to point to a correct time. Call this case Meeting*. In Meeting*, the subject’s belief that she is not late for the meeting essentially depends on a certain falsehood (i.e., the watch has not stopped working) in the following sense: In Meeting*, if the subject were to learn the falsity that leads him to form the (true) belief that he is not late for the meeting, namely, if the subject were to learn that the watch has stopped working, she would have discarded her belief that she was 20 Issues related to this view have received increasing attention from philosophers (cf., e.g., Fitelson 2010; for a recent proponent, cf. Hiller 2013; for a recent objection, cf. Schnee 2015). But for simplicity’s sake, I will only deal with Warfield’s account. 21 In an earlier paper, Hilpinen discusses a similar case (Hilpinen 1988, 163–164). Klein (2008) later calls such falsehoods that seemingly play a role in G-justification “useful false belief”. Klein also does not think that such useful false beliefs are essential to G-justification in the present sense.

On the Distinctive Value of Knowledge

125

not late for the meeting. By contrast, the same does not hold in Meeting. In Meeting, if the subject were to realize the falsity that leads her to believe that she is not late for the meeting, namely, if she were to realize that it is not exactly 2:58 pm, she would still not have discarded her belief that she is not late for the meeting. In other words, in Meeting, the subject’s justified true belief that she is not late for the meeting does not depend essentially on a certain falsehood in a way that mere justified true belief does. The present account, hence, does not predict that the subject lacks knowledge in Meeting. As a result, the present account is compatible with the possibility of knowledge from falsehood.

7 Conclusion Recent discussion of the value of knowledge has been trending toward pessimism with regard to whether or not knowledge is distinctively valuable. Many have concluded that knowledge cannot be distinctively valuable, if epistemic veritism is true, namely, if the value of knowledge is parasitic on the value of truth. Against this trend, I have argued that the three most prominent arguments for such a conclusion—the swamping argument, the general swamping argument, and the ad hoc argument—are all untenable. If what has been said is correct, epistemic veritism is not incompatible with the distinctive value thesis. In fact, if G-justification does not essentially contain a certain falsehood in a way that mere justification does, as the aforementioned discussion shows, then there are good reasons to believe that epistemic veritism implies the distinctive value thesis. Given the initial plausibility of epistemic veritism, the prospect of knowledge being distinctively valuable is bright. Acknowledgements I am grateful to Peter Markie, Matthew McGrath, Kevin McCain, Hanti Lin, Julien Dutant, and Xingming Hu, and two anonymous referees for very helpful comments on the previous drafts of this paper. Earlier versions of this paper had been presented at Central States Philosophy Association Conference at Wayne University (September 2010), Taiwan Philosophical Association Conference at Soochow University (November 2014), and American Philosophical Association (Central Division) at St. Louis (February 2015). I want to thank all participants for comments and discussions. This work is funded by the Ministry of Science and Technology (MOST) of Taiwan (R.O.C.) (MOST 107-2410-H-194-090-MY2).

References Alston, W. P. (1985). Concepts of epistemic justification. The Monist, 68(1), 57–89. BonJour, L. (1985). The structure of empirical knowledge. Cambridge, MA.: Harvard University Press. Clark, M. (1963). Knowledge and grounds: A comment on Mr. Gettier’s paper. Analysis, 24(2), 46–48.

126

K. Y. Lee

David, M. (2001). Truth as the epistemic goal. In M. Steup (Ed.), Knowledge, truth, and duty: Essays on epistemic justification, responsibility, and virtue (pp. 151–169). New York: Oxford University Press. DePaul, M. R. (2009). Ugly analyses and value. In A. Haddock, A. Millar, & D. Pritchard (Eds.), Epistemic value (pp. 112–138). Oxford: Oxford University Press. DeRose, K. (2009). The case for contextualism. Oxford: Clarendon Press. Fantl, J., & McGrath, M. (2009). Knowledge in an uncertain world. Oxford: Oxford University Press. Feldman, R. (2003). Epistemology. Upper Saddle River, NJ: Prentice Hall. Fitelson, B. (2010). Strengthening the case for knowledge from falsehood. Analysis, 70(4), 666–669. Fogelin, R. (1994). Pyrrhonian reflections on knowledge and justification. New York: Oxford University Press. Gendler, T. S., & Hawthorne, J. (2005). The real guide to fake barns: A catalogue of gifts for your epistemic enemies. Philosophical Studies, 124(3), 331–352. Gettier, E. (1963). Is justified true belief knowledge? Analysis, 23, 121–123. Goldman, A. I. (1976). Discrimination and perceptual knowledge. The Journal of Philosophy, 73(20), 771–791. Goldman, A. I., & McGrath, M. (2015). Epistemology: A contemporary introduction. New York: Oxford University Press. Goldman, A. I., & Olsson, E. J. (2009). Reliabilism and the value of knowledge. In A. Haddock, A. Millar, & D. Pritchard (Eds.), Epistemic value (pp. 19–41). Oxford: Oxford University Press. Harman, G. (1973). Thought. Princeton: Princeton University Press. Hiller, A. (2013). Knowledge essentially based upon false belief. Logos and Episteme, 4(1), 7–19. Hilpinen, R. (1988). Knowledge and conditionals. Philosophical Perspectives, 2, 157–182. Hurka, T. (1998). Two kinds of organic unity. The Journal of Ethics, 2, 299–320. James, W. (1992). The will to believe. In Writings 1878–1899 (pp. 457–479). New York: The Viking Press. Jones, W. E. (1997). Why do we value knowledge? American Philosophical Quarterly, 34(4), 423–439. Klein, P. (2008). Useful false beliefs. In Q. Smith (Ed.), Epistemology: New essays (pp. 25–61). Oxford: Oxford University Press. Kvanvig, J. (2003). The value of knowledge and the pursuit of understanding. New York: Cambridge University Press. Lee, K. Y. (2020). On two recent arguments against intellectualism. NCCU Philosophical Journal, 43, 35–68. Lehrer, K. (2000). Theory of knowledge. Boulder, CO: Westview Press. Lycan, W. G. (1977). Evidence one does not possess. Australasian Journal of Philosophy, 55(2), 114–126. Lycan, W. G. (2006). On the Gettier problem problem. In S. Hetherington (Ed.), Epistemology futures (pp. 148–168). Oxford: Clarendon Press. McGrath, M. (2010). Contextualism and intellectualism. Philosophical Perspectives, 24, 383–405. Pritchard, D., Millar, A., & Haddock, A. (2010). The nature and value of knowledge: Three investigations. Oxford: Oxford University Press. Riggs, W. D. (2003). Balancing our epistemic goals. Noûs, 37(2), 342–352. Schnee, I. (2015). There is no knowledge from falsehood. Episteme: A Journal of Social Epistemology, 12(1), 53–74. Sosa, E. (1986). Presuppositions of empirical knowledge. Philosophical Papers, 15, 75–87. Stanley, J. (2005). Knowledge and practical interests. Oxford: Clarendon Press. Swinburne, R. (1999). Providence and the problem of evil. Oxford: Oxford University Press. Warfield, T. (2005). Knowledge from falsehood. Philosophical Perspectives, 19, 405–416. Weintraub, R. (1997). The sceptical challenge. New York: Routledge. Williamson, T. (2000). Knowledge and its limits. Oxford: Clarendon Press.

On the Distinctive Value of Knowledge

127

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. (2003). The search for the source of epistemic good. Metaphilosophy, 34(1/2), 12–28.

The Continuity of Davidson’s Thought: Non-reductionism Without Quietism Claudine Verheggen

Abstract This chapter focuses on the continuity of Davidson’s thought. My goal is to show that two commonly held conceptions of Davidson’s philosophy, one to the effect that there is a shift from radical interpretation to triangulation, in particular, a shift from reductionism to non-reductionism, the other to the effect that Davidson’s non-reductionism is incompatible with constructive philosophizing, are in fact misconceptions. I argue that, though Davidson’s account of meaning, being non-reductionist, does not spell out non-circular sufficient conditions for any particular utterances to have the meaning they have, and for someone to have a language and thoughts at all, it is nonetheless non-quietist, for it does give us necessary conditions for these phenomena to occur. These necessary conditions first emerge from Davidson’s reflections on radical interpretation and are further articulated in his writings on triangulation, which also demonstrate that radical interpretation is but an instance of triangulation, since they make clear that successful radical interpretation in effect requires triangulation. Keywords Davidson · Meaning · Non-reductionism · Quietism · Radical interpretation · Triangulation

1 Introductory Remarks At the centre of Donald Davidson’s philosophy, there has always been the idea that most philosophical questions can be addressed properly only if we first have a theory of meaning or content, that is, a theory of what it is for us to mean what we do by our utterances and to have the propositional thoughts and attitudes we have. And Davidson has always maintained that a good way to construct such a theory is by reflecting on radical interpretation, that is, on what it takes to figure out from scratch This chapter is in its final form and it is not submitted for publication anywhere else. C. Verheggen (B) Department of Philosophy, York University, 4700 Keele Street, Toronto, ON M3J 1P3, Canada e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_8

129

130

C. Verheggen

what an alien speaker means by her utterances and what propositional thoughts and attitudes she has. Thus he has always maintained that reflecting on how to construct a theory that specifies the meanings of an alien speaker’s words, a descriptive theory, will yield a philosophical understanding of the nature of meaning, a foundational theory. Davidson’s approach is motivated, first, by the demand that an elucidation of the nature of meaning should be non-question-begging, hence the inquiry into the interpretation of a speaker whose language and propositional attitudes the interpreter does not know. Second, it is motivated by the assumption that all the features that belong to meaning are essentially public,1 hence the claim that reflecting on how meanings can be assigned to an alien speaker’s utterances and thoughts teaches us much of what there is to know about how meanings are constituted. Most briefly put, then, for Davidson, one good path to the metaphysics of meaning is through its epistemology. On the face of it, it might look as if what Davidson is seeking is an account of meaning that makes no appeal to any semantic notion or context, that is, a reductionist account of meaning. This is how, more often than not, Davidson’s writings on radical interpretation have been understood.2 And this, to my mind, might explain why many commentators have failed to recognize that his thoughts on triangulation, the simultaneous interaction of two speakers with each other and some feature of their shared environment, are continuous with his thoughts on radical interpretation. For, whereas radical interpretation might be understood as a reductive enterprise, the view afforded by the writings on triangulation is unabashedly non-reductionist, since the conclusion of the triangulation argument is that only of someone who has triangulated linguistically with another person and the world they share does it make sense to say that she has a language and thoughts. Indeed, the triangulation argument has been widely attacked for its circularity, and the connected non-reductionism regarded as tantamount to abandoning any hope for a philosophically illuminating account of meaning.3 This is in part why, I suspect, the argument has not been given the attention it deserves. My aim in this chapter is to show that the two conceptions of Davidson’s philosophy sketched above, both to the effect that there is a shift from radical interpretation to triangulation, in particular, from reductionism to non-reductionism, as well as to the effect that Davidson’s non-reductionism is incompatible with constructive philosophizing, are in fact misconceptions. Though Davidson’s account, being nonreductionist, does not spell out non-circular sufficient conditions for any particular utterances to have the meaning they have, and for someone to have a language and thoughts at all, it is nonetheless non-quietist, for it does give us necessary 1 As

he writes: “meaning, and by its connection with meaning, belief also, are open to public determination. I shall take advantage of this fact…and adopt the stance of a radical interpreter when asking about the nature of belief. What a fully informed interpreter could learn about what a speaker means is all there is to learn; the same goes for what the speaker believes” (Davidson 1983, 148). 2 See Hattiangadi (2017) and Bridges (2017) for two recent instances of this interpretation. 3 See, e.g., Pagin (2001) and Bridges (2006).

The Continuity of Davidson’s Thought: Non-reductionism …

131

conditions for those phenomena to occur. These necessary conditions first emerge from Davidson’s reflections on radical interpretation, and are further articulated and expanded in his writings on triangulation, which also demonstrate that radical interpretation is but an instance of triangulation, since they make clear that successful radical interpretation in effect requires triangulation.

2 Davidson’s Brand of Semantic Non-reductionism As he came to express it, Davidson’s starting point when investigating philosophically the nature of meaning is the question, “what is it for words to mean what they do?” (1984, xi).4 It is not the question what it is for sounds and marks to have meaning, let alone the question, what is meaning? This is in itself revealing, for it indicates that the inquiry into meaning is to be conducted within language, scrutinizing signs that are already endowed with meaning, which, in effect, already suggests that no reductionist answer is forthcoming. The question also indicates that the account of meaning may come in two steps, which is exactly how Davidson initially proceeds. For, according to him, what we should be doing when investigating meaning is, initially, reflecting on the construction of a theory that specifies the meanings of an alien speaker’s words, that is, a theory that tells us what someone’s words do mean.5 This theory—call it semantic—is purely descriptive and empirical. But we are reflecting on what it would take to construct it in order ultimately to answer the question what it is for words to mean what they do, in order to say something “philosophically instructive” (ibid.) about the nature of meaning. Call this foundational, philosophical theory meta-semantic. According to Davidson, initially at least,6 how we assign meanings to a speaker’s words from scratch, what the grounds are for our understanding her linguistic behaviour, in short, the construction of the semantic theory, is supposed to tell us everything philosophically interesting there is to know about the nature of meaning. In light of the above, it might be wondered whether the label ‘non-reductionist’ applies both to the semantic theory and to the meta-semantic theory. To my knowledge, however, those who have discussed non-reductionism do not distinguish sharply between the two theories. Thus, in her criticism of Barry Stroud’s defence of

4 Note that my focus throughout the paper will be on language rather than thought, but, for Davidson,

whatever is true of language is true of thought propositionally conceived. The question what it is for words to mean what they do is equivalent to the question what it is for concepts (conceived as elements of propositional thoughts and attitudes) to have the content they have. 5 As we shall see later, strictly speaking, the theory tells us how words contribute to the meaning of the utterances in which they occur. 6 As I suggested above, and as we shall see, this claim will come to be augmented in the writings on triangulation.

132

C. Verheggen

non-reductionism, Hannah Ginsborg has recently written that Stroud confuses two senses in which an explanation of language can be “from outside” meaning7 : [this] rests on a failure to distinguish clearly between two different ideas: the idea that we cannot characterize the meaning of an expression without using that expression or some other expression with the same meaning, and the idea that we cannot characterize the meaning of an expression without using semantical expressions like “means,” “orders,” “says that” and so on. In other words, it fails to distinguish the idea that we cannot characterize meaning without drawing on our own grasp of concepts corresponding to the expressions whose meaning we are trying to characterize, from the idea that we cannot characterize meaning without drawing on our grasp of what it is to mean something by those expressions or to entertain those concepts. (Ginsborg 2011, 163)

Pretty clearly, the first characterization concerns the semantic theory: we cannot specify the meaning of an expression without using an expression that has the same meaning. The second characterization, on the other hand, concerns the meta-semantic theory: we cannot give an explanation of meaning without considering expressions as they are used to mean, order, or say, etc., things, (rather than as they are used merely to get, build, etc., things), that is, without considering expressions in semantic contexts. Few, I believe, would deny the first characterization.8 But nothing follows from this characterization—it is neutral with respect to reductionism. What is reductive or nonreductive is how the specification of meanings is arrived at, whether it indeed makes use of semantic contexts. But this has to do with the construction of the semantic theory and thus is part of the meta-semantic theory. In other words, just by looking at the semantic theory, we cannot tell whether the related account of what it is for words to mean what they do is reductionist or not. This can be done only by looking at the construction of the semantic theory. Davidson’s meta-semantic theory, his philosophical account of meaning, is, as we are about to see, non-reductionist even in the writings on radical interpretation. But that it is then already non-reductionist may have been missed as a result of not distinguishing sharply between semantic theory and meta-semantic theory. As will become crystal-clear when we examine the writings on triangulation, the metasemantic theory is non-reductionist because, in a nutshell, we cannot fully answer the question what it is for words to mean what they do without considering speakers using them meaningfully to communicate with one another (as it is only then, as we shall see, that we can think of their meanings as being fixed and hence specifiable). That is, we cannot fully answer the question from outside a semantic context; we cannot, as I put it earlier, give non-circular sufficient conditions for a speaker’s words meaning what they do. However, the construction of the semantic theory does deliver meta-semantic, philosophical goods. For it reveals that there are several necessary 7 See

Stroud (2012, 2017). I do not myself believe that Stroud is guilty of the confusion. I have discussed Ginsborg’s “partially reductionist” view in Verheggen (2015) and in Myers and Verheggen (2016, Chap. 2). 8 An exception is Michael Dummett when he was seeking a semantic theory that does not “give the interpretation of the language to someone who already has the concepts required” (Dummett 1975, 102).

The Continuity of Davidson’s Thought: Non-reductionism …

133

conditions to be met, such as, most centrally, interacting with others and a shared world, in order for someone to have a language and thoughts.

3 Davidson’s Early Non-reductionism: Radical Interpretation The idea of radical interpretation is first introduced in “Truth and Meaning” (Davidson 1967), which is devoted to the question what kind of theory “would provide an interpretation of all utterances, actual or potential, of a speaker or group of speakers” (Davidson 1984, xv), the question, that is, what form a semantic theory should take. As is well-known, Davidson thinks that a Tarski-style theory of truth would do the trick, as its recursive character could show “how the meanings of sentences depend upon the meanings of words” (Davidson 1967, 23), which in turn would help to show how it is possible to understand any of the infinity of sentences a speaker or group of speakers might utter (Davidson 1973, 127). The basic idea is that a properly constructed theory of truth would enable us to understand any such utterance by giving us its truth-conditions. These would be derived from a finite set of axioms that would tell us, for every primitive semantic expression and for every rule of combination, how they contribute to the truth-conditions, and hence the meaning, of any utterance in which they occur. The idea faced a number of difficulties, some of which are perhaps insurmountable, as Davidson himself acknowledged from the start.9 There may indeed be expressions or grammatical structures that do not fit the truth-conditional mould. But Davidson never doubted that the core of meaning is truth-conditional, in that he never doubted that what determines, at least in part, the meaning of our basic utterances, utterances about perceivable features of the world around us, is what makes these utterances true (though, as we shall see, how this determination is achieved is a complicated matter).10 How then is the theory of truth to be constructed? How is the radical interpreter supposed to proceed in order to understand an alien speaker?11 The radical interpreter has no knowledge of the speaker’s language, and no detailed knowledge of her propositional attitudes, for such detailed knowledge would require her to understand the speaker’s utterances. But there is a lot the interpreter does know and use while doing radical interpretation. Needless to say, she knows her own language and concepts, and how these are connected to the world. She “has the 9 See

Davidson (1967, 35–36), for an initial catalogue of these difficulties. in effect is an early expression of Davidson’s semantic externalism. For more on this, see Myers and Verheggen (2016, Chap. 3), and Verheggen (2017b), in which I focus on the continuity of Davidson’s semantic externalism, from radical interpretation to triangulation. 11 Radical interpretation has had many detractors, among the most virulent of whom are Fodor and Lepore (1993). Though Davidson has said that the criticisms were sometimes based on misunderstandings, he himself acknowledged that “neither [his] terminology nor [his] views have held absolutely steady” (Davidson 1993, 77 fn2). In a constructive spirit, I am myself putting the best possible spin on the thought experiment. 10 This

134

C. Verheggen

concepts of truth, of intention, of belief, of desire, and of assertion (and many, many more). She knows a lot about the world and about how people behave in various circumstances” (Davidson 1993, 81). Moreover, she and the speaker are alike in important respects, in particular, to cite those Davidson mentions, in what they can perceive, how they discriminate, how they are built and how they move (ibid.). Still, since the interpreter cannot yet understand the speaker’s utterances, on what kind of evidence can she start building her theory of interpretation? The answer to this question is also well known. Here are its principal elements. Start by detecting utterances of sentences held true by the speaker in particular circumstances. Sentences that are held true are sentences that express the speaker’s beliefs, and their being held true depends on what they mean for the speaker. If we knew what the speaker believes, we would know what she means by her utterances. If we knew what she means, we would know what she believes. But we know neither. How can we break into this circle? We can do this by employing the principle of charity, thereby assuming, to begin with, that the beliefs her utterances of sentences held true express are the beliefs we ourselves have in the circumstances in which she produces the utterances. That is, we can connect her utterances with features of the world around her that cause us to form certain beliefs and which we also take to cause the beliefs her utterances express. Thus, a rabbit passing by may cause both us and the speaker to form the belief that a rabbit is passing by, which the speaker expresses by uttering “Gavagai!”, which we proceed to interpret as meaning that a rabbit is passing by. Of course we cannot do so immediately. Before we even tentatively give the truth-conditions of the speaker’s utterance of ‘Gavagai’, we need to hear her use the expression in numerous other circumstances, both of the same type—perhaps she initially mistook the rabbit for a cat and actually meant to say that a cat is passing by—and of different types—perhaps she was initially asking about dinner, or commenting on the fur of the animal, or the speed of its run. We will need to hear her use the expression together with other ones, and in contrast with yet other ones. For Davidson, at least in his early writings, it is only when an interpreter knows all the axioms of a theory that she can understand the speaker for whom it is the theory (Davidson 1973, 138). He claimed eventually to have become “more restrained on the topic” (Davidson 1993, 80). Even so, however, it will be necessary for us to witness an extended range of the speaker’s linguistic responses to her environment before we settle on an interpretation of any primitive expression and come up with an axiom for it, that is, with a description of the way the expression contributes to the truth-conditions of the utterances in which it may occur. Doing this will also require that we extend the principle of charity to include not just the assumption that we share many beliefs with the speaker, but also the assumption that her beliefs are by and large rational and her desires by and large reasonable. For only against a background of rationality can a speaker’s behaviour be intelligible and indeed, sometimes, irrational. Leaving the details aside, the question now is, is the meta-semantic theory afforded by reflecting on radical interpretation a reductionist theory? At first blush it might be tempting to think so. The interpreter is after all basing her interpretation on regular connections between the speaker’s utterances and the features of her environment

The Continuity of Davidson’s Thought: Non-reductionism …

135

that cause her to produce them. Moreover, the theory of truth specifies meanings without appealing to any intensional entities or intensional constructions. There is no entity called meaning corresponding to ‘grass’ or ‘is green’, let alone to ‘grass is green’, that we appeal to in order to specify the meaning of ‘Grass is green’. And we have replaced the intensional ‘means that’ in “‘Grass is green’ means that grass is green” with the extensional ‘is true if and only if’.12 These endeavours are of course part and parcel of the attempt to account for the nature of meaning in a nonquestion-begging way. But they might also suggest that the account of meaning we are developing is reductionist, because they suggest that we can build the semantic theory without making use of any semantic context, which, as we shall see shortly, we in fact cannot do. As I said, the semantic theory is itself neutral with respect to reductionism. Thus we should not be misled by the fact that no intensional idioms are used in the specification of the truth-conditions, hence the meanings, of a speaker’s expressions. The relevant question is whether a semantic context is being used while constructing the theory that yields a specification of the truth-conditions. As it turns out, if truth-conditions are to be meaning-giving, the construction of the theory must proceed within a semantic context. The best way to start seeing this is by acknowledging that expressing the axioms of the theory and therefore the truth-conditions of the utterances in which primitive expressions occur in purely extensional terms, that is, using merely co-referring expressions, could not amount to giving their meaning. To take a familiar example, ‘Hesperus’ and ‘Phosphorus’ are co-referring names. Yet they may have different meanings for a speaker. As a result, in building her theory of truth that is to serve as a theory of meaning, the interpreter has to be careful to use a word that will have the same meaning as the speaker’s word and not just a co-referring expression. This, it might be retorted, goes without saying. We all know that there is more to the meaning of words than the items they may refer to. But what this indicates is that, in building her theory of truth, the interpreter must draw on evidence that goes beyond the mere connections between the speaker’s utterances and objects and events in her environment. As I suggested above, she must ascertain the multifarious complex connections that occur among the speaker’s beliefs and other propositional attitudes. Most importantly, as we saw Davidson makes clear, she must draw on the knowledge of her own language and concepts, as well as on what we may assume she knows about the speaker and the world around them. This is to say that whether the account of meaning afforded by radical interpretation is reductionist or not depends, not, to repeat, on whether the radical interpreter uses intensional idioms in specifying the speaker’s meanings, but on whether her own understanding of the words plays an essential role in the specification. What she means by her words is part of the evidence, if we may call it this way, she must draw on. Indeed, as we shall see in detail in the next section, if she is to specify truth-conditions that are meaning-giving, she must actually interact with the speaker linguistically. Thus, I think that we can reconcile Davidson’s saying, on the one hand, that the evidence on which a theory of interpretation is to be constructed “must be describable 12 See

Davidson (1967, 22–23).

136

C. Verheggen

in non-semantic, non-linguistic terms” (Davidson 1974a, 143), and, on the other, that “the idea that the whole evidential base could come from [the relevantly complete physical description of an agent over time] seems to me wrong: some evidence will from the start describe in intentional terms things an agent does” (Davidson 1974b, 282). I take the first remark to emphasize the fact that the philosophical account of meaning is supposed to be non-question-begging. Thus we cannot start building the theory of truth by taking as evidence intensionally described attitudes the speaker manifests or intensionally described responses she makes to her environment. As we saw, what we must start with instead are non-intensionally described utterances of held true sentences and their connections to objects and events in the environment we share with the speaker. The second remark, I think, can be understood to emphasize the fact that, unless the context in which the construction of the theory takes place is also semantic, there is no hope of arriving at a theory of truth that is interpretive. However, the non-reductive aspect of the philosophical account of meaning does not prevent it from being constructive. Central among the conclusions that can be drawn from reflecting on radical interpretation is the kind of perceptual semantic externalism Davidson is advocating, both holistic and historical. As he came to claim, it has always been his view, as he “has for some thirty years been insisting[,] that the contents of our earliest learned and most basic sentences (‘Mama’, ‘Doggie’, ‘Red’, ‘Fire’, ‘Gavagai’) must be determined by what it is in the world that causes us to hold them true” (Davidson 1991a, 200). More generally, it has always been his view that “the contents of our thoughts and sayings are partly determined by the history of causal interactions with the environment” (ibid.).13 The social aspect of this externalism comes to the foreground with Davidson’s writings on triangulation, which also bring out the multiple facets of his non-reductionism.

