Moral psychology and human agency: philosophical essays on the science of ethics 9780198717812, 0198717814, 9780191027642, 0191027642, 9780191787324, 0191787329

Citation preview

Moral Psychology and Human Agency

Moral Psychology and Human Agency Philosophical Essays on the Science of Ethics EDITED BY

Justin D’Arms and Daniel Jacobson

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © The several contributors 2014 The moral rights of the authors have been asserted First Edition published in 2014 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2014938931 ISBN 978–0–19–871781–2 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Acknowledgments The editors would like to thank audiences at the University of Michigan Workshop on Moral Psychology and Human Agency for their helpful comments on many of the chapters in this volume, and Peter Momtchiloff of Oxford University Press for his help in bringing the volume to fruition. This volume and the workshop were supported by a grant from the John Templeton Foundation.

Contents Editors and Contributors

ix

1. Introduction Justin D’Arms and Daniel Jacobson

1

2. Intuitive and Counterintuitive Morality Guy Kahane

9

3. Moral Psychology as Accountability Brendan Dill and Stephen Darwall

40

4. Remnants of Character David Shoemaker

84

5. Knowing What We Are Doing Heidi Maibom

108

6. Meta-Cognition, Mind-Reading, and Humean Moral Agency Julia Driver

123

7. The Episodic Sense of Self Shaun Nichols

137

8. The Motivational Theory of Emotions Andrea Scarantino

156

9. The Reward Theory of Desire in Moral Psychology Timothy Schroeder and Nomy Arpaly

186

10. Does Evolutionary Psychology Show That Normativity Is Mind-Dependent? Selim Berker

215

11. Sentimentalism and Scientism Justin D’Arms and Daniel Jacobson

253

Index

279

Editors and Contributors Editors Justin D’Arms is Professor of Philosophy at the Ohio State University. Daniel Jacobson is Professor of Philosophy at the University of Michigan.

Contributors Nomy Arpaly is Associate Professor of Philosophy at Brown University. Selim Berker is John L.  Loeb Associate Professor of the Humanities at Harvard University. Stephen Darwall is Orrick Professor of Philosophy at Yale University. Brendan Dill is a graduate student in the Department of Linguistics and Philosophy at the Massachusetts Institute of Technology. Julia Driver is Professor of Philosophy at Washington University in St Louis. Guy Kahane is Deputy Director and Research Fellow, Oxford Uehiro Centre for Practical Ethics, Faculty of Philosophy, University of Oxford. Heidi Maibom is Professor of Philosophy and Cognitive Science at University of Cincinnati. Shaun Nichols is Professor of Philosophy at University of Arizona. Andrea Scarantino is Associate Professor in the Department of Philosophy and Neuroscience Institute at Georgia State University. Timothy Schroeder is Professor of Philosophy at Ohio State University. David Shoemaker is Associate Professor in the Department of Philosophy and the Murphy Center at Tulane University.

1 Introduction Justin D’Arms and Daniel Jacobson

There is a philosophical tradition of treating moral psychology as, if not exactly an a priori discipline, at any rate one to be pursued at some remove from empirical fields. In part this is due to some influential but contentious ideas that have shaped the discussion. One is the view that moral demands apply not specifically to humans but to all rational agents capable of recognizing and being motivated by these demands. If so then contingent facts about human psychology seem irrelevant to the most fundamental aspects of morality (though they may play a secondary role in determining how its demands apply specifically to humans). Another is the idea that rationalizing explanations, which seek to understand certain human behavior as action done for reasons, should appeal exclusively to capacities or states—such as belief and desire, understood very abstractly—that are constitutive of rational agency as such. Thus, despite some notable exceptions, much of the philosophy of action abstracts away from contingencies of human cognitive and motivational systems. In recent decades, however, there has been significant growth in the scientific study of ethics from disparate fields, including psychology, biology, linguistics, behavioral economics, and neuroscience. This trend coincides with increasing philosophical enthusiasm for empirical approaches to issues in moral psychology and human agency, to the extent that a split has occurred between those who see the field of moral psychology as a branch of psychology and those who see it as an autonomous branch of moral philosophy. We find ourselves at a distance from both these approaches, doubting whether either can be adequate. Although we see promise in the scientific investigation of ethics, we also think some scientists and empirically minded philosophers overreach in their aspirations. Some of those who claim moral psychology as an empirical discipline derive hasty conclusions about values, reasons, and moral requirements by assuming undefended and implausible normative premises. Others seek to undermine all rationalizing explanations of human behavior and evaluative judgment, on grounds that we find inadequate. While in our view there is no doubt that the details of human thought and motivation can help answer classic philosophical questions, these answers will not be given by scientific inquiry alone. Just how much philosophical progress the new science of ethics can

2  Justin D’Arms and Daniel Jacobson provide, and what are its limitations, remains to be seen. Although empirical study of moral psychology will continue, as it should, philosophers can bring distinctive intellectual tools to bear on the discussion; and we think they can do much to clarify these issues and explore their consequences. This volume attempts just that. The contributors to this volume are philosophers who share our conviction that scientific inquiry is relevant to various classic philosophical questions, while also showing an appreciation for the difficulty of these questions that is not always evident in empirical moral psychology. Beyond those similarities and their shared commitment to analytical rigor, the chapters are diverse in their approach to the philosophical implications of the science of ethics. Two authors (Kahane and Nichols) report here on their own empirical findings. The first (Kahane) responds to one of the most influential but problematic claims in empirical ethics, according to which only consequentialist thinking can be vindicated as rational. The second (Nichols) raises novel questions about the role of memory in two distinct and influential conceptions of the self. The details about human memory, specifically about its deficits in the impairments characteristic of dementia, are claimed (by Shoemaker) to help explain some philosophical puzzles about moral responsibility at the margins. Other authors similarly use recent empirical findings to enrich their thinking about classic philosophical problems. One (Maibom) argues that the question of how much one needs to know about what one is doing, in order to be responsible for an act, is sharpened by attention to studies of situational influences on action. Another (Driver) argues that the capacity for effective moral agency in humans relies on contingent features of our emotional repertoire. Two other papers develop accounts of human motivation that rely on structures and processes that would not be found in every possible rational agent. The first (Schroeder and Arpaly) offers a theory of desire grounded in the neuroscience of the reward system, which is used to answer certain familiar objections to desire-based theories of reasons for action. The second (Scarantino) argues that the emotions are a distinctive species of motivational state quite different from the desires appealed to in standard cases of rational explanation. Finally, several authors resist specific anti-realist or skeptical arguments grounded in evolutionary theory and other empirical claims. It is argued (by Berker) that an influential evolutionary argument against moral realism does not succeed as advertised, and (by D’Arms and Jacobson) that the empirical ethics movement has overstated its case for pessimism about the role of reasons and reasoning in human moral judgment. It is further suggested (by Dill and Darwall) that empirical data support a philosophically prominent account of morality and moral concepts as against the skeptical alternatives more commonly advocated in empirical ethics. Taken together, we think these essays provide an encouraging picture of how careful philosophical engagement with the science of ethics can advance discussion of important questions about morality and agency. What follows are some brief notes on the individual chapters.

Introduction  3 In “Intuitive and Counterintuitive Morality,” Guy Kahane critically examines Joshua Greene’s influential application of the dual-process model of moral thinking. Greene has argued that there are two distinct forms of moral thinking: a fast, intuitive form that is based on alarm-like emotional responses and pushes people to deontological verdicts; and a slower, deliberative form involving other cognitive capacities which leads people to utilitarian conclusions. Thus subjects who (slowly) choose to divert a trolley to save five people at the cost of one are claimed to be reaching a utilitarian conclusion thoughtfully, whereas those who (quickly) refuse to push a man off a bridge in order to achieve the same result reach a deontological conclusion emotionally. Greene takes this to be bad news for deontological approaches to ethics, partly on the assumption that more thinking is better thinking. But Kahane points out that decisions with “utilitarian” content do not necessarily result from utilitarian thinking. He offers an alternative hypothesis to explain why it takes more time to decide to switch the trolley than to decide not to push the man. Since it is more difficult to reach counterintuitive judgments, it takes longer to do so, regardless of their content. Kahane reports on experiments showing that counterintuitive judgments reaching disparate conclusions are based in similar neural processes. He then suggests a compelling explanation of the phenomena that turns Greene’s account on its head. The reason that subjects are slower in reaching the utilitarian decision to divert the trolley is not that they are taking a long time doing the (trivial) utilitarian arithmetic, but that they are weighing disparate moral considerations: in favor of saving innocent lives and against harming innocents. The crucial distinction is not between utilitarian and deontological judgments but between intuitive and counterintuitive ones. Brendan Dill and Stephen Darwall argue, in their collaborative paper, “Moral Psychology as Accountability,” that a large body of empirical evidence supports the view that central moral attitudes, emotions, and motivations are best understood as devices used to hold people (including ourselves) accountable. Thus moral condemnation functions to get putative wrongdoers to hold themselves accountable for their wrongdoing, and moral conscience functions to hold oneself accountable. The authors contrast these accounts with alternative empirical hypotheses about these central moral attitudes that have a more debunking slant, such as the claim that condemnation masks a brute desire to make perpetrators suffer, and that conscience is the egoistic desire to appear moral to others. Like Kahane, they suggest that a closer look at the empirical evidence reveals facts that are friendlier to morality than the skeptical alternatives more typically defended by proponents of the empirical approach to ethics. In addition, this chapter suggests an answer to an outstanding question in moral philosophy and empirical ethics: how to circumscribe morality. Dill and Darwall suggest that what is distinctive to morality is its essential connection to accountability and the motivations that are partly constitutive of it. Their chapter also sets a helpful conceptual backdrop for Shoemaker, who uses the notion of accountability to make sense of some puzzling moral phenomena.

4  Justin D’Arms and Daniel Jacobson Although moral philosophers have made much of psychopathy, a more common but less considered pathology with moral implications has been largely overlooked. In his finely observed paper, “Remnants of Character,” David Shoemaker considers the distinctive issues about moral responsibility and personal identity posed by dementia. This chapter explores questions about responsibility in patients with moderate-stage dementia. A puzzle about this disease is that it often exempts its sufferers from reactive attitudes such as resentment while leaving them capable, it seems, of particular acts of kindness or cruelty. Shoemaker argues that the crucial point about dementia, which explains how and when it excuses, lies in its impairments to memory. The specific memory impairments dementia patients suffer explain why the patients are still sometimes attributable—in that they can be said to act kindly or cruelly—even though they are not accountable for their actions. A central puzzle is how to make sense of the mean actions of a demented person, on the assumption that this meanness is not a new trait issuing from the dementia (which would undermine even attributability), but reveals something that was already present in them though somehow previously overmastered. Shoemaker’s answer is that these actions flow from clusters of commitments once unified in character traits, aspects of which have been ravaged by the dementia. His explanation emerges from the insight that certain traits might be the product of dispositions that persist in the absence of other, tempering dispositions that no longer function. In her contribution to this volume, “Knowing What We Are Doing,” Heidi Maibom investigates the question of when we can hold people responsible for doing things that they do not realize they are doing, and why it can be legitimate to do so. The cases she is interested in are not run-of-the-mill negligence cases, where the agent could grant retrospectively that did something he should not have done (or failed to do something he should). Rather, she focuses on cases where the agent can be expected sincerely to deny that he acted in the way described—typically because situational factors affected the circumstances of his choice—even though in some sense he should realize that he did so act. Maibom considers three plausible candidates for when one is responsible under such circumstances: when the agent would perform the action if he conceived it under the relevant description; when the act is in character; and when the agent should have known that his action is of the relevant kind. She makes a compelling argument that neither of the first two candidates is tenable but the third one is, at least to some degree. As she notes, the trick is fleshing out just what one should have known. She makes progress on this difficult issue by explicating the degree to which moral learning involves the ability to conceive of alternative perspectives on one’s action. The demand for self-reflection turns out to be a crucial component to understanding what we are responsible for doing, even though the degree and limits of that demand cannot be neatly circumscribed. Julia Driver’s chapter, “Meta-Cognition, Mind-Reading, and Humean Moral Agency,” distinguishes moral agency, which underlies distinctively moral action,

Introduction  5 from mere agency arising from beliefs, desires, and intentions. She interprets Hume as drawing this distinction too, and as holding that moral agency requires self-regulation through attitudes of self-appraisal. She thinks Hume’s claim is too strong, and that attitudes of self-appraisal are not necessary conditions on moral agency as such. However, the capacity for such attitudes, which she terms meta-cognition, is crucial for effective moral agency, given other contingent features of human psychology. To be appropriately held accountable by criticism, one needs to have access to the mental states that lie behind one’s behavior in virtue of which one is being criticized. And one needs to be able to regulate those states through higher-order attitudes about them, in order for blame and criticism to do any good. Driver explores recent studies of various affective disorders that contribute to failure of this sort of self-regulation. She argues that some of them diminish the effectiveness of moral agency in humans without obliterating the capacity for such agency. In his chapter, “The Episodic Sense of Self,” Shaun Nichols suggests that ordinary thought displays two distinct conceptions of the self. One is a thick conception, based on the collection of psychological traits that constitute a person’s distinctive personality. The other, thin conception is of the self simply as an enduring subject of experience. This thin conception is delivered to us by the very nature of episodic memory: the form of memory by which we recall events as having happened to us (as opposed to semantic memory, by which we remember facts). The thin, episodic sense of self is independent of the trait conception, in that even great changes in traits do not disrupt your sense of being the same person who experienced the events you remember episodically. Nichols contends that the thin conception of self is central to certain self-conscious emotions. Guilt is triggered by memories of what one has done, for instance, in ways that can be insensitive to whether you are now dissimilar to the earlier person who did them. He finds a tension between ordinary thought about moral responsibility and the role of such episodic memory in generating guilt. Philosophers and ordinary people hold that trait constancy is important to responsibility, Nichols argues, because people generally hold that if someone commits a transgression in youth and then undergoes massive personality changes (of the right sort), her culpability for the action is thereby diminished. Thus guilt is less fitting. Yet the thin sense of self generates guilt in these cases all the same, because people identify with their previous actions through episodic memory in ways that are insensitive to changes in their traits. Nichols argues that while such responses may be unfitting, there are good prudential reasons for them. Guilt serves its function of repairing relationships by motivating people to make amends and acknowledge transgressions—even for actions done by former selves who are quite radically trait-discontinuous. Andrea Scarantino pursues the nature of emotions more generally in “The Motivational Theory of Emotions.” Two influential philosophical approaches are the cognitive and the perceptual theories of emotion. Scarantino argues that neither of them does a good job explaining the central aspect of the psychological role of emotions: namely, the distinctive way in which they motivate behavior. Taking up and

6  Justin D’Arms and Daniel Jacobson expanding on some ideas from the psychologist Nico Frijda, Scarantino emphasizes respects in which emotional motivations involve forms of control precedence that are quite different from standing desires. Emotional motivation clamors for execution, it limits the information one considers in thinking about how to execute it, and it restricts attention in ways that inhibit practical reasoning about how to fit the action within one’s other goals. While not all emotional behavior exhibits these features, any account of emotions must explain these commonplace observations. Scarantino argues that the accounts of emotional motivation given by rival theories are inadequate. Perceptual theories rely on reflexive and hedonistic explanations of how emotions motivate, while cognitive theories understand the motivational role of emotions on the model of belief and desire pairs. But Scarantino shows that this ignores important motivational differences between a bout of fear and the combination of the belief that one is in danger and the desire to avoid it. There are obvious and predictable differences between one person fleeing a threat out of fear and another fleeing it fearlessly, simply to avoid the danger it poses. These differences include the range of options considered, the ability to integrate the escape with other goals, and the ability to attend to or recall various features of the surrounding circumstances. Scarantino lays out a sophisticated alternative, the motivational theory, which better explains these complex phenomena. He goes on to argue that it also meets other desiderata for a general theory of emotions: explaining the intentionality of emotion and providing a sound basis for differentiating emotion types. The next chapter takes up desire, another psychological state central to moral psychology. According to the most widely held theory, desires are simply (sophisticated, cognitively sensitive) behavioral dispositions. But mere behavioral dispositions appear to be inadequate explainers of things that philosophers have invoked desire to explain: love and care, having reasons for action, and the differences between acting on an appetite and acting out of a sense of duty. In their chapter, “The Reward Theory of Desire in Moral Psychology,” Timothy Schroeder and Nomy Arpaly argue against this orthodox theory. Although desire typically motivates behavior, they argue that it is not essentially motivating. Rather, desire is a natural kind that can be picked out by understanding the role it plays in the brain’s reward system. Motivation is one typical effect of a desire, as is pleasure at its satisfaction, but these are not constituents. After explaining their alternative, the reward theory of desire, they go on to explore its implications for a number of important issues in philosophical moral psychology. For example, one standard objection to desire-based theories of reasons for action arises from perverse behavioral impulses, such as compulsions, which do not seem to rationalize the actions they cause. The objection assumes that if a person is behaviorally impelled to act in a coherent way that is responsive to the person’s instrumental beliefs, then what motivates the person must be an intrinsic desire—and if the behavioral impulse is strong, then the intrinsic desire must be a strong one. The rejection of the motivational theory of desire opens up the possibility that some of these behavioral dispositions are caused by something other than a desire, or that the behavioral

Introduction  7 impulses are much stronger than the desires that cause them. Hence, a desire-based theory of reasons need not entail that such impulses provide (strong) reasons to act. The authors argue that the reward theory of desire also helps to handle objections to the idea that motivation by moral duty can be motivation by desire, and to the idea that intrinsic desires provide a grounding for an adequate account of love and care. Selim Berker examines an important argument made by Sharon Street over the course of several articles, in his chapter, “Does Evolutionary Psychology Show That Normativity Is Mind-Dependent?” According to Street, the evolutionary origins of our moral judgments generate profound epistemological problems for normative realists, to which the Humean constructivism she favors is immune. Street’s argument against realism takes the form of a dilemma. Either the evolutionary forces that shaped our moral beliefs influenced those beliefs so as to track the (attitude-independent) moral truth, or they did not. If the realist claims that they did, he must defend an empirically implausible account of the evolution of moral thought, which loses out to better scientific explanations in which the correctness of our moral beliefs play no role. But if he grants that they did not, then he has no basis for supposing that beliefs which were forged to promote our fitness also manage, by sheer coincidence, to track the moral truth. Berker works through Street’s arguments carefully and argues that they do not succeed. He concludes that the distinctively evolutionary challenge in Street’s argument can be met by various species of “third factor” accounts, which posit facts that both (causally) help make various normative judgments adaptive and (metaphysically) help make those judgments true. While Street has offered objections to third factor accounts, Berker argues that that they apply to her own view as well. Moreover, although he grants that these objections raise a profound problem, it does not depend on any claims about evolution. Ultimately Berker concludes that the deepest problem here is the familiar puzzle of how to justify reliance on our most basic cognitive faculties without relying on those faculties in a question-begging manner. Perhaps there is no answer to this problem, but this is not unique to normative theory, has nothing particularly to do with evolution, and challenges Street’s own view as well. The final chapter, “Sentimentalism and Scientism,” is written by the editors. It challenges a central thesis of the empirical ethics movement, which we call pessimism about reasons and reasoning. This is the claim that the reasons people offer in support of their evaluative judgments are merely post hoc rationalizations of conclusions driven by their emotional responses. This pessimism is manifest in Peter Singer’s claim that the science of ethics presents a dilemma between relativism (or some other form of skepticism) and a morality denatured of anything contingently human—in particular, of all emotional influence. We argue that although pessimists are right that emotions influence moral and evaluative thinking in various ways, and some of these influences are deleterious, no pessimistic thesis strong enough to support the dilemma is tenable. In fact, a more philosophically adequate theory, which discriminates between good and bad reasons and reasoning, is at least as well supported by the evidence. This view, rational sentimentalism, is anthropocentric but not relativist or otherwise skeptical.

8  Justin D’Arms and Daniel Jacobson It thus offers a route through Singer’s dilemma that avoids the costs of relativism and denatured rationalism. We then use rational sentimentalism to offer a novel defense of a modest form of retributivism against Joshua Greene’s charge that retributivists are rationalizers, and to press a more familiar challenge to the oppressive social practices that relativism must defend whenever they are sufficiently accepted. Finally, we argue that our anthropocentric view captures certain well-founded reasons that are specific to humans, such as those that issue from the symbolic and expressive aspects of action. Some psychologists ignore or mistakenly reject such reasons, treating commonsense tendencies to act on them as symptoms of magical thinking and taboo. These claims are not deliverances of science, however, but uncharitable assumptions that are symptomatic of scientism.

2 Intuitive and Counterintuitive Morality Guy Kahane

1.  Intuition and Deliberation in Moral Psychology Recent empirical work on moral psychology is often claimed to have yielded surprising and disturbing results. It has discovered, it is said, that moral judgment is not based in reason or conscious reflection, but in immediate intuitions. And it tells us that individuals rarely engage in deliberation when they make moral judgment, and even when they do, their reasoning merely aims to rationalize a pre-established intuitive conclusion. To shift to psychological terms, the emerging picture is that moral judgment is based almost exclusively in processes that are automatic—fast, effortless, and unconscious—and only rarely, if at all, in processes that are deliberative—slow, effortful, and involving explicit conscious thinking (Haidt 2001, 2007, 2012; Greene and Haidt 2002).1 When these findings are discussed, it’s usually assumed that what is surprising about them is the key role they ascribe to intuition in moral judgment. But in itself this is neither surprising nor disturbing. It is hardly news to moral philosophers that intuition plays a central role in our moral lives. Moral philosophers would be surprised, however, to find out that our intuitions are not just starting points for ethical reflection, but also where it invariably ends. A pressing question, then, is whether moral judgment is ever also based in genuine deliberation. An important body of research by Joshua Greene and his colleagues, and research influenced by it, has offered a striking answer to this question. While broadly confirming the view that automatic processing drives the moral judgments of most individuals, it also appears to show that there is a minority whose moral judgments are based on deliberation. Moreover, this minority uses deliberation to arrive at moral conclusions that, far from being mere rationalizations of prior intuitions, actually go counter



For general discussion of these two kinds of psychological processes, see Stanovich 1999; Evans 2007.

1

10  Guy Kahane to their immediate intuitions (Greene et al. 2001; Greene et al. 2004; Greene 2008; Paxton and Greene 2010). What is most striking about this research, however, is that it has tied the psychological divide between automaticity and “controlled” deliberation to moral judgments with opposing contents. For it appears to show that when the majority follow their immediate intuitions, the result is deontological in content, whereas when individuals do engage deliberative reasoning, they arrive instead at contrary utilitarian conclusions.2 This dual process model of moral judgment is meant as a quite general account of moral psychology, dividing the mind into two modes of processing that generate conflicting moral outputs—and moreover moral outputs that have a strong philosophical resonance.3 Needless to say, Greene’s dual model would be of great interest even if it were only a descriptive account of our moral psychology. But much more might be at stake here, since Greene and others have gone on to argue that this theory has dramatic normative implications—that it offers support for utilitarianism and its many counterintuitive implications (Greene 2008; Singer 2005).4 Notice again that what is potentially disturbing about this model for non-utilitarians isn’t the role it ascribes to intuition in deontology. Most (though not all) non-utilitarian views explicitly appeal to intuitions. This is merely a truism, something we don’t need fancy neuroimaging to know.5 What might be disturbing isn’t the presence of intuition, but the supposed absence of deliberation: the claim that deontological judgments 2 In what follows, I will reluctantly follow Greene and pretty much everyone else in this literature in using “utilitarian judgment” to refer to what are really just judgments that, in a given decision context, are more utility maximizing than the alternatives (see Kahane and Shackel 2010). As we shall soon see, this terminology is highly misleading at best. I will also reluctantly follow this literature in using “utilitarianism” to refer to what is at best a very simple form of Act Utilitarianism, understood, moreover, not just as a criterion of rightness but also as a concrete decision procedure (see again Kahane and Shackel 2010). This simple variant of utilitarianism is actually rejected by many contemporary utilitarians, let alone by non-utilitarian consequentialists. But the criticism I will be developing here is meant to apply even if we assume this incredibly narrow understanding of utilitarianism. Finally, again in line with the current literature, I will be using “deontological” to refer merely to views that are non-utilitarian. 3 Greene and others often refer to the view as the “dual process model of moral judgment.” But this label is misleading. Greene isn’t merely applying the general dual process approach to the mind to the moral domain—something that can be done in multiple ways. What is interesting (and controversial) about Greene’s model is the way he ties automaticity and deliberation (aka System 1 and System 2) to distinct moral outputs, something that is in no way entailed by the general dual process approach (see Kahane 2012 for an attempt to precisely state Greene’s model and its different components). When I criticize Greene’s dual process model in what follows, I intend to refer only to this more specific view; I will later actually outline an alternative way to apply the dual process approach to the model domain. 4 For criticism of Singer and Greene’s normative arguments, see Berker 2009; Kahane, 2010. 5 It would be slightly, but only slightly, more surprising to discover that non-utilitarian judgments are driven by emotion. But this is something that some non-utilitarians (think of Bernard Williams) explicitly avow. And the evidence suggesting that non-utilitarian judgments are driven (as opposed to influenced) by emotion is far from conclusive (see, e.g., Huebner et al. 2009). In any event, whether or not such judgments involve emotion, we now know they are responsive to the interaction of a complex range of factors—they are nothing like crude gut reactions. In line with that, in a recent study we found that a pharmacological intervention that reduces the physiological arousal component of emotion (that is, literally the “gut” part of emotion) in fact led to an increase in deontological judgment (Terbeck et al. 2013).

Intuitive and Counterintuitive Morality  11 involve nothing beyond intuition, and that deliberation always (or perhaps nearly always) points in a utilitarian direction. In what follows, I will therefore largely grant the claim that deontological judgment is based in intuition.6 What I will challenge in this chapter is Greene’s more controversial claim that utilitarian judgments are uniquely based in deliberative processing. I will argue that a closer inspection of the evidence supports a very different picture of our moral psychology. My argument will proceed as follows. In section 2, I will provide evidence that there is in fact nothing distinctively utilitarian about deliberation to “utilitarian” conclusions. In section 3, I will go further and argue that such deliberation to “utilitarian” conclusions is in fact best understood as non-utilitarian in character. In section 4, I will argue that when individuals do arrive at counterintuitive conclusions in a way that might actually resemble utilitarian reasoning, these judgments actually turn out to be non-deliberative—and worse. In section 5, I will present findings that suggest that there is actually little connection between so-called “utilitarian” responses to trolley-like dilemmas and genuine concern for the greater good. I  conclude, in section 6, by drawing some general lessons for the cognitive science of ethics. I will not directly consider the possible normative significance of this body of research. But if my argument is even partly successful, then the idea that this empirical research can be used to support utilitarianism will seem highly misguided, at best.

2.  Utilitarian or Counterintuitive? 2.1.  The Evidence Linking Utilitarian Judgment and Deliberative Processing There is a great deal of empirical research that is supposed to support Greene’s dual process model. However, much of this research in fact only provides evidence for the uncontroversial tie between deontological judgment and automaticity, or for the slightly more controversial tie between such judgment and emotion (see e.g. Greene et al. 2001; Greene et al. 2004; Mendez et al. 2005; Koenigs et al. 2007; Ciaramelli et al. 2007; Moretto et al. 2009). The available evidence for deliberative processing in utilitarian judgment is rather more limited. Some of it comes from Greene’s neuroimaging studies. These studies have reported that moral judgments in so-called “impersonal” trolley-like dilemmas— dilemmas to which most people give “utilitarian” answers—recruit greater activity in areas classically associated with cognitive processing (the right dorsolateral prefrontal cortex (DLPFC) and inferior parietal lobe) compared to “personal” dilemmas like the famous Footbridge case, to which most people give “deontological answers” (Greene 6 For further discussion of the relation between deontological judgement, intuition, and emotion, see Kahane 2012.

12  Guy Kahane et al. 2001; Greene et al. 2004).7 Moreover, when subjects do endorse utilitarian solutions to “difficult” personal dilemmas, these judgments turn out to be associated with greater activation in the DLPFC and the dorsal anterior cingulate cortex (dACC), an area implicated in conflict, compared to contrary deontological judgments (Greene et al. 2004). The classic marker of the automatic/deliberative contrast is difference in response times (RT)—deliberative processing should take longer. Although Greene originally reported that utilitarian judgments in personal dilemmas were associated with greater RTs compared to deontological ones (Greene et al. 2001; Greene et al. 2004), these results have not held up when the stimuli used were better (Moore et al. 2008; Greene et al. 2008; see also McGuire et al. 2009; Greene 2009). However, another classical marker is interference through cognitive load: A competing task should affect only responses based in deliberative processing. Greene et al. 2008 report that a cognitive load manipulation raised response times for utilitarian judgments but not for deontological judgments in “high conflict” personal dilemmas. However, contrary to what Greene’s model predicts, such a manipulation did not affect the rates of utilitarian judgment. Nevertheless, in line with the model, more recent studies did find that time pressure increased rates of utilitarian judgment (Suter and Hertwig 2011), whereas stress reduced them (Youssef et al. 2012). A further source of evidence comes from studies tying utilitarian judgement to traits and capacities associated with deliberative processing. In particular, higher rates of utilitarian judgments were associated both with “need for cognition,” a motivational tendency to prefer effortful cognition (Bartels 2008), and with greater working memory capacity, a component of general intelligence (Moore et al. 2008; but see also Koven 2011).8 In addition, exposure of people to non-moral problems that were specifically designed to assess the capacity to suppress a strong intuitive answer and instead endorse a counterintuitive solution (Frederick 2005) also induced greater rates of utilitarian judgment (Paxton et al. 2012).9 7 In “personal” moral dilemmas like Footbridge, one must choose whether to significantly harm someone in an “up close and personal” way, or to let a larger number of people die; “impersonal” dilemmas involve a similar decision, but the harm is done in a more indirect way. In what follows, I will ignore important worries about classifying moral dilemmas in this way (for further discussion, see Kahane and Shackel 2010). 8 Notice however that these studies offer only indirect evidence for the role of deliberative processing in utilitarian judgment, since unlike the studies cited here, these studies don’t directly measure the processes that are involved in generating such judgments. The tie between utilitarian judgment and, say, a motivational tendency to effortful cognition may be explained by greater engagement in effortful cognition when individuals make utilitarian judgments, but it can also be explained by other factors (individuals higher on need for cognition may be different from other people in all sorts of ways that might affect their moral judgment). 9 As I point out in Kahane 2012, n. 10, Moore et al. 2008 and Paxton et al. 2012 (as well as Paxton et al. 2013) tied “utilitarian” judgment to potential markers of deliberative processing only in “personal” dilemmas where—unlike the classic Footbridge case—even if you don’t harm someone to save five, that person will die anyway. This is problematic, since many deontological views would also endorse such a conclusion. For other methodological worries about some of the evidence listed in this section, see Kahane and Shackel 2011 and Berker 2009.

Intuitive and Counterintuitive Morality  13

2.2.  Problematic Inferences There is thus considerable evidence appearing to establish a tie between utilitarian judgment and deliberative processing. There is, however, an obvious gap in this evidence. Greene’s dual process model is clearly meant to be entirely general: It makes a general claim about utilitarian judgment, a claim that is supposed to apply across moral contexts and domains.10 But virtually all of the research taken to support the model has focused on so-called “trolley problems” and similar scenarios where one person must be seriously harmed to save a greater number. Such dilemmas of course relate to a rather unusual part of morality, and are merely one of numerous contexts in which utilitarianism clashes with commonsense intuitions11—to name just a few other examples, act utilitarianism requires complete impartiality, implying both that we shouldn’t give any inherent moral priority to those dear and near to us, and that we have obligations to make extremely demanding sacrifices to help distant strangers in need (I’ll return to this example below); utilitarianism also rejects retributive punishment and the notion of desert quite generally, and similarly gives no intrinsic weight to considerations of distributive justice—and this is just the beginning of a long list.12 So the model currently relies on a huge inductive leap from findings about trolley-like problems to general conclusions about the sources of “utilitarian” and “deontological” judgment. It’s unfortunately common for researchers to report that some study has shown utilitarian judgments to be this or that, where it at most shows this about utilitarian (viz. utility maximizing) judgments in trolley-like dilemmas. We shall later return to this problem. My more immediate aim in what follows, however, is to show that even in the narrow domain of trolley-like problems, this supposed tie between utilitarian judgment and deliberative processing involves a serious misreading of the current evidence. In the next two sections, I will therefore focus on

10 The intended strength of this claim isn’t entirely clear. But the idea seems to be that utilitarian judgments are the favored outputs of deliberative processing, and that contrary deontological outputs are at best rare exceptions, even aberrations. Notice moreover that Greene’s model makes a claim about the causal source of each type of judgment. It is perfectly compatible with deontological judgments sometimes engaging deliberative processing in an extensive way, so long as this processing plays a merely epiphenomenal role in the final output—so long as it is engaged merely to offer a rationalization of the initial intuition. 11 In fact, although Greene and other researchers often present the trolley case as a core part of the traditional case against utilitarianism, this is highly misleading at best. The trolley problem is a problem internal to non-utilitarian ethics, as a look at Foot 1978 and Thomson 1976 would confirm. 12 When we spell out the ways in which utilitarianism clashes with commonsense morality, we need to be careful with the distinction between criterion of rightness and decision procedure (see n. 2). So-called “indirect” utilitarianism, for example, may allow us to give priority to our loved ones, because such a “decision procedure” (i.e. actual moral psychology) will in fact lead to better consequences. But this point, of course, also applies to how we respond to the classical trolley cases (see Kahane and Shackel 2011). As I explained above, I will be following Greene in using “utilitarianism” to refer to a very simple form of Act Utilitarianism—a slightly simplified version, if you want, of the view associated with Peter Singer.

14  Guy Kahane another problematic inference that underlies the dual process model. Greene and others slide from the supposed empirical finding that (1) when individuals make utilitarian judgments in trolley cases, deliberative processing plays a causal role in the generation of these judgments to the conclusion that (2) deliberative processing reflects the distinctively utilitarian character of these judgments. In what follows, I shall argue that this deliberative processing is unlikely to have much to do with the utilitarian content of these judgments. In fact, we shall see that the connection between deliberative processing and utilitarian judgment is rather superficial.

2.3.  An Overlooked Confound? Here is one obvious reason why it is invalid to infer, from evidence associating “utilitarian” judgments with deliberative processing, the conclusion that this processing reflects the distinctive utilitarian character of these judgments. Greene and other researchers typically compare utilitarian judgments that are highly counterintuitive (such as: push the fat man in the Footbridge dilemma) with contrary deontological judgments that are strongly intuitive (e.g. don’t push). They observe various behavioral and neural differences between the two. And they conclude that these differences must reflect differences between utilitarian and deontological judgments. But there is a glaring alternative explanation they have overlooked: These differences might merely reflect differences between counterintuitive and intuitive judgments, quite regardless of the content of these judgments. If this is correct, then the apparent tie between process and content is really just an artefact of the kinds of scenarios that researchers have studied, reflecting nothing very interesting about utilitarian and deontological judgments. Now this alternative explanation wouldn’t be a problem if all counterintuitive judgments were utilitarian, and all intuitive ones deontological. But this is obviously not so, as Kant’s notorious assertion that we mustn’t lie even to prevent murder demonstrates (Kant 1797/1966; this example is actually mentioned in Greene 2008). Most people seem to find it immediately obvious that we should lie in such a case—it is natural to say that they find this “utilitarian” decision intuitive, and the contrary deontological one strongly counterintuitive. And the mere existence of these overlooked pairings of content (utilitarian/deontological) and intuitiveness (intuitive/counterintuitive) already presents apparent counterexamples to Greene’s model. Could Greene deny that there really are these counterexamples? With respect to supposed instances of intuitive utilitarian judgments, he might argue that although some “utilitarian” judgments are immediate and effortless, and based on intuition in some broad sense, they are nevertheless not genuinely automatic—though this would also seem to drain talk of automaticity of much of its content. In any event, Greene himself speculates that utilitarian judgments have their origin in immediate affective

Intuitive and Counterintuitive Morality  15 responses to harm to others (Cushman, Young, and Greene 2010). So he already seems committed to accepting this counterexample, although it’s not clear how exactly to square this with his general model.13 I’m more interested, however, in the possibility of counterintuitive deontological judgments, since they challenge the crucial tie between deliberative processing and utilitarianism. How might Greene challenge this possibility? Greene doesn’t deny that some deontological claims, such as Kant’s remarks on lying, seem counterintuitive. But he seems to treat these as rare cases where a commitment to an explicit philosophical theory leads some philosophers to make counterintuitive judgments (Greene 2008, 65–6; Paxton and Greene 2010). However, non-philosophers plainly make similar deontological judgments, and Greene is surely committed to claiming that such judgments must involve utterly different psychological processes than the counterintuitive utilitarian judgments on which the research has so far focused. One possibility that is in line with Greene’s model is that judgments that appear to be counterintuitive deontological judgments really reflect, not the effortful overcoming of a utilitarian intuition, but unusual affective responses—e.g. an atypically strong aversion to lying. In any case, apparent everyday examples of counterintuitive deontological judgments, and utilitarian judgments based on intuition, have so far been strangely ignored by researchers. What would we find if we examined them? Taken at face value, Greene’s dual process model makes clear predictions. Utilitarian judgments should involve deliberative processing, whether they are (in the loose sense set out above) intuitive or not. And the neural and behavioral correlates of counterintuitive deontological judgments should be utterly different from those of counterintuitive utilitarian ones such as in the Footbridge dilemma. The alternative hypothesis I’ve just offered makes exactly contrary predictions: Counterintuitive judgments should engage similar brain areas whether or not they are deontological or utilitarian; and a parallel prediction follows for intuitive judgments. In a recent neuroimaging study, we set out to test these competing hypotheses.

2.4.  Intuitiveness vs. Content In that study, we first operationalized “intuitiveness” by surveying the unreflective judgments of an independent sample of non-philosophers, and classifying types of judgments as intuitive if they were clearly dominant—if they were made unreflectively by 13 or more out of the 18 independent judges; the contrary judgments were classified as “counterintuitive” (Kahane et al. 2012).14 In addition to the commonly used 13 An even more embarrassing counterexample is provided by a recent study in which Greene himself was involved. That study found, across a wide range of decision contexts, that most subjects find cooperative decisions more intuitive than selfish ones, which typically require greater deliberative effort (Rand, Greene, and Nowak 2012). It’s hard to see how this squares with Greene’s model. Moreover, to the extent that an association with deliberation is supposed to give support to a normative view—as Greene often seems to assume—then these results should strongly support rational egoism. 14 On this operationalization, “intuitiveness” is a population-relative notion, so judging that we are forbidden to push in Footbridge counts as intuitive in this sense even though, as we shall see in section 4, some

16  Guy Kahane dilemmas such as Footbridge where the utilitarian choice is typically counterintuitive, we also used dilemmas, not previously studied, where the deontological choice is counterintuitive, such as refusing to lie to prevent harm. Importantly, although only a minority of these lay judges endorsed refusal to lie in such cases, it is highly unlikely, to put it mildly, that this refusal was based in adherence to some explicit deontological theory. We then used fMRI to compare the neural correlates of responses to these two kinds of dilemmas. Setting aside many points of detail, our main finding can be summarized as follows: The apparent neural and behavioral differences between utilitarian and deontological judgments in trolley-like dilemmas appear to be driven almost exclusively by differences in their intuitiveness, not in their content.15 To focus on the case that is most relevant to our discussion, “utilitarian” judgments such as that it is appropriate to push in Footbridge, and what we can call “ultra-deontological” judgments, such as refusing to lie to prevent harm, were associated with strikingly similar patterns of neural activation, compared to contrary intuitive judgments. In other words, counterintuitive judgments with radically opposing contents were based in highly similar neural processes. These results are plainly incompatible with Greene’s dual process model.16 They show, first, that it’s not the case that deontological judgments are associated with automatic processing, and utilitarian with deliberative processing, across different moral domains.17 But our results also strongly suggest that the differences between utilitarian and deontological judgments reported by previous studies largely reflect differences in individuals may actually lack this intuition. Notice that we don’t define intuitiveness as merely a matter of consensus. The independent judges were explicitly asked for their unreflective response; after reflection, some gave different answers to some of these dilemmas. 15 See Kahane et al. 2012 for the full details of the analysis that supports this conclusion. For a similar conclusion based on a different approach, see Baron et al. 2012. 16 In an empirical response to our study, Paxton et al. 2013 report that responses to the Cognitive Reflection Test, a measure developed to measure counterintuitive thinking in a non-moral context, was associated with increased utilitarian responses both in a personal dilemma and in one of the lying dilemmas we used in Kahane et al. 2012. This is an interesting result that is not predicted by my remarks above. However, even setting aside the fact that only a single dilemma of each type was used in this study, and other methodological worries (see e.g. n. 9), this result does not really address the main result of our study: that similar neural processes were associated with counterintuitive moral judgments of opposing contents, and that the differences between utilitarian and deontological judgments within (appropriate) personal dilemmas were almost entirely explained by differences in intuitiveness. 17 Notice that our main aim was to compare counterintuitive judgments of opposing contents. It wasn’t to demonstrate that in some contexts, deontological judgments also involve deliberative processing. In fact we didn’t find that counterintuitive judgments were generally associated with greater response times, or with activation in classical cognitive areas, compared to contrary intuitive ones. In fact we didn’t find this even when we compared only counterintuitive utilitarian judgments with intuitive deontological ones—the comparison closest to that in Greene et al. 2004. Counterintuitive judgments were, however, associated with greater perceived difficulty, and with activation in the subgenual part of the rostral ACC, an area that has been implicated in affective conflict (Etkin et al. 2006), but also in feeling guilt (Zahn et al. 2009a; Zahn et al. 2009b). So we actually found only limited support for the hypothesis that counterintuitive judgments— including counterintuitive utilitarian judgments in trolley-like cases—are generally associated with deliberative processing.

Intuitive and Counterintuitive Morality  17 intuitiveness even when we consider only “personal” dilemmas like Footbridge cases. So it seems that the neural and behavioral correlates of paradigmatic “utilitarian” judgments merely reflect deliberation to a counter-intuitive conclusion—and not anything distinctively utilitarian.18

2.5.  A Less Exciting Dual Process Model So what we get, instead of Greene’s exciting and controversial dual-process model, is an alternative and rather less surprising dual-process model: The Banal Dual Process Model. Intuitive judgments are generally associated with automatic processing, and counterintuitive judgments with deliberative processing.

Greene has suggested to me that his dual process theory is really meant to essentially make this more general claim about intuitive and counterintuitive judgment. This, however, is certainly not how the theory has so far been presented or understood. If Greene’s claim is simply that many utilitarian judgments are counterintuitive, and that the processes involved in making such judgment merely reflect the processes generally involved in making counterintuitive judgments, then these are near truisms. It is hardly a great surprise that counterintuitive judgments are harder to make, or that they require more deliberative effort. And since we know that some utilitarian judgments are counterintuitive, it is also hardly surprising that these judgments are harder to make . . . But this would give no support to the ambitious theoretical and normative arguments that Greene and others have elaborated on the basis of the dual process model—after all, these truisms imply nothing interesting that is specific to utilitarianism, let alone anything favorable to utilitarianism. It might be replied that even if deliberative processing is generally associated with counterintuitive moral judgments, such processing nevertheless favors utilitarian judgments. If this is a general claim, not specific to trolley-like cases, it’s hard to see what evidence is supposed to support it. How exactly are we supposed to measure such a general tendency?19 It is also important to distinguish this interesting sounding hypothesis from the far less exciting claim in the reverse direction: that utilitarian judgment more often involves deliberative processing simply because utilitarianism issues more counterintuitive conclusions compared to many other moral views. The latter claim is a mere truism. We don’t need fancy neuroimaging to know that. Let me summarize this section. I have argued that even if utilitarian judgments in trolley-like dilemmas are deliberative, this deliberation is merely generic. If these judgments involve deliberative processing, this is merely because they are counterintuitive,

18 In line with this, we also didn’t find common correlates for so-called “utilitarian” judgments across different domains. 19 And see again the Rand, Greene, and Nowak (2012) study mentioned in n. 13, associating deliberation and selfish decisions. Does deliberative processing favor both utilitarianism and rational egoism?

18  Guy Kahane not because they’re utilitarian. And the processes involved are (broadly) the same as those involved in counterintuitive ultra-deontological judgments.

3.  Deliberating about What? 3.1.  “Utilitarian reasoning” and the Footbridge dilemma In this section, I want to take the argument one step further.20 Consider again the problematic inference from evidence associating “utilitarian” judgments with deliberative processing, to the conclusion that this processing reflects the distinctive utilitarian character of these judgments. To decide whether this inference is valid, we need to know what it would mean for such processing to have some distinctive utilitarian character. Unfortunately, however, Greene’s dual process model is surprisingly unclear on this question. Greene sometimes writes that utilitarian judgments are generated by “utilitarian reasoning.” But what does Greene mean when he speaks of “utilitarian reasoning” (or “cost–benefit analysis”) leading, say, to the judgment that it is morally appropriate to push the stranger in Footbridge? Greene does not, of course, think that individuals making such judgments explicitly endorse act utilitarianism or any similar ethical theory. Neither is it plausible that they endorse such a theory implicitly, since virtually all of them make some deontological judgments in some contexts (see Kahane and Shackel 2011). Still, if they are to be said to be making “utilitarian” judgments even in the thinnest sense, their ground for judging that it’s appropriate to push the stranger must be that this would lead to better consequences. And if they reach this conclusion by explicit reasoning, as Greene holds, then they must be reasoning from a corresponding general moral principle. This means that such individuals must be following something at least approximating the following piece of reasoning:21 (1) We are required to impartially maximize well-being. (2) 5 lives > 1 life. Therefore (3) We are required to sacrifice 1 to save 5. If this is what individuals are going through when they arrive at utilitarian conclusions, then they would indeed be making an inference that sets out from a utilitarian premise to a utilitarian conclusion—though I’ll later question whether it’s usefully called utilitarian reasoning. 20 The argument in this section is heavily based on Kahane 2012. Some empirical and conceptual detail has been removed to clarify the argument, and several new points have been added. Still, readers already familiar with (or even persuaded by) the argument of that paper may prefer to skip to section 4. 21 This picture is essentially endorsed in Cushman, Young, and Greene (2010).

Intuitive and Counterintuitive Morality  19

3.2.  Deliberative Processing of What? Let’s suppose for the moment that subjects are really engaged in the piece of reasoning Greene ascribes to them.22 Greene’s main claim is that when subjects engage in such reasoning and reach utilitarian conclusions, they uniquely do so using deliberative processing. This is what the empirical evidence cited above is supposed to show. But we should ask:  What exactly does this deliberative processing reflect? Let’s briefly consider some possibilities. Recognizing a Foundational Principle? If the deliberative processing reflected recognition of the moral principle that we are required to maximize utility then it would perhaps reflect a distinctly utilitarian thinking. But this is highly implausible. Even if someone reached this normative view through effortful reflection when first confronted with, say, the Footbridge case, surely this is something they only need to do once, so this anyway couldn’t account for the greater levels of deliberative processing found in responses to a large set of dilemmas. Notice, moreover, that this is a foundational moral principle—such a principle can serve as a premise for reasoning, but it’s not itself plausibly supported by any kind of inference.23 So to the extent that the deliberative processing associated with utilitarian judgment is supposed to reflect reasoning, it can’t reflect the non-inferential judgment (aka intuition) that we should maximize utility. This example nicely illustrates the limits of thinking about moral psychology using the dual process framework. If the utilitarian principle that we ought to maximize the greater good is non-inferential, then it’s in one good sense intuitive. Yet it is also claimed to be the product of “reason,” rather than anything external to it (cf. Singer 2005; de Lazari-Radek and Singer 2012). Nor need it be immediate: That insight may strike someone after long, effortful reflection. Counting the Numbers? Another possibility, which is often implied by the way Greene and others describe utilitarian judgment, is that the deliberative processing reflects the calculation of “5 lives > 1 life.” The first problem with this suggestion is that this isn’t any kind of utilitarian reasoning. This is just standard non-moral reasoning, used to apply a moral principle—any kind of moral principle.24 It’s easy to vary the degree of non-moral reasoning needed to apply different moral considerations in a given context. Some “utilitarian” questions involve little or no calculation (“Should we give someone in agony a painkiller?”), 22 See Kahane 2012, §2.3 for further reasons to doubt this. 23 This is a point that is often emphasized by utilitarians like Singer—see Singer 2005; de Lazari-Radek and Singer 2012. 24 This point is actually supported by Greene’s own work: Shenhav and Greene (2010) report that so-called “utilitarian” reasoning has the same neural correlates as non-moral reasoning.

20  Guy Kahane and some deontological questions require quite a bit of calculation (“Should someone break a promise to a friend that conflicts with 3 promises to his daughter, when this requires lying twice?”). Responses to such questions are likely to differ in reaction times or DLPFC activation, reflecting obvious differences in complexity. But of course these processing differences would tell us absolutely nothing about the psychology of utilitarianism or deontology. Non-moral reasoning is simply irrelevant to that issue. It follows that if the deliberative processing associated with utilitarian judgment merely reflects the calculation of “5 lives > 1 life,” this would merely be a contingent artefact of the particular stimuli Greene and others are comparing, rather than any kind of interesting feature of utilitarian vs. deontological thinking. In any event, however, it is extremely unlikely that the deliberative processing reflects this simple calculation. Non-utilitarians are not numerically challenged. It is near certain that all subjects considering Footbridge and similar dilemmas make this simple calculation, whether or not they reach a “utilitarian” conclusion. It is astonishing that so many researchers seem to think otherwise. There is a further problem: It is surely the case that no cognitive effort is required to judge that 5 is greater than 1! As we shall see in section 4, there is in fact evidence that engaging in such “utilitarian reasoning” can actually be quick and utterly effortless.

3.3.  So What Does the Deliberative Processing Reflect? The argument I have been developing so far largely converges with the empirical evidence I presented in section 2. It suggests that if utilitarian judgments are associated with deliberative processing, that processing doesn’t reflect any kind of “utilitarian reasoning.” In the previous section, I already argued that this processing really reflects a generic type of deliberation to a counterintuitive conclusion. In the rest of this section, I want to look more closely into what such deliberation might involve in the context of trolley-like dilemmas. Overcoming Conflict? If “utilitarian” judgments in Footbridge involve deliberative processing then that processing couldn’t reflect “utilitarian reasoning.” So it must reflect something else. What could that be? Greene’s own account suggests a further option. After all, his account says that individuals not only engage in such “utilitarian reasoning,” but also need to actively suppress the pre-potent emotional response pushing them in the contrary deontological direction: Deliberation reflects conflict. When deontological intuitions are present, effortful deliberative processing is needed if one is to arrive at a contrary utilitarian judgment.

This is supposed to be why we find greater activation in the dorsal Anterior Cingulate Cortex (dACC), an area associated with conflict, when subjects make utilitarian judgments. As Greene often points out, the dACC is also active in the classical Stroop

Intuitive and Counterintuitive Morality  21 paradigm, where deliberative processing is needed to overcome the pre-potent impulse to name the color word rather than the color in which it is written (Greene et al. 2004, 390). I have so far deliberately ignored this part of Greene’s theory. Greene typically presents the deliberative processing associated with utilitarian judgment as reflecting both “utilitarian reasoning” and the conflict generated by the contrary emotion. The argument so far shows that it can at most reflect the latter. The problem is that talk about emotional conflict and overcoming an emotional response is ambiguous. It can mean two rather different things. The first interpretation, which is implicit in Greene’s view, doesn’t really support his model, and is also implausible. The second, more plausible interpretation is simply incompatible with Greene’s overall theory. I will consider each interpretation in turn. Resisting/Inhibiting a Pre-potent Emotion? If we take the analogy to the Stroop paradigm literally, then the deliberative processing should reflect the effort that individuals are making to resist a spurious distorting influence (see Paxton and Greene 2010). Such outright rejection of one’s intuitions is of course common in genuine utilitarian thinking for the simple reason that utilitarians reject all non-utilitarian moral considerations, and thus all of the intuitions that (at least sometimes) drive them. The first problem with this suggestion is that there is nothing distinctly utilitarian in rejecting some intuition as spurious—this is something that both utilitarians and deontologists do. Indeed, we also often dismiss our intuitions in this way in many non-moral contexts. In any case, since this is a purely negative operation, it tells us nothing positive about the source and nature of utilitarian thinking. Instead of the claim tying utilitarian judgment and deliberative processing, we again get the banal claim that whenever strong intuitions are present, effortful deliberative processing is needed if one is to arrive at a judgment contrary to these intuitions. It should be obvious that this near truism does not support the dual process model as such, let alone the grandiose theoretical and normative conclusions that Greene and others want to derive from it (see again Greene 2008; Singer 2005). In any event, it turns out that this interpretation of the data is also highly implausible. If subjects rejected their aversion to intentionally killing an innocent person as expressing any kind of genuine moral reason, then they should surely reject it as such a reason quite generally. But few subjects make consistently utilitarian decisions across the board even in the restricted context of trolley-style dilemmas—though in section 4 we will consider subjects who at least approximate such a pattern—so when subjects endorse the deontological answer to such dilemmas, are they simply suddenly “overcome” by an emotion they take to be spurious, as people sometimes get confused and make errors in the Stroop paradigm? This is utterly implausible. Moreover, subjects tend to report feeling guilty after making utilitarian judgments in personal dilemmas (Choe and Min 2011),25 and, when 25 More precisely, subjects reported feeling guilty when making utilitarian judgments in an overwhelming majority of personal dilemmas. However, Choe and Min found no correlation between a trait disposition to guilt, and rates of utilitarian judgment.

22  Guy Kahane given the chance, they report finding that both options are somewhat wrong (Kurzban, DeScioli, and Fein 2012). Neither of these results makes much sense if subjects are just resisting an insistent emotional urge which they perceive as utterly spurious. Non-utilitarian Weighing of Opposing Duties The subpersonal talk about conflict generated by a contrary emotion is also compatible with another, and more plausible, picture of the deliberation that most non-philosophers go through when they make “utilitarian” judgments in Footbridge. In most moral dilemmas, we face genuine opposing moral reasons, and try to figure out which of these “wins” in the given context. We don’t treat the reasons we ultimately reject as a mere psychological annoyance we need to repress. This is what makes moral dilemmas difficult: There are strong reasons supporting each of the conflicting choices.26 Once we begin to think of moral dilemmas in this more plausible way, an alternative account of what typically happens in trolley-like dilemmas naturally suggests itself. Subjects start, as I suggested earlier, with a moral reason or principle: (1) We have reason to reduce harm (or Prima Facie Duty to Save). And they count the numbers: (2) 5 lives > 1 life. But they also recognize that (3) There is reason not to intentionally harm an innocent person (or Prima Facie Duty Not to Harm). And at least some of them conclude, after some reflection, that (4) In this context, there is more reason to minimize harm. Which is why they conclude that (5) It is permissible to sacrifice one to save 5. If this is what is going on, then not only is it misleading to describe this as “utilitarian reasoning,” but what we have here is really a paradigmatic example of a distinctly deontological form of deliberation (for a classical defence of this understanding of everyday moral deliberation, see Ross 1930). After all Act Utilitarianism leaves no space for such

26 Of course, even on the utilitarian view, there is some reason not to push the fat man—needless to say, the fat man’s death is a very bad consequence. But this counter-reason cannot generate conflict, or the need for effortful deliberation—on this utilitarian view, the number of lives saved in each of the two options are straightforwardly commensurable. As I have been repeating, there is neither effort nor conflict in concluding that 5 is greater than 1.

Intuitive and Counterintuitive Morality  23 weighing of competing moral duties or reasons: On Act Utilitarianism (understood as a decision-procedure) there is only a single duty, viz. to maximize impartial well-being, and its application to different contexts.

3.4.  Misunderstanding Moral Deliberation Greene often portrays deontological responses as immediate judgments that some act is absolutely wrong, which he contrasts with the sophisticated utilitarian weighing of competing concerns (see e.g. Greene 2008, 64). But this gets things exactly upside down. Most plausible forms of non-utilitarian thinking (including, of course, “commonsense” morality) are not absolutist in this way. Instead, non-utilitarian moral deliberation typically involves precisely the weighing of competing moral considerations (including considerations about consequences), whereas utilitarian deliberation leaves space only for the (non-moral) comparison of the causal consequences of different lines of action. And, to repeat, in cases like Footbridge this utilitarian calculation should be obvious and effortless. This more plausible alternative has so far been overlooked because current research relies on an incredibly narrow conception of what deliberation is or can be. It unthinkingly identifies deliberation with inference, and sharply contrasts it with intuition, as if inference is an utterly independent, competing source of judgment—neglecting to see that the premises for reasoning also need to come from somewhere, that effortful deliberation needn’t involve anything like inference, and, moreover, that there are certain kinds of intuitions and emotions that appear only in deliberative contexts. It is because of that narrow conception of deliberation that Greene and others simply assume that if deliberative processing plays a part in generating “utilitarian” judgments, then this means that this deliberative processing is generating these judgments from scratch, and that it does so by inference—as if the only alternative is for the deliberative processing to produce mere rationalization of some pre-existing emotion or intuition. But deliberative processing can play a genuine causal role in producing moral judgments without necessarily being the source of their content. As described above, deliberative processing can help us decide between competing pro tanto moral reasons (or “prima facie duties”27) by generating an all-things-considered judgment about what ought to be done.28 And that decision can be conscious and effortful without being the result of inference—the judgment that, in some given context, one duty outweighs another is almost certainly also based in intuition! (Notice that this would be a second order intuition resolving a conflict between two first order intuitions. For a classical statement of the view that we need something like intuition to resolve conflicts between opposing principles, see again Ross 1930).29 27 I follow W. D. Ross in speaking of “prima facie duties,” but the relevant reasons are not literally “prima facie” (merely appearing to be genuine reasons), and are best described as “pro tanto” (genuine reasons that can be outweighed in a given context). 28 Nichols and Mallon (2005) show that non-philosophers can easily distinguish between whether an act violated some rule (weak impermissibility) and whether that act was wrong, all things considered. 29 I’m not claiming that no one treats the intuition not to push as spurious, only that most people don’t. There certainly are clear cases where people reject their intuitions/affective reactions as simply morally

24  Guy Kahane It might be objected that subjects couldn’t be going through this deliberative process because, although something like the principle of utility is an explicit principle that people actually cite, the resistance to, e.g., pushing the stranger in Footbridge is an intuition that most people are unable to articulate as an explicit principle (Cushman et al. 2006). But this is irrelevant. We can be conscious of, and torn between, opposing moral considerations even if we can’t articulate them as fully explicit principles.30 This gives us an alternative explanation of the role of deliberative processing in utilitarian judgment: Deliberation as weighing of opposing reasons. “Utilitarian” judgments are generated by deliberative processing reflecting deliberation about the proper weighing of various moral reasons/ principles—some or all of which are deontological in character—when these generate conflicting verdicts about a given situation.

Notice that this account is perfectly compatible with the claim that “utilitarian” judgments in “personal” trolley-like dilemmas are more closely associated with deliberative processing. In many moral contexts, one of the opposing moral reasons or principles can be especially salient, and dominate, even pre-empt, deliberation. It might thus take some effort to see and give proper weight to competing considerations that are less salient.

3.5.  “Utilitarian” Judgment ≠ Utilitarian Reasoning Let me end this section by tying the above argument to a familiar objection to utilitarianism. Consider Bernard Williams’s famous anti-utilitarian example of Jim and the Indians, where Jim is told that if he shoots one Indian, the lives of several others would be spared (Williams in Smart and Williams 1973). Although this is sometime misunderstood, the point that Williams wanted to make with this example wasn’t that we shouldn’t make the “utilitarian” decision to save more lives—he was actually inclined to think we should. Williams’s point was rather that this decision shouldn’t be as easy as is implied by Act Utilitarianism; it should be a difficult, agonizing choice. And indeed, unlike a minority of individuals which (as we shall see in section 4) do find this an easy choice, most people seem to find such “utilitarian” choices difficult in exactly the way Williams described—and not merely because they need to overcome some irritating but spurious “pre-potent response.”31 spurious—one example might be the way that, according to Haidt, liberals regard their disgust reactions to harmless violations (Haidt 2001). 30 In fact the utilitarian principle is itself indeterminate in multiple ways that utilitarians still disagree about. 31 If you could put Bernard Williams in the scanner, and ask him for his view about the Jim and the Indians case, he would engage in agonizing deliberation about what one should do in such a situation. Suppose he ends up endorsing the “utilitarian” conclusion that Jim should shoot one of the Indians to save the greater number. It would be absurd to then say that, since he arrived at this conclusion through effortful deliberation, this shows that utilitarianism is based on more cognitive processing. It would be even more absurd to think that by studying the neural correlates of Williams’s agonizing deliberation, we are learning anything

Intuitive and Counterintuitive Morality  25 Greene et al. 2004 report that utilitarian judgments were associated with greater activation in the insula, which they speculate might reflect repugnance at the sacrifice of the one. Moretto et al. 2009 report that a strong emotional reaction (reflected in skin conductance response) followed utilitarian judgments. And as we saw, Choe and Min 2011 report that utilitarian judgments in most personal dilemmas were reported by subjects to be accompanied by some guilt.32 It’s hard to square these three findings with the common assumption that subjects view the deontological intuition as merely a gut reaction that needs to be resisted—but they nicely fit Williams’s 1965 remarks about the moral residue we feel when we are tragically forced to choose one of two profoundly bad options. The lesson here is that it is a mistake to assume that if a moral judgment is “utilitarian” (in the narrow sense that, in a given moral context, it favors the act that maximizes utility), then this implies that the deliberation that generated it must also be “utilitarian.” Quite the opposite: Such “utilitarian” judgments are typically based in a distinctly deontological form of deliberation that is profoundly incompatible with utilitarianism.33 In this section, I’ve argued that when utilitarian judgments involve deliberation, it’s not just that this deliberation doesn’t reflect anything distinctly utilitarian—it’s a form of deliberation that is resolutely non-utilitarian in character!34 If an association with deliberation is supposed to support a moral theory, as Greene and others think (a rather questionable assumption), then, ironically, this empirical evidence should actually support non-utilitarian views.

4.  Cold or Calculating? “Pure” Utilitarian Judgments are Non-Deliberative 4.1.  Utilitarian Judgment without Deliberation Utilitarians hold that their counterintuitive conclusions are simply what results when, instead of just following our immediate gut reactions, we use moral reasoning to critically scrutinize them (Singer 2005; Unger 1996). This picture of utilitarian psychology

about the psychological sources of utilitarian thinking. I have been arguing in this section that a lot of current research on “utilitarian” judgment falls exactly into this absurd mistake. 32 Utilitarian judgments in some dilemmas were more strongly associated with disgust and other emotions—but these also happen to be the dilemmas that Nick Shackel and I have previously shown to be utterly irrelevant, as now also conceded by Greene (see Kahane and Shackel 2008; Kahane and Shackel 2010). 33 This is just one reason why it’s misguided to suggest, as Greene does, that we take “utilitarian” to refer to whatever processes turn out to underlie judgments with “characteristically utilitarian content” (Greene 2008). 34 Needless to say, if utilitarianism is understood only as a criterion of rightness (and not also as a decision procedure), then it is logically compatible with any form of deliberation—including rational egoism or applying the Categorical Imperative. But Greene’s claims about the origins of utilitarian thinking require treating it, at least in this context, as referring to a distinctive decision procedure.