4 Davidson’s Later Non-reductionism: Triangulation As Davidson came to acknowledge, his writings on radical interpretation do not provide a full answer to the question what it is to for words to mean what they do. As we just saw, they show that the causes of a speaker’s basic utterances play an essential role in determining their meaning, since the radical interpreter has no choice but to consider those causes in trying to understand her linguistic behaviour. She has no choice but to connect the speaker’s utterances to objects and events in their shared environment. And whatever is essential to the attribution of meanings is also essential to their constitution. But the writings on radical interpretation do not spell out just how the specific causes of a speaker’s basic utterances are singled out. A speaker may utter “Gavagai” when no rabbit is in sight, because, for instance, she mistook a cat for a rabbit or she wants to mislead her audience. Moreover, some basic utterances may not be caused by anything currently happening in the vicinity of the speaker. Thus, the 13 Davidson

1997, 71).

expresses this externalism in many places. See, e.g., Davidson (1987, 29; 1988, 44;

The Continuity of Davidson’s Thought: Non-reductionism …

137

meaning-determining causes of basic utterances must be their typical causes: “What determines the contents of…basic thoughts (and what we mean by the words we use to express them) is what has typically caused similar thoughts. But what has typically caused them?” (Davidson 1991a, 201). This is a question Davidson did not address in his early writings.14 He goes on to say that the cause of someone’s response to her environment may in fact be “doubly indeterminate: with respect to width and with respect to distance. The first ambiguity concerns how much [what ‘part or aspect’] of the total cause of [an utterance]…is relevant to [meaning]…The second problem has to do with the ambiguity of the relevant stimulus, whether it is proximal (at the skin, say) or distal” (1999, 129–130). Call these problems, respectively, the aspect problem and the distance problem. Now, it should be made clear that it is the causes of a lifelong solitary person’s responses that, according to Davidson, are doubly indeterminate. Pretty obviously, for the radical interpreter, there is no distance problem to be solved. The causes of a speaker’s basic utterances are to be located, certainly not at the surface of her skin, but in the environment shared by both speaker and interpreter. They are features to which both currently have perceptual access, and thus certainly not elements of the causal chain that led to currently perceived features, such as, say, the birth of the rabbit passing by. In short, for the interpreter at least, the causes of a speaker’s basic utterances are both distal and occurring while the speaker is making the utterances and the interpreter is building a theory of truth for the speaker. However, ascertaining that the causes of a speaker’s utterances are distal is one thing; ascertaining which particular parts or aspects of her environment are causing the speaker’s utterances is another. How is the radical interpreter supposed to do this? Just how is she supposed to make the fine-grained distinctions that are required if she is to understand the speaker correctly? Again, this is a question Davidson never addressed explicitly in his early writings. Moreover, I think that, in the writings on triangulation, he wishes further to extend his enquiry into meaning by shifting his focus from the interpreter’s point of view to the speaker’s point of view. In effect, he wishes to examine how the interpreter solved the aspect problem for herself to begin with. This will in turn shed light on the question how the interpreter is to figure out what aspects of her environment the speaker is responding to. However, this train of thought is not one that Davidson himself ever made explicit. Thus, before addressing the question how the aspect problem is to be solved, I shall pause and retrace chronologically what strike me as the steps of the reasoning. To repeat, early reflection on radical interpretation establishes the general perceptual externalist claim, but it says nothing explicit about how the typical causes of a speaker’s basic utterances are fixed. As a result, it may look as if reflection on radical interpretation leaves room for the possibility of a solitary person having a language, in so far as this is interpretable. However, Davidson explicitly endorsed a social view early in his career, when he first claimed that possession of the concept of belief is needed to have beliefs, and interpersonal communication is needed to 14 Indeed, this is a question most externalists have not addressed. Cf., e.g., Putnam (1975) and Burge

(1986).

138

C. Verheggen

have the concept of belief (Davidson 1975, 170). And he first introduced the idea of triangulation to suggest that engaging in it would yield the concept of objective truth (Davidson 1982, 105). But it was not until he first gave the full triangulation argument (Davidson 1992) that it became clear that he thinks that triangulation is also needed to determine meaning. Indeed, as I see it, it became clear that he thinks that possession of the concept of belief or objective truth15 is essential to the determination of meaning. So, it is not until he first gave the full triangulation argument that he started giving a full account and defence of his social view. This involved addressing the question whether a solitary person could solve the distance and the aspect problems. Presumably, it is because Davidson did not recognize these problems earlier on that the conclusions reached on the basis of radical interpretation were often regarded as compatible with individualism, with the radical interpreter dismissed as a mere “dramatic device” (Glüer 2011, 136),16 and that the writings on triangulation are often alleged to represent a change of mind in that respect. Moreover, making explicit the distance and aspect problems did not make much of a difference to most commentators, since no one was persuaded that solving the distance problem requires a second person, and the aspect problem was by and large ignored, with the result that the essential connection between possession of the concept of belief or objective truth and the determination of meaning was left unacknowledged. So, Davidson’s allegedly new social view was rejected. I agree with commentators that solving the distance problem does not require a second person, but I think that solving the aspect problem does, and this is a problem the radical interpreter obviously has to solve. I thus see the later writings on triangulation as continuous with the early writings on radical interpretation. They demonstrate that Davidson’s view has always been social, and they do this by spelling out how the aspect problem is to be solved. Following Davidson, let us start by considering a lifelong solitary creature, who we imagine is producing regular responses apparently to features of her environment. According to Davidson, the causes of her responses could be anything from stimulations at the surface of her skin all the way to the big bang—which is it going to be? It may be tempting to answer this question by saying, as I suggested above, that it is going to be a feature of her environment that she currently perceives, just as we would answer if asked about the cause of a response we are producing. This singles out a cause that is both distal and contemporaneous. But this tempting answer is given from the interpreter’s point of view. What about the would-be solitary speaker’s? Can the causes of her responses be disambiguated between distal causes and proximal causes, or among several distal causes? Davidson thinks not. He thinks that it 15 Davidson makes no sharp distinction among the concepts of belief, objective truth, objectivity, truth, and error. He certainly thinks that one cannot have one of them without having them all. The basic idea is that of having “the concepts of objects and events that occupy a shared world, of objects and events whose properties and existence is [sic] independent of our thought” (Davidson 1991a, 202), or of being aware of “the contrast between true and false, between appearance and reality, mere seeming and being” (Davidson 1991b, 209), where this awareness may be “inarticulately held” (Davidson 1995, 4). 16 Note that Kathrin Glüer recently changed her mind about this. See Glüer (2018, 227).

The Continuity of Davidson’s Thought: Non-reductionism …

139

is only when a creature has engaged in primitive, i.e., non-linguistic, triangulation17 that distal causes can be isolated. We, on the other hand, may insist that it makes more sense to say that even a solitary creature is responding to currently perceived distal causes, causes that we may share with her, than to say that she is responding to proximal causes or currently inaccessible causes. In other words, we may insist that primitive triangulation may not be needed to solve the distance problem. However, solving the distance problem is of no help in solving the aspect problem, the problem of singling out the aspects of the distal causes that are relevant to determining the meanings of the responses they cause. And it is only when this problem is solved that the meanings of a speaker’s basic utterances are fixed. From our own, third personal and linguistic, point of view, we might again conjecture that the solitary creature is responding to some particular aspects of her environment rather than some others. That is, we might project on her the meanings of the disambiguated responses we would give were we to share her circumstances, though mere projection could not amount to actual interpretation. What reason would we have to think that her responses, however regularly they may seem to be connected to specific aspects of her environment, have in fact the meaning they have for us? More importantly, how could her responses be disambiguated for the solitary creature herself? From her languageless point of view, no aspect in particular will impose itself, as it were. No particular presently perceived aspect will have to be regarded as similar to the aspect that caused her previous response, since no aspect was singled out the first time she produced the response. As Davidson writes, having asked how we can have some idea which causes are being gathered together: “Since any set of causes whatsoever will have endless properties in common, we must look to some recurrent feature of the gatherer, some mark that he or she has classified cases as similar. This can only be some feature or aspect of the gatherer’s reactions…, in which case we must once again ask: what makes these reactions relevantly similar to each other?” (Davidson 2001b, 4–5). This line of thought seems to force on us the following conclusion: as every cause of a would-be speaker’s responses is similar in one respect or another to many others, the only way, it seems, for distal causes, and hence responses, to be disambiguated is by having the would-be speaker herself contribute to this disambiguation. This of course suggests that, since the regular connections that obtain between a would-be speaker’s responses and items in her environment cannot be characterized unambiguously, as long as the characterization is non-semantic, those connections will have to be singled out semantically. In other words, the would-be speaker herself will have to determine, i.e., say, which causes are similar to which. This in turn indicates that a reductionist account of how meanings get fixed cannot be given. Further, if the would-be speaker herself is to determine which causes are similar to which, she must be able to distinguish between what seems to her to be a similar cause and what is in 17 Primitive triangulation is the kind of triangulation even non-linguistic creatures can engage in. It occurs when creatures are reacting simultaneously to each other and to common stimuli in their surroundings, as in Davidson’s example of two lionesses trying to catch a gazelle and coordinating their behaviour by watching each other and the gazelle and reacting to each other’s reactions (Davidson 2001b, 7). Linguistic triangulation is a subset of interpersonal linguistic communication.

140

C. Verheggen

fact a similar cause, where this distinction is per force objective, i.e., independent of what she thinks it is, for fear of annihilating the distinction. To put it differently, she must be able to distinguish between correct and incorrect responses, and she must do this objectively, again for fear of annihilating the distinction. If what she takes to be the correct response, in effect what she takes to be the correct application of an expression, is in fact the correct response or application, then whatever response she gives, however she applies an expression, may be correct, and thus her response or expression not meaningful after all. In short, for the would-be speaker to determine the meaning of her basic responses is for her to fix their causes, and thus for her to fix the meaning of her expressions is for her to fix their conditions of correct application, and this requires possession of the concept of objectivity. Davidson argues that only someone who has triangulated linguistically could be in a position to have the concept of objectivity. The circularity of this claim (in addition to that of the claim broached above) places Davidson’s non-reductionism in the foreground. Someone cannot have a language unless the meanings of her expressions are fixed, and this requires her to have the concept of objectivity. But someone cannot have the concept of objectivity unless she has triangulated linguistically and thus has a language. Davidson’s claim is of course that possession of a language, involving the use of words with fixed meanings, and of the concept of objectivity are intrinsically connected—one cannot have one without the other. It is not as if one gets one and then develops the other; his is not an account of language acquisition, as is to be expected from a non-reductionist account. However, whereas one can make intelligible the idea of triangulating people having both, one cannot make intelligible the idea of a solitary person having either. The main difference between the two cases is this. In a triangular situation, “two (or more) creatures each correlate their own reactions to external phenomena with the reactions of the other. Once these correlations are set up, each creature is in a position to expect the external phenomenon when it perceives the associated reaction of the other. What introduces the possibility of error is the occasional failure of the expectation; the reactions do not correlate” (Davidson 1999, 129). Thus we can understand how people triangulating linguistically with one another are in a position to have the concept of objectivity because we can imagine them disagreeing about some feature of the environment they are talking about and resolving the disagreement in a way that does not depend on the mere say of one or the other interlocutor. Being able to engage in disputes of this kind makes room for their having the idea of different perspectives on the world around them and the idea of this world being as it is independently of how they think it is. But there is no similar situation the solitary person could be in. Even if we were to grant her a language and imagine her in situations where her expectations were not fulfilled and thus somehow disagreeing with herself, the fact of the matter is that she would always have the one and only say as to the resolution of the “disagreement”. Whatever would seem to her to be the truth of the matter would be the truth of the matter, making it impossible for her responses to be governed by conditions of correctness, and thus making it impossible for her to have a language after all. As Davidson put it, “[t]he central argument against [solitary]

The Continuity of Davidson’s Thought: Non-reductionism …

141

languages is that, unless a language is shared, there is no way to distinguish between using the language correctly and using it incorrectly” (Davidson 1991b, 209).18, 19

5 The Continuity of Davidson’s Thought We have encountered three interrelated ways in which Davidson’s writings on triangulation can be seen to be continuous with those on radical interpretation: first, he has always held a social view of language; second, he has always thought that meaning could not be fully explained in reductive, non-semantic terms; and, third, he has nonetheless always offered a constructive account of meaning. Davidson has always held a social view of language not simply because he has always said that one could not have beliefs, and hence a language, unless one had the concept of objectivity,20 and one could not have the concept of objectivity unless one had interacted linguistically with others. What becomes clear, with the focus on the problems, in particular the aspect problem, that the interpreter has to solve in order to be able to understand an alien speaker, is that the interpreter has to engage in linguistic triangulation with the speaker. The interpreter cannot otherwise solve the aspect problem, any more than any one developing a first language could solve the problem on her own. As Davidson writes: the gearing of verbal (and other) responses to situations, events, and objects through the prompting and perception of others plays a key role in the acquisition of a first language, or the learning of a second language in the absence of an interpreter or bilingual dictionary…Learning a first and learning a second language are, of course, very different enterprises…Both, however, depend on similar mechanisms and similar cues. (Davidson 1998, 88)

Radical interpretation thus is an instance of triangulation: [one of the situations in which] triangulation bears on the determination of the contents of our perceptual beliefs [and utterances]…is that in which two participants are equipped with thought and language, but lack a common language. The problem is for each to understand the other: the problem of radical interpretation. (Davidson 2001c, 294)

What the writings on triangulation further reveal and develop is the thoroughly non-reductive character of Davidson’s philosophical account of meaning. With radical interpretation, we saw that the semantic theory an interpreter builds for a speaker cannot specify the meanings of the speaker’s expressions unless the theory is built within a semantic context. With triangulation we further see that nonsemantically characterized connections between a speaker’s responses and items 18 See

also Davidson (1997, 83). that the triangulation argument shows that meaning is essentially social, but not essentially communitarian. That is, it shows that, in order to have a language, one must have understood and been understood by others, but one need not mean the same thing as others by the same words. I discuss this further in Myers and Verheggen (2016, Chap. 3). 20 See, e.g., Davidson (1995, 7; 1999, 124). 19 Note

142

C. Verheggen

in her environment causing these responses could be not be used to specify the meanings of the speaker’s expressions. Indeed, as long as these responses are not shared and expressed in linguistic exchanges, their meanings cannot be fixed. Thus, in the absence of linguistic triangulation, they cannot be fixed for the interpreter, and they could never be fixed for a solitary person. The answer to the fundamental, philosophical question of what it is for words to mean what they do is therefore unabashedly non-reductionist: it is for some words at least21 to have been used in certain ways, that is, to mean certain things, in linguistic triangulation. But this is no quietist non-reductionism. Indeed, in addition to putting the claim that language is essentially social onto firmer ground, reflection on triangulation reinforces the claim that the kind of externalism that is required for the constitution of meaning is both holistic and historical.22 Reflection on radical interpretation first emphasized, rather than demonstrated, the externalist aspect of meaning, for it might be thought that when deciding to reflect on radical interpretation externalism is to some extent assumed, since the thought experiment is premised on the assumption that meaning is essentially public. Reflection on triangulation vindicates these initial assumptions, for it shows that triangulation is needed for meanings to be fixed, and only items that are external to speakers can be triangulated upon. It also vindicates Davidson’s early assumption that possessing the concept of objectivity and using words meaningfully necessarily go hand in hand. For the writings on triangulation make clear that, in order to have a language and thoughts, one must possess the concept of objectivity not just in addition to one’s meanings having to be fixed or determined, but possession of the concept plays an essential role in the determination of meaning to begin with. Contra what most commentators on Davidson have maintained,23 triangulation is needed not first to determine the meaning of one’s words, and second to put one in a position to have the concept of objectivity. Rather triangulation accomplishes both tasks at once.24 All in all, this is quite an impressive list of constructive answers to the question of what it is for words to mean what they do, even though, in the end, the primitiveness of what words mean remains untouched.25

21 Not every word or concept that refers to an external item must have been used in triangulation for its meaning to be fixed. For instance, “the contents of the belief that a guanaco is present [may have been] determined, not by exposure to guanacos, but by having acquired other words and concepts, such as those of llama, animal, camel, domesticated, and so forth. Somewhere along the line, though, we must come to the direct exposures that anchor thought and language to the world” (Davidson 1991a, 197). 22 I have said more on the continuous aspects of Davidson’s externalism in Myers and Verheggen (2016, Chap. 3). 23 See Talmage (1997), Pagin (2001), Lepore and Ludwig (2005), Bridges (2006), Glüer (2006), Ludwig (2011) and Bernecker (2013). 24 Understood in this way, the triangulation argument becomes much less vulnerable to objections. I have said more about this in Myers and Verheggen (2016, Chap. 1). 25 Thanks to Pedro Rui Abreu, Robert Myers, and Olivia Sultanescu for their comments on an earlier draft of this paper.

The Continuity of Davidson’s Thought: Non-reductionism …

143

References Bernecker, S. (2013). Triangular externalism. In E. Lepore & K. Ludwig (Eds.), A companion to Donald Davidson. Wiley Blackwell. Bridges, J. (2006). Davidson’s transcendental externalism. Philosophy and Phenomenological Research, 73(2), 290–315. Bridges, J. (2017). The search for ‘the essence of human language’ in Wittgenstein and Davidson. In Verheggen 2017a. Burge, T. (1986). Cartesian error and the objectivity of perception. In P. Pettit & J. McDowell (Eds.), Subject, thought, and context. Oxford: Oxford University Press. Davidson, D. (1967). Truth and meaning. In Davidson 1984. Davidson, D. (1973). Radical interpretation. In Davidson 1984. Davidson, D. (1974a). Belief and the basis of meaning. In Davidson 1984. Davidson, D. (1974b). Replies to Lewis and Quine. In Davidson 1984. Davidson, D. (1975). Thought and talk. In Davidson 1984. Davidson, D. (1982). Rational animals. In Davidson 2001a. Davidson, D. (1983). A coherence theory of truth and knowledge. In Davidson 2001a. Davidson, D. (1984). Inquiries into truth and interpretation. Oxford: Clarendon Press. Davidson, D. (1987). Knowing one’s own mind. In Davidson 2001a. Davidson, D. (1988). The myth of the subjective. In Davidson 2001a. Davidson, D. (1991a). Epistemology externalized. In Davidson 2001a. Davidson, D. (1991b). Three varieties of knowledge. In Davidson 2001a. Davidson, D. (1992). The second person. In Davidson 2001a. Davidson, D. (1993). Reply to Jerry Fodor and Ernest Lepore. In Stoecker 1993. Davidson, D. (1995). The objectivity of value. In Davidson 2004. Davidson, D. (1997). Indeterminism and antirealism. In Davidson 2001a. Davidson, D. (1998). The irreducibility of the concept of self. In Davidson 2001a. Davidson, D. (1999). The emergence of thought. In Davidson 2001a. Davidson, D. (2001a). Subjective, intersubjective, objective. Oxford: Clarendon Press. Davidson, D. (2001b). Externalisms. In Kotatko et al. 2001. Davidson, D. (2001c). Comments on Karlovy Vary papers. In Kotatko et al. 2001. Davidson, D. (2004). Problems of rationality. Oxford: Clarendon Press. Dummett, M. (1975). What is a theory of meaning? In S. Guttenplan (Ed.), Mind and language. Oxford: Clarendon Press. Fodor, J., & Lepore, E. (1993). Is radical interpretation possible? In Stoecker 1993. Ginsborg, H. (2011). Inside and outside language: Stroud’s nonreductionism about meaning. In J. Bridges, N. Kolodny, & W. Wong (Eds.), The possibility of philosophical understanding: Essays for Barry Stroud. Oxford: Oxford University Press. Glüer, K. (2006). Triangulation. In E. Lepore & B. Smith (Eds.), The Oxford handbook of philosophy of language. Oxford: Clarendon Press. Glüer, K. (2011). Donald Davidson: A short introduction. Oxford: Oxford University Press. Glüer, K. (2018). Interpretation and the interpreter. In D. Ball & B. Rabern (Eds.), The science of meaning: Essays on the metatheory of natural language semantics. Oxford: Oxford University Press. Hattiangadi, A. (2017). The normativity of meaning. In B. Hale, C. Wright, & A. Miller (Eds.), A companion to the philosophy of language (2nd ed.). Wiley. Kotatko, P., Pagin, P., & Segal, G. (Eds.). (2001). Interpreting Davidson. Stanford: CSLI. Lepore, E., & Ludwig, K. (2005). Donald Davidson: Meaning, truth, language, and reality. Oxford: Clarendon Press. Ludwig, K. (2011). Triangulation triangulated. In C. Amoretti & G. Preyer (Eds.), Triangulation: From an epistemological point of view. Frankfurt: Ontos Verlag. Myers, R. H., & Verheggen, C. (2016). Donald Davidson’s triangulation argument: A philosophical inquiry. New York and London: Routledge.

144

C. Verheggen

Pagin, P. (2001). Semantic triangulation. In Kotatko et al. 2001. Putnam, H. (1975). The meaning of ‘meaning’. In his Language, mind and reality. Cambridge: Cambridge University Press. Stoecker, R. (Ed.). (1993). Reflecting Davidson. Berlin, New York: Walter de Gruyter. Stroud, B. (2012). Meaning and understanding. In J. Ellis & D. Guevara (Eds.), Wittgenstein and the philosophy of mind. Oxford: Oxford University Press. Stroud, B. (2017). Davidson and Wittgenstein on meaning and understanding. In Verheggen 2017a. Talmage, C. (1997). Meaning and triangulation. Linguistics and Philosophy, 20, 139–145. Verheggen, C. (2015). Towards a new kind of semantic normativity. International Journal of Philosophical Studies, 23(3), 410–424. Verheggen, C. (Ed.). (2017a). Wittgenstein and Davidson on language, thought, and action. Cambridge: Cambridge University Press. Verheggen, C. (2017b). Davidson’s semantic externalism: From radical interpretation to triangulation. Special issue for the Centenary of Donald Davidson’s Birth, C. Amoretti, M. de Caro, & F. Ervas (Eds.), Argumenta, 3(1), 145–161.

Methodological Considerations in the Triangulation Argument Wai Chun Leong

Abstract In his later works, Davidson argues that an agent cannot have any propositional thought if it has never interacted with another agent in a certain way. This is later known as the triangulation argument. However, this argument is very controversial and not many philosophers accept it. One of the reasons is that there is no consensus on how exactly the argument goes. In this chapter, I try to reformulate this argument, and special attention is paid to the methodological considerations on relevant similarity, which I believe, are often being overlooked in current literature. A tentative modification of the kind of interactions required is also given. Keywords Triangulation argument · Objectivity of thought · Intentionality · Standard of correct application · Joint commitment According to Davidson, a propositional thought has two essential features: objectivity and intentionality. A thought is objective in that it is true or false, and a thought is true or false not because someone believes or disbelieves it. To say this does not imply that every thought has a definite truth value; some may never have one (e.g. the thought that the present king of France is bald may be taken as truth-valueless by some philosophers), but we can at least conceive a situation in which it is true (e.g. the situation in which there is a bald king of France whose existence is not known by many). What is essential to a thought is that the concept of truth applies to it; that is, a thought can be true or false. If we take concepts to be constituents of a thought, concepts are also objective in the sense that they can be correctly or incorrectly applied to objects or events. In the simplest form, a concept Fx is applied correctly to an object a if and only if the thought Fa is true. Whether a thought is true depends primarily on two factors: the propositional content of the thought and the way things are. Thus, we can say that associated to a concept is a standard of correct This chapter is in its final form, and it is not submitted for publication anywhere else. W. C. Leong (B) Department of General Education, Macau University of Science and Technology, Avenida Wai Long, Taipa, Macau, China e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_9

145

146

W. C. Leong

application: whether a concept is correctly applied to an object depends in part on this standard, and this standard of correct application at least partially determines the propositional content of the thought. So the kind of correctness that is relevant to such a standard is the kind of correctness that is relevant to the truth or falsity of a thought. A thought is intentional in that it is about something in reality, so are the concepts that constitute the thought. By contrast, a disposition is not about anything. One cannot say, for example, that a disposition is about what typically causes its triggering. The concept of a cat is about those little creatures in the world. However, one’s startling caused by a cat is not about the cat. Davidson uses the word “disposition” in a very broad sense; it refers not only to physical disposition and mindless instinctual or reflexive responses, but it could also mean any mental activity which does not involve concepts, if we allow such an activity to be called mental. Davidson argues that without certain interactions with another agent, an agent cannot have any propositional thought because the objectivity and the intentionality of thought cannot be explained. This argument is commonly known as the triangulation argument (henceforth: TA). How the argument proceeds exactly is rather elusive. Although it plays a central role in Davidson’s philosophy, the argument seems to have undergone stages of development and refinement in the last decade of his works. To begin with, the argument is sometimes presented as two self-contained arguments, especially in the early stage of the development as the two arguments are not always mentioned together.1 In his later writings, it appears that the two arguments are interdependent and should not be considered separately.2 Moreover, it is often mentioned as a part of a larger argument to other philosophical problems, and thus scattered in various papers on different topics. Depending on the topic, different aspects of TA are being emphasized. However, Davidson later devotes an entire paper, “Externalism” (Davidson 2001a), to TA, but perhaps because it is not included in the five volumes of his collected works, it is relatively neglected by philosophers. It is in this work, I believe, that TA is presented in its most detailed form, and as a unified solution to the issues raised by the two features of thought in question. I, therefore, base my formulation of TA primarily on this paper and treat it as a single argument. The current paper has two aims. I first explain the problems raised by the objectivity and intentionality of thought and introduce TA and explain how it argues that certain interactions between two agents are necessary to an account of these two features. I will also try to identify the assumptions and problems that remain in TA. I will first explain why Davidson thinks that these two features raise problems to an explanation of thought considering only a solitary agent. A solitary agent is not an agent temporarily being cut off from other agents: it is an agent who has never in its life interacted with another similar agent. Then I will explain the kind of interactions required of two agents in order for the agents to engage in a genuine triangulation scenario, before I formulate what I think to be Davidson’s triangulation argument.

1 See, 2 See,

for example, Glüer (2006) and Lepore and Ludwig (2005). for example, Verheggen (2007) and Myers and Verheggen (2016).

Methodological Considerations in the Triangulation Argument

147

1 Objectivity of Thought: The Kripkensteinian Challenge Davidson thinks that Kripke is correct in arguing that having a thought is not having a mere disposition. While a thought is objective in that it can be true or false, a mere disposition is not. One can have a disposition, which can be triggered, but it cannot be true or false. An account of thought must therefore explain how a false thought or belief is possible; otherwise, an essential difference between having a mere disposition and having a thought is left unexplained. To explain errors, therefore, an account of thought must also involve more than an agent’s dispositions. A mere failure to give a response similar to that which the agent has given under similar conditions does not immediately render his deviant response an erroneous one: such a failure may also occur when the agent applies a different concept correctly, or, if the agent does not yet have any concept, when some other conditions of the environment or of the agent himself are not fulfilled (e.g. the agent’s sight is blocked by an object, or the agent is unconscious), thus an error cannot be explained or identified as such a failure. To illustrate, let us consider the differences between a mere disposition and a response that involves conceptual content. Davidson (2001a, pp. 2–4) draws on Kripke’s interpretation of Wittgenstein’s private language argument (Kripke 1982), or KA (Kripkensteinian Argument) for short. According to KA, a response that involves conceptual content cannot be reduced to a mere disposition, for the former has an essential normative element which the latter lacks. What then is this normative element? To illustrate, consider a person who makes the following utterances: 2 @ 2 = 4, 5 @ 7 = 12, 9 @ 13 = 22, 12 @ 25 = 35, 17 @ 31 = 48. Suppose for the moment that this person means the same thing by this symbol “@” in these instances (This supposition, however, violates a methodological constraint of TA, which I will come back in the next section.) Nevertheless, the problem KA rises remains even if this supposition is made: can we say that this person means “addition” by this symbol, so that he made an error in one of these instances? Or can we say instead that he just means something else, something similar to “addition”, such that he correctly applied this concept in all of the instances? Or can we perhaps even say that none of the above is a correct application of the concept he uses, if we take this person to mean something very different from “addition”? The problem is that, given that a concept can be applied incorrectly, no matter what dispositions this person has shown in the past, there is no definite reason to prefer one meaning over another if only the person’s dispositions are considered. All these dispositions can be consistent with any concepts we take this person to mean, as long as we take him as committing an error in the appropriate instance. The list of utterances can even be

148

W. C. Leong

extended indefinitely, but the problem will remain, so the lack of enough utterances is not the essential cause of the problem. The person’s dispositions simply do not imply any constraint strong enough for the determination of the standard of correct application of the concept expressed by “@” in those instances. One remark about KA. KA is not an epistemological argument challenging our justifications to take a response as involving particular concepts. It is a metaphysical argument asking for what determines the fact that a response involves a particular concept rather than another. Although one might take KA as presented originally by Kripke as a skeptical argument for the conclusion that there are no such facts of meaning determination, such as the fact that a certain utterance or sentence means that p, it is not the reason Davidson brings in KA to TA. One misses the point if one tries to defend that given enough observations of an agent’s responses under relevantly similar situation, it will become plausible to say that the agent has a particular concept rather than another. The argument asks for the metaphysical features that determine the concepts involved in a response. To say simply that it is more plausible that a response involves a certain concept rather than another, at best, is to say that there are such features of the response pertaining to, but not specific to, that concept, because those further observed features may be consistent with other concepts. This defense does not answer the question raised by TA. TA does not deny that such features exist, at least not when such features involves more than a single agent. The argument asks for what these features could be. One might contest, as Forbes (1983) seems to do, that an error can be explained by a disposition attributed to an agent. When we say that salt will dissolve in water ceteris paribus, we are adverting to a possibility admitted by this conception, the possibility of interference with the interaction between the relevant state of salt and that of water, whatever these states may be…. (Forbes 1983, p. 230) …if someone has just forgotten to carry, then ceteris paribus he should change his mind about the answer when reminded, or when the problem is given to him again with the sum of the digits in the troublesome column computed separately and the left digit of the answer highlighted. … So the dispositionalist need have no difficulty at all with the idea that one man may intend ‘+’ to stand for addition and make mistakes, while another may mean some non-standard function by ‘+’, even though their ‘+’-behaviour is the same throughout a fixed sequence of tests, for he knows how he could go about confirming which category it is to which a given subject belongs. (Forbes 1983, p. 232)

When we attribute a disposition to an agent, a ceteris paribus clause is associated with it. For instances, I have a disposition to retract my hand if, ceteris paribus, I put my hand on something hot. But when I am unconscious, this response (retracting my hand) would not be triggered under the same conditions. The failure of the triggering of this response while I am unconscious would not nevertheless be a reason to say that I do not have the disposition, for being conscious is one of the normal conditions specified by the ceteris paribus clause. Forbes therefore suggests that a response given by an agent is an error if and only if the response is triggered when the ceteris paribus clause of the disposition we attribute to the agent is not satisfied, but otherwise he would have given a response conforming to the disposition attributed to him under normal conditions. Thus, an agent responds to the question “64 + 32 = ?” by saying

Methodological Considerations in the Triangulation Argument

149

“87” is an error because, say, he misread “32” as “23”, so the ceteris paribus clause is violated, otherwise he would have answered “96”. However, it does not seem to solve our initial problem. Our problem concerns the reasons in virtue of which an agent has a particular disposition formulated in the above way, instead of the explanation of errors with respect to a given disposition. Specifically, it is not a problem that can be solved by merely specifying the underlying physical or biological mechanism that realize the disposition, because it is always possible that the agent has a certain disposition but does not have the concept in question. What is needed is the explanation of the missing link, so to speak, between this concept and the mechanism/the disposition specified with a ceteris paribus clause. One must explain this missing link in a way such that this link and the disposition together determine the standard of correct application of the concept of the agent. While we may have some underlying explanation for the triggering of a disposition (e.g. the brain of the agent is so trained that when certain signals are received by the ear certain movement signals will then be sent to the mouth), an explanation connecting a bodily status and a mental status that could determine a standard of correct application is still missing. Furthermore, an agent’s past responses do not determine a particular disposition and excludes all others, even with a ceteris paribus clause attached. Given a set of responses from an agent, one can always find various dispositions with different ceteris paribus clauses (i.e., the conditions that is considered normal to these dispositions are different) that are compatible to this set of responses. The problem seems to be that the normal conditions specified are applicable to infinitely many instances, while the responses of any given agents are always finite. So finite occurrences of responses do not seem to determine a disposition specified in this way. Besides, there does not seem to be any convincing reason why we must take the already occurred responses of an agent as all under normal conditions. Even though I have not touched anything hot, my disposition of retracting my hand could be triggered due to a misfiring of some signals caused by brain damage. In other words, we do not seem to have reasons to take all the previous responses of an agent as “correct”. If this is the case, even more possible dispositions associated with different conditions which are considered normal can be attributed to the agent. Together with the problem mentioned in the previous paragraph, it seems that our problem cannot be solved by simply manipulating the ceteris paribus clause attached to a disposition. If mere dispositions cannot explain the normative element of thought, what else can?3 One proposal might be to appeal to the rule or procedure the agent tries to follow when he makes the above calculation. An agent maybe said to be implicitly or explicitly following a procedure in figuring out an answer. For example, when working on the question “12 @ 25”, one might put the last digit of the two numbers together and start from 5 and count two more times thus get to the number 7, and 3 It

should be emphasized that when Davidson discusses KA, he is considering a very thin sense of normativity, that is, the standard of correct application of a concept and the truth condition of a thought. Another way to ask this question would therefore be: If mere dispositions cannot explain the correct or incorrect application of a concept, or the truth or falsity of thought, what else can?

150

W. C. Leong

then do the same to the tenth digit of the two numbers and get 3. Putting the two numbers together, we therefore have 37 as an answer. But the same challenge can be raised with regard to this procedure or rule that the agent tries to follow. What determines the agent’s understanding of counting as counting with the sequence of natural numbers instead of another similar sequence of numbers but not exactly the same? Therefore, appealing to further procedures or rules, it seems, simply pushes the problem up one level. Or can the feels one has whenever he understands this symbol “@”, or whenever he tries to use the same concept, determined this standard of correct application? For example, when we try to read a text, we tend to subvocalize the words we read by our eyes, as if we are articulating it out for someone else, just doing it in one’s own mind. We can easily imagine another phonetically indistinguishable language whose semantics are totally different. In fact, whatever feel one has, it is logically possible that this feel is associated to another concept or standard of correct application. So appealing to inner feels is just another infeasible approach. It appears, therefore, that considering only a single agent is not a promising approach to explain the standard of correct application of the agent’s concept. More importantly, Davidson thinks that a methodological problem makes it impossible to give an account of error if we consider only a single agent. If we cannot give an account of error, we cannot explain the standard of correct application of concept. A similar methodological problem also makes Kripke’s original solution to the problem being raised in KA unsuccessful. This methodological problem, therefore, is considered in the next section.

2 Similarity: A Methodological Problem Before considering the solution provided by TA, some considerations on a methodological problem of similarity are necessary. To illustrate, consider the solution given by KA to the above issue. KA states that the linguistic community to which the agent belongs provide the standard of correct application for the agent’s responses. Thus, if typical members of the community answer “37” when asked “12 @ 25”, then the agent’s answer of “35” would be an incorrect answer, because that is not how typical members of the community understand the “@” sign. However, if typical members of the community provide an agent the standard of correct application for his responses, then what, in turn, provides the community with its standard of correct application? If nothing can provide the community with such a standard, however the typical members of the community respond in a novel situation, as long as they respond in a similar way, their responses would be correct. The initial problem would then be merely pushed back onto the level of the community (Boghossian 1989). On the other hand, just as Davidson asks: “[H]ow can the simple fact that two or more people have gone on in the same way introduce the distinction between following a rule and just going on in one way or another? All the sunflowers in the field turn together to face the sun: are they following a rule? Surely not

Methodological Considerations in the Triangulation Argument

151

in the required sense; they have no concepts and so cannot misapply concepts”. (Davidson 2001a, p. 3)

The uniformity of the responses of the typical members in a novel situation does not explain the conceptual content of their responses; otherwise, a field of sunflowers turning together towards the sun would imply that these sunflowers have thoughts about the sun. What is needed is an explanation of the standard of correct application of the agents’ responses, which the mere uniformity of the responses of multiple members in a group cannot provide us.4 More importantly, the assumed similarity of responses of this solution from KA involves a methodological issue: I have emphasized at a number of points the way the notion of relevant similarity can play an essential role without being noticed: thus Kripke, in the attempt to say when a creature’s subsequent response was relevantly similar to an earlier response, helped himself to the notion of the relevantly similar responses of another creature. Burge helped himself to the idea of relevantly similar stimuli. The reason it is so natural to avail ourselves of such concepts is, of course, that they are ours. The trouble is, we move in a circle when we appeal to what is natural for us in order to explain conceptualization itself—to explain, in other words, what is natural for us. (Davidson 2001a, p. 8)

When we say that two things are similar, we often mean that they are similar in a certain aspect. For any two responses, there are always some aspects in which they are similar. They may be both produced by human, uttered in a similar tone, made in similar places, in similar speed, etc., but many of these aspects are not relevant to our issue. The solution from KA requires that two responses be similar in those aspects so that, if two responses are similar in these aspects, one of them can serve as a correct response to which the other response can be compared for its correctness. Similarity in other aspects are then irrelevant. So the two answers of “thirty-seven” and “37” to the question “12 + 25 = ?” are relevantly similar, and if typical member of the English speaking community answer “37” while an agent says “35”, the former can serve as a standard of correct application for the latter and, since the two answers are not similar in the relevant aspect, the agent’s answer is incorrect. This example seems plausible because we assume that “thirty-seven”, “37”, and “35” are verbal responses that refer to two numbers, 37 and 35. But if we make this assumption, it means that we have already assumed that these responses are similar in the aspect that they express the same concepts, or in the case of “35”, a different concept. With this assumption, the solution provided by KA would not be helpful; the solution assumes that the two responses already express certain concepts so that we could compare them, and this would make the whole solution beg the question. So what KA needs is a way to specify these relevant aspects without involving the concepts these responses express. By this I do not mean that KA needs to specify a general aspect so that if two responses are similar in this aspect, they are relevantly similar (that is, similar in that they express the same, or a different, concept.) For instance, if we think of verbal responses, similarity in phonetics does not meet what KA requires, although it is an obvious aspect one might think of. It is neither sufficient 4 See

also Blackburn (1984).

152

W. C. Leong

nor necessary for the similarity of the concepts expressed by verbal responses. It is not sufficient because, besides phonetically ambiguous words, even people in the same linguistic community usually have their own idiolects which are phonetically similar to each other’s, yet the same word could express different concepts for different people. It is not necessary because a typical member in a linguistic community could have his own peculiar accent but words he uses could express the same concepts as his peers would express with the same word. But even if KA requires not the similarity in a general aspect but different aspects for different pairs of concepts and responses, a more serious methodological problem remains. Two responses may be similar in the required sense; i.e., they express the same concepts for the agent, while it is not obvious to us that they are similar, because responses which are not similar in our eyes, may be similar to the agents, and we may also find two responses similar while the agent does not. We cannot, therefore, assume that whatever is similar to the agent in certain situations, we would also find similar, and vice versa. Thus, our finding the agent’s responses similar in certain aspects does not imply that the responses are similar in the relevant sense, that is, similar in that they have the same propositional content, if they have any. An explanation of the relevant similarity of responses must, therefore, take into account what makes the responses relevantly similar. In KA, it is simply assumed that the relevant similarity of the responses of the typical members are given. The same applies to the stimuli of the agents. We cannot assume that two stimuli are relevantly similar to an agent simply because we find them similar. To use the above example, we cannot assume that the two agents understand the question “12 + 25 = ?” in the same way since one may understand it as a simple arithmetic question while the other may take “12” and “25” as referring to two political figures and therefore take the question as political. I will come back to this problem in the next section. Besides, what is an aspect? When we think of an aspect, we usually think of something that can be described with a concept. If it is something that can only be described with a concept, then the problem of specifying the relevant aspect is simply the problem of specifying the relevant concept. Bringing in this kind of aspects does not seem helpful to the current issue. If, instead, we think of an aspect of responses that cannot be specified by any concept, then it seems doubtful to say that it is an aspect at all, since it would be difficult to imagine what is relevant or irrelevant to this aspect, in a way that being red is relevant to the aspect “colour” while being circular is not. One last possibility is that there may be some aspects that are specifiable with concept, but one could entertain it without employing any concept. For instance, the colour red is specifiable with the concept RED, but one could specify this colour by pointing to another instance of a similar colour and say, “this”. However, besides the original aspect problem pointed out by Wittgenstein, whether this pointing and utterance are relevant in our sense again incurs our initial problem: we cannot simply assume that the pointing or utterance are similar to previous ones in a relevant way, so that they express the same concept or propositional content, if they have any. It seems, therefore, involving similarity of aspect does not help with our problem.

Methodological Considerations in the Triangulation Argument

153

This methodological problem requires also that an explanation of the similarity of responses and stimuli must be observer independent, in the sense that the explanation does not depend on judgements of similarity from the perspective of a mere observer. For one observer may find certain responses or stimuli similar while another observer may not find those responses or stimuli similar. The two observers would then attribute different concepts to the agent of the responses, so that a response is correct for one observer while incorrect for another. Thus, a mere observer’s judgement of the similarity of an agent’s responses or stimuli cannot be an element that determines the standard of correct application of the agent’s concept.

3 Intentionality and Distal Causes What makes a concept or a thought about what it is about? Our concept of, say, table, is a concept of those objects which we call “table” instead of the stimulus on our sense organs caused by them. What makes a concept or a thought about anything in the world? Consider a solitary agent A, who has never interacted with any other agents, and a single instance of his responses. Can one say that O, an object in the world which is also one of the many causes of his response, is what A responds to? Davidson’s answer is no, because of two ambiguities generated by the concept of cause: one in terms of width, one in terms of distance: The first ambiguity concerns how much of the total cause of a belief is relevant to content. … The second problem has to do with the ambiguity of the relevant stimulus, whether it is proximal (at the skin, say) or distal. (Davidson 1999, p. 13)

In terms of width, nothing could determine which of the many causal chains contributing to the total causes of the agent’s response are relevant to his response. The causes of his responses are numerous and nonlinear (e.g. causes of the conditions of the ambient lights, of the freedom from blockage between O and the agent, of the agent’s visual system, of the well-functioning of the agent’s brain, etc.). There are no particular reasons to take one of these causes as what the agent responds to; each of these causes is but a link in a web of causal chains, and nothing about any of them could determine itself as what the agent responds to. Conversely, the concept of O is about the object O, but the total cause of a belief that is partially formed by the concept of O includes plenty of causes that are not even causally related to what the concept or the belief is about. In terms of distance, there are many distal causes and proximal causes in a causal chain which together lead from what a concept is about to the agent’s response. We can say that the agent responds to O, but we can also say that he responds to the proximal stimulus on his, say, retina, which is caused by O, and there seems to be no reason for us to say the former instead of the latter.5 5 Some

critics think that we do have reasons to say that the agent responds to the distal instead of the proximal cause. See, for example, Bridges (2006) and Burge (2010).

154

W. C. Leong

The difference between these two kinds of ambiguities can be seen in the following example. Suppose an agent always gives similar responses to a cow then the total causes of his responses include, mostly likely, the ground on which a cow stands, simply because the ground supports the presence of a cow. If his responses are about cows, not the ground on which a cow stands, then the problem of the ambiguity of width concerns how to exclude the latter from the former as the relevant cause of the agent’s responses. When we succeed in pinpointing cows as causes that are relevant to what his responses are about, we still need to identify cows (distal causes) instead of, say, the stimuli they cause on the retinas of the agent (proximal causes), as what his responses are about. This is the problem of the ambiguity of distance. When multiple instances of the responses of A are considered, we could appeal to the principle that similar stimuli elicit similar responses and try to find out the relevant stimuli by finding out the similar causes in cases where A gives similar responses. It may seem then that the similar causes in these instances would be what A responds to in these instances. However, the methodological problem of similarity affects this proposal too. In order to identify which responses are similar in the relevant way, we must first solve the methodological problem of relevant similarity of responses mentioned above. Even if this problem is solved, there is still a problem about the relevant similarity of stimuli. Not every cause that is similar in a given aspect is relevant to the propositional content of the agent’s response. The aspects in which two causes are similar may be irrelevant to why the agent makes certain responses, and the agent may not even notice these aspects at all. We cannot assume, in addition, that what we find similar in two occasions are also relevantly similar to the agent. For example, in the English-speaking world, blue and green are usually taken as two different colours. But Japanese speakers have a colour term, “ao青”, which basically covers both blue and some parts of green, so it seems that Japanese speakers have a different colour concept. So there does not seem to be a unique classification of colours we have to follow. Our concepts of colours seem to be determined partially by reasons other than features of the colours themselves, as colour concepts vary from culture to culture. Although there seems to be biological reasons why we classify things in certain ways (e.g. we classify tigers together and exclude cats from it), these reasons do not explain individual difference or even cultural differences of systems of concepts (e.g. different cultures have different colour concepts; i.e., they classify colours differently). Non-biological reasons thus seem to play a role too in the determination of our concepts. Therefore, we cannot assume that an agent must classify things in the way an observer does. So it seems that whether two things are relevantly similar to an agent depends on the agent’s response to them. In other words, whether two things are relevantly similar is entirely up to the agent, in that no matter what responses the agent gives, there is no standard or criteria for us to say that he makes a mistake, even if the agent does follow a standard or criteria. One might think that the problem is not as serious as it sounds, since we only need to find out whether it is, say, the colour or the shape of an object that an agent attends to when he finds two things similar. But the range of possible aspects to which an agent attend is much wider, we cannot assume that the agent attends to an aspect which is readily specifiable with concepts

Methodological Considerations in the Triangulation Argument

155

familiar to us. Theoretical speaking, the extension of a concept may consist of any combination of objects. One could have a concept whose extension includes several chairs, a number, a few kinds of animals, or what have you. Similarly, what an agent finds similar may appear totally arbitrary to us, but even so, we are not in a position to say any of the agent’s responses are incorrect, for we have no criteria to say whether the agent is applying a concept incorrectly, or applying a different concept correctly. In this sense, therefore, whether something is similar to an agent is entirely up to the agent. To clarify, the above consideration is not to try to challenge the intuition that human beings find more or less the same things similar. Instead, it is to argue that in explaining what determines the content of an agent’s concepts, we cannot explain what is similar to the agent in an observer dependent way, be it for the agent’s responses or what the agent responds to, or simply assume a correct classification of what an agent finds similar. For example, suppose for objects a, b, and c, an agent A gives the responses Ra, Rb, and Rc. Suppose also that a theory T 1 takes Ra and Rb as similar, while Rc not, so T 1 attributes to A a concept C 1 that has a and b, but not c, in its extension. In T 1 , what justifies the attribution of C 1 to A is the similarly of a and b, which is explained by the similarity of Ra and Rb. But what, in turn, explains the similarity of Ra and Rb to A? T 1 cannot simply assume that Ra and Rb are similar to A. For other theories could classify A’s responses in other ways and, as a result, classify what are similar to A differently. If a theory simply assumes that a certain classification of responses is correct, it begs the question by assuming partially the extension of the concept the agent has or will come to acquire, through classifying instances that are relevant to the concept. If one defends that a certain classification is the most natural to us, as observers, then what A finds similar would then depend on what is natural to observers, and, as discussed above, it seems to be implausible to say that what concepts an agent has depends on those who are observing her. Whether something is relevantly similar to something else for A depends therefore on whether A’s responses to them are relevantly similar, i.e., whether he treats them similarly in his own view. For the same reason, whether A’s responses are relevantly similar (such that A makes these responses because he takes those stimuli as similar), in turn, cannot be determined by mere similarity of the responses in a given aspect. We must once again take into account whether the agent takes his own responses as relevantly similar. Since an agent usually does not respond to his own response, again, it seems that there is no way to explain exactly what an agent responds to, when only a single agent is considered. But, one might contest, so far nothing requires that the agent’s responses must be observable by another agent or an observer. Thus, the responses of an agent may just be (non-propositional) feels the agent has, that is, the kind of feels an agent has when he finds two things similar. But whatever kind of responses they are, they are in general accessible to the agent himself. So the agent must be able to discriminate between what he responds to and what he finds similar. However, this complain does not explain why the agent’s responses are to distal causes or proximal causes, which is what needs to be explained in the distance problem. For illustration, consider a series of instances in which an agent gives responses, which he takes as similar, to

156

W. C. Leong

something he also takes as similar. In each of these instances, there is a proximal cause pi , (say, the stimulus on the agent’s retina caused by a cow) and a distal cause d i (e.g. a cow). The series of the proximal causes P = p1 , p2 , … must have a similarity pattern in certain aspect, so does the series of distal causes D = d 1 ,d 2 ,…. Their similarity patterns make them both candidates of what the agent responds to. But given simply that the responses are similar to the agent, it is still indeterminate whether it is the series of P or of D that is what the agent finds similar. To sum up, the objectivity and the intentionality of thought are the two features which Davidson argues that a theory cannot explain if only a solitary agent is considered. In the following sections, I will explain how Davidson argues that a multi-agent setting, that is, a triangulation scenario, is necessary for any adequate account of these features.