26  Guy Kahane of course sits nicely with Greene’s dual process model, which portrays lay people who make utilitarian judgments as thinking harder than the rest. In the previous sections of this chapter, I  have cast some doubt on this claim. In this section, I want to take the argument yet one step further. I will point to evidence that strongly suggests that there are in fact some individuals who arrive at utilitarian conclusions through a process that more closely resembles a utilitarian decision procedure. The problem is that it appears that these utilitarian judgments are non-deliberative—they are driven, not by greater cognitive effort, but by an affective deficit. Such individuals are not more “calculating” in the sense of being more rational. They are simply colder. This alternative picture of “utilitarian” psychology is actually supported by a growing body of evidence. Some of this evidence comes from clinical populations. Patients with lesions in the vmPFC (Ciaramelli et al. 2007; Koenigs et al. 2007; Moretto et al. 2009) and with frontotemporal dementia (Mendez et al. 2005), conditions associated with deficits in empathic concern and social emotion and with disordered social behavior, exhibit increased rates of utilitarian judgment in emotionally-loaded moral dilemmas, apparently because such patients lack the normal aversive response to harming—they lack, if you want, the intuition that it’s wrong to push the stranger in Footbridge. Now this result is taken by Greene and others to be a key bit of evidence supporting the dual process model: If the absence of social emotion leads to an increase in utilitarian judgment, this supports the claim that deontological judgment is based in emotion (see e.g. Greene 2008). This, however, is only one half of the dual process model. What hasn’t been noticed so far is that vmPFC patients at the same time raise a worry about the other, more interesting part of the model. I have earlier argued that to the extent that “utilitarian” judgment is associated with controlled processing, that processing doesn’t reflect any kind of “utilitarian reasoning.” (For isn’t it just obvious that it’s better to save 5 than to save 1?) This argument actually generates a straightforward empirical prediction: that such reasoning (i.e. the psychological processes directly pushing us in the direction of pushing the fat man in Footbridge) can actually be fast and effortless. And indeed, so it is in vmPFC patients. A recent study has shown that in vmPFC patients, utilitarian judgments in personal dilemmas were associated both with weaker skin conductance responses, and with shorter reaction times, compared to healthy subjects (Moretto et al. 2009). That is, once you remove the opposing force of the deontological consideration against directly harming someone, the “utilitarian” solution becomes pretty intuitive.35 35 It is also worth noting that the area of the DLPFC that Greene et al. 2004 reported to be associated with utilitarian judgment is damaged in many of these VMPFC patients (Moll and de Oliviera 2007). Since other parts of the DLPFC are intact, this doesn’t itself show that they are incapable of controlled processing, or that they are not engaged in such processing when making utilitarian judgments. What this does strongly suggest is that, if the DLPFC activation reported in Greene et al. 2004 reflects the controlled processing healthy subjects engage in when they make utilitarian judgments in Footbridge, then this controlled processing couldn’t

Intuitive and Counterintuitive Morality  27 This finding may not seem so significant. After all, this evidence relates to rare and unusual brain damage. It tells us nothing about how ordinary, healthy individuals arrive at utilitarian judgments. But more recent evidence shows that this pattern of response is actually far from unusual. To begin with, several recent studies report greater rates of utilitarian judgment in individuals high on antisocial traits such as psychopathy (Glenn et al. 2010; Bartels and Pizarro 2011; Koenigs et al. 2012; Wiech et al. 2013; though see also Cima et al. 2010; Glenn et al. 2009). These findings have been overinterpreted by the press. The Economist, for example, reported one such study by announcing that “Goodness has nothing to do with it: Utilitarians are not nice people.”36 This is of course a blatantly fallacious inference. From the fact that some not so nice people are utilitarian, it hardly follows that all utilitarians are not nice people. This claim also relies on the terrible inference I mentioned earlier, from “utilitarian” responses to trolley-like dilemmas to general claims about utilitarian judgment. But some defenders of utilitarianism are happy to make that terrible inference when it associates utilitarian judgment with something more flattering! However, it would also be a mistake to assume that these results merely relate to a tiny, abnormal minority. “Proper” psychopaths are indeed a tiny minority, but some of the studies I  just cited actually examined ordinary individuals who happen to have psychopathic and similar antisocial traits—individuals who are a bit colder and harsher than the rest of us. And anyway these findings are just part of a much broader trend in recent research suggesting that utilitarian judgments in healthy individuals are often rooted in an atypically weak or even absent aversion to harming others. For example, higher rates of utilitarian judgment in normal, healthy individuals have been associated with lower rates of trait empathy (Crockett et al. 2010; Choe and Min 2011), as well as with higher levels of testosterone (Carney and Mason 2010), which often leads to reduced empathic concern (Hermans et al. 2006). Another study found that a disposition to utilitarian judgment was associated, not with greater cognitive ability, but with lower emotional intelligence, and specifically with a difficulty with reasoning thoughtfully about one’s emotions (Koven 2011). More directly, a strong “utilitarian” disposition was associated with reduced aversive reactivity to harming others, as indexed by peripheral vasoconstriction (Cushman et al. 2012), with reduced skin conductance response (Moretto et al. 2009), and, most importantly for us, with significantly lower response times (Greene et al. 2008). In other words, even in ordinary, healthy individuals, a strong disposition to utilitarian judgment in “personal” moral dilemmas is often due, not to increased critical reflection that has overcome deontological intuitions, but to the absence of these intuitions,

reflect a utilitarian cost–benefit analysis, since these patients can reach this utilitarian conclusion without engaging this brain area.

The Economist, 24 September 2011.

36

28  Guy Kahane and more generally, to reduced empathic concern. And when ordinary people are “colder” in this way, their utilitarian judgments also become pretty fast and effortless. Stated more generally, the conclusion we should draw is the following: Deliberative effort as function of deontological commitment. The more weight one gives to deontological constraints on harming, the more deliberative effort is needed to conclude that one should set them aside in a given moral context; the less seriously one takes such constraints, the more effortless and non-deliberative “utilitarian” judgments become.

In other words, the degree to which deliberative effort is involved in so-called “utilitarian” judgments really reflects the strength of one’s deontological commitments! Without such commitments, “utilitarian” judgments require little or no deliberation.

4.2.  Utilitarian Judgment: Hot or Cold Let us return now to the neuroimaging study we discussed in section 1. One of our findings was that counterintuitive moral judgments in general, as well as counterintuitive utilitarian judgments in particular, were perceived as more difficult, and were associated with increased activation in the anterior cingulate cortex, and specifically in its subgenual part, compared to contrary “intuitive” moral judgments (Kahane et al. 2012). This is an area of the brain that has been repeatedly implicated in empathic concern and guilt (Decety, Michalska, and Kinzler 2012; Green et al. 2010; Moll et al. 2006; Zahn et al. 2009a; Zahn et al. 2009b). This result is in line with the finding mentioned earlier that counterintuitive utilitarian judgments in healthy individuals are typically followed by stronger skin conductance response (Moretto et al. 2009), indicating an emotional reaction. That reaction seems to be one of guilt, or something close, given that individuals also typically report feeling guilty when making utilitarian judgments in most personal dilemmas (Choe and Min 2011). This tie between counterintuitive utilitarian judgments and emotion is in itself compatible with Greene’s model, since these emotional responses seem to follow such judgments rather than cause them. But as I have argued in section 3, what these results do strongly suggest is that most individuals take their deontological intuitions at face value, as representing genuine constraints on harming others that, in some unusual cases of “life and death” decisions, might nevertheless be outweighed by strong competing considerations. This is why, for most people, such decisions appear to be accompanied by guilt, or a similar aversive response to the violation of a moral norm. But in a separate study based on the same neuroimaging data, we found that this may not be true of all individuals (Wiech et al. 2013). To begin with, we found that higher rates of utilitarian judgment were associated with “psychoticism” (or “tough headedness”), a trait associated with lack of emotionality and empathic concern, hostility, aggression, and non-conformity to social norms (Eysenck 1976).37 Individuals 37 We did find that need for cognition was also associated with greater rates of utilitarian judgment, but, importantly, need for cognition and psychoticism were statistically independent (Wiech et al. 2013), suggesting they may be independent factors in utilitarian judgment. However, it is worth noting that we did not

Intuitive and Counterintuitive Morality  29 higher on psychoticism have been shown to perceive media violence as more comical and enjoyable, and show rapid habituation to violent material (Bruggemann and Barry 2002), and reduced aversion to killing enemies in a video game (Ravaja et al. 2008). Unsurprisingly, psychoticism is also strongly correlated with psychopathy (Hare 1981; Shine and Hobson 1997), a construct often seen as lying on a continuum with psychoticism (Corr 2010; Eysenck 1992) although, unlike psychopathy, psychoticism primarily aims to capture a general dimension of personality in the normal population. Interestingly, we found that psychoticism was at once positively correlated with the number of utilitarian judgments and negatively correlated with activation in the subgenual anterior cortex. Put together with the evidence cited here, these findings seem to suggest that although most individuals find utilitarian judgments aversive, individuals higher on psychoticism do not. We already know that vmPFC patients don’t feel guilt when they violate social norms (Krajbich et al. 2009), and this is of course also true of psychopaths. In other words, it seems that individuals who are strongly disposed to utilitarian solutions to “personal” dilemmas don’t just lack intuitions against directly harming others, but also do not find such harm morally problematic. They do not perceive it as a regrettable (if sometimes necessary) violation of an important moral constraint. These “colder” individuals appear to arrive at utilitarian judgments just by considering the consequences. They are not inhibited by deontological constraints against directly harming others, or need to engage in any deliberation to decide whether some significant benefit outweighs these constraints in a given context. Unsurprisingly, these individuals also tend to more consistently endorse utilitarian solutions to “personal” dilemmas compared to other individuals, who may endorse such solutions in a few cases but not in most others.38 The “colder” individuals consider only the consequences because they do not recognize deontological constraints against directly harming others, nor, of course, find it aversive to violate such constraints. In these respects, these individuals arrive at utilitarian judgment in ways that are more closely in line with something resembling a utilitarian decision procedure—these individuals take only consequences into account, and nothing else.39 find any association between need for cognition and increased activation in the DLPFC or any other marker of effortful cognition. 38 It is hardly surprising that a colder approach to moral problems, where various deontological constraints are suppressed or ignored, will tend to lead to more “utilitarian” conclusions. This is actually a familiar feature of decision-making in emergency situations such as triage. In ordinary medical contexts, doctors and nurses don’t just try to maximize the outcome as a utilitarian would—they will usually try to save the lives of even those who are gravely injured, and give some attention even to relatively minor injuries. But these deontological considerations will be brushed aside in situations of emergency, where there is an imperative to focus limited resources on those who will benefit the most. We can call such practices “utilitarian,” but these imperatives are really just part of commonsense morality. This point is overlooked by Greene and others because they conflate deontology and absolutism, the view that deontological constraints must never be broken. They thus fail to see that many judgments that they classify as “utilitarians” are really just expressions of commonsense, non-utilitarian morality. I am grateful to the editors for these points. 39 I described such individuals as “colder.” They are colder in the sense that they exhibit little empathic concern, and don’t find the idea of harming others emotionally aversive. But they needn’t be entirely

30  Guy Kahane Now it’s not generally true that people who make utilitarian judgments aren’t nice. Our study actually suggests that antisocial tendencies are the driving force behind utilitarian judgment only in some individuals (Wiech et al. 2013)—“utilitarian” judgment may have more than one source. But it does appear that many of those who are more strongly disposed to make utilitarian judgments, and who make such judgments in ways that more closely resemble a utilitarian decision procedure, really aren’t that nice. In fact, these individuals are disposed to utilitarian solutions because they aren’t so nice. The important thing, though, is that these utilitarian judgments are non-deliberative. These individuals may lack common emotions and intuitions, but it doesn’t follow from this that they are thinking harder when presented with these moral dilemmas. They just respond to them differently. One further lesson of these results is that we need to be more careful with the way we think about counterintuitive judgments. Claims of intuitiveness and counterintuitiveness are ambiguous. In common philosophical and everyday use, they are often, explicitly or implicitly, tied to a population. When someone says that something is counterintuitive, they are not merely reporting that it goes against their own intuitions. They are assuming that it goes against the intuitions of many or most others. Certain utilitarian claims are often said to be counterintuitive in exactly this sense. They are claimed to go against the strong intuitions of most people. That this is so in some cases is now supported by quite a bit of empirical evidence (see e.g. Cushman et al. 2006). But it doesn’t follow that when someone makes a utilitarian judgment that is counterintuitive in this sense, then this means that this person is going against their own intuitions—that they are using reason to override their first, intuitive response. Here we would be sliding to a person-relative sense of “intuitive” and “counterintuitive.” In fact, a claim might be both counterintuitive in the population-relative sense, yet also be intuitive for some individuals. And this seems to be the case for many utilitarian judgments. These judgments reflect unusual (or absent) intuitions, not the “rational” overcoming of common intuitions. Let me summarize this section. In section 3, I argued that when “utilitarian” judgments are driven by deliberation, this deliberation is non-utilitarian. In this section, I further argued that when individuals make “pure” utilitarian judgments, these judgments are non-deliberative! These individuals endorse counterintuitive conclusions, but these conclusions are counterintuitive only relative to the intuitions of the majority—for these individuals, these utilitarian conclusions are intuitive.40

unemotional. In fact, several studies have tied utilitarian judgment to feeling angry (Choe and Minn 2011; Ugazio et al. 2012; see also Plaisier and Konijn 2013 on how anger can increase tolerance of antisocial behaviour). Interestingly, Jonathan Baron, a self-avowed utilitarian, independently described moral anger as the dominant utilitarian emotion (see Baron 2011).

This point can be missed if we mistakenly identify intuition (and automaticity) with emotion.

40

Intuitive and Counterintuitive Morality  31

5.  For the Greater Good? In section 4, I offered a very different picture of the psychology of utilitarian judgment than that defended by Greene, and assumed by utilitarians such as Peter Singer. Utilitarianism is often presented as the reasoned and systematic generalization of natural human empathic concern (Hare 1981; Singer 1979). Yet it ironically turns out that so-called “utilitarian” judgments are often driven, not by rational reflection or “generalized benevolence” (Smart 1961), but by a deficit in empathic concern and an indifference to harming others. This result might seem puzzling. In section 4, I have highlighted the tie between a strong tendency to utilitarian judgment and reduced aversion to harming others. And I claimed that if we want to study the psychological processes associated with a form of moral thinking in lay people that at least resembles a kind of utilitarian decision procedure, we should look at this category of individuals, who appear to just consider the consequences of their actions, and to not be inhibited (normatively and psychologically) by the nature of the violent act that they are asked to contemplate doing. But it might be objected that such individuals aren’t likely to be following a genuine utilitarian outlook. Psychopaths aren’t likely to be concerned with the greater good. They are likely to be concerned only with their own good. This is correct. In fact, in a further set of studies, we set out precisely to investigate whether there really is any relationship between so-called “utilitarian” judgments in trolley-like dilemmas and any kind of genuine concern for the greater good (Kahane et al. under review). We found that greater rates of “utilitarian” judgment were positively correlated not only with psychopathic tendencies and reduced empathic concern, but also with explicit endorsement of rational egoism—while being negatively correlated (albeit only marginally) with whether individuals identify with the whole of humanity. In addition, we found that rates of “utilitarian” judgment were also associated with seeing uncontroversially immoral acts in a business context (e.g. stealing money from one’s company) as less wrong. We also found that although individuals higher on psychopathy do appear to show a “utilitarian” tendency in their responses to “personal” moral dilemmas, their responses to such dilemmas are in fact highly sensitive to considerations of self-interest. Thus, they show an even stronger utilitarian tendency when the “utilitarian” act would also benefit them—i.e. they would be among the five people saved when someone is sacrificed “for the greater good”—while if they are given the choice whether to sacrifice themselves or someone else to save five others, they overwhelmingly prefer to sacrifice another person (individuals low on psycho­ pathy show exactly the reverse pattern of response). Utilitarianism tells us to selflessly promote the greater good—to transcend our narrow personal concern and care about the good of all human beings, even all sentient beings, as if from the point of view of the universe. But those who tend to most strongly endorse “utilitarian” solutions to “personal” dilemmas turn out to largely care about themselves . . . In one obvious sense, utilitarianism and rational egoism are

32  Guy Kahane diametrically opposed. But on reflection, this convergence shouldn’t be so surprising. Both utilitarianism and rational egoism are teleological views, concerned with maximizing the good (in one case, the greater good, in the other, one’s own good). Both often come across as rather tough-headed, unsentimental, and calculating. And both treat common moral norms, intuitions, and sentiments as no more than useful instruments, to be followed only when this leads to better consequences. Is it really so surprising that when egoists are asked for their view about a moral dilemma, and given either a deontological option or a more calculating “utilitarian” one, they would tend to prefer the latter?41 The results I  have just described further highlight the antisocial dimension in so-called “utilitarian” judgment. You might still think that, although this is an unpleasant association, we might be able to get around it. We just need to bracket this antisocial element, and this will reveal those judgments that are truly utilitarian—those responses that are really driven by greater deliberation, and by genuine concern for the good.42 But don’t get your hopes up. In the same series of studies, we also investigated the relation between “utilitarian” judgment and a range of views and attitudes drawn directly from the work of Peter Singer—if someone tends to think that, for example, you should push the stranger in Footbridge, would they also think that it’s wrong to eat meat, or that we have obligations to people in need in developing countries? Would they think that failing to donate money to save the lives of children in poor countries is wrong in the way it is wrong to refuse to step into a pool to save a drowning child? The answer is no. We found that there is no relation between a “utilitarian” tendency in personal dilemmas and any of these marks of a genuine concern for the greater good. It’s not just that individuals with such a “utilitarian” tendency don’t think that, say, eating meat is wrong. They don’t even find eating meat, or failing to help children in poor countries, a bit more wrong than those who don’t have such a tendency. And importantly, this lack of relation remained even when we controlled for the antisocial dimension in so-called “utilitarian” judgment.43 41 In line with these results, a recent study found that individuals who identify as political libertarians turn out to have lower empathic concern, identify least with humanity as a whole, and don’t place as much moral weight on considerations of harm and the welfare of others (Iyer et al. 2012). Needless to say, the moral views of such libertarians are clearly very far from utilitarianism—libertarianism is in certain respects a very extreme deontological view! Yet it still turns out that libertarians are also more “utilitarian” in personal moral dilemmas. . . 42 Conway and Gawronski (2013) suggest an interesting method to try to distinguish positive “utilitarian” tendencies from mere reduced aversion to harm. Their approach, however, doesn’t address any of the issues I have been pressing in this chapter. Worse, they use a number of highly inappropriate dilemmas, thereby serious compromising their results. Their analysis strangely assumes, for example, that utilitarians would be opposed to abortion in a young single woman who has no income and is clearly not ready to have children. 43 It’s important to emphasize that the complaint isn’t that people who make “utilitarian” judgments in trolley-like moral dilemmas don’t explicitly endorse and follow a utilitarian theory, or even that they aren’t inarticulate clones of Peter Singer. We were not expecting these people to think that we should make the extremely demanding sacrifices required by utilitarianism. But in what sense are they expressing a genuine utilitarian disposition if they don’t even have a slightly stronger tendency to more impartial altruism compared to others? You might still insist that judging that we should push the stranger in Footbridge to save five

Intuitive and Counterintuitive Morality  33 But these results don’t threaten the argument I have developed in the previous section. They only strengthen it. As I have pointed out repeatedly, it is a mistake, and a common mistake, to treat so-called “utilitarian” responses to trolley-like dilemmas as any kind of general measure of a genuine utilitarian outlook. To repeat: This type of dilemma reflects merely one particular point where utilitarianism clashes with commonsense morality (Kahane and Shackel 2010). Worse, such dilemmas focus only on a narrow and fairly negative aspect of utilitarianism, its willingness to cause severe harm to individuals when this leads to a better outcome. They entirely ignore the more positive, altruistic, and impartial core of utilitarianism. The thrust of my argument in this chapter is that the two main driving forces of apparently “utilitarian” judgment in such dilemmas turn out to be either a form of deliberation that is explicitly non-utilitarian, or a form of moral response that does, in this context, at least resemble a utilitarian decision procedure (i.e. a concern for nothing besides the consequences of actions) but actually reflects a broadly selfish and antisocial tendency. Individuals who are low on empathic concern are not concerned about the greater good, but they exhibit a more unsentimental, hard-headed, and calculating approach to moral questions, an approach which manifests itself in a more utilitarian outlook in the unusual context of “personal” dilemmas, even if not in real life.44 Utilitarianism is a rather peculiar moral view, with a recent history, and which never spread beyond a small minority. It is not a view that is common in the general population, so it is not surprising that we don’t find much evidence for it in the lay population. If we want to study the psychological correlates of genuine utilitarianism, we should study actual utilitarians—and I think there are few of these in the subject pool of most experiments. When Greene tells us that utilitarian judgment is uniquely based in deliberation and reasoning, this can seem like good news to utilitarians; the emerging link between utilitarian judgment and antisocial tendencies is less flattering. But true utilitarians should neither cheer the relationship between “utilitarian” judgments and “rational” deliberation, nor feel discomfort about the sinister associations between such judgments and psychopathy—for, contrary to appearances, the so-called “utilitarian” responses of non-philosophers to trolley-like dilemmas may, in the end, have rather little to do with utilitarianism.

is appropriately called a “utilitarian” judgment. But judging that we should help the elderly lady across the street is also a “utilitarian” judgment in that sense. Yet studying the neural correlates of the latter judgment would surely tell us nothing at all about utilitarianism qua ethical outlook. 44 Some philosophers with strong utilitarian inclinations have actually confessed to me that they see themselves as low empathy, unsentimental types. It is an open empirical question whether most proper utilitarians start out as high empathy types that, through a process of reflection, come to reject various intuitions, or whether they are low empathy types who are attracted, at an abstract level, to the systematic, tough-headed nature of utilitarianism, without feeling much natural concern for the welfare of others.

34  Guy Kahane

6. Conclusion According to Greene’s dual process model of moral judgment, our moral psychology is divided between opposing forces: an emotional, intuitive deontological half, and a rational, deliberative utilitarian half. In this chapter, I have argued that this provocative picture of our moral psychology is mistaken. I first argued that there is in fact nothing distinctively utilitarian about deliberation to “utilitarian” conclusions: They merely reflect generic deliberation to a counterintuitive conclusion. I then went further and argued that such deliberation to “utilitarian” conclusions in trolley-like dilemmas is in fact non-utilitarian in character. It involves, not “utilitarian reasoning,” but the weighing of opposing moral principles, a process that itself relies on intuition, at a second order level. I next argued that when individuals do arrive at counterintuitive conclusions in a way that might resemble utilitarian reasoning, these judgments actually turn out to be non-deliberative—and to reflect a strong antisocial, self-centered tendency. For these individuals, utilitarian solutions to trolley-like dilemmas don’t require deliberation, because for them such solutions are actually intuitive. Finally, I presented evidence that strongly suggests that there is actually little connection between so-called “utilitarian” responses to trolley-like dilemmas and genuine concern for the greater good, even when we set aside the antisocial dimension that drives some of these responses. This shouldn’t be surprising: that someone is willing to dismiss (or otherwise discount) a single specific deontological constraint against harming others in highly unusual contexts hardly shows that this person has any special concern for the greater good in any other context, let alone that they have the slightest inclination to make (or even just endorse) the demanding sacrifices required by a genuine utilitarian outlook. Over the past decade or so, an ever growing empirical literature has been devoted to studying so-called “utilitarian” judgment in trolley-like dilemmas. The argument of this chapter is that much of this literature has been based on a series of mistaken assumptions—mistaken assumptions about the nature of the relation between deliberation and intuition, about what “utilitarian reasoning” might involve, and what moral deliberation is actually like, about the different senses in which a judgment might be intuitive or counterintuitive, about what it is to genuinely face a moral dilemma, and finally, about what it even means to make a “utilitarian” judgment. We should welcome the use of philosophical concepts, distinctions, and examples in the cognitive science of ethics, and it is only natural that when such concepts and distinctions are adapted to an empirical setting, they may change some of their meaning, or lose some of their philosophical precision. We should even welcome bold attempts to draw ambitious normative conclusions from such empirical research. But the distinctions that philosophers draw in their armchairs often have important psychological implications, and when these distinctions are overlooked or ignored, empirical research can be led into error. Perhaps worse, when empirical research assumes problematic or even simply mistaken understandings of basic philosophical notions, it is led to overlook important possibilities, and to ignore key questions.

Intuitive and Counterintuitive Morality  35 Dual process models may be the height of fashion at the moment, but trying to divide everything in the mind into “automatic” or “controlled” isn’t any kind of advance. Nor is it useful to crudely oppose intuition and deliberation, or emotion and reason. We already have a richer, and more precise, vocabulary to describe the different ways in which we form our moral judgments. Moral deliberation often draws on intuition, and it isn’t always, or even often, a kind of inference. In fact, moral deliberation not only often takes its starting point from intuition, it sometimes actually requires intuition to reach its conclusion—yes, there are intuitions (and for that matter emotions) that only make sense in the context of deliberation. Current empirical research hasn’t even begun to explore this territory.45

Acknowledgments This chapter draws on prior work, including Kahane and Shackel 2010; Kahane et al. 2012; Wiech et al. 2012, and especially the argument outlined in Kahane 2012—the present chapter reprises and develops that earlier argument, and then takes it several steps forward. I would like to thank my collaborators on some of these earlier papers, and especially Nick Shackel and Katja Wiech. I am also extremely grateful to the editors for excellent comments on an earlier draft of this chapter. I presented much of the material in this chapter at the workshop on Moral Psychology and Human Agency at the University of Michigan, as well as at events in Bielefeld and Munich. I am also grateful to the audiences in these events for many useful suggestions. Work on this chapter was supported by a University Award from the Wellcome Trust (WT087208MF) and by a grant from the Volkswagen Foundation.

References Baron, J. 2011. Utilitarian emotions:  Suggestions from introspection. Emotion Review, 3(3), 286–7. Baron, J., Gürçay, B., Moore, A. B., and Starcke, K. 2012. Use of a Rasch model to predict response times to utilitarian moral dilemmas. Synthese, 189(1), 107–17. Bartels, D. 2008. Principled moral sentiment and the flexibility of moral judgment and decision making. Cognition, 108, 381–417. Bartels, D. M., and Pizarro, D. A. 2011. The mismeasure of morals: Antisocial personality traits predict utilitarian responses to moral dilemmas. Cognition, 121(1), 154–61. 45 Once we start to think about deliberation in this richer way, numerous interesting empirical questions come into view: How are different moral (and other practical) reasons represented, aggregated, and integrated in the brain? How does our cognitive system keep tab both of different kinds of reasons (e.g. relating to beneficence, fidelity, justice, etc.) and of whether they support or oppose some particular option in a choice situation? (After all reasons of different types can simultaneously support and oppose a given option.) Does the brain represent the number of reasons in favor of some option independently of their strength? (A single weighty reason can outweigh numerous weak ones.) How does the brain represent and process the distinction between pro tanto reasons and all-things-considered ones (or “oughts”)? Is the outcome of deliberation merely a function of the context-invariant strength of each reason, or is a function of the holistic constellation of reasons in each particular situation, as Ross and Dancy argue?

36  Guy Kahane Berker, S. 2009. The normative insignificance of neuroscience. Philosophy and Public Affairs, 37, 293–329. Bruggemann, J. M., and Barry, R. J. 2002. Eysenck’s P as a modulator of affective and electrodermal responses to violent and comic film. Personality and Individual Differences, 32, 1029–48. Carney, D. R., and Mason, M. F. 2010. Decision making and testosterone: When the ends justify the means. Journal of Experimental Social Psychology, 46, 668–71. Choe, S. Y., and Min, K.-H. 2011. Who makes utilitarian judgments? The influences of emotions on utilitarian judgments. Judgment and Decision Making, 6, 580–92. Ciaramelli, E., Muccioli, M., Làdavas, E., and di Pellegrino, G. 2007. Selective deficit in personal moral judgment following damage to ventromedial prefrontal cortex. Social Cognitive and Affective Neuroscience, 2, 84–92. Cima, M., Tonnaer, F., and Hauser, M. D. 2010. Psychopaths know right from wrong but don’t care. Social Cognitive and Affective Neuroscience, 5, 59–67. Conway, P., and Gawronski, B. 2013. Deontological and utilitarian inclinations in moral decision making: A process dissociation approach. Journal of Personality and Social Psychology, 104(2), 216–35. Corr, P.  J. 2010. The psychoticism–psychopathy continuum:  A  neuropsychological model of core deficits. Personality and Individual Differences, 48, 695–703. Crockett, M. J., Clark, L., Hauser, M. D., and Robbins, T. W. 2010. Serotonin selectively influences moral judgment and behavior through effects on harm aversion. Proceedings of the National Academy of Sciences, 107, 17433–8. Cushman, F., Gray, K., Gaffey, A., and Mendes, W. B. 2012. Simulating murder: The aversion to harmful action. Emotion, 12(1), 2–7. Cushman, F., Young, L., and Greene, J. D. 2010. Our multi-system moral psychology: Towards a consensus view. In J. M. Doris (ed.), Oxford Handbook of Moral Psychology. Oxford: Oxford University Press. Cushman, F., Young, L., and Hauser, M. 2006. The role of conscious reasoning and intuition in moral judgment: Testing three principles of harm. Psychological Science, 17, 1082–9. Decety, J., Michalska, K. J., and Kinzler, K. D. 2012. The contribution of emotion and cognition to moral sensitivity: A neurodevelopmental study. Cerebral Cortex, 22(1), 209–20. Etkin, A., Egner, T., Peraza, D. M., Kandel, E. R., Hirsch, J. 2006. Resolving emotional conflict: A role for the rostral anterior cingulate cortex in modulating activity in the amygdala. Neuron, 51, 871–82. Evans, J. 2007. Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255–78. Eysenck, H. J. 1976. Psychoticism as a dimension of personality. London: Hodder & Stoughton. Eysenck, H. J. 1992. The definition and measurement of psychoticism. Personality and Individual Differences, 13, 757–85. Foot, P. 1978. The problem of abortion and the doctrine of double effect. In Foot, Virtues and Vices. Oxford: Blackwell. Frederick, S. 2005. Cognitive reflection and decision making. Journal of Economic Perspectives, 19(4), 25–42. Hermans, E. J., Putman, P., and Van Honk, J. 2006. Testosterone administration reduces empathetic behavior: A facial mimicry study. Psychoneuroendocrinology, 37, 859–866.

Intuitive and Counterintuitive Morality  37 Glenn, A. L., Koleva, S., Iyer, R., Graham, J., and Ditto, P. H. 2010. Moral identity in psychopathy. Judgment and Decision Making, 5, 497–505. Glenn, A. L., Raine A., and Schug R. A. 2009. The neural correlates of moral decision-making in psychopathy. Molecular Psychiatry, 14, 5–9. Green, S., Lambon Ralph, M. A., Moll, J., Stamatakis, E. A., and Zhan, R. 2010. Selective functional integration between anterior temporal and distinct fronto-mesolimbic regions during guilt and indignation. Neuroimage, 54, 1720–6. Greene, J.  D. 2008. The secret joke of Kant’s soul. In W. Sinnott-Armstrong (ed.), Moral Psychology: The Neuroscience of Morality, pp. 35–79. Cambridge, MA: MIT Press. Greene, J. D. 2009. Dual-process morality and the personal/impersonal distinction: A reply to McGuire, Langdon, Coltheart, and Mackenzie. Journal of Experimental Social Psychology, 45, 581–4. Greene, J. D., and Haidt, J. 2002. How and where does moral judgment work? Trends in Cognitive Science, 6, 517–23. Greene, J. D., Morelli, S. A., Lowenberg, K., Nystrom, L. E., and Cohen, J. D. 2008. Cognitive load selectively interferes with utilitarian moral judgment. Cognition, 107, 1144–54. Greene, J. D., Nystrom, L. E., Engell, A. D., Darley, J. M., and Cohen, J. D. 2004. The neural bases of cognitive conflict and control in moral judgment. Neuron, 44, 389–400. Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M., and Cohen, J. D. 2001. An fMRI investigation of emotional engagement in moral judgment. Science, 293, 2105–7. Haidt, J. 2001. The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review, 108, 814–34. Haidt, J. 2007. The new synthesis in moral psychology. Science, 316: 998–1002. Haidt, J. 2012. The Righteous Mind. New York: Pantheon. Hare, R. M. 1981. Moral Thinking. Oxford: Oxford University Press. Huebner, B., Dwyer, S., and Hauser, M. 2009. The role of emotion in moral psychology. Trends in Cognitive Sciences, 13(1), 1–6. Iyer, R., Koleva, S., Graham, J., Ditto, P., Haidt, J. 2012. Understanding libertarian morality: The psychological dispositions of self-identified libertarians. PLoS ONE, 7(8): e42366. doi:10.1371/ journal.pone.0042366 Kahane, G. 2012. On the wrong track:  Process and content in moral psychology. Mind and Language, 25(5), 519–45. Kahane G., Everett, J., Earp, B., Farias, M., and Savulescu, J. Under review. Kahane, G., and Shackel, N. 2008. Do abnormal responses show utilitarian bias? Nature, 452, 7185. Kahane, G., and Shackel, N. 2010. Methodological issues in the neuroscience of moral judgment. Mind and Language, 25, 561–82. Kahane, G., Wiech, K., Shackel, N., Farias, M., Savulescu, J., and Tracey, I. 2012. The neural basis of intuitive and counterintuitive moral judgement. Social, Cognitive and Affective Neuroscience, 7, 393–402. Kant, I. 1797/1966. On a supposed right to lie from philanthropy. In M. J. Gregor (ed.), Practical Philosophy, pp. 611–15. Cambridge: Cambridge University Press. Koenigs, M., Kruepke, M., Zeier, J., and Newman, J. P. 2012. Utilitarian moral judgment in psychopathy. Social, Cognitive and Affective Neuroscience, 7(6), 708–14. Koenigs, M., Young, L., Adolphs, R., Tranel, D., Cushman, F., Hauser, M., and Damasio, A. 2007. Damage to the prefrontal cortex increases utilitarian moral judgements. Nature, 446, 908–11.

38  Guy Kahane Koven, N. S. 2011. Specificity of meta-emotion effects on moral decision-making. Emotion, 11(5), 1255–61. Krajbich, I., Adolphs, R., Tranel, D., Denburg, N. L., and Camerer, C. F. 2009. Economic games quantify diminished sense of guilt in patients with damage to the prefrontal cortex. Journal of Neuroscience, 29, 2188–92. Kurzban, R., DeScioli, P., and Fein, D. 2012. Hamilton vs. Kant: Pitting adaptations for altruism against adaptations for moral judgment. Evolution and Human Behavior, 33, 323–33. de Lazari-Radek, K., and Singer, P. 2012. The objectivity of ethics and the unity of practical reason. Ethics, 123, 9–31. McGuire, J., Langdon, R., Coltheart, M., and Mackenzie, C. 2009. A reanalysis of the personal/impersonal distinction in moral psychology research. Journal of Experimental Social Psychology, 45(3), 577–80. Mendez, M. F., Anderson, E., and Shapira, J. S. 2005. An investigation of moral judgement in frontotemporal dementia. Cognitive and Behavioral Neurology, 18, 193–7. Moll, J., Krueger, F., Zahn, R., Pardini, M., de Oliveira-Souza, R., and Grafman, J. 2006. Human fronto–mesolimbic networks guide decisions about charitable donation. Proceedings of the National Academy of Sciences, 103(42), 15623–8. Moll, J., and de Oliveira-Souza, R. 2007. Moral judgments, emotions and the utilitarian brain. Trends in Cognitive Sciences, 11, 319–21. Moore, A. B., Clark, B. A., and Kane, M. J. 2008. Who shalt not kill? Individual differences in working memory capacity, executive control, and moral judgment. Psychological Science, 19, 549–57. Moretto, G., Làdavas, E., Mattioli, F., and di Pellegrino, G. 2009. A psychophysiological investigation of moral judgment after ventromedial prefrontal damage. Journal of Cognitive Neuroscience, 22(8), 1888–99. Nichols, S., and Mallon, R. 2005. Moral dilemmas and moral rules. Cognition, 100(3), 530–42. Paxton, J. M., Bruni, T., and Greene, J. D. 2013. Are “counter-intuitive” deontological judgments really counter-intuitive? An empirical reply to Kahane et  al. (2012). Social Cognitive and Affective Neuroscience, nst 102. Paxton, J. M., and Greene, J. D. 2010. Moral reasoning: Hints and allegations. Topics in Cognitive Science, 2, 511–27. Paxton, J. M., Ungar, L., and Greene, J. D. 2012. Reflection and reasoning in moral judgment. Cognitive Science, 36, 163–77. Plaisier, X. S., and Konijn, E. A. 2013. Rejected by peers—Attracted to antisocial media content:  Rejection-based anger impairs moral judgment among adolescents. Developmental Psychology, 49(6), 1165. Rand, D.  G., Greene, J.  D., and Nowak, M.  A. 2012. Spontaneous giving and calculated greed. Nature, 489(7416), 427–30. Ravaja, N., Turpeinen, M., Saari, T., Puttonen, S., and Keltikangas-Jarvinen, L. 2008. The psychophysiology of James Bond:  Phasic emotional responses to violent video game events. Emotion, 8, 114–20. Ross, W. D. 1930/2002. The Right and the Good. Oxford: Oxford University Press. Shenhav, A., and Greene, J. D. 2010. Moral judgments recruit domain-general valuation mechanisms to integrate representations of probability and magnitude. Neuron, 76, 667–77.

Intuitive and Counterintuitive Morality  39 Shine, J., and Hobson, J. 1997. Construct validity of the Hare Psychopathy Checklist, Revised, on a UK prison population. Journal of Forensic Psychiatry and Psychology, 8, 546–61. Singer, P. 1979. Practical ethics. Cambridge, New York: Cambridge University Press. Singer, P. 2005. Ethics and intuitions. Journal of Ethics, 9, 331–52. Smart, J. J. C. 1961. An Outline of a System of Utilitarian Ethics. Melbourne: Melbourne University Press. Stanovich, K. 1999. Who is Rational? Studies of Individual Differences in Reasoning. Mahwah, NJ: Lawrence Erlbaum. Suter, R. S., and R. Hertwig. 2011. Time and moral judgment. Cognition, 119(3), 454–8. Terbeck, S., Kahane, G., McTavish, S., Levi, N., Savulescu, J., Cowen, P., Hewstone, M. 2013. Beta-adrenergic blockade reduces utilitarian judgment. Biological Psychology, 92(2), 323–8. Thomson, J. J. 1976. Killing, letting die, and the trolley problem. The Monist, 59(2), 204–17. Ugazio, G., Lamm, C., and Singer, T. 2012. The role of emotions for moral judgments depends on the type of emotion and moral scenario. Emotion, 12(3), 579. Unger, P. 1996. Living High and Letting Die: Our Illusion of Innocence. Oxford: Oxford University Press. Wiech, K., Kahane, G., Shackel, N., Farias, M., Savulescu, J., and Tracey, I. 2013. Cold or calculating? Reduced activity in the subgenual cingulate reflects decreased aversion to harming in counterintuitive utilitarian judgment. Cognition, 126, 364–72. Williams, B. A. O. 1965. Ethical consistency. Proceedings of the Aristotelian Society (Supplement), 39, 103–24. Williams, B. A. O. 1973. A critique of utilitarianism. In J. J. C. Smart and B. A. O Williams (eds.), Utilitarianism: For and Against. Cambridge: Cambridge University Press. Youssef, F.  F., Dookeeram, K., Basdeo, V., Francis, E., Doman, M., Mamed, D., Maloo, S., Degannes, J., Dobo, L., Ditshotlo, P., and Legall, G. 2012. Stress alters personal moral decision making. Psychoneuroendocrinology, 37, 491–8. Zahn, R., Moll, J., Paiva, M., Garrido, G., Krueger, F., Huey, E. D., and Grafman, J. 2009a. The neural basis of human social values:  Evidence from functional MRI. Cerebral Cortex, 19, 276–83. Zahn, R., de Oliveira-Souza, R., Bramati, I., Garrido, G., and Moll, J. 2009b. Subgenual cingulate activity reflects individual differences in empathic concern. Neuroscience Letters, 457(2), 107–10.

3 Moral Psychology as Accountability Brendan Dill and Stephen Darwall

1. Introduction When moral psychology exploded a decade ago with groundbreaking research, there was considerable excitement about the potential fruits of collaboration between moral philosophers and moral psychologists. However, this enthusiasm soon gave way to controversy about whether either field was, or even could be, relevant to the other (e.g., Greene 2007; Berker 2009). After all, it seems at first glance that the primary question researched by moral psychologists—how people form judgments about what is morally right and wrong—is independent from the parallel question investigated by moral philosophers—what is in fact morally right and wrong, and why. Once we transcend the narrow bounds of quandary ethics and “trolleyology,” however, a broader look at the fields of moral psychology and moral philosophy reveals several common interests. Moral philosophers strive not only to determine what actions are morally right and wrong, but also to understand our moral concepts, practices, and psychology. They ask what it means to be morally right, wrong, or obligatory: What distinguishes moral principles from other norms of action, such as those of instrumental rationality, prudence, excellence, or etiquette (Anscombe 1958; Williams 1985; Gibbard 1990; Annas 1995)? Moral psychologists pursue this very question in research on the distinction between moral and conventional rules (Turiel 1983; Nichols 2002; Kelly et al. 2007; Royzman, Leeman, and Baron 2009) and in attempts to define the moral domain (e.g., Haidt and Kesebir 2010). Moral psychologists also research the question of what motivates moral behavior (e.g., Batson 2008), a question that philosophers have been debating from Plato’s story of Gyges’ ring (Plato c.380 BC/1992) to the present day (Nagel 1970; Foot 1972; Stocker 1976; Herman 1981; Gauthier 1986; Korsgaard 1986; Smith 1994; Shafer-Landau 2003; Rosati 2006). And right in step with the “affect revolution” in moral psychology (Haidt 2003), there has been burgeoning philosophical interest in the nature and significance of the moral emotions (Strawson 1968; Taylor 1985; Watson 1987; Wallace 1994; D’Arms and Jacobson 2000; Velleman

Moral Psychology as Accountability  41 2001; Hieronymi 2004; Sher 2006; Scanlon 2008; Hurley and Macnamara 2010; Bagnoli 2011; Bell 2013; Coates and Tognazzini 2013). These three areas of common interest—moral motivation, the moral emotions, and the distinctiveness of morality—present an underappreciated opportunity for fruitful collaboration between moral philosophers and moral psychologists. As a step in this direction, this chapter argues that a recent philosophical proposal regarding the nature of morality provides significant insights into moral psychology. The philosophical framework we present makes substantive psychological predictions which are well confirmed by the empirical literature. Our central claim is that the psychology of morality, especially that of the moral motives and emotions, is best understood within a conception of morality—morality as accountability—that one of us has defended (Darwall 2006; 2013a; 2013b). According to morality as accountability, moral right and wrong involve accountability conceptually.1 In taking an action to be morally wrong, for example, we take it to be something we are justifiably held accountable for doing through distinctive emotions and attitudes that P. F. Strawson called “reactive attitudes,” such as condemnation or indignation when the agent is someone else, or guilt when it is we ourselves (Strawson 1968). Strawson argued convincingly that reactive attitudes like moral blame involve a distinctive way of regarding people, whether others or ourselves, that holds them accountable for complying with demands we take to be legitimate. Darwall (2006) explores the distinctively interpersonal or “second-personal” structure of these attitudes and argues that the central moral concepts of obligation or duty, right, and wrong are all essentially tied to second-personal attitudes and practices. The philosophical theory of morality as accountability can be seen as a response to the question of what it is for an action to be morally right or wrong. Questions of the moral rightness or wrongness of an action must be carefully distinguished from other normative, ethical, and even moral questions. For example, the question of what is the most desirable life for a human being or what makes for human well-being is clearly a normative issue. But even if a good human life were to include morally right conduct as an essential element, the normative question of what makes a human life desirable is not itself a question of moral right and wrong. Neither is the question of what is estimable, or worthy of emulation, pride, and praise a moral issue in this sense. Many estimable accomplishments and traits are distinctively moral, of course, but many have relatively little to do with morality—artistic achievements, for example. And even if

1 We are not the first to emphasize the connection between morality and accountability. In psychology, this connection has been extensively explored in the research of Philip Tetlock (e.g., Tetlock 2002), as well as in Jonathan Haidt’s recent book (Haidt 2012). In philosophy, this connection has been previously emphasized in the work of Allan Gibbard (1990), Gary Watson (1987, 1996), and R. Jay Wallace (1994), among others. Unfortunately, we do not have the space to survey this important work here. Our aim is to add to this literature by tying these philosophical and psychological threads together, specifically by applying the philosophical account offered in Darwall (2006) to the psychological question of the contents of the moral conscience and condemnation motives. Thanks to Jonathan Haidt for pressing us to clarify this point.

42  Brendan Dill and Stephen Darwall moral virtue and conduct were the only thing worthy of admiration, meriting admiration or contempt would nonetheless remain a different thing from being morally obligatory or prohibited. Setting to one side these other morally relevant concepts and questions, let us focus on the “deontic” concepts of moral obligation or duty, right, and wrong. These concepts are interdefinable. If something is a moral obligation or duty, it follows that it would be morally wrong not to do it, and conversely. When people say that something would be “morally right,” they can mean one of two things: either that it is morally obligatory or that it is not wrong, hence not obligatory not to do. We are interested in what it is for an action to be morally obligatory. Since one way of referring to an action as a moral duty or obligation is to say that it is something one “morally ought” to do, and since a popular philosophical approach to “oughts” in general is in terms of normative reasons favoring something, a natural thought is that what it is for an action to be a moral duty is for there to be most moral reason to do it. Following a line of argument similar to G. E. Moore’s famous “open question argument,” however, we can see why, however plausible this may seem on theoretical grounds, it cannot be correct (Moore 1903/1993). It seems clear that two people can be completely agreed that the balance of moral reasons most favors an action, but nonetheless coherently disagree about whether the action is morally required or obligatory, that is, whether it would be morally wrong to fail to perform it. For example, an act utilitarian might hold that it would be wrong not to perform a certain action that, though it would require a massive sacrifice for the agent (say, significant harm and the risk of death), would produce greater good overall than any other act available to her in the circumstances. A critic might agree with the act utilitarian that the act would be the morally best thing the agent could do, but nonetheless hold that morality does not require an agent in such a circumstance to produce such an outcome at such a personal cost. Such a sacrifice, the critic might hold, would be an unreasonable one for the agent to be asked to bear to produce the outcome and, consequently, that act utilitarianism “demands too much” to be a plausible theory of moral duty, even if it were the correct theory of moral reasons. There can be, the critic holds, supererogatory acts that are above and beyond the call of moral duty, whereas the act utilitarian holds there to be no such category, thinking that our moral duty is always to do whatever will produce the best consequences overall.2 Regardless of who, act utilitarian or critic, would be right about this, it seems obvious that the two could coherently disagree about this question. But if that is so, then our interlocutors cannot both mean by “morally obligatory” being what there is most moral reason to do, since they could be agreed that the self-sacrificial, optimific act is what there is most moral reason to do in the situation but disagreed about whether this act is morally obligatory, or wrong not to do.



Scheffler (1982) discusses this debate at length, ultimately siding with our critic.

2

Moral Psychology as Accountability  43 Darwall (2006) argues that the element that is lacking in the idea of what morality favors but essential to the deontic ideas of moral duty, obligation, or requirement, and moral prohibition or wrong, is that of moral demands with which we are accountable for complying. It is no part whatsoever of the idea that there is good reason, even good or best moral reason, to do something that the agent is in any way answerable to others for failing to do it and justifiably held answerable with reactive attitudes such as moral blame. This is, however, essential to the deontic ideas of moral obligation, right, and wrong. It is a conceptual truth that conduct is morally wrong if, and only if, the agent would justifiably be regarded with the attitude of moral blame by any agent (including himself), were he to undertake it without excuse. What our act utilitarian and critic were disagreeing about is whether the agent in question would be blameworthy if she were to fail to undertake the self-sacrificial but morally optimific action without excuse. We turn now to the notion of accountability and to the role and character of reactive attitudes like moral blame in holding people accountable. A central theme of Strawson’s (and of Darwall 2006) is that reactive attitudes like moral blame involve a way of regarding someone that is distinctively “inter-personal” or “second personal.”3 Reactive attitudes differ from other critical attitudes, like contempt or disdain, in implicitly making a (putatively) legitimate demand of their objects and holding their objects accountable for complying with it (Strawson 1968, 85; see also Watson 1987 and Wallace 1994; for a dissenting view, see Macnamara 2013). Critical attitudes like contempt and disdain also apply an implicit standard, against which their objects are found wanting, but they do not implicitly summon their objects to hold themselves accountable for complying with this standard in the same way that reactive attitudes do. Unlike these other critical attitudes, moral blame comes with an implicit RSVP, an implicit demand for accountability and acknowledgment of the legitimacy of this demand. Whereas shame internalizes contempt, guilt reciprocates blame by acknowledging the legitimacy of its implicit demand, and it is itself a form of holding oneself accountable for complying with this demand. It is through this analysis of the moral emotions that our philosophical proposal makes direct contact with moral psychology. Contained in the second-personal analyses of guilt and blame are specific psychological hypotheses about the cognitive presuppositions and motivational consequences of these emotions. We shall now turn to the psychological theses that we will defend throughout the remainder of this chapter, explaining how they follow from the philosophical framework laid out above. Following a model of emotion that is popular among both psychologists (Baumeister et al. 2007) and philosophers (Hurley and Macnamara 2010), we take complex, conscious emotions such as guilt and blame to have cognitive and motivational components in addition to their basic phenomenological and physiological “feel.” The cognitive component of an emotion is the state of affairs the emotion represents as 3 “Inter-personal” is Strawson’s term. Darwall (2006) uses “second-personal” to emphasize that the logical structure of these attitudes mimics the grammatical second person: They have an implicit addressee.

44  Brendan Dill and Stephen Darwall obtaining. The motivational component of an emotion is the goal or motive that the emotion activates. A  clear example of both these components is provided by fear, which portrays its object as dangerous (cognitive component) and motivates the subject to avoid that danger (motivational component). The second-personal analyses of guilt and blame imply hypotheses about the cognitive and motivational components of these two emotions. Blame portrays its object (some other person) as having committed a moral wrong without excuse. And guilt makes the same “charge” concerning oneself. These are the cognitive components of blame and guilt, respectively. We can characterize the motivational component of both emotions as the desire to hold the wrongdoer accountable, whether that person is oneself (in the case of guilt) or someone else (in the case of blame). Let us elaborate on this claim. To hold someone accountable to an obligation just is to make a (putatively legitimate) moral demand of that person. When a moral wrong has already been committed, one can hold a perpetrator accountable by pressing the demand that was flouted via expressions of blame and reproach. The implicit goal of these condemnatory actions is to get the perpetrator to hold himself accountable. How can a perpetrator hold himself accountable? By regarding his actions as condemnable in the same way that an outside party would, and responding appropriately to this fact. He can take responsibility by acknowledging and internalizing the wrongness of his action. This process of holding oneself accountable always requires (i) accepting that one did wrong and is blameworthy, (ii) not merely believing that, but also having the attitude of self-blame or guilt, and (iii) internalizing the standard that one’s wrongful action violated, and thus being motivated to comply with this standard in the future. Since this internalization will also motivate one to counteract the wrong that was done, holding oneself accountable will often involve (iv) acknowledging one’s guilt to others; (v) taking steps to ensure one’s own future compliance with the violated standard; (vi) taking steps to demonstrate one’s intention to comply with said standard to others; (vii) accepting punishment or sanction for one’s wrongdoing; and (viii) compensating and making amends with the victims of one’s wrongdoing, if there are any. By performing some or all of these actions, the perpetrator holds himself to the very demand that he had previously shirked.4 We can now formulate our claims about the motivational components of guilt and blame more precisely. Guilt motivates its subject to hold herself accountable by making the very demand of herself that she flouted in doing wrong; adequately holding oneself accountable will involve performing some or all of the actions listed in the previous paragraph. Blame motivates its subject to get the wrongdoer to hold himself accountable. People pursue this motive by holding the perpetrator accountable themselves,

4 Note that holding oneself accountable sometimes involves actions that require others’ participation: e.g. (iv), (vi), (vii), and (viii) on this list. This means that holding oneself accountable is not necessarily something one can do all by oneself.

Moral Psychology as Accountability  45 pressing the violated demand with verbal reproach, expressions of outrage, and punishment. The motives that accompany blame and guilt are, we claim, the fundamental motives driving moral behavior. Following DeScioli and Kurzban (2009), we distinguish two primary moral motives: conscience, the motive to regulate one’s own behavior by moral norms, and condemnation, the motive to respond to others’ moral wrongdoing with behaviors such as reproach and punishment. Our claim is that the motivational components of blame and guilt, as described above, are one and the same as the motives of moral condemnation and conscience, respectively. Regarding condemnation, this is a straightforward claim: The motive that accompanies blame, i.e. the motive to get the wrongdoer to hold herself accountable, is the drive behind condemnatory behavior. Regarding conscience, it is less obvious what our claim amounts to. This is because guilt is a backward-looking emotion, responding to a wrong action one has already performed, while conscience is more intuitively conceived as a forward-looking motivation to avoid doing wrong in one’s future actions. However, the conscience motive and the condemnation motive can each arise in both backward-looking and forward-looking contexts. The backward-looking condemnation and conscience motives are, respectively, the motivational components of blame and guilt, driving one to respond to wrongdoing by holding the perpetrator accountable to the demand that was already violated (whether that perpetrator is oneself or someone else). The forward-looking motives of conscience and condemnation aim to hold people to moral demands before they are violated. In forward-looking contexts, the condemnation motive is the goal to hold others to the demands of morality in their future conduct, while the conscience motive is the goal to hold oneself to the demands of morality in one’s own future conduct. We can now state our psychological theses in full, beginning with conscience. Looking forward, the conscience goal is to fulfill one’s moral obligations, or equivalently, to comply with the moral demands to which others may legitimately hold one accountable. Looking backward, after one has already committed moral wrong without excuse, the conscience goal is to hold oneself accountable, internalizing the moral demand that one flouted, making amends, and demonstrating to others one’s future intent to comply. This backward-looking conscience motive is the motivational component of the emotion of guilt, the cognitive component of which is the belief that one has committed moral wrong. Looking forward, the condemnation goal is to hold others to the demand that they not do moral wrong. Looking backward, after another person has already committed moral wrong without excuse, the condemnation goal is to get the perpetrator to hold herself accountable for her wrongdoing. One pursues this goal by holding the perpetrator accountable oneself, pressing the demand that was flouted via verbal reproach, expressions of blame, and even punishment. This backward-looking condemnation motive is the motivational component of the emotion of blame, the cognitive component of which is the belief that some other person has committed moral wrong.

46  Brendan Dill and Stephen Darwall We do not of course claim that accountability-based motivations and attitudes are the only mental states that can motivate moral conduct. Neither do we make any specific claims about these motives’ strengths relative to other motives in human psychology. Rather, we claim only that the conscience and condemnation motives exist, have the contents we have described, and crucially, are unique in non-coincidentally motivating moral behavior. While any motive may, in Kant’s phrase, “fortunately ligh[t]‌upon what is in fact . . . in conformity with duty” (Kant 1785/1996: 53, Ak. 4: 398), only motives and attitudes, like those we will be discussing, that involve holding oneself to standards of moral right and wrong can non-accidentally motivate an agent to avoid what would be morally wrong in her own view. Our theory posits a deep unity to the moral motives. Whether the object of concern is one’s own actions or another person’s, in the future or the past, the moral motives active in each of these contexts can all be described as motives to uphold the demands of morality. The philosophical account of morality as accountability has brought us, with some elaboration, to a unified psychological theory of the moral motives and emotions. We will now argue that our psychological theses are supported by the existing experimental data. In section 2, we will defend our theory of condemnation; in section 3, we will defend our theory of conscience. In section 4, we will turn to the implications of our framework for the distinctive nature of morality itself.

2.  Moral Condemnation 2.1.  Section Prospectus What is the motive driving morally condemnatory behavior? When we condemn, reproach, censure, sanction, or punish someone for doing wrong, what are we trying to accomplish? Our answer to this question is that people are motivated to respond to perceived moral wrongdoing with condemnatory actions primarily in order to get the perpetrator to hold herself accountable for her wrongdoing. Call this the accountability theory of condemnation. In this section, we will argue that the accountability theory better accounts for the experimental data on condemnation than any of the available alternatives. We will begin by considering two intuitively plausible theories of condemnation, which we call the egoistic theory and the deterrence theory. Due to evidence which we shall review (§2.2 and §2.3), most psychologists have rejected both of these theories. We will then turn to the theory of condemnation that seems to be currently most popular, which we call the retributive theory (§2.4). Directly comparing the retributive theory with our accountability theory, we shall argue that the evidence strongly favors the accountability theory. We conclude by addressing a recent challenge to the idea that genuinely moral condemnation exists at all (§2.5).

Moral Psychology as Accountability  47

2.2.  The Egoistic Theory The egoistic theory says that moral condemnation is motivated by a goal to attain a self-interested benefit of some kind. Different egoistic theories posit different self-interested motives underlying condemnatory behavior. The three most prominent views hold that condemners seek material benefit, reputational benefit, and mood improvement. We will consider each proposal in turn. The material benefit theory says that condemnatory sanctions are applied as negative incentives with the aim of motivating others to provide material goods to the condemner. This was once taken to be the default explanation of punishment behavior in economic games such as the Public Goods game. However, studies of these very games have refuted this version of the egoistic theory. Subjects are willing to pay a fee to punish non-cooperative players even when doing so guarantees a net material loss, since they will have no further interactions with the individuals whom they punish (Fehr and Gächter 2000; Fehr, Fischbacher, and Gächter 2002; Turillo et al. 2002; Fehr and Fischbacher 2004). If subjects were merely pursuing material or financial gain, they would not pay to punish in such conditions. The theory that condemnatory behavior aims for reputational benefit has more empirical support. One study has shown that subjects will pay more of their own money to punish a free rider when their decision will be made known to others than when it is anonymous (Kurzban, DeScioli, and O’Brien 2007). The proposed explanation of this result is that others will view a person more positively for punishing a non-cooperative other, and so people condemn in order to secure this social approval.5 However, several studies have shown that people are willing to punish at cost to themselves even in totally anonymous conditions, which offer no opportunity for reputational gain or loss (Turillo et al. 2002; Fehr and Fischbacher 2004; Gächter and Herrmann 2008). The reputation-based egoistic theory cannot account for the motive to punish in these anonymous contexts. The final egoistic theory we will consider says that people condemn wrongdoers in order to “let off steam,” relieving the negative affect caused by the transgression and thus improving their mood. This hypothesis has been tested directly by Gollwitzer and Bushman in a recent paper titled “Do Victims of Injustice Punish to Improve their Mood?” (Gollwitzer and Bushman 2011). Two studies answer this question with a resounding no. Both studies confronted subjects with a free-riding perpetrator and gave them the opportunity to punish him or her. Some of these subjects were led to believe that attempts to improve their negative mood would be ineffective or unnecessary. If condemnation were driven by a mood regulation goal, then subjects in this experimental condition would not be motivated to punish the perpetrator, since doing so would be pointless. (Indeed, a previous study using the same 5 We venture that the best explanation for this result is that publicity made these subjects more motivated to do what they perceived to be morally right, i.e. punish the free rider, since accountability to others activates the conscience motive (as we argue in §3.2.2).

48  Brendan Dill and Stephen Darwall experimental design has demonstrated that non-morally motivated aggression is sometimes driven by a mood regulation goal; see Bushman, Baumeister, and Phillips 2001). However, Gollwitzer and Bushman found no significant difference in punishing behavior between the subjects in the experimental and control conditions. This finding tells strongly against the idea that condemnation is driven by an egoistic goal to improve one’s mood. As each of the three most plausible egoistic proposals faces powerful empirical objections, we conclude that we should reject the egoistic theory of condemnation as a whole.

2.3.  The Deterrence Theory The deterrence theory of condemnation says that condemnatory behaviors are motivated by the goal to deter people from future immoral behavior. This theory is given some intuitive support by the fact that deterrence benefits seem to provide an appealing justification for condemnatory behavior, one which many people explicitly endorse (Ellsworth and Ross 1983; Carlsmith, Darley, and Robinson 2002; Carlsmith 2008). However, people’s actual condemnatory behavior is insensitive to potential deterrence benefits, a robust finding that is fatal to the deterrence theory (Baron, Gowda, and Kunreuther 1993; Baron and Ritov 1993; Darley, Carlsmith, and Robinson 2000; Sunstein, Schkade, and Kahneman 2000; Carlsmith, Darley, and Robinson 2002; Carlsmith 2006; Carlsmith 2008; Carlsmith and Sood 2009; Keller et al. 2010). In most studies testing the deterrence theory, subjects are told about a crime and are asked to make a judgment regarding how severely the perpetrator of that crime should be punished. The experimenters vary the descriptions of the crime according to some factor that is relevant to the deterrence benefits of the punishment, such as the publicity that the punishment will receive. The deterrence theory predicts that subjects will judge that a more severe punishment is appropriate when the potential deterrence benefit is high, and that a less severe punishment is appropriate when the potential deterrence benefit is low. The findings in all of these studies contradict this prediction: The severity of punishment subjects judge to be appropriate is simply insensitive to variations in potential deterrence benefit. In some studies, subjects even assign substantial punishments when doing so will have harmful effects in addition to not having any deterrent effect (Baron, Gowda, and Kunreuther 1993; Baron and Ritov 1993). This robust pattern of insensitivity to deterrence benefits demonstrates that condemnatory behavior is not motivated by the goal to deter future wrongdoing.

2.4.  The Accountability Theory vs. the Retributive Theory We take the most compelling alternative to our proposal to be what we call the retributive theory of condemnation. This theory claims that condemnatory behavior is motivated by the goal to cause harm to the perpetrator in proportion with the blameworthiness of the perpetrator’s wrongdoing. From the retributive perspective, the

Moral Psychology as Accountability  49 punishment imposed on the perpetrator is an end in itself, and the sole end condemners are after. In contrast, the accountability theory views punishment as a means to the end of getting the perpetrator to hold himself accountable for his wrongdoing. On our view, merely punishing the perpetrator is not sufficient to satisfy the condemnation motive—the perpetrator must also acknowledge his wrongdoing, feel remorse, make amends, etc., in order for the condemner’s goal to be fulfilled. 2.4.1.  Evidence for the Retributive Theory The retributive theory has traditionally been presented as an alternative to the deterrence theory. As a result, empirical tests of the retributive theory have focused on testing it against the deterrence theory. The results of these experiments heavily favor the retributive theory. In addition to finding that the severity of punishment subjects judge appropriate is not sensitive to deterrence benefits, these studies have found that subjects’ punishment judgments are sensitive to the perceived blameworthiness of the perpetrator (Baron, Gowda, and Kunreuther 1993; Baron and Ritov 1993; Darley, Carlsmith, and Robinson 2000; Sunstein, Schkade, and Kahneman 2000; Carlsmith, Darley and Robinson 2002; Carlsmith 2006; Carlsmith 2008; Carlsmith and Sood 2009; Keller et al. 2010). The more blameworthy the subjects deem the offense, the more severe the punishment they recommend—a pattern predicted by the retributive theory, but not by the deterrence theory. Crucially, however, these studies do not support the retributive theory over the accountability theory, or vice versa. Both theories predict that people will judge more severe punishments to be appropriate for more blameworthy crimes. On our favored view, punishment is a part of the larger process of holding the perpetrator accountable. By imposing a punishment on the perpetrator, the condemner communicates to the perpetrator the blameworthiness of his offense, thus pushing him to internalize this blameworthiness, feel remorse, and hold himself accountable. By accepting the punishment, the perpetrator can censure himself and thereby demonstrate his commitment to the moral standard that he violated.6 Imposing and accepting punishment are actions of the same kind as imposing and accepting verbal reproach—the former is simply a more severe version of the latter. So, just as more blameworthy actions warrant a harsher reproach, they also warrant harsher punishments. Hence the accountability theory predicts that more blameworthy actions will motivate more severe punishments, just as the retributive theory does. So we shall set the results confirming this prediction aside, and turn to experiments that can tell between these two theories.

6 This characterization is supported by interview data on punishment in romantic relationships: “punishment sends a signal that something is wrong and a relationship rule has been broken; it helps to ‘educate’ an offending partner about the hurt partner and his or her needs . . . Notably, several participants also described the function of punishment as a ‘test’ for the relationship. If a punished partner responds with empathy and remorse, and does not retaliate in turn, then this is a reliable sign of commitment to the relationship” (Fitness and Peterson 2008, 262).

50  Brendan Dill and Stephen Darwall 2.4.2.  Evidence for the Accountability Theory The retributive and accountability theories make different predictions about the end state that, when attained, satisfies the condemnation motive. If the retributive theory is correct, the condemnation motive should be satisfied once the perpetrator has adequately suffered for his wrongdoing. If the accountability theory is correct, the condemnation motive should be satisfied when and only when the perpetrator has adequately held himself accountable for his wrongdoing. Thus the two theories make different predictions regarding cases where the perpetrator has been punished but has not held himself accountable by acknowledging his blameworthiness and displaying remorse.7 If the condemnation motive is satisfied under these conditions, this finding would favor the retributive theory. If the condemnation motive is not satisfied under these conditions, that finding would favor the accountability theory. It would provide further support for the accountability theory if, in addition, the condemnation motive is satisfied when the perpetrator has acknowledged his blameworthiness and displayed remorse in addition to being punished. To evaluate these predictions, we turn first to the research literature on forgiveness. McCullough (2001), quoting Webster’s Dictionary, tells us that to forgive is “to give up resentment against or the desire to punish” (Forgive 1983, 720). If forgiveness is the giving up of blame, then it should typically result from the satisfaction and resultant deactivation of the condemnation motive.8 This means that the accountability and retributive theories’ predictions regarding the satisfaction conditions of the condemnation motive entail predictions regarding the conditions under which forgiveness typically occurs. Looking at the forgiveness literature through this lens, the accountability theory’s predictions are overwhelmingly supported. One of the most robust findings from this research is that forgiveness usually occurs when and only when the perpetrator has adequately demonstrated remorse by acknowledging guilt, apologizing, and/or offering compensation (McCullough, Worthington, and Rachal 1997; McCullough et al. 1998; Gold and Weiner 2000; Bottom, Gibson, and Daniels 2002; de Jong, Peters, and De Cremer 2003; Schmitt et al. 2004; Zechmeister et al. 2004; Kelley and Waldron 2005; Bachman and Guerrero 2006; McCullough et al. 2009; Fehr, Gelfand, and Nag 2010; Hannon et al. 2010; Leonard, Mackie, and Smith 2011; Tabak et al. 2011). In a meta-analysis of over 175 studies, Fehr, Gelfand, and Nag (2010) found that the extent to which the perpetrator apologizes predicts the degree of the victim’s forgiveness with a total correlation coefficient of r = .42. This was one of the strongest effects they found. The hypothesis implied by the retributive theory, that forgiveness is predicted by the severity of the punishment received by the offender, did not have 7 The two theories also make different predictions for the case where the perpetrator hasn’t been punished (in any usual sense) but has adequately held himself accountable (including to others). 8 Since it is possible to voluntarily forgive a perpetrator who has neither been punished nor held herself accountable, this will not always be the case. However, as with the deliberate relinquishment of other unsatisfied goals, forgiving before one’s condemnation motive has been satisfied is a difficult endeavor requiring considerable self-control; we should thus expect such cases to be relatively rare.

Moral Psychology as Accountability  51 enough support to even make it onto the list of 22 predictive factors tested in Fehr et al.’s meta-analysis. A closer look at the studies demonstrating the effects of apology on forgiveness shows that apologies only elicit forgiveness when they are perceived as a sincere expression of the perpetrator’s remorse and commitment to improved conduct. A mere apology is less effective than an apology combined with substantive compensation (Bottom, Gibson, and Daniels 2002). In fact, without adequate amends, an apology can backfire, resulting in less forgiveness (Zechmeister et al. 2004). A study that pulled apart the various elements of an apology showed that forgiveness was most likely when the perpetrator admitted fault, admitted the damage that was done, expressed remorse, and offered compensation; only then could the perpetrator ask for forgiveness without this request backfiring (Schmitt et al. 2004). These studies show that apologies lead to forgiveness when and only when they are seen as demonstrating that the perpetrator has fully held himself accountable for his wrongdoing. Given the premise that forgiveness is usually caused by the satisfaction of the condemnation motive, this implies that the condemnation goal aims to get the perpetrator to hold himself accountable. Beyond the forgiveness literature, we find even more direct experimental tests of the retributive and accountability theories’ predictions. A study titled “The Paradoxical Consequences of Revenge” reports the following surprising result: Though people expect to feel better after they have punished someone who has wronged them, they in fact feel worse than they would if they had not punished at all (Carlsmith, Wilson, and Gilbert 2008). Carlsmith et al. put subjects in a Public Goods game staged to have an offender who encouraged others to cooperate and then did not cooperate herself. Some subjects were given the opportunity to punish this free rider (punishment condition) while other subjects had no opportunity to punish (no-punishment condition). After doling out punishment, the punishment condition subjects had no further interactions with or communication with the perpetrator; so, crucially for our purposes, there was no opportunity for the perpetrator to take responsibility or signal remorse. After the study, subjects in the punishment condition reported significantly more negative affect than subjects in the no-punishment condition. In addition, subjects who had punished the free rider reported ruminating about the offender more than subjects who had not been given the opportunity to punish. Carlsmith et al.’s discussion focuses on the implications of these results for our understanding of affective forecasting. However, we think the study has more direct relevance to the retributive theory. According to the retributive theory, merely punishing the offender should be sufficient to fulfill the condemnation goal. Research on goal pursuit has shown that two important signatures of goal fulfillment are positive affect and inhibition of goal-relevant concepts; in contrast, negative affect and increased accessibility of goal-relevant concepts are signals of goal frustration (Chartrand 2001; Förster, Liberman, and Higgins 2005; Förster, Liberman, and Friedman 2007; Liberman, Förster, and Higgins 2007; Denzler, Förster, and Liberman 2009). Therefore, the negative affect and ruminating thoughts experienced

52  Brendan Dill and Stephen Darwall by Carlsmith et al.’s subjects after punishing strongly indicates that merely punishing the offender did not fulfill their motive to condemn, contrary to the retributive theory’s predictions. Two other studies report a similar pattern (Gollwitzer and Denzler 2009; Gollwitzer, Meder, and Schmitt 2010). In both studies, subjects were given the opportunity to punish a confederate who had treated them unfairly. After doling out punishment, one group of subjects received a message from the offender communicating his understanding that he deserved the punishment he had received (the “understanding” condition), while the other group either received no communication or received an actively unrepentant message from the offender. A manipulation check showed that subjects perceived the “understanding” message to “not only [contain] an admittance of harm and fault, but also an expression of remorse, an apology, and, most strikingly, a compensation offer” (Gollwitzer, Meder, and Schmitt 2010, 370). In other words, the “understanding” condition subjects took their offender to be holding himself accountable. Confirming the accountability theory’s predictions, Gollwitzer et  al.’s subjects showed clear signs of goal fulfillment when they received the “understanding” message, but not when they merely punished the offender. In the first study (Gollwitzer and Denzler 2009), subjects showed significantly decreased automatic accessibility of aggression-related words after receiving the “understanding” message; as we have said, this is a strong indicator of goal fulfillment. Subjects who merely punished the offender showed no such decrease in accessibility, contrary to the retributive theory’s prediction. In the second study (Gollwitzer, Meder, and Schmitt 2010), subjects reported how satisfied they felt after punishing the offender. Subjects who received the “understanding” message expressed significantly greater feelings of satisfaction than subjects in the “no understanding” condition, who, crucially, were no more satisfied than those who had not punished the offender at all. Contra the retributive theory, mere punishment did not satisfy subjects if it was not accompanied by the offender’s holding himself accountable. Confirming the accountability theory, punishment did satisfy subjects if, and only if, it was accompanied by a message that those subjects saw as showing that the perpetrator was taking responsibility—admitting fault, expressing remorse, and apologizing. Empirical tests of the predictions made by the accountability and retributive theories confirm the predictions of the accountability theory while disconfirming the predictions of the retributive theory. We have also seen (in §2.2 and §2.3) that the egoistic and deterrence theories of condemnation face significant empirical challenges as well. On this basis, we claim that the accountability theory is the best-supported theory of moral condemnation currently available. This concludes our argument for the accountability theory of condemnation.

2.5.  The Existence of Genuine Moral Condemnation We conclude our discussion of moral condemnation by considering a recent challenge to the existence of genuinely moral condemnation offered by C. Daniel Batson and

Moral Psychology as Accountability  53 colleagues (Batson et al. 2007; Batson, Chao, and Givens 2009; O’Mara et al. 2011). In a series of studies, Batson et al. have found that people experience and express great outrage in response to moral violations when the victim is themselves, someone in their group, or someone with whom they empathically identify, but show much less outrage at moral wrongdoing when the victim is someone else, outside of their group, with whom they do not empathize (Batson et  al. 2007; Batson, Chao, and Givens 2009; O’Mara et al. 2011; see also the similar results in Yzerbyt et al. 2003; Bernhard, Fischbacher, and Fehr 2006; and Gordijn et al. 2006). They conclude from this data that the outrage experienced by their subjects cannot be moral outrage, because moral outrage would not differentiate in this way between offenses against oneself and one’s group members on the one hand, and offenses against strangers on the other. While we do not dispute Batson et  al.’s findings, we think the conclusion they draw from these findings is mistaken. Moral outrage can take both personal and impersonal forms: The personal resentment felt by a victim of wrongdoing is no less a form of moral blame than the impersonal indignation felt by a third-party bystander (Strawson 1968; Darwall 2012). The fact that people feel personal moral outrage more intensely than they feel impersonal moral outrage does not impugn the moral content of either emotion. Rather, we think this result is best explained by the simple fact that stimuli must be emotionally salient to produce a strong emotional response. Wrongs committed against oneself or members of one’s group are more emotionally salient, so quite understandably, they produce a more intense response of moral outrage. A similar point holds for altruistic motivation: It can both be the case that people are sometimes genuinely driven by an altruistic motive to improve another’s well-being for its own sake (Batson and Shaw 1991) and that people are much more likely to experience this altruistic motive when another person’s welfare is made emotionally salient to them by empathic perspective-taking (Batson and Shaw 1991) or identifiability (Small and Loewenstein 2003). In general, the fact that manipulations of emotional salience affect the intensity of a motive or emotion should not affect the conclusions we draw about the content of that motive or emotion. We think Batson et  al. are right to emphasize the importance of distinguishing between non-moral anger and genuinely moral condemnation. However, we do not think that impersonality is the feature that distinguishes moral outrage from non-moral anger. Instead, we propose that the distinction between moral and non-moral anger is best understood in terms of the differing motivations that accompany these emotions. As we have argued in this section, moral anger drives its subject toward a unique goal: to make the wrongdoer hold himself accountable to the moral demand he flouted. This goal does not seem to be shared by non-moral anger, which aims instead at regulating mood (Bushman, Baumeister, and Phillips 2001), acquiring social status (Griskevicius et al. 2009; Wenzel et al. 2008), or simply inflicting harm for its own sake (Denzler, Förster, and Liberman 2009). By showing that the condemnation motive has the essentially moral aim of holding a perpetrator to a

54  Brendan Dill and Stephen Darwall moral demand, as opposed to the morally neutral aims of attaining egoistic benefit, deterrence, or retribution, we have provided a basis for a principled distinction between moral and non-moral anger. This is a crucial point: Previous major reviews on the moral emotions have treated anger as a unitary psychological state (Haidt 2003; Hutcherson and Gross 2011), and thus have missed a distinction of fundamental importance for moral psychology.9

3.  Moral Conscience 3.1.  Section Prospectus What is the motive driving morally conscientious behavior? When we “do the right thing,” what are we trying to accomplish? Our answer to this question is that morally conscientious behavior is driven by moral conscience, an intrinsic desire to comply with moral demands to which one may be legitimately held accountable, or equivalently, to comply with one’s moral obligations. This is the accountability theory of moral conscience. The most popular alternative theory of moral conscience is a view we call the approval theory. Approval theorists are skeptical about the existence of genuine moral conscience, maintaining that instead of being motivated to in fact comply with our moral obligations, we are motivated only to appear as if we are complying with our moral obligations (Batson 2008). The approval theory denies that human beings have any intrinsic desire to fulfill their moral obligations, instead claiming that moral behavior is driven by an instrumental desire to appear moral in order to gain the egoistic benefits of good repute. The accountability theory, in contrast, holds that while people may well be motivated to appear moral and gain a good reputation, these are not the only motives driving moral behavior; in addition, human beings have an intrinsic desire to uphold their moral duties. The approval theory is composed of two major theses. First, it claims that morally conscientious behavior is motivated by what we will call the approval motive, a desire to gain the moral approval and avoid the moral disapproval of one’s peers. Second, the approval theory claims that the moral conscience motive does not exist. In contrast, the accountability theory is not similarly committed to denying the existence of the approval motive. We think it is obvious that human beings desire approval and fear disapproval, and that this motive will sometimes contribute to the production of moral behavior. Our controversial further claim is that human agents also desire to uphold 9 An illuminating exception to this trend is Lemay, Overall, and Clark (2012), whose distinction between “hurt” and “anger” closely corresponds to our own proposed distinction between moral and non-moral anger, with what they call “hurt” corresponding to what we call “moral anger,” and what they call “anger” corresponding to what we call “non-moral anger.” Our proposal moves one step beyond Lemay et al.’s in hypothesizing that the moral/non-moral anger distinction applies not only to the condemnatory emotions of victims, but to those of third parties as well.

Moral Psychology as Accountability  55 morality for its own sake: People are sometimes motivated by genuine moral conscience. So the disagreement between the accountability and approval theories boils down to the question of whether moral conscience exists. Some might worry that the approval theory as we have described it is a straw man opponent, since we have saddled this theory with the very strong negative claim that moral conscience does not exist. A  more plausible version of the approval theory might acknowledge that the moral conscience motive exists, but claim that it plays a minor role in producing moral behavior compared with the far more powerful approval motive. (A view like this is advocated in Haidt 2012.) This hybrid view presents a more formidable opponent, which our arguments do not directly address.10 However, we think the more radical version of the approval theory we address is worth arguing against, even if it is a straw man. Just as philosophers argue against skepticism about our knowledge of the external world, not because anyone actually holds this view, but in order to find a more secure foundation for this knowledge, it is worthwhile to argue against skepticism about moral conscience even if no one actually holds this view, in order to find a more secure foundation for our theory of moral conscience. To this end, we will focus on the stronger version of the approval theory, which denies the existence of moral conscience.11 The accountability theory does not merely affirm the existence of moral conscience, however: It also provides a theory of the content of this conscience motive. A more orthodox view of moral conscience takes the idea of moral obligation as primitive, saying that what it is for an agent to represent a rule as a moral obligation is just for her to include it in her internal list of moral principles. An agent determines what principles make it onto this “moral rule list” by exercising her individual faculty of moral judgment. Regarding what distinguishes the rules that an agent takes to be her moral obligations from any other rules, all that seems to be said is that these rules are somehow internally stamped with a “morality label” that picks them out as the rules relevant to moral conscience. On this view, the conscience goal aims to have one’s behavior conform to those rules that one has internally labeled as morally obligatory. Contrast this orthodox view of moral conscience with the accountability theory. Rather than leaving the concept of moral obligation as primitive, the accountability theory provides an analysis of what it is to represent a rule as a moral obligation. As stated in the introduction, we hold that to regard a rule as a moral obligation just is to regard it as a demand to which any agent may legitimately hold one accountable with the reactive attitude of blame if one violates the demand without excuse. Thus we can elaborate upon the content of the conscience motive as follows: The goal to 10 However, our arguments do have some significance for the hybrid view. We argue in §3.2 that patterns of behavior that are usually ascribed to the approval motive can be explained at least as well by appeal to the conscience motive. Our arguments thus undermine any argument based in this data for the view that the approval motive is more powerful than the conscience motive, since the data may be explained by the conscience motive as well. 11 Thanks to Jonathan Haidt for raising this concern.

56  Brendan Dill and Stephen Darwall comply with one’s moral obligations is one and the same as the goal to comply with those demands to which one may legitimately be held accountable. As we shall see (especially in §3.2.2), this elaborated theory of the conscience motive has explanatory resources that the more orthodox view of conscience does not. It is important to distinguish the conscience motive as the accountability theory construes it from the motive posited by the approval theorist. The approval motive can be described as the motive to comply with those demands to which one is actually being held accountable by others; or to avoid incurring others’ blame and disapproval. In contrast, the conscience motive we posit is the motive to comply with those demands to which one would legitimately be held accountable by others; or to avoid warranting others’ blame and disapproval. The crucial difference is between wanting to avoid being blamed (approval motive) and wanting to avoid being blameworthy (conscience motive). The present section will be an extended argument in favor of the accountability theory over the approval theory. The second part of the section (§3.2) will be dedicated to rebutting arguments for the approval theory; the final part (§3.3) will present a positive argument for the accountability theory based in data on guilt and shame. Before we proceed, however, we wish to mention and set aside a third view of moral conscience that, although initially appealing, is ultimately untenable. This is the idea that the moral conscience motive is a motive of altruistic compassion for others based in empathy. Though an empathy-based altruistic motive has been shown to exist and often leads to morally right actions such as helping those in need (Batson and Shaw 1991), this motive does not have the right kind of content to either count as a moral conscience motive or explain all cases of morally conscientious behavior. The goal of empathy-based altruism is to help the person for whom empathy is felt, considered independently of the morality of doing so; thus this motive has no intrinsic moral content. In line with this conceptual point, at least one study has experimentally dissociated the empathy-based altruistic motive to help a specific person from the moral conscience motive to treat people fairly (Batson et al. 1995). This study also illustrates our second point, that there are many kinds of morally conscientious behavior that are not plausibly explained by altruistic concern for others’ welfare, such as people’s concern for fairness, hierarchy, authority, promissory and contractual obligations, and religious strictures. Thus, though empathy-based altruism may well sometimes help to motivate moral behavior, it cannot itself be the moral conscience motive. What the moral conscience motive is, and whether it exists, are the questions to which we now turn.

3.2.  Arguments for the Approval Theory In this section, we will present what we take to be the two strongest arguments for the approval theory, and argue that the accountability theory has the resources to rebut both arguments.

Moral Psychology as Accountability  57 3.2.1.  First Argument: Moral Hypocrisy and Moral Licensing The most explicit and direct arguments for the approval theory in the literature are based upon two empirical findings: the moral hypocrisy effect and the moral licensing effect. The moral hypocrisy effect is the finding that, under certain conditions, subjects will strive to appear moral without undertaking the costs of actually acting morally (Batson et al. 1997; Batson, Thompson, and Seuferling 1999; Batson and Thompson 2001; Batson, Thompson, and Chen 2002; Batson, Collins, and Powell 2006; Batson 2008). The moral licensing effect is the finding that subjects will behave less morally immediately after engaging in behavior that makes them appear moral (Monin and Miller 2001; Cain, Loewenstein, and Moore 2005; Khan and Dhar 2006; Effron, Cameron, and Monin 2009; Sachdeva, Iliev, and Medin 2009; Mazar and Zhong 2010; Kouchaki 2011; Effron, Miller, and Monin 2012; Effron, Monin, and Miller 2013; Merritt et al. 2012). Both of these findings appear to strongly favor the approval theory because they seem to show that people are more motivated to appear moral than they are to actually be moral. Though this prima facie appearance is strong, we think that upon reflection, these two findings do not in fact support the approval theory over the accountability theory. To see why, consider the difference between the motive to appear moral to others and the motive to appear moral to oneself. If the moral licensing and hypocrisy effects showed that subjects are less motivated to act morally when they merely appear moral to others, this would provide unambiguous support for the approval theory. Such a result would show that the motive driving moral behavior is satisfied when social approval has been secured, even if the agent is aware that she has not fulfilled her moral obligations. This would imply that the motive driving moral behavior is a motive to secure social approval, not a motive to fulfill one’s moral obligations. On the other hand, if the moral licensing and hypocrisy effects show instead that subjects are less motivated to act morally only when they appear moral to themselves, this result is compatible with the existence of moral conscience and thus with the accountability theory. For an agent to appear moral to herself just is for it to seem to her that she has fulfilled her moral obligations. If the agent is motivated by genuine moral conscience, this motive will be satisfied when it appears to her that she has fulfilled her moral obligations. As we have already noted (in §2.4.2), it is a domain-general feature of motivation that a goal is suppressed or “turned off ” when it appears to the agent that the goal has been fulfilled (Förster, Liberman, and Higgins 2005; Förster Liberman, and Friedman 2007; Liberman, Förster, and Higgins 2007; Denzler, Förster, and Liberman 2009). Thus we can predict that when it appears to an agent that she has fulfilled her moral obligations, her conscience motive will be fulfilled and thus suppressed immediately thereafter. This would make room for more selfish motives to govern the agent’s behavior. Thus by affirming the existence of genuine moral conscience, the accountability theory can predict and explain why, when an agent appears moral to herself, she will act less morally immediately afterwards.

58  Brendan Dill and Stephen Darwall So, it seems that whether the moral hypocrisy and moral licensing effects ground an argument for the approval theory over the accountability theory depends crucially on whether these effects involve appearing moral to others or appearing moral to oneself. The research on these effects strongly supports the latter conclusion: Moral hypocrisy and moral licensing occur when and only when subjects appear moral to themselves, not to others. We will now review this research, beginning with moral hypocrisy. In the moral hypocrisy paradigm, subjects are given the opportunity to assign tasks to themselves and another participant, where one task is clearly much more enjoyable than the other. These subjects are told that the fairest choice is to flip a coin, giving oneself and the other participant equal chances of receiving the better task. About half of subjects choose to flip the coin; but crucially, these subjects still overwhelmingly assign themselves the better task (90 percent; Batson et al. 1997). This is the moral hypocrisy finding: These subjects flip the coin, which makes them appear moral, but do not obey the coin flip’s results when it gives them the less enjoyable task, thus avoiding the costs of actually being moral. Despite their unfair behavior, subjects who flip the coin subsequently rate their own behavior as having been significantly more moral than subjects who do not flip the coin (Batson 2008). By flipping the coin, these subjects are somehow fooling themselves into thinking that they did the right thing. Batson and colleagues predicted that drawing subjects’ attention to their own behavior would eliminate the moral hypocrisy effect by blocking this self-deception. This is what they found: When subjects in this paradigm are placed in front of a mirror (a manipulation that has been shown to increase self-awareness), those who flipped a coin gave themselves the more enjoyable task only 50 percent of the time, abiding by the flip’s results (Batson, Thompson, and Seuferling 1999). We can sum up these findings as follows: When people are not attending to their behavior, they may convince themselves that they have behaved morally when in fact they have not (moral hypocrisy). However, as soon as someone pays enough attention to her behavior to notice that it is immoral (the mirror effect), she is motivated to adjust her behavior to conform to her moral standards. Clearly, Batson et al.’s moral hypocrisy effect depends on subjects’ appearing moral to themselves, not to others; thus it does not provide evidence that favors the approval theory over the accountability theory. The evidence on moral licensing points in the same direction. The original moral licensing paradigm showed that subjects are more likely to make implicitly racist or sexist hypothetical hiring decisions after being given an opportunity to explicitly disagree with racist or sexist statements (Monin and Miller 2001). The affirmations of anti-racist/anti-sexist values that produced this licensing effect were performed in a written questionnaire that subjects were told was to be private and anonymous; so, these subjects should not have seen their affirmations as having any influence over their appearance to others. A  follow-up study showed that the moral licensing effect emerges just as powerfully when the value-affirming survey (the experimental

Moral Psychology as Accountability  59 manipulation) and the job selection survey (the dependent measure) are administered by two different experimenters. These findings seem to establish that the moral licensing effect is not explained by subjects’ establishing their “moral credentials” to others, since they could not reasonably take their anonymous value affirmations to have any effect on anyone else’s opinion of them. Instead, the moral licensing effect seems best explained by the value affirmations causing subjects to appear moral to themselves. At least two studies have found that the moral licensing effect is statistically mediated by a positive change in how morally good subjects judge themselves to be (Sachdeva et al. 2009; Kouchaki 2011). One study obtained a moral licensing effect merely by having subjects write a story about themselves that contained morally positive words, thus affirming their own moral goodness (Sachdeva et al. 2009). Privately writing such a story has no effect on one’s appearance to others, so it must produce licensing by making subjects appear moral to themselves. Thus, we submit that the moral licensing effect, like the moral hypocrisy effect, is caused by subjects’ appearing moral to themselves, not to others, and thus is quite compatible with the existence of moral conscience. Thus we conclude that the approval theory is not supported over the accountability theory by the experiments demonstrating moral hypocrisy and moral licensing, as both effects are quite compatible with the existence of moral conscience. In fact, these findings can be seen as supporting the accountability theory over the approval theory, rather than vice versa. For while the accountability theory has a ready explanation for why appearing moral to oneself should decrease subsequent moral motivation (since it fulfills the moral conscience goal), the approval theory does not. We should not expect a priori that the goal to attain the moral approval of others should be satisfied when one appears moral only to oneself; rather, it seems an approval-seeking agent would remain vigilant until her moral credentials were public. Batson explains the importance of securing self-approval as follows: “If I can convince myself that serving my own interests does not violate my principles, then I can honestly appear moral and so avoid detection without paying the price of actually upholding the principles” (Batson 2008, 53). This hypothesis might be able to explain why self-approval is necessary for the satisfaction of the approval motive, but would not explain why it is sufficient. If deceiving yourself into approving of your own actions is a means to attaining the moral approval of others, then merely attaining self-approval should not be sufficient to satisfy the approval motive, as the moral licensing studies indicate it is. The approval motive should only be satisfied once one has gained others’ approval in addition to one’s own. More generally, we don’t need to attribute such a nefariously self-manipulating motive to human beings to explain their susceptibility to the sort of self-deception Batson’s studies have revealed. A simpler explanation for this self-deception is that the moral conscience motive is only one motive among many, including selfish motives, which compete for control over cognition and behavior. The dominant motive at any time biases attention and cognition so as to suppress alternative, incompatible motives

60  Brendan Dill and Stephen Darwall (Shah, Friedman, and Kruglanski 2002). So if, as seems likely, selfish motives to gain money and avoid tedious, difficult tasks are originally dominant in the experimental context, they will bias attention to avoid goal-discrepant thoughts such as “it would be unfair to disobey the coin flip and give myself the better task.” Only when moral considerations are sufficiently attention-grabbing to bring the conscience motive to the fore—as Batson et al.’s mirror ensured—will these selfish cognitive biases give way to morally motivated thought and action. We thus submit that the accountability theory gives a better explanation of the moral hypocrisy and moral licensing findings than the approval theory can provide. 3.2.2.  Second Argument: The Dependence of Conscience on Social Norms A more general, and more worrisome, argument for the approval theory goes as follows. If people are motivated by genuine moral conscience, as the accountability theory claims, then we should expect their moral behavior to be best predicted by their beliefs about what is morally right and wrong. If, on the other hand, morally conscientious behavior is solely driven by the motive to gain social approval, as the approval theory claims, then we should expect people’s moral behavior to be best predicted by whether and how their actions are being judged by others. A survey of the experimental literature overwhelmingly confirms the approval theory’s prediction that moral behavior depends in this way on social context, while providing little support for the idea that explicit moral beliefs predict moral behavior. Across many studies with various methodologies, the data indicates that whether subjects are motivated to conform to a moral standard depends primarily upon whether there are other people watching and holding them to that standard. In other words, H. L. Mencken seems to have been on target when he declared that “conscience is the inner voice that warns us somebody may be looking” (Mencken 1949). And thus, the argument goes, the approval theory is confirmed, and the accountability theory falsified. We do not dispute the empirical premises of this argument. The data, which we will review presently, clearly indicates that moral behavior is powerfully influenced by the presence or absence of actual social accountability. However, we will argue that the accountability theory can explain this data at least as well as the approval theory can. The first line of relevant findings shows that people’s behavior in many morally relevant domains is strongly predicted by their perceptions of the norms of approval and disapproval that hold in their social environment. This result has been demonstrated across many domains of moral behavior, including charitable donation (Reingen 1982), pro-environmental behavior (Reno, Cialdini, and Kallgren 1993; Kallgren, Reno, and Cialdini 2000; Schultz et al. 2007; Goldstein, Cialdini, and Griskevicius 2008), contraception use (Fekadu and Kraft 2002), voting (Gerber, Green, and Larimer 2008; Gerber and Rogers 2009), intergroup cooperation (Paluck 2009), and a wide range of criminal behaviors (e.g. vandalization, Zimbardo 1969; tax evasion, Steenbergen, McGraw, and Scholz 1992; and many others, Grasmick and Green 1980; Tittle 1980;

Moral Psychology as Accountability  61 Kahan 1997). Summarizing the results on criminal behavior, Dan Kahan observes: “the perception that one’s peers will or will not disapprove exerts a much stronger influence than does the threat of a formal sanction on whether a person decides to engage in a range of common offenses—from larceny, to burglary, to drug use” (1997, 354). These findings are corroborated by laboratory experiments on cooperation in economic games, especially the public goods game. Cooperation in these games increases dramatically when the participants are allowed to punish others for free riding by subtracting from their money (Fehr and Gächter 2000). One might think that these subjects are cooperating simply out of self-interest, to avoid losing money by incurring sanctions. However, further studies have indicated that punishment motivates cooperation by means of its expression of disapproval rather than its financial incentives. One line of studies shows that holding the free rider accountable by merely expressing indignation is more effective than material punishment in motivating cooperation (Ostrom, Walker, and Gardner 1992; Masclet et al. 2003; Noussair and Tucker 2005; Ule et al. 2009; Janssen et al. 2010; see also Orbell, Van de Kragt, and Dawes 1988). In addition, punishment without moral disapproval is ineffective in motivating cooperation. Fehr and Rockenbach (2003) found that when a punishment is imposed to enforce an obviously illegitimate, selfish demand, subjects will cooperate less than control subjects who faced no sanctions at all. These findings indicate that what motivates subjects to cooperate are not material punishments, but rather the condemnation they express. The approval theorist will say that the above results are best explained by the fact that agents are concerned only with their peers’ approval. If people were genuinely motivated by conscience, the approval theorist may reason, they would follow their moral convictions regardless of whether doing so attracts the approval or disapproval of others. Perhaps this conditional would hold true of an ideally rational, cognitively unlimited agent motivated solely by moral conscience. But if we consider imperfect creatures like ourselves, who have highly limited cognitive capacities and must negotiate between many competing motives, we can see why moral behavior might depend on social norms even if it is motivated by genuine moral conscience. The defender of moral conscience can argue that since human beings are fallible judges of moral rectitude, it makes rational sense for us to look to others for guidance as to what is morally right and wrong. Human beings are epistemically dependent upon and highly deferential to others regarding descriptive matters even as mundane as the relative lengths of lines on paper (Asch 1955), and so it should be no surprise that they are similarly deferential in matters of moral judgment (Berkowitz and Walker 1967). Since people’s praise and blame are good indicators of their moral beliefs, doing what others tend to praise and avoiding what others tend to blame will be a good heuristic strategy for doing what is right and avoiding what is wrong. Therefore, unless the stakes are high enough to merit going beyond this heuristic and engaging in effortful deliberation to form an independent moral judgment, a genuine conscience

62  Brendan Dill and Stephen Darwall motive will often produce conformity to the social–moral norms expressed by others’ approval and disapproval. However, this line of thought cannot on its own provide a fully satisfactory response to the approval theorist’s skeptical argument. For the dependence of moral behavior on social norms cannot be fully explained in terms of moral belief. Sometimes, social accountability motivates behavior that violates a person’s explicit moral beliefs. The most famous and disturbing demonstration of this fact is provided by Stanley Milgram’s obedience experiments, in which a majority of subjects were willing to follow the experimenter’s orders to deliver painful electric shocks to another participant. When asked about this situation, most people say that it is morally wrong to continue to shock the victim past the point of danger, and insist that they would defy the experimenter’s orders out of moral conviction. But when this conviction is put to the test, most people will obey the experimenter to the point of torturing another person, rather than acting on their moral belief that doing so is wrong (Milgram 1974). Further evidence that the influence of social accountability overreaches that of moral belief is provided by Kwame Anthony Appiah’s fascinating recent study of moral revolutions (Appiah 2010). Appiah investigates three rapid, society-wide changes in moral behavior:  the abandonment of dueling in nineteenth-century England, the abandonment of footbinding in China, and the abolition of slavery. He summarizes his observations as follows: Arguments against each of these practices were well known and clearly made a good deal before they came to an end . . . Whatever happened when these immoral practices ceased, it wasn’t, so it seemed to me, that people were bowled over by new moral arguments. Dueling was always murderous and irrational; footbinding was always painfully crippling; slavery was always an assault on the humanity of the slave. (Appiah 2010, xii)

Instead of being abandoned on the basis of moral argument, Appiah contends that these practices were defeated by social disapproval. The adoption of dueling by working-class men made it appear vulgar and ridiculous, forcing aristocratic gentlemen to disdain the practice that once embodied their elite code of honor. Similarly, the centuries-old Chinese practice of binding women’s feet suddenly disappeared once China was mocked for it on the international stage. The British labor class spearheaded the abolitionist movement because they saw slavery as an insult to the dignity of their profession, manual labor. What ended these immoral practices were real, on-the-ground norms of social approval and disapproval. These social norms determined moral behavior independently from moral belief: People did not act on their moral beliefs that dueling and footbinding were wrong until these practices were also socially condemned. Doesn’t this show that it was the fear of disapproval, not genuine moral conscience, that motivated these revolutions in moral behavior? Not necessarily. Since the conscience motive is just one desire among many that compete for control over an agent’s behavior, it will only fully govern behavior when

Moral Psychology as Accountability  63 some feature of the agent’s situation makes moral considerations sufficiently salient. The accountability theory holds that moral obligations are represented as legitimate interpersonal demands enforced by warranted attitudes of blame. So, what could be better placed to make one’s moral obligations salient than actually expressed interpersonal demands and blame? Being actually held to moral standards by others makes one’s accountability for complying with warranted demands much more salient; being actually blamed by others gives one a vivid experience of one’s blameworthiness. So, even if conscience is a desire to avoid blameworthiness rather than actual blame, and to comply with those demands that are justified rather than those that are actually made, actual blame and actual demands can spur moral behavior by means of making motivationally salient the legitimate demands to which one is subject and the blameworthiness that their violation would entail. In short, we claim that moral behavior depends on one’s actual context of social ­accountability because salient cues of social accountability are usually required to activate the conscience motive sufficiently for it to control behavior. There is some independent confirmation for the hypothesis that cues of social accountability automatically spur conscience into action. Two studies have found that merely presenting an image of two eyes elicits significantly more moral behavior from subjects. One study found that subjects playing the Dictator Game on a computer with stylized eyes in the background are significantly more generous than controls (Haley and Fessler 2005). The second study replicated this result in a real-life situation: Subjects were given the opportunity to serve themselves freely available coffee, which they were asked to pay for in an “honesty box.” When a poster displaying a pair of eyes was placed behind the coffee dispenser, subjects paid almost three times as much for their coffee as when a control poster was displayed (Bateson, Nettle, and Roberts 2006). The empirical premises of the approval theorist’s skeptical argument can all be alternatively explained by the mechanism that underlies these priming studies. Consider first the studies showing that moral behavior tracks perceptions of social approval. Over and above the non-negligible influence of social norms on moral belief, these norms will exert independent influence on moral behavior. The awareness that others will disapprove of littering, for instance, will make motivationally salient the fact that littering is wrong, and thereby activate a conscience-based desire not to litter (Reno Cialdini, and Kallgren 1993; Kallgren, Reno, and Cialdini 2000). Similarly, the blame expressed by punishments of free riding in the public goods game makes vivid the fact that free riding is blameworthy, and thereby activates a conscience-based desire to cooperate. A similar explanation can be applied to Appiah’s historical findings. Though there were well-known moral arguments against dueling, footbinding, and slavery, the fact that these practices were socially condoned made these abstract moral principles easy to forget. (The practice of eating meat may be a modern-day analogue.) Furthermore, these practices were so culturally entrenched that refraining from them attracted active social disapproval. This generated an appearance of obligation to participate in the practice that

64  Brendan Dill and Stephen Darwall was far more motivationally salient than any countervailing moral arguments. These practices ended when the social sanctions for non-participation broke down, as when participation in dueling was no longer a mark of high social status, and when the practices themselves came to attract social condemnation, as when China was mocked for footbinding on the international stage. The first change dissipated the powerful appearance of an obligation to participate in the practice; the second change made the moral reprehensibility of the practice salient enough to motivate disengagement from it. Consider, finally, the Milgram experiments. Milgram’s subjects were willing to undertake the aversive task of delivering shocks because of the forceful demands of the experimenter, to whom they were held personally accountable. Even if the subjects believed these demands to be on balance unjustified, they nonetheless were in the grip of a strong appearance of their being justified (cf. Gibbard 1985, 15–17). These subjects disobeyed the experimenter, however, once their accountability to the shock victim was made more salient. In the first study, the subjects who did disobey the experimenter did so only once the victim protested by banging on the wall. When the victim was placed in the same room as the subjects, rendering those subjects directly accountable to the victim as well as to the experimenter, obedience of the experimenter decreased by almost 40 percent, to a minority (Milgram 1974). And when a confederate subject, also ordered to deliver shocks, vocally defied the experimenter, thereby undermining the authority of his demands, subjects overwhelmingly defied the experimenter as well (36 out of 40; Milgram 1965). In sum, though subjects in the Milgram paradigm violated their own reflective moral convictions, this may be because their conscience motives were hostage to the powerful moral appearances generated by the demands of the experimenter.12 We therefore conclude that the dependence of morally conscientious behavior on social norms is compatible with the accountability theory’s claim that such behavior is motivated by genuine moral conscience. First, social norms of approval and disapproval can serve as a heuristic guide to moral rightness and wrongness, thus influencing people’s moral behavior via their moral beliefs. Second, and more importantly, being held accountable to an actual demand by one’s peers automatically activates the conscience motive to comply with legitimate demands. In the absence of social accountability, moral considerations may not be salient enough for the conscience motive to overpower other amoral motives (as in the cultures Appiah studied). And when agents are held accountable to demands they reflectively believe to be unjustified, this generates a strong appearance of moral obligation, which may be more motivationally potent than reflective moral belief (as with Milgram’s subjects). Thus the accountability theory can explain the data reviewed in this section, and thereby rebut the second argument for the approval theory. The demonstrated dependence of moral behavior on social accountability is compatible with both the accountability theory and the approval theory, and so does not support one over the other.

For further discussion of the role of accountability in the Milgram studies, see Darwall (2006, 162–71).

12

Moral Psychology as Accountability  65 Note, however, that this data does support the accountability theory over accounts of moral conscience that do not essentially implicate social accountability. The orthodox view of conscience we discussed in §3.1, which takes moral norms to be represented as mere intrapersonal standards rather than essentially interpersonal demands, cannot offer the same explanation of this data that the accountability theory can. Only by positing a conceptual link between moral obligations and social accountability can one predict that the conscience motive is selectively activated by cues of social accountability; and this prediction was essential to our explanation of the Milgram and Appiah findings. In other words, it is in virtue of the accountability theory’s unique thesis that moral obligations are represented as legitimate interpersonal demands that this theory is able to rebut the approval theorist’s arguments for skepticism about moral conscience. Insofar as one affirms the existence of moral conscience, then, one must acknowledge its essential link to social accountability.

3.3.  Argument for the Accountability Theory: Guilt as Backward-Looking Moral Conscience The major lesson of the discussion so far seems to be that the accountability and approval theories are hard to pull apart. The data that has previously been taken to support the approval theory—the moral hypocrisy and moral licensing effects, and the dependence of moral behavior on social norms—can be explained with equal adequacy by the accountability theory. So, the data reviewed so far seems equally compatible with the existence of moral conscience as construed by the accountability theory, and the brand of skepticism about moral conscience offered by the approval theory. However, the data we have so far considered has focused exclusively on forward-looking moral behavior, where agents are concerned with avoiding future violations of moral norms. Prospects for answering our question look more promising if we concentrate on how agents respond after they have committed moral wrong. Though they are difficult to pull apart in forward-looking contexts, the accountability theory and approval theories make sharply divergent predictions regarding this backward-looking moral behavior. Consider, first, the approval theory. From the perspective of the approval motive, wrongful behavior is a PR disaster to be managed. After doing wrong, approval-driven agents should seek to mitigate the negative consequences of their wrongdoing for their reputation. Strategies for accomplishing this end include distancing oneself from the wrongdoing, providing excuses, pinning the blame on someone else, or simply avoiding social attention altogether. In contrast, the accountability theory holds that agents who have done wrong will also be driven by a moral conscience motive to hold themselves accountable for their wrongdoing.13 This will involve taking responsibility for the wrongful action rather than trying to deflect responsibility to someone else. It will involve seeking out the 13 It is compatible with the accountability theory that people who have done wrong may simultaneously experience both the moral conscience motive to hold themselves accountable and the approval motive to

66  Brendan Dill and Stephen Darwall victim of one’s actions to apologize, express remorse, and make amends, perhaps by giving compensation. It will also involve a re-energized vigilance against immoral behavior, driven by the recommitment to moral standards involved in holding oneself accountable for violating those standards. Thus the moral conscience and social approval motives should produce very different patterns of behavior after an agent has committed a moral wrong. We submit that both behavioral patterns occur, and respectively accompany the emotions of guilt and shame. Guilt is the emotional reaction to wrongdoing characteristic of the moral conscience motive, and it leads to the behaviors involved in holding oneself accountable. Shame is the emotional reaction to wrongdoing characteristic of the social approval motive, and it leads to the behaviors involved in managing one’s reputation. These claims are confirmed by empirical research on guilt and shame. The literature on guilt and shame is enormous, so we will simply state the most common findings without detailing the evidence behind them (for a helpful review, see Tangney, Stuewig, and Mashek 2007). We begin with guilt. First and foremost, guilt leads its subject to take responsibility for her wrongdoing (McGraw 1987; Tangney et al. 1992; Roseman, Wiest, and Swartz 1994; Baumeister, Stillwell, and Heatherton 1995; Mandel and Dhami 2005; Fisher and Exline 2006; Tracy and Robins 2006; Tangney, Stuewig, and Mashek 2007). Second, guilt motivates its subject to make amends with the victim of wrongdoing by apologizing (Roseman et al. 1994; Baumeister et al. 1995), making reparations (Roseman, Wiest, and Swartz 1994; Lickel et  al. 2005; Brown et al. 2008; Zebel et al. 2008; Gino, Gu, and Zhong 2009; Čehajić-Clancy et al. 2011; de Hooge et al. 2011), striving to correct future behavior (Baumeister, Stillwell, and Heatherton 1995; Tangney et  al. 1996; Amodio, Devine, and Harmon-Jones 2007; Hopfensitz and Reuben 2009; Orth, Robins, and Soto 2010; Stillman and Baumeister 2010), and even self-punishing (Bastian, Jetten, and Fasoli 2011; Nelissen 2012; Inbar et  al. 2013). Guilt is characterized by other-directed empathy and concern for the victim of wrongdoing (Niedenthal, Tangney, and Gavanski 1994; Leith and Baumeister 1998; Tangney, Stuewig, and Mashek 2007; Basil, Ridgway, and Basil 2008; Yang, Yang, and Chiou 2010). Finally, guilt, whether dispositional or occurrent, leads its subject to behave more morally in general (Regan, Williams, and Sparling 1972; Cunningham, Steinberg, and Grev 1980; Montada and Schneider 1989; Tangney, Wagner, Hill-Barlow, Marschall, and Gramzow 1996; Quiles and Bybee 1997; Millar 2002; Stuewig and McCloskey 2005; Tangney, Stuewig, and Mashek 2007; Hopfensitz and Reuben 2009; Kochanska et al. 2009; Cohen et al. 2011; Polman and Ruttan 2012; for a dissenting view, see de Hooge et al. 2011). In stark contrast with guilt, shame leads its subject to avoid responsibility for her wrongdoing by distancing herself from the event and blaming others for her wrongdoing (Tangney et al. 1992; Tangney, Miller, et al. 1996; Ferguson et al. 1999; Johns, manage the damage to their reputations. Remember: We only claim that the moral conscience motive exists, not that the approval motive doesn’t.

Moral Psychology as Accountability  67 Schmader, and Lickel 2005; Lickel et al. 2005). Rather than motivating apology and reconciliation, shame leads to negative interpersonal consequences such as aggression and social withdrawal (Tangney et al. 1992; Tangney, Wagner, et al. 1996; Orth et al. 2010; Cohen et al. 2011). Rather than eliciting empathy and other-directed concern, shame is associated with a focus on one’s self-image and public reputation (Niedenthal et al. 1994; Tangney 1995; Smith et al. 2002; Bagozzi, Verbeke, and Gavino 2003; Lickel et al. 2005; Tangney, Stuewig, and Mashek 2007). As shame is centered on one’s overall reputation, it causes its subject to focus on her general traits; while guilt, centered on one’s accountability for a particular wrongful action, draws its subject’s attention to her actions rather than her overall traits (Niedenthal, Tangney, and Gavanski 1994). Finally, unlike guilt, neither dispositional nor occurrent shame has positive associations with moral behavior (Tangney, Wagner, et al. 1996; Quiles and Bybee 1997; Tangney, Stuewig, and Mashek 2007, 354; Cohen et al. 2011). In sum, while shame motivates those behaviors we would expect to be produced by the approval motive, guilt has been shown in many empirical studies to produce exactly the behaviors we would expect to be produced by a genuine conscience motive. Thus we take the existence of a genuine backward-looking moral conscience motive, and its independence from the approval motive, to be demonstrated by the evidence showing the existence of guilt and its independence from shame. Does this mean that we should accept the existence of forward-looking moral conscience as well? We think so. In fact, we hold that backward-looking and forward-looking moral conscience are simply manifestations of a single moral conscience motive in different contexts. The conscience goal has the same content in both contexts—to fulfill one’s moral obligations and so do one’s part in upholding morality—but achieving this goal requires different actions depending on whether one has already done wrong, or merely needs to avoid future wrongdoing. Thus evidence for the existence of backward-looking moral conscience is pari passu evidence for the existence of forward-looking moral conscience. This unity thesis regarding backward-looking and forward-looking conscience is supported by the data just reviewed showing that guilt leads to more moral behavior in general, even in areas unrelated to the guilt-inducing action (see especially Regan, Williams, and Sparling 1972). If backward-looking and forward-looking conscience are two manifestations of the same motive, then the activation of the conscience motive in a backward-looking context should also lead to more moral forward-looking behavior. We thus conclude that the experimental evidence on guilt demonstrates the existence of moral conscience in general. This completes our argument for the accountability theory of conscience. We have argued that genuine moral conscience exists:  The objections that have motivated moral conscience skepticism are not sound (§3.2), and the empirical research on guilt and shame demonstrates the existence of the conscience motive while dissociating it from the egoistic motive to gain social approval (§3.3). Furthermore, we have argued that the content of the conscience motive should be understood in terms of

68  Brendan Dill and Stephen Darwall social accountability. Moral obligations are represented as legitimate interpersonal demands; thus the conscience motive is a motive to comply with warranted interpersonal demands to which one may be legitimately held accountable. This unique insight of the accountability theory explains the widely confirmed observation that moral behavior depends on social context, and specifically depends upon the demands to which agents are actually held accountable by their peers (§3.2.2). Conscience motivates us to be moral, and to be moral is to be accountable.

4.  What is Distinctive about Morality? We have contended that conceiving of morality in terms of accountability provides a unified understanding of the emotions, attitudes, and motives that are targeted on acting rightly and avoiding moral wrong. The implicit goal of all moral motivation is to hold people—oneself or others—accountable for compliance with moral requirements. Moral emotions and attitudes all aim to uphold morality, conceived as demands with which we are accountable to one another for complying as equal moral persons. In this final section, we wish to consolidate these points by refocusing on what is distinctive about morality as an ethical conception, and, consequently, on what distinguishes the psychological items we have discussed that concern it. It is important to emphasize again that facts about morality and moral right and wrong are normative and thus distinct from any descriptive psychological or social fact. Morality, in this normative sense, is what moral judgment and motivation are about. In a recent article on “Morality,” Jonathan Haidt and Selin Kesebir define “moral systems” as “interlocking sets of values, virtues, norms, practices, identities, institutions, technologies, and evolved psychological mechanisms that work together to suppress or regulate selfishness and make cooperative social life possible” (Haidt and Kesebir 2010, 800). By “values” and “norms,” Haidt and Kesebir evidently mean psycho-social phenomena. In this sense, a value or norm consists in people valuing something or in their holding or accepting some norm. To hold or accept a norm or value, however, is to be committed to something normative. It is to hold that something or other is valuable or, for example, that some kinds of conduct are morally wrong. In this latter case, it is to hold a normative belief or attitude about morality, for example, that some kind of conduct really is morally wrong. We might think of morality, in the sense we have been discussing, as consisting in valid moral norms that make moral judgments and beliefs true or false. Moreover, as we also noted at the outset, not all normative beliefs and attitudes concern morality. What makes for a desirable life or promotes human welfare, for example, is a normative question, but it is not, in itself, a moral issue. Even if acting morally is an essential part of well-being, the proposition that it is so is not a proposition of morality. These observations raise two questions regarding the distinctiveness of morality. First, what distinguishes moral facts from other domains of normative fact? Second,

Moral Psychology as Accountability  69 what distinguishes the psychological attitudes, emotions, and motivations that are concerned with morality—our moral psychology—from other, non-moral normative and evaluative attitudes? Since the attitudes that make up our moral psychology are attitudes about morality, our answer to the first of these questions will constrain our answer to the second. Haidt and Kesebir’s definition of “moral systems” offers an answer to our second, descriptive question. However, we think that this definition fails to capture what is distinctive about moral psychology, because it fails to capture what is distinctive about morality as a normative concept. Morality is a more specific normative notion than that of just any norms and values that are concerned with “suppress[ing] or regulat[ing] selfishness and mak[ing] cooperative social life possible.” Take, for example, norms of esteem and honor that define a hierarchy of status in an honor society. We might imagine these working to regulate selfishness and foster forms of cooperation. However, the normative ideas that would be involved would be those of the honorable and the estimable, of what warrants honor, deference, and esteem, on the one hand, and contempt or disdain, on the other. These are different normative ideas from those involved in morality, as can be seen quite vividly by considering Nietzsche’s critique of morality in On the Genealogy of Morality (Nietzsche 1887/2006). Although Nietzsche sharply criticizes morality’s concepts of moral “good” and “evil,” he has no objection to the concepts of “good” and “bad” of an “aristocratic” ethos that structures a hierarchy of status and honor. According to Nietzsche’s etymology, the term “good” originated with the “nobles” themselves to connote qualities that fit someone for high status and to contrast with qualities that are lowly and base (Nietzsche 1887/2006, 11). We don’t have to accept Nietzsche’s etymology to recognize the conceptual distinction he is marking between ideas of the noble and the base, on the one hand, and those of moral right and wrong, good and evil, on the other. This normative conceptual distinction is reflected, moreover, in the psychological difference between contempt and shame, on the one hand, and blame or condemnation and guilt, on the other. What is base or low is what warrants contempt and shame, or at least the form of shame that portrays one to oneself as contempt portrays one to the person who views one with contempt, namely, as contemptible, low, or base. Culpable wrongdoing, on the other hand, is what warrants the accountability-seeking emotions of condemnation or guilt, where guilt portrays one to oneself as condemnation portrays one to someone who condemns one, namely, as worthy of condemnation or moral blame. In theory, an aristocratic honor code might hold to be contemptible or low the very same actions that morality condemns as morally wrong. Take Haidt and Kesebir’s examples of selfishness and uncooperativeness. Morality condemns excessive selfishness or free riding on the cooperative efforts of others as morally wrong. But we can easily imagine an aristocratic ethos holding selfish free riding to be contemptible or base also. Our point is that whereas morality condemns such conduct in terms of accountability, as blameworthy lacking excuse, the normative notions involved in an

70  Brendan Dill and Stephen Darwall honor code are fundamentally different. The attitudes and emotions they bring into play, contempt and shame, differ from the accountability-seeking attitudes implicated in morality. When a noble looks down on a serf with contempt, he hardly aims to have the serf hold himself accountable for his contemptible state. The emotion that responds to contempt is not guilt, but shame, which, as we have noted (§3.3), shows itself in very different ways than guilt does. Moreover, just as moral norms differ conceptually from those of honor and esteem, so also are they conceptually distinct from norms of purity and disgust. It is a favorite claim of Haidt’s that liberals tend to think of morality in terms of fairness and equality, whereas conservatives also include loyalty, respect for authority, and purity among morality’s “foundations” (Haidt and Kesebir 2010, 821–3). Whereas liberals are skeptical of “disgust-based” notions of purity, Haidt claims that disgust is a moral emotion and, therefore, that purity is no less a foundation of morality than are equality and fairness. For purposes of discussion, we can simply stipulate that an action that violates a purity taboo, consensual incest, for example, is morally wrong, indeed that it is wrong for that very reason. Our point is that the proposition that incest is morally wrong is a different kind of normative claim than the proposition that it is impure, or violates a taboo, or that it is disgusting. For present purposes, we can even allow that there is a genuine normative concept of the disgusting—of what justifies disgust or to which disgust is a fitting response—that is distinct from the concept of what actually causes disgust. Our point remains that the proposition that it is morally wrong to do something disgusting is an additional moral claim that goes beyond the claim that such an action warrants disgust. It is the claim that the disgusting action also warrants the accountability-seeking attitudes of condemnation and guilt.14 Consideration of these examples shows that Haidt and Kesebir’s functional definition of morality is too broad. Norms of honor and contempt, or of purity and disgust, count as “moral systems” by their definition, insofar as they help to suppress selfishness and promote cooperation. However, lumping these together into a single category with moral norms of blame and guilt elides an important distinction: between norms that simply evaluate people, such as those of honor and purity, and those that hold people responsible, in the sense of asking for a response. It is this latter feature that, we claim, makes morality distinctive. Thus we propose an alternative definition of morality. Morality, understood as a normative phenomenon, is that set of demands to which we may legitimately hold one another accountable with blame, and hold ourselves accountable with guilt. Moral attitudes, emotions, and motives are those the contents of which essentially refer to morality so understood. A particular society’s morality, understood as a descriptive 14 We can even agree that there is such a thing as moral disgust. However, it seems that any such response itself presupposes the concepts of moral right and wrong and, therefore, on our analysis, the independent idea of actions warranting the accountability-seeking emotions of condemnation and guilt.

Moral Psychology as Accountability  71 phenomenon, is that set of norms to which the members of the society actually hold one another accountable with blame, and hold themselves accountable with guilt. We should stress that according to morality as accountability, what makes a set of normative beliefs and practices moral beliefs and practices is not their contents—the actions they prescribe and proscribe—but the distinctively accountability-seeking attitudes and responses they take those actions to warrant. The definitions of “morality” to which Haidt and Kesebir object are all content-based definitions, such as Turiel’s definition of moral judgments as “prescriptive judgments of justice, rights, and welfare pertaining to how people ought to relate to each other” (Turiel 1983, 3). Haidt and Kesebir’s objection is that these characterizations of morality typically limit the potential contents of moral beliefs to those endorsed by Western liberals. Our account avoids this objection, since it is content independent. To illustrate, notice that we can agree with Haidt and Kesebir that conservatives’ beliefs and attitudes are no less moral, that is beliefs about morality, than are liberals’, regardless of whose beliefs are more correct. What is more, we can say what the difference between the conservative and the liberal amounts to: a disagreement about what standards we can hold each other accountable to. A liberal might be just as disgusted by some behavior as a conservative, but disagree with a conservative’s belief that the conduct is morally wrong, since only the latter thinks that the behavior warrants blame as well as disgust. What is distinctive about morality as a normative idea, therefore, is neither its content nor its performing the functions of regulating social order, curbing selfishness, or enabling social cooperation in just any way. It is the specific way morality purports to regulate. Morality’s distinctiveness, and its distinctive form of social regulation, are explained by its conceptual tie to mutual accountability and to the accountability-seeking attitudes of blame and guilt. We have been arguing that appreciating this fact enables a unified and explanatory psychological account of moral thought, emotion, and motivation.

Acknowledgments We thank Jonathan Haidt, Joshua Knobe, and three anonymous reviewers for helpful comments and criticism.

References Amodio, D. M., Devine, P. G., and Harmon-Jones, E. 2007. A dynamic model of guilt: Implications for motivation and self-regulation in the context of prejudice. Psychological Science, 18(6), 524–30. Annas, J. 1995. Prudence and morality in ancient and modern ethics. Ethics, 105(2), 241–257.

72  Brendan Dill and Stephen Darwall Anscombe, G. E. M. 1958. Modern moral philosophy. Philosophy, 33(124), 1–19. Appiah, K. A. 2010. The Honor Code: How Moral Revolutions Happen. New York: W.W. Norton & Company. Asch, S. E. 1955. Opinions and social pressure. Scientific American, 193(4), 31–5. Bachman, G. F., and Guerrero, L. K. 2006. Forgiveness, apology, and communicative responses to hurtful events. Communication Reports, 19(1), 45–56. Bagnoli, C. (ed.) 2011. Morality and the Emotions. Oxford: Oxford University Press. Bagozzi, R. P., Verbeke, W., and Gavino, J. C. J. 2003. Culture moderates the self-regulation of shame and its effects on performance: The case of salespersons in the Netherlands and the Philippines. Journal of Applied Psychology, 88(2), 219–33. Baron, J., Gowda, R., and Kunreuther, H. 1993. Attitudes toward managing hazardous waste: What should be cleaned up and who should pay for it? Risk Analysis, 13(2), 183–92.

Baron, J., and Ritov, I. 1993. Intuitions about penalties and compensation in the context of tort law. Journal of Risk and Uncertainty, 7(1), 17–33. Basil, D. Z., Ridgway, N. M., and Basil, M. D. 2008. Guilt and giving: A process model of empathy and efficacy. Psychology and Marketing, 25(1), 1–23. Bastian, B., Jetten, J., and Fasoli, F. 2011. Cleansing the soul by hurting the flesh: The guilt-reducing effect of pain. Psychological Science, 22(3), 334–5. Bateson, M., Nettle, D., and Roberts, G. 2006. Cues of being watched enhance cooperation in a real-world setting. Biology Letters, 2, 412–14. Batson, C. D. 2008. Moral masquerades: Experimental exploration of the nature of moral motivation. Phenomenology and the Cognitive Sciences, 7, 51–66. Batson, C. D., Chao, M. C., and Givens, J. M. 2009. Pursuing moral outrage: Anger at torture. Journal of Experimental Social Psychology, 45(1), 155–60. Batson, C. D., Collins, E., and Powell, A. A. 2006. Doing business after the fall: The virtue of moral hypocrisy. Journal of Business Ethics, 66(4), 321–35. Batson, C. D., Kennedy, C. L., Nord, L. A., Stocks, E. L., Fleming, D. A., Marzette, C. M., Lishner, D. A., Hayes, R. E., Kolchinsky, L. M., Zerger, T. 2007. Anger at unfairness: is it moral outrage? European Journal of Social Psychology, 37(6), 1272–85. Batson, C. D., Klein, T. R., Highberger, L., and Shaw, L. L. 1995. Immorality from empathy-induced altruism: When compassion and justice conflict. Journal of Personality and Social Psychology, 68(6), 1042–54. Batson, C. D., Kobrynowicz, D., Dinnerstein, J., Kampf, H., and Wilson, A. 1997. In a very different voice: Unmasking moral hypocrisy. Journal of Personality and Social Psychology, 72(6), 1335–48.

Moral Psychology as Accountability  73 Batson, C.  D., and Shaw, L.  L. 1991. Evidence for altruism:  Toward a pluralism of prosocial motives. Psychological Inquiry, 22, 107–22. Batson, C. D., and Thompson, E. R. 2001. Why don’t moral people act morally? Motivational considerations. Current Directions in Psychological Science, 102, 54–7. Batson, C. D., Thompson, E. R., and Chen, H. 2002. Moral hypocrisy: Addressing some alternatives. Journal of Personality and Social Psychology, 83(2), 330–9. Batson, C. D., Thompson, E., and Seuferling, G. 1999. Moral hypocrisy: Appearing moral to oneself without being so. Journal of Personality and Social Psychology, 77(3), 525–37. Baumeister, R. F., Stillwell, A. M., and Heatherton, T. F. 1995. Personal narratives about guilt: Role in action control and interpersonal relationships. Basic and Applied Social Psychology, 17(1), 173–98.

Baumeister, R. F., Vohs, K. D., DeWall, C. N., and Zhang, L. 2007. How emotion shapes behavior:  feedback, anticipation, and reflection, rather than direct causation. Personality and Social Psychology Review, 11(2), 167–203. Bell, M. 2013. Hard Feelings: The Moral Psychology of Contempt. New York: Oxford University Press. Berker, S. 2009. The normative insignificance of neuroscience. Philosophy & Public Affairs, 37(4): 293–329. Berkowitz, L., and Walker, N. 1967. Laws and moral judgments. Sociometry, 30(4), 410–22.

Bernhard, H., Fischbacher, U., and Fehr, E. 2006. Parochial altruism in humans. Nature, 442, 912–15. Bottom, W., Gibson, K., and Daniels, S. 2002. When talk is not cheap: Substantive penance and expressions of intent in rebuilding cooperation. Organization Science, 13(5), 497–513. Brown, R., González, R., Zagefka, H., Manzi, J., and Cehajic, S. 2008. Nuestra culpa: Collective guilt and shame as predictors of reparation for historical wrongdoing. Journal of Personality and Social Psychology, 94(1), 75–90. Bushman, B. J., Baumeister, R. F., and Phillips, C. M. 2001. Do people aggress to improve their mood? Catharsis beliefs, affect regulation opportunity, and aggressive responding. Journal of Personality and Social Psychology, 81(1), 17–32. Cain, D. M., Loewenstein, G., and Moore, D. A. 2005. The dirt on coming clean: Perverse effects of disclosing conflicts of interest. The Journal of Legal Studies, 34(1), 1–25. Carlsmith, K. M. 2006. The roles of retribution and utility in determining punishment. Journal of Experimental Social Psychology, 42(4), 437–51. Carlsmith, K.  M. 2008. On justifying punishment:  The discrepancy between words and actions. Social Justice Research, 21(2), 119–37.