4 Triangulation Let me start with the objectivity of thought and the standard of correct application of concepts. The solution of KA aims to temporarily isolate the instance of an application of a concept whose correctness we are evaluating from other instances that, as a whole, partially determine the content of the concept. This way of putting the aim of the solution could sound ambiguous. This is not to say that KA tries to treat the former as a non-content-determining instance. It is designed to accept the possibility that, for each of the actual applications of the concept, it is possible that it does not determine which concept is employed in other similar responses. The reason for the isolation is that, if every single application of the concept at the same time also determines the content of the concept, then whatever it is applied to, it would become part of the extension of the concept, because the application is contentdetermining. In principle, therefore, no application could be incorrect. For there to be incorrect application of a concept, this temporary isolation is necessary. The challenge is how to make the separation between correct and incorrect applications of the concept in a non-ad hoc, non-arbitrary way. KA attempts to do this by taking the responses of typical members of the linguistic community in similar situations as content-determining conditions, so any other deviant applications could be taken as incorrect. As mentioned above, however, this answer is not methodologically satisfactory, since it already assumes that some responses are relevantly similar. Although the original solution from KA, according to Davidson, cannot explain the objectivity of thought and the standard of correct application of concepts, the direction of taking other agents into account is basically right, for an essential element is introduced: This is, of course, right, and it points us in the direction of Kripke’s Wittgensteinian answer to the problem of rule following. It is a brute fact—a fact about each brute, including you and me—that each reacts to stimuli as it does; but it is a suggestive fact that many of these brutes react in more or less the same ways. It is a suggestive fact, but what exactly does it suggest? As we just noticed, mere crowd behavior in itself can’t explain conceptualization;

Methodological Considerations in the Triangulation Argument

157

it can’t even explain error. What it can do, however, is make space for something that can be called error: room for error is created by cases in which one individual deviates from a course of action when the crowd does not. (Davidson 2001a, p. 5)

In a simplified triangular scenario consisting of minimally two agents and one shared focus, Davidson argues, the concept of error becomes applicable; when an agent’s response differs from his previous responses (differs in a sense specified below) while that of the other agent does not, one of their responses could be said to be incorrect. This scenario does not show who is wrong, for this is not a sufficient condition for attributing errors. What is important, however, is that this scenario allows our explanation to introduce a standard of correct application as something shared between the two interacting agents, and to take one of their responses as applying this standard of correct application correctly, while the other incorrectly. Such a triangular scenario thus, in Davidson’s words, “makes space” for the application of the concept of error, but it is still possible that the agents simply do not always find the same thing similar—one of them simply changes his response at that instance. There is, therefore, still no way to account for: … failure to apply a concept correctly, given that what one person might count as an error may just be another person applying a different concept. Kripke’s suggestion is that if a learner fails to apply a concept (or word) as his teacher would, the learner has made a mistake. Unfortunately this does not distinguish between failure to apply a concept correctly and applying a different concept correctly—the very distinction in need of explication. (Davidson 2001b, pp. 143–144)

This problem, formulated in terms of the methodological problem of similarity, is that it fails to distinguish between relevantly dissimilar responses and irrelevant responses. When two responses are not relevantly similar, besides the possibility that one of the responses is correct while the other is incorrect, it is also possible that they are both correct, when the two responses just do not involve the same concepts. This distinction is crucial because without it, what is left is only the distinction between relevantly similar responses and non-relevantly similar responses, no responses can be said to be incorrect. If no responses can be said to be incorrect, then no responses can be said to be correct either, and no standard of correct application or concepts would then be possible. In order to solve this problem, Davidson makes one of the most crucial and controversial claims of the triangulation argument: if an agent does not have the concept of error, the agent does not have concepts or thoughts (henceforth: NCE). NCE states that without an agent’s own application of the concept of error, none of his responses can have any propositional content; thus, the agent cannot have any thoughts, because none of his responses can be said to be correct or incorrect. It does not, however, state the even stronger claim that whenever an agent makes a response, he must be conscious of the possible error in order for that response to have propositional content. As long as the agent has the general awareness that a thought or a belief can be false, he satisfies the requirement of NCE. Davidson’s justification of NCE is that, if an agent does not have the concept of error, nothing can distinguish between the agent’s mere disposition and thought; since even mindless creatures can

158

W. C. Leong

have sophisticate dispositions, an agent who does not have the concept of error has nothing that can explain his having a thought instead of a mere disposition. NCE is often seen as an over-intellectualized thesis, for it is said that one need not have the concept of error in order to commit an error, just as one need not have the concept of breathing in order to breath. I believe, however, that NCE is defensible and will come back to it in the next section, but for the moment I will simply assume it. If one accepts NCE, the next question would be: how could an agent acquire the concept of error? Although triangulation is not designed to provide a full answer to this question, it is supposed to explain what is necessary for the acquisition of this concept. In the triangular scenario mentioned above, there is no interactions between the agents, so they may not be aware of their being in such a situation, or even the presence of each other. For each of the agents, therefore, this scenario may have no difference from a solitary scenario. Interactions between the agents are thus crucial; they enable the agents to be benefited from the triangular situation. Before I explain further, I shall illustrate the triangulation scenario first, as it is not identical to the triangular scenario in KA. Since Davidson has never been explicit about the formulation of a triangulation scenario, the following is my attempt to explain it. Davidson says: In this primitive triangle of two creatures reacting to a common stimulus and to each other’s reactions to that stimulus we have necessary conditions for the concept of error to arise. The conditions, to spell it out, are repeated reactions to shared situations found similar by two or more creatures each simultaneously observing the other’s reactions, these reactions in turn being found similar by each creature to reactions previously observed in the other in the shared situations found similar. The space needed for error to make sense appears when the correlation breaks down for one creature, but not for the other (or others): one creature responds to a shared situation as before (meaning, in a way the other creature has found similar in the past), but the other does not. (Davidson 2006, p. 1066)

Consider a simplified single instance first. Suppose two agents A and B in the same scenario both give a response which is caused by the same (distal) object or event OE1 (which in turn causes proximal causes pOEA1 for A, pOEB1 for B), and A gives the response RA1 , while B the response RB1 . At the same time, they both observe each other and their responses. At this stage, since there is only a single instance, the two agents cannot compare the other’s response to a previous one, the two agents cannot be said to be responding to OE1 or some other objects or events, or the proximal causes these objects caused. This is therefore not yet a scenario of triangulation (Fig. 1). As similar instances go on, the two agents may together encounter objects or events OE2 , OE3 , etc., the presence of which causes them to find something similar to something each of them encountered before. Thus, A finds iOEA1 , iOEA2 , iOEA3 , etc., similar, while B finds iOEB2 , iOEB3 , etc., similar. At this stage, however, we cannot say whether iOEAi or iOEBi are distal causes or proximal causes, since the intentionality problem has not yet been solved. We cannot, as a result, say whether A and B are responding to the same thing in any of these instances. That is, we cannot say whether iOEA1 is identical to iOEB1 . Besides their responses (RAx and RBx ) to iOEAx and iOEBx , each of them also finds the other’s responses in these instances

Methodological Considerations in the Triangulation Argument

159

OE

pOEA1, A

pOEB1, B

Fig. 1 A single instance of pre-triangulation scenario

similar to the other’s previous responses. Thus, A finds RB1 , RB2 , and RB3 etc. similar, and B finds RA1 , RA2 , RA3 , etc. similar. Notice that in the above quote from Davidson, it is assumed that the two agents are responding to the same stimulus; that is, it is assumed that iOEA1 is identical to iOEB1 in my example, for Davidson is trying to explain what makes thought objective. Since my purpose here is to present the triangulation scenario in a more general form, so that it can be used also to explain the intentionality of thought, I make no such assumption here. Nevertheless, A may then associate B’s responses, RBx , to iOEAx , and takes RAx and RBx as responses to the same thing. Similarly, B may also associate A’s responses, RAx , to iOEBx , and takes RBx and RAx as responses to the same thing. Now each agent takes both his and the other’s responses in each of these instance as having a shared focus, although we cannot say even in this stage whether they in fact respond to the same thing in any of these instances, for there is still no way to compare what they each take themselves and others as responding to (i.e. whether iOEAx and iOEBx are the same thing.) For A, therefore, both agents respond to iOEAx , while for B, they both respond to iOEBx . If we see the triangulation scenario in either agent’s perspective, the situation can be illustrated as follows (Fig. 2).

Fig. 2 An instance of pre-triangulation scenario in each agent’s perspective

160

W. C. Leong

In summary, these are the requirements for two agents A and B to engage in a triangulation scenario: (a) Both A and B find something similar, that is, iOEAx and iOEBx , respectively, which are, or are caused by, OEx , but we cannot yet say whether iOEAx and iOEBx are the proximal or distal causes. In other words, we cannot yet say whether iOEAx is pOEAx or OEx , or iOEBx is pOEBx or OEx . (b) For A, items of RAs are similar, and B finds them similar too. (c) For B, items of RBs are similar, and A finds them similar too. (d) A associates iOEAs to B’s responses (RBs) as what B responds to, after repeated observations. (e) B associates iOEBs to A’s responses (RAs) as what A responds to, after repeated observations. For Davidson, these complicated requirements are needed to solve the methodological problem about similarity. These requirements make sense of the comparison of one agent’s responses to that of the other, not only on our part, but also in the perspectives of the agents. For such a comparison to make sense, the two agents must be seen (by themselves) as having a shared focus. Although when an agent finds two things similar, the features on which he focusses is not directly available to the other agent, they can establish (d) (for A) and (e) (for B). The two series are established by both agents’ recognition of similarity. So not only must each of the agents find items in a series similar (iOEAs for A, iOEBs for B), what they take as the other agents’ responses to these items must also be found by each other to be similar. In a sense, we cannot yet say that the two agents respond to the same things, for in these formulations, items in the iOEA series may not be the same as those in the iOEB series, since at this stage there does not seem to be any reason why we must take iOEA or iOEB to be distal causes, for iOEAs are merely recognized by A as something B also responds to, the same is true of iOEBs. Once (d) and (e) are established, they can then compare each of their own responses to the other’s to see if their responses match; their responses match each other’s if and only if whenever A and B are both present in the same scenario, and A responds with RAx , B would also respond with RBx , and vice versa. RAx and RBx are not needed to be similar for either agent, so this formulation allows agents to have responses not taken by them to be similar to that of the other agent’s, as long as the responses given by the same agent are taken to be similar. For instances, whenever a lion is present, one agent always runs, while the other always climbs up a nearby tree. Note also that because of (b) and (c), RAx and RBx are for both agents similar to other items in the same sequence RA1 , RA2 , … and RB1 , RB2 , … respectively. These requirements of (b) and (c) imply that not only must the agents’ responses match each other’s, they must also be aware of each other’s responses. If they are aware of each other’s responses (RAx and RBx , call these first-order responses), they can then give meta-responses to these first-order responses (RA[RBx ] and RB[RAx ]); i.e., they can respond to each other’s responses to things they find similar which are caused by the external world. In the following, I will focus on the conditions in which the concept of error could arise for the two agents.

Methodological Considerations in the Triangulation Argument

161

Before I proceed, however, it is important to distinguish four kinds of triangulation scenarios. I call them the primitive scenario, the pre-conceptual scenario, the learning scenario, and the conceptual scenario. The primitive scenario can happen among primitive creatures (i.e. creatures with no propositional thought), as long as they can interact with other creatures. Creatures in this kind of scenarios need not be aware of any OE series; i.e., they need not associate others’ responses to items or events in the world. The pre-conceptual scenario is the most important for the triangulation argument, for the argument focuses on this scenario, and it is the scenario that is closest to, but not yet arrived at, a stage in which both agents have concepts. Agents in this scenario satisfy all the descriptions from previous paragraphs: the series, iOEA, iOEB, RB, and RA are already established; thus, the agents are in a position to be aware of the discrepancy between a response from a series of previous responses. Yet, the agents do not have any concepts or thoughts. The triangulation argument proceeds from this scenario and argues that before either agent can have thoughts, certain conditions (that is, (a)–(e)) must be fulfilled. A typical example of the learning scenario is a child beginning to learn its first language from a teacher. It is an asymmetric scenario: one of the agents involved has thoughts, while the other does not. Davidson uses this scenario for illustration and argues why error does not make sense in the very first few instances of learning one’s very first concepts. The last kind of scenarios is mentioned the least in TA, for it is basically the scenario of radical interpretation. Agents in this kind of scenarios already have thoughts and concepts, only that they do not understand each other’s speech. In the following sections, the triangulation scenario mentioned is the pre-conceptual scenario, since the solution to both problems raised by objectivity and intentionality are about certain necessary conditions for the emergence of thoughts from thoughtless agents.

5 Error and the Concept of Error Now that the triangulation scenario is in place we can explain how, for Davidson, the acquisition of the concept of error is possible. He says that during the early stages of the triangulation scenario, before the agents associate each other’s responses to something they find similar, no error can be explained: Nothing in this story suggests the necessity of thought, conceptualization, or error, even if the creatures occasionally fail to act in concert. … The problem is to put the creatures in a position to think this. Something must be added which will make the difference between the cases where the creatures act in unison, and the cases where they do not, available to the creatures themselves. This can be done by adding a further element of association or natural induction: each creature associates the other creature’s responses with stimuli from the shared world. (Perhaps this does not have to be learned in some cases.) Let us assume that our two creatures, A and B, can often observe one another when both can observe cows. Now suppose that creature A responds to cows by uttering the sound ‘Cow’; creature B then associates cows with the utterances of ‘Cow’ by A. Creature B similarly associates cows with A’s utterances of ‘Vache’. A and B are now each in a position to notice occasions on which the responses of the

162

W. C. Leong

other differs from the other’s usual responses (perhaps one is badly placed for seeing a cow; or one has better eyesight or a keener sense of smell). When I speak of responses differing, I mean: A’s responses are not taken by B to be the same as A’s earlier responses, while B’s responses are the same as B’s earlier responses. The basic triangle is now complete: there are causal connections between cows and our two reacting creatures, by way of perception, and there are causal connections between the creatures, also by way of perception. Finally, these two sets of connections are correlated, for each creature associates cows with the cowreactions of the other creature, except, of course in the relatively rare case of error. (Davidson 2001a, p. 6)

Davidson argues that the following scenario must occur before the agents are aware of error: Both agents must first associate each other’s responses to something caused by items in the iOEAx (for A) and iOEBx (for B) series, so that (d) and (e) are satisfied. Then one of the agents, say A, is aware of iOEAn , and responds with RAn ; A may then expect B to respond to iOEAn with RBn (note that (b) and (c) must also be satisfied here). But B gives a response, MB, which A does not take to be similar to any of B’s previous responses in the RBx series. A then may take iOEAn as, in B’s eyes, something not similar to items in the iOEAx series, and has make the contrast necessary to the concept of error, that is, whether iOEAn is or is not similar in the relevant way to the items in the iOEAx series, available to A. Likewise, B notices that A gives a response, which for B is similar to items in the RAx series, to something that B does not find similar to items in the iOEBx series. A’s response is a response that B does not expect; as a result, the contrast necessary to the concept of error is available to B too. Why is the agents’ mutual awareness of the triangular situation necessary for their acquisition of the concept of error? Here Davidson seems to assume that the concept of error can only be acquired in or after encountering a situation in which applying the concept of error would make sense; i.e., it would make sense to say that an error occurs in such a situation, even though an error does not actually occur. If this is the case, then since only in a triangular situation explained above does it make sense to attribute a mistake to an agent’s response, it follows that the triangular situation is necessary for the acquisition of the concept of error. Unsurprisingly, however, this assumption does not go unquestioned. For instance, among others [e.g. Briscoe (2007)]. Montminy (2003) suggests that there are several ways a solitary agent who has never interacted with other agents can acquire the concept of error: by retrospection, unfulfilled expectation, or imagined alternative view. There are at least three ways in which it can come to understand that one of its beliefs is wrong. First, a solitary creature can compare its current reaction to a past one. Robinson [a solitary agent] observes a bent stick that is half immersed in water. His impression is that the stick is curved. But later on, he observes the same stick from a different angle, and it now seems straight. This pair of observations can enable Robinson to realize that what seems to be the case may be different from what is the case, since at least one of his two impressions must be wrong. A second method available to the solitary creature is to try to see if its current reaction satisfies a generalization it has so far admitted. Since every time he throws a stone in the pond, it sinks, Robinson expects the stone he is about to throw to sink as well. But today the pond is frozen and, to Robinson’s surprise, the stone rebounds on the water’s surface. This observation goes against Robinson’s previously held generalization according to which

Methodological Considerations in the Triangulation Argument

163

every stone thrown in the pond sinks and can thus allow him to distinguish between his expectations and what actually happens. Thirdly, the solitary creature can compare its actual reaction to one it would have if it observed the same situation from a different perspective. It can indeed imagine how things would appear if its distance or angle were different, if lighting conditions were not the same, etc. The horse that Robinson sees in the distance seems to have three legs. But he can imagine how this horse would appear if his viewpoint were different; maybe it would seem to have four legs. Such a “thought experiment” can enable Robinson to realize that there may be some discrepancies between the way things are and the way they appear to him. (Montminy 2003, p. 38 my emphasis)

Montminy believes that an agent can compare his current perceptual experience with one in his memory and find out that what seems to be the case may not in fact be the case; an agent can also expect something to happen while in fact it does not; finally, an agent can imagine that something looks differently than it actually appears to him when viewed from another angle. In all these situations, Montminy believes, an agent could acquire the concept of error as applying to something in his current mental states or something in his memory, expectation, or imagination. C. Verheggen, however, defends that these ways to acquire the concept of error or objectivity is available to an agent only when he already has thoughts, i.e., already has the concepts in question: These methods would all be well and good provided the solitary creature could think or talk to begin with – provided, that is, it had beliefs about present situations which it could compare with past or imagined situations, or with admitted generalizations, and provided, of course, it could remember, imagine and generalize to begin with. But again, if I am right, the solitary creature is in no position to do any of this precisely because it cannot have the concept of objectivity. (Verheggen 2007, p. 102)

I think Verheggen is right that the way Montminy describes these situations presupposes that the agent has thoughts already, for Montminy argues that these methods are ways in which an agent “can come to understand that one of its beliefs is wrong”. In other words, it is a presupposition that the agent has beliefs. Furthermore, in Montminy’s descriptions, normative concepts such as “wrong”, “actually”, and “seem” are used to describe the content of what the agent, Robinson, can realize. If these descriptions are true of the agent, then it means that the agent already has these concepts. It begs the question to say that these are things the agent can do, for the issue here is exactly how they could acquire these concepts. If these descriptions are intended to describe what the agent could acquire after certain processes, then what these processes are, whether interaction with a second agent must be part of these processes, is precisely what is at issue. Without explaining how these concepts can be acquired in the first place, Montminy’s story of the solitary agent cannot serve as an example for the case that a solitary agent can acquire the concept of error. But even if Montminy’s proposed ways can be reformulated without attributing beliefs or concepts to the agents and describe what he finds out in terms of things the agent finds similar, I think these ways cannot help the agent to acquire the concept of error. Montminy’s account does not provide us any way to answer the Kripkensteinean challenge, so no standard of correct application can be attributed to the agent’s

164

W. C. Leong

responses, and no responses of the agent can be said to be an error. As a result, since the concept of error also has its own standard of correct application, the agent cannot make any error in applying the concept of error, which means the agent cannot have the concept of error in the first place. For illustration, we can construct a triangulation scenario in terms of an agent with his former self, so as to see its difference from the two-agent triangulation scenario.6 The two “agents” in this case would then be the agent’s former self and his current self. The “interaction” between the two “agents” would then be very different: the current agent could not observe his former self and associate his response to an object or event in the external world, as he would do to another agent. To see what is wrong with this proposal, we can consider a reductio. Suppose the agent acquires the concept of error in one of the proposed ways. Now what makes the concept he has the concept of error instead of a very similar yet different concept? His past dispositions of applying the concept could not, as mentioned above in the Kripkenstein argument, explain the normative constraint of the concept he employed; that is, his dispositions cannot determine the standard of correct application of his concept. Nor can the procedure or rule he thinks he follows when he applies the concept determine such a standard, because the same challenge can be put against the procedure itself. Similarly, his raw feels in his mind whenever he applies the concept cannot determine a standard of correct application, since in principle there is no reason why he cannot have the same feels while applying another concept. In a nutshell, there is nothing that can determine the agent’s concept as the concept of error instead of a very similar concept. Since KA is a metaphysical or conceptual challenge rather than an epistemological one, Montminy’s proposal, we should conclude, fails to explain how the concept of error thus acquired has any content at all, and that the agent cannot acquire the concept of error in the proposed ways. Montminy’s proposal does not offer an answer to the Kripkenstein challenge. There is another problem facing Montminy’s proposal. Since the agent’s former self cannot negotiate with his current self or be aware of his current self’s awareness of the mismatch of his two responses, whether it is his former response or his current response that can be considered as correct is entirely up to his current meta-awareness of the responses. Suppose the agent now finds iOEn similar to iOEa , and remember that he does not find them similar before. Presumably, the agent now would have a meta- “judgement” about his previous “judgement” that the latter is incorrect. But the problem is, should he decide to take his previous “judgement” to be correct and the current one incorrect, there is no standard of correct application to say whether his meta- “judgement” is correct or incorrect. Since the correctness of his current and previous “judgement” depends on his current meta- “judgement”, for which there is no standard of correct application; it seems that whether his current or previous “judgement” is correct is entirely up to the current agent after all. It is thus implausible to say that the concept of error, or any other concept, can be acquired in the ways Montminy proposes. 6 One

can also construct a triangulation scenario in terms of an agent with his imagined self or one in his expectation, the situation is, I believe, essentially the same.

Methodological Considerations in the Triangulation Argument

165

However, even though an isolated agent cannot acquire the concept of error, why is the concept of error essential for the agent’s acquisition of other concepts? Or in other words, why should we accept NCE, the claim that the possession of the concept of error is necessary for an agent to have thoughts and other propositional attitudes? Here is Davidson’s answer: In order to understand a proposition, one must know what its truth conditions are ... To know the truth conditions of a proposition, one must have the concept of truth. To have the concept of truth is to have the concept of objectivity, the notion of a proposition being true or false independent of one’s beliefs or interests. In particular, then, someone who has a belief, who holds some proposition to be true or false, knows that that belief may be true or false. In order to be right or wrong, one must know that it is possible to be right or wrong. (Davidson 1995, p. 211)

For Davidson, therefore, to believe a proposition that p one must know the truth conditions of p. To know the truth conditions of p, one must have the concept of truth, which, in turn, requires one to understand the concept of objectivity; i.e., p is true or false independent of one’s beliefs or interest. Therefore, if one believes that p, then one knows that p may be true or false; that is, it is possible to commit an error. This is a typical response from Davidson on the issue. But it simply assumes without arguing for the crucial claim that to know the truth conditions of p one must have the concept of truth and of error, which is essentially the claim of NCE. Over the years, understandably, not many philosophers are convinced by the argument. One may challenge, for instance, that even if we grant that to understand a proposition is to know its truth conditions, why must one have the concept of truth in order to know its truth conditions? When we think that snow is white, we think of snow, the colour white, and the certain relationship between the two. We need the concept of truth only when we reflect on, say, a sentence or a proposition with certain content, and judge that whether it is true or false. In short, the concept of truth and other related concepts such as the concept of error, objectivity, belief, etc., are second-order concepts which one does not need to have in order to entertain firstorder thoughts. Such a critic may hold that a second person is therefore not needed at all, or that only a potential, instead of actual, second person is needed. They all agree, however, that actual interaction with another agent is not essential. I believe, however, NCE is not as implausible as it may first seem, and actual interaction with a second agent in a triangulation scenario seems to be necessary for an agent to have thoughts. I will not be able to argue for NCE in this paper, however, interested reader can see Verheggen (2007, 2013) and Myers and Verheggen (2016) for arguments for NCE. I will now provide some of my own reasons for accepting it. To recap, the Kripkensteinian reason that a disposition is not a thought is that no error is possible without consideration of another agent: whatever an agent does, depending on which standard of correct application, we choose to compare his responses to, what the agent does can be accordingly said to be correct or incorrect, thus, in order words, nothing can determine its correctness or incorrectness. Davidson claims that interactions with another agent in triangulation make possible the situation in which the concept of error can be correctly applied, so the acquisition of the concept of error is possible. What Davidson does not say, however, is

166

W. C. Leong

how an error becomes an error, and how it can be identified by the agents. Therefore, on the one hand, we lack an account of how errors are determined, and on the other hand, Davidson leaves unexplained how the concept of error contributes to the “emergence” of thought. Here is what I propose. I think there is an essential relation between the concept of error and the determination of error when an agent first learns a concept: an error is an error because it is identified as an error by the agent. This is not to say that every error must be identified as an error by someone, it is just the claim that without the agent’s own identification of error no response can be seen as an error at the initial stage of concept acquisition. This is not a claim of temporal priority, but a conceptual necessity. To illustrate, recall the triangulation scenario above with the agent B’s response not being seen by A as relevantly similar to any of the items in the RBs series. If they do not come to some kind of agreement on B’s response, nothing can determine whether it is A or B who is making a mistake. For even if one of them could make a decision to, say, ignore the other’s response, the decision is entirely up to the agent himself. The agent himself may have certain psychological impulse to make one decision instead of the other, but even if he decides otherwise, nothing can determine that that decision is an incorrect one. By contrast, if they agree on whose response is a mistake, then it is an error not entirely up to one of the agents. They would then make a decision, so to speak, which partially constitutes the content of the concepts they employed, by agreeing on an instance in which the concepts are applied incorrectly. A constitutive connection between the agents’ responses and the standard of correct application of their thoughts and concepts is thus established. It is constitutive to the agent’s thinking of these thoughts and the related concepts, instead of the propositional content of the thoughts per se. Or, to put it differently, it is constitutive to the relationship between the agent’s mental propositional attitudes and particular propositions. After an agent acquires the concept of error through such scenario, he could discover or invent new concepts with the ones he already has, without interacting with another agent in a way relevant to these concepts. This proposal clearly implies actual interactions with another agent. But the question is, what kind of agreement could this be? I believe that the kind of agreement required is what M. Gilbert calls joint commitment (Gilbert 1989, 2014). A certain kind of joint commitment between at least two agents can determine the objects or events to which the concept associated to this commitment can correctly apply. But this kind of joint commitment is not always explicit agreement (although an explicit agreement is, for Gilbert, a joint commitment), it still depends on the agents’ own sense of similarity. This proposal, while in line with Davidson’s thesis that intentional concepts are not reducible to extensional concepts, can also explain the difference between having a thought and having a mere disposition. A creature can have highly sophisticated discriminative skill, and the ability to treat various objects under divergent conditions in similar ways. They must have a complicated mechanism that enables them to do so, which can be described in extensional terms. But to describe a creature in terms of thoughts and beliefs basing on mere its discriminative mechanism is no more plausible to describe a computer with propositional thought: we can say a

Methodological Considerations in the Triangulation Argument

167

computer thinks that I am not authorized to use it when I fail to enter the correct password, or it wants to know if I am authorized to use the programme by asking me for a password. By attributing thoughts to the programme or the computer, we can describe its behaviours in a convenient way, but we do not really think that contemporary programmes or computers can have propositional thoughts as we do. Besides the objectivity of thought, my proposal can also support TA’s explanation of the intentionality of thought. I give a brief account of this in the next section.

6 Shared Causes and Intentionality Davidson says: Take an example. I observe a lioness stalking a gazelle. I describe this situation in this way by correlating two aspects of the situation: gazelle behavior and lion behavior (the lioness changes course to intercept the gazelle; the gazelle changes course to evade the lioness). I can confirm my theory that the gazelle is the object salient for the lioness by (among other possibilities) noting the responses of other lionesses to gazelles on other occasions. But even this further evidence leaves it up to me to decide what the object of the lioness’s attention is; I decide on the basis of what catches my attention. (Davidson 2001a, p. 6)

Recall that the problem raised by intentionality is that there is no observer independent reason to say whether it is the distal or the proximal causes that an agent responds to when he makes a response. Even if we compare his current response with his previous ones, both the distal causes and the proximal causes of these responses can be said to be similar. However, we have only considered the single-agent scenario, so now we turn to multi-agent scenarios. Davidson continues: If a second lioness joins the first in pursuit of the gazelle, I can eliminate such complete dependence on my own choice of salient object in this way: I class together the responses of lioness A with the responses of lioness B in the same places and at the same times. The focus of the shared causes is now what I take to be the salient object for both lionesses. I no longer have to depend on my own choice of the relevant stimulus of the lionesses’ behavior. In the exceptional case where responses differ I can say: one of the lionesses has erred. (Davidson 2001a, pp. 6–7)

When two non-interactive agents are considered, the situation is essentially the same for each of the agents as when they are alone, so it is still indeterminate whether they respond to distal causes rather than to proximal causes. But now in this scenario with two interacting agents, it makes sense to speak of shared causes of the two agents’ responses. What Davidson calls a shared cause is a cause that is common to the responses from two distinct agents, so in this sense, a single agent cannot share a cause with his former self. If we classify two agents’ responses together, since the proximal causes cannot be shared, the shared cause of their responses, if there is one, must be a distal cause. Therefore, if we take the things to which the agents respond to be a shared cause, we can say that the agents respond to a distal cause. What the agents respond to, therefore, no longer depends on our choice.