74  Brendan Dill and Stephen Darwall Carlsmith, K. M., Darley, J. M., and Robinson, P. H. 2002. Why do we punish? Deterrence and just deserts as motives for punishment. Journal of Personality and Social Psychology, 832, 284– 99. Carlsmith, K. M., and Sood, A. M. 2009. The fine line between interrogation and retribution. Journal of Experimental Social Psychology, 45(1), 191–6. Carlsmith, K. M., Wilson, T. D., and Gilbert, D. T. 2008. The paradoxical consequences of revenge. Journal of Personality and Social Psychology, 95(6), 1316–24. Čehajić-Clancy, S., Effron, D. A., Halperin, E., Liberman, V., and Ross, L. D. 2011. Affirmation, acknowledgment of in-group responsibility, group-based guilt, and support for reparative measures. Journal of Personality and Social Psychology, 101(2), 256–70. Chartrand, T.  L. 2001. Mystery moods and perplexing performance:  Consequences of succeeding and failing at a nonconscious goal. Manuscript submitted for publication. Coates, D. J., and Tognazzini, N. A. (eds.). 2013. Blame: Its Nature and Norms. Oxford: Oxford University Press. Cohen, T. R., Wolf, S. T., Panter, A. T., and Insko, C. A. 2011. Introducing the GASP scale: A new measure of guilt and shame proneness. Journal of Personality and Social Psychology, 100(5), 947–66. Cunningham, M., Steinberg, J., and Grev, R. 1980. Wanting to and having to help:  Separate motivations for positive mood and guilt-induced helping. Journal of Personality and Social Psychology, 38(2), 181–92. Darley, J. M., Carlsmith, K. M., and Robinson, P. H. 2000. Incapacitation and just deserts as motives for punishment. Law and Human Behavior, 246, 659–83. D’Arms, J., and Jacobson, D. 2000. Sentiment and value. Ethics, 110 (4): 722–48. Darwall, S. 2006. The Second-Person Standpoint:  Morality, Respect, and Accountability. Cambridge, MA: Harvard University Press. Darwall, S. 2012. Bipolar obligation. In R. Shafer-Landau (ed.), Oxford Studies in Metaethics, vol. 7, pp. 333–67. Oxford:  Oxford University Press. Darwall, S. 2013a. Morality, Authority, and Law:  Essays in Second-Personal Ethics I. Oxford: Oxford University Press. Darwall, S. 2013b. Honor, History, and Relationship:  Essays in Second-Personal Ethics II. Oxford: Oxford University Press. de Hooge, I. E., Nelissen, R. M. A., Breugelmans, S. M., and Zeelenberg, M. 2011. What is moral about guilt? Acting “prosocially” at the disadvantage of others. Journal of Personality and Social Psychology, 100(3), 462–73. de Jong, P. J., Peters, M. L., and De Cremer, D. 2003. Blushing may signify guilt: Revealing effects of blushing in ambiguous social situations. Motivation and Emotion, 27(3), 225–49.

Moral Psychology as Accountability  75 Denzler, M., Förster, J., and Liberman, N. 2009. How goal-fulfillment decreases aggression. Journal of Experimental Social Psychology, 45(1), 90–100. DeScioli, P., and Kurzban, R. 2009. Mysteries of morality. Cognition, 112(2), 281–99. Effron, D. A., Cameron, J. S., and Monin, B. 2009. Endorsing Obama licenses favoring Whites. Journal of Experimental Social Psychology, 45(3), 590–3. Effron, D. A., Miller, D. T., and Monin, B. 2012. Inventing racist roads not taken: The licensing effect of immoral counterfactual behaviors. Journal of Personality and Social Psychology, 103(6), 916–32. Effron, D. A., Monin, B., and Miller, D. T. 2013. The unhealthy road not taken: Licensing indulgence by exaggerating counterfactual sins. Journal of Experimental Social Psychology, 493, 573–8. Ellsworth, P. C., and Ross, L. 1983. Public opinion and capital punishment: A close examination of the views of abolitionists and retentionists. Crime & Delinquency, 29(1), 116–69. Fehr, E., and Fischbacher, U. 2004. Third-party punishment and social norms. Evolution and Human Behavior, 25(2), 63–87. Fehr, E., and Gächter, S. 2000. Cooperation and punishment in public goods experiments. The American Economic Review, 90(4), 980–94. Fehr, E., Fischbacher, U., and Gächter, S. 2002. Strong reciprocity, human cooperation, and the enforcement of social norms. Human Nature, 13(1), 1–25. Fehr, R., Gelfand, M. J., and Nag, M. 2010. The road to forgiveness: A meta-analytic synthesis of its situational and dispositional correlates. Psychological Bulletin, 136(5), 894–914. Fehr, E., and Rockenbach, B. 2003. Detrimental effects of sanctions on human altruism. Nature, 422(6928), 137–40. Fekadu, Z., and Kraft, P. 2002. Expanding the theory of planned behavior: The role of social norms and group identification. Journal of Health Psychology, 7(1), 33–43. Ferguson, T. J., Stegge, H., Miller, E. R., and Olsen, M. E. 1999. Guilt, shame, and symptoms in children. Developmental Psychology, 35(2), 347–57. Fisher, M. L., and Exline, J. J. 2006. Self-forgiveness versus excusing: The roles of remorse, effort, and acceptance of responsibility. Self and Identity, 5(2), 127–46. Fitness, J., and Peterson, J. 2008. Punishment and forgiveness in close relationships: An evolutionary, social-psychological perspective. In J. P.  Forgas and J. Peterson (eds.), Social Relationships: Cognitive, Affective, and Motivational Processes, pp. 255–69. New York: Taylor & Francis Group, LLC.

76  Brendan Dill and Stephen Darwall Foot, P. 1972. Morality as a system of hypothetical imperatives. The Philosophical Review, 81, 305– 16. “Forgive.” 1983. In Webster’s New Universal Unabridged Dictionary. New York: Dorset & Baker. Förster, J., Liberman, N., and Friedman, R. S. 2007. Seven principles of goal activation: A systematic approach to distinguishing goal priming from priming of non-goal constructs. Personality and Social Psychology Review, 11(3), 211–33. Förster, J., Liberman, N., and Higgins, E. T. 2005. Accessibility from active and fulfilled goals. Journal of Experimental Social Psychology, 41(3), 220–39. Gächter, S., and Herrmann, B. 2008. Reciprocity, culture and human cooperation:  Previous insights and a new cross-cultural experiment. Philosophical Transactions of the Royal Society B:  Biological Sciences, 364(1518), 791–806. Gauthier, D. 1986. Morals by Agreement. Oxford: Oxford University Press. Gerber, A. S., Green, D. P., and Larimer, C. W. 2008. Social pressure and voter turnout: Evidence from a large-scale field experiment. The American Political Science Review, 102(1), 33–48.

Gerber, A. S., and Rogers, T. 2009. Descriptive social norms and motivation to vote: Everybody’s voting and so should you. The Journal of Politics, 71(1), 178–91. Gibbard, A. 1985. Moral judgment and the acceptance of norms. Ethics, 96(1), 5–21. Gibbard, A. 1990. Wise Choices, Apt Feelings:  A  Theory of Normative Judgment. Cambridge, MA: Harvard University Press. Gino, F., Gu, J., and Zhong, C. B. 2009. Contagion or restitution? When bad apples can motivate ethical behavior. Journal of Experimental Social Psychology, 45(6), 1299–302. Gold, G. J., and Weiner, B. 2000. Remorse, confession, group identity, and expectancies about repeating a transgression. Basic and Applied Social Psychology, 22(4), 291–300. Goldstein, N. J., Cialdini, R. B., and Griskevicius, V. 2008. A room with a viewpoint: Using social norms to motivate environmental conservation in hotels. The Journal of Consumer Research, 35(3), 472–82. Gollwitzer, M., and Bushman, B. J. 2011. Do victims of injustice punish to improve their mood? Social Psychological and Personality Science, 00(0), 1–9. Gollwitzer, M., and Denzler, M. 2009. What makes revenge sweet: Seeing the offender suffer or delivering a message? Journal of Experimental Social Psychology, 45(4), 840–4. Gollwitzer, M., Meder, M., and Schmitt, M. 2010. What gives victims satisfaction when they seek revenge? European Journal of Social Psychology, 41(3), 364–74. Gordijn, E.  H., Yzerbyt, V., Wigboldus, D., and Dumont, M. 2006. Emotional reactions to harmful intergroup behavior. European Journal of Social Psychology, 36(1), 15–30.

Moral Psychology as Accountability  77 Grasmick, H. G. and Green, D. E. 1980. Legal punishment, social disapproval and internalization as inhibitors of illegal behavior. The Journal of Criminal Law and Criminology, 71(3), 325–35. Greene, J.  D. 2007. The secret joke of Kant’s soul. In W. Sinnott-Armstrong (ed.), Moral Psychology, vol. 3:  The Neuroscience of Morality:  Emotion, Disease, and Development, pp. 359–71. Cambridge, MA: MIT Press. Griskevicius, V., Tybur, J. M., Gangestad, S. W., Perea, E. F., Shapiro, J. R., and Kenrick, D. T. 2009. Aggress to impress: Hostility as an evolved context-dependent strategy. Journal of Personality and Social Psychology, 96(5), 980–94. Haidt, J. 2003. The moral emotions. In R. J. Davidson, K. R. Scherer, and H. H. Goldsmith (eds.), Handbook of Affective Sciences, pp. 852–70. Oxford: Oxford University Press. Haidt, J., and Kesebir, S. 2010. Morality. In S. T.  Fiske, D. T.  Gilbert, and G. Lindzey (eds.), Handbook of Social Psychology, 5th edn., pp. 797–832. Hoboken, NJ: John Wiley & Sons, Inc.

Haidt, J. 2012. The Righteous Mind:  Why Good People Are Divided by Politics and Religion. New York: Random House, Inc. Haley, K. J., and Fessler, D. M. T. 2005. Nobody’s watching? Subtle cues affect generosity in an anonymous economic game. Evolution and Human Behavior, 26(3), 245–56. Hannon, P.  A., Rusbult, C.  E., Finkel, E.  J., and Kamashiro, M. 2010. In the wake of betrayal: Amends, forgiveness, and the resolution of betrayal. Personal Relationships, 17(2), 253– 78. Herman, B. 1981. On the value of acting from the motive of duty. The Philosophical Review, 90(3), 359–82. Hieronymi, P. 2004. The force and fairness of blame. Philosophical Perspectives, 18(1), 115–148.

Hopfensitz, A., and Reuben, E. 2009. The importance of emotions for the effectiveness of social punishment. The Economic Journal, 119(540), 1534–59. Hurley, E. and Macnamara, C. 2010. Beyond belief: Toward a theory of the reactive attitudes. Philosophical Papers, 39(3), 373–99. Hutcherson, C. A., and Gross, J. J. 2011. The moral emotions: A social–functionalist account of anger, disgust, and contempt. Journal of Personality and Social Psychology, 100(4), 719–37.

Inbar, Y., Pizarro, D. A., Gilovich, T., and Ariely, D. 2013. Moral masochism: On the connection between guilt and self-punishment. Emotion, 13(1), 14–18. Janssen, M. A., Holahan, R., Lee, A., and Ostrom, E. 2010. Lab experiments for the study of social-ecological systems. Science, 328(5978), 613–17. Johns, M., Schmader, T., and Lickel, B. 2005. Ashamed to be an American? The role of identification in predicting vicarious shame for anti-Arab prejudice after 9–11. Self and Identity, 4(4), 331–48.

78  Brendan Dill and Stephen Darwall Kahan, D. M. 1997. Social influence, social meaning, and deterrence. Virginia Law Review, 83(2), 349–95. Kallgren, C. A., Reno, R. R., and Cialdini, R. B. 2000. A focus theory of normative conduct: When norms do and do not affect behavior. Personality and Social Psychology Bulletin, 26, 1002–1012. Kant, I. 1785/1996. Groundwork of the Metaphysics of Morals. In Practical Philosophy, trans. and ed. M. J. Gregor. Cambridge: Cambridge University Press. References are to page numbers of the Preussische Akademie edition. Keller, L.  B., Oswald, M.  E., Stucki, I., and Gollwitzer, M. 2010. A closer look at an eye for an eye:  laypersons’ punishment decisions are primarily driven by retributive motives. Social Justice Research, 23(2–3), 99–116. Kelley, D. L., and Waldron, V. R. 2005. An investigation of forgiveness-seeking communication and relational outcomes. Communication Quarterly, 53(3), 339–58. Kelly, D., Stich, S., Haley, K. J., Eng, S. J., and Fessler, D. M. T. 2007. Harm, affect, and the moral/ conventional distinction. Mind and Language, 22(2), 117–31. Khan, U., and Dhar, R. 2006. Licensing effect in consumer choice. Journal of Marketing Research, 43, 259–66. Kochanska, G., Barry, R. A., Jimenez, N. B., Hollatz, A. L., and Woodard, J. 2009. Guilt and effortful control:  Two mechanisms that prevent disruptive developmental trajectories. Journal of Personality and Social Psychology, 97(2), 322–33. Korsgaard, C. 1986. Skepticism about practical reason. The Journal of Philosophy, 83(1), 5–25.

Kouchaki, M. 2011. Vicarious moral licensing: The influence of others’ past moral actions on moral behavior. Journal of Personality and Social Psychology, 101(4), 702–15. Kurzban, R., DeScioli, P., and O’Brien, E. 2007. Audience effects on moralistic punishment. Evolution and Human Behavior, 28(2), 75–84. Leith, K. P., and Baumeister, R. F. 1998. Empathy, shame, guilt, and narratives of interpersonal conflicts: Guilt-prone people are better at perspective taking. Journal of Personality, 66(1), 1–37. Lemay, E. P., Overall, N. C., and Clark, M. S. 2012. Experiences and interpersonal consequences of hurt feelings and anger. Journal of Personality and Social Psychology, 103(6), 982–1006.

Leonard, D. J., Mackie, D. M., and Smith, E. R. 2011. Emotional responses to intergroup apology mediate intergroup forgiveness and retribution. Journal of Experimental Social Psychology, 47(6), 1198–206. Liberman, N., Förster, J., and Higgins, E. T. 2007. Completed vs. interrupted priming: reduced accessibility from post-fulfillment inhibition. Journal of Experimental Social Psychology, 43(2), 258–64.

Moral Psychology as Accountability  79 Lickel, B., Schmader, T., Curtis, M., Scarnier, M., and Ames, D. R. 2005. Vicarious shame and guilt. Group Processes & Intergroup Relations, 8(2), 145–57. McCullough, M. E. 2001. Forgiveness: Who does it and how do they do it? Current Directions in Psychological Science, 106, 194–197. McCullough, M. E., Rachal, K. C., Sandage, S. J., Worthington, E. L., Brown, S. W., and Hight, T. L. 1998. Interpersonal forgiving in close relationships: II. Theoretical elaboration and measurement. Journal of Personality and Social Psychology, 756, 1586–603. McCullough, M., Root, L., Tabak, B., and Witvliet, C. 2009. Forgiveness. In C. R.  Snyder and S. J.  Lopez (eds.), Oxford Handbook of Positive Psychology, 2nd edn., pp. 427–36. New York: Oxford University Press. McCullough, M. E., Worthington, E. L., and Rachal, K. C. 1997. Interpersonal forgiving in close relationships. Journal of Personality and Social Psychology, 732, 321–36. McGraw, K. M. 1987. Guilt following transgression: An attribution of responsibility approach. Journal of Personality and Social Psychology, 532, 247–56. Macnamara, C. 2013. Taking demands out of blame. In D. J. Coates and N. A. Tognazzini (eds.), Blame: Its Nature and Norms, pp. 141–61. Oxford: Oxford University Press. Mandel, D. R., and Dhami, M. K. 2005. “What I did” versus “what I might have done”: Effect of factual versus counterfactual thinking on blame, guilt, and shame in prisoners. Journal of Experimental Social Psychology, 41(6), 627–35. Masclet, D., Noussair, C., Tucker, S., and Villeval, M. C. 2003. Monetary and nonmonetary punishment in the voluntary contributions mechanism. The American Economic Review, 93(1), 366–80. Mazar, N., and Zhong, C.  B. 2010. Do green products make us better people? Psychological Science, 21(4), 494–8. Mencken, H. L. 1949. A Mencken Chrestomathy. New York: Knopf. Merritt, A. C., Effron, D. A., Fein, S., Savitsky, K. K., Tuller, D. M., and Monin, B. 2012. The strategic pursuit of moral credentials. Journal of Experimental Social Psychology, 48(3), 774–7.

Milgram, S. 1965. Liberating effects of group pressure. Journal of Personality and Social Psychology, 1(2), 127–34. Milgram, S. 1974. Obedience to Authority: An Experimental View. New York: Harper & Row Publishers, Inc. Millar, M. 2002. Effects of a guilt induction and guilt reduction on door in the face. Communication Research, 29(6), 666–80. Monin, B., and Miller, D. T. 2001. Moral credentials and the expression of prejudice. Journal of Personality and Social Psychology, 81(1), 33–43. Montada, L., and Schneider, A. 1989. Justice and emotional reactions to the disadvantaged. Social Justice Research, 3(4), 313–44. Moore, G. E. 1993. Principia Ethica, ed. T. Baldwin. Cambridge: Cambridge University Press. Nagel, T. 1970. The Possibility of Altruism. Oxford: Oxford University Press.

80  Brendan Dill and Stephen Darwall Nelissen, R. M. A. 2012. Guilt induced self-punishment as a sign of remorse. Social Psychological and Personality Science, 3(2), 139–44. Nichols, S. 2002. Norms with feeling:  Towards a psychological account of moral judgment. Cognition, 84(2), 221–36. Niedenthal, P.  M., Tangney, J.  P., and Gavanski, I. 1994. “If only I  weren’t” versus “if only I hadn’t”: distinguishing shame and guilt in counterfactual thinking. Journal of Personality and Social Psychology, 67(4), 585–95. Nietzsche, F. 1887/2006. On the Genealogy of Morals and Other Writings: Revised Student Edition, trans. C. Diethe, ed. K. Ansell-Pearson. Cambridge: Cambridge University Press. Noussair, C., and Tucker, S. 2005. Combining monetary and social sanctions to promote cooperation. Economic Inquiry, 43(3), 649–60. O’Mara, E. M., Jackson, L. E., Batson, C. D., and Gaertner, L. 2011. Will moral outrage stand up? Distinguishing among emotional reactions to a moral violation. European Journal of Social Psychology, 41(2), 173–9. Orbell, J. M., Van de Kragt, A. J., and Dawes, R. M. 1988. Explaining discussion-induced cooperation. Journal of Personality and Social Psychology, 54(5), 811–19. Orth, U., Robins, R. W., and Soto, C. J. 2010. Tracking the trajectory of shame, guilt, and pride across the life span. Journal of Personality and Social Psychology, 99(6), 1061–71. Ostrom, E., Walker, J., and Gardner, R. 1992. Covenants with and without a sword: Self-governance is possible. The American Political Science Review, 86(2), 404–17. Paluck, E. L. 2009. Reducing intergroup prejudice and conflict using the media: A field experiment in Rwanda. Journal of Personality and Social Psychology, 96(3), 574–87. Polman, E., and Ruttan, R.  L. 2012. Effects of anger, guilt, and envy on moral hypocrisy. Personality and Social Psychology Bulletin, 38(1), 129–39. Plato. c.380 BC 1992. Republic, rev. edn., trans. G. M. A. Grube and C. D. C. Reeve. Indianapolis: Hackett. Quiles, Z. N., and Bybee, J. 1997. Chronic and predispositional guilt: Relations to mental health, prosocial behavior, and religiosity. Journal of Personality Assessment, 69(1), 104–26. Regan, D. T., Williams, M., and Sparling, S. 1972. Voluntary expiation of guilt: A field experiment. Journal of Personality and Social Psychology, 24(1), 42–5. Reingen, P.  H. 1982. Test of a list procedure for inducing compliance with a request to donate money. Journal of Applied Psychology, 67(1), 110–18. Reno, R. R., Cialdini, R. B., and Kallgren, C. A. 1993. The transsituational influence of social norms. Journal of Personality and Social Psychology, 64(1), 104–12.

Moral Psychology as Accountability  81 Rosati, C. S. 2006. Moral motivation. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, Fall 2008 edn. Roseman, I. J., Wiest, C., and Swartz, T. S. 1994. Phenomenology, behaviors, and goals differentiate discrete emotions. Journal of Personality and Social Psychology, 67(2), 206–21. Royzman, E.  B., Leeman, R.  F., and Baron, J. 2009. Unsentimental ethics:  Towards a content-specific account of the moral-conventional distinction. Cognition, 112(1), 159–74.

Sachdeva, S., Iliev, R., and Medin, D. L. 2009. Sinning saints and saintly sinners: The paradox of moral self-regulation. Psychological Science, 20(4), 523–8. Scanlon, T. 2008. Moral Dimensions: Permissibility, Meaning, Blame. Cambridge, MA: Harvard University Press. Scheffler, S. 1982. The Rejection of Consequentialism:  A  Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions. Oxford: Oxford University Press. Schmitt, M., Gollwitzer, M., Förster, N., and Montada, L. 2004. Effects of objective and subjective account components on forgiving. The Journal of Social Psychology, 144(5), 465–86.

Schultz, P. W., Nolan, J. M., Cialdini, R. B., Goldstein, N. J., and Griskevicius, V. 2007. The constructive, destructive, and reconstructive power of social norms. Psychological Science, 18(5), 429–34. Shafer-Landau, R. 2003. Moral Realism: A Defence. Oxford: Clarendon Press. Shah, J. Y., Friedman, R., and Kruglanski, A. W. 2002. Forgetting all else: On the antecedents and consequences of goal shielding. Journal of Personality and Social Psychology, 83(6), 1261–80.

Sher, G. 2006. In Praise of Blame. Oxford: Oxford University Press. Small, D.  A., and Loewenstein, G. 2003. Helping a victim or helping the victim:  Altruism and identifiability. Journal of Risk and Uncertainty, 26(1), 5–16. Smith, M. 1994. The Moral Problem. Malden, MA: Blackwell Publishing Ltd. Smith, R. H., Webster, J. M., Parrott, W. G., and Eyre, H. L. 2002. The role of public exposure in moral and nonmoral shame and guilt. Journal of Personality and Social Psychology, 83(1), 138–59. Steenbergen, M. R., McGraw, K. M., and Scholz, J. T. 1992. Taxpayer adaptation to the 1986 tax reform act: Do new tax laws affect the way taxpayers think about taxes? In J. Slemrod (ed.), Why People Pay Taxes: Tax Compliance and Enforcement. Ann Arbor: University of Michigan Press. Stillman, T. F., and Baumeister, R. F. 2010. Guilty, free, and wise: Determinism and psychopathy diminish learning from negative emotions. Journal of Experimental Social Psychology, 46(6), 951–60. Stocker, M. 1976. The schizophrenia of modern ethical theories. The Journal of Philosophy, 73(14), 453–66. Strawson, P. F. 1968. Freedom and resentment. In P. F. Strawson and G. Ryle (eds.), Studies in the Philosophy of Thought and Action, pp. 71–96. London: Oxford University Press.

82  Brendan Dill and Stephen Darwall Stuewig, J., and McCloskey, L. A. 2005. The relation of child maltreatment to shame and guilt among adolescents: Psychological routes to depression and delinquency. Child Maltreatment, 10(4), 324–36. Sunstein, C. R., Schkade, D., and Kahneman, D. 2000. Do people want optimal deterrence? The Journal of Legal Studies, 29(1), 237–53. Tabak, B.  A., McCullough, M.  E., Luna, L.  R., Bono, G., and Berry, J.  W. 2011. Conciliatory gestures facilitate forgiveness and feelings of friendship by making transgressors appear more agreeable. Journal of Personality, 80(2), 503–36. Tangney, J.  P. 1995. Recent advances in the empirical study of shame and guilt. American Behavioral Scientist, 38(8), 1132–45. Tangney, J. P., Miller, R. S., Flicker, L., and Barlow, D. H. 1996. Are shame, guilt, and embarrassment distinct emotions? Journal of Personality and Social Psychology, 70(6), 1256–69. Tangney, J. P., Stuewig, J., and Mashek, D. J. 2007. Moral emotions and moral behavior. Annual Review of Psychology, 58(1), 345–72. Tangney, J. P., Wagner, P., Fletcher, C., and Gramzow, R. 1992. Shamed into anger? The relation of shame and guilt to anger and self-reported aggression. Journal of Personality and Social Psychology, 624, 669–675. Tangney, J. P., Wagner, P. E., Hill-Barlow, D., Marschall, D. E., and Gramzow, R. 1996. Relation of shame and guilt to constructive versus destructive responses to anger across the life­ span. Journal of Personality and Social Psychology, 70(4), 797–809. Taylor, G. 1985. Pride, Shame, and Guilt: Emotions of Self-Assessment. Oxford: Clarendon Press. Tetlock, P. E. 2002. Social functionalist frameworks for judgment and choice: intuitive politicians, theologians, and prosecutors. Psychological Review, 109(3), 451–71. Tittle, C. R. 1980. Sanctions and Social Deviance: The Question of Deterrence. New York: Praeger Publishers. Tracy, J. L., and Robins, R. W. 2006. Appraisal antecedents of shame and guilt: Support for a theoretical model. Personality and Social Psychology Bulletin, 32(10), 1339–51. Turiel, E. 1983. The Development of Social Knowledge:  Morality and Convention. Cambridge: Cambridge University Press. Turillo, C. J., Folger, R., Lavelle, J. J., Umphress, E. E., and Gee, J. O. 2002. Is virtue its own reward? Self-sacrificial decisions for the sake of fairness. Organizational Behavior and Human Decision Processes, 89(1), 839–65. Ule, A., Schram, A., Riedl, A., and Cason, T. N. 2009. Indirect punishment and generosity toward strangers. Science, 326(5960), 1701–4. Wallace, R. J. 1994. Responsibility and the Moral Sentiments. Cambridge, MA: Harvard University Press.

Moral Psychology as Accountability  83 Watson, G. 1987. Responsibility and the limits of evil: Variations on a Strawsonian theme. In F. D.  Schoeman (ed.), Responsibility, Character, and the Emotions:  New Essays in Moral Psychology. Cambridge: Cambridge University Press. Watson, G. 1996. Two faces of responsibility. Philosophical Topics, 24(2), 227–48. Wenzel, M., Okimoto, T. G., Feather, N. T., and Platow, M. J. 2008. Retributive and restorative justice. Law and Human Behavior, 32(5), 375–89. Williams, B. 1985. Ethics and the Limits of Philosophy. London: Fontana Press. Velleman, J. D. 2001. The genesis of shame. Philosophy and Public Affairs, 30(1), 27–52. Yang, M.-L., Yang, C.-C., and Chiou, W.-B. 2010. When guilt leads to other orientation and shame leads to egocentric self-focus:  Effects of differential priming of negative affects on perspective taking. Social Behavior and Personality: An International Journal, 38(5), 605–14.

Yzerbyt, V., Dumont, M., Wigboldus, D., and Gordijn, E. 2003. I feel for us: The impact of categorization and identification on emotions and action tendencies. British Journal of Social Psychology, 42(4), 533–49. Zebel, S., Zimmermann, A., Tendayi Viki, G., and Doosje, B. 2008. Dehumanization and guilt as distinct but related predictors of support for reparation policies. Political Psychology, 29(2), 193–219. Zechmeister, J. S., Garcia, S., Romero, C., and Vas, S. N. 2004. Don’t apologize unless you mean it:  A  laboratory investigation of forgiveness and retaliation. Journal of Social and Clinical Psychology, 23(4), 532–64. Zimbardo, P. G. 1969. The human choice: Individuation, reason, and order versus deindividuation, impulse, and chaos. In W. J. Arnold and D. Levine (eds.), 1969 Nebraska Symposium on Motivation, pp. 237–307. Lincoln, NE: University of Nebraska Press.

4 Remnants of Character David Shoemaker

In compiling his list of pleas that we take to warrant exemptions from the community of morally responsible agents, P. F. Strawson included the following: “ ‘He’s only a child,’ ‘He’s a hopeless schizophrenic,’ ‘His mind has been systematically perverted,’ and ‘That’s purely compulsive behavior on his part.’ ”1 He also mentioned as exempt those who are “peculiarly unfortunate in . . . formative circumstances.”2 These are descriptions of agents we typically exclude from the whole range of reactive attitudes like resentment, he noted, insofar as they are “psychologically abnormal” or “morally undeveloped.”3 But what is the precise nature of the abnormality or underdevelopment at stake here, and why does it exempt? Strawson did not explore any of the details of these cases in a way that would help us very much in answering such questions.4 The consequence of this incompleteness is that we also did not get from him an adequate account of the specific features or capacities that might be necessary for full-fledged, or paradigm, responsible agency. To explain, Strawson claims that our reactive attitudes are unified in virtue of their targeting agents’ quality of will, which is “the good or ill will or indifference of others towards us, as displayed in their attitudes and actions.”5 So in resenting someone for elbowing me in the gut, say, I am holding him responsible not for his action but for his ill will toward me in performing that action. If his elbow hit me by accident (and I perceive this), my resentment would (appropriately) be suspended, given that I no longer perceive his quality of will to be poor—even though the pain in my gut remains the same as it would have had he elbowed me with ill intent. Appealing to quality of will

1 P. F. Strawson, “Freedom and Resentment,” in Gary Watson (ed.), Free Will, 2nd edn. (Oxford: Oxford University Press, 2003), 78. 2 Strawson, “Freedom and Resentment,” 79. 3 Strawson, “Freedom and Resentment,” 79. 4 He fully owns up to this point. As he puts it (p. 79), “I must deal here in crude dichotomies and ignore the ever-interesting and ever-illuminating varieties of case.” It is partially my aim in this chapter to explore one of those illuminating varieties. 5 Strawson, “Freedom and Resentment,” 80; emphasis in original.

Remnants of Character  85 thus seems to do the relevant work in explaining why we excuse agents from the reactive attitudes in individual cases (but continue to regard them as responsible agents more generally). However, the theory begins to unravel once we consider more closely the above sorts of marginal cases, those in which the agents are allegedly exempt from the world of responsible agency altogether.6 Exemptions implicate general incapacities, and so on Strawson’s theory, if quality of will is the essence of paradigm responsible agency, then marginal agents ought to be exempted in virtue of their being incapacitated with respect to it. But what does it mean to lack the capacity for “good or ill will or indifference”? And why precisely do our marginal agents lack it? To see the worry more clearly, consider those who are schizophrenic and those who suffer from compulsions (e.g., Obsessive Compulsive Disorder or kleptomania). Both fall under the broad rubric of the “psychologically abnormal,” but they each seem to have very different abnormalities: The former seem unable to recognize certain reasons (they lack a “moral sense”7), whereas the latter seem unable to act on certain reasons. In addition, not any old psychological abnormality eliminates the capacity for quality of will, as we do not exempt those with genius-level IQs, photographic memories, or extraordinary empathic skills. So it would seem that only some subset of psychological abnormalities is exempting, and different abnormalities exempt for different reasons. And the same must be true for “moral underdevelopment”: some children as young as 3 years old can make the so-called “moral/conventional distinction,”8 so insofar as they are exempt it must be for some other, more relevant, feature. But what is that feature, when is it developed, and why is it relevant? It looks as if we have to investigate each marginal case individually and in detail in order to discover the distinct sufficient conditions for exemption, and so to discover the necessary conditions for paradigm agency. Nevertheless, even if we can identify the different exempting conditions via this process, we have a further problem to contend with, namely, our attitude toward so-called exempt agents in many marginal cases isn’t wholehearted. As Gary Watson noted in making a different point, “A child can be malicious, a psychotic can be hostile, a sociopath indifferent, a person under great strain can be rude, a woman or man ‘unfortunate in formative circumstances’ can be cruel.”9 Watson’s point was that these exempted agents still exhibit quality of will in some sense, so a plausible theory of responsibility cannot be constructed solely by appeal to quality of will; other factors (e.g., history, autonomy, authenticity, control, knowledge, etc.) are needed to tell the whole 6 For more on this distinction between excusing and exempting conditions in Strawson’s account, see Gary Watson, Agency and Answerability (Oxford: Oxford University Press, 2004), 223–5. 7 Strawson, “Freedom and Resentment,” 86. 8 See J. Smetana, “Preschool children’s conceptions of transgressions: effects of varying moral and conventional domain-related attributes,” Developmental Psychology, 21 (1985), 18–29; J. Smetana, “Preschool children’s conceptions of moral and social rules,” Child Development, 52 (1981), 1333–6; and M. Hollos, P. Leis, and E. Turiel, “Social reasoning children and adolescents,” Journal of Cross Cultural Psychology, 17 (1986), 352–74. 9 Watson, Agency and Answerability, 228.

86  David Shoemaker story of exemptions. As I have argued elsewhere, I believe there is a better response to this conundrum, one appealing to a wider, pluralistic understanding of “quality of will.”10 Here, though, I simply want to focus on the ambivalence Watson’s cases bring out: Children, while surely exempt from certain types of “harsher” responses like resentment, are not neatly exempt from other sorts of negative response insofar as they are malicious; the hostility of the psychotic still generates some negative responses in us, even if we recognize others to be inappropriate; and while resenting psychopaths does seem pointless, their indifference may still be deeply objectionable in ways that rouse other sorts of responses we may not deem inappropriate. The way it feels to many is that even if these marginal agents are not responsible in an important sense, they might nevertheless be responsible in another. There are thus two burdens that those, like me, who are attracted to a Strawsonian approach to responsibility must take up. First, we must explain why, precisely, the listed marginal agents are exempt, i.e., what is their specific psychological abnormality, or the nature of their moral underdevelopment, such that it grounds their exemption? Second, we must also explain (or explain away) the ambivalence such cases may arouse in us, the feeling that while the agents are exempt in one sense, they may not be in another. Several have already taken up some of these burdens for the marginal cases Strawson explicitly mentioned.11 My aim in this chapter is to take up these burdens for a marginal case Strawson did not mention but surely would have, had he been writing today. It is the now tragically familiar case of dementia. There are three important reasons to explore it. First, those who suffer dementia dramatically illustrate just the phenomena at 10 David Shoemaker, “Qualities of Will,” Social Philosophy & Policy, 30 (2013), 95–120. 11 Gary Watson explored the troubling case of Robert Harris, a horrifying murderer with a horrifying childhood, in Agency and Answerability, 219–59. See also Susan Wolf, “Sanity and the metaphysics of responsibility,” in Watson, Free Will, 372–87. For a more empirical exploration of the ambivalence felt towards those with morally deprived backgrounds, see David Faraci and David Shoemaker, in both “Insanity, responsibility, and the deep self: The case of JoJo,” Review of Philosophy & Psychology, 1 (2010), 319–32; and “Huck vs. JoJo:  The moral ignorance (a)symmetry,” forthcoming in Oxford Studies in Experimental Philosophy 1 (Oxford: Oxford University Press). The case of children has been explored in Watson, Agency and Answerability, 229–31; and Michael Tiboris, Youth and Diminished Responsibility (doctoral dissertation, University of California San Diego, 2012). Psychopathy has been explored by many, including me. See, e.g., Gary Watson, “The trouble with psychopaths,” in R. Jay Wallace, Rahul Kumar, and Samuel Freeman (eds.), Reasons and Recognition: Essays on the Philosophy of T. M. Scanlon (Oxford: Oxford University Press, 2011), 307–31; Neil Levy, “The responsibility of the psychopath revisited,” Philosophy, Psychiatry, & Psychology, 14 (2007), 129–38; Heidi Maibom, “The mad, the bad, and the psychopath,” Neuroethics, 1 (2008), 167–84; Ishtiyaque Haji, “Psychopathy, ethical perception, and moral culpability,” Neuroethics, 3 (2010), 135–50; Matt Talbert, “Blame and responsiveness to moral reasons: Are psychopaths blameworthy?” Pacific Philosophical Quarterly, 89 (2008), 516–35; and David Shoemaker, “Moral address, moral responsibility, and the boundaries of the moral community,” Ethics, 118 (2007), 70–108, as well as “Attributability, answerability, and accountability: Toward a wider theory of moral responsibility,” Ethics, 121 (2011), 602–32. For discussion of the insane, see T. M. Scanlon, “The significance of choice,” in Sterling McMurrin (ed.), The Tanner Lectures on Human Values, 8 (Salt Lake City: University of Utah Press, 1988), 149–216. For discussion of those under great strain, or who are “not themselves,” see Watson, Agency and Answerability, 231–3; and Erin Kelly, “What is an Excuse?” in D. Justin Coates and Neal Tognazzini (eds.), Blame: Its Nature and Norms (Oxford: Oxford University Press, 2012), 244–62.

Remnants of Character  87 issue. The various impairments constituting dementia come in degrees, and in its early stages dementia does not tend to exempt its sufferers from responsibility (although its various symptoms may excuse them from time to time), while in its end stages dementia sufferers are completely exempt from responsibility of any kind. In its middle stages, however, dementia often leaves its victims in just the sort of marginal position I have articulated: On the one hand, they do seem incapacitated for good or ill will in some important sense, yet on the other hand, they still occasionally display qualities like “meanness” or “kindness.” Our exemptions of them during this “perverse sweet spot”12 do not feel wholehearted. Second, the case of dementia has been quite underexplored in the philosophical literature to this point, and to the extent it has been investigated, it has been with respect to topics other than interpersonal moral responsibility.13 But given the prolific and insidious role that dementia has come to play in disabling or altering many, many relationships, it is high time for a detailed philosophical investigation into how and why it does so, particularly with respect to the responsibility responses that tend to structure such relationships in the first place. Finally, I think there are some details specific to dementia that reveal several unrecognized or underappreciated conditions of responsibility, as well as some suggestive insights into the nature of character. In what follows, then, I will first provide the details about the impairments specific to dementia, and then I will discuss which of these are, and which are not, relevant to the exemptions of those who suffer from it during the “perverse sweet spot.”14 I will pay particular attention to the role of memory loss and the way it has been thought to implicate loss of identity. I reject this implication, so I will explain why and then articulate the reason dementia does exempt (when it does) as well as what this suggests for the conditions of paradigm agency. In the second part of the chapter, I will take up the question of whether those with dementia might nevertheless be responsible in a different sense, vindicating our ambivalence about them. What matters, I will suggest, is that they are still capable of revealing features—remnants—of their characters in a way that appropriately grounds certain responsibility responses to them. Of course, these might just be fragments or shards of a disintegrated character, and so not genuine character revelations any more. In grappling with this worry, I will suggest a new way of thinking about character that may have significant explanatory power for theorizing about some forms of responsibility. Before concluding, I will respond to a few objections about my overall story, drawn in part from those few others who have taken the details of dementia seriously. 12 I owe this apt phrase to the editors. My thanks to them also in pressing me to be clearer about the stage under discussion. 13 See Annette Dufner, “Should the late stage demented be punished for past crimes?” Criminal Law and Philosophy, 7 (2013), 137–50; and Agnieszka Jaworska, “Respecting the margins of agency: Alzheimer’s patients and the capacity to value,” Philosophy & Public Affairs, 28 (1999), 105–38. 14 And it should be understood that, throughout, I am talking about the moderate stages of dementia, those illustrating the “perverse sweet spot.” I will, however, occasionally issue reminders of this fact.

88  David Shoemaker

Varieties of Dementia “Dementia” is a catchall term for a number of different symptoms caused by a wide variety of brain disorders. The official DSM-5 statement on it has replaced the term “dementia” with “major or minor neurocognitive disorder,” and its diagnostic criteria for the generic major form are as follows: A. Evidence of significant cognitive decline from a previous level of performance in one or more cognitive domains (complex attention, executive function, learning and memory, language, perceptual-motor, or social cognition) . . . B. The cognitive deficits interfere with independence in everyday activities . . . C. The cognitive deficits do not occur exclusively in the context of a delirium. D. The cognitive deficits are not better explained by another mental disorder (e.g., major depressive disorder, schizophrenia).15 By far the most insidious and widespread purveyor of these evils is Alzheimer’s disease, which affects one in ten Americans over the age of 65, and nearly half of those over the age of 85. It is a progressive dementia, one that eventually impairs all of the aforementioned brain functions, and because of its familiar presence in many of our lives (and because of its rough similarity to many of those forms with a different etiology), it is the form on which I will focus here.16 What makes this form of dementia distinctive is that, in its major form, it requires both memory impairment and impairment in at least one other of the specified cognitive domains: language function (aphasia), the ability to activate or execute intact motor capacities (apraxia), complex attention, impaired social cognition, or a disturbance in planning, organizing, sequencing, or abstracting abilities (executive functioning).17 Here is the uncontroversial datum with which I will begin: There is a stage in the progression of the disease, once the dementia is sufficiently advanced, where dementia sufferers (DSs) typically stop being held responsible (e.g., blamed), and we think this cessation to be appropriate. In the first part of my investigation, I will try to figure out what individual or combined incapacities could render this practice appropriate and why. In the second, I will argue that there is nevertheless a remnant of a different kind of responsibility even in advanced DSs, derived from their remnants of character.

15 American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, DSM-5 (Washington, DC: American Psychiatric Association, 2013), Code 331.83. 16 While many forms of dementia do have numerous features in common, one form of dementia—frontotemporal dementia (FTD)—involves some striking differences. Specifically, patients with FTD seem to behave much like psychopaths. They have lost a kind of emotional empathy that makes it very difficult for them to make a variety of moral judgments. See Mario Mendez et al., “An investigation of moral judgement in frontotemporal dementia,” Cognitive and Behavioral Neurology, 18 (2005), 193–7. While FTD patients would also be exempt from accountability, it would be for different reasons than those I will articulate for most of the standard cases of dementia, so I will set their cases aside here. For a clue into how I would view them, though, see my “Moral address, moral responsibility, and the boundaries of the moral community.” 17 DSM-5, Code 331.9.

Remnants of Character  89

Impairments and Incapacitations Why does the plea “She’s demented” exempt its subject when it does, and what does this exemption consist in? It is not a mere excusing condition, an explanation for why an otherwise responsible agent shouldn’t be held responsible just here and now. It instead reflects the agent’s general status as external to the moral community in some respect. On the Strawsonian story, it reflects the fact that its subject is no longer appropriately susceptible to the classic set of reactive attitudes: resentment, indignation, and guilt on the negative side, and gratitude on the positive side.18 Adopting standard terminology, we can gloss this point as follows: To be exempt from the moral community is to not be accountable, where being accountable is just a function of being appropriately liable to being held to account via expression of the reactive attitudes.19 What grounds the general suspension of reactive attitudes to some agents, on Watson’s Strawsonian story, is that their incapacities generate intelligibility constraints on the demand for good will implicit in the expressions of reactive emotions.20 So even if one has the capacity for quality of will in some sense, one might still be exempt to the extent that one cannot understand this demand. Without the possibility of such comprehension, the intelligibility of making such a demand in the first place is absent. If we think, therefore, that the expressed reactive emotions involved in holding someone accountable are communications of the basic moral demand, then if middle-stage DSs are exempt from being held accountable, they must be so in virtue of some condition that makes them unable to comprehend these sorts of emotional communications. But if this is the right story, then the various individual dementia symptoms we have discussed start to seem irrelevant. Why, precisely, would those specific inabilities undermine the agent’s ability to comprehend the sorts of moral demands being expressed via reactive responses? Impairments in my ability to name objects, or recognize faces, or move my arms, or organize and plan do not seem to implicate my ability to understand your emotional communications to me at all. Of course, it might seem that impairments in social cognition or language would do the trick, but this is also not clear. Impairments to one’s social cognition, after all, do not necessarily involve impairments to one’s understanding of emotional demands; rather, they could 18 As Angela Smith has pointed out to me (in private correspondence), the status of DSs may actually be more complicated (and interesting) than this. During the middle stages of Alzheimer’s dementia, DSs have “good days” and “bad days,” so they may actually alternate between being inside and outside the moral community, rather than being outside, full stop. Smith is surely right about this possibility, so I should be clear that I am interested solely in the issue of why it is that DSs are exempt from the classic set of reactive attitudes when they are exempt. Exemptions, after all, may be global or local, and for some DSs they may merely be local (obtaining only on the bad days). On DSs’ good days, getting off the hook (when they do!) will be a matter of merely being excused. 19 See, e.g., R. Jay Wallace, Responsibility and the Moral Sentiments (Cambridge, MA: Harvard University Press, 1994), and Michael McKenna, Conversation and Responsibility (New York: Oxford University Press, 2012). 20 Watson, Agency and Answerability, 228–9. See also, McKenna, Conversation and Responsibility, esp. ch. 4.

90  David Shoemaker just involve impairments to one’s ability to avoid various priming or attribution biases in reasoning or to one’s solving of cooperative problems. And impairments to one’s language abilities are typically verbal or signal impairments, not necessarily emotional impairments, that is, losses in one’s ability to understand the point of emotional demands like resentment. The only impairment left, then, is that to memory, which must occur on any diagnosis of Alzheimer’s dementia.21 Could memory loss alone thus be the key to exemptions from accountability?

Memory, Identity, and Accountability In John Locke’s famous inaugural discussion of the issue, moral responsibility (what Locke did call “accountability”) was thought to be linked inexorably with the diachronic identity of persons. What makes me the person I am at any given time, for Locke, is my self-reflecting consciousness, so my identity as a person across time consists in an extension of that consciousness to other times: To the extent that I now can have a consciousness of some past experience or action in the same way I did in having experienced or performed it initially, I am identical to that experiencer or agent.22 My consciousness of some past action, then, makes it mine: “whatever has the consciousness of present and past actions, is the same person to whom they both belong.”23 And only actions that belong to me in this way are those for which I am “justly accountable.”24 To explain: When I am held accountable now for some past action, I, a persisting agent, must now have some sort of intimate connection to that past (typically momentary) action. This relation, whatever it amounts to, is what action-ownership consists in, and it is surely at least necessary for accountability: It could not at all be ­appropriate to hold me to account for some action that wasn’t mine.25 So what is “consciousness,” precisely, the element Lockean action-ownership consists in? Most theorists take Locke’s to be a story about memory, so to the extent that I can remember some past action, it is rendered mine for purposes of being held accountable. But if past-directed consciousness really is just memory, then it is neither

21 Technically, the impairment is to memory and learning, but as this impairment is conjunctive, and as I think what does the real exempting work of the two is memory, I will simply focus on it. 22 This way of putting it clearly begs the question if what’s being given is a criterion of personal identity: “I was the one who performed some past action just in case I have a consciousness of my having performed it.” But Locke’s own formulation (again, if it’s taken to be a criterion of personal identity) does clearly beg the question: “For as far as any intelligent being can repeat the idea of any past action with the same consciousness it had of it at first, and with the same consciousness it has of any present action; so far it is the same personal self.” John Locke, “Of identity and diversity,” in John Perry (ed.), Personal Identity (Berkeley: University of California Press, 1975), 40; emphasis mine. 23 Locke, “Of identity and diversity,” 45; emphasis mine. 24 Locke, “Of identity and diversity,” 45. 25 It will probably not be sufficient, however, as sometimes I may be excused from actions that are nevertheless mine in some sense.

Remnants of Character  91 necessary nor sufficient for action-ownership.26 Consider necessity: If, while drunk, I beat someone up for no good reason, that action is still mine even if I have absolutely no recollection of it after sobering up. Consider sufficiency: If a single memory trace of my beating someone up were transplanted into someone else’s brain, that person would not yet own that action for purposes of being held accountable. Insofar as memory cannot be a criterion of action-ownership, impairments to memory alone cannot exempt from accountability. It is important to notice that the objections to Locke just raised are to the alleged connection between memory and action-ownership, the latter of which Locke seemingly failed to distinguish from identity, thinking that whatever makes some past action mine is just what makes that past agent me. But even if memory cannot deliver action-ownership, couldn’t it still deliver identity? No, for the reason Bishop Butler famously articulated: Memory cannot be constitutive of identity precisely because memory presupposes identity. My memory of some past action merely reveals that its agent was me, given that I can remember only my own actions.27 Ingenious science-fiction-based responses to this charge aside,28 it has persistent resonance primarily because memory still seems to be of merely epistemic, not metaphysical, value: It just reveals me to myself, rather than serving to unite my past and present stages. This is the real Butlerian lesson, I think, and contemporary attempts to incorporate memory into psychological criteria of personal identity keep on missing it. What they conflate is a criterion of identity with a sense of identity, and memory, it seems, is only relevant to the latter.29 If we have eliminated memory from contention in delivering either the conditions of action-ownership or the conditions of identity, it would seem like it plays no serious role at all, then, in the conditions for exempting DSs. But as it turns out, memory plays quite a crucial, albeit subtle and complicated, role. It is a role exclusive to accountability, though, not action-ownership or identity itself, and—while it cannot alone exempt—when taken in combination with other symptoms of dementia, as well as the sense of identity, it is the key to accountability exemptions.

Memory and Accountability As earlier detailed, on Watson’s Strawsonian story of accountability, agents are exempt to the extent that emotional expressions to them of the basic demand are unintelligible, 26 The arguments in this paragraph are drawn from my “Responsibility and the self,” in Shaun Gallagher (ed.), Oxford Handbook of the Self (Oxford: Oxford University Press, 2011); and “Responsibility without identity,” Harvard Review of Philosophy, 18 (2012), 109–32. 27 See Joseph Butler, “Of personal identity,” in Perry, Personal Identity, 99–105. 28 See, e.g., Derek Parfit, Reasons and Persons (Oxford: Oxford University Press, 1984), 219–23, for the “q-memories” solution, which itself is drawn from Sydney Shoemaker, “Persons and Their Pasts,” American Philosophical Quarterly, 7 (1970), 269–85. 29 On the notion of the sense of identity, see Stanley B. Klein and Shaun Nichols, “Memory and the sense of personal identity,” Mind, 121 (2012), 677–702.

92  David Shoemaker and so pointless as forms of communication. If impairments to memory are to play a role in exemptions, then, they would have to render this sort of communication unintelligible. In what follows, I will try to explain how this might be so. Let us start by revising the sort of scenario that eliminated memory as a necessary condition of attributability. Suppose that I take a sleeping pill that takes an hour to knock me out, and I know that its only side-effect is that I will not later remember anything that happens during the half hour prior to its knocking me out.30 Crucially, this pill does not alter my ordinary cognitive, deliberative, or volitional abilities in the least. Suppose, then, that during that last half hour after taking the pill one night, my simmering hatred for my slightly noisy neighbor reaches its boiling point, so I decide to beat him up before going to bed. The fact that I genuinely don’t remember doing this the next day is irrelevant to the fact that that action is mine for accountability purposes. Indeed, my neighbor will surely, and appropriately, resent me for beating him up (and communicate that resentment), and bystanders will appropriately be indignant with me for my behavior. My lack of memory of the incident doesn’t at all get me off the accountability hook. Why not? Locke’s objection to my being held to account (punished, in his language) in such a scenario is that there would then be no difference “between that punishment, and being created miserable. . .”31 This is supposed to emphasize the importance of personal identity in this arena: To punish someone who has no consciousness of some past action would be just as immoral as bringing someone into existence who has no connection whatsoever to the crime and then incarcerating him. Now it’s certainly true that, from the internal perspective of each such person, there would be no difference: Both would feel unfairly put-upon. But from our perspective, the external perspective, there would still be a huge difference, given that there were still plenty of other connections between the first guy and the crime, in particular, causal continuity of all the other psychological elements—cognitive, deliberative, and volitional—that contributed to the commission of the crime in the first place (and so would have contributed to my beating my neighbor up regardless of whether I had taken the sleeping pill).32 In such a case, where the only thing missing was memory of the crime, Butler’s point would again come to the fore: All that would be lost would be the agent’s epistemic access to the crime (his sense of identity with the criminal agent), but nothing about its being his crime would necessarily go missing. One can thus deserve a particular fate without having any consciousness of what made it one’s fate. But if expressions of the reactive emotions are communicative, and the agent doesn’t remember what he did to render appropriate such reactions, how could their intelligibility be sustained through this sort of memory loss? The answer lies, I think, in the

30 So I do remember taking the pill. This case is an adaptation of one presented by Parfit, Reasons and Persons (pp. 287–8) to very different effect. 31 Locke, “Of identity and diversity,” 51. 32 I say much more about this sort of “volitional continuity” in “Responsibility without identity.”

Remnants of Character  93 fact that there are multiple communicative functions involved in the deployment of the Strawsonian reactive attitudes. Watson discusses only the very general function of communicating the basic demand for goodwill to the targets of expressions of reactive emotions, but “communicating the basic demand” can involve both backward- and forward-looking aspects, despite the fact that, in most explications, Strawsonians (and Strawson himself) have focused only on the former.33 In expressing my resentment to you, I certainly communicate my anger over your violation of the basic demand, in which case I am drawing your attention to the action you own, pointing out how it involved the expression of ill will to me, and conveying how this makes me feel. In doing so, I am (at the least) demanding that you acknowledge what you did to me, and this is backward-looking.34 But I also communicate something in addition, namely, “Here’s a reminder of the basic demand for goodwill, and you will be well-served to heed it in the future.” In other words, not only may I be communicating my anger over your violation of the basic demand for goodwill (and so demanding that you acknowledge your violation), but also I may be dramatically reiterating what that basic demand is so that you may better keep it in mind for any of our future dealings. Consequently, even if your loss of memory of some past action undermines the first communicative function, it does nothing to undermine the second. To the extent that expression of the reactive emotions may be intelligible as long as it fulfills either communicative function, it can still be intelligible strictly with respect to its forward-looking aspects when directed at the person who has genuinely forgotten performing the action that sparked such a response.35 One might worry that bringing in a forward-looking communicative function returns us to the objectionable social efficiency (consequentialist) justification for our responsibility practices to which Strawson initially objected as lacking in humanity.36 33 Victoria McGeer, in a very recent article (“P. F. Strawson’s consequentialism,” forthcoming in Oxford Studies in Agency and Responsibility 2), does stress the forward-looking functions, but she is in the distinct minority, and she may also be guilty of going too far in the other direction. I favor an account that emphasizes the importance of both functions. 34 I  discuss and defend this “acknowledgment” theory of the communicative function of responsibility-anger in both “Qualities of Will” and Responsibility from the Margins (unpublished book manuscript). 35 To what end, though? In this case, for instance, is it just to get the agent to think twice about taking the sleeping pill in the future? After all, he simply doesn’t remember having disregarded the demand in this case. So is the communication of the demand just aimed at getting him to steer clear of conditions that will cause him to forget whether he has complied with the demand? This would be an awfully strange communicative function. I don’t think the communication is problematic in this way, however, because, first, it’s not as if the agent could securely deny that he violated the demand, given his memory of taking a pill whose effects he knew about. He doesn’t know for sure what he did after taking the pill, so the most he can say is that he simply can’t remember violating the basic demand. Indeed, as I will argue shortly, the ability he has to identify with that demand-violating agent allows him to take on the mantle of demand-violator despite his lack of memory. But more importantly for our purposes here, the idea is that, when he has taken the pill in the future, it won’t affect his memory of pre-pill events, so he will be able to remember having received the stern reiteration of the demand even while under the pill’s spell, and it will hopefully have the effect of getting him to adhere to the demand at that time. 36 See, e.g., Strawson, “Freedom and resentment,” 89–90. See also McGeer, “P. F.  Strawson’s consequentialism.”

94  David Shoemaker But Strawson’s objections to these justifications of “Optimists” were to their emphases on policy and social regulation, which, while rendering responsibility compatible with determinism, nevertheless required an objectivity of attitude, a viewing of their targets as objects to be manipulated and controlled, and it was here that humanity was lacking. What was lost, precisely, was the engagement between humans in genuine interpersonal relationships. But to live within these relationships (and who could live without them, being human?) is to be committed to the personal reactive attitudes of resentment, gratitude, and the like, and only by attending to such attitudes can we “recover from the facts as we know them a sense of . . . all we mean, when, speaking the language of morals, we speak of desert, responsibility, guilt, condemnation, and justice.”37 Now these are indeed all backward-looking concepts. But even if all we mean in speaking of these concepts is captured by the backward-looking communicative features of the reactive attitudes, this does not mean that all we mean in expressing these attitudes is captured by their backward-looking communicative features. For it is obvious that we issue forward-looking demands and expectations of our friends and loved ones intended to get them to change their behavior precisely as ways of engaging with them personally. In reminding my friend of the terms of our relationship, I am reminding him of what continuing our interpersonal engagement must consist in, and this is a reminder to which I expect autonomous compliance. To the extent that he is my friend, I do not merely want to directly and immediately cause his behavior to conform to my demand for goodwill, to wield my emotional arsenal solely to manipulate and control him; instead, I want to cause him to remember my demand precisely as the demand of a friend when he is deciding how to treat me in the future, in the hope that he will consider such a demand to matter to him in his autonomous deliberations. This is the difference between treating him as an object to be manipulated and a person to be beseeched. The forward-looking communicative feature of reactive attitudes, therefore, may be (indeed, must be) included in our analysis of them, and we can easily do so without losing our humanity in the process. This isn’t yet to say, however, that their backward-looking communicative function is actually missing in the sleeping pill case. To see why it might not be, it is very important to note that, while responses to wrongdoing (say) by their very nature take place after the fact of wrongdoing, there can be considerable variation in how long after the fact they take place, and this delay may sometimes cause confusion in our views of the conditions rendering such responses appropriate. Imagine two variations on our case, then. In the first variation (A) my neighbor and some bystanders respond immediately after the assault with reactive emotions (with resentment and indignation, respectively). In the second variation (B) they don’t respond until the next morning (perhaps the assaulted neighbor is unconscious until morning, and then he tells others straight away upon waking up). At the time of their response in (A), I  still remember the



  Strawson, “Freedom and resentment,” 91; emphasis in original.

37

Remnants of Character  95 assault, but at the time of their response in (B) I do not. In (A), my neighbor’s expressed resentment presumably has both communicative features: It serves to communicate his anger at my violation of the basic demand for goodwill as well as to dramatically reiterate that demand as something to be paid closer consideration in the future.38 Why, though, could this not also be the case with the neighbor’s expressed resentment the next morning in (B)? If the backward-looking aim of the expressed resentment is to communicate his anger at me for assaulting him and get me to acknowledge what I did, that aim can be met as long as there is some sort of essential connection between me in the morning and the action the night before, something that ties the assault to me-now. Memory is typically what would reveal this connection to me, providing me with the sense of identity. But my neighbor’s communicated message could also well resonate with me in memory’s absence, particularly if I can still fully identify with the assaulter of the night before, i.e., to fully understand what it was like for him at the time from the inside, to feel what he was feeling, and to appreciate, endorse, and recognize the force of the considerations that moved him. Where my current volitional structure is continuous with, and causally dependent on, his, I can actually “see myself ” doing what he did in a way that both provides me with a relevant sense of identity and renders my actual memory of having done so redundant. This is how the communicated message of my neighbor can still intelligibly maintain its backward-looking force. And as it obviously meets the conditions of its forward-looking aspect as well—I can understand the emotional force of his expression of the demand in a way that can get me to reshape my present and future deliberations about how to treat him—such emotional expression may fulfill both of its communicative functions in the sleeping pill scenario regardless of the time at which it is deployed. Turn, then, to the memory impairments of dementia. To this point, I have focused solely on memory loss with respect to some specific past action. But the memory impairment of dementia is much broader, applying to one’s actions, experiences, and judgments, and it constitutes an ongoing problem. One can never be sure that what is experienced will ever be remembered again. But now we can see how both communicative aims of expressions of the reactive emotions may be undermined in such patients. If I  can neither remember what I  did nor be expected to remember your remonstrations of me when engaged in future deliberation, then it looks as if it might well be pointless (as a form of communication) for you to express your reactive emotions to me in the first place. Now one could try to tell a story about identification like the one told above to rescue at least the backward-looking communicative function. After all, at the time of remonstration, couldn’t the memory-impaired patient at least identify with the past agent,

38 As it turns out, I  won’t remember it, but that’s irrelevant to whether his expression serves its forward-looking communicative function, to dramatically impress on me the importance of the basic demand in a way that I can understand and take up with respect to its forward-looking dimension at the time of the expression. More on this point below.

96  David Shoemaker even if she couldn’t remember precisely what she did, such that she-now could still be tied to the action in a way that makes her victim’s expressed resentment (at this later point in time) a sensible form of communication to her? She could, I have suggested, without any additional impairments. But memory impairments alone are insufficient for a diagnosis of dementia, remember; needed in addition are impairments to at least one of the other five psychological abilities. While the impairments of Alzheimer’s ultimately run in varying degrees across all of the named abilities, what I want to suggest likely does the real work in exempting moderately demented patients from accountability—when they are so exempt—is, in addition to general memory impairment, impairment to executive functioning. The reason is that this general capacity includes more specific capacities to engage in identification of the sort I have already argued is necessary for expressions of the reactive emotions, in the absence of memory, to achieve both their backward- and forward-looking communicative aims; in particular, it includes the deliberative and abstraction capacities that are implicated in the ability to place oneself in different circumstances, either in the position of some future affected self or in the position of some past acting agent. To the extent that one’s ability to imaginatively project oneself forward or backward in this way is damaged, though, so too will be the point of any communication presupposing that ability.39 Nevertheless, the varying degrees of the impairments of dementia might seem to provide possible counterexamples to the story I am telling here. For instance, someone might have no (or very mild) executive impairments while having wholly pervasive memory impairments. These memory impairments might be sufficient to destroy any real notion of the agent’s self and so may also single-handedly seem to undermine the possibility of identification with past or future selves. Alternatively, someone’s memory might be more or less intact but she might have such severe executive capacity impairment that any sort of reflection, abstraction, planning, sequencing, etc. is off the table, and so it is this impairment that single-handedly undermines the possibility of remembering what she did. These are both possible scenarios, but they actually serve to make my point, which is that it is a capacity for the sense of one’s agential identity that is necessary for accountability.40 This sense may be provided by memory or identification (involving robust projective imagination). Now while it is most fully achieved via a combination of memory and identification, sometimes severe enough damage to one or the other of these capacities alone may be sufficient to undermine both. But in 39 Again, I am focused only on that “perverse sweet spot” in which DSs have crossed the threshold of non-accountability, yet they remain agents toward which we may still have what can be described as ongoing “responsibility responses.” 40 Why “agential” and not “personal” identity? The latter typically refers to diachronic numerical identity, a unique (one–one) relation between a person at t1 and a person at t2. The sense of identity to which I am referring could easily attach to a one–many relation, however. Just think of two fission products remembering the pre-fission self: They would both have access to what it felt like “from the inside” to perform various actions. I want to allow, then, that “agential identity” may not have the same metaphysical constraints as numerical identity. The “sense” in question also refers to memories of, and identifications with, agents, for those are the entities to whom our responsibility responses are relevant.

Remnants of Character  97 any event, it is the overall capacity for the sense of one’s agential identity that has been undermined, and that’s my focus. Consequently, if you can neither remember some past action nor identify with its agent, the backward-looking communicative aim of my expression of resentment to you over your having violated the basic demand cannot be met. And if you will not remember my expression of resentment in your future deliberations, nor are you able to deliberate effectively in the formulation of a plan now to adhere to my resentful demand in the future (by projecting yourself into that future), the forward-looking aim of that expression cannot be met. With both communicative functions of expressions of the reactive emotions effectively blunted by a proposed target’s dementia, holding that agent accountable would be inappropriate. And to the extent the relevant psychological impairments obtain, the agent cannot sensibly be regarded as a member of the moral community susceptible to these expressions. This is why DSs are exempt from accountability.

Attributability and the Remnants of Character And yet . . . Even when DSs are exempt from accountability in the middle stages of the disease, it may still seem true of them that they express ill—or good!—will in their actions in some sense. Even if there is no communicative point to our emotional remonstrations of them, it still seems to be true that they are, occasionally, mean, kind, generous, sour, pinched, or cheerful. And the fact that they are this way may render various responsibility responses to them fitting, even if some “accountability” responses to them remain unfitting, and so even if it would be inappropriate to express the fitting responses with the communicative aims earlier discussed. Consider a case independently of dementia. Suppose that, while I can recognize the interests of others, I never take them seriously as reasons. Indeed, I often deliberately discount the interests of others. I am a cruel person, and you may both judge me to be cruel and respond to me with contempt. These seem clearly to be responsibility responses. They may not be the sort of responses we have been discussing (e.g., holding to account via expressed resentment), but they are nevertheless responses to my character qua practical agent, and they will typically ground various other (responsibility) responses to me, such as your avoiding me, or your not trusting me, or your failing to respond to my greetings with your normal good cheer.41 Gary Watson has labeled this form of responsibility attributability.42 On Watson’s construal, for Φ to be attributable to one is for Φ to reflect one’s deep self, where this is constituted by one’s evaluative 41 These are responses in line with the Scanlonian understanding of blame. See T. M. Scanlon, Moral Dimensions (Cambridge, MA: The Belknap Press of Harvard University Press, 2008), ch. 4. 42 On the original distinction between accountability and attributability, see Watson, Agency and Answerability, 259–88. In my “Attributability, answerability, and accountability,” I articulate a third distinct face of responsibility (answerability), and I also present the more ecumenical understanding of attributability I am about to introduce here.

98  David Shoemaker system, the set of one’s evaluative ends and commitments.43 I have elsewhere argued for a broader, more ecumenical model of the deep self, one that includes both evaluative commitments and the emotional dispositions constituting one’s set of cares.44 I will simply stipulate that ecumenical account here. The important point, though, is this: Responses to attributability-responsibility, after qualitative assessments of Φ itself are in (i.e., was Φ good or bad?), are typically aretaic, including emotional responses targeting agential character (e.g., admiration, disdain, contempt), and predications describing features of the agent’s character given the attribution of Φ to him or her, e.g., “cowardly,” “brave,” “miserly,” “cruel,” “callous,” “generous,” “kind,” and so forth.45 This approach leaves the door open for DSs to be attributability-responsible too. Their dementia, even if advanced, may not necessarily undermine their ability to have a motivational structure that grounds various attributions, and so aretaic emotional responses and predications. Here examples of positive behavior may initially be more resonant. Suppose that when I visit my demented grandfather, he reaches for a chocolate in the candy jar and brings it over to me, despite the great physical effort it requires. After accomplishing this, he is on to the next things: moving a bowl on his kitchen counter half an inch to the left, gazing out the window at a preening cardinal, turning to ask me who I am. My expressing real gratitude to him at that point has no communicative point (it may well be highly confusing).46 He cannot remember what he just did, he will not remember how I respond to what he just did, he cannot identify with that moments-ago agent, and he cannot deliberate about the future (and so identify with some expected future self). He lacks the sense of his agential identity required to be accountable. Nevertheless, what he did was kind—perhaps he took the fact that it would give me pleasure to be a reason to bring me a chocolate—and if he does this generally, it reflects on a kindness that is a trait of his character: He is kind. These sorts of predications are anything but unfamiliar to us. Our demented parents can be loving, patient, generous, considerate, or brave. And these assessments are deployed more confidently when they are traits continuous with those from their pre-demented days. “She was always patient like this,” we may say when watching our mothers with our own attention-deficient children. “He’s still so considerate,” we may say about my grandfather to another chocolate-receiving guest. Of course, matters aren’t always so rosy. Many demented parents and grandparents remain just as cranky or cantankerous as ever. Some are stingy, some are callous, some are narcissistic, some are difficult, some are inconsiderate, and some are even cruel. And these too may be longstanding traits, persisting through the dementia and 43 Watson, Agency and Answerability. See also his “Free Agency,” in Agency and Answerability, 13–32. 44 See especially “Ecumenical responsibility,” forthcoming in Randolph Clarke, Michael McKenna, and Angela M. Smith (eds.), The Nature of Moral Responsibility (Oxford: Oxford University Press); and Responsibility from the Margins. 45 Watson, Agency and Answerability, 266, 270–1. 46 I, along with a few others, take gratitude to be the positive correlate of resentment. See also Scanlon, Moral Dimensions, 151.

Remnants of Character  99 expressed repeatedly in actions and attitudes. One can still be an old jerk despite the ravages of Alzheimer’s. It may be communicatively pointless to blame such a person for being an old jerk, but it still may be true that he is one. Of those psychological features relevant to our enterprise, then, character seems the last to go. Dementia, even advanced, may leave intact enough of who one was to ground a continuing, albeit limited, range of assessments of genuine responsibility. It is to these remnants of character that they attach, just as it is to these remnants of character that we—their children, friends, and caregivers—often cling. While DSs are no longer members of the accountability community—we can no longer make intelligible normative demands of them—they are not outside the world of responsibility altogether. And insofar as the attributability of which they are eligible may still inform and shape the ways in which we continue to respond (and not just react) to them, it could well constitute the last remaining threads of our interpersonal relationships with them.

Traits of Character and Commitment Clusters To this point I have been discussing character remnants, persisting psychological traits of practical agents that may continue to ground attributability responses even through several stages of dementia. Our confidence in the appropriateness of these responses is typically bolstered by the ongoing familiarity of the traits in question. But what of the more challenging (and equally familiar) cases in which the formerly kind demented person’s actions start to seem “mean,” say, in a way never seen before?47 Perhaps she now lashes out with contempt at the petty flaws of others, whereas prior to her disease she was always patient and indulgent. Is this new meanness attributable to her? And if so, what does this say about the nature of character generally? There seem to be only three ways to answer such questions, and the challenge posed by such cases is that these answers either seem implausible or not very amenable to the thesis I have advanced. First, we might say that this agent developed a new character trait—meanness—after the onset of dementia. This is a rather implausible possibility, however: If character traits are constituted by one’s cares and commitments, then developing new cares and commitments would require a kind of abstraction, planning, and anticipation that may no longer be possible (or at least not possible to a sufficient degree) once the various executive impairments have taken place. Second, we might say that this agent had the mean character trait all along but that her meanness is just now being revealed for the first time, given that the dementia robbed her of her control over its expression. But this is implausible too, as it would have required a kind of control throughout her life up until now that is beyond the range of most humans (and so the implausibility comes in thinking of just how many DSs display such “revealed” meanness in contrast to the likely numbers of those who could realistically exercise 47 Thanks to Angela Smith for raising this important worry. “Mean” is in scare quotes precisely because what’s at issue now is whether such behavior really is mean at all, a reflection of a character trait.

100  David Shoemaker such control throughout their lives). Further, even if the dementia undermined the kind of control capacity heretofore necessary to resist expressing one’s meanness, this sort of expression does not typically occasion the kind of rue occasioned by other instances of losing control. For example, Alzheimer’s patients who lose control over their bodily functions are deeply ashamed. But there is typically no such shame or even regret experienced when meanness is expressed by the formerly kind. This would suggest that the “suddenly disinhibited” character trait story is unlikely (although I will return to a variation of this possibility later). This seems to leave us with our last option, the one that is not amenable to my account of the attributability-responsibility of DSs: Perhaps what happens in such cases is that dementia has just ravaged the agent’s character, so that what behaviors now occur are simply first-order desire-based responses to stimuli, responses which then yield the sorts of immediate joys or frustrations associated with the simple desire satisfaction or dissatisfaction of young children and even some non-human animals. This would mean that the DS’s so-called “meanness” isn’t meanness at all, as it is just disconnected from her “real” self (which Alzheimer’s has destroyed anyway) and so we should not respond to it as such. But then, one might say, why shouldn’t we think the same thing of those “traits” that do seem to persist, viewing the agent’s behavior instead as bearing only an accidental resemblance to her character-grounded behavior of old, not flowing from any real self simply because no such self exists anymore? I think the key to responding to this important challenge is to delve more deeply into the nature of character. I have simply stipulated that it consists in one’s set of cares and commitments, and while this may be uncontroversial, it is so only because the nature of the cares and commitments has been left somewhat open (and so may invite very different interpretations). So let me start there. I take cares to be dispositions to respond in sync emotionally to the up-and-down fortunes of the cared-for object.48 Commitments (to the extent they are distinct from cares) are dispositions to behave or react in certain ways with respect to the advancement or defense of the ends or objects one has judged good or worthy of defense. Now I explicitly agree that the formation of new cares and commitments is probably unlikely once dementia has set in, given the impairments to executive capacities. But already-formed cares and commitments don’t necessarily require such executive capacities to persist, insofar as they are essentially dispositions, triggered by various thoughts and perceptions that may survive executive impairment. Both caring and commitment dispositions may easily be triggered by a perception that the object of one’s care or commitment is being threatened, say, and the habitual nature of character traits thereby revealed may produce action that is fairly immediate and direct, with little or no need for any real kind of deliberation. When a loved one is in peril, the impulse to save him or her tends to move one fairly automatically, unmediated by



See my “Caring, Identification, and Agency,” Ethics, 114 (2003), 88–118.

48

Remnants of Character  101 executive thinking or planning. To have the latter would indeed be to have one thought too many.49 It thus seems quite possible that some cares and commitments may persist through several stages of dementia.50 But some cares and commitments may well not do so. Obvious candidates will include commitments whose objects or appropriate response behaviors require certain abstraction or planning capacities. Religious commitments, for instance, may fall by the wayside with the onset or advancement of dementia, insofar as they demand an ability to conceive a very abstract object of devotion.51 Prudential caring may be lost as well: It may no longer matter to the DS that her lack of regular exercise will affect her ability to play with her grandchildren in a few years. Certain kinds of moral commitments may fade away too: A committed Kantian would be unlikely to maintain his theoretical stance once the ability to universalize his maxims—or see the point of doing so—was lost. One may still be a morally good person absent the ability to theorize about morality, however. Kindness, for instance, doesn’t require thoughts about either the nature or the demands of kindness; it may merely require the disposition to respond in helpful ways to certain people. At issue, then, is how to make sense of meanness in DSs who were never mean, on the assumption that it is probably not a new character trait. My suggestion is that, when such behavior isn’t merely expressed frustration, it may flow from remnants of certain caring or commitment clusters, some of whose members have been lost via dementia. To explain, the pressure for psychological integrity and the elimination of cognitive dissonance usually produces harmonious relations among our cares and commitments,

49 See Bernard Williams, “Persons, character, and morality,” in Amelie Oksenberg Rorty (ed.), The Identities of Persons (Berkeley, CA: University of California Press, 1976), 197–216. 50 One might well wonder about the extent to which commitments could persist like this. For one thing, it would seem that we need to be much more active with respect to them than cares, and that this sort of activity isn’t compatible with their being merely habitual. In addition, commitments would seem to require certain sorts of sophisticated cognitive machinery, involving the making of evaluative judgments, and these sorts of capacities would be unlikely to survive dementia. And even if they did, it is difficult to see how such judgments could persist across time in a habitual manner. (Thanks to Justin D’Arms for raising a version of this worry.) I don’t think this is necessarily the case, however. As I go on to discuss in the text, I agree that there will certainly be many commitments that are lost via dementia, but these will tend to be those that have pretty abstract objects. Commitments having more concrete objects could still persist as a matter of habit, though. Consider a commitment to eating good food. This explicitly requires a judgment of worth, but it may have been so internalized that it persists through dementia, and it is also dispositional, called up in the agent upon being presented with food of any kind. Nevertheless, even if I am wrong about this, the traits with their source in the agent’s caring (which are emotional dispositions that may operate independently from judgments of worth) can still persist habitually through dementia, and insofar as these may be sufficient to ground some traits of motivational character, they may be sufficient to ground attributability-responsibility. 51 For this sort of example, see Jaworska, “Respecting the margins of agency,” 124. Jaworska talks here in terms of values, and she also notes the ways in which some values may persist through dementia (see her n. 41 on p. 124), but her notion of values and my notion of the cares and commitments constituting character traits are importantly similar in many respects. I am indebted to her discussion, and I say more about the positioning of my account relative to hers later in this chapter.

102  David Shoemaker and they often function to temper or bolster each other. Caring for something52 could manifest itself in a very wide variety of attitudes and behaviors; which ones are manifested is typically delimited by other cares in one’s psychological network. To take an example, if the only thing I cared about were my wife, and there were no other psychological elements on which my volitional structure depended, then my behavior to protect her or advance her interests would know no bounds: I would constantly attend to her, fend off threats with spectacular violence if need be, do whatever was necessary (and I mean whatever was necessary) to advance her interests, and so forth. But as it happens, I also care about morality, my own needs and desires, and other people. These other cares serve to restrict the wide variety of ways in which my caring for my wife is manifested. Losing those other cares, however, would correspondingly loosen those restrictions. Were I to stop caring about other people, the degree to which I attended to my wife’s interests would dramatically increase. Were I to stop caring about morality, the kinds of things I would do to advance her interests would considerably widen to include the nefarious. Were I to stop caring about myself, the circle of acceptable behavior would surely expand even further to the illegal and self-destructive. Our cares and commitments seem to operate in clusters, then, and it is some of these clusters that best pick out what we think of as “character traits.” So courage might be thought of as involving (at least) a strong care about the defense of certain people (or ideals), tempered by a care about survival (so as not to be reckless) that is nevertheless somewhat weaker than it is in other agents. And kindness might be thought of as involving (at least) a strong caring about helping others in need, tempered by (typically weaker) cares about one’s own well-being (so that one doesn’t just give all one’s resources away). What this account suggests might be going on in the case of DSs with seemingly new character traits, then, is that some members of a previously harmonious cluster of cares and commitments were lost to dementia, and this loss served to unshackle their volitional possibilities in a way that generated mean behaviors. But these behaviors would not be flowing from new character traits; rather, they would be flowing from persisting clusters of cares and commitments with fewer members. And the lost members would most likely be those requiring the executive capacities typically ravaged by dementia: the abilities to abstract and plan. Consider a real-life case of someone who was both deeply religious and loved his wife dearly. Pre-dementia he never would have considered suicide (viewing it as seriously immoral), but post-dementia, when his wife died, he repeatedly expressed the wish to die.53 To the extent he lost his religion to the dementia, the restrictions were removed on the scope of acceptable behavior in response to the loss of his remaining object of care. Or consider someone who was financially responsible pre-dementia, but post-dementia went out and bought the fancy truck he had always wanted, a purchase that would require great financial sacrifices for 52 From here on out, I will talk only in terms of cares, but this should be read as a gloss on “cares and commitments.” 53 Jaworska, “Respecting the margins of agency,” 107.

Remnants of Character  103 him and his family.54 To the extent he may have lost his ability to engage in prudential planning, the prior restrictions on buying his dream vehicle were removed. What I am suggesting is that a story along these lines could explain a DS’s mean turn as well. My hunch is that this might well be true most typically for those who were previously nothing but kind solely due to a religious or moral commitment. Once the abstraction abilities necessary to religious or moral understanding were lost as part of their executive impairments, the restrictions on some formerly unthinkable behavior were also lifted. Crucially, that’s not to say that they had a straightforward mean disposition all along; rather, it’s simply to say that altered clusters of cares and commitments may produce behavior judged to be mean in contrast with the DS’s earlier behavior. Caring about being honest is typically tempered by a commitment to protect the feelings of others. Losing that tempering commitment may yield true but hurtful statements, which will be construed as mean, especially coming from the mouth of someone who had previously been so kind. Or caring about comfort may produce a kind of belligerent demandingness if not tempered by a certain sort of respectful attitude (something that may well be lost in light of executive impairments), and again, this too will often be construed as meanness in the formerly kind.55 Ironically, though, this story might return us to the thought (one I deemed implausible) that sometimes persisting character traits are actually revealed during dementia. Consider someone whose spouse treated him poorly throughout their married lives but who remained solicitous and polite to her until the onset of his dementia. As the disease progressed, though, he started to tell her off in nasty ways.56 Had he suddenly developed a mean, vicious streak? Now on the one hand, a significant aspect of my story still seems plausible here: A powerful inhibition of some sort was probably undermined by his dementia. But on the other hand, a different part of my story may seem implausible: It is less likely that the loss of the inhibition altered any existing care-commitment clusters (character traits) than it is that some persisting clusters were suddenly allowed new expression. In other words, why isn’t it more plausible to say here that his vicious character trait (if it is vicious) was hidden all along, and was only revealed by dementia? Indeed, his lack of shame or regret for this behavior might support such a story: His true self is being revealed, after all. In response, this case brings out the important difference between the mere loss of volitional control and the loss of a care about volitional control. Suppose our husband did for years repress any lashing out—despite his desires to do so—because he cared 54 Jaworska, “Respecting the margins of agency,” 107. 55 Chandra Sripada offers a different take on the matter. Once key cares or commitments in a character cluster are genuinely lost (and not replaced), he thinks, aretaic predications (and attributability-responsibility judgments generally) no longer apply. (See his “Self-expression: A deep self theory of moral responsibility,” unpublished manuscript.) I don’t quite see, however, why this should be the case. The patient who lost his wife (and his religion) had some quite profound feelings, and they shaped his motivational structure in important ways. Indeed, his remaining cares fully explain what he was doing as a practical agent, in precisely the sort of way his more robust character clusters pre-dementia did. 56 I am grateful to the editors and to Don Loeb for suggesting a possibility along this line.

104  David Shoemaker strongly about not being that sort of person. People may then have come to admire him as a man of great restraint, as someone who was indulgently patient. Here the relevant cluster consisted, we might say, in his caring about revenge for insults tempered by his (stronger) caring about not being the sort of husband who lashes out at his wife (one doesn’t count as restrained unless there is something one is restraining). Suppose we found a way to disconnect this latter caring from structuring his motivations, i.e., a way to strip him of the volitional control he cared about by blunting the effectiveness of this tempering care. He then cruelly lashes out. Here we would expect regret or shame, for it would still be true of him that he is an indulgently patient man, a man of great restraint; he was just the subject of volitional manipulation. So while “patience” and “restraint” may still be attributable to him, his vengeful expression is not. Dementia doesn’t just blunt the motivational efficacy of certain of one’s cares however; it also often eliminates those cares. So suppose our husband’s dementia did just that, eliminating his caring about being a certain sort of husband (which it would likely do to the extent that this is a fairly abstract aim). Now he might be expected to lash out, and without regret. But now “patience” and “restraint” are no longer attributable to him, whereas his vengeful expression is. But this is neither because a new character trait was developed nor because an old character trait was finally being revealed, exactly. Rather, it was because an old character trait transmogrified via the loss of a key care in its cluster. A character cluster remains, then, albeit a fundamentally altered one. These are obviously complicated issues, though, and I have presented a fairly speculative story. But one aspect of it to which I am committed is that certain character traits, like meanness, may sometimes just be the product of certain dispositions in the absence of other typically tempering dispositions, and not, as many would have it, straightforward dispositions to be mean (perhaps in light of judgments of its worth). Meanness is, at least in part, behavior in pursuit of some end coupled with ignoring the feelings of others or judging them not to matter in such pursuits. Whenever the feelings of others cease to matter, therefore, meanness may erupt, and this predicate may truly apply regardless of the target’s executive incompetence or inability to understand any additional blaming responses, and regardless of the source of any such inability, as long as some character clusters are being expressed in the process. Meanness may be attributable to one, even if one is no longer accountable for being mean.

Objections Before concluding, I want to illustrate and reiterate some of the main points of the account by noting and responding to two final objections. First, one might well think that what I have presented is just a lot of superfluous puffery, given that we can simply assign DSs to the same exemption category as children (aren’t DSs really just going through a “second childhood” anyway?). Strawson noted that children were exempt by being “morally undeveloped.” But what this must really mean is that they are exempt in virtue of the fact that they just can’t comprehend the basic demand for good will. But

Remnants of Character  105 neither can DSs, so why couldn’t we have explained their exemption from accountability without wading through the bits about memory and identification? My first, very brief, point in response is that children present a much trickier case than assumed by the objection. As noted earlier, most children from the age of 3 can distinguish between so-called moral and conventional normative judgments,57 and many children are also able to recognize when a demand is being made of them and how they ought to respond. Determining their actual level of moral development or normative (in)capacity, as well as their actual sources, is very difficult, so it’s not as if we can make a simple appeal to some such analysis and just apply it to the case of DSs, for I doubt that any such (uncontroversial) analysis exists. My second and more responsive answer is to point to a crucial difference between children and DSs: Those who suffer from dementia had at one time, and still have operative remnants of, a full-blown character; children, at least young ones, haven’t yet established any sort of full-blown character at all. The psychological traits and dispositions that together constitute an agent’s deep self for purposes of attributability are fairly robust, consisting of developed cares and evaluative commitments. These require far more mature psychological faculties than children have. While children may have certain character tendencies and predispositions—perhaps enough to explain why they seem to exhibit quality of will “in some sense”—they don’t have sufficiently developed deep selves for actions to be properly attributable to them.58 DSs, on the contrary, may well continue to have sufficiently robust character traits to be attributability-responsible for the attitudes or actions flowing from them. These have been long-developed, maintained, and ingrained, and there is nothing about (some of) them that couldn’t in principle persist through a diagnosis of dementia. Again, our agential characters may be one of the last things we get and also one of the last things to go. A second objection might be drawn from one of the only other philosophical discussions of dementia around, Agnieszka Jaworska’s insightful account of Alzheimer’s patients and the puzzle of advance directives. Such directives are “living wills” drawn up to specify how physicians are to treat one if one becomes disabled or hospitalized for any number of reasons. The puzzle has been what to do in cases where one preference is expressed by the original author of the directive and a contrary preference is expressed by a demented subject some time later. Whose preference is authoritative in that case? While most exploring the issue have defended the view that the signer’s preferences are authoritative, Jaworska makes a persuasive case for occasionally respecting the preferences of the demented subjects. Through discussion of a series of moving and detailed real-life examples, she argues that even those with advanced dementia

57 For an explanation of my “so-called” language, a defense of my skepticism that there is any such distinction, at least in terms psychologists have presupposed, see my “Psychopathy, responsibility, and the moral/ conventional distinction,” The Southern Journal of Philosophy, 49, Spindel Supplement (2011), 99–124. 58 See also Wolf, “Sanity and the metaphysics of responsibility.”

106  David Shoemaker may continue to value, an ability she too notes as one that may be retained long after other decision-making capacities have faded away. But insofar as it is valuing that provides the essential foundation for autonomy (by producing principles for conduct), and insofar as the execution of the DSs’ values in the world may helpfully be assisted or implemented by others, DSs may be sufficiently autonomous that their current preferences about medical treatment could be more authoritative (in virtue of being autonomously made) than those of their earlier directive-signing selves.59 The objection that might be drawn from her analysis (but is not, I stress, one that Jaworska herself makes!) is that, insofar as autonomy is thought to be a crucial condition of moral responsibility, and DSs might have Jaworska-style autonomy, any moral responsibility they have might be established, contrary to my story, directly (and more simply) via appeal to their autonomy. Autonomy is certainly the central buzzword in debates about advance directives, so I suppose it behooves Jaworska to establish autonomy on behalf of her Alzheimer’s patients so as to generate in its wake their preferential authority in such practical matters. And autonomy, true enough, has been taken by many to be crucial to moral responsibility. But one important implication of my project is that I don’t think so much can be delivered by a mere label. What obviously matters are the precise conditions in which said autonomy consists. And when we look directly and closely at the various possible candidates, we can see that the persisting abilities to value and care about things—to take up or be committed to an evaluative stance on various aspects of the world—may, in the face of dementia, ground one kind of responsible agency even though dementia has eliminated another. In other words, in the arena of responsibility, at least, there isn’t anything like the all-or-nothing deliverances of autonomy; indeed, aiming solely at such a one-size-fits-all answer to our various questions blinds us to a much more nuanced and complicated picture of human agency. Whether the same is actually true for the practical issue of advance directives, or other bioethical puzzles, is a matter I leave for others to explore.

Conclusion I have argued that, while advanced DSs are exempt from the world of moral demands— exempt from accountability—they may not be exempt from all facets of responsibility; in particular, they may often be attributability-responsible and perform actions or exhibit attitudes that reflect on their agential character and thus ground aretaic emotions and predications. The reason they are exempt from accountability, however, reveals something important about the conditions of such for paradigm agents like us: to be accountable requires that one be able to comprehend the communications essential to being held accountable. In addition to basic comprehension capacities (i.e., one



Jaworska, “Respecting the margins of agency,” e.g., p. 130.

59

Remnants of Character  107 mustn’t be a young child or a psychotic), though, one must also have the capacities of either memory or identification, capacities which generate one’s sense of agential identity. When DSs have significant impairments to both such capacities (either directly or indirectly), they are exempt from accountability in virtue of these impairments to that sense.

Acknowledgments My thanks to Daniel Jacobson, Michael McKenna, Nick Sars, Angela Smith, and Tamler Sommers for their helpful remarks on a previous draft. I’m also grateful to the insightful discussion provided by several members of the audience when I presented an earlier version of this paper at the University of Washington in February 2012, especially Ann Baker and Janice Moskalik. The paper also benefited tremendously from the questions and remarks offered by my fellow participants in the June 2012 University of Michigan workshop on Moral Psychology and Human Agency, sponsored by the John Templeton Foundation. In particular, I would like to thank Selim Berker, Julia Driver, Pamela Hieronymi, Don Loeb, Heidi Maibom, Tim Schroeder, Walter Sinnott-Armstrong, and Nicole Smith. For extended and especially helpful discussion at the same workshop, I’m grateful to Justin D’Arms, Shaun Nichols, and Chandra Sripada. And while I have already expressed my gratitude to them individually, I am also grateful to the editors collectively for their excellent and helpful comments on the penultimate draft.

5 Knowing What We Are Doing Heidi Maibom

Suppose you are standing with a group of your friends outside a busy restaurant waiting for your table. A group of old people leaves and walks up the road. Suddenly one of them falls to the ground. There is a pause. No one moves. Then someone says: “He fell over, maybe we should go help?,” to which another replies: “No, they’ll take care of it.” The conversation resumes briefly. The waiter comes out and lets you know that your table is ready. As you walk in, you notice that the person is still on the ground while the other old people are attempting to help him to his feet. Are you and your friends blameworthy for your inactivity? After all, it is evident that every person in that group is very old and in a poor position to help the fallen man. It is lucky that the old people do not harm themselves while attempting to do so. All of you, on the other hand, are relatively young and spry. Now pretend that you are the chair of an academic department in the US perusing Lakisha Johnson’s CV. If you are like most, you judge her to be less qualified than a male candidate with an English sounding name, but with the same CV. The candidate is in the unfortunate position of sounding African American, and applying while being a woman. However, you certainly don’t take yourself to be discriminating against her. You are a self-avowed feminist after all. You think she is a less good candidate. You are, we may say, the victim of stereotyping as much as the candidate is; only the consequences are more dire for her. Can we blame you for undervaluing her qualifications? On the one hand, it seems that failing to help those in need and discriminating against others are things people do, and if they are not cognitively or volitionally impaired, they can be held responsible for it. On the other, they do not conceive of their actions as we do. The bystanders genuinely thought that the old person did not need help and the chair really did believe that Lakisha is less qualified that Robert. It therefore seems unfair to hold them responsible for actions that they were unaware of performing. Can we hold people responsible for what they do not know that they do? On the third hand, as it were, it certainly seems as if the youngish, spry people and the chair

Knowing What We Are Doing  109 ought to have known what they were doing. If so, they can be held responsible for their actions. The difficulty is encapsulated in some of the debate surrounding employment discrimination cases (e.g., Fiske and Borgida 2008; Williams 2003). Given the available evidence that much discrimination is due to stereotyping based in unconscious bias, how do we evaluate defendants’ guilt? Stereotyping appears to be an automatic process, i.e. “unintentional, uncontrollable, efficient, autonomous, and outside conscious awareness” (Fiske and Borgida 2008, 127). The trouble is that there is a tendency to think of responsibility in terms of awareness, both in the law and in philosophy. People talk as though an agent’s responsibility extended only as far as her awareness (Sher 2009, 4). This is not meant to exclude cases of neglect, but these instances are downplayed. This is nowhere clearer than when John Martin Fischer and Mark Ravizza maintain that “an agent is responsible only if she . . . knows (or can reasonably be expected to know) the particular facts surrounding her action” (etc.) (1993, 8). The parenthesis suggests that cases where the agent does not know what he or she is doing are the exception and should not mislead the reader when she thinks of the main cases, which almost invariably concern reasons and considerations that the agent is aware of culminating in action that the agent conceives of under the description or descriptions relevant to moral assessment. But even if they are exceptions—and more argument is needed to show that they really are—the justification of why agents can be held responsible for such actions must connect in some interesting way to their responsibility in ordinary cases. In this chapter, I consider whether we are responsible for acting under situational influences that we are unaware of at the time of action, such as stereotyping or standing by when others are in need. Put somewhat bluntly, the question is whether we are sometimes responsible for doing what we do not know that we do. I argue that we are sometimes responsible for acting under situational influences though we do not know what we do at the time. Accounts that take character, identification, or a deep self to be central to responsible agency cannot account for responsibility for acting under such influences, as these are largely impersonal. Instead, I develop a version of the idea that an agent could and should have known what she was doing. I think of this in terms of “inferential routes” to a relevant description of the action. To be responsible, agents must have the capacity to know what they do, but not necessarily, or even primarily, in a direct, private, first personal, or introspective way. There is no need for the agent to be directly aware of her action under a relevant description. Rather, it is the ability to conceive of our action from a perspective other than our own that is central to responsibility; to know what we do as from the outside. Or so I shall argue.

1.  Not Knowing What We Do A person who leaves her dog in a hot car or a distracted driver who does not give himself enough time to merge with traffic are both typically held responsible for their

110  Heidi Maibom actions (Sher 2009). Yet, they did not perform the action we hold them responsible for under that description. In fact, it is less what they do and more what they failed to do that we are concerned about. Their fault is one of inaction or omission. Omission is tricky. We cannot but omit to do an infinite number of things. C’est la vie, you might say. But though we cannot help but not do a great number of things, we do have some control over which particular actions we omit to perform. Typically omissions are only thought to be relevant if the agent has knowledge relevant to the situation or could reasonably have acquired such information. The person who leaves her dog in her car on a hot day can reasonably be expected to know that she should leave the windows ajar or not leave the dog in the car at all. Her husband, although he is only a block away, cannot be expected to come to the dog’s rescue, as he is unaware of its predicament and of the facts surrounding it. In cases of neglect, agents fail to do something they ought to have done without awareness of such failure. The woman does not think “to hell with the dog,” for instance, thus leaving it to its fate. Neglectful inaction is typically “undertaken” for no reason at all. That is, there might be a reason that the agent fails to consider performing the action that we judge he should have performed, but more often than not he does not have a reason not to perform it. According to George Sher (2009), neglectful action is a prototypical example of agents not knowing what they are doing, yet being responsible for it. It is a curiously opaque situation: The agent is unaware of omitting to do something. We should agree with Sher, I think, that agents can be held responsible for neglectful actions even though it means that they are sometimes responsible for doing what they do not know that they do. The types of neglect Sher talks about are cases where the subject is distracted, preoccupied, overly focused on other tasks, etc. So it does seem quite reasonable to say that the agent could have known what he was doing and, if the action is morally significant, we can praise or blame him for it. But what should we say of cases where the agent is unaware of what she is doing in a more principled way? Typical neglect cases are cases where if the agent considered the situation—“the dog is in the car and it’s hot and sunny,” or “the other drivers on the road won’t have time to avoid me if I speed up to merge”—then the agent would have acted differently (assuming that they are only cases of neglect). But the cases we are considering are ones in which the agent does consider the situation, but fails to see that his or her action falls under a description that makes it blameworthy. It is a description that the agent is typically disposed to deny is appropriate. So where the neglectful agent can be made to see what he neglected to do pretty easily, the person who fails to help someone in need or who discriminates against a candidate is apt to question our description of his action. Acting under unconscious psychological influences seems different from being neglectful. Once we want to hold people responsible for actions that they do not conceive of under that description, we are faced with the problem of limiting the potentially infinite number of descriptions any action can fall under to just those an agent is responsible for. The most straightforward way to do this is to anchor such descriptions in

Knowing What We Are Doing  111 what the agent has conscious access to. The agent either knows what she is doing, or she has access to that information. In the neglect cases, it is a matter of directing her attention to other features of the situation. She can consciously access the information she requires to act better, but she does not. But there are many cases where the best or most salient explanation of what we do is not consciously accessible to us in this way. Situational influences, of which the bystander effect and stereotyping are examples, are instances where conscious access to a morally relevant description of what one does is affected (in ways soon to be specified). Some such influences are surprising, but hardly morally disquieting. Richard Nisbett and Timothy Wilson (1977) found that people tend to show a right side bias when asked to choose the best quality item in a row (position effect) and that they tend to value a mug that they own more than an identical one (endowment effect, see also Kahneman, Knetsch, and Thaler 2005; Kahneman 2011). Subjects were not only unaware of these effects, but would also categorically deny that they had any influence on their actions. One might fairly claim that they did not know what they were doing. However, determining the relevant description of their action—how to weigh their account of what they were doing against psychologists’ descriptions of their actions—may seem relatively unimportant. Other types of situational influences are more troubling. Bib Latané and John Darley discovered that people are much less likely to help a person in need if other people are present (Latané and Darley 1970). People are also less likely to help if they are in a hurry (Darley and Batson 1973), and more likely to help if in a good mood (Carlson, Charlin, and Miller 1988). Yet, these are not aspects of the situation that people reference when explaining their reasons. In bystander scenarios, people typically say that they did not help the person in need because they were not quite sure: what happened, whether the person was in need of help, or whether the situation was serious enough to require intervention (Latané and Rodin 1969). People’s tendency to cooperate is also easily influenced by the environment they are in. Describing a prisoner’s dilemma as the Wall Street Game as opposed to the Community Game has large effects on cooperation. Under the former description, only about a third of subjects chose to cooperate compared to two-thirds when the game was presented under the latter description (Ross and Ward 1996). Stereotyping is now a well-known phenomenon. It affects all aspects of group belonging: ingroup vs. outgroup, men vs. women, white vs. black, white vs. Asian, high socioeconomic status vs. low socioeconomic status, old vs. young, and so on (for overviews see Kunda 1999 and Fiske and Taylor 2013). For instance, the prevailing stereotypes associated with women are that they are submissive, more intuitive than rational, understanding, empathetic, bad at math, emotional, and so on. Women and African Americans are assumed to be less qualified than men (Superson 1999) even if the evaluator has the same information about both, for example identical CVs (Steinpreis, Anders, and Ritzke 1999; Bertrand and Mullainathan 2004; Moss-Racusin et al. 2012). This is thought to be due to stereotyping, (hopefully) mostly of an unconscious kind.

112  Heidi Maibom In the above cases there is reason to be morally concerned about situational effects. People see certain features of the situation as salient or ignore other, morally important, features, without having conscious access, even in principle, to the fact that this is occurring. The types of explanations that psychologists typically give of such situational effects do not have a counterpart in what the subject can become aware of by introspection. This is why someone like Timothy Wilson (2002) says that much of the time we do not know what we are doing, or why we are doing it. There is some truth to this. Stated like this, though, it suggests a vision of people staggering about like zombies operated by situational and other factors as if by a puppet master. But just because people don’t conceive of their action under a certain description doesn’t mean that they have no idea what they are doing. They may conceive of their actions under a different description. In stereotyping, for instance, the subject takes himself to be doing whatever he is doing without undue influence of extraneous factors. In his own mind, the chair judges our female African-American candidate to be less qualified than her white male peer because she is less qualified. The bystander effect distinguishes itself from stereotyping by causing, or contributing to, inaction. In John Darley’s and Dan Batson’s (1973) cases, where subjects rushed past a person slumped in a doorway seeming to be in dire straits, many maintained that they hardly even noticed the person in need, so busy were they getting to where they were going. In Latané and Rodin’s (1969) study where the experimenter pretended to have an epileptic seizure behind a curtain separating her from the experimental subjects, who were waiting for her, people claimed not to know whether she really needed help. In bystander cases, people typically consider the other person’s situation and decide—incredibly—that the person is not in need of help, or something along those lines. A bystander who does not help need not quibble about the description of his action as “you did not help.” Instead, she might maintain that the person was not in need of help, that it was not clear that he needed help, or that she did not know that he needed help. After all, we do not help others most of the time. Indeed, often others would not want us to help them. Sometimes it would be inappropriate to do so. It may therefore be better to describe the agent’s failure to help as failing to help a person who is in need of help. This would help distinguish the agent’s description of her action and that of others (other non-bystanders, as it were). The debate, then, seems to be about whether or not the person who was not helped was in need of help. But this cannot be quite right either. We do not blame people for not rushing to the help of someone who, say, has just unwittingly consumed a deadly poison, but is not yet showing symptoms. It must therefore be relatively obvious that the agent is in need of help. Prototypical bystander cases seem fairly straightforward. Think back to the experimenter thrashing around on the floor in the next room. Surely, she is in need of help! Just last week, I saw a man tumble out of his wheelchair onto a busy road. As I was on my balcony on the 13th floor, I assumed people at the ground level would come to his aid quicker than I would. But as no one stopped to help, I descended 13 floors and

Knowing What We Are Doing  113 crossed half a block to be the first person there to help. He had been flailing around on the street for 5–10 minutes in the rush hour! If we assume that the pedestrians and the drivers would all say that they were unsure whether the person needed help—in line with subject reports from other bystander effect studies—they are surely self-deceived. Presumably, the most obvious assumption when seeing a man flailing in the street next to his wheelchair is that he is in need of help. Yet, there is no particular reason to think that bystanders are lying when they avow their ignorance of the neediness of the person. Why is this so hard for people to see when they are in the company of others? The situationist literature suggests that we are far more sensitive to others’ actions and judgments than we think. Solomon Asch (1951; 1956) found that people are likely to override their own, correct, judgments of the comparative length of lines if the majority of the people they are with judge otherwise. One interpretation of the bystander effect, therefore, is that since we look to others as guides to what to think and do, and they do nothing, we don’t do anything either. Perhaps we are responding to what we assume is a norm of non-helping or we assume that others are not helping because help is not required or desirable. When we say that we did not help because we were unsure about whether the subject needed help or not, it may be interpreted as a reasonable shorthand for this process. It strikes me as quite plausible that bystanders typically do not help because they are unsure of whether the subject is really in need of help or whether it is appropriate to help. This uncertainty arises from observing that others do not help, but they probably do not help because they are unsure of what to do given others’ inactivity. Most likely, they would help were they aware that they were subject to the effect. I know knowledge of the effect had that effect on me. This fact makes it clear that there is an essential lack of access here for most people. And this is where it gets tricky. For if we say that it is obvious to the agent that the patient is in need of help, the agent would disagree. That is the very point about the bystander effect; people question the need of the other person. In order to hold bystanders responsible, it cannot simply be that it was relatively obvious that the other was in need of help, because one might always ask: “obvious to whom?” There must be some sort of connection between the mind of the agent and the obviousness of the need of the other. If we give up on the idea of conscious awareness, there are a couple of candidates for the responsibility fixing relation between the agent and the action: (i) the agent would agree with performing the action if she conceived of it under the relevant description (or even weaker: the agent would not refrain from performing the action if she thus conceived of it); (ii) the act is in character; or (iii) the agent should and could have known that her action is an action of the relevant kind, and if blameworthy, should have refrained from performing it. Should we hold people who neglect to help others in need, people who discriminate against others unknowingly, etc. responsible? It seems to me that we should, or at least that we sometimes should. Some cases are clearer than others. Think of the toddler Wang Yue in Foshan, China, who wandered from her house onto a busy laneway where

114  Heidi Maibom she was run over by a car. After the motorist leaves the scene without helping, a truck runs over her legs as she lies immobilized on the street. She is eventually picked up and taken to a hospital by an old lady. In the meantime, however, she lies crying and bleeding on the road for a full seven minutes while eighteen people skirt around her. No one helps. She subsequently died from her injuries. In this case, there seems to be little doubt that the eighteen are responsible and blameworthy for not helping. I have discussed the bystander effect at length, but similar reflections apply to the case of discrimination. Evidently, Lakisha is unfairly evaluated. Yet, the chair is un­aware of the effects that stereotyping has on his evaluation and regards it as fair. This type of covert, unconscious discrimination is the most serious obstacle facing women and minorities today (at least in the West). As in the bystander effect, it seems compelling to hold people responsible for discriminating against others even when they do not know that they are doing so. How can we make this intuition fit with a decent account of responsibility? That is the topic of the next section.

2.  Responsibility for Situational Influences As previously mentioned, there are three good candidates for responsibility attribution that circumvent the need for conscious access: (i) the agent would agree with performing the action if she conceived of it under the relevant description (or: the agent would not refrain from performing the action if she thus conceived of it); (ii) the act is in character or expresses the agent’s deep self; and (iii) the agent should and could have known that her action is an action of the relevant kind, and if blameworthy, should have refrained from performing it. Here I rule out (i) and (ii) as decent contenders for holding people responsible in situations where they are under situational influences. I then develop the remaining candidate in the following section. The first option is easy to dismiss as a candidate for holding people responsible who are subject to situational effects. Bystanders are not likely to agree that they should refrain from helping someone in need. If they conceived of their action in this way, they would stop, i.e. they would aid the person in need (by and large). Though blatant discrimination exists, most of the discrimination women and members of racial minorities face in the West these days is of an implicit sort. Discriminators, too, are likely to disagree that it is OK to discriminate against others. Many, or most, of those who engage in implicit discrimination would cease to do so were they to think of their actions as discriminatory. The problem is, evidently, that we are demanding some form of conscious or explicit agreement, whereas situational effects mostly operate implicitly. But perhaps some sort of identification will succeed where explicit agreement fails (e.g., Frankfurt 1971). So-called “Deep Self ” views (Wolf 2003) base responsibility in identification with the reasons that bring about the action, or perhaps with what the action represents (e.g., Frankfurt 1971). This is quite a popular approach. Crucial to responsibility, on

Knowing What We Are Doing  115 such a picture, are the deeper structures of the self that have to do with self-evaluation, self-transformation, and longer-term goals and projects. Here responsibility is inextricably linked to our sense of who we are, who we want to be, and what we embrace or reject in ourselves. To make such views accommodate cases of neglect, conscious avowal of one’s deeper self cannot be necessary. People who are inconsiderate, for instance, typically do not identify with this tendency. Nevertheless, they rarely stop and think about how others think or feel in response to their actions. And it is this fact more than any act of identification, actual or counterfactual, that makes us critical of them. Perhaps one can think of it as passive acceptance. So it is not enough to object, if blamed for something, that the action flowed from motives that are not part of who we want to be. We want to avoid holding people responsible for truly alien ideas and motives. In the case of being inconsiderate, the person is arguably responsible for it to the extent that she consistently fails to take the relatively easy and obvious steps to avoid being inconsiderate. This lack of consideration can, therefore, be thought to form part of her deep self in a suitably expanded framework. But how, you might ask, is such a framework to make sense of the sorts of unconscious situational motivations and influences that we have seen examples of above? Although John Doris (2002) worries that the impersonal nature of situational influences sits uncomfortably with the rather rich notion of a deep self, he argues that minor adjustments to this picture will accommodate responsibility for acting under such influences. Instead of attempting to fit situational effects into a narrative structure of who we take ourselves to be or who we want to be, Doris introduces the notion of plans and long-term commitments. Through our plans and projects we plot a route through situationist space. By making it a habit to fly busily from one commitment to another, to hang out with a bad crowd, and so on, we betray a commitment or lack of commitment to certain ways of living. Someone who hurries past a person in need is blameworthy because his being hurried reflects a way of life that he endorses and though he does not endorse not helping others because he is in a hurry, he does endorse the type of lifestyle that predictably leads to such effects. So, situational influences can be understood in a personal way after all, Doris maintains. And where they cannot, it might indeed be inappropriate to assign praise or blame to a subject for acting under such influences. It seems to me that Doris strikes an uneasy balance between situationism and Deep Self views. The literature contrasts the person and the situation so as to stress the impersonal nature of situational influences. An action performed under situational pressures is not an action that we should understand in terms of the person’s character, but in terms of the situation that she is in. If we consider the bystander effect again, the literature suggests that had there not been other people present, then the person would likely have helped. This is not to say that character can play no role in such situations, but where it does, it is more likely to do so in cases of helping. Susceptibility to situational influences has little to do with character, life plans, etc. It is a more universal feature of people (though culture does, of course, have an effect). It is therefore difficult to

116  Heidi Maibom reconcile situationism with an identificationist account, since such accounts focus on the personal. It seems that we are in need of another account of responsibility. Sher (2009) fleshes out “should have known/been aware of the action under the relevant description” in option (iii) above in terms of the causal aspects of a subject’s reasons. He argues that if an action springs from “some combination of his own constitutive attitudes, dispositions, and traits” (p. 87) and it is a failure to live up to some applicable standard, we can hold the agent responsible for it. For something to be constitutive of a person, it must be “among the elements of the system whose causal interactions determine the contents of the conscious thoughts and deliberate activities in whose absence he would not count as responsible at all” (p. 121). Are bystanders who do not help responsible on this account? It is hard to say. On the one hand, people’s sensitivity to the judgments of others is causally related to the conscious capacity for deliberation that marks what it is to be a person. So it seems that they can be held responsible. On the other hand, the idea that the action must spring from “some combination of his own constitutive attitudes” etc. suggests that a personal element remains in Sher that conflicts with the spirit of situationism, much like Doris’s account does. If I am right that we can hold someone responsible though we think she was subject to the bystander effect (or engaged in implicit stereotyping), it looks like we are giving up linking responsibility to a person’s long-term plans and policies, her character, or her deep self. What a person is responsible for need not be something that she is aware of, or that pertains to her character; it need not be (causally) related to her life-plan, her policies, or her values. There is nothing personal about being subject to the situational effects outlined by the social psychology literature. In fact, in many cases, there is something deeply impersonal about it. These are tendencies that we share with many or, in some cases, most people. They are part of who we are, of course, but not in a way that individuates us as persons. It may be part of our inner, deeper core, but it is a core that we have in common with many others.

3.  Responsible Self-Reflection Let us remind ourselves why having conscious access to relevant aspects of one’s actions (or omissions) is thought to be so important. Ceteris paribus an agent who has conscious access to such aspects is linked to it directly by her understanding. She knows what she does; she owns her action. Epistemic access is also important since if she does not know what she does or is proposing to do, she lacks an important reason to change her course of action. Thus, conscious access gives you rational control and authenticity or ownership, all else being equal, which are ingredients in responsibility. But conscious access cannot account for responsibility for aspects of actions or outcomes that the agent is not aware of. It excludes simple cases of negligence. So if we are to account for why people are responsible for such actions, we must widen the scope of the epistemic condition to include the possibility of awareness, suitably defined. As

Knowing What We Are Doing  117 we saw, Fischer and Ravizza demand that the agent “could reasonably be expected to know” (1993, 8) the relevant facts pertaining to her action. This seems to be a fair demand in instances of negligence where, in effect, we are requiring the agent to attend to relevant consciously accessible information (“I am leaving the dog in the car, it’s hot, who knows how long I’m going to be,” etc.). This line of thinking is the most promising of the possibilities that we have considered so far. The way I’m going to develop it, however, takes it in a different direction by stressing an agent’s capacity for reasonable self-reflection and scrutiny of motives over conscious access. When we hold people responsible for neglect, we reflect a demand for cognitive effort on the part of the subject that goes beyond what she might be naturally disposed to expend. So neglect points in the direction that I want to take us. Consider moral education. Children have relatively little authority over how their actions are interpreted. They are blamed for actions that they did not conceive of under the description that their parents think of them. A child does not usually get off the hook by denying that he was hurting his sister and only trying to get his toy back. A delicate balance is struck whereby children are blamed for actions that they did not conceive of under that description, but not before they are able to conceive of that action as falling under such a description. Doing otherwise is pointless, not to mention unfair. Being blamed for doing something that one was not entirely aware of doing, and accepting such blame, is part of learning how to be a moral agent. It is an ongoing process. We never quite finish learning. Life as a social being is an ongoing negotiation among individuals about how to conceptualize situations, events, and actions. How to describe an action is not simply a function of how the agent conceives of it, or how those affected by it, or those observing it, conceive of it. (And these conceptions are sometimes incommensurable.) It is also relative to a culturally sanctioned way of conceiving of certain types of actions: their most socially salient aspects, and so on. A canonical description of an action relevant to the assignment of praise or blame is typically an interaction between these potentially different perspectives on the very same trajectory through cognitive or behavioral space. Sometimes the socio-cultural perspective is paramount. Stealing, lying, cheating, hurting, and killing are all dominant descriptions of actions, which are morally and socially sanctioned, and we are expected to consider whether or not our actions fall under them. We cannot excuse ourselves by pointing to our own perspective on the action in such cases. At other times the individual’s perspective plays a prominent role. I cannot be blamed for giving my friend strychnine if unbeknownst to me someone has laced my bottled water with it, but I can be blamed for neglecting to consider that leaving a dog in a car on a hot day might kill it. Returning to the bystander effect again, in many cases the person should have been aware that there was something suspect about her easy acceptance of the idea that the subject is not in need or that it is not appropriate to help. People flailing around in rooms or on streets are surely in need of help. A further complicating factor in accepting the bystander’s defense of her not helping is that most people help if they are the

118  Heidi Maibom only ones to witness the event. And they do so because they think that the person is in need of help. So, it is hard to make out that bystanders are incapable of seeing the subject as being in need or seeing that help is appropriate. In fact, it very much seems to be the case that they should have seen the situation in this way. The conclusion seems inescapable. In many instances of the bystander effect—where the subject is in somewhat serious need of help and there is some obligation to help—the inactive bystander is responsible and, therefore, the proper subject of blame. We may be more understanding towards him because we realize that many or most of us would have acted similarly, but understanding why he did not help is not the same as absolving him of responsibility. Considering one’s actions under a variety of different descriptions requires a degree of self-reflection. We expect others to critically examine themselves, their actions and patterns of action, and to question their own interpretation of them. Deluding oneself or failing to be properly reflective about one’s actions is something we can be held responsible for, assuming that we possess the relevant capacities. As long as there is an inferential route from a person’s values and beliefs, or from values and belief relatively easily accessible to her, to an important conceptualization of her action, we may hold her responsible (Maibom 2013). The male academic who professes to be a feminist, yet excludes female academics, ignores their opinions, and fails to take their concerns seriously, is deluded, but not necessarily in a way that is impenetrable to him. For he can think about his actions in a more objective fashion, or from a perspective other than his own, and consider what motives may underlie them. He does not. But this is not because he is unable to conceptualize his action in such a way as to see it as being wrong. Why should it matter that the conceptualization of the action is important and how should we understand “important” anyway? Because of the potentially infinite number of descriptions an action falls under, one must by necessity restrict the range that a person can be responsible for. As we have seen, being aware of performing the action under the relevant description falls within that range, but does not exhaust it. For this excludes neglect. So the relevant description of the action may not be one that the agent is aware of, but it must be within her grasp. She can conceive of this description without undue epistemic hardship. But surely there are all sorts of descriptions that an action falls under that an agent could conceive of without straining herself cognitively! That is, the agent can come to consider each one individually, but we cannot expect her to think of very many such descriptions in her deliberations, let alone all of them. This is where “important” comes in. We can say nothing terribly precise about what is an important description and what is not. Descriptions that are closely linked to moral prohibitions and prescriptions are obvious candidates. Descriptions that people affected by our actions or that reasonable objective observers would give are also candidates for descriptions that an agent can be expected to consider. The more severe the consequences of our actions, the higher the demand for self-reflection. Whereas it might be reasonable to absolve us of the responsibility for bruising someone’s feelings—on the basis that the demands for self-reflection would be too onerous in this case—we are reasonably held responsible

Knowing What We Are Doing  119 for injuring or killing someone even if awareness of this possibility would have required significant reflection. How does all this apply to unconscious states and processes, you might ask. We have the capacity to reflect on our actions from the outside, as it were. Ordinary human interaction forces other perspectives on us from early on. We can imagine how someone else would describe the action we consider performing. When people get angry with us, it is often because they see our actions in ways that we do not. They may be wrong, they may be right. In either case, we are in a position to consider whether such descriptions are reasonable. The situation is symmetrical, of course. We have all experienced feeling fobbed off with an unlikely story about somebody’s motives, which is supposed to alter our conception of her action. We suspect her of being insincere. This complex interaction between our own and others’ perspective on thoughts, motives, and actions sets the scene for the ability to reflect critically on what we think, say, and do. It should not, therefore, be ruled out that a person is responsible even if she does not conceive of her action as being of the relevant kind. For instance, the inconsiderate person or the bystander is responsible if he fails to be properly self-reflective. We have to be careful, of course, that our conditions on responsible agency are not too strict. We cannot demand rigorous self-reflection of a sort that would put anyone but Socrates or the Buddha to shame. It is unlikely that people possess the intellectual energy for such exercise, and even if they did, it is not at all clear that it would be best expended on such an enterprise. The demand for self-reflection is a function of the severity of the consequences of not engaging in it. This is no different, however, from what we require when the moral stakes are high generally: We demand greater scrutiny, more care, etc. from the people faced with difficult moral choices. If a person makes a hash of his life by wandering blindly through it, never stopping to examine his actions or reasons, to question himself, he is responsible for failing to exercise his capacity for self-reflection. Someone who tends to regard female candidates as automatically less qualified than male candidates ought to question his judgment. Determining whether, or the degree to which, someone is responsible for unconscious mental activity may often not be straightforward. Nevertheless, the fact that it seems reasonable to hold people responsible in many such cases suggests that we should not translate the demand for knowing what we are doing into a demand for conscious access to what we are doing under the relevant description. My suggestion takes the ability for self-reflection and access to inferential routes to socio-morally important descriptions of our actions to be basic to responsibility and moral agency. Conscious awareness is important, yes, because the inferential route culminates in the agent thinking about her action under relevant descriptions. However, we should not think of the central cases of such a process as involving conscious access to information as from a first person perspective. We are not simply “looking in” to see whether we have considered all the relevant information. Often, we must consider different ways of thinking about our actions, thereby engaging our ability to take up

120  Heidi Maibom different perspectives on our own actions. As we develop as moral agents, we achieve facility with classifying our actions as belonging to relevant moral kinds. At the outset of our journey, however, it is often by taking another perspective on ourselves and our actions that we come to see actions as belonging to certain (moral) kinds. Much of the time, situations and our actions within them are nuanced, complex, and do not present themselves to us with ready-made descriptions or in categorical terms. We need to do a bit of work conceptualizing them. This work is often done by perspective taking. We consider what we do from the perspective of others. We do not arrive at the relevant description(s) of our action by introspection, but by seeing our action as from the outside. What matters is that we are able to see our action as others would see it, i.e. from their perspective(s). Thinking about our own actions in a morally and socially productive way often, even typically, involves thinking about them as we would think of others’ actions, from a standpoint outside our own agency, and in the absence of the curious privacy that typically characterizes our decision process. This brings out the impersonal nature of the suggestion, and highlights its fit with situationism. The capacity for self-reflection, for perspective taking, and so on is entirely impersonal and applies to moral agents across the board. There is, of course, a personal touch insofar as the inferential route is not indifferent to the agent’s capacities, her knowledge, the situation she is in, and so on. Whereas one agent can be held responsible, another might not be depending on the inferential routes available to her. Nevertheless, the capacity for self-reflection is impersonal and universal. It has nothing to do with the kind of person someone is.

4. Conclusion What I have been concerned with here is sometimes called the epistemic condition on responsibility. I have argued that the best way of extending this condition to apply to prima facie responsible actions, such as those that result from people being subject to various situational effects, is in terms of the capacity for self-reflection. Conscious access to the socio-morally relevant perspective(s) on an action is relevant in a limited way. It comes from the experience of interacting with others, taking their perspectives, reflecting on one’s actions more broadly, and so on, and allows for fluid and adept action. But the required experience for responsible action comes from examining one’s actions as from the outside. We need flexibility in how to conceive of our actions. We must be able to conceive of them from someone else’s perspective and according to a publicly sanctioned description, and we need the ability to scrutinize motives or reasons for action. We require the ability to take a perspective on what we are doing that is not purely our own. Identification or the possession of a deep self is not central to this view of responsibility. We are responsible for more than what flows in any direct way from our characters, who we want to be, our long-term goals and projects. Scrutiny of motives and capacity for flexible conceptualization of our actions and their

Knowing What We Are Doing  121 consequences has little to do with the kinds of persons we are. It is a universal human capacity to reconceptualize our proposed actions in categories or terms that are publicly available and important or that capture how others will see our actions.

References Asch, S. E. 1951. Effects of group pressure upon the modification and distortion of judgment. In H. Guetzkow (ed.), Groups, Leadership, and Men. Pittsburgh: Carnegie Press. Asch, S.  E. 1956. Studies of independence and conformity. Psychological Monographs, 70(9), 1–70. Bertrand, M., and Mullainathan, S. 2004. Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review, 94, 991–1013. Carlson, M., Charlin, V., and Miller, N. 1988. Positive mood and helping behavior: A test of six hypotheses. Journal of Personality and Social Psychology, 55, 211–29. Darley, J., and Batson, C. 1973. From Jericho to Jerusalem: A study of situational and dispositional variables in helping behavior. Journal of Personality and Social Psychology, 27, 100–8. Doris, J. 2002. Lack of Character: Personality & Moral Behavior. New York: Cambridge University Press. Fischer, J., and Ravizza, M. (eds.). 1993. Perspectives on Moral Responsibility. Ithaca, NY: Cornell University Press. Fiske, S., and Borgida, E. 2008. Providing expert knowledge in an adversarial context: Social cognitive science in employment discrimination cases. Annual Review of Law and Social Science, 4, 123–48. Fiske, S., and Taylor, S. 2013. Social Cognition: From Brains to Culture. London: SAGE Publishers. Frankfurt, H. 1971. Freedom of the will and the concept of a person. Journal of Philosophy, 68, 829–39. Kahneman, D. 2011. Thinking Fast and Slow. Toronto: Doubleday Canada. Kahneman, D., Knetsch, J. L., and Thaler, R. H. 2005. Experimental tests of the endowment effect and the Coase theorem. In D. Kahneman, J. L. Knetsch, and R. H. Thaler (eds.), Negotiation, Decision Making, and Conflict Management, pp. 92–115. Northampton, MA: Edward Elgar Publishing. Kunda, Z. 1999. Social Cognition: Making Sense of People. Cambridge, MA: MIT Press. Latané, B., and Darley, J. 1970. The Unresponsive Bystander:  Why Doesn’t He Help? New York: Appleton-Century Crofts. Latané, B., and Rodin, J. 1969. A lady in distress: Inhibiting effects of friends and strangers of bystander intervention. Journal of Experimental Social Psychology, 5, 189–202. Maibom, H. 2013. Values, sanity, and responsibility. In D. Shoemaker (ed.), Oxford Studies in Responsibility, pp. 263–83. New York: Oxford University Press. Moss-Racusin, C., Dovidio, J., Brescoll, V., Graham, M, and Handelsman, J. 2012. Science faculty subtle gender bias favors male students. PNAS, 109, 16474–9. Nisbett, R., and Wilson, T. 1977. Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231–59.

122  Heidi Maibom Ross, L., and Ward, A. 1996. Naïve realism in everyday life: Implications for social conflict and misunderstanding. In T. Brown, E. Reed, and E. Turiel (eds.), Values and Knowledge, pp. 103– 35 Hillsdale, NJ: Erlbaum. Sher, G. 2009. Who Knew? Responsibility without Awareness. New York: Oxford University Press. Superson, A. 1999. Sexism in the classroom: The role of gender stereotypes in the evaluation of female faculty. APA Newsletter on Feminism and Philosophy, 99, 46–51. Steinpreis, R., Anders, K., and Ritzke, D. 1999. The impact of gender on the evaluation of the curricula vitae of job applicants and tenure candidates: A national empirical study. Sex Roles, 41, 509–28. Williams, J. 2003. The social psychology of stereotyping: Using social science to litigate gender discrimination cases and defang the “cluelessness” defense. Employee Rights & Employment Policy Journal, 7, 401. Wilson, T. 2002. Strangers to Ourselves:  Discovering the Adaptive Unconscious. Cambridge, MA: The Belknap Press of Harvard University Press. Wolf, S. 2003. Sanity and the metaphysics of responsibility. In G. Watson (ed.), Free Will, 2nd edn., pp. 372–87. New York: Oxford University Press.

6 Meta-Cognition, Mind-Reading, and Humean Moral Agency Julia Driver

 . . . if a creature be generous, kind, constant, compassionate; yet if he cannot reflect on what he himself does or sees others do, so as to take notice of what is worthy or honest; and make that notice or conception of worth and honesty to be an object of his affection; he has not the character of being virtuous; for thus, and no otherwise, he is capable of having a sense of right or wrong; a sentiment or judgment of what is done, through just, equal, and good affection, or the contrary. (Lord Shaftesbury, An Inquiry Concerning Virtue or Merit, R, 174)

Early moral sentimentalists, such as Shaftesbury and Hume, associated moral agency with a capacity for a special sort of self-appraisal, or self-regulation; moral evaluation correspondingly required of the evaluator the capacity to reflect on the mental states of others. Shaftesbury regarded these capacities as crucial not only for moral agency, but for virtue. This is one very significant and underappreciated way in which Hume differs from other sentimentalists such as Shaftesbury. Hume made distinctions— granted, implicit—between what we would term moral appraisability and moral accountability, as well as distinctions between being appraisable and being capable of appraisal. Only beings that possess the capacity for self-regulation via appraisal are moral agents, though other beings who act in self-directed ways via beliefs and desires are agents, albeit simple agents. Over the last twenty years in the moral psychology empirical literature a great deal of work has been done in investigating these capacities. In the psychology literature a distinction is made between meta-cognition (mental states regarding one’s own mental states) and mind-reading (detecting and having ones own mental states that make reference to the mental states of others). My aim in this chapter is to discuss the sentimentalist account of self-regulation at the heart of its account of moral agency in light of this literature, focusing on a potential problem for the view

124  Julia Driver that arises when we consider various models of how meta-cognitive appraisal works in relation to mind-reading.1 One issue I’ve discovered in reading this literature is that it is sometimes not clear what capacity, exactly, is being discussed. For the sake of clarity I would like to distinguish four “meta-capacities” that are in play: 1. The capacity to detect the mental states of others: mind-reading 2. The capacity to approve or disapprove of the mental states of others 3. The capacity to detect one’s own mental states: introspection or self-directed mind-reading 4. The capacity to approve or disapprove one’s own mental states:  a type of meta-cognition I should note that these are my own way of drawing distinctions between the relevant capacities I would like to discuss in relation to moral agency. Moral agency, as opposed to mere or simple agency, is the agency that underlies distinctly moral action. This is understood in contrast to other sorts of agency. So, for example, what I term “mere” agency is the sort of agency that one sees in animal behavior as well as a good deal of human behavior. This is action that is the result of features of psychology such as beliefs and desires, motives, intentions. However, mere agency is not regulated fully in the same way as moral agency. I may regulate mere agency in the following way: I desire an ice cream cone, and believe that I can get one at the local supermarket. It turns out that the local supermarket has stopped selling ice cream cones. Once I find this out, I change my plans. That’s a kind of regulation in which agency is sensitive to new information. This is a kind of self-regulation—for example, one adjusts one’s desires relative to the new information. However, the sort of self-regulation involved in distinctly moral agency involves approval or disapproval of the mental states themselves. And the approval/disapproval is of a distinctly moral sort—rather than, say, aesthetic.2 This is one reason why moral agency is characterized by meta-capacities. However, my claim is not that it is a necessary condition for moral agency. My claim is that it is very important—given other features of our psychology—for effective moral agency. I suspect that it is, in fact, necessary for limited agents and will consider challenges to that claim towards the end of the chapter. However, one of my goals is to show that there can be defects to moral agency that don’t obliterate moral agency itself, but,

1 A note about terminology: Psychologists use “meta-cognition” to refer to the capacity to think about one’s own thoughts, whereas “mind-reading” refers to the capacity to detect the mental states of others. Both forms of meta-reflection involve thinking about thoughts—either one’s own or others. I will be using “meta-cognition” broadly to include not simply having certain beliefs about one’s own mental states, but also to having certain emotional responses to those mental states. The same for “mind-reading.” 2 The phenomenology of moral and aesthetic approval/disapproval may be similar, but in other respects they are distinct.