168

W. C. Leong

But which instances of their responses should we consider as relevant and should be classified together? Although the above consideration can eliminate the observer dependency of relevant stimuli of responses, the observer dependency of relevant responses is still here. Davidson thus proposes: The challenge is to put the lions in a position to distinguish these cases. To do this we have to eliminate the dependence on my arbitrary (or interested) choice of relevantly similar responses on the parts of the lionesses. A further element enters when the lions cooperate to comer their prey. Each watches the other while both watch the gazelle, noting the other’s reactions to the changes of direction. (Davidson 2001a, p. 7)

Under this account what responses are relevantly similar are decided by the agents involved, by whether they find the responses similar. In Davidson’s example, the two agents (the lionesses) also respond to each other’s responses to a shared focus (“noting the other’s reaction to the changes of direction”), their cooperation of hunting their prey enables them each to associate each other’s responses to a share focus. They each may find the other’s responses similar with regard to their cooperation, instead of responses out of, say, fear of a hunter. One might, however, object that the decision to take what the agents respond to as a shared cause is still arbitrary, or even ad hoc: since we know that a proximal cause cannot be shared, our choice to take the agents as responding to a shared cause is no better than to simply say that the agents respond to a distal cause. Their cooperation may encourage the agents to take themselves as responding to a shared focus, but taking themselves as responding to a shared focus does not imply that they are actually responding to a shared focus. One could say that they are in fact responding to each of their proximal causes. We, therefore, still lack a non-observer dependent explanation. What we need then is an independent reason to take the agents’ responses as responses to a shared cause. This worry is appropriate, if we consider the case from the perspective of each of the two agents separately. In fact, from the above description, the agent we considered may just be, in a sense, an observer of each other, that is, they respond to the other agent’s responses as if the other is just some mindless object. For there to be genuine interactions between the two agents, therefore, both agents must respond to the other’s responses in a special way. This special way is, I think, the kind of collaboration in which multiple agents try to achieve an aim together, instead of an agent trying to do one’s own in a multi-agent setting. In such a collaboration each agent is jointly committed with each other to do certain things. The agents thus share one and the same aim. Therefore, if two agents are in such a collaboration, this would be a very strong reason to take their responses as responses to the same cause because one must be aware of the collaboration in order to collaborate; the agent’s awareness of the shared cause and their associations of each other’s responses to this shared cause would become essential elements in the intentionality of their concepts. But this is a work I leave for a future paper. This is how I take TA as arguing that certain interactions with another agent are necessary for an agent to have propositional thought, which is both objective and intentional. As mentioned above, there are a lot of objections of TA among philosophers. Some of them even seem very convincing given Davidson’s presentation of

Methodological Considerations in the Triangulation Argument

169

TA. Nevertheless, I believe TA is correct and it is defensible from these objections, but only with some reinforcement. I hope my current formulation of TA at least would make it more plausible than it may first seem.

References Blackburn, S. (1984). The individual strikes back. Synthese, 58(3), 281–301. Boghossian, P. A. (1989). The rule-following considerations. Mind, 98(392), 507–549. Bridges, J. (2006). Davidson’s transcendental externalism. Philosophy and Phenomenological Research, 73(2), 290–315. Briscoe, R. E. (2007). Communication and rational responsiveness to the world. Pacific Philosophical Quarterly, 88(2), 135–159. Burge, T. (2010). Origins of objectivity. Oxford: Oxford University Press. Davidson, D. (1995). The problem of objectivity. Tijdschrift voor filosofie, 203–220. Davidson, D. (1999). The emergence of thought. Erkenntnis, 51, 7–17. Davidson, D. (2001a). Externalisms. In P. Kotatko, P. Pagin, & G. Segal (Eds.), Interpreting Davidson (pp. 1–16). Standford CA: CSLI Publications. Davidson, D. (2001b). What thought requires. In J. Branquinho (Ed.), The foundations of cognitive science. Oxford: Clarendon Press. Reprinted in Problems of rationality (pp. 135–149). Oxford: Oxford University Press. 2004. Davidson, D. (2004). Problems of rationality. Oxford: Clarendon Press. Davidson, D. (2006). The perils and pleasures of interpretation. In E. Lepore & B. Smith (Eds.), The Oxford handbook of philosophy of language (pp. 1057–1068). Oxford: Oxford University Press. Forbes, G. (1983). Scepticism and semantic knowledge. Proceedings of the Aristotelian Society, 84, 223–237. Gilbert, M. (1989). On social facts. New Jersey: Princeton University Press. Gilbert, M. (2014). Joint commitment: How we make the social world. New York, NY: Oxford University Press. Glüer, K. (2006). Triangulation. In E. Lepore & B. Smith (Eds.), The Oxford handbook of philosophy of language (pp. 1006–1019). Oxford: Oxford University Press. Koˇtátko, P., Pagin, P., & Segal, G. (2001). Interpreting Davidson. Stanford, CA.: CSLI Publications. Kripke, S. A. (1982). Wittgenstein on rules and private language: An elementary exposition. Oxford: Oxford University Press. Lepore, E., & Ludwig, K. (2005). Donald Davidson: Meaning, truth, language, and reality: Oxford: Clarendon Press. Lepore, E., & Smith, B. C. (2006). The Oxford handbook of philosophy of language. Oxford: Oxford University Press. Montminy, M. (2003). Triangulation, objectivity and the ambiguity problem. Crítica, 35, 25–48. Myers, R. H., & Verheggen, C. (2016). Donald Davidson’s triangulation argument: A philosophical inquiry. London: Routledge. Verheggen, C. (2007). Triangulating with Davidson. The Philosophical Quarterly, 57(226), 96–103. Verheggen, C. (2013). Triangulation. In E. Lepore & K. Ludwig (Eds.), A companion to Donald Davidson (pp. 456–471). Chichester: Wiley Blackwell.

Meaning, Evidence, and Objectivity Olivia Sultanescu

Abstract This chapter addresses the question of what, according to the conception of meaning offered by Donald Davidson, makes expressions meaningful. It addresses this question by reflecting on Kathrin Glüer’s recent response to it. It argues that Glüer misconstrues both the evidence for meaning that the radical interpreter must rely on and the way in which the principle of charity must be deployed. The articulation of the correct construal of the evidence and the principle reveals the thoroughly non-reductionist aspect of Davidson’s conception of meaning. This aspect becomes even clearer in his later work, through the articulation of the triangulation argument. The chapter shows how this argument, which is reconstructed in accordance with Claudine Verheggen’s interpretation of it, helps answer the initial question. It ends with a brief discussion of the question whether the conception has the resources to account for the objectivity of meaning. Keywords Radical interpretation · Triangulation · Foundational semantics · Non-reductionism

1 Introduction What is it for words to be meaningful? This question is at the core of the metaphysical inquiry into meaning. One way to proceed towards articulating an answer is to start from a very plausible thought, namely, that it is distinctive of meaningful utterances that they can be understood, not just by the speaker making them, but also by an interlocutor. Few would object to this claim. But some philosophers pursue this idea further. They say that meaningful utterances are essentially understandable by another speaker. This is tantamount to the claim that meaning is essentially public. This chapter is in its final form, and it is not submitted for publication anywhere else. O. Sultanescu (B) Society of Fellows in the Liberal Arts, The University of Chicago, Gates-Blake Hall, 5845 S Ellis Avenue, Chicago, IL 60637, USA e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_10

171

172

O. Sultanescu

One famous proponent of this claim is Donald Davidson. As he puts it, “that meanings are decipherable is not a matter of luck; public availability is a constitutive aspect of language” (1990b, 314). Thus, for Davidson, to investigate the decipherability of meaning is in effect to make progress in the metaphysical inquiry into meaning; it is to answer the question of what it is for words to mean what they do (Davidson 1984, xv). But how are we to investigate decipherability? Davidson proposes that we do this by addressing two other questions. First, what sort of knowledge would allow one to decipher the meaning of the utterances of someone who speaks an alien language and whose states of mind one does not know? Second, how might one acquire that knowledge? His answers to these two questions appeal to the notion of a Tarski-style theory of truth for a speaker. It is knowledge of such a theory that would allow an interpreter to understand any utterance made by the speaker, and it is by attending to the speaker’s holding sentences true in particular circumstances, and taking her to be by and large right, that is, taking the sentences she holds true to be by and large true, that we may construct such a theory. This is a very rough summary of some of the main claims defended in Davidson’s papers on radical interpretation. But even once the decipherability of meaning is elucidated in this way, we might still find ourselves wondering what makes expressions meaningful in the first place. This is the topic that I shall pursue in this paper. It is also the topic that Kathrin Glüer pursues in a series of recent works.1 I shall sketch what I take to be the right answer to the above question by reflecting on Glüer’s own response to it. First, I shall outline Glüer’s answer and say why I think we should be suspicious of it as a reconstruction of Davidson’s account. Then, I shall try to show that she misconstrues both the evidence for meaning that the radical interpreter must rely on and the way in which the principle of charity must be deployed. The correct construal of the evidence and the principle suggest a thoroughly non-reductionist conception of meaning. The nonreductionist aspect of Davidson’s project becomes much clearer in his later work, through the articulation of the triangulation argument. I shall explain that argument by relying on Claudine Verheggen’s reconstruction of it.2 I shall conclude by briefly addressing the question, which Glüer also addresses, whether the account provides a satisfactory conception of the objectivity of meaning. Before delving into the details of Glüer’s view, I should note that, while I find her reconstruction of some of the aspects of the Davidsonian project inadequate, both on philosophical and on interpretive grounds, I take her attempt to articulate the idea that meaning is “an evidence-constituted property” (Glüer 2011, 27) in a way that shows that, contrary to the received view, Davidson is not an interpretivist, to be extremely valuable.

1I

shall focus on Glüer (2011, 2017, 2018). Myers and Verheggen (2016, Chap. 1). See also Verheggen’s contribution to this volume for a more detailed argument to the effect that, contra many interpreters, Davidson’s commitment to non-reductionism spans his entire philosophical career.

2 See

Meaning, Evidence, and Objectivity

173

2 Radical Interpretation and the Metaphysics of Meaning If we endeavour to shed light on the nature of a phenomenon, we might begin by asking, simply, what it is for something to be an instance of it. But how to go about answering such a question is not always obvious. As I have said, Davidson’s strategy is to tackle the question of the nature of meaning indirectly, by addressing two other questions about its decipherability. And, to repeat, once the answers are laid out, that is, once the decipherability of meaning is fully elucidated, we might still find ourselves looking for an articulation of the determinants of meaning, or for an answer to the question of what makes something a meaningful expression.3 Let us begin by surveying Glüer’s reconstruction of the Davidsonian answer to this question.4 Glüer claims that when elucidating the determination of a phenomenon, we should distinguish between three elements, namely, the base, the target, and the principle of determination. The base is composed of the facts that play the role of determinants; the target consists in that which is thereby determined; the principle governs the relation of determination.5 The latter, which is “surprisingly often overlooked in discussions of meaning determination” (Glüer 2018, 252n2), is a requirement for any account that seeks to illuminate it. This is due to the fact that “we need to know by what principle the items in the determination target are determined by those in the determination base” (Glüer 2017, 75). Merely saying that X determines Y is insufficient to provide a full elucidation of how, specifically, Y is determined, for there are many ways of mapping instances of X to instances of Y, many ways in which the latter could metaphysically yield the former. In fact, the requirement of articulating the principle of meaning determination is, according to Glüer, “one of the main lessons to be drawn from Wittgenstein’s rule-following considerations” (Glüer 2018, 252n2), considerations to which we shall return. For now, our question is how Davidson might be taken to conceive of the base, target, and determination relation in his elucidation of meaning. To answer this question, Glüer focuses exclusively on the radical interpretation papers, which contain Davidson’s reflections on the procedure through which a radical interpreter constructs a theory of meaning for a speaker whose language and attitudes are not known to her.6 A crucial part of the radical interpreter’s endeavour involves her recording the sentences that the speaker holds true. This is something that the interpreter is in a position to do prior to grasping the meaning of those 3 Mark

Greenberg helpfully distinguishes between full constitutive accounts of a phenomenon and modest constitutive accounts. The former “purports to give an account of what X is” (2005, 303). The theorist who denies the possibility of such an account might still try, in a more modest spirit, to specify what makes something an instance of X in particular cases. Importantly, the distinction between full and modest accounts is independent from that between reductive and non-reductive accounts. A reductive account is a full constitutive account that is non-circular. But full constitutive accounts need not be non-circular. Davidson clearly aims for a full constitutive account that is non-reductive. 4 As I said before, my focus will be on Glüer (2011, 2017, 2018). 5 See also Pagin (2002), on which Glüer’s discussion of this tripartite distinction is based. 6 See, especially, Davidson (1973, 1974, 1975, 1976).

174

O. Sultanescu

sentences and without knowing anything about the speaker’s states of mind. (I shall say more about this later.) On Glüer’s view, the Davidsonian determination base consists of facts about sentences held true by the speaker or, in her own words, facts about “attitudes of holding true towards uninterpreted sentences” (Glüer 2011, 137). Once the interpreter has collected sufficiently many such facts, she may proceed to construct a theory of meaning. The theory of meaning is a theory of truth, a theory that yields truth conditions for any of the sentences of the speaker’s language by specifying the contributions that basic expressions and the rules for combining them make to these truth conditions. The only way to construct such a theory, for Davidson, is by deploying the principle of charity, which, in Glüer’s words, “counsels the radical interpreter to interpret the alien speaker such that his beliefs are true and coherent where plausibly possible” (2011, 9). The principle of charity is the principle of meaning determination. So, “attitudes of holding true thus form the base on which the principle of charity metaphysically determines meanings” (2011, 137). This way of conceiving of the base, target, and principle of determination is, Glüer thinks, compulsory in the light of Davidson’s commitment to the public nature of meaning. On her view, to say that meaning is essentially public, or that public availability is a constitutive aspect of meaning, is to commit oneself to the idea that meanings are determined by the evidence for them. In other words, the facts that determine the meaning of expressions for a speaker must be the facts that serve as evidence for the theory of meaning that a radical interpreter might build for her: The data for the t-theory that the radical interpreter is after therefore have an epistemicometaphysical double nature. They epistemically support the theory, but at the same time, Davidson tells us, they “entirely determine” the very thing the theory is a theory of. The data available to the radical interpreter thus form the determination base for meaning. Meaning is an evidence-constituted property. (Glüer 2017, 82)

Before discussing the problems that Glüer raises about Davidson’s account of meaning determination, there is an aspect of her reconstruction that is worth emphasizing. The basic tenets of this reconstruction are, first, that the epistemic base is the same as the metaphysical base and, second, that the epistemic base—and, thus, the determination base—consists entirely of “facts about which sentences speakers hold true under which observable circumstances” (2011, 135). The determinants of meaning are non-semantic. Glüer writes: The relation the principle of charity establishes is one of supervenience: On the Davidsonian picture, linguistic meanings (and belief contents) are determined by non-semantic facts. But the relation is one of supervenience only; according to Davidson, there is no chance of any reduction of the semantic to the non-semantic, or the intentional to the non-intentional. His is a non-reductive naturalism. (Glüer 2011, 137)

It is undeniable that, according to Davidson, the fact that a speaker holds a sentence true can be detected by an interpreter who has no knowledge of the meaning of that sentence, indeed no knowledge of any of the speaker’s intentional attitudes. As he puts it in one of the early papers, “it is an attitude an interpreter may plausibly be taken to be able to identify before he can interpret, since he may know that a person intends to express a truth in uttering a sentence without having any idea what

Meaning, Evidence, and Objectivity

175

truth” (1973, 135).7 The detection of these attitudes allows the radical interpreter to get started on the task of devising a theory of meaning. But does this really suggest that these facts, which are detected in this way, exhaust the metaphysical base for meaning? In the next section, I shall argue that it does not. For now, I shall very quickly outline what I take to be a source of tension in Glüer’s account, one which should make us wonder whether her reconstruction gives us a complete account of the determinants of meaning. As we have seen, according to Glüer, the facts in virtue of which expressions have the meanings they do are the facts that properly serve as evidence for an interpreter. However, it is undeniable, and I take it that Glüer would not deny, that nothing could properly serve as evidence for an interpreter unless some meaning facts already obtain. Facts about speakers’ holding sentences true cannot obtain unless speakers already mean something by those sentences. But, then, given that this is the case, it is not easy to make sense of the idea that the facts that properly serve as evidence for an interpreter are metaphysically more basic than meaning facts. This gives us reason to be cautious. Glüer might reply that this is precisely what it is for meaning to be evidenceconstituted, and for the interpreter to be the “truly ingenious device illustrating the Davidsonian take on the public nature of meaning and its metaphysical foundations” (2011, 8). Perhaps it is not easy to make proper sense of this idea, but proper sense of it we must try to make if we are to follow Davidson in taking the perspective of the interpreter to be metaphysically basic, or so her thought would go. However, from the claim that meaning is evidence-constituted, it does not follow that the evidence is altogether devoid of meaning facts.

3 The Evidence for Meaning For Davidson, meaning is, to use again Glüer’s apt way of putting it, an evidenceconstituted property. The problem, I shall argue, is not with her suggestion that the public nature of meaning commits us to the view that the basis on which we might know what someone means by her expressions must be constitutive of her meaning it. Rather, the problem, as we shall see in this section, has to do with the way in which she construes both the evidence for the construction of a theory of meaning and the application of the principle of charity. In pursuing the project of radical interpretation, the interpreter necessarily relies on her own attitudes, her own language, and her own conception of what it is to be a speaker and an agent. The early papers of Davidson’s, on which Glüer focuses, might not make fully perspicuous the significance of the fact that the interpreting is, necessarily, done by someone who already has a language and thoughts, as well as an understanding 7 Later on, he writes that “the interpreter can detect one or more nonindividuating attitudes. Examples

of the kind of special attitude I have in mind are: holding a sentence true at a time, wanting a sentence to be true, or preferring that one sentence rather than another be true” (1991b, 211).

176

O. Sultanescu

of what it is to be an agent and speaker. Davidson insists that the evidence that the interpreter must start from cannot presuppose knowledge of the speaker’s intentional attitudes, which is why it must consist solely in facts about sentences held true. Such facts are, as we have seen, non-semantic insofar as they can be detected without any knowledge of the content of the speaker’s attitudes, and stated without specifying the meaning of her sentences. This is the only way to ensure that the elucidation of meaning that Davidson offers is not question-begging. But this claim concerns the evidence that is about the speaker; it is not a claim about the totality of the evidence that the radical interpreter might rely on.8 After all, it is indisputable that the interpreter already has a language, and that she is already a thinker. She “has the concepts of truth, of intention, of belief, of desire, and of assertion (and many, many more). She knows a lot about the world and about how people behave in various circumstances” (Davidson 1993, 81). But, one might wonder, does this show that she must rely on these concepts and on this knowledge in order properly to execute her interpretative project? To begin with, as Verheggen explains in her contribution to this volume, what the interpreter means by her words is part of that on which she must draw in order to construct a theory of meaning that is genuinely meaning-giving. Second, the interpreter’s own understanding of people, acquired through previous engagements with them, does play, indeed must play, a significant part in the project of interpreting someone from scratch. Recall that the execution of this project requires that the interpreter take the speaker to be largely right in her beliefs about the world as well as in her engagements with it. It also requires that the interpreter attribute to her mistakes whenever it is plausible to think that mistakes have been made. To take just one example, the interpreter should be more inclined to attribute a mistake to a speaker if she notices that the speaker is distracted. This is due to the fact that she should know, from previous encounters with others, that distracted people are more likely to make mistakes.9 Consequently, talk of optimizing agreement, which “involves weighting mistakes or disagreements and minimizing a theory’s overall score” (Glüer 2017, 80n22), and which pervades Glüer’s articulation10 of the radical interpreter’s task, is unsuitable for capturing the way in which radical interpretation must unfold. This idea of optimization suggests a picture on which the interpreter’s endeavour is fundamentally one of calculation: she simply assigns weights to mistakes, and she compares those weights in order to reach interpretative decisions. But, we might ask, what content could a principle for assigning relative weights to mistakes have? According to Glüer, one such principle is that according to which “being wrong on simple observational matters such as whether it rains around one is more destructive [and thus weightier] than disagreement on highly theoretical matters” (Glüer 2017, 80n22). Thus, the radical interpreter should generally opt for attributing to the speaker false beliefs about theoretical matters rather than false beliefs about observational ones. 8 See

also Myers and Verheggen (2016, 123). is an example of Davidson’s. See (1995, 50). 10 See also, e.g., Glüer (2011, 79). 9 This

Meaning, Evidence, and Objectivity

177

But, surely, we can imagine plenty of cases in which the interpreter is better off not pursuing this path, that is, cases in which disagreement on basic, observational matters helps make better sense of the speaker—for instance, the case, mentioned earlier, in which the speaker is distracted. Glüer might of course respond by insisting that there is nothing in her account that rules out these cases, for the general principles that she outlines are ceteris paribus principles. However, whether a general principle applies in a particular case cannot be settled independently of considerations that already presuppose that meaning facts obtain; that is, it cannot be settled outside a conception of what it is to be a speaker and an agent. As Glüer herself says, “agreement must be optimized, and that involves weighting mistakes according to how destructive they are for understanding” (2011, 79, emphasis added). In fact, as Davidson came to emphasize more and more, the goal of interpretation is not that of maximizing truth or of optimizing agreement, but rather that of securing understanding. But understanding the speaker requires attributing to her errors that are themselves understandable. As Davidson puts it, “the fallings off from rationality in others [must be made] intelligible” (Davidson 1995, 50). Crucially, which errors come to seem intelligible to the interpreter will largely depend on the context, as well as, much more significantly, on the interpreter’s own conception of how people—qua rational agents—think, speak, and act. And this is simply not something that could be captured by principles for assigning relative weights to mistakes; it is not something that could be codified in this manner. All of this would seem to suggest not only that the evidence for meaning, and thus the so-called determination base, is, contra Glüer, not non-semantic, but also that the very application of the principle of charity is constrained by considerations about the semantic domain—about what it is to have a mind, what it is to be capable of making mistakes, and so on. To repeat an earlier point, this is not to say, of course, that the evidence that directly concerns the speaker is not non-semantic; as Davidson insists, it is mandatory that we conceive of the interpreter as having no prior knowledge of semantic facts about the speaker. But this is compatible with the interpreter’s reliance on semantic evidence throughout—on facts about what people, including herself, mean, think, and do. The radical interpreter always constructs her theory from within the semantic domain, as it were. This also gives us a clue about how to approach the imagined case, which was first discussed by Peter Pagin and is further developed by Glüer, in which a radical interpreter named Casey cannot detect any pattern with respect to the uses of a particular expression ‘’ made by an alien speaker whom she refers to as Alien. Casey deals with this conundrum by defining the set of correct uses for that expression as the set of uses that have actually been made by Alien. More specifically, Casey defines the predicate ‘’ by cases: “true of objects that Alien applies  to, false of objects that Alien withholds  from, and for all objects b unconsidered by Alien, F is true of b just in case b is a rocket” (Pagin 2013, 236). Casey builds much of the theory of meaning for Alien by deploying this strategy. The theory is one that, Pagin thinks, meets the constraints of the principle of charity, for it maximizes the truths that Alien possesses; it renders his view maximally accurate. Nevertheless, Pagin

178

O. Sultanescu

argues, the theory does not make it possible for Casey to understand Alien (2013, 237). However, given that the principle of charity is a principle that demands that the interpreter make interpretative decisions “partly on normative grounds by deciding what, from his point of view, maximizes intelligibility” (Davidson 1991b, 215), there is reason to doubt that Casey meets its constraints in this case. For we do not improve, let alone maximize, intelligibility if we depict someone in a way that makes it impossible for her to have been wrong in any of her past judgements about certain observational matters; in fact, we seem to erode it. An agent is the sort of creature who is likely wrong at least on some occasions and who lives with the recognition of this fact. Glüer agrees that Casey’s theory does not allow us to understand Alien (2017, 86), but she does not seem to have doubts about whether the theory is the product of a proper application of the principle of charity or the outcome of a properly carried out interpretative project. To my mind, doubts are in order here. Thus, meaning is evidence-constituted, but the evidence cannot be counted as evidence for meaning unless it includes semantic facts. It must be granted that this has the air of a paradox. But it is, ultimately, the manifestation of Davidson’s nonreductionism. His commitment to non-reductionism emerges most clearly in his triangulation argument, which I shall discuss in the next section.

4 The Triangulation Argument For Davidson, the radical interpreter cannot carry out her project of interpreting the speech of another without relying on semantic considerations. This is, in a nutshell, what I take the previous two sections to have shown. The upshot, as we have seen, is that we cannot say what it is to use expressions meaningfully, and we cannot specify the facts in virtue of which expressions are used meaningfully, without presupposing meaning. In other words, the upshot is a more ambitious form of non-reductionism than Glüer seems to allow for. The argument that definitively shows that Davidson adopts such a non-reductionism is known as the triangulation argument. I shall try to offer a very brief reconstruction of the argument, which very closely follows the interpretation developed by Verheggen.11 The starting point for the argument is the claim that the meanings of our most basic utterances about the world are constituted, at least in part, by the external features 11 See

Myers and Verheggen (2016, Chap. 1), for a detailed account of the argument. See also Verheggen (2007) for her initial defence of the claim, which is central to her reconstruction of the argument, that linguistic meanings and propositional contents are bound to remain indeterminate in an individual who lacks the recognition of the distinction between correctness and incorrectness, truth and falsity. Verheggen’s interpretation of the triangulation argument as a unified argument contrasts with Glüer’s, who thinks that there are “a number of more or less different versions of the triangulation idea” (2011, 235) in Davidson’s writings. These, according to her, amount to different arguments, among which are the argument from content determination and the argument from objectivity (2011, 232–244).

Meaning, Evidence, and Objectivity

179

that typically cause those utterances.12 Thus, to give an account of what it is for our basic utterances to be meaningful is, in part, to give an account of what it is for a cause of an utterance to be meaning-constituting. One difficulty is that, for any type of utterance, there are indefinitely many candidates for the role of typical cause that might be invoked; the typical causes of utterances are ambiguous or indeterminate. Take, for example, a type of utterance that is made whenever the speaker is in the presence of an oval item. It might seem as though it must be about ovalness, given that ovalness is its apparent cause. And yet, the causes may be numerous. They include not only, say, patterns of light on the speaker’s retinas, but also, more significantly, other features that typically characterize the oval items encountered by the individual, such as, say, having a texture, or having a colour, or being tasty, or simply being an object. Thus, the indeterminacy has two dimensions, which Davidson distinguishes: ...the cause is doubly indeterminate: with respect to width, and with respect to distance. The first ambiguity concerns how much of the total cause of a belief [or utterance] is relevant to content. (...) The second problem has to do with the ambiguity of the relevant stimulus, whether it is proximal (at the skin, say) or distal. (Davidson 1999, 129–130)

It is not uncommon for the two dimensions of indeterminacy to be conflated. It is also not uncommon to assume that once the indeterminacy pertaining to distance is removed, there is no indeterminacy left. However, it is the indeterminacy pertaining to “width” that is the most significant. To repeat, every external cause of a type of utterance can be seen as having more than one aspect, and thus can be viewed in more than one way. For a determinate aspect to be fixed as the aspect of the object or event to which a creature is responding is, roughly, for the standards of correctness that govern that type of utterance, and thus for its meaning, to be determined. Let us call this, following Verheggen, ‘the aspect problem’.13 This problem would persist even if the distance problem were solved, for singling out the object or event to which the individual is responding does not determine the way in which she conceives of it. It would seem, then, that the individual herself must contribute to the singling out of the meaning-constituting aspect. She herself must conceive of the item in some determinate way. But, Davidson thinks, she cannot acquire such a conception on her own. For an aspect to be singled out is for the distinction between it being correctly and incorrectly said to pertain to an object to be intelligible. But if an individual deals with the world entirely on her own, whatever she takes to be correct may, by her own lights, count as correct. The distinction between correctness and incorrectness is intelligible only if it has a certain independence from how the individual draws it. Otherwise, in Ludwig Wittgenstein’s words, “one would like to say: whatever is going to seem correct to me is correct. And that only means that here we can’t talk about ‘correct’” (1953, Sect. 258). That only means, we might continue, that in the 12 This view is known as ‘semantic externalism’. See, e.g., Davidson (1991a, 2001b). See Verheggen

(2017) for a defence of the claim that a commitment to externalism can be found in Davidson’s early papers and not just in the later ones. 13 See Myers and Verheggen (2016, 18–22). Verheggen is the first one to have insisted on, and to have elaborated on the significance of, the difference between the two problems for a proper understanding of Davidson’s view.

180

O. Sultanescu

case of an individual who deals with the world entirely on her own, the idea that her responses are correct or incorrect, that is, that they are about determinate aspects of their external causes, is unintelligible. We can make sense of the idea that the individual singles out the aspects of the world that her utterances and thoughts are about only if we can make sense of the individual’s recognition of the possibility of her being mistaken, that is, only if we can make sense of her having the concept of truth, which, for Davidson, is equivalent to the concept of objectivity, the idea of things being a certain way regardless of how they are taken to be. For him, she is in a position to have this concept only if she has interacted with another individual and, simultaneously, shared features of the world. Such triangular interactions make room for tensions between their reactions, opening each of them to the possibility of appreciating the perspective of the other. But to appreciate the perspective of another is precisely to recognize the distinction between how things are and how they seem to be, between correctness and incorrectness—it is to have the concept of truth. For this to be the case, the tension between the two individuals’ reactions must turn into disagreement about factual matters—into conversation about the world they share. For “what gives each [individual] the concept of the way things are objectively is the base line formed between the creatures by language” (1982, 105). To summarize, someone is in a position to produce meaningful utterances only if she recognizes that things are a certain way independently of her perspective on them, a recognition which cannot be had unless she has triangulated linguistically with another individual and features of the world. It becomes clear that triangular interactions must pertain to the semantic realm in order to be able to have the role of securing the concept of truth for a speaker, and thus her capacity to mean things by her expressions. In Davidson’s words, “unless the base line of the triangle, the line between the two agents, is strengthened to the point where it can implement the communication of propositional contents, there is no way the agents can make use of the triangular situation to form judgements [and produce meaningful utterances] about the world” (1999, 130). In order for triangulation to render expressions meaningful, it must already involve meaningful uses of expressions, as it were. Thus, the triangulation argument shows semantic reductionism to be hopeless. For it shows that we cannot fully account for meaning facts, we cannot say what it is in virtue of which expressions are meaningful, by appealing to goings-on that are entirely devoid of meaning. What makes it the case that someone’s expressions are meaningful are the very facts of her having engaged in meaning-involving triangulation.14 Granting that the triangulation argument assigns significance to the individual’s possessing the concept of objectivity, one might wonder whether the overall account of meaning delivers a robust conception of the objectivity of meaning itself. For, if the triangulation argument goes through, we shall have to accept that the aspects of the world picked out by individuals in triangular exchanges are precisely the aspects 14 See

Myers and Verheggen (2016, 33–35) for a discussion of the circularity of the account of meaning afforded by the triangulation argument.

Meaning, Evidence, and Objectivity

181

that endow their expressions with meaning. We must thus think of basic triangular exchanges as those in which the applications of the triangulating individuals are correct. But, then, a use’s being taken to be correct in such a context seems to be constitutive of its being correct. To return to Wittgenstein’s warning, if the correctness of the application cannot be disentangled from the taking of the application to be correct, in what sense, if any, can we really talk about correctness? More generally, it would seem that, to appeal to John McDowell’s remarks, “the notion of investigation-independent patterns of application” of an expression, patterns that extend “independently of the actual outcome of any investigation, to the relevant case” (McDowell 1984, 222), cannot be vindicated. In a recent paper, Glüer raises a version of this problem. She claims that the Davidsonian account of meaning determination (as she conceives of it) does not succeed in answering Wittgenstein’s rule-following challenge, because it fails to render meaning “fully and pristinely objective” (Glüer 2017, 88). An account of meaning determination is not fully and pristinely objective when the relation of metaphysical determination—which, according to Glüer, is the principle of charity— “contains a ‘subjective’ element” (Glüer 2018, 233). This is the topic of the next (and last) section.

5 The Objectivity of Meaning As we saw earlier, Glüer claims that Davidson’s account of radical interpretation leaves room for a scenario of the sort that Casey finds himself in. Recall that it is open to Casey to think that “an object (that Alien has considered) satisfies  precisely in case Alien has applied  to it” (Glüer 2017, 87). Glüer concludes that Davidson’s account is unable to rule out scenarios in which whatever the speaker takes to be correct counts as correct, and thus it is unable to ensure that meaning is objective. Indeed, she thinks that “objectivity couldn’t be much further away” (Glüer 2017, 87) from such an account. Glüer suggests that the only way out of this problem is to ensure that the interpreter is in principle able to overcome any difficulty of the sort that Casey encounters. Consequently, the interpreter must be able to detect patterns in the applications of the speaker’s expressions. This is tantamount to saying that she must be sensitive to, or that she must have the natural capacity to discern, the features of the world to which the speaker is reacting. But this entails that the only properties “eligible to be meant”, as Glüer puts it, are properties that the radical interpreter could, in principle, detect. In other words, the worldly features that are relevant to meaning must be features that are within the perceptual reach of the radical interpreter. Facts about the sensitive capacities of the interpreter thus become crucial to meaning determination.15 But, then, Glüer suggests, it is doubtful that the conception of objectivity that it yields can 15 All of this suggests that the radical interpreter is not “merely a dramatic device”, but rather “an essential element of the meaning determining relation” (2018, 227). Glüer changed her mind about

182

O. Sultanescu

be satisfactory. In particular, she thinks that the account cannot provide an answer to Wittgenstein’s paradox about rule-following.16 I obviously cannot spell out the paradox here. Suffice it to say that Glüer takes it to involve a challenge to articulate a constitutive account of meaning. Meeting the challenge requires, in part, meeting its objectivity constraint. Glüer articulates that constraint as follows: Meaningful expressions have conditions of correct application. And, in many cases at least, the correctness of an application has to be an objective matter if there is to be any correctness at all. The correctness of the use of our expressions must have a certain sort of independence from us—it must, for instance, be independent of our actual applications, judgments, and other relevant kinds of reactions. (Glüer 2017, 85)

That Glüer thinks that only answers that accommodate this conception of objectivity are “congenial to Wittgenstein” (2017, 83) is surprising. As many commentators have argued, one lesson of Philosophical Investigations is precisely that the aspiration for an account that preserves this sort of independence from us is as misguided as it is alluring.17 At any rate, whether it is congenial or not to Wittgenstein, it is certainly uncongenial to Davidson. Davidson does not make a secret, so to speak, of the fact that shared capacities for discrimination are necessary for meaning metaphysically to arise. In the absence of discriminatory capacities that are largely similar, which allow individuals to react in similar way to similar things, triangulation, which is central to meaning constitution, cannot so much as get off the ground.18 Indeed, it is only if one fails closely to consider the triangulation papers, and to grant them a role in the project of making the best sense of Davidson’s views, that one might think that there is a genuine question as to whether detectable similarity by the interpreter is necessary for the Davidsonian account of meaning. I conclude by emphasizing that Glüer is preoccupied with an important question, which has to do with the sense in which the standards of correctness yielded by triangular transactions are objective. After all, one consequence of the triangulation argument is that the investigation-independence of the patterns of correctness of our expressions, to borrow McDowell’s term again, turns out to be essentially dependent on our investigative pursuits in triangular scenarios. This thought might seem baffling, even incoherent. I cannot address the question of its apparent incoherence here; this will have to be left for another occasion. But here is a hint as to how I think the answer should go. The Davidsonian picture allows us to say that whether or not an application is correct remains an investigation-independent matter. The key here is to recognize that the investigation-independence of correct applications does not require that the pattern of correctness extends “of itself” (McDowell 1984, 256) to new cases. Indeed, as McDowell intimates, the idea that the pattern extends of itself, whether the radical interpreter is merely a dramatic device. She initially thought that it is (2011, 136–138), and later came to deny it. 16 See Wittgenstein (1953, Sect. 201). 17 See, e.g., Pears (1988). 18 He says, for instance, that “what makes communication [and hence meaning] possible is the sharing, inherited and acquired, of similarity responses” (1990a, 61) and that “a condition for being a speaker [and thinker] is that there must be others enough like oneself” (1992, 120).

Meaning, Evidence, and Objectivity

183

without any contribution from us, is to be rejected. But accepting that we contribute to the constitution of the pattern, and thus to the way in which it extends, need not commit us to the idea that correctness is to be equated with the product of our investigations. It is the fact of our being investigating creatures that matters for the constitution of correctness, not the actual outcomes of our investigations. Acknowledgments I am very grateful to Claudine Verheggen for her comments on earlier drafts of this paper and for our many conversations about Donald Davidson’s conception of meaning. I would also like to thank Christopher Campbell, William Child, Robert Myers, and Sam Steadman for discussion.

References Davidson, D. (1963). Actions, reasons, and causes. In Essays on actions and events. Oxford: Clarendon Press, 1980. Davidson, D. (1973). Radical interpretation. In Davidson 1984. Davidson, D. (1974). Belief and the basis of meaning. In Davidson 1984. Davidson, D. (1975). Thought and talk. In Davidson 1984. Davidson, D. (1976). Reply to Foster. In Davidson 1984. Davidson, D. (1982). Rational animals. In Davidson 2001a. Davidson, D. (1984). Inquiries into truth and interpretation. Oxford: Clarendon Press. Davidson, D. (1990a). Meaning, truth, and evidence. In Truth, language, and history. Oxford: Clarendon Press, 2005. Davidson, D. (1990b). The structure and content of truth. The Journal of Philosophy, 87(6), 279–328. Davidson, D. (1991a). Epistemology externalized. In Davidson 2001a. Davidson, D. (1991b). Three varieties of knowledge. In Davidson 2001a. Davidson, D. (1992). The second person. In Davidson 2001a. Davidson, D. (1993). Reply to Jerry Fodor and Ernest Lepore. In Ralf Stoecker (Ed.), Reflecting Davidson. Berlin-New York: Walter de Gruyter. Davidson, D. (1995). The objectivity of values. In Davidson 2004. Davidson, D. (1999). The emergence of thought. In Davidson 2001a. Davidson, D. (2001a). Subjective, intersubjective, objective. Oxford: Clarendon Press. Davidson, D. (2001b). Externalisms. In P. Kotatko, P. Pagin & G. Segal (Eds.), Interpreting Davidson. Stanford, California: CSLI Publications. Davidson, D. (2004). Problems of rationality. Oxford: Clarendon Press. Glüer, K. (2011). Donald Davidson: A short introduction. Oxford: Oxford University Press. Glüer, K. (2017). Rule-following and charity: Wittgenstein and Davidson on meaning determination. In C. Verheggen (Ed.), Wittgenstein and Davidson on language, thought, and action. Cambridge: Cambridge University Press. Glüer, K. (2018). Interpretation and the interpreter. In D. Ball & B. Rabern (Eds.), The science of meaning: Essays on the metatheory of natural language semantics. Oxford: Oxford University Press. Greenberg, M. (2005). A new map of theories of mental content: constitutive accounts and normative theories. Philosophical Issues, 15(1), 299–320. McDowell, J. (1984). Wittgenstein on following a rule. Reprinted in Mind, value, and reality. Cambridge, Mass.: Harvard University Press, 1998. Myers, R. H., & Verheggen, C. (2016). Donald Davidson’s triangulation argument: A philosophical inquiry. New York: Routledge.

184

O. Sultanescu

Pagin, P. (2002). Rule-following, compositionality and the normativity of meaning. In D. Prawitz (Ed.), Meaning and interpretation. Stockholm: Almqvist & Wiksell. Pagin, P. (2013). Radical interpretation and the principle of charity. In E. Lepore & K. Ludwig (Eds.), A companion to Donald Davidson. Oxford: Wiley Blackwell. Pears, D. (1988). The false prison: A study of the development of Wittgenstein’s philosophy (Vol. Two). Oxford: Clarendon Press. Verheggen, C. (2007). Triangulating with Davidson. The Philosophical Quarterly, 57(226), 96–103. Verheggen, C. (2017). Davidson’s semantic externalism: From radical interpretation to triangulation. Argumenta, 3(1), 145–161. Wittgenstein, L. (1953). Philosophical investigations (4th edn.). Edited by P. M. S. Hacker and Joachim Schulte. Translated by G. E. M. Anscombe, P. M. S. Hacker & Joachim Schulte. Oxford: Wiley-Blackwell.

Davidson’s Triangulation Argument in the Logic of Actions Syraya Chin-Mu Yang

When people talk to each other, they never say what they mean. They say something else and you’re expected to just know what they mean. Alan Turing (The Imitation Game Quotes. n.d. Quotes.net. https://www.quotes.net/mquote/1097787.)

Abstract In this chapter, I provide a formal machinery to complement the epistemic background in Davidson’s triangulation argument by presenting a structural modelling of the argument in the framework of an epistemic logic of actions. I focus on two aspects. Firstly, to guarantee that the propositional content of what the speaker says in the ongoing communication is justified true beliefs, I suggest that the speaker’s utterances must be assertions governed by the knowledge norm, assuming Davidson’s characterization of knowledge in terms of holistically justified true beliefs. Secondly, I claim that ongoing communication succeeds only if what the speaker says becomes common knowledge. Accordingly, the required underlying system for the proposed logic of actions will be a logic of common knowledge, in which the notion of common knowledge is characterized in terms of the knowledge account of assertion. A semantics appropriate for the logic of common knowledge will be given. I further propose that the transition from the speaker’s initial assertion to common knowledge can be illuminated by virtue of individual agents’ learning and the group learning, resulting from what the hearer learns from the speaker’s assertion and what the speaker learns from the hearer’s response, and the following interactions between them. Two act-operators for individual agents’ learning and group learning are introduced; the required semantic rules are stipulated. Keywords Triangulation argument · Logic of actions · Common knowledge · Knowledge account of assertion · Learning · Group learning This chapter is in its final form and it is not submitted for publication anywhere else. S. C.-M. Yang (B) Department of Philosophy, National Taiwan University, 1, Sec. 4, Roosevelt Road, Taipei 10617, Taiwan e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 S. C.-M. Yang and R. H. Myers (eds.), Donald Davidson on Action, Mind and Value, Logic in Asia: Studia Logica Library, https://doi.org/10.1007/978-981-15-7230-2_11

185

186

S. C.-M. Yang

1 Introduction Donald Davidson has devoted his early work mainly to the theory of action and the philosophy of language, independently. Starting with a most noted paper (Davidson 1963), the primary concern of Davidson’s philosophy of action is to elaborate ‘the role of causal concepts in the description and explanation of human action’. Shortly later, inspired by Tarski’s axiomatization of the theory of truth, Davidson proposes a truth-conditional theory of meaning, contending that the meaning of a sentence (of a given natural language) can be characterized by virtue of its truth conditions. The idea is that to know the meaning (roughly, the propositional content) of a sentence is to know its truth conditions. But with his repudiation of the role that social convention plays in the Wittgensteinian use theory of meaning and his rejection of the application of the notions of satisfaction and assignment in Tarski’s formal semantics to natural language, Davidson proposes the well-known radical interpretation, according to which the hearer (interpreter) is said to know the meaning of a sentence s, only if it is uttered by a speaker, and the hearer can interpret s appropriately by attributing to the speaker some justified beliefs that are true, or at least held to be true. The attributed beliefs will be then taken to be the truth conditions of s. That said, Davidson offers no specification to elucidate how radical interpretation can be carried out. Things become worse when Davidson has to embrace a holistic conception of justification, which might denounce the metaphysical import of the notion of truth characterized by the application of Tarski’s T-scheme. Davidson thus needs to show that the alleged truth conditions are not purely semantic. To deal with these issues, Davidson has later developed an evolutionary program, known as the triangulation argument, which is a simplified, ideal model for communication amongst rational agents, focusing on the interactions between the speaker, the hearer, and their understanding of, and responses to, the external world. When uttering a sentence s of the language in use, the speaker is assumed to hold a certain collection of (holistically) justified true beliefs, referred to as the speaker’s prior theory, which suffices to serve as evidence for a certain justified true belief that the speaker intends to communicate by uttering a sentence s. Simultaneously, the speaker would also assume that the hearer may hold some true beliefs, referred to as the hearer’s prior theory. Based on these assumptions, the speaker would then think that the hearer may understand what she says. Now, the ongoing communication succeeds only if the hearer can appropriately attribute to the speaker certain true beliefs, which would justify the propositional content of what the speaker says. We may then take the attributed true beliefs as the truth conditions of the very sentence s. The triangulation argument thus illuminates how the hearer can understand what the speaker utters. Still, there is no guarantee that what the speaker utters is a justified true belief, although the speaker’s prior theory is presupposed. Moreover, communication is essentially a dynamic process of interactions amongst agents. In particular, verbal communication is possible only if the propositional contents of what the speaker

Davidson’s Triangulation Argument in the Logic of Actions

187

utters become common knowledge, which can only be rendered through the interactions amongst agents. It seems to me that Davidson’s appeal to the speaker’s passing theory and the hearer’s passing theory is intended to serve for this goal. However, no mechanism of the transition from what the speaker says to common knowledge has been shown. In this paper, I provide a formal machinery to complement the epistemic background in Davidson’s triangulation argument. To this end, I will construct a semantic framework for a dynamic epistemic logic of actions with some additional modal operators (act-operators) to signify certain specific actions, so that the intrinsic properties of the specified actions, and the reasoning about the consequences of these actions, can be explicitly axiomatized. The idea is to regard the speaker’s utterances in an ongoing communication as speech acts of some sort. Also, the hearer’s understanding of whatever the speaker utters can be signified in terms of a second type of action, namely learning—what the hearer understands is precisely what she learns from the speaker’s utterance. Now, model-theoretically, we may take an action to be a mapping from an initial state of the world (wherein the very action is to be done) to a set of (possible) states of the world (showing what will happen after the action is done). Typically, by virtue of such a dynamic structural analysis of the semantic framework for a desired logic of actions, we may explore how the agents’ actions change others’ epistemic states and how this process brings about successful communication. For simplicity’s sake, we focus on verbal communications and their impacts on the agents’ epistemic states. A logic of actions thus constructed would provide a theorization of the information that the agents may offer/gather in an ongoing communication based on the transition/change of agents’ epistemic states. I start with a survey of anti-realism in the theory of meaning, which paves the way for Davidson’s triangulation argument. Then I construct a comprehensible framework for a logic of actions suitable for modelling the triangulation argument. If by uttering a sentence, the speaker intends to communicate some of her belief, the utterance is to serve as an outward formulation of what she believes. Following Frege’s footsteps of taking assertions to be the outward formulation of our inward judgments, we may take assertion as the kind of speech acts that a speaker is performing when uttering a sentence in an ongoing communication. Accordingly, the language in use needs an act-operator for assertion, say ‘Ai ’ (for each i in the group of agents G). Moreover, to assure that what the speaker utters are justified true beliefs, a certain norm should be imposed on the speaker’s assertion. Now, given Davidson’s (implicit) idea of taking holistically justified true beliefs to be the characterization of the concept of knowledge, it would be appealing to endorse the knowledge norm (or knowledge account) of assertion.1 This would guarantee that whatever the speaker utters in a successful communication can only be what she knows, or (holistically) justified true beliefs. Next, I argue that an ongoing communication succeeds only if the hearer can offer an appropriate interpretation to turn the propositional content of what the speaker 1 See

Williamson (2000). Also, Yang (2014), for a model-theoretical defence of the knowledge account of assertion.

188

S. C.-M. Yang

asserts into common knowledge. This line of thought suggests that the desired logic of actions needs an epistemic logic of common knowledge to serve as the underlying system, wherein the concept of common knowledge can be characterized in terms of the knowledge account of assertion. To formulate explicitly the transition from the proposition content of the speaker’s assertion to common knowledge, we appeal to a second basic action in the logic of actions, namely learning. This is a multipletiered task. At the initial stage, what the hearer understands/knows upon the presence of the speaker’s assertion can be characterized in terms of the hearer’s learning of the speaker’s assertion. But notice that the hearer’s learning per se at this stage is not sufficient for giving rise to the required common knowledge. The speaker has to know simultaneously and spontaneously that the hearer has already known the propositional content of her assertion. Again, what the speaker knows at this stage can be specified in terms of the speaker’s learning of the hearer’s response to his assertion. Accordingly, we need a second type of act-operators for individual agents’ learning, say ‘Li ’ (for each agent i in G), to signify what the hearer understands and what the speaker knows regarding the hearer’s response. Still, sometimes, the hearer has to learn something more from the speaker’s response again, and so on, until eventually the required common knowledge is obtained. For ease of discussion, we will introduce an additional act-operator for ‘group learning’, say ‘LG ’, to show how a piece of common knowledge can result from the speaker’s assertion at the initial state via group learning. As far as the required underlying system for the desired epistemic action logic is concerned, I shall adopt an epistemic logic of common knowledge, the semantics of which I proposed previously (Yang 2016). The models will be constructed; the knowledge account of assertion and the assertion account of common knowledge will be specified in the constructed framework. The required semantic rules for the additional act-operators Ai , Li , and LG will be stipulated, respectively. Then I show how the transition of what the speaker asserts to common knowledge can be carried out.

2 The Pursuit of an Antirealist’s Theory of Meaning In the tradition of philosophy, realism has been playing a dominant view in the theory of meaning, which claims that there are entities of some sorts, independent of, or supervenient upon, human minds to serve as the meanings of linguistic expressions. A variety of entities have been posited, such as universals, John Lock’s ideas, Frege’s thoughts, Russell and Wittgenstein’s propositions, facts, states of affairs, immediate-experience, sensory-impressions, or sense-data/the given, Prior’s intensional objects, Armstrong’s truth-makers, etc. However, misgivings over the intelligibility of the posited entities (meaning) remain. By contrast, antirealists opt to explain how linguistic expressions can be meaningful without the appeal to posited abstract entities. A notable starting point for contemporary anti-realism is Wittgenstein’s (1953) use theory of meaning, which takes the meaning of an expression to

Davidson’s Triangulation Argument in the Logic of Actions

189

be its use. ‘Don’t ask for the meaning, ask for the use’, as the slogan highlighted. Following this line of thought, ordinary language philosophers paid much more attention to speech acts (Austin 1963, 1975) and speakers’ intention (Grice 1989). Perhaps the most salient antirealist theory of meaning can be found in Quine’s behaviouristic program of radical translation, according to which talk of the meaning of a sentence s of a language L S makes sense only if an L S -speaker utters s and the hearer (the translator), based on her understanding of (the content of) what the speaker uttered via speaker’s obvert behaviours, can construct a translation manual M, containing a sentence s* which translates s into the hearer’s own language, say L h . Quine concludes that ‘meaning is what an assertion has in common with its translation s* ’ (Quine 1960 [2015]: 32, also 1992). Clearly, to explain what it is to make a sentence meaningful, all that is required is to pay attention to the interaction between the agents involved, including the speakers’ intention and speech acts, and the hearer/translator’s responses. The meaning of a sentence is what is shared in common by the speaker and the hearer. So far so good. However, the appeal to the speaker’s intention and speech acts, and the hearer’s understanding based on the speaker’s obvert behaviour, may weaken the objectivity of the meaning of the language. In particular, if the meaning of a sentence has something to do with our knowledge of the external world, a metaphysical conception of truth, or some other equivalent concepts, is indispensable. After all, metaphysically, ‘truth’ is just a synonym of ‘factiveness’: saying that a sentence is true is no more and no less than saying that what the sentence depicts is a fact— ‘truth hinges upon reality’, as Quine famously says. For externalists, a metaphysical concept of truth is what anchors the thought onto the reality (the external world). Unfortunately, such a concept of truth seems to fade away along the aforementioned antirealist’s approaches. To restore the substantial role that the metaphysical conception of truth plays in a theory of meaning, Davidson (1967) brought in Tarski’s (1933 [1983], 1944 [1986]) axiomatization of the theory of truth and apply the well-established T-scheme to linguistic sentences—for a given sentence s, ‘s’ is true if and only if p, where p is the interpretation of the sentence s. However, since Davidson does not adopt the Tarskian notion of satisfaction based on the interpretation of the basic non-logical symbols and the assignment of variables, the required truth conditions for sentences in natural language cannot be characterized in the Tarski-style semantics. An alternative approach to the truth condition of the given sentence is called for. This is the primary concern of Davidson’s truth-conditional theory of meaning. Following the footsteps of Austin, Grice, and Quine, Davidson takes it that talk of the meaning of a sentence s of a language makes sense only if a speaker utters s, intending to communicate a certain true belief, say B, that the speaker has in mind or holds to be true. (Hereafter, we use B(s) to signify the belief that the speaker has in mind when uttering a sentence s.) Davidson further assumes that the speaker holds some evidential-based beliefs sufficient to justify B(s). Now upon the understanding of what the speaker utters, the interpreter (i.e. the hearer) may appropriately ascribe/attribute some beliefs to the speaker. At this stage, to make sure that the interpreter is able to attribute beliefs to the speakers, Davidson has to appeal to

190

S. C.-M. Yang

the Principle of Charity, according to which, the interpreter and the speaker share something in common. Accordingly, the hearer could then assume that the speaker holds such and such beliefs as the evidence for her belief B(s). The ongoing communication succeeds only if the attributed beliefs serve as the evidence for the speaker’s belief B(s). It is then striking that the collection of beliefs taken as evidence for B(s) is precisely what justifies the truth of B(s), which is tantamount to the truth conditions of s. The interpreter’s attribution of beliefs to the speaker thereby provides a specification of the truth conditions of the sentence s. Thus, for the interpreter, an appropriate attribution of beliefs to the speaker is no more and no less than an assignment of the meaning to the very utterance of s, which in turn shows that the interpreter knows the meaning of the sentence s. Accordingly, to know the meaning (roughly, the propositional content) of s is to know its truth condition. Davidson was soon aware that something is still missing. For one thing, there is no way to show that the specified truth conditions do have a certain metaphysical import, rather than merely a semantic treatment. If truth conditions lack metaphysical import, the appeal to the truth conditions would not help at all. Moreover, the Principle of Charity, and a fortiori, the attribution of beliefs to the speaker, cannot be wellgrounded. Things would be worse if we accept Davidson’s ‘no-language thesis’—the view that ‘there is no such thing as a language, not if a language is anything like what many philosophers and linguists have supposed. There is therefore no such thing to be learned, mastered, or born with’ (Davidson 1986 [2005a]: 107). The nolanguage thesis implies that linguistic expressions have no fixed meaning. Talk of the meaning of a sentence makes sense only if the sentence is uttered during an ongoing communication, which can only be characterized by virtue of the dynamic interactions amongst the speaker, the hearer, and their responses to each other. Apparently, some more convincing argument is called for, in addition to the appeal to the attribution of beliefs. This line of thought has led Davidson to develop the well-known triangulation argument, to which we turn our attention now.