Meta-Cognition, Mind-Reading, and Humean Moral Agency  125 rather, sharply reduce effectiveness of that agency. This helps to diagnose certain criticisms of sentimentalist accounts of moral agency since some writers confuse being a moral agent with being an effective moral agent. One example is presented by Jesse Prinz’s attack on a kind of sentimentalism that is committed to what he calls the “Humean Precondition Thesis”—which holds that empathy or sympathy is a precondition to moral judgment (Prinz 2011).3 Prinz discusses several ways this precondition claim can be understood. One version is as a kind of epistemic precondition claim in which empathy (or sympathy, as Hume calls it) is needed by the agent to acquire information needed to make a moral judgment (and presumably be guided by such). Prinz views empathy as a kind of emotional contagion, and, of course, there is much more to it than that if one looks at all the ways Hume employs “sympathy” in his writings. Sympathetic engagement with others is sometimes a matter of feeling what they feel, but is also a matter of feeling for them. In terms of its epistemic value it alerts the observer to putative sources of value and disvalue. It needn’t do this infallibly to play a really crucial role in effective agency, and it is not a requirement of moral agency per se or the capacity to make any sort of moral judgment at all. Prinz also makes the familiar point that mere emotion—whether it is pure contagion or the result of making an inference—is frequently distorted by bias, prejudice, mood, and a multitude of other factors. But as I’ve noted elsewhere, this is something the early sentimentalists were well aware of. This is the whole reason for providing a corrective mechanism such as the general point of view, or an ideal observer standard. What is often ignored in these sorts of criticisms of the sentimentalist model is the role of meta-capacities of evaluation. A person who is sympathetic inappropriately has the chance to self-regulate and judge, through a variety of corrective mechanisms, that the sympathy is inappropriate. This may change the feeling of sympathy, but then again, it may only result in the judgment that the feeling is not appropriate and ought not be acted on. That is why such capacities are crucial for moral agency, though not simple agency.

Agency and Moral Agency Moral agents can be held accountable for their actions in ways that it would not make sense to hold mere agents accountable. I will not argue for the view here, but I hold agents to be creatures capable of acting in self-directed ways—through a combination of beliefs and desires. For this reason, I believe that many animals are agents, though



I argue against this view in Driver (2011b).

3

126  Julia Driver not moral agents.4 Holding accountable is a crucial feature of our critical practices, and is distinct from evaluating, or appraising, which is also a significant feature of our critical practices. Some argue that holding accountable is warranted, not by the consequences of holding accountable, but rather by whether or not (1) the being deserves to be held accountable and (2) the being holding accountable has the right kind of authority (Strawson 1962). However, this account of the conditions that need to be met in specific cases of holding accountable is compatible with the view that such a system of holding accountable does serve our interests; that it is justified pragmatically, as well as only warranted in the ways discussed by Strawson. Accountability misfires (just as harming the innocent cannot be thought of as genuine punishment) if the person being held accountable has not done anything wrong; but the system is justified normatively on the basis of how critical practices function to promote things like social utility. Holding accountable, as opposed to simply evaluating (as one might to oneself, in private), has as paradigmatic the case in which someone is confronted about doing something wrong. To be appropriately held accountable, on the Humean view (and in common sense), one needs to have access to the mental states that are being criticized. A person who cannot access his own mental states will not be in a position to understand criticism of his or her own mental states. Rocks have no mental states, and so there is no point in blaming a rock for falling on one’s foot. But there may be other beings that have mental states, and yet lack any access to those mental states (let’s assume here very extreme cases in which there is no inferential access as well as no direct access). In such a case, again, blame would do no good (leaving aside its general good effects, but then those might hold for blaming rocks as well). Indeed, I think it quite plausible that one thing we want to accomplish in holding accountable is the generation of meta-cognitive states—beliefs, for example, that one’s motives may have been vicious; shame at recognition of those motives, and so forth. This kind of self-appraisal requires access to one’s own mental states. Sometimes this self-appraisal is spontaneous, but it is also prompted by being held accountable by another person. The self-appraisal need not be something that is cognitively sophisticated in the sense of involving deep reflection. Sometimes, of course, people do engage in conscious reflection on their own characters. More often, though, we regulate via cues that we get from others we are interacting with. A frown, a lifted eyebrow, a skeptical sound— these things can lead the moral agent to adjust attitudes and behavior. For example, a person may become extremely angry at a perceived slight, but then question the appropriateness of the depth of the anger based on the responses others have to her expressions of anger. Signs of surprise, alarm, and distancing can be subtle and need not involve any verbal evaluation at all. Navigating the social environment effectively requires this kind of sensitivity. Of course, we don’t inevitably change our attitudes in response to how we perceive others perceiving those attitudes—we may continue

I discuss this in more detail in Driver (2011a).

4

Meta-Cognition, Mind-Reading, and Humean Moral Agency  127 to think ourselves justified in our responses. But the process of self-regulation often works like this—not consciously in the sense of being consciously deliberative, but automatically. We may have access to considerations that we were responsive to after the fact (and in this way we are like third person observers of our past selves), but at the time the regulation will be automatic. However, one feature of the simple model has been challenged by work suggesting that we do not engage in direct introspection, and that, instead, what we term metacognition is actually self-directed mind-reading. Peter Carruthers has recently argued that there is no such thing as introspective access to our own mental states (except for direct sensory perceptions such as pain). This means that in self-regulation the agent is not typically proceeding with direct access to mental states which he or she then evaluates. There is almost always an essentially interpretive element, which may itself involve some kind of proto-evaluation—that is, evaluation that is not conscious. This is fine for the sentimentalist—indeed, it may even strengthen the sentimentalist claim for empirical accuracy, since part of the interpretation is likely to involve or engage our emotions and conative commitments (Carruthers 2011). Another crucial feature of moral agency in the sentimentalist tradition is the capacity to care about the well-being of others. On a Humean model, it is the capacity for sympathetic engagement with others. Without caring (and, possibly, second-order caring needed for self-regulation), it isn’t possible for the person to self-regulate in the appropriate way, by approving or disapproving of her own mental states on the basis of what we take to be morally significant reasons—that is, reasons having to do with the well-being of others. A misunderstanding of Hume’s views on sympathy has led to mistaken criticisms of the sentimentalist approach—usually having to do with a conflation of different ways in which we use the words “empathy” and “sympathy.” For example, writers such as Jeannette Kennett argue that Hume must be mistaken, otherwise persons with autism, a kind of “empathy deficit,” would not count as moral agents, and yet they are clearly moral agents (Kennett 2002). But Kennett’s criticism depends upon mixing up the different ways in which sympathy works—as a way to get information, and as a source of our affective responses to others.5 These are quite different. Persons with autism have trouble picking up on the mental states of others, and this can interfere with their effectiveness as agents when it comes to navigating social interactions. However, their affective dispositions are just as Hume indicates for moral agents. It is, rather, psychopaths who are defective in terms of affective dispositions. They lack the capacity to sympathetically engage, though they may be very skilled at figuring out what other people want, and thus epistemically proficient.6 However, one key difference between psychopaths who lack empathy and persons with normal affect 5 I discuss this criticism of Kennett’s views on Hume in Driver (2011b) and Driver (2013). 6 Very many psychopaths are also poor with respect to this skill. Psychopathy is measured along twenty distinct parameters, and an overall score that puts someone in the psychopathy range can reflect deep inabilities to detect what others want. However, there are psychopaths who are strong epistemically, but score high overall due to deep failures to connect affectively with others and due to impulsive and criminal tendencies.

128  Julia Driver is that psychopaths may be able to determine what others are thinking, or what they find plausible, and yet not understand emotions—they don’t pick up on emotions in the contagion sense. This is a kind of epistemic limitation, as well, since they won’t have direct access to the emotions. This could explain why some are baffled at the anger their victims. The significance of these functions can be seen when we examine in more detail the ways in which moral agency can be either missing, or in some way defective. A great deal of interesting data can be obtained by studying patients displaying lack or loss of affect. In fact, some psychologists are writing about lack of affect itself, since it is seen as associated with several mental illnesses, though many manifest differently with different illnesses or conditions (Cohen et al. 2012).

Affective Disorders There are very many psychological disorders relating to affect. The ones I discuss here are associated with loss of affect. That is, these are disorders that are marked by the agent failing to exhibit or feel emotion, to the point that motivation is inhibited. There are other ways in which one can have an affective disorder. For examples, some disorders, such as mania, may be characterized by overly positive emotional responses. Though I haven’t seen discussion of this in the empirical literature, there are probably some people who suffer from a surfeit of sympathy. Imagine someone who grieves a disproportionately long time over the death of a pet goldfish. Or, consider someone who obsesses over the misfortunes of others—genuinely serious misfortunes—to the extent that he cannot function properly. Generally, these deficits are partly responsible for self-regulation failure. Sometimes self-regulation failure has no moral side: It may make an agent imprudent, for example, as Damasio’s case of Elliot exhibits. Elliot was a person who, like Phineas Gage, suffered frontal lobe damage that affected his subsequent behavior. He scored highly on intellectual tests—he had good working memory and normal intelligence. However, when it came to actually making decisions and acting on them—particularly when it came to actions in social settings—he did very badly. After initial puzzlement, Damasio noticed that Elliot’s emotions seemed out of sync with his experience: Elliot was able to recount the tragedy of his life with a detachment that was out of step with the magnitude of the events. He was always controlled, always describing scenes as a dispassionate, uninvolved spectator . . . He was not inhibiting the expression of internal emotional resonance or hushing inner turmoil. He simply did not have any turmoil to hush. (Damasio 1995, 44)

The failures are practical, and these sorts of failures will often also have moral repercussions. Failure to follow through can be imprudent, but can also lead to failures in responsible behavior, such as a failure to keep one’s promises, or live up to other obligations.

Meta-Cognition, Mind-Reading, and Humean Moral Agency  129 There are broadly three ways in which there are affective failures. As in Elliot’s case, one can seem to be utterly lacking in the appropriate affect. However, there are other cases where affect is present, but out of proportion, and other cases where affect is present and yet seems misdirected somehow. Hume discusses ways in which we intuitively view the passions as somehow unreasonable, or mistaken. He needs to account for this, since on his system passions cannot be “true” or “false,” strictly speaking. Passion is not “unreasonable” since reason is concerned with truth and falsehood. However, this doesn’t mean that passions cannot be defective, and, loosely speaking, deemed unreasonable. He writes:  . . . ’tis only in two senses, that any affection can be call’d unreasonable. First, When a passion, such as hope or fear, grief or joy, despair or security, is founded on the supposition of the existence of objects, which really do not exist. Secondly, When in exerting any passion in action, we chuse means insufficient for the desired end, and deceive ourselves in our judgment of causes and effects. (Hume 2011, T 2.3.3)

So, if someone is angry because he falsely believes someone has harmed him, that anger is unreasonable, and defective. If someone’s anger is so intense that it causes that person to be confrontational, when confrontation is actually counter to that person’s aims, then this also means, on Hume’s view, that the emotion is unreasonable, and defective.7 There are cases Hume doesn’t discuss in which emotions seem unreasonable—as when they are out of proportion. The angry person may not be mistaken in a belief about the conditions causing his anger, and it may even be that the anger does not run counter to his own aims—even so, the anger may be excessive. One important role for emotional self-regulation is to make sure that we feel the emotions we ought to feel, and as we ought to feel them. To some extent—in the cases Hume discusses, where an emotion is based on a false belief—the self-regulation will be a form of cognitive self-regulation. The reasonable agent may question whether or not he or she has adequate evidence to support the belief in question. But the proportionality cases don’t seem to work like this. Some cases can be handled similarly—Hilary may question her anger’s appropriateness in light of her beliefs about her own goals, and so forth. That is, she may regulate her anger out of a belief that such strong feelings of anger will put people off, and thus make her goal of having friends more difficult to attain. Certainly that is one way in which disproportionate emotional responses can be diagnosed as disproportionate. But this doesn’t exhaust the cases. For example, a person may realize at the time she experiences the anger that it is disproportionate. Indeed, this is one factor that goes into self-regulation. What makes it disproportionate in this case? There are a variety of ways one could go in accounting for this. One very intuitive way would be to say that emotional responses should reflect value appropriately. If someone experiences more intense grief at the loss of a beloved goldfish than at the loss of a beloved

7 This defect might be reducible to the first, if what is involved is a false belief that anger is not counterproductive.

130  Julia Driver parent, this is likely disproportionate in some way—either too much grief for the goldfish, or too little for the parent, or both. Of course, we could be appealing to a person’s beliefs about value here. But we may not be—in cases where self-regulation kicks in the person may well believe that the goldfish doesn’t deserve such a response, and yet feel it anyway. The second-order attitude is informed by the belief, not the first-order emotional response. So, perhaps this is another way in which Hume would allow for emotions as “unreasonable.”

Acquired Sociopathy and Loss of Affect Hume himself had the rather optimistic view that a capacity for sympathy was universal in human nature.8 Unfortunately, at least some psychopaths seem to present counterexamples to that claim. However, Hume is on firmer ground here if the claim is restricted to normal humans—humans with the capacities that they, as social beings, evolved to have. Jeannette Kennett reminds us that in discussing psychopathy we need to avoid images of psychopaths that predominate in popular culture. Most actual psychopaths act in ways “characterized by impulsivity and irresponsibility; they are habitual liars, are indifferent to the rights of others, and display lack of remorse for wrongdoing. When they do something wrong, they do not really see what the fuss is about . . .” (Kennett 2002, 341). Thus, though they are lacking in sympathy, they suffer from many other deficits as well. Much as been written on psychopaths and their lack of affect. There are very interesting intermediate cases in the empirical literature. These are cases what Antonio Damasio termed “acquired sociopathy” of which psychopathy is the developmental correlate. I will discuss persons who have suffered severe personality changes after suffering brain trauma to the orbitofrontal cortex. These persons developed normally as human beings prior to the brain damage. After the brain damage they engaged in socially inappropriate behavior, exhibiting lack of appropriate affect, in ways similar to psychopathic behavior, though these individuals would not score highly enough on the standard diagnostic test to be classified as psychopaths. However, part of the reason for this is that their deficits are primarily in the lack of affect categories. Jennifer S. Beer et al. note that the empirical literature supports the view that “patients with orbitofronal damage behave with strangers in ways that are more appropriate for interactions with close others. Patients with orbitofrontal damage tease strangers in inappropriate ways and are more likely to include unnecessary personal information or tangential information when answering questions” (Beer et al. 2006, 871). But what Beer et al. themselves discovered is that these patients were not able to integrate first

8 I discuss Hume’s views on sympathy, and the different senses employed in different parts of his work, in Driver (2013) and Driver (2011a).

Meta-Cognition, Mind-Reading, and Humean Moral Agency  131 person and third person self-monitoring. That is, at the time they behaved inappropriately they were not able to see that they were behaving inappropriately, but when they later viewed a video tape of their own behavior they did become embarrassed, and agreed they behaved inappropriately. They suffer a major deficit in self-monitoring when they are unable to react appropriately while actually engaged in social interaction. Their main deficits occur in the categories of affect and self-monitoring. One theory accounting for the inappropriate social interactions is Damasio’s view that such persons cannot properly interpret their emotional responses. Another theory that Beer discusses in her research places more importance on self-monitoring deficits: It “emphasizes the importance of the orbitofrontal cortex . . . or the ability to evaluate one’s behavior in the moment in reference to higher order goals or the reactions of other people. . .” (Beer et al. 2006, 872). Persons with orbitofrontal cortex damage would be embarrassed by the video tapes of their social mistakes. Beer suggests that this may offer a kind of therapy for them—though the efficacy of video tape training has not been tested. One explanation is that taking a third person perspective may help people evaluate not just others, but themselves as well, especially when “on-line” self-monitoring is somehow defective. Further, it may offer a kind of back-up, or complimentary, system. Evaluating and monitoring are the same kind of thing, and monitoring is very important to moral agency.

Clinical Depression and Loss of Affect Psychopathy is not the only psychological condition marked by some form of inhibited affect. Some persons suffering from clinical depression also report loss of normal affect. These cases differ from the acquired-sociopaths in that the persons can often at the time they are acting note that they lack normal affect—there is no disconnect between the first and third person.9 Clinical depression is often characterized by anhedonia. Generally this is understood as the failure to derive pleasure from one’s experiences. However, there is a good deal of disagreement over exactly how to characterize anhedonia. One way to test for the presence of anhedonia is to look for “reduced capacity to sustain positive emotion.” In one study depressed individuals were shown both positive, negative, and neutral images while having their brains scanned via fMRI. They were also asked to try to enhance the mood presented in the images. As a result, the study’s designers found that—among other things—it was much more difficult for them to sustain positive emotion, though not negative emotion (Heller et al. 2009). Further, this worries or bothers the people experiencing it. They realize that there is something wrong with their responses, and often feel motivated to do something about it.

9 DSM-IV lists among the cluster of symptoms associated with a major depressive episode: “Markedly diminished interest or pleasure in all or almost all activities.” This is referred to as “anhedonia.”

132  Julia Driver There has been some work in the philosophical literature on desires to desire. One philosopher has suggested that we have a desire to have desires as a way of analyzing Schopenhauer’s pessimism.10 Let’s assume that a morally good person desires to do the right thing, and, not only that, desires to have that very desire. (1) Maria desires to desire to act in a morally appropriate way. (2) Maria desires to desire to help those in need. (3) Maria desires to desire to help Arthur, because he needs her help and he is deserving. A person suffering from depression might possess all of these desires, and lack each of the first order desires. If a person is like Maria in (1) she has a standing desire to be morally good. This standing desire is itself motivating. A person who is depressed and who is not experiencing first order desires to act in ways that are morally good—perhaps he is indifferent to the needs of others, watches reports on TV of disasters, and feels nothing, and so forth—may nevertheless notice that fact about himself and be bothered by that because it conflicts with the second order desire. The second order desire along with the realization that to fulfill that desire he needs treatment, will motivate seeking treatment. It may be that case that with extreme versions of psychopathy we don’t have such standing second order desires. At best one might get puzzlement at the mismatch between the psychopaths first order desires and what seem to be the standing second order desires of normal agents. Recently some researchers on anhedonia have argued for a more diverse understanding of how it functions, arguing that in some cases, anyway, anhedonia should be understood in terms of “impaired ability for normative decision-making” (Treadway and Zald 2010). What is particularly interesting here is the claim that part of the impairment has to do with misrepresenting “the value of future rewards in a consistent manner” (Treadway and Zald 2010, 4). At least some part of the impairment has to do with various ways depressed individuals fail to model value. They don’t pick up on it at all, or they misrepresent it either overestimating or underestimating future costs and benefits. Other studies also suggest that human decision-making is adversely affected by changes in the orbitofrontal cortex that inhibit the agent’s ability to form desires on the basis of good information about prospective benefits. The researchers in this study hypothesized that in healthy people reward cues are picked up on and are attended with the appropriate weight. This allows them to make good decisions about whether or not they ought to pursue a particular course of action. When this is lacking, their possible choices are not attended with the right kind of reward cues—in other words, there is valuable information that they are not picking up on, and this might at least be partly responsible for their failure to derive pleasure and their failure to be motivated to pursue a particular course of action that external observers can see ought to be rewarding for them (Rogers et al. 2003). This also has implications for their

See Fernandez (2006).

10

Meta-Cognition, Mind-Reading, and Humean Moral Agency  133 ability to engage in reliable affective forecasting—that is, they are unable to effectively predict how they will feel if they pursue a particular course of action. These are all impediments to effective agency. When the deficits involve an inability to feel appropriately in the face of value that we typically take to be moral—e.g., well-being indicators for other people—these are deficits in effective moral agency. Self-regulation is not working properly. Based on the literature, we can divide the defects in practical reasoning associated with anhedonia into three broad categories: (1) Motivational anhedonia: Because the agent fails to represent a future experience as valuable, the agent fails to be properly motivated to engage in beneficial behavior (Treadway et al. 2009). (2) Failure to register rewards: This is what some researches refer to as a failure in “hedonic response.” What would normally be a pleasurable experience, or a rewarding experience, is not registered as such by the depressed agent. (3) Deliberative anhedonia:  This involves a failure to represent pleasurable responses in one’s deliberations about what to do. Deliberative anhedonia results in actions that can be bad for the agent, and bad for others. Let us suppose that self-regulation must involve the ability to pick up on bits of value in the world.11 Let us also suppose that when we are talking about morality that value will involve the well-being, the flourishing, of others. If there is a problem detecting when such value is at stake, there will be a problem with effective moral agency. Whether this is a deep moral problem will depend on one’s normative commitments. Consider how this issue has arisen in the literature on certain forms of moral particularism. Particularists such as John McDowell and Lawrence Blum have made much of the role moral perception plays in guiding action (McDowell 1979; Blum 1991).12 In the specific case of particularist moral theory the contrast is with rules or decision procedures as guides to action. I have no such contrast in mind, but do think that the issue of moral perception is important to understanding how agency can go wrong. To see this, we need a distinction between the sort of caring about the well-being of others that I call “general” beneficence. All by itself this is insufficient to guide behavior appropriately. We also need the agent to be able to pick up on what is morally significant. We can determine that the agent has the right kind of caring attitude if we know that “were the agent to pick up on what was morally significant, the agent would act appropriately” is true. That is, that there is genuine caring depends upon this counterfactual or conditional test. It is important because it marks the distinction between two completely different sorts of moral defect: insensitivity and cruelty. The agent who cares, but fails to detect moral value, is insensitive, not cruel. The agent who detects moral 11 We can leave unexplored, for now, the basis for that value. 12 On my view, these writers conflate moral perception with virtue. It is crucial to effective agency, as I will argue, but that is a different issue.

134  Julia Driver value, and doesn’t care—or doesn’t care in the right way, anyway—is cruel. This person might see the suffering of others as suffering, and actually be happy about it. But even if not happy, the seeing and perceiving of suffering as suffering, and the failure to appreciate it morally, is a defect that is distinct from the failure to pick up on the suffering in the first place. Really, then, there are three states that are in some way morally defective along this parameter: insensitivity to evidence: one cares, but doesn’t act on the caring because one doesn’t see any suffering; another sort of insensitivity where one doesn’t care at all, and may or may not perceive suffering; and cruelty, where one cares in a way—but in a twisted way, where the suffering of others actually gives one pleasure, so one’s reward mechanism is pathological. Why treat the affective capacity as so important? Here I will appeal to traditional philosophical arguments bolstered by evidence from the psychology of emotions and conative states. It is the affective capacity that is absolutely crucial to motivation. Without it there is no agency. However, without a well-functioning epistemic capacity, there is no effective agency. The two interact. As noted earlier, patients with a depressive disorder fall short in a variety of ways. Some are quite poor when it comes to what psychologists call “affective forecasting.” That is, they lack the ability to properly or accurately predict their own future affective states with respect to relevant alternative courses of action. There can be third person manifestations of this as well. It is an interesting empirical issue as to how much these defects go together. I predict that, as with the cases of acquired sociopathy discussed earlier, they can come apart. Affective forecasting is crucial for self-regulation on the sentimentalist model of self-regulation. Consider the following cases: (Mary) Mary suffers from Major Depressive Disorder (MDD). This results in apathy as she cannot foresee that one course of action (vacationing in Mexico) over another course of action (spending her vacation in bed at home) would result in greater amounts of pleasure for her. (Bob) Bob is unable to foresee that one course of action (attending his parents’ thirtieth wedding anniversary) would make them happier than another course of action (sending them an anniversary card). The failure in affective forecasting in Bob’s case is the sort of insensitivity I mentioned earlier. Persons with MDD may also have this sort of insensitivity, but in principle it is separable from first personal decisional anhedonia. I think that in cases such as Bob’s we also suspect that one reason why he doesn’t do a good job of affective forecasting is that he just doesn’t care enough. But, again, in principle the failure to care is separable from the failure to pick up on sources of value. Failure of affective forecasting has seriously negative implications for an agent’s ability to effectively plan, especially when it comes to moral planning, or considering courses of action from the moral perspective. One cannot plan effectively without being able to predict how people will feel following one’s action. In the case of Bob, he is a blunderer, even if he is a well-intentioned one. We may deem him virtuous if our view of virtue is

Meta-Cognition, Mind-Reading, and Humean Moral Agency  135 purely internal, if all we care about is whether or not he cares, or whether or not his basic desires are the sort we can endorse. Nevertheless, we can also view his agency as itself defective because, in spite of being well-intentioned, he is systematically wrong when it comes to detecting value and potential sources of value. This relates to the earlier issue of the impact autism has, also on affective forecasting. One source of the difficulty can lie in the empathy deficit disorder itself, where the deficit in question involves an epistemic failure tied to perspective taking. In this case a person may know that, for example, embarrassing situations are bad for people because they cause pain and social anxiety, but not be able to put themselves in situations where they can anticipate what is embarrassing for another person. This sort of failure to detect disvalue is different from the failure to understand at the start that embarrassment is bad. Some failures to detect value/disvalue involve a kind of perversity of tastes. Sometimes, as in the case of extreme psychopathy, the perversity is due to a failure to care positively about the well-being of others at all. Sometimes, there might be a milder social deficit in which the agent cares abstractly, but fails to appreciate the full range of factors that harm others (such as embarrassing situations). And, sometimes, the agent cares, the agent realizes the embarrassment harms people, but then can’t put himself in the shoes of another to figure out what the person would view as embarrassing, or what would, objectively, be embarrassing for that person. I think most would agree that the really extreme deficit is the failure to care at all. The sentimentalist intuition behind this is the view that epistemic failures, while unfortunate, and while they do impact effectiveness, don’t speak to the core of a person’s character in the same way. They seem remedial in a way that the character core is not. This, in turn, can be cashed out in terms of value commitments. Self-regulation of affective states offers a chance of self-remediation. Frequently, it is not conscious and is simply prompted by cues in the environment which get us to respond by tamping down our emotions or revving them up. In any case, reflective or not, a standard is being employed by the agent. The standard may be set by the people around her, or by an idealization of herself (“That’s not the way I want to be”). But any criticism of the sentimentalist account of moral agency must consider not just the way the first order emotions go wrong, but the ways in which the second order beliefs and emotions about those emotions goes wrong. Any account of the sentimentalist view of moral agency which ignores self-regulation either descriptively or as providing a standard by which our emotional responses may be criticized just is not discussing the strongest, most plausible, view of sentimentalism. We clearly do engage in self-correction. Whether, of course, that self-correction is in the right direction will depend on the correctness of the standard itself.

Acknowledgments I would like to thank members of the Workshop on Moral Psychology held at the University of Michigan in June 2012, and particularly Dan Jacobson and Justin D’Arms for their detailed written comments on an earlier draft. I  would also like to thank

136  Julia Driver members of SLEW (the Saint Louis Ethics Workshop) for very helpful discussion of an earlier draft of the paper.

References Beer, J., et. al. 2006. Orbitofrontal cortex and social behavior: Integrating self-monitoring and emotion-cognition interactions,. Journal of Cognitive Neuroscience, 18, 871–9. Blum, L. 1991. Moral perception and particularity. Ethics, 101, 701–25. Carruthers, P. 2011. The Opacity of Mind: An Integrative Theory of Self-Knowledge. Oxford: Oxford University Press. Cohen, A.  S., et. al. 2012. On the boundaries of blunt affect/alogia across severe mental illness: Implications for research domain criteria. Schizophr. Res. 140(1–3), 41–5. Damasio, A. 1995. Descartes’ Error. New York: Harper. Driver, J. 2011a. A Humean account of the status and character of animals. In T. Beauchamp and R. Frey (eds.), The Oxford Handbook of Animal Ethics. Oxford: Oxford University Press. Driver, J. 2011b. The secret chain:  A  limited defense of sympathy. The Southern Journal of Philosophy, 49(1), 234–8. Driver, J. 2013. Moral Sense and Sentimentalism. In Roger Crisp (ed.), The Oxford Handbook of the History of Ethics. Oxford: Oxford University Press. Fernandez, J. 2006. Schopenhauer’s pessimism. Philosophy and Public Affairs (November), 646–64. Heller, A. S., Johnstone, T., Shackman, A. J., Light, S., Peterson, M. J., Kolden, G. G., Kalin, N. H., and Davidson, R. J. 2009. Reduced capacity to sustain positive emotion in major depression reflects diminished maintenance of fronto-striatal brain activation. PNAS Early Edition, 2 December. Hume, D. 2011. A Treatise of Human Nature, ed. D. F. Norton and M. Norton. Oxford: Oxford University Press. Kennett, J. 2002. Autism, empathy, and moral agency. The Philosophical Quarterly, 52, 340–57. McDowell, J. 1979. Virtue and reason. The Monist, 62, 331–50. Prinz, J. 2011. Against empathy. The Southern Journal of Philosophy, 49, 214–33. Rogers, R. D. et al. 2003. Tryptophan depletion alters the decision-making of healthy volunteers through altered processing of reward cues. Neuropsychopharmacology, 28, 153–62. Strawson, P. F. 1962. Freedom and resentment. Proceedings of the Aristotelian Society, 48, 1–25. Treadway, M., and Zald, D. 2010. Reconsidering anhedonia in depression: Lessons from translational neuroscience. Neuroscience and Behavioral Reviews, 35(3), 537–55. Treadway, M. T. et al. 2009. Worth the “EEfRT”? The Effort Expenditure for Rewards Task as an objective measure of motivation and anhedonia. PLOS ONE, 4(8).

7 The Episodic Sense of Self Shaun Nichols

When social psychologists talk about the concept of self, they generally mean a collection of psychological traits. Here is a characteristic description: The self-concept refers to a person’s understanding of what she “is like” as a person, that is, what personality characteristics she manifests, what idiosyncratic abilities and proclivities define her as an individual, and to what extent she regards herself positively (i.e. has high or low self-esteem). (Mitchell 2009, 247)

This is a thick way of conceiving the self, chock full of aspirations, convictions, affections, and affiliations, and it is, of course, a useful way to think about the self. Features of this self-concept predict a person’s general well-being, success in life, social desirability, and a host of other important practical concerns. Philosophers have often been interested in a much thinner conception of self. A thin notion of self is championed by Reid and Butler, who maintain that the traits that one happens to have should be construed as belonging to the self, rather than constituting the self. Kant, on the other hand, maintains that a thin conception of self, the I, is “the poorest of all representations” (Kant 1968, B408), and Kant thinks this conception engenders the illusions of rational psychology—illusions that there is a self that is simple, substantial, and identical across time. It is a matter of some controversy whether Kant should be read as speaking only about the rational psychologists (e.g., Descartes), or whether he thinks that ordinary people have a thin I-concept that leads to a lay illusion of identity across time. Regardless of the exegetical issue, we can explore whether people do in fact have a thin self-concept that is to be sharply distinguished from the trait notion of self. In section 1 of this chapter, I’ll argue that ordinary people exploit both the psychologist’s trait conception and a very thin conception of self, which resonates with philosophical discussions. This thin conception of self is linked, I suggest, with episodic memory. In section 2, I distinguish the sense of identity that is delivered by episodic memory from the biological conception of identity. Section 3 will look at moral

138  Shaun Nichols emotions and the representation of self. And in the final section, I will explore the propriety of judgments and feelings that are based on the episodic sense of self.

1.  Memory and Representations of Self Our interest is primarily in how people think about the self across time. Since memory will be an important player, a brief review is in order. Within explicit or declarative memory, philosophers and psychologists distinguish between two fundamental types of memory: episodic and semantic. Episodic memory is memory of experience—you remember having an experience. For instance, when you recall the experience of having dinner last night, that recollection will be an episodic memory. Semantic memory, by contrast, houses your knowledge of facts, like your knowledge that George Washington was the first US President and that Detroit is in Michigan. Of course, one can have this kind of factual knowledge about oneself, e.g., I know that I was born in Montana. This does not come from my recalling the experience of my birth. Rather, this is part of semantic memory.1 One characteristic feature of episodic memory is that it delivers www:  what, where, and when. I remember having dinner, at a downtown restaurant, last night at 7. Although www is typical of episodic memory, the standard characterization holds that in addition to www, one must also remember the experience as happening to oneself. The memory has to present the experience as having happened to me (Tulving 1985). The idea here goes back at least to William James, who wrote, “Memory requires more than the mere dating of a fact in the past. It must be dated in my past. In other words, I must think that I directly experienced its occurrence” (1890/1950, 650). In addition, contemporary memory theorists hold that many of the same processes that are implicated in episodic memory are also involved in episodic foresight— considering what will (or might) happen in the future (e.g. De Brigard et  al. 2013; Suddendorf and Corballis 2007; Wheeler, Stuss, and Tulving 1997). In support of this view, theorists note that patients with deficits in episodic memory typically show related deficits in episodic foresight (for a review, see Szpunar 2010). In addition, there are important similarities in the phenomenology of episodic memory and foresight (e.g., D’Argembeau and Van der Linden 2004). Harlene Haynes and colleagues summarize the prevailing view about episodic memory as holding that: (1) it involves a form of mental time travel in which the rememberer travels backward or forward in mental time to relive a past experience or to consider possible future scenarios, and (2) it is accompanied by conscious awareness that the event happened to “me” or will happen to “me” that does not accompany retrieval of other kinds of memories (Hayne et al. 2011, 344).



Reid already recognized this distinction, he just refused to call such factual knowledge “memory” (1785, 231).

1

The Episodic Sense of Self  139

1.1.  Patient Studies It will come as no surprise that people are good at reporting their own traits. When given a list of trait adjectives and asked to indicate the extent to which they possess those traits, people respond consistently over time, and they respond in ways that cohere with the judgments of their friends and relatives. What is surprising is that this trait self-knowledge turns out to be robust across catastrophic memory loss. Indeed, patients retain trait self-knowledge even when they have no episodic memory at all. Following a stroke, patient DB could not report a single episode from his life, but he had intact knowledge of his traits (Klein et al. 2004). Similarly, following a motorcycle accident, KC had a complete absence of episodic memory—when asked to remember something, anything, from his past, KC reported that it was just blank (Tulving 1985, 4). Nonetheless, KC had accurate knowledge of his traits (Tulving 1993). Indeed, across a wide range of neural dysfunction and injury, including Alzheimer’s, autism, and schizophrenia, patients show a remarkably robust knowledge of their traits (Altinyazar, Metz, and Klein forthcoming; Klein and Lax 2010). The fact that trait self-knowledge is preserved even in people with no episodic memory provides strong reason to think that trait self-knowledge is stored in semantic memory. What about the notion of self implicated in episodic memory? Endel Tulving, the most prominent contemporary advocate of the idea that the self is represented in episodic retrieval, offers the following suggestion for how to think about the self in episodic memory: We can think of self as the traveler who engages in mental time travel. Like other components of the system that is episodic memory, self too is defined in terms of its properties, and in terms of its relations to other components of the system. (Tulving 2005, 15)2

As cognitive scientists, we might want to define the self in terms of its properties and relations. But is that how the self is conceived in episodic memory and episodic foresight? The answer, I suggest, is no. Instead, episodic recollection presents the self in a way that floats free of traits and trait representations. Even when people have dramatic changes in their traits, this does not disrupt the sense of identity delivered by episodic recollection. Consider HM, who lost all ability to form new episodic memories after having his hippocampus surgically removed to alleviate his epileptic seizures. Following his surgery, HM could no longer remember anything for more than a couple of minutes. 2 Neisser (1988) also characterizes the self in episodic memory (what he calls the “extended self ”) as a composite: “that self can be thought as a kind of cumulated total of such memories: the things I remember having done and the things I think of myself as doing regularly” (47). And Kihlstrom (1997) offers the following proposal: “Within the network-theory framework, we can begin to see what the self might look like. Viewed as a declarative knowledge structure, the self can be defined as one’s mental representation of his or her own personality, broadly construed. . . . So, my own mental representation of self consists of interconnected nodes representing my name, the names of people intimately associated with me, and my physical, demographic, and personality characteristics” (453).

140  Shaun Nichols However, decades later, he would revel in recollecting a few episodes from his youth. One of HM’s biographers writes: “He tells of living in South Coventry, Connecticut, where he could go behind the house to shoot guns. He had a rifle, ‘One with a scope!’ he says, always enthusiastic at this point in his story. ‘And I had handguns too! A .38 and a. 32’ ” (Hilts 1995, 138). The difference between the child with the guns and the post-surgical adult are profound. After surgery, HM couldn’t take care of himself, didn’t know where he was living, didn’t know whether his parents were dead or alive. But that did not preclude him from identifying with the young man with a scope on his rifle. A somewhat fuller example comes from Floyd Skloot, a novelist who suffered significant brain damage from viral encephalitis. Skloot subsequently began writing— hauntingly—about his life as a brain-damaged person. One of the first things he stresses is how much brain damage changes a person: “Not just how we see or speak, how we feel or think, what we know or recall, but who we are is no longer the same. Without warning, without choice, we are Other” (2003, pp. ix–x). The lesions in his brain, he says, produce “lesions in my Self, fissures in the thought processes that result from this damage to my brain. When the brain changes, the mind changes— these lesions have altered who I am” (pp. 3–4). Despite the fact that he stresses how his self has changed, Skloot goes on in later essays to talk about episodes from his childhood. One particularly striking passage describes his childhood games with baseball cards: I teased Dodgers catcher Roy Campanella about the way he wore his hat, but secretly smashed and folded my own Dodgers hat until the crown resembled Campy’s. I asked Pirates pitcher Paul La Palme if his teammates ever called him Paula. . . . I liked to sniff the cards, which held a scent of bubblegum from their packaging . . . (pp. 169–70)

Skloot continues reminiscing about these experiences for several paragraphs, never giving anything other than the impression that the experiences present as experiences that he had, undiminished by the ravages of his brain lesions. Cases like HM and Skloot are particularly dramatic, but the basic point can be seen with a little phenomenological exercise. Think of some salient event from your youth, say, your first kiss. Now think of a salient event from a couple of years ago—a job offer, admission to grad school, a personnel action. Presumably your traits have changed substantially in the interval. But in your immediate recollection, does it feel any less you in the former case than in the latter? Although I’ve changed radically since my first kiss, when I remember it, it doesn’t seem like those trait changes disrupt my immediate identification with that adolescent standing outside the gate of my girlfriend’s yard. Furthermore, knowledge of trait continuity isn’t sufficient for this sense of identification, as indicated by a further brain damage case (Klein and Nichols 2012). At age 43, RB was in a bicycle accident that caused significant head trauma. After the injury, RB was able to recollect scenes from his past, but they did not have the sense of experiences that happened to him. In RB’s words, he lacked “ownership” for those experiences.

The Episodic Sense of Self  141 What I realized was that I did not “own” any memories that came before my injury. I knew things that came before my injury. In fact, it seemed that my memory was just fine for things that happened going back years in the past . . . But none of it was “me.” It was the same sort of knowledge I might have about how my parents met.

It’s clear that the memories had much in common with normal episodic memory— they carried the what, where, and when: I could clearly recall a scene of me at the beach in New London with my family as a child. But the feeling was that the scene was not my memory. As if I was looking at a photo of someone else’s vacation.

He described another example, this time from his University days: I can picture the scene perfectly clearly . . . studying with my friends in our study lounge. I can “relive” it in the sense of re-running the experience of being there. But it has the feeling of imagining, [as if] re-running an experience that my parents described from their college days. It did not feel like it was something that really had been a part of my life. Intellectually I suppose I never doubted that it was a part of my life. . . . But that in itself did not help change the feeling of ownership.

RB has the experience of episodes from his past, he just lacks the sense that they happened to him. The natural explanation for this is that somehow his recollection fails to bring with it the episodic sense of self. In addition, although RB lacks this sense of identity with the subject of past experiences, he retained his knowledge of his own traits throughout (Klein and Nichols 2012). All of this suggests that episodic memory carries with it a sense of self that is quite different from the trait conception stored in semantic memory. Radical changes in one’s traits do not uproot the sense of identification delivered by episodic memory. In addition, this sense of self can be disrupted even when trait self-knowledge is preserved, suggesting that there is a double dissociation between the trait sense of self and the episodic sense of self.

1.2.  Field and Observer RB isn’t the only patient to exhibit a sense of detachment from one’s experience memory. There are similar cases in the annals of neurology (e.g. Talland 1964; Stuss and Guzman 1988). Furthermore, people with post-traumatic stress disorder (PTSD) often report a related kind of detachment from the person who experienced the traumatic events (McIsaac and Eich 2004; Kenny and Bryant 2007). This detachment has been explained in terms of a perspectival distinction in episodic memory. A range of experiments show that experience memories can be recalled from the field (first person) or observer (third person) perspective (Nigro and Neisser 1983; Robinson and Swanson 1993; McIsaac and Eich 2002; McIsaac and Eich 2004). In one early experiment, participants were asked to recall events that fall into certain classes (e.g., studying,

142  Shaun Nichols swimming, being embarrassed, giving a public presentation), and then asked to indicate, for each memory, which of the following held for the memory: [Observer] In your memory, you imagine the scene as an observer might see it. Such an observer would see you as well as other aspects of the situation. [Field] In your memory, you imagine the scene from your original point of view, not as an external observer would see it. [Neither] Neither of the above perspectives fits your memory of this situation.

In this simple classification task, participants categorized slightly over half of the memories as field and slightly over 1/3 as observer (Nigro and Neisser 1983, 471). It’s widely assumed that episodic memory naturally presents experiences from the field (first person) perspective, but because of the reconstructive capacities of memory (see, e.g., Loftus and Palmer 1974), we can come to have a memory as-of from an observer perspective (e.g., McIsaac and Eich 2004, 248). The two different perspectives can also be experimentally manipulated. McIsaac and Eich had participants engage in activities and then later recollect the experiences. In one condition participants were told to see it again through their own eyes; in the other condition, participants were told to remember the experience as if they were viewing the scene as a detached observer (McIsaac and Eich 2002, 147). Subsequent coding of responses showed that participants in the observer condition were more likely to mention personal appearance. By contrast, those in the field condition were more likely to use first person pronouns (148). It’s plausible that retrieving a memory from the observer perspective will deflate emotional responses, and in fact, participants recollecting from the observer perspective are less likely to report affective reactions (McIsaac and Eich 2002, 148). Turning now to PTSD patients, research indicates that recalling a traumatic event from the observer perspective is experienced as less emotional (McIsaac and Eich 2004, 250). And the clinical literature is full of accounts of traumatized people reporting feeling dissociated from themselves. For instance, in a study of journalists who witnessed the execution of Robert Harris in 1992, 40 percent of the journalists assented to the statement “I experienced myself as though I were a stranger” and 53 percent assented to the statement “I felt distant from my own emotions” (Freinkel, Koopman, and Spiegel 1994, 1337; see also Cardena and Spiegel 1993).3 Such distancing—whether in the moment or in recollection—is interpreted as a strategy for avoiding unpleasant emotions (McIsaac and Eich 2004, 252; Kenny and Bryant 2007; Lemogne et al. 2009; Wilson and Ross 2003). This offers a broader context for understanding RB’s deficit. There is a particular kind of self-conception that is typically delivered in episodic memory. But

3 Note that these clinical studies do not carefully distinguish between whether (i) the original experience was encoded in a non-dissociative way but then retrieved as dissociative or (ii) the original experience was experienced as dissociative during encoding and is also experienced as dissociative at retrieval.

The Episodic Sense of Self  143 that kind of representation can be circumvented either through brain damage (RB) or strategic avoidance (in PTSD). To summarize, people have (at least) two conceptions of the self. There is a trait-based conception of self as composed of various character traits, values, emotional dispositions, and so forth. This conception of self is stored in semantic memory and preserved across catastrophic amnesia. But people also have a conception of self that is not defined in terms of traits. Such a conception is implicated in episodic memory. It is robust across the knowledge of massive trait changes (HM, Skloot), and it can be disrupted even when the trait sense of self is largely preserved (RB, PTSD).

1.3. Representations Trait conceptions of self are relatively well understood. From a psychological perspective, we can characterize the traits themselves, and the personality psychology tradition indicates that we get considerable predictive power from the trait attributions (for a review, see Funder 2001).4 Ordinary people have some knowledge of their own traits and often think of themselves in terms of their traits. In addition, we have good models of how this trait self-conception is represented—it is stored in semantic memory as a pre-computed summary, poised for quick retrieval (e.g., Klein et al. 2002). We also know that changing how people think about the trait-self has significant impacts on decision-making. In a series of studies, Dan Bartels and colleagues have shown that getting people to think that their trait-self is highly mutable affects behavior on temporal discounting tasks. When people think that the trait-self is constantly changing, they are more impatient—they prefer to take a lower amount of money soon rather than wait for a greater amount of money later (Bartels and Urminsky 2011; Bartels and Rips 2010; Bartels, Kvaran, and Nichols 2013). The non-trait sense of self we find in episodic memory is poorly understood. Psychologists have had next to nothing to say about the representation of this conception.5 Although psychologists have had little to say on the episodic sense of self, the nature of the representation might be explicated by drawing on philosophical discussions concerning indexical expressions like “I,” “now,” and “here.” First, there is an important role in action for the cognitive correlates of these linguistic expressions. To adapt an example from John Perry, if Fred thinks Fred is about to be hit by a boulder, that won’t lead Fred to dodge unless he also thinks I am Fred. Or, more simply, Fred won’t dodge unless he thinks I am about to be hit by a boulder (Perry 1977, 494; Perry 1979). Second, theoretically, at least, it seems that one could wake up with total amnesia 4 The predictive value of the attributions is significant, but it might be less than people normally think (Doris 2002). 5 There is discussion of the representation of self in episodic memory. Perhaps the most influential model comes from Martin Conway. But the representation of self is not the thin self-representation identified by Kant. Rather, once again, the representation is rich. On Conway’s view, the “conceptual self ” comprises “complex episodic memories” which themselves comprise “simple episodic memories,” and the conceptual self includes self-image, life stories, and beliefs/values/attitudes (2009, 2310).

144  Shaun Nichols and token an “I” representation in appropriate ways, for instance, by thinking “where am I?” (cf. Perry 1977). An associated description in terms of traits is neither necessary (as illustrated by the amnesia case) nor sufficient (as illustrated by the boulder case) for tokening the representation. Let’s adopt the notation I* for this mental representation.6 It’s a small step to maintain that this same representation—I*—is also the representation implicated in episodic memory and foresight. In the Perry-style examples, the I* cannot be identified with traits; similarly, trait continuity is neither necessary (HM, Skloot) nor sufficient (RB, PTSD) for tokening the episodic sense of self.7 There is an asymmetry here between I* and the concepts corresponding to the closely related indexicals “here,” and “now.” As with I*, the mental representations corresponding to “here” and “now” cannot be replaced with descriptions and retain their cognitive roles (e.g., Perry 1977; Perry 1979). For instance, believing that the meeting starts now has a different effect than believing that the meeting starts at noon; believing that the meeting will be here has a different effect than believing that the meeting will be in the library. This parallels the fact that believing that Shaun is about to be attacked by a cougar has a different effect than believing that I* am about to be attacked. Despite these similarities, in episodic recall, there is an interesting dissimilarity. In episodic retrieval, instead of now, we get a specific when, and instead of here, a specific where. By contrast, I* doesn’t get replaced with descriptive content. This feature of the I* in episodic memory enables the sense of identity that accompanies typical episodic memory. When I remember my first kiss, this comes with a sense of identity with that distant and qualitatively different individual. Let’s call this the episodic sense of identity.

2.  The Animal and the I* Thus far, I’ve argued that there is a special sense of self implicated in episodic memory which is not tied to traits and which plausibly implicates a special kind of mental representation—I*. When I recall an experience from the past, this typically yields an episodic sense of identity with that past individual, even if there have been extensive trait changes. This account of the episodic sense of identity overlaps in important ways with animalist views of personal identity. According to animalism, the self is identical to the organism, and it’s likely that the I*, i.e., the non-trait conception of self, (as a matter of fact) roughly corresponds with the organism that tokens the representation. When episodic recollection returns a representation of the form I* had experience P,

6 This notation is adopted as a nod to Castañeda’s important work in the area, though his use of the notation is somewhat different (1967). 7 I leave open whether I* is actually part of the basic experience of having an episodic memory or whether the I* is a representation that characteristically gets joined with the experience. There are interesting issues here for cognitive science. For instance, if other mammals have the phenomenology associated with episodic recollection, does that mean that they have the I* representation? These questions haven’t been explored in the memory literature, even among those (e.g. James and Tulving) who promote the importance of the self in episodic memory. But in any case, it isn’t necessary to settle these issues for present purposes.

The Episodic Sense of Self  145 typically, it will be true that this organism had experience P.8 It might seem that there is a quick argument from here to a substantive conclusion about the self. Contrary to Kant, Buddha, and Parfit, the I that was thought to be so problematic is just the organism. This would count as a vindication of an animalist account of personal identity (Olson 1997; Shoemaker forthcoming). The animalist doesn’t win so easily. For the episodic sense of identity is importantly different from the biological conception of identity. First, knowledge of animal identity does not carry the episodic sense of identity. We see this most dramatically in the case of RB, who has full knowledge that the organism RB was on that beach in New London. Yet he lacks the sense that he was there. The dissociation occurs in the other direction as well. People sometimes have an episodic sense of identity that is flatly at odds with their knowledge of animal identity. The most striking cases are memories of past lives. We tend to think of such cases as quite bizarre, but they actually fit naturally into the broader context of what we know about the reconstructive nature of episodic memory. A long tradition of work in memory shows that people can easily be led to have false episodic memories (see, e.g., Schacter 2001). In some cases, the memories are false in the details—subjects are led to misremember certain aspects of an experience, e.g., whether a car crash captured on video involved broken glass. In other cases, people can be led to remember having done something that, in fact, they never did at all. Getting people to imagine vividly having done something often leads them to remember actually having done it (Goff and Roediger 1988; Thomas and Loftus 2002; Rajagopal and Montgomery 2011).9 As it happens, ordinary people can often be led to believe in past lives through hypnotic suggestion. In a series of experiments, Spanos hypnotized subjects (all of whom were undergraduate students with moderate to high levels of suggestibility) and played a pre-recorded tape that indicated that the subject was going back in time, concluding with: “You are now in a different life, living in another life that you have lived before in another time. You are now reliving that other life that you lived once before in a different time” (Spanos et al. 1991, 310). After this suggestion, subjects were asked questions like “What name can I call you by?” and “Where are you?” Remarkably, over 30 percent of the subjects responded in ways that indicated that they thought they were having an experience of a past life. This group of participants did not differ on measures of psychopathology from the remaining 70 percent who did not think they were having experiences of a past life (Spanos et al. 1991, 311–12). Subsequent studies showed that the details of the subjects’ reported past lives was heavily influenced by the presentation offered by the experimenters. For instance, subjects who were told that there 8 Of course, as I shall discuss, there are lots of reconstructive processes in memory. It might be that often my episodic recollections really bear no relation to the experiences of this organism. But the evidence certainly doesn’t undermine the idea that in the typical case, an episodic memory will correspond to a past experience of the organism. 9 Some maintain that the plasticity of memory is an adaptation that facilitates modifying memories for relevance (e.g., Lee 2009).

146  Shaun Nichols was a high incidence of child abuse in earlier times were more likely to recall a past life in which they suffered child abuse (Spanos et al. 1991, 315). Indeed, these results on memories of past lives are interpreted as further evidence for the reconstructive nature of memory (e.g. Loftus and Polage 1999). Of course, outside of the laboratory, many people report experiences of past lives. In recent work, researchers recruited people who claimed to have memories of a past life. They gave these subjects a standard task for inducing false memories (Roediger and McDermott 1995) and compared the responses of these subjects with a control group of subjects who did not think they had memories of a past life. Those who had memories of past lives reported more false memories on this standard memory task (Meyersburg et al. 2009, 402; see also Peters et al. 2007). This suggests that part of the reason these people have the memory of a past life is that their memory formation processes are especially susceptible to reconstructive influences. The point of going through the role of memory reconstruction in past life memories was just to shore up the idea that memories of past lives actually fit into a well-established research tradition on the construction of false memories. But the important lesson for us is simply that the sense of identity that is operative in these false memories is operating in the face of explicit knowledge of biological discontinuity. So, just as knowledge of biological continuity isn’t sufficient to issue in the episodic sense of identity, neither, it seems, is knowledge of biological discontinuity sufficient to destroy the episodic sense of identity. An episodic memory as-of before one’s biological life can still carry the episodic sense of self and the concomitant impression of identity across organisms.

3.  Moral Emotions and the Episodic Sense of Identity Thus far, I’ve argued that there are two kinds of representations of self. One of them is given in terms of traits; the other representation of self is the impoverished I*, and this representation is typically delivered in episodic recollection. It is this latter representation that yields the sense of identity across decades and massive trait transformations. Let’s turn now to the representation of self in moral emotions. Several central moral emotions, including guilt, shame, and pride are also identified as “self-conscious emotions.” Contemporary emotion theorists maintain that a representation of the self is implicated in the activation of such “self-conscious” emotions. In Lazarus’ influential appraisal framework, he maintains that a critical component of the appraisal of pride is that “credit is to oneself ”; for guilt and shame, the appraisal must include that “blame is to oneself ” (1991, 271, 242–3; see also Lewis et al. 1989). Tracy and Robins put the point more generally: The primary distinctive characteristic of self-conscious emotions is that their elicitation requires the ability to form stable self-representations (“me”), to focus attention on those representations

The Episodic Sense of Self  147 (i.e., to self-reflect; “I”), and to put it all together to generate a self-evaluation. (Tracy and Robins 2007, 191)

Even some emotions that are not categorized as “self-conscious,” like anger, might critically involve self-representations. Recalling that the offense was demeaning to me can be important to activating anger. Indeed, Lazarus’ appraisal theory has ego-involvement as a dimension for appraisal for all emotions (1991).10 Although emotion researchers maintain that self-representations are required to trigger guilt, pride, and shame, they have not explored which self-representations are required. Is it the trait sense of self or the thin conception of self? For at least some cases, it seems clear that the thin conception of self suffices to trigger a moral emotion. To take an unflattering personal example, I remember hitting my babysitter over the head with a large door spring. It’s nearly forty years later, but whenever I recall that incident, I feel guilty. And I am positive that the first thing I would do, if I saw her now, is apologize for the head bashing. Obviously my traits are significantly different from those of the 9-year-old (for one thing, I am pretty sure I wouldn’t hit anyone over the head with a door spring). However, this does not obviate the immediate guilt reaction that ensues on the recollection. In this case, the I* suffices to trigger the moral emotion. This case suggests that guilt can be activated with I*, but it would be rash to draw global conclusions. One interesting possibility is that in different cases, different kinds of self-representations are implicated in self-conscious emotions. For instance, when I remember the public display in my 7th grade gym class of my running time, I remember that I was ashamed of how slow I was. Although I remember feeling shame then, I certainly don’t feel ashamed now. The fact that I have the episodic sense of identity with that chubby 7th grader doesn’t make me feel shame now. In that case, the episodic sense of identity doesn’t activate the self-conscious emotion. One obvious difference between the guilt case and the shame case is that the former involves an act and the latter a trait, and perhaps this difference in content explains the difference in how the episodic sense of identity interacts with the emotions. There is a nest of interesting issues here, but for now the important point is just that in at least some cases, moral emotions are activated by the episodic sense of self.

4.  Philosophy and the Episodic Sense of Identity The issue of personal identity lies at the intersection of metaphysics and ethics. A familiar assumption is that if we can determine the correct metaphysical account of personal identity, we will be able to draw inferences about when it is appropriate to hold someone responsible. In the simple case of a transgression, it’s appropriate for me to

10 It’s intuitively plausible that a representation of self is critical to self-conscious emotions, but it should be acknowledged that there is not much research narrowly directed at this issue (cf. Tracy and Robins 2004a, 111).

148  Shaun Nichols feel guilt only if I am the person who committed the transgression. Obviously there is a wide range of complications and a wide range of issues implicated at this intersection between metaphysics and ethics. In this final section, I will explore how the nature of the episodic sense of identity affects some of these issues.

4.1.  The Episodic Sense of Identity and the Metaphysics of Identity Philosophers have railed against the viability of non-trait conceptions of self. As mentioned above, Kant says that the “I” is “the poorest of all representations” (B408.7). On Kant’s view, the poverty of the I-representation is responsible for a range of illusions concerning the self. Here’s Nagel, apparently channeling Kant, on the issue: The concept of the self seems suspiciously pure—too pure—when we look at it from inside. . . . When I consider my own individual life from inside, it seems that my existence in the future or the past—the existence of the same “I” as this one—depends on nothing but itself. . . . (Nagel 1986, 33; see also Locke 1731, Book II, chapter xxvii, section 14; Kant 1968 B407– B432; Nagel 1979, 200)

This “too pure” conception of self is plausibly precisely the conception of self implicated in the episodic sense of identity. And as a foundation for a metaphysics of identity, the episodic sense of identity seems problematic. The most obvious problem with taking the episodic sense of identity to reveal the correct metaphysics of personal identity is fission (e.g. Parfit 1984). Identity is a 1–1 relation, and the episodic sense of identity doesn’t respect that. Imagine each hemisphere of the brain being put into a different body. If this happens then, assuming that each hemisphere contains the episodic memory of bashing Leanne over the head, each fission fragment will have the episodic sense of identity with the 9-year-old.11 The fragments can’t both be identical to me (they are different individuals who might be having different current experiences), yet each of them would have a robust episodic sense of identity with a single past person.12 This episodic sense of identity is plausibly encapsulated—even if you knew that you were one fission fragment, this would not (I suggest) destroy the sense of identity delivered by episodic recollection of a pre-fission event. The problem of fission is the starkest way of seeing the limitations of the episodic sense of identity as a foundation for metaphysics of identity. But there is a further metaphysical oddity in the episodic sense of identity. The sense of identity is insensitive to temporal discontinuity. For physical objects, disruptions in spatio-temporal continuity engender disruptions in the impression of identity (see, e.g., Cheries et al. 2009). But as far as my episodic sense of identity goes, it seems that I’m identical with a door-spring-wielding 9-year-old, and there is nothing in this sense of identity that 11 Fission is a thought experiment, of course. But in terms of what we know about memory and the sense of identity, the assumptions about our reactions to fission (e.g., that each descendant will have the sense of identity) are very plausible. 12 In philosophy, one move to try to accommodate fission is to take a four-dimensional approach to the issue (e.g., Lewis 1976; Sider 2003). I will leave that matter aside here.

The Episodic Sense of Self  149 demands uninterrupted continuity of self between then and now. Again, this seems encapsulated. Even if I came to believe that I didn’t exist during the interval, episodic recollection would still generate the sense of identity. The representation delivered by episodic memory doesn’t care about uninterrupted continuity. Of course, there is a familiar philosophical move to try to secure continuity—one can maintain that episodic memory reveals specific experiences of an underlying substance, and that substance endures without interruption. But there is nothing in the episodic sense of identity that presupposes continuity. At a minimum, episodic memory and the associated conception of the self make for a very imperfect guide to metaphysics. The system is not designed to get at metaphysical truth about personal identity.

4.2.  The Episodic Sense of Identity and Moral Emotions Even though the episodic sense of identity is not a good basis for a metaphysics of personal identity, we have seen that self-conscious moral emotions are sometimes activated by the episodic sense of identity, even after massive trait changes. How are we to evaluate the propriety of such activation? The verdict here will depend on background assumptions about the normative propriety of various emotions. D’Arms and Jacobson provide a useful framework for exploring this issue. Like many contemporary moral psychologists, they defend a sentimentalist approach to metaethics. But since they defend a “rational” sentimentalism, their view might gain wider acceptance than more thoroughgoing anti-rationalist approaches to ethical justification (e.g. Nichols 2013; Prinz 2007). D’Arms and Jacobson argue that emotions or sentiments are right to feel (under certain circumstances), if they involve deep and wide human concerns. “Deep concerns,” they write, “are those that are firmly entrenched in their possessors, such that it would be either impossible or extremely costly to excise them.” According to D’Arms and Jacobson, a concern is wide, when it is “firmly enmeshed in our web of psychological responses,” that is, when “the object of a concern prompts a variety of evaluative attitudes, not just a single emotion or desire; when desire for it (or aversion to it) arises in many different situations; when it is implicated in the ability to get or avoid many other things people care about; when its pursuit or avoidance grounds disparate actions and plans” (2005, 116). With this general approach, the broad issue about the propriety of self-conscious moral emotions depends on whether those emotions involve concerns that are deep and wide.13 To determine whether the emotions involve deep and wide concerns, it’s helpful to consider the likely function of these self-conscious emotions. In a somewhat ironic development, contemporary researchers think that the functions of self-conscious

13 There is one important way in which my appropriation of D’Arms and Jacobson might diverge from their own proposal. I do not mean to rely on the idea that the deep and wide concerns are constitutive of the specific emotions. Rather, what’s important for the following discussion is the idea that the self-conscious emotions are systems that are dedicated to advancing concerns that are deep and wide.

150  Shaun Nichols emotions turn out to be largely social: “self-conscious emotions evolved primarily to promote the attainment of specifically social goals” (Tracy and Robins 2007, 6). Shame serves to motivate behavior to avoid social disapproval (e.g. Fessler and Haley 2003). The function of guilt is to repair and sustain relationships after doing harm to a loved one (Baumeister, Stillwell, and Heatherton 1994). The function of pride is to maintain or enhance status in light of an achievement (Tracy and Robins 2004b). The social functions of self-conscious emotions provide ample reason to think that these emotions do indeed promote deep and wide concerns (see also D’Arms and Jacobson 2005, 116ff). Each of the self-conscious emotions plays an important role in promoting a successful social life. Guilt encourages relationship reparation with intimates; pride serves to sustain or enhance social status; shame helps to decrease social disapproval. These reactions are all relatively deep in that it would be hard to eradicate the emotions entirely. In addition, each of these individual social concerns—intimate relationships, social status, community approval—is a wide concern. Even when we’re not feeling guilty we care about sustaining our close relationships. Even when we’re not feeling proud, we care about our status. And even when we’re not ashamed, we care a great deal about how our community thinks of us. More generally, all of these emotions address social life, and it takes no philosophical invention to note that a flourishing social life is among the deepest and widest human concerns. Thus, the self-conscious emotions involve concerns both deep and wide. Now we turn to the issue of moral emotions triggered by distant episodic memories. Return to my memory of beating my babysitter. The episodic recollection of that event triggers guilt. And it’s plausible that this actually serves the deep and wide concerns of guilt. If I see Leanne now, and I don’t show guilt for the head bashing, then this will be bad for my current social interaction with her (at least on the plausible assumption that she remembers being cracked over the head by an ill-behaved 9-year-old). If I didn’t feel like I did the action, I wouldn’t feel guilty for it, and then I wouldn’t display the socially functional behaviors. The behaviors are functional because other people naturally track organisms, not feature clusters. The agent needs to negotiate the social world and be moved by the shameful, pride-worthy, and guilt-befitting events of its past as a persisting organism. As a result, the fact that the episodic sense of identity is not tied to psychological traits plausibly facilitates the social function. It is socially beneficial that episodic memory delivers a conception of self that will continue to activate the self-conscious emotions even after radical trait changes. Again, this is because even though my traits change, my community tracks what this organism did.14 One way to think about this is robustly pluralistic. At the end of section 3, I suggested that while guilt is triggered by episodic memories of distant transgressions, shame is

14 The organism is a much better proxy than feature sets for the self that is tracked by others (at least when it comes to actions). This might be a heuristic—it might be that some kind of unspecified Cartesian soul would be a better fit, if only because of its complete flexibility. But as heuristic, organism is a better proxy than feature cluster.

The Episodic Sense of Self  151 not triggered by episodic memories of distant shameful traits. Thus, one might maintain that for purposes of morally evaluating a person’s actions, it’s simply correct to think about the self in a way that is not trait dependent, whereas for the purposes of morally evaluating a person’s traits, it’s correct to think about the self in a way that is trait dependent. Alternatively, one might think, with Kant, that the trait-independent sense of self is deeply defective. Indeed, perhaps the dominant view in philosophy is that traits are what matter for ethical questions like responsibility—if I commit an immoral act in my youth and undergo massive trait changes, then my culpability for the act is thereby diminished. Recent studies show that ordinary people sometimes react in a similar way. When led to believe that self-traits change a great deal over time, participants think that they deserve less punishment for a past misdeed (Tierney et al. 2014). But we have also seen that guilt is activated by the episodic sense of self despite massive trait changes. Thus it seems that the fundamental tension between the two senses of self re-emerges in the ethical domain. The trait sense of self seems to be central to issues like responsibility, but the episodic sense of self generates guilt even when the trait-self (and the representation thereof) has changed radically.15 There might be rational reasons for thinking that guilt about distant transgressions isn’t warranted given that my traits have changed so much. But there are prudential reasons for thinking it’s a good thing that guilt is triggered by the episodic sense of self. Let’s take it as a standing fact that others do and will continue to track each other as organisms. People will continue to attribute social transgressions, moral transgressions, and accomplishments to organisms, not feature sets. At any particular time slice of a person, among the deep and wide concerns of that person are his social standing, the quality of his relationships, and so forth. And, as we’ve seen, there is reason to think that the self-conscious emotions are mechanisms that function to promote those concerns. Importantly, they promote those concerns even if they implicate a metaphysically suspicious representation of self. Take my guilt about hitting Leanne. Even if that is driven by a peculiar feature of episodic sense of identity, the reaction is facilitating things I care about right now. By presenting the self as persisting despite massive trait changes, episodic memory facilitates the social function of the self-conscious emotions. Even if I am not the same person as the babysitter-beater, my community will treat me as such. And at any given point, I care a lot about how that community treats me. The fact that I feel guilty motivates me to make amends and to acknowledge the transgression, revealing both my awareness that it was a transgression as well as 15 The option set here can be put in terms of revisionism. A fully non-revisionist proposal would hold that there are two different kinds of selves and different ethical concerns key on these different kinds. For instance, one might hold that there is a trait-self and a Cartesian soul, not defined by traits. A radically revisionist view would maintain that the non-trait sense of self is hopeless, and all reactions that are insensitive to continuity of psychological traits are defective. A moderate revisionism might hold that the biological conception of self is a sufficiently close approximation to the non-trait sense of self; this would facilitate a revisionist pluralism on which the trait sense of self is the relevant notion of self for some ethical concerns (like shame concerning traits), and the biological conception is the relevant notion of self for other ethical concerns (like guilt for distant transgressions).

152  Shaun Nichols a certain amount of contrition; showing such awareness and contrition can help improve my relationship. So, while it might be rational to think that the self changes as traits change, it can be prudent to feel guilt even after the trait-self changes.

Acknowledgments I’m grateful to John Doris and Dave Shoemaker for comments on an earlier draft of this chapter. Special thanks to Felipe De Brigard, Justin D’Arms, and Dan Jacobson, for extensive feedback on the chapter.

References Altinyazar, V., Metz, M., and Klein, S. Forthcoming. Systems of Self-Knowledge and Schizophrenia:  Evidence for the Unusual Robustness of Trait Self-Knowledge in Patients Suffering from Schizophrenia. Bartels, D., Kvaran, T., and Nichols, S. 2013. Selfless giving. Cognition 129, 392–403. Bartels, D., and Rips, L. 2010. Psychological connectedness and intertemporal choice. Journal of Experimental Psychology: General, 139, 49–69. Bartels, D., and Urminsky, O. 2011. On intertemporal selfishness: The perceived instability of identity underlies impatient consumption. Journal of Consumer Research, 38, 182–98. Baumeister, R. F., Stillwell, A., and Heatherton, T. F. 1994. Guilt: An interpersonal approach. Psychological Bulletin, 115, 243–67. Burke, B.  L., Martens, A., and Faucher, E.  H. 2010. Two decades of terror management theory:  A  meta-analysis of mortality salience research. Personality & Social Psychology Review, 14, 155–95. Butler, Joseph 1736/1819. Of personal identity. In The Analogy of Religion, pp. 211–15. London: Allman and Sawers. Cardena, E., and Spiegel, D. 1993. Dissociative reaction to the San Francisco Bay Area earthquake of 1989. American Journal of Psychiatry, 150, 474–8. Castañeda, H. N. 1967. Indicators and quasi-indicators. American Philosophical Quarterly, 4(2), 85–100. Cheries, E. R., Mitroff, S. R., Wynn, K., and Scholl, B. J. 2009. Do the same principles constrain persisting object representations in infant cognition and adult perception? The cases of continuity and cohesion. In B. Hood and L. Santos (eds.), The Origins of Object Knowledge. Oxford: Oxford University Press. D’Argembeau, A., and Van der Linden, M. 2004. Phenomenal characteristics associated with projecting oneself back into the past and forward into the future: Influence of valence and temporal distance. Consciousness and Cognition, 13, 844–58. D’Arms, J., and Jacobson, D., 2005. Anthropocentric constraints on human value. Oxford Studies in Metaethics, 1, 99–126. De Brigard, F., Addis, D., Ford, J., Schacter, D., and Giovanello, K. 2013. Remembering what could have happened:  Neural correlates of episodic counterfactual thinking. Neuropsychologica. doi: 10.1016 Doris, J. 2002. Lack of Character. Cambridge: Cambridge University Press.

The Episodic Sense of Self  153 Fessler, D., and Haley K. 2003. The strategy of affect: Emotions in human cooperation. In P. Hammerstein (ed.), Genetic and Cultural Evolution of Cooperation, pp. 7–36. Cambridge, MA: MIT Press. Freinkel, A., Koopman, C., and Spiegel, D. 1994. Dissociative symptoms in media eyewitnesses of an execution. American Journal of Psychiatry, 157, 1335–9. Funder, D. C. 2001. Personality. Annual Review of Psychology, 52, 197–221. Goff, L., and Roediger, H., III. 1998. Imagination inflation for action events: Repeated imaginings lead to illusory recollections. Memory & Cognition, 26, 20–33. Hayne, H., Gross, J., McNamee, S., Fitzgibbon, O., and Tustin, K. 2011. Episodic memory and episodic foresight in 3- and 5-year-old children. Cognitive Development, 26(4), 343–55. Hilts, P. 1995. Memory’s Ghost. New York: Simon & Schuster. James, W. 1890/1950. Principles of Psychology, vol. 1. Mineola, NY: Dover. Kant, I. 1968. Kant’s Critique of Pure Reason, trans. N. Kemp Smith. New York: St Martin’s. Kihlstrom, J.  F. 1997. Consciousness and me-ness. In J. D.  Cohen and J. W.  Schooler (eds.), Scientific Approaches to Consciousness, pp. 451–68. Mahwah, NJ: Erlbaum. Kenny, L., and Bryant, R. 2007. Keeping memories at an arm’s length: Vantage point of trauma memories. Behaviour Research and Therapy, 45, 1915–20. Klein, S., Cosmides, L., Tooby, J., and Chance, S. 2002. Decisions and the evolution of memory: Multiple systems, multiple functions. Psychological Review, 109, 306–29. Klein, S., German, T., Cosmides, L., Rami, G. 2004. A theory of autobiographical memory:  Necessary components and disorders resulting from their loss. Social Cognition, 22, 460–90. Klein, S., and Lax, M. 2010. The unanticipated resilience of trait self-knowledge in the face of neural damage. Memory, 18, 918–48. Klein, S., and Nichols, S. 2012. Memory and the Sense of Personal Identity. Mind, 121(483), 677–702. Lazarus, R. S. 1991. Emotion and Adaptation. New York: Oxford University Press. Lee, J. L. 2009. Reconsolidation: Maintaining memory relevance. Trends in Neurosciences, 32, 413–420. Lemogne, C., Bergouignan, L., Piolino, P., Jouvent, R., Allilaire, J., and Fossati, P. 2009. Cognitive avoidance of intrusive memories and autobiographical memory: Specificity, autonoetic consciousness, and self-perspective. Memory, 17, 1–7. Lewis, D. 1976. Survival and identity. In A. Rorty (ed.), Identities of Persons, pp. 17–40. Berkeley and Los Angeles: University of California Press. Lewis, M., Sullivan, M. W., Stanger, C., and Weiss, M. 1989. Self development and self-conscious emotions. Child Development, 60, 146–56. Locke, J. 1731: An Essay Concerning Human Understanding. London: Edmund Parker. Original work published 1690. Loftus, E. F., and Palmer, J. C. 1974. Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior, 13(5), 585–9. Loftus, E. F., and Polage, D. C. 1999. Repressed memories: When are they real? How are they false? The Psychiatric Clinics of North America, 22, 61–71. (P. Resnick, Guest ed.). McIsaac, H. K., and Eich, E. 2002. Vantage point in episodic memory. Psychonomic Bulletin & Review, 9, 146–50.

154  Shaun Nichols McIsaac, H. K., and Eich, E. 2004. Vantage point in traumatic memory. Psychological Science, 15, 248–53. Meyersburg, C., Bogdan, R., Gallo, D., and McNally, R. 2009. False memory propensity in people reporting recovered memories of past lives. Journal of Abnormal Psychology, 118, 399–404. Mitchell, J. 2009. Social psychology as a natural kind. Trends in Cognitive Sciences, 13(6), 246–51. Nagel, T. 1979. Subjective and objective. In Mortal Questions. Cambridge: Cambridge University Press. Nagel, T. 1986. The View from Nowhere. Oxford: Oxford University Press. Neisser, Ulric. 1988. Five kinds of self-knowledge. Philosophical Psychology, 1, 35–59. Nichols, S. 2013. Brute retributivism. In T. Nadelhoffer (ed.), The Future of Punishment. New York: Oxford University Press. Nigro, G., and Neisser, U. 1983. Point of view in personal memories. Cognitive Psychology, 15, 467–82. Olson, E. 1997. The Human Animal:  Personal Identity without Psychology. Oxford:  Oxford University Press. Parfit, D. 1984. Reasons and Persons. New York: Oxford University Press. Perry, J. 1977. Frege on demonstratives. Philosophical Review, 86, 474–97. Perry, J. 1979. The problem of the essential indexical. Nous, 13, 3–21. Peters, M., Horselenberg, R., Jelicic, M., and Merckelbach, H. 2007. The false fame illusion in people with memories about a previous life. Consciousness and Cognition, 16, 162–9. Prinz, J. 2007. The Emotional Construction of Morals. Oxford: Oxford University Press. Rajagopal, P., and Montgomery, N. 2011. I imagine, I experience, I like: The false experience effect. Journal of Consumer Research, 38(3), 578–93. Reid, T. 1785. Essays on the Intellectual Powers of Man. Edinburgh: Bell & Robinson. Robinson, J., and Swanson, K. 1993. Field and observer modes of remembering.  Memory, 1, 169–84. Roediger, H., and McDermott, K. 1995. Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, & Cognition, 21, 803–14. Schacter, D. 2001. The Seven Sins of Memory. Boston: Houghton Mifflin. Shoemaker, D. Forthcoming. The stony metaphysical heart of animalism. In S. Blatti and P. Snowdon (eds.), Animalism. Oxford: Oxford University Press. Sider, T. 2003. Four-Dimensionalism. Oxford: Oxford University Press. Skloot, F. 2003. In the Shadow of Memory. Lincoln, NE: University of Nebraska Press. Spanos, N., Menary, E., Gabora, N., DuBreuil, S., Dewhirst, B. 1991. Secondary identity enactments during hypnotic past-life regression:  A  sociocognitive perspective. Journal of Personality and Social Psychology, 61, 308–20. Stuss, D. T., and Guzman, D. A. 1988. Severe remote memory loss with minimal anterograde amnesia: A clinical note. Brain and Cognition, 8, 21–30. Suddendorf, T., and Corballis, M. 2007. The evolution of foresight: What is mental time travel, and is it unique to humans? Behavioral and Brain Sciences, 30, 299–313. Szpunar, K. 2010. Episodic future thought: An emerging concept. Perspectives in Psychological Science, 5, 142–62. Talland, G.  A. 1964. Self-reference:  A  neglected component in remembering. American Psychologist, 19, 351–3.

The Episodic Sense of Self  155 Thomas, A. K., and Loftus, E. F. 2002. Creating bizarre false memories through imagination. Memory & Cognition, 30, 423–31. Tierney, H., Howard, C., Kumar, V., Kvaran, T. and Nichols, S. 2014. How many of us are there? In J. Sytsma (ed.), Advances in Experimental Philosophy of Mind. New York: Continuum Press. Tracy, J. L., and Robins, R. W. 2004a. Putting the self into self-conscious emotions: A theoretical model. Psychological Inquiry, 15, 103–25. Tracy, J. L., and Robins, R. W. 2004b. Show your pride: Evidence for a discrete emotion expression. Psychological Science, 15, 194–7. Tracy, J. L., and Robins, R. W. 2007. The self in self-conscious emotions: A cognitive appraisal approach. In J. Tracy, R. Robins, and J. Tangney (eds.), The Self-Conscious Emotions: Theory and Research, pp. 3–20. New York: Guilford. Tulving, E. 1985. Memory and consciousness. Canadian Psychology/Psychologie Canadienne, 26, 1–12. Tulving, E. 1993. Self-knowledge of an amnesic individual is represented abstractly. In T. K. Srull and R. S. Wyer (eds.), Advances in Social Cognition, vol. 5, pp. 1–49. Hillsdale, NJ: Erlbaum. Tulving, E. 2005. Episodic memory and autonoesis. In H. Terrace and J. Metcalfe (eds.), The Missing Link in Cognition: Origins of Self-reflective Consciousness, pp. 3–56. New York: Oxford University Press. Wheeler, M., Stuss, D., and Tulving, E. 1997. Toward a theory of episodic memory: The frontal lobes and autonoetic consciousness. Psychological Bulletin, 121, 331–54. Wilson, A., and Ross, M. 2003. The identity function of autobiographical memory: Time is on our side. Memory, 11, 137–49.

8 The Motivational Theory of Emotions Andrea Scarantino

1. Introduction Having an emotion customarily involves appraising a stimulus a particular way, feeling a particular way, and being motivated to act a particular way. These three aspects offer distinctive entry points for explaining what emotions are. Philosophers have so far built their theories primarily around appraisals and feelings. As a result, two research programs have dominated the philosophy of emotions over the past forty years: Cognitivism and Perceptualism. I will argue that neither research program has been able to explain how emotions motivate us to act. This is an especially significant omission because it is ultimately what we do when we emote that produces significant personal and social consequences. To remedy this state of affairs, I begin by illustrating three features of emotional motivation that neither Cognitivism nor Perceptualism have been able to provide an integrated account for: impulsivity, flexibility, and bodily underpinnings. I then introduce a new theory of emotions that offers a cohesive explanation for these three features of emotional motivation. A preliminary statement of what I call the Motivational Theory of Emotions (MTE) is that emotions are action control systems designed to prioritize the pursuit of some goals over others. As modern day Cognitivism was inspired by the appraisal tradition in psychology (e.g. Arnold 1960; 1970) and modern day Perceptualism by the Jamesian theory of emotions (James 1884; Damasio 1994), the theory of emotions I propose here is inspired by Frijda’s (1986; 2007) psychological theory of emotions as states of action readiness with control precedence. MTE will be shown to not only explain how emotions motivate better than Cognitivism and Perceptualism, but to also solve two standard problems in the philosophy of emotions:  the problem of intentionality (how are emotions “about” the world?) and the problem of differentiation (how are any two emotions different from one another?). Since MTE can solve these familiar problems, and does significantly better on motivation, I conclude that it should be considered a promising new competitor in the philosophy of emotions.

The Motivational Theory of Emotions  157

2.  Three Markers of Emotional Actions It is common practice in science and folk psychology to explain various phenomena by appealing to emotions. Matt’s heartbeat increased because he was excited. Susan displayed a gape face because she was disgusted to have found an insect in her soup. Roger recoiled because he was afraid of the suddenly looming car. Cassandra slapped her lover because she was angry after having read a flirtatious email he sent to another woman. Stan wrote a pointed letter of complaint about his boss because he was angry about having been insulted at a board meeting the day before. These phenomena include involuntary bodily changes (increased heartbeats, gape expressions), reflex actions (recoiling at a suddenly looming object), intentional actions done “in the heat of the moment” (slapping someone), and intentional actions done while no longer engrossed in the emotional incident (writing a letter of complaint one day after having been insulted). Although all of these phenomena must be explained, my initial focus will be on what I take to be the most prototypical of emotional actions, namely intentional actions done “in the heat of the moment” such as slapping someone out of anger. I will call these impulsive emotional actions. Once I formulate an account of impulsive emotional actions, I will further develop it to account for the other types of actions that are caused by emotions. I will now introduce several markers of prototypical emotional actions, organized under the headings of impulsivity, flexibility, and bodily underpinnings, to which I collectively refer as the phenomena of emotional motivation. Any viable theory of emotions will need to provide a convincing account of these interrelated phenomena.

2.1. Impulsivity When we act emotionally we typically act impulsively. Impulsive emotional actions are well exemplified by things like “punching someone in a bar brawl, running away in fright, hugging one’s fellow supporters in joy at the victory of one’s football team, and so on” (Pacherie 2001).1 Following Frijda (2010), I emphasize two characteristics of impulsive emotional actions in particular. The first is that in such actions there is a “sense of urge,” which comprises “both an expectation of gain after completing [the action], and haste to fulfill it” (Frijda 2010, 571). I understand urgency as the “preference for early action over late action” (Elster 1 Impulsive emotional actions belong to two different classes. One class is formed by what I call instrumental impulsive actions, namely actions performed as a means to a goal that is constitutive of the emotion. For example, we punch someone out of anger with the goal of removing the obstacle he or she represents, and we run away out of fear with the goal of achieving safety. The other type is what I call displaced impulsive actions, exemplified by actions like piercing holes in the picture of a hated enemy or kicking a door out of anger. Hursthouse (1991) called actions of this sort arational actions, and argued that they are intentional, not done for a reason, and best explained by being in the grip of an emotion. In Scarantino and Nielsen (MS), we offer an alternative explanation of arational actions according to which such actions are displaced in the ethological sense of having goals unrelated to the constitutive goals of the emotions that cause them. We argue that displaced actions result from conflicting or thwarted motivations.

158  Andrea Scarantino 2010, 275). Impulsive emotional actions clearly manifest this temporal preference. There is urgency in the punching out of anger, urgency in the kissing out of lust, urgency in the running out of fear. A second characteristic of impulsive emotional actions is the “use of only part of the available cues that might indicate the adequacy of action” (Frijda 2010, 571). Pacherie (2001, 76) has referred to this aspect as the “shortsightedness” of impulsive actions. I emphasize that we should not straightforwardly equate shortsightedness with irrationality. In some circumstances, the rational thing to do is precisely to engage in an action that uses only part of the available cues for adequacy. This is because using all of the available cues is incompatible with acting quickly, and acting quickly is what rationality demands. In other circumstances, however, the shortsightedness of impulsive actions leads to negative consequences, because the emoter fails to consider key information. My examples will focus primarily on cases of the latter kind. I distinguish between two sources of shortsightedness: reduced information gathering and information processing biases. I will label the conjunction of these elements as partial informational access. In impulsive emotional actions, the time spent collecting information about what to do is limited by one’s preference for acting sooner rather than later. As a result, emoters often fail to consider the long-term consequences of what they do. For example, impulsively slapping someone in a bar fight is not preceded by careful consideration of the long-term consequences—legal, moral, and practical— of engaging in an act of violence. Besides spending little time collecting information, emoters engaged in impulsive emotional actions also manifest various cognitive biases in the way they process information. Their ability to focus attention, recruit memories, draw inferences, evaluate evidence, have accurate perceptions, assign probabilities, assess risks, and so on are all potentially impacted by processing information while in the grip of an emotion (Loewenstein and Lerner 2003; Elster 2010). For example, when we wake up in a panic to the realization that our house is on fire we may not remember that it is freezing outside and fail to pick up our coats on the way out.2 Or we may fail to properly evaluate the evidence about where the fire is most ravaging and end up walking right into it. Or we may mistake the sounds of the crackling fire for human screams. Or we may overestimate the risks involved in going back into the house to retrieve vital financial information.

2.2. Flexibility When we act emotionally we typically also manifest some degree of behavioral flexibility. Flexibility is related to the fact that the same emotion can lead to a number of different actions depending on the circumstances. In fear, we may run, but we may also freeze, shut a door, make a phone call, brandish a gun, keep as quiet as possible, and so



I thank the volume editors for this example.

2

The Motivational Theory of Emotions  159 on. Similarly, we may slap someone out of anger, but also push them, yell at them, spit on them, threaten them verbally, report them to the police, and so on. Emotional flexibility, however, is constrained. Although there are many different things we can do when we are in the grip of a certain emotion, not everything goes. For example, people who are in the grip of fear of a tiger coming their way do not generally take a nap, nor do they approach the tiger to pet it. Similarly, people who are in the grip of anger towards their lover do not generally start juggling balls in the air, nor do they invite their lover out for a nice dinner.3 Another aspect of emotional flexibility to be explained is that not every instance of emotion leads to action. First, we can be afraid or angry, and do nothing about it. Second, there seem to be some emotions like sadness, grief, and depression that by their very nature seem to lead to actively refraining from actions of all kinds rather than to engaging in specific actions.

2.3.  Bodily underpinnings When we act emotionally we also typically manifest concurrent bodily changes that we do not voluntarily choose. These prominently include changes in facial, vocal, and postural expressions, and changes in the autonomic nervous system (Ekman 1999). For example, when we freeze while in the grip of fear our freezing action is often accompanied by involuntary expressions such as upper eyelids raised and lips stretched horizontally and by autonomic physiological changes such as increased heartbeats and sweaty hands. When we slap someone while in the grip of anger our act of slapping is often accompanied by involuntary expressions such as fixed stare, eyes widened, and bared teeth and by autonomic physiological changes such as increased heartbeats and tremors. Two features of emotional bodily changes must be emphasized. The first is their variability. Although many in the history of emotion science have tried to unveil a one-toone correspondence between emotion types and types of bodily changes, there is at this point strong evidence that emotions like fear, anger, disgust, and so on are highly variable both at the level of facial, vocal, and postural expressions and at the level of autonomic bodily changes (e.g. Barrett 2006; Scarantino, forthcoming). The second relevant aspect is that bodily changes are often associated with distinctive emotional feelings (e.g. Goldie 2000). For instance, the feelings associated with fear often involve subjective experiences associated with trembling, sweating, having a dry mouth, and so on. In what follows, I argue that neither Cognitivism nor Perceptualism can provide a satisfactory explanation for the phenomena of emotional motivation I have illustrated in this section. I will then formulate a new theory of emotions designed to account for

3 Similar points are raised by Price (2006). I thank the volume editors for emphasizing the importance of flexibility constraints on emotional actions.

160  Andrea Scarantino such phenomena while addressing some of the standard puzzles in the philosophy of emotions.

3.  Can Cognitivism Make Sense of Emotional Actions? The core idea of Cognitivist theories of emotions is that the identity of an emotion is essentially tied to the cognitions that it involves.4 We can distinguish two main versions of this idea. The first is Judgmentalism, according to which emotions are judgments. The second is Belief and Desire (B&D) Cognitivism, according to which emotions are combinations of beliefs and desires.5 Since I consider the latter variety of Cognitivism to offer a more promising account of emotional motivation, I will focus on it exclusively.6 According to B&D Cognitivism, “emotions are identified with (certain sets of) beliefs and desires” (Marks 1982).7 B&D Cognitivism presupposes the Humean Theory of Motivation, according to which “the psychological states that constitute motivations are pairs of . . . desires and . . . beliefs” (Smith 2010, 153) and desires and beliefs have what Hume called “distinct existences.” As Smith (2010) goes on to explain, this distinctness amounts to the fact that the “possession of a psychological state of the one kind doesn’t entail his possession of a psychological state of the other kind” (155). The “distinct existences” idea has been cashed out in contemporary philosophy with the proposal that beliefs and desires have different directions of fit (Searle 1983). The 4 This is what in previous publications I have called Constitutive Cognitivism, to be distinguished from two other varieties of cognitivism: Etiological Cognitivism, which holds that emotions are caused by cognitions of a particular sort, and Representational Cognitivism, which holds that emotions are intentional states of a particular sort (Scarantino 2010). 5 In psychology, the label “Belief and Desire Theory of Emotions” (BDTE) covers a somewhat larger domain of theories. Reisenzein (in press) has recently distinguished between three main versions of BDTE in psychology: the causal view, according to which beliefs and desires are causal preconditions of emotions but not constituent parts of them, the part–whole view, according to which beliefs and desires are parts of emotions, and the fusion view, according to which emotions result from the fusion between beliefs and desires. What I focus on in this chapter is a version of the part–whole view. 6 Prominent judgmentalists like Robert Solomon and Martha Nussbaum have argued that emotions motivate because desires are either conceptually connected to (e.g., Solomon 2003) or caused by (e.g., Lyons 1980; Nussbaum 2001) the sorts of judgments with which emotions are identified. I find these accounts eminently ad hoc, because no convincing explanation is given of why emotion judgments are conceptually or causally connected with desires by virtue of their content. On the contrary, it seems quite possible to hold an emotion judgment—e.g., the judgment that something is dangerous—without any desire—e.g., the desire to avoid dangerous things—being involved. I will not try to convince the skeptical reader of this point in this chapter, because even if a good account of why and how emotional judgments are conceptually or causally associated with desires were available, Judgmentalism would still fail to capture the phenomena of emotional motivation for the reasons that apply to B&D Cognitivism. For further critical analysis of Judgmentalism, see Scarantino (2010). 7 In this chapter, I will not consider hybrid versions of the belief and desire theory like, for instance, Goldie’s (2000) proposal that emotions are combinations of feelingful beliefs and desires. This theory combines elements of B&D Cognitivism with elements of the feeling theory. For a critical analysis of how Goldie’s theory applies to emotional actions, see Scarantino and Nielsen (MS).

The Motivational Theory of Emotions  161 direction of fit of a mental state specifies whether such mental state aims to represent how things are (mind-to-world direction of fit) or aims to represent how things are to be (world-to-mind direction of fit). Beliefs are the paradigmatic mind-to-world attitudes because they aim at being true, whereas desires are the paradigmatic world-tomind attitudes because they aim at being realized. B&D Cognitivism identifies emotions with pairs of what I call emotion-constitutive beliefs and desires.8 Here is an example of such combinations: Fear of a tiger is the belief that the tiger is dangerous and the desire to avoid dangerous things

The belief involved may be labeled a core relational theme (CRT) belief, because it registers the obtaining of what Lazarus (1991) called core relational themes such as dangers for fear, offences for anger, losses for sadness, moral violations for guilt, failures to live up to an ego ideal for shame, and so on.9 To explain emotional actions, B&D cognitivists must introduce a further posit, namely what I call an action-explaining belief and desire pair, and make the assumption that it was caused by the emotion-constitutive belief and desire pair. Suppose for instance that we want to explain why Matt started running out of fear of a tiger. First, we would have to posit something like the following action-explaining belief and desire pair: Matt has a desire to get away from the tiger and believes that by running he will get away from the tiger

Second, we would have to assume that this action-explaining combination of beliefs and desires was caused by the emotion-constituting combination that instantiates being afraid of a tiger, namely believing that the tiger is dangerous and desiring to avoid dangerous things. The key problem for B&D Cognitivism is that acting on the basis of belief and desire pairs is neither necessary nor sufficient for manifesting the phenomena of emotional motivation I have illustrated in section 2. I will first consider the problem of non-necessity, and then the problem of non-sufficiency. 8 Several other proposals have been made concerning what the relevant beliefs and desires should be. For instance, Green (1992) has argued that fearing that a tiger will attack me is constituted by believing with certainty that a tiger will attack me and desiring that a tiger not attack me. This will not do. One clearly need not believe with certainty in order to be afraid. I can have a low degree of belief that a tiger will attack me and still be afraid of it. Reisenzein (in press) has suggested instead that a person fears p if she desires not-p and is uncertain about p. This formulation encounters a second problem, namely that it is not sufficient for fearing that p that one believes that p and desires that not-p. For example, one can believe that an insect got into one’s soup and desire that it did not happen, without being afraid of the insect (one will be disgusted instead). The most promising version of B&D Cognitivism connects the content of the emotion-constituting belief with core relational themes, as I go on to suggest in the body of the chapter. 9 Examples of emotion-constitutive belief and desire pairs can quickly be constructed by making these core relational themes the content of the relevant beliefs. Thus, we can think of anger about Mario’s behavior as the belief that Mario’s behavior instantiates an offense and the desire to not be offended, guilt about having forgotten to pay back a loan from a friend as the belief that forgetting to pay back a loan from a friend is a moral violation and the desire that one does not commit a moral violation, and so on.

162  Andrea Scarantino It has often been remarked that emoters in the grip of some emotion E may lack the belief-desire combo with which B&D Cognitivists identify E. The most commonly discussed missing ingredient is the relevant CRT belief. For example, fear of a snake in a snake phobic patient is often not accompanied by the belief that the snake is dangerous, and yet persists in motivating snake avoidance. Emotions are in this sense recalcitrant to reason (D’Arms and Jacobson 2003; Benbaji 2012). Furthermore, impulsive emotional actions often occur so quickly that it is highly doubtful that they are preceded by the formation of a belief and a desire, and by the mental act of putting the two together in a practical inference. In non-emotional cases, this problem is handled by positing the notion of an intention-in-action—e.g., the intention-in-action to run now—that is formed on the fly and causes the action without being in turn caused by a belief and desire pair (Searle 1983). Intentions-inactions are contrasted with forward-looking (or prior) intentions, which are intentions to act at a later time and are generally formulated at the conclusion of a standard process of practical reasoning (Searle 1983; Bratman 1987; Pacherie 2008). This solution is not available to the B&D Cognitivist, because unless the emotion-constitutive belief and desire pair causes the intention-in-action, the resulting action will no longer be an emotional one. And as it is doubtful that impulsive emotional actions are preceded by the formation of action-explaining belief and desire pairs, it is a fortiori doubtful that they are preceded by the formation of emotion-constituting belief and desire pairs that cause intentions-in-action. A less familiar problem for B&D Cognitivism is the non-sufficiency of belief and desire pairs for generating emotional actions. I call this the Unemotional Twin Action Problem: Whatever belief, desire, and intention combos may be said to motivate an emotional action, there will be a non-emotional “twin” version of the action caused by the very same belief, desire, and intention combos, namely a version that fails to manifest the right sort of impulsivity, flexibility, and bodily underpinnings. If so, belief, desire, and intention combos are never a sufficient cause of emotional actions.

Consider Regular Matt and Twin Matt. By hypothesis, these two Matts are similar in all relevant respects—e.g. IQ, moral values, age, physical ability, eyesight, etc.—except for the fact that Regular Matt is an ordinary guy who works in a bank whereas Twin Matt is a Delta Force marine trained not to become afraid in dangerous circumstances.10 Twin Matt and Regular Matt are visiting a circus after hours and realize that a tiger has escaped her enclosure. Let us assume that they both form a desire to get away from the tiger and believe that they will achieve such objective by running away from the tiger, so they both run away. Furthermore, suppose that these action-explaining beliefs

10 This example and the discussion that follows are inspired by a similar example I found in Nash (1989). I thank the volume editors for pressing me to discuss this issue.

The Motivational Theory of Emotions  163 and desires are caused in both Matts by the belief that the tiger is dangerous and by the desire to avoid dangerous things. One good feature of this explanation is that it captures an important aspect of the flexibility of emotional actions. Since the same emotion-constituting belief and desire pair can cause different action-explaining belief and desire pairs, B&D Cognitivism has the resources to explain why some forms of fear lead to running whereas others lead to climbing, shooting a gun, standing still, and so on, depending on the circumstances. All such actions can reasonably be caused by the desire to avoid danger while believing one is in danger. On the other hand, the B&D account does not have the resources to distinguish between emotional and non-emotional ways of acting flexibly. Being motivated by a belief and desire pair is not sufficient for acting impulsively, for manifesting the constraints on flexibility that derive from impulsivity, and for displaying the bodily underpinnings typical of emotions. Delta Force-trained Twin Matt coldly assesses the situation, remembering where the exit he just passed through is located. Without breaking a sweat or trembling, he quickly moves towards such exit, while picking up a small child he finds on his way out. Once Twin Matt realizes that the exit door is closed from the outside, he recalls having seen a documentary indicating that standing still without looking at the tiger is the best way to neutralize it, and so he proceeds to remain entirely motionless while holding the child as the tiger approaches. Regular Matt, on the other hand, starts to tremble and sweat profusely, forgets where the closest exit is located even though he just passed through it, mistakes a shadow for a second approaching tiger, and starts running in a random direction while trampling upon a small child and leaving him face down on the ground. After he reaches a faraway exit, he realizes that the exit door is closed from the outside. Although he also remembers watching a documentary on the advantages of standing still when a tiger is around, he cannot bring himself to do it, and so he frantically climbs a small tree well within the reach of the tiger while the tiger approaches. Regular Matt’s actions display the urgency, partial informational access, constrained flexibility, and bodily underpinnings of prototypical emotional actions, whereas Twin Matt’s actions do not. And yet by hypothesis the two Matts are motivated by the very same belief and desire pairs. It follows that being motivated by the sorts of belief and desire pairs posited by B&D Cognitivism is not sufficient for being motivated to act emotionally. A B&D Cognitivist may reply that it is not the job of action-explaining belief and desire pairs to account for the emotionality of actions: Their job is simply to explain why an action, rather than a happening, is taking place. Supplemental explanations may be offered to capture the emotional element that mere action-causing belief and desire pairs miss. For instance, the B&D Cognitivist may suggest that emotional actions are those in which action-causing belief and desire pairs are in turn caused by emotion-constitutive belief and desire pairs. This rejoinder would not help, because

164  Andrea Scarantino actions caused by belief and desire pairs that have other belief and desire pairs as their causes are still not going to display the sort of urgency and partial informational access that is characteristic of emotional actions. 11

4.  Can Perceptualism Make Sense of Emotional Actions? The core idea of Perceptualism is that the identity of an emotion is essentially tied to the perceptions that it involves. The two most promising strategies to account for the phenomena of emotional motivation within this research program are what I call Reflexive Perceptualism and Valenced Perceptualism.12 Reflexive Perceptualism is the theory first defended by William James and further developed by Antonio Damasio. James (1884; 1890)  famously proposed that emotions are feelings generated by the perception of bodily changes. Crucially, he assumed that these bodily changes—which on his view include expressions (e.g., grimacing), 11 A further line of response B&D Cognitivists could explore is that emotions involve strong desires, and that the strength of desire accounts for the phenomena of emotional motivation. This is the line proposed by Marks (1982), who suggested that what is distinctive about emotions is precisely that they involve strong desires. “If the primary desire is very strong,” Marks writes, the emoter “may be too hasty in some of his inferences, or make a logical misstep, or make a literal misstep for failure to attend carefully to where he is putting his feet, and so on” (235). His view is also that when the desire is strong “a number of physiological changes . . . are quite naturally connected to the B/D set” (235). This reply is not convincing, because there is no good reason to think that Twin Matt’s desires to avoid danger and to run away are any weaker than Regular Matt’s desires. On the Humean picture of motivation, the strength of a desire hinges on whether it motivates action when competing with other, incompatible desires. For example, a desire that X be realized is stronger than a desire that Y be realized just in case the agent is motivated to do X rather than to do Y. A strong desire tout court is consequently one that is motivationally stronger than a great many other competing desires. But both Regular Matt and Twin Matt are acting on a desire to avoid danger that is strong in this absolute sense, and clearly stronger than any competing desires they may have had when they first saw the tiger, since they both run towards the exit. The difference is that whereas Regular Matt goes about realizing his strong desires to avoid danger and run towards the exit emotionally, Twin Matt goes about realizing them non-emotionally. A final line of defense the B&D Cognitivist could explore is to embed into beliefs and desires whatever properties of emotional motivation the account is missing. This may for instance involve suggesting that the desires involved in emotion are not only “strong” but “urgent,” “information-blocking,” “bodily,” and so on. Besides having a patently ad hoc character (see Scarantino 2010 for a general critique of this common cognitivist strategy), this solution would still leave B&D Cognitivism with the problem that beliefs and desires appear to be non-necessary for the instantiation of an emotion, due to the problems of recalcitrance and speed of impulsive emotional actions. 12 A third and distinctly less promising strategy is what I call Primitivist Perceptualism. It amounts to the stipulation that the sorts of perceptions involved in emotions have the ability to motivate to action. For example, Doring (2003, 224) recently rejected the Humean assumption that all motivation consists of pairs of beliefs and desires, suggesting instead that “the motivational force of an emotion . . . [is] to be explained in terms of the feeling-dimension of the emotion.” On Doring’s (2003) view, fear of a tiger is the felt representation of the tiger as dangerous, namely a combination of an “affect” and of an intentional representation. This sort of perceptual primitivism is problematic for several reasons. First, since not every feeling is motivational (e.g., the feeling of one’s chair on the back while sitting), the reason why having emotional feelings leads to motivation is left obscure and must be accepted as a brute fact. Second, assuming that emotions do not motivate unless they are felt precludes the possibility of unfelt yet motivating emotions, a contentious commitment better avoided in light of some recent behavioral evidence for unconscious emotions (e.g., Winkielman and Berridge 2004). Thirdly and most importantly, this explanation encounters a version of the

The Motivational Theory of Emotions  165 autonomic changes (e.g., trembling) and full-fledged actions (e.g., striking)—follow directly the perception of some exciting fact. They follow it directly in the sense that “peculiarly conformed pieces of the world’s furniture” will “fatally call forth”—i.e. bring about in a reflex-like fashion—the bodily reactions whose perception is the emotion (James 1890, 191). Damasio (1994; 2003) rejected James’s commitment to the felt dimension of emotions by allowing perceptions of bodily changes to be both conscious and unconscious. He also expanded what counts as a bodily change, including for instance neural changes to the somatosensory cortex (he called these “as if ” bodily changes). However, Damasio preserved the Jamesian account of how emotions motivate, describing emotional responses as “automatic and largely stereotyped” (Damasio 2003, 35). He explicitly assimilated their mechanism of operation to that of other automated regulatory processes such as the startle reflex or pain behaviors. The problem with Reflexive Perceptualism is that it completely misses the boat on impulsive emotional actions. Although some emotional actions such as jumping sideways out of fear of a suddenly looming car, or turning around with bared teeth and clenched fists out of anger when suddenly poked in the back are indeed reflex-like, most emotional actions do not work that way. As we have seen, they are impulsive rather than automatic, flexible rather than stereotyped, and involve variable rather than rigid bodily changes. On Damasio’s (2003) view, flexibility enters the picture only after full-fledged practical reasoning enters the picture.13 This is a mistake, because impulsive emotional actions already manifest flexibility, even though such flexibility is constrained by their characteristic impulsivity. A second strategy for making sense of the phenomena of emotional motivation is represented by what I  call Valenced Perceptualism. Its core insight is that emotions motivate because they contain a valenced—i.e. positive/pleasant or negative/ Unemotional Twin Action problem. It may well be the case that both Regular Matt and Twin Matt entertain the felt representation of the tiger as dangerous, but only Regular Matt reacts to it with the sort of impulsivity, flexibility, and bodily underpinnings that instantiate a prototypical emotional action. Simply being motivated by a felt representation appears compatible with both emotional and non-emotional ways of acting, just like being motivated by a belief and desire pair is. 13 Damasio’s (1994; 2003) signature proposal is that emotions and practical reasoning are closely interrelated. According to the somatic marker hypothesis, when normal subjects consider options in practical reasoning they elicit memories of past emotions experienced in comparable situations. These emotion-related memories lead to the activation of somatic markers, which are “gut feelings” marking options as positive or negative in light of their expected emotional consequences. Damasio emphasizes that the activation of somatic markers “is not a substitute for proper reasoning [but plays] an auxiliary role, increasing the efficiency of the reasoning process and making it speedier” (Damasio 2003, 148). Occasionally, somatic markers signal so strongly that the process of practical reasoning is interrupted and a certain option is rejected or chosen right away. More commonly, somatic markers produce incentive signals and alarm signals that are further weighed in the process of practical reasoning. The somatic marker hypothesis predicts that the disruption of either the machinery of emotions or the machinery of emotion-related memory will lead to the disruption of the machinery of practical reasoning. This is because if decision-makers cannot rely on memories of past emotional experiences, they will be unable to automatically label options positively or negatively in light of their expected emotional consequences, and their decision-making process will be adversely affected (e.g.,

166  Andrea Scarantino unpleasant—component. The most sophisticated proposal in this vein comes from Jesse Prinz (2004). On his view, emotions are combinations of embodied appraisals and valence markers. Embodied appraisals are consciously or unconsciously perceived bodily changes—autonomic, expressive, behavioral, and/or neural changes—that are about core relational themes (Prinz 2004). They are about core relational themes in the sense that such themes are represented by bodily changes. In turn, bodily changes represent core relational themes in light of a teleosemantic account of representation, according to which mental states represent what they have the function of being elicited by/correlate with (Prinz 2004, 54; Dretske 1988). On this view, fear is about dangers because it has the function of being elicited by dangers, anger is about offences because it has the function of being elicited by offences, shame is about failures to live up to an ego ideal because it has the function of being elicited by failures to live up to an ego ideal, and so on. Valence markers, on the other hand, are positive and negative reinforcers that tell the organism, respectively, “more of this” or “less of this.”14 On this view, fear of a tiger is a combination of an embodied appraisal consisting of “a racing heart and . . . other physiological changes” (Prinz 2004, 69) that have the function of being elicited by dangers, and a negative valence marker that says “less of this.” Importantly, “this” refers not to the external stimulus, but to the emotion itself. Prinz (2004) argues that valence markers are associated with “inner state goals” (194) in the sense that they tell us to either continue how we are feeling (positive valence markers) or to change how we are feeling (negative valence markers). As a result, “emotions do not impel actions directly” (228). They do so indirectly, because valence markers are ultimately “commands to sustain or eliminate a somatic state by selecting an appropriate action” (229). According to Prinz, this indirect link with actions makes emotions motives rather than motivations. A “motive provides a reason for action . . . [whereas] a motivation is that which impels us to act” (Prinz 2004, 193). On this view, fear provides a reason for running, but it does not impel us to run. Motivations, on the other hand, “[tell] us to change how we are acting” by directly specifying “action goals” (194). Thus, motivations are associated with injunctions to act rather than injunctions to change how we feel. Prinz uses hunger as an example of a motivation. Hunger has in common with fear the fact that it contains an embodied

they will be unable to quickly eliminate options that have previously led to strongly negative consequences). The jury is still out on whether the somatic marker hypothesis is confirmed (see, e.g., Dunn, Dangleish, and Lawrence 2006; Reimann and Bechara 2010). What is clear is that impulsive emotional actions do not result from a standard process of practical reasoning, whatever role somatic markers may play in it. 14 Prinz (2004) argues that most emotions are characterized by unique valence markers. For instance, fear is associated with a negative reinforcer, whereas happiness is associated with a positive reinforcer. On the other hand, surprise is associated with both negative and positive reinforcers depending on the circumstances.

The Motivational Theory of Emotions  167 appraisal (it represents undernourishment) and a negative valence marker (less of this!), but it differs from fear because it also contains an action command (eat!). A good feature of Prinz’s (2004) account is that it easily captures the bodily underpinnings of emotional actions. Since emotions contain embodied appraisals, their bodily aspect is straightforwardly accounted for.15 The trouble begins when it comes to accounting for the impulsivity of prototypical emotional actions. This is because we can once again imagine Regular Matt and Twin Matt to be motivated by a negative valence marker, but to go about its elimination in, respectively, an emotional and a non-emotional way. Although Prinz did not address this problem, a tentative solution for it can be found within the confines of his own theory. What we must assume is that valence markers can differ in intensity, and that emotions typically contain strongly valenced markers. If we think of emotions as containing highly negative or highly positive valence markers, this can arguably make sense of the urgency and partial informational access characteristic of impulsive emotion actions. For example, fear of a tiger would contain a highly negative valence marker (LESS OF THIS! LESS OF THIS!) with the potential for eliciting a preference for early rather than late action, a limited investment in information gathering and cognitive biases in information processing, presumably driven by the distraction created by attending to a strongly negative or strongly positive marker. The main problem with Prinz’s (2004) Valenced Perceptualism is that it posits the wrong kind of flexibility for emotional actions. On the positive side, if emotions contained valence markers that set inner goals, this would lead to multiple action goals that satisfy the same inner goal depending on the circumstances. For instance, the goal of no longer being afraid of a tiger (an inner goal) could be achieved by fleeing the tiger, by climbing a tree, by shooting the tiger, and so on. This is an improvement over James and Damasio, who assumed that emotional responses are always automatic and stereotyped. On the negative side, if fear simply contained a command to select whatever action achieves the inner goal of eliminating fear, such inner goal could just as well be achieved by ingesting a fear-calming pill or by running towards the dangerous stimulus so as to perish more quickly. But fear of a tiger does not motivate us to the pursuit of such action goals, even though they would indeed serve the inner goal of eliminating fear. Rather, fear motivates us to pursue the goal of avoiding the tiger in a variety of flexible ways. Once the danger constituted by the tiger is avoided, fear dissipates, but this does not show that fear has the inner goal of self-elimination any more than

15 Even though Prinz (2004) is tempted by the idea that each emotion type is associated with unique bodily changes or at least a unique range of bodily changes, his account is compatible with the view that different tokens of the same emotion type manifest variability in terms of the bodily changes associated with them. A more troubling aspect of Prinz’s (2004) theory is that a great many emotions do not seem to involve distinctive bodily changes at all (e.g., guilt at having forgotten a friend’s birthday). This suggests that embodied appraisals are typical rather than necessary for the instantiation of an emotion.

168  Andrea Scarantino the disappearance of hunger due to eating shows that hunger has the inner goal of self-elimination. The central ingredient missing from Prinz’s theory is the idea that emotions already contain action goals rather than inner goals. Once we understand in what sense having an emotion already involves having an action goal, a new theory of emotions as action control systems becomes available. According to it, bodily changes are organized towards the pursuit of an action goal and the perception of bodily changes becomes a side show to the main act of goal pursuit. The rest of my chapter will be devoted to developing just such a theory.

5.  The Motivational Theory of Emotions The core idea of the theory of emotions I develop in this chapter is that the identity of an emotion is essentially tied to a prioritized tendency to action (or inaction) with the function of being elicited by a core relational theme. I call it the Motivational Theory of Emotions, because it replaces the primacy of the appraisal and feeling aspects of emotions with the primacy of their motivational dimension. The theory comprises four component parts, which I illustrate in turn: (1) Prioritized action and inaction tendencies, (2) A two-level action control structure that provides a unified account of the phenomena of emotional motivation, (3) An analysis of emotional intentionality, (4) An account of how emotions differ from one another.

5.1.  Prioritized Action and Inaction Tendencies Many in the history of emotion theory have suggested that a significant aspect of emotions is that they have an impact on our tendencies to act. Aristotle already had described anger as an impulse for revenge caused by an appraisal of slight and accompanied by unpleasant feelings, even though in his description of emotions other than anger he neglected to specify the action tendencies involved to focus on the appraisals and feelings associated with emotions. An early statement of the primacy of actions over feelings can be found in Dewey’s (1895) theory of emotions, according to which an emotion crucially involves “a disposition, a mode of conduct, a way of behaving” (16). To exemplify, Dewey remarked that when we say that “John Smith is very resentful at the treatment he has received, or is hopeful of success in business, or regrets that he accepted a nomination for office,” what we mean is not “chiefly” that “he has a certain ‘feel’ ” but rather that he has “assumed a readiness to act in certain ways” (Dewey 1895, 16). The behaviorist movement, both in its psychological (Watson 1919; Skinner 1953) and in its philosophical (Ryle 1949) wings, strongly emphasized the connection between emotions and behavior (to the detriment of all other components of emotions), but it failed to understand emotions as the causes of behavior. As Skinner (1953, 160) put it, the “emotions are excellent examples of the fictional causes to which we commonly

The Motivational Theory of Emotions  169 attribute behavior.” Furthermore, behaviorists had a limited understanding of the complexity of the behaviors associated with emotions. For example, in Watson’s (1919) theory all emotional behaviors are conditioned or unconditioned reflexes, a view that, as we have seen, fails to do justice to most emotional actions. The best worked out account of the relation between emotions and actions in contemporary emotion theory can be found in the work of the psychologist Nico Frijda (1986; 2007; 2010).16 His trademark contribution is the thesis that emotions are states of action readiness with the distinctive property of control precedence. The Motivational Theory of Emotions (MTE) I propose in this chapter is heavily indebted to Frijda’s work, but improves upon it in various ways. Among other things, MTE emphasizes the importance of a two-level control structure of emotional actions, it expands on Frijda’s original theory to account for reflex-like emotional actions and for planned emotional actions, and, most importantly, it endows emotions with intentionality, an ingredient sorely missing from Frijda’s account. Frijda’s (1986) orienting thought is that emotions are types of action tendencies (see also Arnold 1960). Action tendencies are in turn understood as “states of readiness to execute a given kind of action,” where the kind of action “is defined by . . . [the] . . . end result aimed at” (70). I refer to such an end result as the relational goal of the emotion. Frijda (1986) suggests for instance that the “action tendency of anger is interpreted as a tendency to regain control or freedom of action [henceforth, an attack tendency]— generally to remove obstruction” (88). Fear is associated with the action tendency of “avoidance,” associated with the relational goal of achieving one’s “own inaccessibility” (henceforth, one’s own safety) with respect to a certain stimulus. Disgust is associated with the action tendency of “rejecting” (henceforth, expelling), characterized by the relational goal of “removal of object.” These examples indicate that relational goals are abstract goals that need to be situated in a concrete context in order to guide bodily changes. This is typical of most goal-oriented processes, including non-emotional intentional actions. When we decide to get to school by 10am in order to attend a talk, the overarching action goal of getting to school by 10am can be achieved through a variety of situated goals (e.g., taking a bus at 9:20am, taking the subway at 9:30am) (cf. Pacherie 2008). Each of these situated goals can in turn be achieved by a variety of motor goals that directly guide bodily changes. For simplicity of reference, I will distinguish between the relational goal of an emotion and its relational sub-goals, understood as the collection of situated and motoric goals by which the relational goal can be achieved. Frijda (1986) distinguishes between two types of actions associated with emotions:  physical actions involving motor control (e.g., insulting) and mental actions involving exclusively mental control (e.g., thinking about insulting). He emphasizes that emotions often lead to mental actions without leading to physical actions, as when 16 Kovach and Delancey (2005) have also offered a philosophical account of emotions that emphasizes their ability to motivate. I discuss some aspects of their theory in Scarantino and Nielsen (MS).

170  Andrea Scarantino we fantasize about slapping our boss without actually engaging in any physical action against him or her. Frijda’s (1986) main contribution is having argued that the sorts of action tendencies with which emotions should be identified have control precedence. Absent control precedence, the identification of emotions with action tendencies would be ill-conceived, because there are innumerable action tendencies that have nothing to do with emotions (e.g., the tendency to scratch one’s head when thinking hard, the tendency to over prepare for exams, etc.). Here is how Frijda (1986, 78) describes control precedence: Action tendencies—and action readiness changes generally—clamor for attention and for execution. They lie in waiting for signs that they can or may be executed; they, and their execution, tend to persist in the face of interruptions; they tend to interrupt other ongoing programs and actions; and they tend to preempt the information-processing facilities . . . Evidently, then, action tendencies are programs that have a place of precedence in the control of action and of information processing. We therefore say: Action tendencies—action readiness changes generally—have the feature of control precedence.

The key idea here is that what gives an action tendency its emotional character is that it seizes control of the emoter with respect to his or her mental and physical actions. Although Frijda did not organize the various components of control precedence into subcategories, it is useful to do so. I propose that an action tendency acquires control precedence—henceforth, it becomes prioritized—by virtue of the two functional components of Precedence and Preparation. Precedence refers to the fact that an action tendency takes precedence over other actions and states of action readiness. This is manifested by the fact that prioritized action tendencies interrupt other processes in their pursuit of a relational goal, clamor for attention, persist in the face of interruptions, and pre-empt access—in memory, inference, perception, etc.—to information. Preparation refers to the fact that in a state of prioritized action tendency we are not simply ready for action, but actively preparing for it. This is manifested by the fact that such states tend to be accompanied by preparatory bodily changes, to clamor for execution, and to wait for signs that they may be executed. The prioritization of an action tendency comes in degrees. This is to say that an action tendency can involve a higher or lower degree of Precedence and/or Preparation. A maximal degree of prioritization will be typical of prototypical instances of an emotion such as being afraid of a charging tiger, which will involve an avoidance tendency with maximal Precedence and maximal Preparation. Being afraid of a potentially hostile crowd prior to giving a talk, on the other hand, will involve a significantly lower degree of Precedence and Preparation, even though the avoidance tendency associated with it still “clamors for attention and for execution” to some extent. As with every degree property, at some point it will become indeterminate whether or not a given action tendency is prioritized, namely whether it involves enough

The Motivational Theory of Emotions  171 Precedence and/or Preparation to be endowed with control precedence. This may give rise to states of action readiness that, by the lights of the Motivational Theory of Emotions, neither fully qualify as emotions nor fully qualify as non-emotions. The account I have described so far faces two problem cases. The first is constituted by emotions that are energizing, but do not seem to involve the pursuit of any specific relational goal. A standard example is that of existential joy, which does not seem to predispose the emoter to pursue any relational goal in particular. The second problem case is constituted by emotions such as sadness, grief, and depression that appear to considerably reduce one’s willingness to act and pursue all kinds of relational goals. Frijda’s solution to these two problems is to introduce alongside the notion of an action tendency the notion of a “state of action readiness as such,” which includes both what he calls “activation modes” like joy and “null states” like sadness, grief, and depression. On Frijda’s (1986, 71) view, “emotions . . . can be defined as modes of relational action readiness,” which can take either the form of an action tendency, an activation mode, or a null state. I propose we adopt a slightly different taxonomy. The difference between joy, sadness, and fear, I submit, is not that fear has a relational goal whereas joy and sadness lack it, but that the relational goals of joy and sadness are significantly less specific than the relational goal of fear. For joy, I submit, the goal is to relate as such. As Fredrickson and Cohn (2008) put it, joy “creates the urge to play, push the limits, and be creative” (782). The joyful person is ready to engage in an open range of actions, and actively prepares for this open engagement with the world with a generalized state of arousal. For sadness, grief, and depression, on the other hand, the goal is not to relate as such. The sad/grieving/depressed person is disengaged from the world in an undifferentiated fashion, in the sense that there is not much of anything that they wish to do.17 So I  propose we distinguish between focused action tendencies such as fear and disgust, which have specific relational goals, unfocused action tendencies such as joy, which has the generic relational goal of relating as such, and inaction tendencies, which have the generic relational goal of not relating as such. What makes focused action tendencies, unfocused action tendencies, and inaction tendencies members of a theoretically unified category is that, insofar as they are emotions, they all instantiate control precedence. This is to say that fear, joy, and sadness all seize control of the emoter in the sense that the relational goals of, respectively, achieving safety, relating as such, and not relating as such acquire precedence over other goals, clamor for attention and execution, and pre-empt information processing while the body prepares for, respectively, fearful actions, undifferentiated engagement, and undifferentiated disengagement.

17 These remarks are inspired by Frijda’s (2008, 72) comment that emotions that correspond to states of action readiness other than action tendencies “have no aim other than to relate or not to relate in general.”

172  Andrea Scarantino

5.2.  How The Two-Level Control Structure of Emotional Actions Generates Impulsive Actions To understand how emotions lead to emotional actions, the first thing to realize is that, with one notable exception,18 emotional actions result from a two-level structure of control. One level is constituted by the action and inaction tendencies with control precedence I have just illustrated, which are special-purpose motivational structures. The other level is constituted by a set of general-purpose capacities I refer to as rational control, which determines if and how the prioritized (in)action tendency is manifested. In a seminal book-length treatment of action, Gallistel (1980) pointed out that two-level structures of control are the primary means to achieve behavioral flexibility in the animal kingdom. Although Gallistel (1980) did not explicitly consider emotions, we can think of them as good examples of what he called central motive states, namely states that “lay down a frame or general direction for behavior by selectively potentiating coherent sets of behavioral options” (322), with the “lower levels fill[ing] . . . in details within the general pattern” and giving “behavior its flexibility, its ready adaptation to momentary circumstances.” Once prioritized action (or inaction) tendencies have laid down a general direction for behavior and suitably potentiated (or depotentiated) sets of behavioral options, emoters exercise two types of rational control.19 Compatibility control involves monitoring that the emotion’s relational goals and sub-goals are compatible with the emoter’s other goals and value system. For example, an emoter may ask herself whether getting back at her boss after being insulted during a meeting is compatible with her career goals. Or the emoter may ask whether withdrawing from all actions out of depression is compatible with other goals such as getting a graduate degree, making one’s partner happy, enjoying the simple pleasures of life, and so on. Additionally, the emoter may ask herself whether a particular way of going about fulfilling the emotion’s relational goal—e.g. cursing at the boss right there and then in the middle of the meeting, slapping him after the meeting, etc.—is compatible with her goals and values. The second type of rational control, executive control, involves securing that the emotion’s relational goal is translated into a set of sub-goals that is instrumentally adequate. For example, once the decision to slap the boss after the meeting is made, executive control monitors the specific bodily changes involved in the slapping action, turning it into an appropriate set of motor goals, and seeing to it that the slapping action is completed successfully. Since sadness, grief, and depression lead to inaction, executive control is significantly curtailed in such emotions, in the sense that it is limited to the sort of executive control involved in actively refraining from action.

18 The exception is constituted by emotional reflexes, which I discuss shortly. 19 The two levels of rational control I introduce for emotions are inspired by Pacherie’s (2008) discussion of “tracking control” and “collateral control” in garden-variety actions.

The Motivational Theory of Emotions  173 This two-level control structure—the level of prioritized (in)action tendencies and the level of compatibility and executive control—can ground a unified explanation for the phenomena of emotional motivation on which Cognitivism and Perceptualism stumbled. Impulsive emotional actions come about whenever the interaction between rational control and a prioritized action tendency quickly leads to an intention-in-action, namely an intention to act now so as to achieve the relational goal of the action tendency. This intention-in-action causes and guides bodily movements so as to translate the relational goal—e.g., achieving safety—into a sequence of executable relational sub-goals—e.g., reaching an exit, opening a door, shooting a gun, etc. The impulsivity and bodily underpinnings of prototypical emotional actions, which B&D Cognitivism had a hard time accounting for, follow directly from the fact that the intention-in-action is formed and acted upon while an action tendency with control precedence is up and running. As a result, impulsive emotional actions manifest urgency, namely a preference for early versus late action, partial informational access, because the investment in information gathering and its quality are constrained by the pre-existence of a prioritized action tendency, and bodily underpinnings, because one of the elements of control precedence is bodily preparation. On this view, the difference between Regular Matt and Twin Matt is that only the former is seized by a prioritized avoidance tendency that “clamors for attention and execution,” constraining Regular Matt’s ability to perform compatibility and executive controls. Twin Matt, on the other hand, is relying on standard practical reasoning, without having to deal with a state of action readiness that monopolizes attention, pre-empts access to relevant information, prepares the body for action, etc. The flexibility of impulsive emotional actions results from the fact that a prioritized action tendency only determines an abstractly described relational goal, allowing compatibility and executive control to fill in the details. On this picture, emotional flexibility is limited by two constraints that neither Cognitivism nor Perceptualism fully captured. The first is that the actions one selects while in the grip of emotion tend to have sub-goals of the relational goal of the action tendency. This is why fearful emoters do not take fear-stopping pills nor walk towards tigers, contrary to what Prinz’s theory would predict. These actions are not means to the relational goal of achieving safety, even though they are means to the inner goal of eliminating fear. The second constraint is that the presence of a prioritized action tendency makes the selection of certain actions hard to implement. This is why, contrary to what B&D Cognitivism would predict, a fearful agent like Regular Matt would have trouble standing still next to a tiger even if he believed that the best way to achieve safety was standing still and desired to achieve safety. A powerful behavioral tendency to avoid the tiger makes standing still while the tiger approaches hard to achieve.20 20 Although I have so far focused on flexibility constraints related to impulsivity, another source of constraint on flexibility is constituted by the very bodily changes involved in prioritized action tendencies. As Zhu and Thagard (2002, 31) have pointed out, “[i]‌t is hard to put a thread through the eye of a needle when

174  Andrea Scarantino

5.3.  Explaining Planned Emotional Actions, Actions Done out of Expected Emotions, and Reflex-Like Emotional Actions The two-level structure of control I have introduced can help us understand other emotional phenomena. In particular, it can explain why emotions may not lead to any actions at all and why emotions may lead to actions that are far in time and space from the initial emotional incident. First, the interaction between prioritized action tendencies and rational control can lead to what I call the extinction of the action tendency. It is often the case that the output of compatibility control is the conclusion that pursuing the relational goal of the action tendency is incompatible with the other goals and values of the emoter. For example, if my boss insults me at a board meeting and as a result I form a tendency to attack him, I may quickly conclude that I should not attack him in any form, either now or at a later time, because having a successful career at the firm is more important to me than getting even. In this case, the effect of rational control is to extinguish the prioritized action tendency, a process a careful observer may pick up on by spotting bodily leakages at the level of expressions and autonomic changes, the extinction of which may take a little time (Ekman 1999). The interaction between prioritized action tendencies and rational control can also have as an outcome the elicitation of a forward-looking intention with the same relational goal of the prioritized action tendency (or with a relational sub-goal of it). As I mentioned in section 3, a forward-looking intention is an intention to act at a later time. For example, rational control may lead me not to extinguish the attack tendency I have formed towards my boss, but rather to form a forward-looking intention to write a letter of complaint about him tomorrow (relational sub-goal) or, more generally, to get back at him in some way at a later time (relational goal). Once these forward-looking intentions cause actions, the result is what I call an emotional planned action, a form of emotional action that manifests maximal flexibility and relies on the full force of practical reasoning.21 Other examples include taking legal action against one’s landlord out of anger one month after a heated confrontation, buying a gift for a friend out of gratitude one week after having received his visit at the hospital, writing a letter of complaint to the IRS out of anger every April 15 for years after having been audited. These sorts of planned actions are properly called emotional just in case three conditions apply. First, the forward-looking intention that causes emotional planned actions you are in a state of rage or anxiety, simply because you cannot accurately control your hands in such a mood.” 21 Pacherie (2001) refers to what I have called planned emotional actions as semi-deliberate emotional actions. Her account differs from mine with respect to the role played by action tendencies. Whereas her proposal is that in semi-deliberate actions “the action tendency gets converted into a conscious intention to pursue a certain goal” (2001, 80), my view is that the action tendency must continue to exist and exercise some degree of control precedence after the formation of the forward-looking intention for the resulting planned action to qualify as emotional.