3 The Triangulation Argument It is somewhat interesting to observe that although Davidson has developed the triangulation argument in a series of papers from the early 1980s to the early 2000s, he offers no well-organized version of the argument.2 However, Davidson has occasionally summarized the main theme of the argument. For instance, he noted, 2 More specifically, the development of the triangulation argument can be traced mainly via a series

of papers as follows: ‘A unified theory of thought, meaning and action’ (1980)—presumably the starting-point, ‘Rational animals’ (1982), ‘A nice derangement of epitaphs’ (1986)—a somewhat informal version, ‘Three varieties of knowledge’ (1991), ‘The second person’ (1992)—perhaps the most detailed and well-organized version, ‘The social aspect of language’ (1994)—showing the acceptance of Quinean social externalism and a rejection of Wittgenstein’s appeal to conventional social externalism, and ‘Externalisms’ (2001d)—a more updated version, in particular, for a substantial explanation concerning how thought and language are anchored to the world.

Davidson’s Triangulation Argument in the Logic of Actions

191

[T]o have thoughts, and so to mean anything in speaking, it is necessary to understand, and be understood by, a second person … the objectivity which thought and language demand depends on the mutual and simultaneous responses of two or more creatures to common distal stimuli and to one another’s responses. This three-way relation amongst two speakers and a common world I call ‘triangulation’. (Davidson 1991 [2001c]: xv)

Recently, Myers and Verheggen also highlight the underlying thought of the argument by claiming that The triangulation argument purports to show that only someone who has interacted linguistically with another person and the world they share, i.e. only someone who has actually triangulated, could have a language and thoughts. (Myers and Verheggen 2016: 1, the authors’ original italics)

In other words, only when rational agents have actually triangulated, they could have a language to communicate verbally. And the thought they have in mind makes the linguistic expressions in use meaningful. For the sake of discussion, I outline a structural analysis of the triangulation argument in what follows. • The tripod. The constituents of the tripod include a speaker, a hearer (the interpreter), and the external world. Note that the speaker and the hearer are supposed to be rational agents. • The Principle of Charity (the fundamental version). With a rejection of Quine’s appeal to observation sentences and patterns of stimuli, Davidson presupposes a much more comprehensive version of the Principle of Charity, which states that human agents by nature have in common a large number of true beliefs. In particular, Davidson argues that if the majority of our beliefs are false, communication would be impossible. • Speaker’s intention. To activate communication, the speaker should have the intention of sharing with the hearer ideally a certain justified true belief, or at least a belief held as true. • The speaker’s speech acts—uttering sentences. For the speaker, holding beliefs is an inward epistemic state, which is essentially private, subjective, and hardly observable. To communicate a belief verbally, the speaker must perform a certain obvert action, a speech act in character, typically by uttering a sentence, say s, of the language in use to serve as an outward formulation of the speaker’s inward mental state of holding the very belief (i.e. B(s)). • The speaker’s prior theory. When the speaker intends to communicate a belief B(s), typically a justified true one, to the hearer, the speaker by and large holds an appropriate collection of (holistically) justified true beliefs which could serve as evidence for B(s) being a justified true belief. Here, a collection of justified true beliefs is assumed simply because evidence for justified true beliefs must be justified true beliefs as well. Davidson calls the collection of justified true beliefs that the speaker holds as evidence for the belief B(s) the speaker’s prior theory. • The hearer’s prior theory. When the speaker has the intention of communicating the propositional content of B(s) to the hearer, the speaker, apart from having her own prior theory, must be also motivated by the presupposition that the hearer

192

•

•

•

•

S. C.-M. Yang

would have a corresponding collection of (holistically) justified true beliefs which the hearer could take as evidence for the belief B(s). Of course, the hearer may not have all the beliefs that the speaker assumed. However, to make the ongoing communication succeed, the hearer must hold a certain substantial portion of them, which suffices, perhaps together with some of her own (holistically justified true) beliefs, to serve as evidence for B(s). The collection of justified true beliefs characterized in this way is referred to as the hearer’s prior theory. The Principle of Charity (particular cases). When the speaker’s intention of communicating a belief to the hearer is triggered, she would assume that the hearer might have already shared with her some justified true beliefs in common, which would pave the way for the hearer knowing what the speaker utters, namely the belief that the speaker holds. Simultaneously, upon the presence of the speaker’s utterance for a belief B(s), the hearer would also assume that the speaker holds some justified true beliefs as the evidence for the belief B(s). Otherwise, the hearer would be in no position to attribute any belief to the speaker. At this stage, Davidson appeals to the Principle of Charity again to articulate the legitimacy of the aforementioned assumptions. One can see clearly that by the Principle of Charity thus characterized, the speaker’s and the hearer’s prior theories would share some justified true beliefs in common. This would also guarantee the legitimacy of the hearer’s attribution of beliefs to the speaker. Attribution of beliefs, the assignment of meaning, and truth conditions. In an ongoing communication, upon the presence of the speaker’s utterance of a sentence (say, s), the hearer shows her understanding of what the speaker says (e.g. B(s)) by attributing to the speaker some beliefs that the hearer believes that the speaker would hold as the evidence for the belief B(s). Clearly, the hearer’s attribution of beliefs to the speaker not only signifies the hearer’s understanding of whatever the speaker utters but also specifies the evidence for the speaker’s belief B(s). Admittedly, for Davidson, the hearer’s understanding of the speaker’s uttering a sentence s for expressing the belief B(s) is tantamount to the assignment of the meaning to the very sentence s. It follows that the hearer’s attribution of beliefs to the speaker can be treated as the hearer’s assignment of the meaning to what the speaker utters (i.e. the sentence s). Now, if the attributed beliefs can be taken as the evidence (for the speaker) for the belief B(s) being justified true, the attributed beliefs can be in turn treated as the truth conditions of s. That is, to know the meaning of a sentence s (viz. the understanding of the speaker’s utterance—the attribution of beliefs to the speaker) is to know its truth conditions. The transition from the speaker’s belief to common knowledge. The ongoing communication succeeds only if the hearer attributes appropriate beliefs to the speaker. And so, not only the hearer knows the meaning of what the speaker utters, the speaker also knows that the hearer knows what the speaker intends to show, and so on. Eventually, the propositional content of what the speaker utters thus becomes common knowledge. The speaker/hearer’s passing theory. So far, Davidson has shown that with regard to the speaker’s utterance of s, the speaker’s prior theory and the hearer’s prior theory must share something in common, including the evidence for B(s). In

Davidson’s Triangulation Argument in the Logic of Actions

193

particular, the appropriate collection of beliefs attributed to the speaker must be common knowledge. Also, when the ongoing communication succeeds, what the speaker utters would become common knowledge as well. And of course, more common knowledge follows from the attributed beliefs and B(s). The totality of common knowledge obtained in this way seems to be what Davidson calls the speaker/hearer’s passing theory. At this point, a brief remark on Davidson’s notions of justified true beliefs and knowledge is helpful. As is well-known, Davidson takes as the starting point an intuitive assumption that we are rational agents with a large number of true beliefs. And by the Principle of Charity, we share with others a collection of true beliefs about (i) one’s own minds, (ii) others’ minds, and (iii) the external world. Davidson, as an empiricist, insists that all true beliefs must be empirically justified. Inspired by Quine, a holistic justification for true beliefs would be more appealing than any version of the principle of verification. Accordingly, Davidson maintains that every true belief should be holistically justified to the extent that even a true belief about one’s own minds has to be justified by certain knowledge of the external world and knowledge of others’ minds. Justified true beliefs are then objective and, more importantly, can be anchored to the external world. It is striking that Davidson implicitly takes holistically justified true beliefs as the characterization of knowledge. Presumably, this is the reason why Davidson claims that human agents, as rational animals, have three varieties of knowledge, which are mutually interdependent, and none of which can be reduced to the other two. Hence, all these three varieties are indispensable.3 Of course, Davidson accepts margin for error so that he often prefers to talk about beliefs, and sometimes merely ‘propositions held as true’. For the sake of simplicity, we may say from now on that the propositional content of what the speaker intends to communicate when uttering a sentence is a certain piece of knowledge that the speaker has. We may sum up that the triangulation argument displays the specification of the meaning of the sentence that the speaker utters, and a fortiori, the thought that the speaker has in mind. ‘[A]n interconnected triangle … constitutes a necessary condition for the existence of conceptualization, thought and language, so Davidson writes’ (Davidson 2001d: 7). In particular, since the justification of true beliefs must involve some common knowledge about the external world, a metaphysical conception of truth can be restored. Moreover, the transition from the speaker’s initial beliefs to common knowledge is obtained by virtue of a dynamic interaction between the speaker, the hearer, and their responses to each other. This is what Davidson concludes when he writes ‘[t]he social scene provides a setting in which we can make sense of the 3 A more explicit remark from Davidson’s later work: ‘What skeptics doubt is whether we have good

reason to hold those of our beliefs that happens to be true, that is, whether we have knowledge’. (Davidson 2001d: 1). As a matter of fact, this line of thought is close to the so-called defeasibility theory which defines ‘knowledge’ as true justified belief that is stable under belief revision with any new evidence: ‘if a person has knowledge, then that person’s justification must be sufficiently strong that it is not capable of being defeated by evidence that he does not possess’ (Pappas and Swain 1978: 27).

194

S. C.-M. Yang

transition from mere dispositions (of the speaker’s) to conceptualization (objective to the speaker and the interpreter)’ (Davidson 2001d: 8, my italics). Radical interpretation is precisely a dynamic process of the required transition from a specific mental state, in which the speaker holds a certain belief based on his prior theory, to some other’s mental state, in which both the hearer and the speaker share some common knowledge based on their passing theory.4

4 A Diagnosis—What’s Missing in the Triangulation Argument? The forgoing analysis of the triangulation argument shows that, apart from the Principle of Charity, two salient phenomena are involved in a successful communication. Firstly, what the speaker intends to communicate is essentially a justified true belief, or a belief held as true. But how can we be sure that when the speaker utters a sentence, the propositional content of her utterance must be a justified true belief? Davidson simply takes this for granted. It would be better to impose some constraints on the speaker’s utterances so that the epistemic force of the speaker’s utterances would guarantee that the propositional content of what the speaker says is a justified true belief. In fact, Davidson is aware of this requirement but admits that there is no way out, as he notes later: There are no linguistic conventions or rules which determine the force of an utterance, and so no regular relation between force and the literal truth or meaning of what is said. We cannot learn what sets apart utterance that aim at literal truth by the study of their force. (Davidson 2005b: 123)

Secondly, the triangulation argument shows that when communication succeeds, the hearer and the speaker would share some justified true beliefs in common, which is essentially the so-called passing theory. In particular, the propositional contents of the speaker’s initial utterance must become common knowledge. Still, Davidson does not say much about what role common knowledge plays in the triangulation argument (apart from the Principle of Charity). Accordingly, he does not explain how to implement the transition from the speaker’s beliefs to common knowledge. The aforementioned analysis suggests that if we take the speaker’s utterance as a kind of speech act, and if we take the transition from the speaker’s belief to common knowledge as the results of the interaction between the speaker and the hearer, the triangulation argument intrinsically involves two kinds of actions. The first one is concerned with the speaker’s speech act to initiate an ongoing communication. In order to show that ideally the speaker intends to communicate only justified true beliefs, or knowledge, a certain norm that governs the speaker’s speech act should be articulated. A second type of action indicates that the hearer can get to 4 It

would not be surprised to notice a remark in the same line of thought from van Benthem (2010: 135): ‘Common knowledge high-lighted, “mixes “factual” information with “social” information about the information of others”’.

Davidson’s Triangulation Argument in the Logic of Actions

195

know/understand what is embodied in the speaker’s speech act, and that the speaker can get to know/understand what the hearer knows in turn. From an epistemological point of view, believing/knowing is a kind of epistemic (mental) state, and an agent can show that she has actually learned of some belief/knowledge by specifying a transition of the change of her epistemic state wherein she has not yet believed/known something to a new epistemic state in which she has already got the very piece of belief/knowledge. That is to say, the hearer’s understanding of whatever the speaker says can be characterized in terms of the impact of the speaker’s speech acts on the hearer’s epistemic states; and this can be in turn characterized by the hearer’s learning something from the speaker’s speech acts. And so, given that the hearer knows something from speaker’s speech act, there must be a certain mechanism showing that the hearer does learn what is embodied in the speaker’s speech act. Also, the speaker must be able to learn of something from the hearer’s learning of what she asserts. This action of the speaker’s is very crucial to successful communication, as Davidson remarks: But it does make a significant difference if the creature associates the responses of other creatures with aspects of the world to which it also responds … (Davidson 2001d: 12)

We need a mechanism to describe the hearer’s learning (from the speaker’s speech act) and the speaker’s learning (of the hearer’s responses to her speech act). One can see clearly that if these two types of actions can be formulated explicitly within an epistemic setting, we would have a more illuminating version of the triangulation argument. In what follows, I will provide a formal machinery to characterize the required epistemic setting by presenting a structural modelling of the triangulation argument in the framework of an epistemic logic of actions, as the actions involved explicitly change the cognitive/epistemic states of the speaker and the hearer. It is appealing to take a logic of common knowledge as the required underlying system for the desired epistemic logic. Since our primary concern here is with the construction of a framework for a logic of actions suitable for the triangulation argument, we shall then focus only on the so-called epistemic actions, hence epistemic action logic.

5 The Appeal to Epistemic Action Logic: A Proposal 5.1 A Relational Account of Actions in Epistemic Action Logic In general, an epistemic action logic, a dynamic logic in character, takes a certain epistemic logic, e.g. an S5-logic of knowledge, as the required underlying system. The language in use is an extension of the language for the underlying system by adding some extra modal operators to signify the specified actions, referred to as

196

S. C.-M. Yang

‘act-operators’.5 The intrinsic properties of agents’ actions can thus be formulated explicitly and reasonings about the impacts of actions can be axiomatized. Working on the framework of possible-worlds semantics, we characterize an action in terms of a mapping from an initial (epistemic) state, wherein the action in play is to be performed, to a (possibly empty) set of (possible) states wherein the results of the impacts of the action can be displayed. Let M be an epistemic model for a multi-agent system for a group of agents G, that is, a complex of the form. M = S, {Ra }a∈G , V P , where • S, a non-empty set of states, and • {Ra }a∈G , a set of accessibility relations for each member a ∈ G (more specifically, {Ra }a∈G , signifying labeled transitions for successful executions of the action under investigation performed by agents in the group G), and • V P , an assignment of a truth value to each propositional letter in P so as to specify a set of true propositional letters in every state. We now consider an action, to be signified by a modal operator ‘α’, which is about to be executed by an agent in a state (referred to as the initial state), say s[α] , in M. To the initial state s[α] , a set of states accessible from s[α] , say s1 , …, sn , … (denoted by Ss[α] ) is associated in that each sk in Ss[α] displays a possible consequence of the execution of an action α. And so, the totality of Ss[α] can be treated as the result of the impacts of the action α in M. In general, there are three noticeable cases to be considered: 1. If Ss[α] is empty, the action α has been aborted in S. 2. If Ss[α] is a singleton, the action α is deterministic on S. 3. If Ss[α] has more than one element, the action α is non-deterministic on S. Model-theoretically, it is then reasonable to characterize the action α in terms of a set of ordered pairs, a subset of S × S, defined by {si[α] , sk | for any si[α] ∈ S, and sk ∈ Ssi[α] }. Semantically, the binary relation thus constructed can be treated as the intended interpretation of the act-operator α.6 In particular, if we take speech acts as dynamic interaction amongst a group of agents, an action executed by an agent can be characterized by a specification of the change of agents’ epistemic states resulting from the completion of the action. It is in this sense that the dynamic aspect of the triangulation argument, especially the process of radical interpretation, can be specified in a possible-worlds-semantics-based framework for an epistemic action logic. 5 Sometimes,

the language thus constructed for an epistemic action logic is referred to as a polymodal language, which includes two or more types of modal operators, typically, a certain type of epistemic operators for agents’ epistemic states such as ‘knowing’ (e.g. ‘Ki ’) and ‘believing’ (e.g. ‘Bi ’), and some more modal operators for other types of modalities, such as act-operators for agents’ actions. 6 For more details, see Vaughan Pratt, ‘Logic & action’, in van Benthem, et al. (2016); also van Ditmarsch, et al. (2008), especially Chap. 5.

Davidson’s Triangulation Argument in the Logic of Actions

197

5.2 Act-Operators for Assertion, Learning, and Group Learning We have suggested that to construct an epistemic action logic appropriate for a structural analysis of the triangulation argument, two types of act-operators are required. The first one is intended to signify the speaker’s initial utterance so that the propositional content of what the speaker utters can be formulated and guaranteed to be justified true beliefs, or knowledge. Historically, this line of thought can be traced back to Frege (1879, 1918–19) who recognizes that when an agent intends to formulate her judgments explicitly by linguistic expressions, some sort of assertoric force must be imposed on the agent’s linguistic formulation. According to Frege, judgment is a kind of inward mental activity, which appears to be private and subjective in character. To guarantee that the propositional content of an affirmative judgment is sharable and objective, some sort of outward formulation of judgments is required. Frege suggests that the mode of linguistic expressions qualified for this purpose should be entitled as assertion due to the assertoric force which is imposed on the linguistic expression in use. The imposed assertoric force can thereby guarantee that the propositional contents of affirmative judgments are objective truths.7 Following Frege’s footsteps, Davidson takes a further move and notes that What we do in making an assertion is represent ourselves as believing what we say, … Assertion, like issuing an order or asking a question, is a force we give to some utterances, and an utterance has a certain force only if the speaker intends his or her audience to be aware of the force and so of the intention. (Davidson 2005b: 123)

It is then appealing to stipulate that assertion can serve as the type of speech act that the speaker performs when uttering a sentence in communication. Accordingly, we need a set of unary act-operators, say ‘Ai ’ (for each i ∈ G, referred to as assertionoperators), and ‘Ai p’ read ‘the agent i asserts p’, for a sentence p. Still, the epistemic force of the speaker’s assertion per se is not sufficient for guaranteeing that what the agent intends to communicate are justified true beliefs. To meet such a requirement, a certain normative constraint should be imposed on the speaker’s assertion. Williamson (2000) examines three kinds of norms (or rules, or accounts) of assertion, including the truth norm (‘One must: assert p only if p is true’), the warrant norm (‘One must: assert p only if one has warrant to assert p’) and the

7 In

Begriffsschrift, Frege’s introduced the well-known logical notation ‘’, claiming that As a constituent of the sign , the horizontal stroke combines the symbols following it into a whole; assertion, which is expressed by the vertical stroke at the left end of the horizontal one, relates to the whole thus formed. (1879: 2, original italic)

So ‘p’ can be read as ‘It is asserted that the (propositional) content of the sentence p is an affirmative judgment, or is judged as true’. This can be viewed as the origin of the truth norm/account of assertion. But if affirmative judgments can be viewed as the kind of epistemic state equivalent to knowledge, this can be treated as the origin of the knowledge norm/account of assertion.

198

S. C.-M. Yang

knowledge norm (‘One must: assert p only if one knows p’), and defends the last one.8 Here, I have no intention to dwell on the issue concerning which norm is the most plausible. If we accept Davidson’s implicit characterization of knowledge in terms of holistically justified true belief, we may stick to the knowledge norm/account of assertion. This would guarantee that whatever the speaker utters must be what she knows. Model-theoretically, this indicates that the semantic treatment of the assertion-operator Ai should meet the knowledge norm of assertion. In addition, we need a second type of act-operator, say ‘Li ’ (referred to as ‘learning-operators’, for each i ∈ G, read ‘the agent i learns that …’), to signify the individual agents’ learning so that what the agent i knows or understands upon the presence of some other’s actions (e.g. the speaker’s assertion) can be explicitly specified by virtue of what the agent i learns. In particular, if a learning-operator Li is operating on certain actions, say β, it indicates that the agent i knows something by learning from the execution of β. It is in this sense that the hearer’s understanding of the propositional content of what is embodied in the speaker’s assertion can be specified by the hearer’s action of learning. The learning-operators for individual agents can also be generalized so as to cover all agents of the whole group. Recall Davidson’s brief remark: ‘Each creature associates the other creature’s responses with stimuli from the shared world’ (Davidson 2001d: 6). Accordingly, we may have an act-operator for group learning, say ‘LG ’. The result of group learning should be something shared by all agents, which is essentially just common knowledge. The remainder of this paper is devoted to the construction of a framework wherein the semantic rules for these two types of act-operators can be stipulated appropriately, and the transition of what the speaker intends to communicate to common knowledge can be specified explicitly in terms of individual agents’ learning and group learning.

5.3 A Semantic Framework for the Underlying System—A Logic of Common Knowledge We have proposed that when a speaker utters a sentence in an ongoing communication, she is making an assertion, which is governed by the knowledge norm. Moreover, the propositional content of what the speaker asserts would become common knowledge. This line of thought indicates that the most promising underlying system for the desired epistemic action logic should be a logic of common knowledge in which the concept of common knowledge is characterized in terms of the knowledge account of assertion. In Yang (2013), I constructed a collection of models for a knowledge-first epistemic logic of knowledge, referred to as TW-models, which can justify the main theses of knowledge-first epistemology that Williamson (2000) proposes. Here, let 8 But see Weiner (2005) for a defence of the truth norm. Also there are some variants of the warrant

norm including Rational Credibility Norm (Douven 2006), Reasonable-to-Believe Norm (Lackey 2007), and Supportive Reasons Norm (McKinnon 2013), to mention a few.

Davidson’s Triangulation Argument in the Logic of Actions

199

us briefly describe these models and the related semantic treatment for knowledge in what follows. Let us start with the language L K , which is a language for a multi-agent epistemic logic of knowledge (for a group G of agents), generated by the following BNF: ϕ::= p |¬ϕ| ϕ → ψ| Ki ϕ (for each i ∈ G)

Here, ‘Ki ϕ’ means ‘the agent i knows ϕ’. Let P be the set of all propositional letters and Form (L K ), the set of all formulas of L K . A TW-model for a multi-agentepistemic logic of knowledge for a group G of agents is a complex: M = S, R, {δ i }i∈G , V P , where

• S, a non-empty set of states; • R ⊆ S × S (the required accessibility relation, to be specified in accordance with the axioms admitted)9 ; • V P : (S → P → {T, F}), an assignment of a truth value to each propositional letter in every world; • δ i : S → ℘(Form(L K )), a function from S to the power set of Form(L K ) such that for any s ∈ S, δ i (s) ⊆ {ϕ | M, s  ϕ, ϕ ∈ Form(LK )}. Intuitively, ‘ϕ ∈ δ i (s)’ read as ‘the agent i is actually in a position to know ϕ in the state s. Clearly, for any propositional letter p M, s  p iff V P assigns a truth value T to p at s.

Semantic rules for negation and material implication are standard. Now, the semantic rule for Ki can be stipulated as follows: (KS ) M, s  Ki ϕ iff ∀t ∈ S(Rst → M, t  ϕ) ∧ ϕ ∈ δ i (s).

The first condition, ∀t ∈ S(Rst → M, t  ϕ), is standard, simply following Hintikka’s proposal that knowledge can be viewed as ‘truth throughout the logical space of possibilities that the agent consider relevant’ (Hintikka 1962), or in Williamson’s words, ‘we know p only if p is true in nearby cases’ (Williamson 2000: 17). The second condition, ϕ ∈ δ i (s), indicates the requirement that to know ϕ, the agent must actually be in a position to know ϕ in the given state. This extra condition is required simply because, according to Williamson, for a given proposition p, p’s being true in all relevant states is not sufficient for an agent i to know p; rather, the agent must be actually in a certain condition (i.e. in a good position) so that when the case obtains, the agent can actually get to know what obtains. More 9 For

simplicity, S5 models are most favourite for an epistemic logic of knowledge. Accordingly, one may prefer to assume R to be an equivalent relation to validate the characteristic formula for S5-models. However, it is noteworthy that an ideal epistemic logic for human agents’ knowledge seems more likely to be stronger than S4 but weaker than S5, the required accessibility relation would meet reflexivity and transitivity but hardly to be equivalence. In particular, when we move to an epistemic action logic, S4 will be much more appealing. Of course, these issues remain open to disputes. And in this paper my primary concern is methodological. Hence I will leave these issue untouched.

200

S. C.-M. Yang

straightforwardly, each function δ i classifies a set of propositions in each state s for each agent i, indicating that the agent i is actually in a position to know ϕ in s, if ϕ ∈ δ i (s). Later, in Yang (2014), I constructed extensions of TW-models for a logic of assertion which validate the knowledge account of assertion. The required language, L KA , is generated by the following BNF: ϕ::= p |¬ϕ| ϕ → ψ| Ki ϕ | Ai ϕ| Aϕ| (for all i ∈ G)

‘Ai ϕ’ means that the agent i asserts ϕ; and ‘Aϕ’ simply means that ϕ is an assertion. As Davidson (2001a: 90) rightly points out, there are no conventions governing the formation of agents’ intentions. So I can only add to TM-models a primitive function, λi , which intends to capture the idea that assertion is a kind of intentional speech act in that if one asserts something, one has the intention to do so. Let Form(L KA ) be the set of all formulas of the language. The required models are essentially extensions of TW models of the form: M = S, R, {δ i }i∈G , {λi }i∈G , V P , where

• S, R and V P are the same as above; • δ i : S → ℘(Form(L KA )), a function from S to the power set of Form(L KA ) such that for any s ∈ S, δ i (s) ⊆ {ϕ | M,s  ϕ, ϕ ∈ Form(LKA )}. Again, ‘ϕ ∈ δ i (s)’ read as ‘the agent i is actually in a position to know ϕ in the state s. • λi : S → ℘(Form(L KA )), a function from S to the power set of Form(LKA ) such that for each s ∈ S, λi (s) ⊆ Form(L KA ), and λi (s) ⊆ δ i (s). Note that a formula ϕ ∈ λi (s) is to mean that the agent i has the intention to asserts ϕ in s. And the condition ‘λi (s) ⊆ δ i (s)’ shows that when the agent i has an intention to assert ϕ, i is actually in a position to know ϕ. Now, we have the following semantic rules for assertion: (Ai S ) M, s  Ai ϕ iff ∀t ∈ S(Rst → M, t  Ki ϕ) ∧ ϕ ∈ λi (s) ∧ Ki ϕ ∈ δ i (s); (AS ) M, s  Aϕ iff for some agent i ∈ G M, s  Ai ϕ.

The extra condition ‘Ki ϕ ∈ δ i (s)’ shows that if one asserts ϕ, then one not only knows ϕ but also actually is in a position to know that one knows ϕ. This would justify the knowledge account of assertion. Now let us move to the required underlying system, an epistemic logic of common knowledge based on the knowledge account of assertion. In Yang (2016), I further construct extensions of the models for a logic of assertion to serve as models for a logic of common knowledge. The required language, L KAC , is generated by the following BNF: ϕ::= p |¬ϕ| ϕ → ψ| Ki ϕ| Ai ϕ| Aϕ|E G ϕ| C G ϕ (for each i ∈ G)

‘E G ϕ’ read as ‘Everyone in G knows ϕ’, and ‘C G ϕ’ means that ‘ϕ is common knowledge to everyone in G’. A model for the underlying system is exactly the same as models for the logic of assertion we described above with additional interpretations for E G and C G , respectively. The semantic rules for ‘E G ’ is also standard:

Davidson’s Triangulation Argument in the Logic of Actions

201

(E G S ) M, s  E G ϕ iff M, s  Ki ϕ for all i ∈ G

There I examined three popular accounts of common knowledge in epistemic logic, including 1. The iterated approach (C I ) C G ϕ:= ϕ ∧ E G ϕ ∧ E G E G ϕ ∧ … ∧ … and so on, ad infinitum. 2. The fixed-point approach (C FP ) C G ϕ ↔ E G (ϕ ∧ C G ϕ). 3. The shared-environment approach (due original to Clark and Marshall 1981): Having common knowledge ϕ, just in case there is a situation s such that M, s  ϕ, M, s  Ka ϕ, also, M, s  Kb ϕ.10 I have shown that the iterated approach is too week and the formulation itself is not a well-formed formula in the language in use, but the fixed-point approach is too commanding and so arguably committed to the fallacy of circular definition; while the shared-environment approach seems not comprehensible enough. Instead I proposed an assertion account of common knowledge based on the knowledge account of assertion: ϕ is common knowledge to a group G of agents if and only if everyone knows that ϕ is true and everyone knows that ϕ is an assertion, in symbols, C G ϕ ↔ E G (ϕ ∧ Aϕ). The semantic rule for C G ϕ can be so stipulated: (C G S ) M, s  C G ϕ iff M, s  E G (ϕ ∧ Aϕ)

This completes the semantics for the required underlying system. We now turn our attention to the semantic framework for the desired epistemic action logic.

5.4 The Semantic Framework for the Desired Epistemic Action Logic Firstly, the language L EA is an extension of L KAC , a poly-modal language in character, which is composed of two sorts of linguistic expressions, namely a set of (descriptive) formulas and a set of descriptions for the specified actions, generated by the following BNF: ϕ::= p | ¬ϕ | ϕ → ψ | Ki ϕ | Aϕ | E G ϕ | C G ϕ | [α]ψ (p ∈ P; i ∈ G) α::= ?ϕ| Ai α | Li α | LG α]

Here, ‘α’ is a modal construct standing for an action-modality (i.e. a linguistic expression containing some act-operators and possibly some descriptive formulas) to signify the operation of some action. And ‘[α]ψ’ indicates that after every successful 10 For

more details, see Barwise (1988), Clark and Marshall (1981).

202

S. C.-M. Yang

completion of an action α, ψ holds in all subsequent states (or states accessible from the initial state). Note that α can be a basic modal construct, composed of a basic act-operator and a formula/statement (such as ‘?ϕ’, showing that the act-operator ‘?’ operates on descriptive formulas). More often, α can be a compound modal construct, composed of more than one act-operators, such as ‘Li β’ (here β is a certain modal construct with some other act-operators for some actions), in which the act-operator ‘Li ’ operates on the results of the executions of β. For example, we may take ‘Li ’ as the well-known learning-operator, the modal construct ‘Li β’ will signify what the agent i learns from the execution of the action β. It is noteworthy that both of the language L KAC and the language L EA include a set of epistemic-operators Ai ’s (for each agent i ∈ G), respectively, each of which behaves just like Ki . But, in the former, ‘Ai ϕ’ is a descriptive formula, showing that ϕ is an assertion, a formula asserted by the agent i; whereas, in L EA , Ai is taken as an act-operator, operating on the results of some other actions just as in the compound act-operator ‘[Ai ?ϕ]’. Recall that in the logic of assertion, we do have M, s  Aϕ iff for some agent i ∈ G M, s  Ai ϕ. If we at the same time keep ‘Ai ϕ’ as a descriptive formula in L EA , we would have M, s  [Ai ?ϕ]Ai ϕ automatically (a basic property of models for the logic of actions), which implies M, s  [Ai ?ϕ]Aϕ trivially. To avoid confusion, the descriptive formulas of the form ‘Ai ϕ’ (for any i ∈ G) will be dispensed in L EA , and only ‘Aϕ’ is retained. This would not weaken the expressive power of the language in use. Let us now turn our attention to the required models. A model for the desired logic of action is exactly the same as those for the underlying system we decribed above (see p. 200) with additional appropriate interpretation for act-operators. The semantic rules for atomic propositional letters and truth-functors are standard; semantic rules for the epistemic operators are the same as stipulated in the underlying system as well. The basic semantic rule for an act-operator α is also standard as follows. ([α]S )

M, s  [α]ϕ iff for all M , and for all s , (M, s) [|α|] (M , s ) implies M , s  ϕ.

Note that the model-theoretical notation ‘M, s  [α]ϕ’ reads: In a model M, after the execution of the action α in a state s, ϕ holds in all subsequent states, that is, ϕ is true in all states accessible from s. Also, (M, s) [|α|] (M , s ) intends to specify all possible new states resulting from the change of agents’ epistemic states due to the execution of α in s. Here (M, s) stands for the initial state of the performance of α. When α is performed, we may take α as the kind of modal operator which puts some constraints on the accessible states. And so, M means any model resulting from M with the imposed constraints, and s means any state accessible from s in the resulting models. Clearly, (M , s ) is a possible state rendered by the execution of α. More specifically, M = S, R , δ i  (i ∈ G) , V  , where S  = [|ϕ|]M = {s ∈ S | M, s ϕ}; R = R ∩ ([|ϕ|]M × [|ϕ|]M ), δi



λi



∈G

= δ i [|ϕ|]M ,

∈G

= λi [|ϕ|]M ,

V  p∈P = V p ∩ [|ϕ|]M .

Davidson’s Triangulation Argument in the Logic of Actions

203

The act-operator ‘?’ (read as ‘test’) is a basic act-operator in epistemic action logic, operating directly on all formulas so that the modal construct ‘?ϕ’ shows that ϕ is tested as true. When a formula is tested as true at a state s in a model M, we need only consider the states of the models where in ϕ holds. Thus, the standard semantic rule for ‘?ϕ’ is as follows: ([?ϕ]S ) (M, s) [|?ϕ|] (M , s ) iff M = [|ϕ|]M , ∅, V P | [|ϕ|]M and s = s .

Accordingly, M, s  [?ϕ]ϕ

holds trivially. The action of test plays a significant role in epistemic action logic in that being rational animals, the agents aim at truth. In particular, to make communication successful, whatever the speaker asserts must be truths. This is precisely what the action ‘[Ai ?ϕ]’ intended to capture. Now let us turn our attention to the semantic rule for the act-operator Ai for asserting. Based on the knowledge account of assertion, when the speaker i asserts ϕ, i is actually in a position to know ϕ in all accessible states s ; hence ϕ ∈ δ i (s ) for all s in all M . And of course, the speaker’s action of asserting ϕ shows clearly that Aϕ holds not only in s (the facticity of asserting) but also in all accessible (subsequent) states s (action-preserving). Hence, it suffices to propose the semantic rule for the characteristic application of ‘Ai ’ to render assertion-formulas of the form Aϕ. This can be stipulated as follows: ([Ai ]S ) M, s  [Ai β]Aϕ iff there are some θ i ’s such that for each θ, M, s  [β]θ, and for all M , and for all s , (M, s) [|Ai β|] (M , s ) implies that M , s  K i θ, and the conjunction of some Ki θ  s implies Ki ϕ, also, ϕ ∈ λi (s ) (hence, M , s  Aϕ).

We can now focus on the semantic rules for the learning-operator Li and LG . In epistemic action logic, the act-operator Li for learning usually occurs in compound modalities of the form ‘Li β’, where β is an action, showing what the agent i has learnt from the execution of β. The group learning LG . (i.e. M, s  [LG β]ϕ) shows what all agents in G learn in common, directly or indirectly, from the execution of β. Recall that to show that M, s  [Li β]ϕ, we have to show that for all M , and for all s , (M, s) [|Li β|] (M , s ) implies M , s  ϕ. Also, the learning-operator Li intends to specify what the agent i learns from the execution of the action β. And this would show that, given that for any θ such that M, s  [β]θ, that is, for all M , and for all s ∈ S  , (M, s) [|β|] (M , s ) implies M , s  θ, (M, s) [|Li β|] (M , s ) implies that M , s  Ki θ: To say that an agent learns what holds from an action β in a state is tantamount to saying that the agent knows what holds in all subsequent states after the execution of β.

204

S. C.-M. Yang

Accordingly, we may focus on characterisitc applications of a learning operator to render formulas of the form K i ϕ to show what the given agent learns and set as the required semantic rule for Li : ([Li ]S* ) M, s  [Li β] Ki ϕ iff there are some θ i ’s such that for each θ, M, s  [β]θ, and for all M , for all s , (M, s) [|Li β|] (M , s ) implies that M , s  Ki θ, and the conjunction of some Ki θ implies ϕ also, ϕ ∈ δ i (s )

(hence, M , s  Ki ϕ).

In particular, if M, s  [β]θ, then M, s  [Li β]Ki θ. It is noteworthy that an indexed learning-operator, say Li , may operate on a certain action which is signified by some other indexed operator, say A j β. Clearly, if i = j, Li Ai β would indicate that the agent i learns something from his own action Ai β. This suggests that the agent’s self-learning directly from the execution of her own action is possible, and for the agent, the previous action is an action of epistemic transparency, or luminosity in Williamson’s term. However, for those who reject direct self-learning, a more complicated semantic treatment is called for: ([Li ]S ) Let β be an act-operator, say of the form β n … β 1 , prefixed with an indexed actoperator β n . When i = n, ([Li ]S* ) holds; otherwise (i.e. i = n), M, s  [Li β]ϕ iff M, s  [β]ϕ.

We can further generalize the action of individual agents’ learning to the group learning, LG , which intuitively says that every agent in the group has leant, directly or indirectly, from a certain action which has been performed already. Thus, we may have a semantic rule for group learning LG as follows: M, s  [LG β]ϕ

iff for every agent i ∈ G, there is a certain (possibly empty) finite sequence L1 , …, Ln such that M, s  [Li Ln … L1 β]ϕ holds.

This completes the construction of the proposed semantic framework for an epistemic action logic suitable for a formal machinery to complement the epistemic background in Davidson’s triangulation argument. Now, let us show how to apply this framework to the triangulation argument, in particular, how to formulate the transition of the proposition content of what the speaker asserts to common knowledge.

6 From the Speaker’s Assertion to Common Knowledge Since the triangulation argument is primarily concerned with only two rational agents, i.e. the speaker (say, a) and the hearer (say, b), we may assume G = {a, b}. Also, only the speaker a asserts something, we may, for the time being, consider only the act-operator Aa .

Davidson’s Triangulation Argument in the Logic of Actions

205

The idea then is to formulate explicitly that after the speaker a asserts ϕ in a state s, what would result in all possible consequent states (i.e. all states s accessible from s). Clearly, the knowledge account of assertion implies that after the speaker asserts ϕ, in all subsequent states the speaker knows ϕ. That is, (M, s) [|Aa ?ϕ|] (M , s ) implies M , s  Ka ϕ. Also, both ϕ, due to the facticity of knowledge, and Aϕ hold in all subsequent states. One can see clearly that the semantic rule we put forth shows that for all M and for all s , (M, s) [|Aa ?ϕ|] (M , s ) implies 1. M , s  Ka ϕ. 2. M , s  ϕ 3. M , s  Aϕ (Action-description). Hence, 4. M, s  [Aa ?ϕ]Ka ϕ (the knowledge account of assertion) 5. M, s  [Aa ?ϕ]ϕ (the facticity of knowledge) 6. M, s  [Aa ?ϕ]Aϕ (basic property of the act-operator ‘asserting’). At this stage, it is tempting to claim that M, s  [Aa ?ϕ] Ka Aϕ as well, given that for all M , for all s , M , s  Aϕ. It seems then straightforwardly to maintain that for all M , for all s , M , s  Ka Aϕ; hence M, s  [Aa ?ϕ] Ka Aϕ. However, it is questionable whether assertion is transparent, or luminous. Adhering to the luminosity of assertion, the speaker a would know that she is making an assertion when asserting ϕ. Hence, the speaker must be actually in a position to know that she is making an assertion, hence Aϕ ∈ δ a (s ) for all s in all M . Thus, M, s  Ka Aϕ holds automatically. By contrast, some might reject the luminosity of assertion. It can be argued that the speaker might not be actually in a good position to know that she is asserting ϕ when she actually asserts ϕ. Clearly, if the luminosity of assertion is assumed, M, s  [Aa ?ϕ] Ka Aϕ holds. Otherwise, M, s  [Aa ?ϕ] Ka Aϕ may not hold although M, s  [Aa ?ϕ]Aϕ is a truism. As a matter of fact, in this case the speaker knows that she is asserting ϕ only when she has learnt from the hearer’s response to her assertion. Given that M, s  [Aa ?ϕ]Aϕ holds, we want to show that M, s  [LG Aa ?ϕ]C G ϕ, which states that both the speaker a and the hearer b learn that ϕ is common knowledge, when the speaker a asserts ϕ. It suffices to show that M, s  [Lb Ln … L1 Aa ?ϕ]C G ϕ, assuming that each j (1 ≤ j ≤ n) can only be a or b, and M, s  [La Lm … L1 Aa ?ϕ] C G ϕ, assuming that each l (1 ≤ l ≤ m) can only be a or b. Let us start with the hearer’s action [Lb Aa ?ϕ], which indicates what the hearer b learns from the speaker’s assertion of ϕ. And our goal is to show M, s  [Lb Aa ?ϕ] C G ϕ, given the assertion account of common knowledge, namely (CKA) C G ϕ ↔ E G (ϕ∧Aϕ). That is, we have to show M, s  [Lb Aa ?ϕ] E G (ϕ∧Aϕ).

It suffices to show in turn that for all M , for all s , M , s  E G (ϕ∧Aϕ), i.e. for all M , for all s , M , s Ka (ϕ∧Aϕ) ∧ Kb (ϕ∧Aϕ). Now, given M, s  [Aa ?ϕ] Aϕ, we have shown that for all M and for all s , (M, s) [|Aa ?ϕ|] (M , s ) implies,

206

S. C.-M. Yang

7. for all M , for all s , M , s  (ϕ∧Aϕ). By ([Li ]S ), we do have for all M , for all s , (M, s) [|Lb Aa ?ϕ|] (M , s ) implies 8. for all M , for all s , M , s  Kb (ϕ∧Aϕ); Still, we have to show that 9. for all M , for all s , M , s  Ka (ϕ∧Aϕ). We have shown M , s  Ka ϕ; but we do not have M , s  Ka Aϕ because of the rejection of the luminosity of assertion. So neither Ka Aϕ would hold automatically, nor Ka Aϕ could follow from what the hearer has learnt. So we are in no position to claim that M , s  [Lb Aa ?ϕ] Ka Aϕ. Nor could we claim that M, s  [Lb Aa ?ϕ] E G (ϕ∧Aϕ). However, the hearer may learn this indirectly via some of the speaker’s action, in particular, the speaker’s learning from the the hearer’s response to his assertion. So let us focus on the action [La Lb Aa ?ϕ]. The established semantic rules show clearly that 10. M, s  [La Lb Aa ?ϕ] Ka (ϕ∧Aϕ), as M, s  [Lb Aa ?ϕ](ϕ∧Aϕ) (by (8) and [Li ]S ) 11. M, s  [La Lb Aa ?ϕ]Kb (ϕ∧Aϕ), as M, s  [Lb Aa ?ϕ] Kb (ϕ∧Aϕ) (by (7)) Hence, 12. M, s  [La Lb Aa ?ϕ] E G (ϕ∧Aϕ) (by (10) and (11)) That is, 13. M, s  [La Lb Aa ?ϕ] C G ϕ. Still, we have shown that the hearer cannot learn that C G ϕ holds directly from the speaker’s assertion that ϕ holds. However, the construction of compound actionmodalities suggests that the hearer may learn something more via the speaker’s response to her (the hearer’s) response to what the speaker asserts. So let us consider the act-modality Lb La Lb Aa ?ϕ. Clearly, the established semantic rules would tell us that the followings hold as well: 14. M, s  [Lb La Lb Aa ?ϕ]Ka (ϕ∧Aϕ) (by (10)) Also, 15. M, s  [Lb La Lb Aa ?ϕ]Kb (ϕ∧Ka ϕ) (by (11)) Hence, 16. M, s  [Lb La Lb Aa ?ϕ]E G (ϕ∧Aϕ) (by (14) and (15)) It follows 17. M, s  [Lb La Lb Aa ?ϕ] C G ϕ. Together with (13),

Davidson’s Triangulation Argument in the Logic of Actions

207

18. For all M , and for all s , (M, s) [|LG Aa ?ϕ|] (M , s ) implies M , s  C G ϕ. That is, 19. M, s  [LG Aa ?ϕ] C G ϕ. This completes our formulation of the transition of the speaker’s assertion to common knowledge via the individual agents’ learning and group learning in the proposed semantic framework for an epistemic action logic. Owing to the limit of space, I could not present a formalism of the desired logic of actions. However, I believe that the proposed semantic framework, taken as a formal machinery, suffices to complement the epistemic background for the triangulation argument. Methodologically, this approach still falls within the main theme of Davidson’s philosophy of action which elaborates ‘the role of causal concepts in the description and explanation of human action’. Acknowledgements Previous versions of this paper were presented at Philosophy Programme Seminars (Victoria University of Wellington), Current Projects Seminars (The University of Sydney), the Third Taiwan Metaphysics Colloquium (National Taiwan University), and the Workshop ‘From Computation to Agency’ (Tsinghua University). I thank all organizers of these events and participants for their valuable comments. Special thanks to Fenrong Liu, Johan van Benthem, Pascal Engel, Nicholas J. J. Smith, Ed Mare, Kok Yong Lee, and Run-June Wang.

References Austin, J. L. (1963). Performative—Constative. In C. E. Caton (Ed.), Philosophy and ordinary language (pp. 22–33). Urbana: University of Illinois Press. Originally written in French and presented at a conference held at Royaumont in March 1958. English translation by G. J. Warnock. Austin, J. L. (1975). How to do things with words (2nd ed.) (1st ed., 1962). Oxford: Oxford University Press. Barwise, J. (1988). Three views of common knowledge. In TARK’88: Proceedings of the 2nd Conference on Theoretical Aspects of Reasoning About Knowledge (pp. 365–379). San Francisco, CA: Morgan Kaufmann Publishers. Clark, H. H., & Marshall, C. R. (1981). Definite reference and mutual knowledge. In A. K. Joshi, B. L. Webber, & I. A. Sag (Eds.), Elements of discourse understanding (pp. 10–63). Cambridge: Cambridge University Press. Davidson, D. (1963 [2001a]). Action, reasons, and causes. The Journal of Philosophy, 60(23), 685–700. Reprinted in Davidson 2001a: 3–20. Davidson, D. (1967 [2001b]). Truth and meaning. Synthese, 17(1), 304–323. Reprinted in Davidson 2001b: 17–36. Davidson, D. (1980 [2004]). A unified theory of thought, meaning, and action. Originally entitled as ‘Toward a unified theory of meaning and action’. Grazer Philosophische Studien, 11, 1–12. Reprinted in Davidson 2004: 151–166. Davidson, D. (1982 [2001c]). Rational animals. Dialectica, 36(4), 317–327. Reprinted in Davidson 2001c: 95–105. Davidson, D. (1986 [2005a]). A nice derangement of epitaphs. In E. LePore (Ed.), Truth and interpretation: Perspectives on the philosophy of Donald Davidson (pp. 433–446). Oxford: Basil Blackwell. Reprinted in Davidson 2005a: 89–107.

208

S. C.-M. Yang

Davidson, D. (1991 [2001c]). Three varieties of knowledge. In A. P. Griffiths (Ed.), Royal Institute of Philosophy supplement (Vol. 30, pp. 153–166). Cambridge: Cambridge University Press. Reprinted in Davidson 2001c: 205–220. Davidson, D. (1992 [2001c]). The second person. Midwest Studies in Philosophy, 17(1), 255–267. Reprinted in Davidson 2001c: 107–121. Davidson, D. (1994 [2005]). The social aspect of language. In B. McGuiness & G. Oliveri (Eds.), The philosophy of Michael Dummett (pp. 1–16). Amsterdam: Kluwer Academic Publishers. Reprinted in Davidson 2005a: 109–125. Davidson, D. (2001a). Essays on actions and events (1st ed., 1980). Oxford: Clarendon Press. Davidson, D. (2001b). Inquiries into truth and interpretation (1st ed., 1984). Oxford: Clarendon Press. Davidson, D. (2001c). Subjective, intersubjective, objective. Oxford: Clarendon Press. Davidson, D. (2001d). Externalisms. In P. Kotatko, P. Pagin, & G. Segal (Eds.), Interpreting Davidson (pp. 1–26). Stanford, CA: CSLI Publications. Davidson, D. (2004). Problems of rationality. Oxford: Clarendon Press. Davidson, D. (2005a). Truth, language, and history. Oxford: Oxford University Press. Davidson, D. (2005b). Truth and predication. Cambridge, MA: Harvard University Press. Douven, I. (2006). Assertion, knowledge, and rational credibility. Philosophical Review, 115, 449– 485. Frege, G. (1879). Begriffsschrift, eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. Halle a/S: Verlag von Louis Nebert. A new edition of English translation entitled: Conceptual notation and related articles. Edited and translated, with a biography and introduction, by T. W. Bynum (1st ed., 1972). Oxford: Clarendon Press. 2000. Frege, G. (1918–19). The thought: A logical inquiry. Mind (New Series, 1956), 65(259), 289–311. Grice, H. P. (1989). Studies in the way of words. Cambridge, MA: Harvard University Press. Hintikka, J. (1962). Knowledge and belief—An introduction to the logic of the two notions. Ithaca: Cornell University Press. Reissued version compiled by V. F. Hendricks & J. Symons. London: College Publications. 2005. Lackey, J. (2007). Norms of assertion. Noûs, 41, 594–626. McKinnon, R. (2013). The supportive reasons norm of assertion. American Philosophical Quarterly, 50, 121–135. Myers, R. H., & Verheggen, C. (2016). Donald Davidson’s triangulation argument. London: Routledge. Pappas, G., & Swain, M. (Eds.). (1978). Essays on knowledge and justification. Ithaca, IL: Cornell University Press. Quine, W. V. (1992). Pursuit of truth (Revised ed). (1st ed., 1990). Cambridge, MA: Harvard University Press. Quine, W. V. (1960 [2015]). Word and object (2nd ed.) (1st ed., 1960). Cambridge, MA: The MIT Press. Tarski, A. (1933 [1983]). The concept of truth in the languages of the deductive sciences. Original version in Polish, in Prace Towarzystwa Naukowego Warszawskiego, Wydzial III Nauk Matematyczno-Fizycznych (Vol. 34). Warsaw. Expanded English translation in Tarski’s Logic, semantics, metamathematics, papers from 1923 to 1938 (J. Corcoran, Ed.) (pp. 152–278). Indianapolis: Hackett. Tarski, A. (1944 [1986]). The semantic conception of truth: And the foundations of semantics. Philosophy and Phenomenological Research, 4(3), 341–376. Reprinted in S. R. Givant & R. N. McKenzie (Eds.), Alfred Tarski—Collected papers, Vol. 2: 1935-44. (pp. 661–699). Basel: Birkhäuser Verlag. van Benthem, J. (2010). Modal logic for open minds. Stanford, CA: CSLI Publications. van Benthem, J., van Ditmarsch, H., van Eijck, J., & Jaspars, J. (2016). Logic in action. Internet. 2016. New edition. November 23. 2016. Electronic book. Available from http://www.logicinac tion.org.

Davidson’s Triangulation Argument in the Logic of Actions

209

van Ditmarsch, H., van der Hoek, W., & Kooi, B. (2008). Dynamic epistemic logic. Dordrecht, The Netherlands: Springer. Weiner, M. (2005). Must we know what we say? Philosophical Review, 114, 227–251. Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press. Wittgenstein, L. (1953 [2009]). Philosophical investigations (4th ed.) (P. M. S. Hacker & J. Schulte, Eds.) (1st ed., 1953). Oxford: Wiley-Blackwell. Yang, S. C.-M. (2013). TW-models for logic of knowledge-cum-belief. In J. Brendle, R. Downey, R. Goldblatt, & B. Kim (Eds.), Proceedings of the 12th Asian Logic Conference (pp. 314–337). Singapore: World Scientific Publishing Co. Yang, S. C.-M. (2014). A defence of the knowledge norm of assertion: From a model-theoretical perspective. In T.-W. Hung (Ed.), Communicative Action: Selected Papers of the 2013 IEAS Conference on Language and Action (pp. 33–50). Berlin: Springer. Yang, S. C.-M. (2016). Common knowledge and the knowledge account of assertion. In S. C.-M. Yang, D.-M. Deng, & H. Lin (Eds.), Structural analysis of non-classical logics (pp. 253–278). Berlin: Springer.