The Motivational Theory of Emotions  175 aims to satisfy the relational goal of an earlier prioritized action tendency, or one of its relational sub-goals. So if the relational goal of the attack tendency characteristic of anger is removing an obstacle, the forward-looking intention to write a letter of complaint about the boss the morning after is still a means to realize the goal of removing an obstacle. Secondly, the forward-looking intention that causes emotional planned actions is formed while a prioritized action tendency is up and running. As a result, forward-looking intentions of this sort may partake at least in part in the shortsightedness of impulsive emotional actions, because they are formed “in the heat of the moment.” But unlike intentions-in-action, which immediately cause actions, forward-looking intentions can be revised. This is why we often give up on forward-looking intentions formed during emotional incidents: Although they passed the compatibility control performed while in the heat of the moment, they do not pass the compatibility control performed coldly on the day after. Thirdly, the prioritized action tendency must not disappear completely after the forward-looking intention has been formulated. For example, I  may form a forward-looking intention to write a letter of complaint about my boss the day after, but still spend part of the night obsessively fantasizing about slapping the boss in front of everyone. As a result, in planned emotional actions, when the forward-looking intention finally causes an intention-in-action, an action tendency with some degree of control precedence—perhaps involving only mental actions and little to no bodily preparation—should still be up and running.22 Emotional planned actions are not to be confused with actions done out of expected emotions.23 Examples include ingesting a beta blocker prior to a public speech out of the expectation that one will be afraid without it, marrying a high school sweetheart out of the expectation that one will be happy ever after, turning down an opportunity to cheat on one’s wife with an attractive woman out of the expectation that one will later on experience guilt. Actions done out of expected emotions rely on our ability to use predicted future emotions as input in a standard process of practical reasoning.

22 This is compatible with the fact that the prioritized action tendency may not be up and running at various points during the period between the emotional incident and the emotional action (e.g., during sleep). A borderline case would be one in which I rely on a forward-looking intention formed in the heat of the moment and such intention shares the relational goal of an earlier prioritized action tendency, but the prioritized action tendency has dissipated entirely by the time the forward-looking intention causes an action. I remain neutral in this chapter on whether such planned actions should also be called emotional (Pacherie 2001 suggests that they should). 23 Actions done out of expected emotions are a species of what Pacherie (2001) calls fully deliberate emotional actions. Other species of such genus include using memory and the imagination to elicit emotions (e.g., remembering past slights in order to become angry) and using expected emotions as a commitment device (e.g., announcing one will quit smoking in order to use the expected guilt that will result from infringement to strengthen one’s commitment). I do not consider the use of memories of past emotions, imaginings of future emotions, or expectations of future emotions to constitute a species of emotional actions because such uses do not constitute by themselves actions with control precedence, so I reject Pacherie’s (2001) suggestion that there exist fully deliberate emotional actions.

176  Andrea Scarantino My view is that actions done out of expected emotions are not emotional actions, even though they presuppose the ability to have emotions and to predict future emotions. In ordinary English, the distinction between emotional actions (impulsive or planned) and non-emotional actions done out of expected emotions is often blurred. Some may describe the action of taking a beta blocker as being done out of fear (rather than out of expected fear), and the action of refraining from cheating as being done out of guilt (rather than out of expected guilt), even though in neither case an actual emotion has taken place. The final challenge is that of explaining reflex-like emotional actions such as automatically recoiling out of fear of a suddenly looming object or automatically expelling a cockroach from one’s mouth out of disgust. In this case, the two-level control structure I have described does not kick in. This is because there is no formation of an action tendency, and action follows directly the perception of the stimulus, without any input from consistency control and executive control. Frijda’s (2007, 31) view is that reflexes are “emotional” in an inverted commas sense at best, but I disagree. This is because, as Frijda himself pointed out, reflexes exhibit the highest degree of control precedence: Reflex-like emotional actions take priority over all other competing processes and involve bodily preparation synchronous with motor execution. In this sense, they are relevantly similar to prioritized action and inaction tendencies. At the same time, there are many reflexes that have nothing to do with emotions. The knee reflex and the blinking reflex are good examples of non-emotional reflexes. I propose that we distinguish reflex actions that are emotional from reflex actions that are not emotional by focusing on their relational goals: A reflex action is an emotional action just in case it shares the relational goal of a prioritized action tendency. In other words, for reflex actions to count as emotional, there must exist impulsive emotional actions with the same relational goal. This is why the knee reflex and the blinking reflex are not emotions. There is no prioritized action tendency that shares the goal of either reflex. On the other hand, the avoidance reflex activated by suddenly looming objects is an emotional action because it shares with prioritized avoidance tendencies the relational goal of achieving safety. It is time to take stock. I have argued that the Motivational Theory of Emotions (MTE) outperforms Cognitivism and the Perceptualism when it comes to accounting for the phenomena of emotional motivation. This, however, is not sufficient to consider it a viable philosophical theory of emotions. MTE must also provide solutions for two standard problems in the philosophy of emotions. One is the problem of intentionality, which requires explaining why emotions are about their core relational themes (e.g., why is fear about dangers?). The other is the problem of differentiation, which requires explaining how emotions differ from one another (e.g. how is guilt different from shame?). In what follows, I will illustrate how MTE deals with these two problems.

The Motivational Theory of Emotions  177

5.4.  The Intentionality of Emotions Suppose that fear is a prioritized avoidance tendency with the relational goal of achieving safety, as suggested by Frijda (1986). This fact per se does not explain why fear is about dangers. Similarly, the fact that anger is a prioritized attack tendency with the relational goal of removing an obstacle does not explain in what sense anger is about offenses. In order to become a workable philosophical theory of emotions, MTE must explain the intentionality of emotions, also known as their ability to represent. Cognitivists and Perceptualists follow a similar strategy to account for it. First, they assume that emotions play two representational roles: (a) A descriptive role: Emotions represent what obtains. (b) An imperative role: Emotions represent what needs to obtain. Second, they assume that these two roles are played by distinct components of emotions. For B&D Cognitivists, the two components are beliefs and desires, which play respectively the descriptive and the imperative role. On this proposal, what explains the intentionality of emotions is that they are partially constituted by beliefs to the effect that the core relational theme (CRT) of the emotion is instantiated. On this view, fear is about dangers and anger is about offences because they are partially constituted by, respectively, the CRT belief that danger is at hand and the CRT belief that one has been offended. For Prinz (2004), the two components are embodied appraisals and valence markers, which play respectively the descriptive role and the imperative role. On his proposal, what explains the intentionality of emotions is that they are partially constituted by embodied appraisals that have the function of being elicited by core relational themes. On this view, fear is about dangers and anger is about offenses because they each contain embodied appraisals with the function of correlating with, respectively, dangers and offenses. I share with B&D Cognitivists and perceptual theorists like Prinz the view that emotions play both a descriptive and an imperative role. Since Frijda’s (1986; 2007) original theory only accounted for the imperative side, it needs to be supplemented with a descriptive side. What I reject is the further assumption that descriptive and imperative roles are played by distinct components of emotions. I propose instead that emotions combine descriptive and imperative roles into a unified whole. On the view I  propose, emotions have what I  have previously called a mind-to-world-to-mind (or dual) direction of fit: they represent how things are (mind-to-world) and how things are to be (world-to-mind) at the same time (Scarantino 2010). Millikan (2004) has labeled representations of this sort as pushmi-pullyu representations, arguing that they “represent facts and . . . represent goals, both at once” (157). What we need to do is to make explicit that (in)action tendencies and action reflexes do not merely represent relational goals: They also represent facts. For the purposes

178  Andrea Scarantino of this chapter, I accept Prinz’s (2004) teleosemantic framework for explaining how emotions represent facts, namely that they do so by having the function of being elicited by them. But I supplement it with the idea that emotions have at the same time the function of achieving a certain relational goal. Thus, emotions have what I call an informational-cum-motivational function. According to MTE, what explains the intentionality of emotions is that they are (in) action tendencies or action reflexes with the informational-cum-motivational function of achieving relational goals while correlating with core relational themes. On this view, fear is about dangers because it is a prioritized avoidance tendency/reflex with the informational-cum-motivational function of achieving the relational goal of one’s own safety while correlating with dangers. Anger is about offenses because it is a prioritized attack tendency/reflex with the informational-cum-motivational function of achieving the relational goal of removing an obstacle while correlating with offences. Sadness is about losses because it is a prioritized inaction tendency with the informational-cum-motivational function of achieving the relational goal of not relating as such while correlating with losses. I understand informational-cum-motivational functions in the etiological sense, namely as past effects that explain the current presence of the function bearer (cf. Wright 1973; Wouters 2003). Saying that emotion E has the informational-cum-motivational function of achieving relational goal G while correlating with core relational theme T is therefore saying that the past effects of achieving G in the presence of T explain why E was selected for, relative to some selection process S (e.g., natural selection, cultural selection). For instance, saying that fear has the informational-cum-motivational function of achieving the relational goal of one’s own safety while correlating with dangers amounts to saying that the past effects of attaining one’s own safety in the presence of danger explain why fear was selected for. Functions understood etiologically can have normative import, because they characterize what a function-bearer is supposed to do. A corollary is that fear will be defective in the absence of danger, anger will be defective in the absence of offences, sadness will be defective in the absence of losses. In all such cases, emotions do not prioritize relational goals in the presence of those core relational themes that explain why prioritizing such goals in the past was selected for.

5.5.  What Emotions Are and How They Differ from One Another I am finally in a position to formulate the definition at the heart of the Motivational Theory of Emotions: (MTE): An emotion is a prioritizing action control system, expressed either by (in)action tendencies with control precedence or by action reflexes, with the function of achieving a certain relational goal while correlating with a certain core relational theme.

The flowchart (Figure 8.1) summarizes the causal sequence of events posited by MTE (the chart focuses on action rather than inaction tendencies). Emotional episodes start

The Motivational Theory of Emotions  179 Physical or mental event

Appraisal

Activation of prioritizing action control system

Emotional reflex

Prioritized action tendency

Impulsive emotional action

Planned emotional action

Compatibility & executive control

Extinction

Figure 8.1 

with an initial physical event (e.g., being insulted) or mental event (e.g., recollecting being insulted) whose appraisal activates the action control system with which I identify the emotion (e.g., the anger system). The action control system in turn can either directly lead to a reflex action or to an action tendency with control precedence (e.g., an attack tendency), i.e. one that takes precedence over all other action modalities and involves bodily preparation. As soon as the prioritized action tendency is up and running, rational control enters the picture, performing compatibility and executive controls. The outcome of such controls can be an impulsive emotional action (e.g., striking), the formation of a forward-looking intention that will eventually lead to a planned emotional action (e.g., writing a letter of complaint), or the extinction of the prioritized action tendency. MTE adds three main ingredients to Frijda’s (1986; 2007) original analysis. First, action reflexes are explicitly included in the domain of emotional actions proper along with prioritized action and inaction tendencies. Second, the identity of an emotion becomes defined in part by its function. Third, the function of emotions is understood in informational-cum-motivational terms, i.e. achieving a certain relational goal in the presence of a certain core relational theme. Before explaining how MTE distinguishes between different emotions, it is worth emphasizing that appraisals and feelings no longer play a definitional role in MTE.

180  Andrea Scarantino This marks a major discontinuity in the philosophy of emotions, because it shifts the focus of analysis from how emotions evaluate (Cognitivism) and how emotions feel (Perceptualism) to how emotions affect action. It would be a mistake, however, to conclude that appraisals and feelings have no important role to play in MTE. Appraisals are the standard elicitors of prioritized (in)action tendencies and action reflexes. This is because stimuli do not cause emotions per se, but contingently on how they are appraised. I have argued elsewhere that emotional appraisals lie on a continuum (Scarantino 2010). On one end, we have forms of appraisal that operate as a modular input system (e.g., fast, effortless, mandatory, informationally encapsulated, with limited central access and fixed neural architecture). An example is the appraisal of threat involved in reflex fear, which is automatically elicited by suddenly looming objects and loud noises. On the other hand, we have forms of appraisal that are centrally driven (e.g., slow, effortful, non-mandatory, cognitively penetrable, with full central access, without fixed neural architecture). An example is the appraisal of threat performed by an airplane pilot who gradually, on the basis of various subtle cues, realizes that there is something seriously wrong with the plane, which elicits a prioritized avoidance tendency. Feelings are also often involved in the activation of the action control systems with which I identify emotions. What is felt according to MTE is that a certain (in)action tendency has acquired control precedence.24 This complex feeling results from a variety of co-occurring subjective experiences, some associated with the interruption of ongoing actions, some associated with the focusing of attention, some associated with the occurrence of facial expressions, some associated with the occurrence of autonomic bodily changes, and so on. A basic difference between MTE and traditional theories of emotions is that appraisals and feelings are no longer essential components of emotions. First, exceptions to the rule that prioritized (in)action tendencies and action reflexes are caused by appraisals are allowed. Non-standard mechanisms for the elicitation of prioritized (in) action tendencies and reflexes include chemical induction, brain manipulation, and facial feedback (Izard 1993). Second, prioritized (in)action tendencies and action reflexes can occur in the absence of concurrent feelings. Candidate examples include unconscious prioritized (in) action tendencies in humans and prioritized (in)action tendencies in simple animals like sea slugs whose ability to have subjective experiences in the first place is highly doubtful. We now have all the ingredients required to explain how MTE differentiates between different emotions. It does so by means of three elements: (i) emotion-specific (in)action tendencies or action reflexes, (ii) emotion-specific core relational themes,

24 Reflexes are too quick for feelings to play a role in their operation, even though feelings often follow emotional action reflexes, as when one starts feeling fear several seconds after a car accident.

The Motivational Theory of Emotions  181 Table 8.1  Emotion

(In)action tendency/ action reflex

Relational Goal

Core Relational Theme

Anger Fear Sadness

Removal of obstruction One’s own safety Not relating as such

Offense Danger Loss

Joy Disgust Guilt

Attacking Avoiding Undifferentiated disengagement Open engagement Expelling Repairing relationship

Positive Event Contamination Moral transgression

Shame

Disappearing

Relating as such Removal of object Making up for a flawed behavior Hiding a flawed self

Failure to live up to an ego ideal

and (iii) emotion-specific relational goals. Table 8.1 offers some tentative examples for specific emotions.25 Let us consider the distinction between shame and guilt by way of example (cf. Lewis 2008). According to MTE, shame is a prioritized disappearance tendency/reflex with the function of achieving the relational goal of hiding a flawed self while correlating with failures to live up to an ego ideal. Guilt is a prioritized reparation tendency/reflex with the informational-cum-motivational function of achieving the relational goal of making up for a flawed behavior while correlating with moral transgressions. This account of their difference is compatible with guilt and shame being indistinguishable at the level of bodily changes, or lacking bodily changes entirely, and being indistinguishable at the level of feelings, or lacking feelings entirely. For example, it may well be that some forms of guilt and shame only consist of mental actions without distinctive bodily signatures, and that some forms of guilt and shame are associated with no subjective experiences at all. But as long as they are associated with prioritized action tendencies that differ in the way I have indicated, the ground of difference between them is not threatened. Furthermore, this account of the difference between guilt and shame is compatible with significant variation in the complexity of the information processing involved in the appraisals that elicit guilt and shame. Some varieties of guilt and shame (e.g., victim’s guilt, shame generated by the presence of high ranking members of one’s group) will be elicited by appraisals that display several of the markers of a modular input

25 The chart combines Lazarus’ (1991) account of core relational themes (with the addition of contamination as the core relational theme for disgust; see Curtis, DeBarra, and Aunger 2011) with suggestions by several authors on the action tendencies and relational goals of specific emotions (e.g., Frijda 1986 on anger, fear, and disgust; Lewis 2008 on shame and guilt; Bonanno 2001 on sadness; Fredrickson and Cohn 2008 on joy).

182  Andrea Scarantino system (e.g., fast, informationally encapsulated, effortless), whereas other varieties of guilt and shame (e.g., guilt about one’s financial success, shame about having been caught violating a promise) will be elicited by appraisals that are centrally driven (e.g. slow, informationally penetrable, effortful). The ground of difference between guilt and shame I have outlined is even compatible with the possibility that guilt and shame may occasionally not involve appraisals at all, as in the (so far) theoretical possibilities of guilt and shame being elicited by direct brain stimulation, chemical induction, and facial feedback. Finally, MTE offers a recipe for distinguishing between theoretically more homogeneous sub-types of the same folk psychological emotion category. The recipe is to ask what states of affairs emotions have the function of correlating with and what relationship with the environment they have the function of bringing about. In principle, this process can lead to the realization that different instances of the same folk psychological emotion have the function of correlating with different states of affairs and the function of bringing about different relational goals, potentially leading to splitting the original folk category into sub-types. Disgust is a good candidate for this sort of splitting. Some instances of disgust appear to have the function of producing an action tendency/reflex of oral expulsion while correlating with physically contaminating items such as poisons and parasites, whereas other instances of it appear to have the function of producing an action tendency/reflex of social rejection while correlating with morally polluting items such as racism and hypocrisy (Rozin, Haidt, and McCauley 2008; Kelly 2011). In such a case, it may be theoretically useful to distinguish between two sub-types of disgust—e.g., core disgust and moral disgust—each defined by its own triad of fine-grained action tendencies, core relational themes, and relational goals.

6. Conclusion In this chapter I have introduced a new theory of emotions—the Motivational Theory of Emotions (MTE). On the view I propose, emotion types are defined and differentiated by their distinctive (in)action tendencies/action reflexes, their distinctive relational goals, and their distinctive core relational themes, on the assumption that the informational-cum-motivational function of each emotion is to achieve the relational goal when the core relational theme is instantiated. MTE explains what I have called the phenomena of emotional motivation—impulsivity, flexibility, and bodily underpinnings—better than the two currently most popular philosophical theories of emotions, namely Cognitivism and Perceptualism. MTE also accounts for the variety of emotional actions associated with emotions, of which I have distinguished three species: reflex-like emotional actions, impulsive emotional actions, and planned emotional actions.

The Motivational Theory of Emotions  183 Finally, MTE has solutions to offer to the problems of intentionality and differentiation, explaining, respectively, what endows emotions with aboutness and what makes any two emotions different from one another. In light of these achievements, I conclude that the Motivational Theory should be considered a promising new competitor in the philosophy of emotions.

Acknowledgments I want to especially thank Dan Jacobson and Justin D’Arms for their generous support throughout the development of some of the central ideas of this chapter. Their meticulous and insightful comments on a previous draft have had a major influence on the current structure of the chapter. I want to acknowledge the importance of a one-month residency I spent at the University of Michigan, Ann Arbor, discussing emotions with Dan and Justin. If only philosophy could always be this much fun! I also benefited from a Workshop on Moral Psychology and Human Agency held in June 2012 at the University of Michigan, Ann Arbor. Both my residency and the workshop were sponsored by the “Science of Ethics” Templeton Foundation grant. I am grateful to Justin D’Arms for inviting me to present a version of this chapter to the participants to his seminar on emotions at Ohio State University in Fall 2012. Their comments were very helpful. Finally, I want to thank my colleagues Eddy Nahmias and Neil van Leuween, as well as the students of my seminar on emotions in Spring 2013 at Georgia State University for their valuable feedback on a previous draft. The errors, of course, remain all mine.

References Arnold, M. B. 1960. Emotion and Personality, vol. 1. New York: Columbia University Press. Arnold, M. B. (ed.). 1970. Feelings & Emotions. San Diego, CA: Academic Press. Barrett, L. F. 2006. Are emotions natural kinds? Perspectives on Psychological Science, 1, 28–58. Benbaji, H. 2012. How is Recalcitrant Emotion Possible? Australasian Journal of Philosophy. doi: 10.1080/00048402.2012.699078 Bonanno, G.  A. 2001. Grief and emotion:  A  socio-functional perspective. In M. Stoebe, R. O. Hanson, W. Stroebe, and H. Shut (eds.), Handbook of Bereavement Research: Consequences, Coping, and Care, pp. 493–516. Washington, DC: American Psychological Association. Bratman, M.  E. 1987. Intention, Plans, and Practical Reason. Cambridge, MA:  Cambridge University Press. Curtis, V., DeBarra, M., and Aunger, R. 2011. Disgust as an adaptive system for disease avoidance behaviour. Philosophical Transactions of the Royal Society B: Biological Sciences, 366, 389–401. Damasio, A. 1994. Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Grossett/ Putnam. Damasio A. R. 2003. Looking for Spinoza: Joy, Sorrow, and the Feeling Brain. New York: Harcourt.

184  Andrea Scarantino D’Arms, J., and Jacobson, D. 2003. The significance of recalcitrant emotions (or anti-quasijudgmentalism). In A. Hatzimoysis (ed.), Philosophy and the Emotions. Cambridge: Cambridge University Press. Dewey, J. 1895. The theory of emotion. (2) The significance of emotions. Psychological Review, 2, 13–32. Döring, S. 2003. Explaining action by emotion. Philosophical Quarterly, 211, 214–30. Dretske, F. 1988. Explaining Behavior. Cambridge, MA: MIT Press. Dunn, B. D., Dangleish, T., and Lawrence, A. D. 2006. The somatic marker hypothesis: A critical evaluation. Neuroscience and Biobehavioral Reviews, 30, 239–71. Ekman, P. 1999. Basic emotions. In T. Dalgleish and M. Power (eds.), Handbook of Cognition and Emotion, pp. 45–60. Chichester: John Wiley and Sons Co. Elster, J. 2010. Emotional choice and rational choice. In P. Goldie (ed.), Oxford Handbook of the Philosophy of Emotion, pp. 263–281. Oxford: Oxford University Press. Fredrickson, B. L., and Cohn, M. A. 2008. Positive emotions. In M. Lewis, J. M. Haviland-Jones, and L. F. Barrett (eds.), Handbook of Emotions, 3rd edn., pp. 777–96, New York: Guilford Press. Frijda, N. H. 1986. The Emotions. Cambridge: Cambridge University Press. Frijda, N. H. 2007. The Laws of Emotion. Mahwah, NJ: Erlbaum. Frijda, N. H. 2008. The psychologists’ point of view. In M. Lewis, J. M. Haviland-Jones, and L. F. Barrett (eds.), Handbook of Emotions, 3rd edn., pp. 68–87, New York: Guilford Press. Frijda, N. H, 2010. Impulsive action and motivation. Biological Psychology, 84, 570–9. Gallistel, C.  R. 1980. The Organization of Action:  A  New Synthesis. Hillsdale, NJ:  Lawrence Erlbaum Associates, Inc. Goldie, P. 2000. The Emotions: A Philosophical Exploration. Oxford: Clarendon Press. Green, O.  H. 1992. The Emotions:  A  Philosophical Theory. Dordrecht:  Kluwer Academic Publishers. Hursthouse, R. 1991. Arational actions. The Journal of Philosophy, 88(2), 57–68. Izard, C. E. 1993. Four systems for emotion activation: Cognitive and noncognitive processes. Psychological Review, 100, 68–90. James, W. 1884. What is an emotion? Mind, 9, 188–205. James, W. 1890. The Principles of Psychology. New York: Holt. Kelly, D. 2011. Yuck! The Nature and Moral Significance of Disgust. Cambridge, MA: The MIT Press. Kovach, A., and De Lancey, C. 2005. On emotions and the explanation of behavior. Nous, 39(1), 106–22. Lazarus, R. S. 1991. Emotion and Adaptation. New York: Oxford University Press. Lewis, M. 2008. Self-conscious emotions: Embarrassment, pride, shame, and guilt. In M. Lewis, J. Haviland-Jones, and L. Feldman Barrett (eds.), Handbook of Emotions, 3rd edn., pp. 742–56. New York: Guilford Press. Loewenstein, G., and Lerner, J. 2003. The role of affect in decision-making. In R. J. Davidson (ed.), Handbook of Affective Sciences, pp. 619–42. New York: Oxford University Press. Lyons, W. 1980. Emotion. Cambridge: Cambridge University Press. Marks, J. 1982. A theory of emotions. Philosophical Studies, 42, 227–42. Millikan, R. G. 2004. Varieties of Meaning: The 2002 Jean Nicod Lectures. Cambridge, MA: MIT Press. Nash, R. 1989. Cognitive theories of emotion. Nous, 23, 481–504.

The Motivational Theory of Emotions  185 Nussbaum, M. 2001. Upheavals of Thought: The Intelligence of Emotions. Cambridge: Cambridge University Press. Pacherie, E., 2001. The role of emotions in the explanation of action. European Review of Philosophy, 5, 53–91. Pacherie, E., 2008. The phenomenology of action:  A  conceptual framework. Cognition, 107, 179–217. Price, C. 2006. Fearing Fluffy: The content of an emotional appraisal. In Graham MacDonald and David Papineau (eds.), Teleosemantics. Oxford: Oxford University Press, 208–28. Prinz, J. 2004. Gut Reactions: A Perceptual Theory of Emotion. Oxford: Oxford University Press. Reimann, M., and Bechara, A. 2010. The somatic marker framework as a neurological theory of decision-making: Review, conceptual comparisons, and future neuroeconomics research. Journal of Economic Psychology, 31, 767–76. Reisenzein, R. In press. What is an emotion in the belief-desire theory of emotion? In F. Paglieri, L. Tummolini, R. Falcone, and M. Miceli (eds.), The Goals of Cognition: Essays in Honor of Cristiano Castelfranchi. London: College Publications. Rozin, P., Haidt, J., and McCauley, C. R. 2008. Disgust. In M. Lewis, J. M. Haviland –Jones, and L. F. Barrett (eds.), Handbook of Emotions, 3rd edn., pp. 757–76. New York: Guilford Press. Ryle, G. 1949. The Concept of Mind. London: Hutchinson. Scarantino, A. 2010. Insights and blindspots of the cognitivist theory of emotions. British Journal for the Philosophy of Science, 61(4), 729–68. Scarantino, A. 2012. How to define emotions scientifically. Emotion Review, 4(4), 358–68. Scarantino, A. Forthcoming. Basic emotions, psychological construction and the problem of variability. In J. Russell and L. Barrett (eds.), The Psychological Construction of Emotion. New York: Guilford Press. Scarantino, A., and Nielsen, M. (MS). Kicking doors and piercing dolls: Why emotions make us do foolish things. Searle, J. 1983. Intentionality. Cambridge: Cambridge University Press. Skinner, B. F. 1953. Science and Human Behavior. New York: Macmillan. Smith, M. 2010. Humeanism about motivation. In T. O’Connor and C. Sandis (eds.), A Companion to the Philosophy of Action, pp. 153–8. Oxford: Wiley-Blackwell. Solomon, R. 2003. Not Passion’s Slave. Oxford: Oxford University Press. Watson, J. B. 1919. Psychology from the Standpoint of a Behaviorist. Philadelphia: Lippincott. Winkielman, P., and Berridge, K.  C. 2004. Unconscious emotion. Current Directions in Psychological Science, 13, 120–3. Wouters, A. G. 2003. Four notions of biological function. Studies in History and Philosophy of Biological and Biomedical Sciences, 34(4), 633–68. Wright, L. 1973. Functions. Philosophical Review, 82, 139–68. Zhu, J., and Thagard, P. 2002. Emotion and action. Philosophical Psychology, 15, 19–36.

9 The Reward Theory of Desire in Moral Psychology Timothy Schroeder and Nomy Arpaly

In this chapter, we describe and defend the reward theory of desire, and then explore some of its consequences for moral psychology.1 At least since Plato, moral psychologists have worked hard to understand the nature of desire, and to incorporate desires into their larger moral-psychological frameworks. Naturally, one’s theory of desire makes a difference to what it is plausible to hold about other phenomena. According to the widely held motivational theory of desire, to desire that p is to tend to be motivated to bring it about that q, when one believes that q will make it more likely (or will bring it about) that p.2 The motivational theory is clean and clear, and captures a good deal of what is commonly thought about desires. It also raises a large number of worries for anyone who thinks that desire has an important role in moral psychology. According to the motivational theory, there is nothing more to desire (in its essence) than a sophisticated, cognitively sensitive behavioral disposition. Behavioral dispositions, however, do not appear to be adequate explainers of things desire has been said to explain: love and care, having reasons for action, or the differences between acting on an appetite and acting out of a sense of duty (to name just three worries). That is, the motivational theory appears to be too clean, too antiseptic, even, to allow desires to explain important moral-psychological phenomena. The moral psychologist who thinks desires deserve a central role in her moral psychology has several options. She may defend the clean and clear motivational theory against its critics. She may keep the motivational theory but hold that desires only 1 The first reward theory of desire to our knowledge is found in Dretske (1988, ch. 5). For a full treatment of the idea see Schroeder (2004). Objections to Schroeder are articulated in Brook (2006), De Sousa (2006), Latham (2006), Thagard (2006), with a response by Schroeder as well (Schroeder 2006). A similar approach to the present one is taken, with different results, in Morillo (1990). We have also benefited from Peter Railton, who is working on related ideas. 2 See, for example, Smith (1994).

The Reward Theory of Desire in Moral Psychology  187 explain love, generate reasons, and so on when they also have further, supplementary features. And she may reject the motivational theory in favor of some other theory that seems to do better with the problem cases. In this chapter, we suggest that the reward theory of desire offers a promising way of pursuing the third strategy. Rejecting the motivational theory of desire and replacing it with the reward theory opens many new avenues for the philosopher who sees desires having a central role in theories of love and care, of acting for reasons generally, of acting for moral reasons specifically, and so on. The reward theory of desire does not make any of the most familiar effects of desires into the essence of desire. Desires—or, to be precise, intrinsic desires, which are the only desires we will discuss—have various effects. They have motivation effects, as well as emotional effects (effects on when pleasure and displeasure are felt) and cognitive effects (for example, effects on where attention is directed). All of these effects, on the reward theory, remain common but inessential causal consequences of having desires. In human beings temporarily subjected to powerful non-rational influences on their thoughts, feelings, and actions, intrinsic desires can exist while failing to cause their familiar effects on motivation or feelings or thoughts, or while causing excessively strong or weak forms of their familiar effects. People who are very sleepy, over-caffeinated, deprived of the lithium they rely on, and so on are people whose motivational, emotional, and cognitive states can substantially vary from what would be expected, given what they intrinsically desire, though what they intrinsically desire and how much they desire it remains unchanged. And in unhealthy human beings, and perhaps aliens, these gaps between intrinsic desires and their usual effects can be chronic or permanent. An addict who greatly intrinsically desires things that can only be gained by ceasing to use cocaine, and only moderately intrinsically desires things that can be gained by continuing to use cocaine, can nonetheless live for years with a tremendous number of the motivational, emotional, and cognitive signs of someone with the opposite priorities. An alien who greatly intrinsically desires justice might lack the capacity to be motivated at all, or to feel anything at all, or both, without desiring justice any less. The theory that makes the foregoing true takes a fair bit of explaining. The way in which the reward theory of desire entails the result that desire’s most familiar motivational, emotional, and cognitive signs are the causal consequences of possessing intrinsic desires is through holding that the essence of desire is something unfamiliar, and through holding that this unfamiliar essence nonetheless counts as the essence of desire because desires form a natural kind. Just as H2O is the unfamiliar essence of water, so (we will argue) states of the reward system are the unfamiliar essence of desire. Explaining this unfamiliar essence of desire is the first job of this chapter. We will then proceed to show that this unfamiliar putative essence of desire plays the causal roles that we assume desires play: that desires as theorized by the reward theory of desire turn out to play the roles that desires are thought to play as commonly understood. Next we go on to argue that it is at least reasonable to think of desires as forming a

188  Timothy Schroeder and Nomy Arpaly natural kind, and so it is at least reasonable to think that desires are something distinct from the effects (on motivations, feelings, and thoughts) by which we know them from the armchair. We finish by suggesting ways in which the reward theory of desire might ameliorate a range of problems facing desire-centered moral psychology.

1.  The Reward and Punishment Systems The reward system is an unconscious part of the mind that is, in virtue of its particular states, a normal cause of overt and covert action, of positive and negative feelings, and of the cognitive effects most associated with intrinsic desires. The reward system is also the only thing that is a common cause of all of these effects (indeed, it is the only thing that is even close to a common cause of all these effects). And the reward system is a psychological natural kind.3 The reward system’s configurations are thus a reasonable starting point for a theory of intrinsic desire. The reward system constitutes things as rewards, and we will explain in some detail what reward is in this context, in just a moment. We will also explain the opposite of reward, namely punishment. While scientists do not understand punishment as well as reward, a wealth of indirect evidence exists that the brain also has a punishment system that mirrors the operations of the reward system, just as mathematical models of reward and punishment would predict. Hence we will talk about the reward and punishment systems, though most of our evidence will be drawn from what is known about the reward system. Focusing on the reward and punishment systems could give rise to a number of different theories of intrinsic desire. We hold the following. Reward Theory of Desire: To have an intrinsic desire regarding it being the case that p is to constitute p as a reward or a punishment.

Intrinsic desires come in two sub-types: appetites and aversions. To have an intrinsic appetitive desire that p is to constitute p as a reward. To have an intrinsic aversion to p is to constitute p as a punishment.4

To constitute a (conceivable) state of affairs p as a reward is to make it the case that p counts as a reward, in the relevant sense, and similarly for punishments. And the distinction between appetitive desires and aversions is, roughly, the familiar distinction between desiring that things be the case because we want them for their own sakes, and

3 At least, it is by the reasonable test that it is a kind that scientists find useful, at the relevant level of investigation, for correct explanation, prediction, and control of psychological phenomena. The reward system is implemented by a biological natural kind: the brain’s dopamine system. But it is still possible to distinguish the psychological natural kind (a specific sort of learning system) from the biological natural kind (the specific groups of cells and neuromodulators that implement this learning in animals like ourselves). 4 This theory is fully articulated in Schroeder (2004, ch. 5). For more on the distinction between appetites and aversions, see Schroeder (2004, chs. 1, 5).

The Reward Theory of Desire in Moral Psychology  189 desiring that things not be the case because we have an aversion to them in themselves.5 Thus, understanding the reward theory of desire is little more than understanding the relevant senses of “reward” and “punishment.” There are two sensible ways to understand each of “reward” and “punishment” in the reward theory of desire. One is the ordinary way of understanding “reward” and “punishment,” according to which a child can be rewarded for sitting still at the dinner table by being granted permission to stay up late, and punished for squirming by being sent to bed early. The other is a technical way of understanding “reward” and “punishment” which is used in certain mathematical theories of learning and in certain parts of neuroscience.6 Surprising as it might seem, the reward theory of desire is true on both understandings. However, if one is seeking a theory of desire that explains what is less basic in terms of what is more basic, then “reward” and “punishment” need to be understood in the technical, mathematical, scientific sense. This is because a full explanation of how being granted the right to stay up late can be a reward in the ordinary sense of “reward” will ultimately appeal to the technical sense of “reward.”7 To understand the technical senses of “reward” and “punishment” it is best to start with the related notion of reward learning.8 Reward learning is most famously exemplified by the operant conditioning of behavior in experimental animals: by rats trained to run mazes for cheese, essentially. The association of reward learning with rats and mazes is understandable but unfortunate. It is understandable because behaviorists were the first scientists to discover reward learning and give it a theoretical treatment.9 It is unfortunate because any form of learning associated with behaviorism is likely to be tainted by the association, stigmatized as a sort of learning that, while it might explain something about rats in mazes, is not the sort of learning of interest to someone who wishes to understand the complexities of the human mind. As section 2 will show, however, the stigma is misplaced. The reward learning system is in fact the cause of many phenomena that are absolutely central to understanding the complexities of the human mind, in both some of the ways it is like the minds of other mammals and in some of the ways in which it is distinctive. So what is reward learning? Human beings learn in many ways. We learn by listening, by observing others, by trying it out for ourselves, by repeatedly practicing until awkward actions become smooth, precise, and efficient. Different sorts of learning are

5 The evidence that this distinction is significant, and not merely terminological, is discussed in Schroeder (2004, chs. 1, 5). 6 This technical sense is related to, but distinct from, the technical sense of “reward” long used in the theory of operant conditioning. The differences will become clear as this particular technical sense is further characterized. 7 This is fully defended in Schroeder (2004, ch. 2). 8 Here we will use the phrase “reward learning” to encompass what is sometimes also called “contingency-based learning:” the sort of learning that takes into account both rewards and punishments as initiators of learning. This is an extension of the most common use of “reward learning,” which excludes consideration of punishment. 9 The investigation of operant conditioning begins with B. F. Skinner. See Skinner (1938).

190  Timothy Schroeder and Nomy Arpaly made possible by different mechanisms. There is one mechanism that seems to underlie learning by remembering and recalling events (such as who said what at the last departmental meeting),10 there is a different mechanism that seems to underlie learning to perform complex sequences of actions smoothly and fluently (as when riding a bicycle),11 and so on. Reward learning is another sort of a learning, distinct from these, and implemented by yet another mechanism. Reward learning is a form of learning that has an opportunity to take place when one mental event participates in causing another. Such causings happen all the time: Groups of edge perceptions participate in causing perceptions of squareness, thoughts that a claim is universal in scope participate in causing thoughts that surely there are exceptions, perceptions of the approach of a dance partner participate in causing efforts to take matching steps backward, and so on.12 So there are hundreds, if not thousands, of opportunities at each moment for reward learning to have its effects. In reward learning, there is the causing of one mental state by another, and then that causal sequence is followed by the unconscious release of a signal in the brain, a signal that takes one of three forms. One form causes the disposition of the first mental state to produce the second to increase (all else being equal). This sort of signal can be called a positive learning signal, though there is nothing positive about it beyond the fact that it increases the strengths of the relevant disposition. The second form of the signal causes no change in dispositions; this can be called a neutral learning signal. And the third form of the signal causes the disposition of the first mental state to produce the second to decrease (all else being equal). This sort can be called a negative learning signal. Imagine Juan is trying to keep up with a more talented dance partner. He perceives the approach of his dance partner, and this perception participates in causing an effort to take a matching step backward.13 If this causal sequence is followed by a positive learning signal, then the ability of that sort of perceptual event to cause that sort of matching step backward will be strengthened. The probability of Juan making that sort of step backward in that sort of situation in the future has just increased. On the other hand, if the causal sequence is followed by a negative learning signal, then the ability of perceptions of advancing dance partners to cause Juan to take matching steps backward will be weakened. The probability that such a step will be taken in that sort 10 Episodic learning, made possible (it seems) by a process called “long-term potentiation.” See, e.g., Shastri (2002). 11 Theories of skill-learning based on back-propagation of error and recurrent decorrelation are compared in Porrill, Dean, and Stone (2004), for example. 12 Of course, there is dispute within the philosophy of mind about mental causation. The details of this debate are not particularly important for our purposes. Since it is clear that reward learning takes place in some sense that is suitable to scientific purposes, the precise metaphysics of how it takes place can be left to other philosophers without worrying too much about the details. 13 This description of things, with a perception causing an effort, is not meant to suggest that Juan does not participate “as an agent” in his movement. Even when Juan takes a step backward while acting as a full-blooded agent, there is still a causal chain leading to the final mental event that causes the movement of the leg.

The Reward Theory of Desire in Moral Psychology  191 of situation in the future has just decreased. And if it is followed by a neutral learning signal then nothing within Juan changes (at least, not through this mechanism). If a positive learning signal appears every time Juan reacts to his partner’s approach with a matching step backwards then Juan will, in one ordinary sense of “learn,” learn to react to his partner’s approach with a matching step backwards. Learning in this sense is not analogous to learning how to ride a bicycle or to paint but to the sort of learning we refer to when we say, “she learned to speak more in class,” “she learned to taste the ‘chocolate’ notes in red wines,” or “by the time a hundred people had asked him whether as an Irishman he liked Enya’s music, he had learned to hate Enya’s music.” The obvious question at this point is: What determines whether a positive, neutral, or negative learning signal will be released in a creature capable of reward learning? There are two main parts to the answer. The first part is that the signal coming from the reward learning system is influenced by how the organism perceptually and cognitively represents the world to be. The representation of certain states of affairs causally affects the reward system such that it directly increases the chance that a positive learning signal will be released, and this is what makes these represented states of affairs rewards (technically, “positive rewards”).14 That is, this is what makes them rewards in the technical sense relevant to this form of learning. To other represented states of affairs, the reward system does not directly react, and these are neither rewards nor punishments. And a failure to represent a reward increases the chance of generating a negative learning signal. This is what makes the failure of that state of affairs to obtain a punishment (technically, a “negative punishment”). What holds for rewards holds in parallel for punishments. The representation of certain states of affairs causally affects the punishment system such that it directly increases the chance that a negative learning signal will be released, and this is what makes these represented states of affairs punishments (technically, “positive punishments”). And the punishment system responds to failure to represent punishments by increasing the chance of generating a positive learning signal (technically, this is a “negative reward”). Note that the labels “reward” and “punishment,” like “positive” and “negative,” have nothing to do with feelings, self-conscious stances toward states of affairs, the intentions of would-be rewarding and punishing agents, judgments of goodness or badness, or the actual goodness or badness of any state of affairs. Particularly important to note, given the associations of the word “reward,” is the fact that the labels “reward,” “punishment,” “positive,” and “negative” have nothing to do with feelings of pleasure or displeasure (or pain). Within the theory of reward learning, these terms’ meanings ultimately come down to variations on the strengthening and weakening of causal connections, nothing more. If ideas of goodness, or feelings of pleasure, are at all related 14 If a representation that p only increases the chance of a positive learning signal being released because it causes a representation that q, then the former representation does not cause the relevant effect directly.

192  Timothy Schroeder and Nomy Arpaly to rewards in a given individual, it is because goodness, or pleasure, happens to be a reward in the technical, learning-theoretic sense for that individual.15 Returning to Juan, recall that he perceives the approach of his partner and that this perception causally participates in causing a matching step backward on his part. Suppose that what follows this step is that Juan’s partner smiles and the dance continues smoothly. This dancing, or smile (or what it signifies), might be a (positive) ­reward or a (positive) punishment in the sense relevant to reward learning, and it all depends on how Juan’s brain responds. If Juan’s brain responds by releasing a negative learning signal, the sort of signal that weakens the power of the perceptual state to cause matching steps backwards, then Juan’s brain has treated being smiled at (or continuing to dance smoothly16) as a punishment. On the other hand, if Juan’s brain responds by releasing a positive learning signal, one that strengthens the power of the perceptual state to cause matching steps backward, then Juan’s brain has treated being smiled at (or continuing to dance smoothly) as a reward. Most animals like ourselves respond to food (when hungry) and water (when thirsty) in the “positive” way, with the sort of reward learning that increases the strengths of just-used causal connections between mental states.17 And most human beings respond to things like money, success at solving problems, our favorite music, being cooperated with, donating to charities, and seeing our loved ones with the same sort of learning.18 Food, water, money, success, hearing “A Day in the Life” by The Beatles, being cooperated with, giving to a good cause, seeing a loved one—these are all commonly rewards, in the technical sense, for us. And similarly, most animals like ourselves respond to being deprived of food or water, or to (for instance) having their tails pinched in the “negative” way, with the sort of reward learning that decreases the strengths of just-used causal connections between mental states. Being denied supper or a turn to drink from the water fountain or being pinched are, commonly, punishments for us. And all these things are rewards or punishments because of the way representation of them disposes us to learn. 15 It has become common lore among parents that bored children will sometimes learn to do what gets them attention, even attention for bad behavior, even when bad behavior leads to bad feelings. This is a good illustration of the sort of learning involved in reward learning: Certain patterns become stronger because of their consequences (parental attention is a reward for most children), even when those consequences are quite independent of what is pleasant (or thought of as “good,” for that matter). The fact that other consequences (the scowling, the yelling, etc.) do not undo the reward learning suggests that parental attention is a stronger reward than parental yelling is a punishment, or that the timing of the reward (the attention is immediate) vs. punishment (the yelling often follows) matters to the learning process, or both. 16 Or something that such states of affairs seem to Juan to instantiate or make more likely. 17 Actually, it is probably better to think of animals like ourselves as intrinsically desiring a state of homeo­ static balance with respect to things like blood sugar, blood salinity, blood oxygen, and core body temperature—and perhaps also sex hormones or something tied to them. When we have deficiencies in these things, we have frustrated intrinsic desires for our homeostatic set points; when we have excesses of these things we also have frustrated intrinsic desires (as when one feels overfed, and there is an unpleasantness to it that goes beyond mere distension of the stomach or worry about appearance). 18 For experimental evidence, see, e.g., Stellar and Stellar (1985), Knutson et al. (2001), Johnsrude et al. (1999), Salimpoor et al. (2011), Rilling et al. (2002), Moll et al. (2006), and Aron et al. (2005).

The Reward Theory of Desire in Moral Psychology  193 Again, it is worth keeping in mind that the kind of learning under discussion is not just behavioral learning. In a person for whom hearing “A Day in the Life” is a reward, hearing this song might cause perceptual learning (so that one hears nuances that others cannot hear), learned patterns of involuntary attention (so that one by default attends to nuances that others can in principle hear, but to which they must make an active effort to attend), and more (if one often hears songs such as “A Day in the Life” while listening to a particular radio station, that might promote retention of that station’s call sign in one’s memory—so long as the mental connection between hearing the call sign and one’s memory is followed by hearing “A Day in the Life” in the pattern required to strengthen that connection). Hidden, perhaps, in the above description of reward learning is a fact of central importance to moral psychology: What makes a conceivable state of affairs p count as a reward or punishment for a given person (or any creature) is constituted by the contingent power of that person’s representation that p to contribute to the calculation of a reward or punishment signal. Most of us are born constituting sweet taste experiences, full stomachs, dry bottoms, and the like as rewards: Representations of these things contribute significantly in the production of reward signals, right from birth, as a result of nature’s design. And most of us come to constitute the taste of chocolate, and hearing “A Day in the Life,” as rewards as we get older: These representations come to have the power to directly promote the production of reward signals, though they were not disposed to do so from birth.19 But there is no state of affairs that is, by its very nature, such that representing it must generate a positive or negative reward signal.20 And there is no evidence to suggest that deliberation or other cognitive processes can, entirely on the basis of their contents, change what is constituted as a reward or punishment.21 The little that is known about how the brain comes to treat p as a reward or punishment when it did not before suggests that this happens when p is appropriately associated with antecedent rewards or punishments. When a young child finds that she gets things that are already rewards, such as cookies and positive parental attention, more often when everyone is cooperative, she is unconsciously changed so that she comes to constitute cooperation as a reward also.22 So far, no other natural mechanism for changing what is constituted as a reward is known. 19 There is some evidence that there is a specific region of orbito-frontal cortex that acts as a sort of junction box, where input from perception and cognition is either passed forward to the reward system or not, depending on the particular neural details. See, e.g., Schultz, Tremblay, and Hollerman (2000). 20 Unless one imagines a state of affairs constituted in part by the fact that it causes this sort of response, of course. 21 Thoughts can change whether one experiences an event as a failure to get X or an opportunity to better learn to Y, but this reframing relies on facts about whether X or Y is constituted as a reward—facts that the cognitive reframing does nothing to change. 22 This is a staple of the behaviorist literature on reward learning, but on the assumption that performing a guessing task successfully is or instantiates rewards for experimental subjects the effect has also been demonstrated over the short term in an experimental context by non-behaviorists: See Johnsrude et al. (1999), where a mere statistical association between being told “you got that right” and an abstract image caused greater liking for the associated image in human subjects.

194  Timothy Schroeder and Nomy Arpaly The second part of the explanation of how a positive, neutral, or negative learning signal is produced involves a calculation. The mere representation of a state of affairs that is a reward in the technical sense does not guarantee that a positive learning signal will be generated, and similarly for the representation of a punishment: Such representations are inputs into the calculation of what is sometimes known as a “prediction error.” It is easier, though, to think of what is being calculated as the difference, at each moment, between the expected amount of net reward or punishment in the world and the actual amount. That is, to produce reward learning, an organism must be doing three things. It must have an expectation regarding the net amount of reward to be found in the world, it must be evaluating the actual net amount of reward found in the world, and it must at each moment be taking the latter and subtracting the former from it.23 If the result of the calculation is positive, then a positive learning signal is released (and the more positive the result of the calculation, the stronger the positive learning signal). If the result of the calculation is neutral, then a neutral learning signal is released. And if the result is negative, then a negative learning signal is produced. Thus, in reward learning: Amount of reward learning signal = actual net rewards minus expected net rewards.

Hence, Juan’s partner’s smile causes less reward learning in him if he completely takes it for granted (viscerally expects) that his partner will smile at him and more if he was not taking such smiles for granted, assuming that smiles from this partner are (or signify to him other things that are) rewards in the first place. What we viscerally expect can often be thought of as what we are used to at a given time. It appears that in animals like us there is one system that calculates rewards only, and another system that calculates punishments only, with the two systems having reciprocal connections. (Note that this is the sort of detail we have generally suppressed for the sake of exposition.) Our speculation is that the punishment system performs the same calculation as the reward system, simply substituting “punishment” for “reward,” but this is still awaiting full empirical confirmation. The calculation of the difference between actual and expected reward or punishment looks like an intellectually challenging mathematical task, and not something that could be carried out unconsciously in human beings, much less monkeys or rats, but in fact neural computations of this sort are quite tractable. (Here, note, we move from the psychological level to its realization at the neural level.) A simple subtractive calculation can be performed unconsciously by some neurons exciting one region of the brain, while other neurons inhibit activity in that same region. Crudely, the resulting activity can be the difference between the sum of the exciting input and the sum of the inhibiting input. When all these states of neural activity represent magnitudes (on the same linear scale), then the resulting level of activity in the third group of neurons will represent a magnitude that is as great as the first magnitude minus the second.

See, e.g., Sutton and Barto (1998) for a textbook treatment.

23

The Reward Theory of Desire in Moral Psychology  195 These sorts of implicit calculations are ubiquitous in the brain, and not in themselves problematic. There might nonetheless seem to be something problematic about expectation. Surely this, at least, is a sophisticated task requiring explicit reasoning and judgment. But there are simpler ways to form expectations of reward than going through a thorough conscious reasoning process. The net rewardingness of a given moment can be expected on entirely associative grounds, without deliberation. A group of neurons that learn purely by association can determine the net rewardingness that has been associated, in the past, with the present perceived and conceived states of affairs, and this value (expressed, one assumes, in the rate of firing of this group of neurons) can serve as the expected level of net reward for that moment.24 This sort of implementation of expectations does not result in a perfectly accurate prediction of the amount of net reward that should obtain, of course. It might be that, in the past, circumstances like one’s current circumstances (perhaps, sitting at a Formica countertop in a Midwestern diner) have not been associated with tasting a perfect croissant, yet one consciously expects to taste such a croissant because one knows it is a morsel of such a croissant that one is about to eat. If tasting a perfect croissant is a reward for one, then one’s associations will lead to an expectation of less reward than one might have predicted consciously. Nonetheless, an unconscious, associative expectation system has substantial advantages over a conscious expectation system. Its expectations are continuously available without putting other cognitive activity on hold, are automatically available (regardless of how little one might feel like making more predictions), and can be formed using intellectual resources available to any creature capable of representing its environment and forming associations. So there are some virtues to this manner of calculating expected reward, in addition to there being faults. No doubt these virtues are responsible for the natural selection of the particular neural mechanisms we have (and share with other animals), but this particular evolutionary story is not really important. The degree to which a state of affairs is rewarding or punishing is constituted by these same connections that make the states of affairs rewards or punishments in the first place. A connection between the capacity to represent p and the reward system might be a very strong connection, the sort that will make a large positive contribution to the calculation of net reward. This would be how p gets to count as a big reward for the creature so constructed. But the connection could always be stronger or weaker, contributing a greater or lesser value to the calculation of net reward when p is represented. And thus how much of a reward p is can range on a continuum from being barely a reward at all to being by far the number one reward possible for the creature in question. 24 See, e.g., Schultz, Dayan, and Montague (1997). The scientific investigation of these details is continuing. The point of these speculations is simply to show that what is being described is possible and sensible; the details might well be different from what is being suggested.

196  Timothy Schroeder and Nomy Arpaly Putting all these pieces together gives a clear picture of reward learning. Imagine an organism constantly perceiving and cognizing its environment. Through a combination of genes and environment, some of the states of affairs it perceives and conceives are constituted by its reward learning system as greater or lesser rewards, while others are treated as neutral. The net sum of these rewards comes to some amount of net reward at each moment. At the same time, the organism also unconsciously, associatively predicts a given amount of net reward for that moment based on its past associations with similar conditions. The net reward of the moment has subtracted from it the predicted amount of net reward, and the result is either a positive value, zero, or a negative value. If the result is a positive value, then a positive learning signal is released, and it causes certain causal connections between mental states (those causal connections that were just used) to strengthen. If the result is a negative value, then a negative learning signal is released, and it causes certain causal connections between mental states (again, those that were just used) to weaken. And if the result is zero, then a neutral learning signal is produced and nothing changes. And as with reward, so too with punishment. The upshot of reward and punishment learning is that the mental processes—perceptual, intellectual, agentive, and so on—that lead to rewards (so defined) and the avoidance of expected punishments (so defined) become more dominant over time in the agents’ perceptions, thoughts, and actions, while those leading to punishments and to missing out on expected rewards play a smaller role over time. So long as there is some non-random pattern to the distribution of rewards and punishments in the world of the organism, the organism will come over time to perceive, think, and act in ways that are progressively more likely to lead to rewards and less likely to lead to punishments.25 Given the nature of reward and punishment, it follows that, according to the reward theory of desire, what it is to intrinsically desire that p is for the reward system to respond to representations of p in a way that directly increases the chance of a positive reward learning signal being generated. That is, Juan’s constituting p as a reward (in the technical sense) is the same thing as Juan’s desiring that p. His intrinsic (appetitive) desire for the positive regard of his dance partner is his responding (unconsciously, and not as an agent) to representations of this regard in a way that makes a positive contribution to the calculation of actual reward and, if the value of that calculation exceeds what was predicted, makes a positive contribution to the generation of a positive reward learning signal. Intrinsic aversions to it being the case that p are the mirror reflection of this arrangement in punishment instead of reward. And for one of Juan’s intrinsic appetites or aversions to be very strong is for the representation of its content by Juan to make a large contribution to the calculation of the overall learning signal,

25 For a more mathematical treatment of reward learning theory and contingency-based learning more generally, see, e.g., Sutton and Barto (1998).

The Reward Theory of Desire in Moral Psychology  197 while for an intrinsic desire to be weak is for the representation of its content to make a very small contribution. Unconsciously responding to p as a reward in the technical, learning-theoretic sense does not sound very much like desiring that p, and responding in a manner that greatly increases the chance of one sort of unconscious learning does not sound very much like greatly desiring that p. Although it might be true that we have the vague sense that there are unconscious learning processes, and that giving people treats (things that they intrinsically desire) can assist some of these unconscious learning processes, this is hardly the first fact about intrinsic desires that will leap to anyone’s mind. Similarly, the combination of an explosive gas with the gas we need to breathe does not sound like water. Yet water really is a combination of hydrogen and oxygen. This is our approach to thinking about intrinsic desires. Though intrinsic desires are not familiar to us as states of the reward and punishment systems, states of the reward and punishment systems do what intrinsic desires do, and so there is good reason to have a reward theory of desire. So far, we have said only what the reward and punishment systems are and what they do that makes them the sorts of things they are in themselves. So it is time to show that in addition to driving their own special form of learning, they also play the full range of roles associated with intrinsic desires: to show that it is states of the learning system that do what intrinsic desires do.

2.  The Reward System Causes What Desires Cause The reward system deserves the attention of philosophers because it causes what intrinsic desires cause. The system that controls reward learning causally influences actions, feelings, and (to a lesser extent) cognitions, and it is the only system that causally influences all of these in anything like the manner expected of intrinsic desires.26 In this section, we will show that this is so. To do so, we will rely primarily on neuroscience, especially findings from neuroanatomy and neurophysiology. That is, we will descend from the psychological level to the level of neural realization. By doing so, we believe, we can make the most secure case possible for the claimed causal roles of the reward system. After all, it is at the level of neural realization that cause and effect can be observed in the most direct, theory-neutral manner, as neuroscientists observe physical connections between discrete regions of the brain, and observe the causal powers of these connections. Begin with action production. The reward system’s impact on action production is both profound and very well studied, and much of what is known is very standard textbook material.27 26 This argument is made in full in Schroeder (2004, chs. 2, 3, and 4). 27 The impact of the punishment system on action is less well understood, and is not textbook neuroscience. However, if one line of contemporary thought is borne out and the brain’s serotonin system implements its punishment system, then there is a long history of empirical study of the way that the punishment

198  Timothy Schroeder and Nomy Arpaly The brain’s realization of the reward system sends some of its output to a key integrative structure for action production, known as the dorsal striatum. The dorsal striatum receives further input from all of the brain’s sensory and cognitive regions. Finally, it receives information about what basic actions (such as opening a hand, uttering a syllable, . . .)28 are primed for performance at the moment. Thus, information about what is the case and what is intrinsically desired combines in the dorsal striatum with information about what basic actions are available to be performed. The output of the dorsal striatum directly controls the regions of the brain known as motor and pre-motor cortex. These regions contain the brain’s capacity to command the performance of simple bodily movements, and their activity fully dictates the voluntary bodily movements people make. The motor cortex and pre-motor cortex can be thought of as made up of many different “keys,” like a multi-tiered organ keyboard, with each key capable of producing a basic bodily movement (such as opening a hand or uttering a syllable). The keys are primed by what is being thought and perceived at any given moment, but the actual “pressing” of the keys is under the control of the dorsal striatum, and so of belief and perception, on the one hand, and the reward and punishment systems, i.e., intrinsic desires, on the other. The reward system is not an optional component of action production: In its most extreme form, Parkinson’s disease kills all of the cells of the part of the reward system that contributes directly to action production, and the result is complete paralysis.29 This is not to say that a positive reward signal is required for action: Obviously, one does not need to constitute pinching one’s forearm as a reward in order to pinch one’s forearm. But the action production system is not designed to work independently of the reward system. If one is to pinch one’s arm, the production of this action through the dorsal striatum etc. must proceed in a manner that is at least somewhat sensitive to the strengths of reward signals (positive, neutral, or negative) that are being generated. (The specific way in which cognitive input, the reward signal, and the intrinsic structure of the dorsal striatum work together to select specific actions is still being studied.) What is important about the paralysis that can result from Parkinson’s disease is that it is not relieved by beliefs about what is right or good or reasonable. No such belief suffices to restore the capacity to act on its own, independently of what is constituted as a reward (i.e., intrinsically desired). Another striking way in which the reward system is a compulsory contributor to action is that there are known conditions that can override the output of the dorsal striatum, and so cause movements independently of the reward and punishment system contributes to inhibiting action and freezing movement, just as the reward system contributes to releasing action and promoting movement. See, e.g., Deakin (1983), Soubrié (1986). 28 Though we will not explore it in this chapter, there is substantial empirical evidence for Danto’s category of basic actions, which is essentially Davidson’s category of primitive actions. See Danto (1963; 1965), Davidson (1980), and any textbook account of the neuroscience of action and, e.g., the “motor homunculus.” See also Schroeder (2004, ch. 4). 29 Langston and Palfreman (1995) is a sophisticated lay account of this phenomenon.

The Reward Theory of Desire in Moral Psychology  199 systems (i.e., independently of intrinsic desires), but these conditions produce mere movements, not actions. For instance, direct electrical stimulation of parts of the brain can override the output of the dorsal striatum, and thus ignore the normal causal contribution of intrinsic desires to action, but these movements are experienced by their makers as not their own, and most philosophers would agree: These are not actions performed by the person, but mere movements forced out of the agent by a curious neuroscientist. Tourettic tics also appear to sometimes override the normal operation of the integrative action-production structure, forcing movements (not actions) out of the person with the disorder.30 The effect of the reward system on mental actions is not textbook neuroscience. Here the claim that these actions are influenced by the reward system is based on a limited number of research findings, but these are nonetheless suggestive. The best-studied mechanisms for mental action are those that are responsible for our capacities to visualize and to talk to ourselves. These mechanisms rely on distinct structures in the frontal lobes of the brain which can causally influence the same regions of the brain that are directly involved in seeing (for visualizing) and hearing (for talking to ourselves).31 So the question arises: what controls the activity of these specialized structures in the frontal lobes of the brain? It appears to be the same combination of reward signal and cognitive information (intrinsic desire plus belief) as serves overt action. The same neural structures in dorsal striatum that take input from belief plus desire and send output to control overt action also send output to the specialized structures in the frontal lobes with which we are concerned. If this output is not for the voluntary control of visualizing and talking to oneself, it would be quite a surprise. Having considered actions, let us consider feelings—another thing that desire seems to cause. To keep things simple, begin with feelings of pleasure and displeasure. The evidence that increases in the activity of the reward system cause increases in feelings of pleasure is widespread and powerful, and returns us to the realm of textbook neuroscience. Many pleasure-causing drugs (including the illegal ones) are known to act directly or indirectly upon some part of the reward system,32 electrical stimulation of parts of the reward system appears sufficient to cause pleasure under certain conditions,33 conditions that produce subjective pleasure also produce increased activity in the reward system,34 and so on.35 Furthermore, expected but not actualized rewards cause drops in the reward signal, and these same events cause feelings of displeasure.

30 Though it is important to note that Tourette syndrome is a complicated disorder, and not every Tourettic tic is involuntarily produced. See Schroeder (2005) for more on this. 31 A lovely overview is found in Baddeley (2003). 32 A nice pair of experiments is Liechti et al. (2000), Liechti and Vollenweider (2000). Schroeder (2004, ch. 3) contains a review. 33 See, e.g., Stellar and Stellar (1985). 34 An elegant experiment showing this is reported in Salimpoor et al. (2011). 35 For an overview, see Schroeder (2004, chs. 2 and 3).

200  Timothy Schroeder and Nomy Arpaly Evidence that increases in the activity of the punishment system causes displeasure is suggestive, though much less well established.36 If there is room to explain feelings of pleasure and displeasure through the actions of the reward and punishment systems, then what about the wider range of positive and negative feelings? What about feelings of fear and anger, or schadenfreude, or bliss? These feelings contain pleasure and displeasure. Insofar as these are feelings, they go beyond pleasure and displeasure in how one’s body feels while in their grips. Fear without the feeling of a pounding heart or hair standing on end does not feel much like fear; bliss without a feeling of muscles relaxing throughout one’s body does not feel much like bliss; and so on. Not surprisingly, these bodily changes are not independent of the reward and punishment systems. There are outputs from the punishment system that reach the amygdala, a neural structure that has a famous role as the central “junction box” causing the bodily states most associated with fear and anger and disgust (it is involved in feelings linked to other emotions as well).37 Thus, the reward and punishment systems have causal connections of the sort required to bring about the feelings we expect to be brought about by getting, or being denied, what one intrinsically (appetitively) desires (or is averse to). Finally, consider the cognitive effects associated with intrinsic desires, among which are changes in attention and some types of motivated irrationality. There is evidence that these effects are also commonly generated by the reward and punishment systems. Unfortunately, this evidence is the furthest from textbook ready; it is all the product of evolving research programs. Nonetheless, it is suggestive. A little should be said about both attention and motivated irrationality. Insofar as the direction of attention is a voluntary (not necessarily premeditated) action, one would predict that it would be under the control of the dorsal striatum and hence the reward and punishment systems. Evidence for this can be found by considering the deficits that people with Parkinson’s disease show in the voluntary direction of attention. In various experiments, it has been found that people with Parkinson’s disease, who share in common damage to their reward systems, have difficulty directing or redirecting their attentions when the direction of such attention is up to the person in the experiment. For instance, in one such experiment, subjects were asked to hold their eyes fixed on a particular target, but cued to pay attention (“covertly”) to a location nearby the target; this voluntary direction of attention made a discrimination task easier. But people with Parkinson’s disease were found to not improve on the discrimination, showing that they were having trouble keeping their attentions fixed on the cued locations.38 People with Parkinson’s disease thus appear to have difficulties directing their attentions where they want.

36 For a start, the reader might consult Lowry et al. (2008) for a review. 37 For a review of what we know about the amygdala, see LeDoux (2000). On the connection of the punishment system to the amygdala, there are a number of works cited in Lowry et al. (2008). 38 See Sampaio et al. (2011), where every effort is made to ensure that only difficulty in controlling attention explains the experimental findings. Older work focused specifically on the voluntary direction of attention

The Reward Theory of Desire in Moral Psychology  201 As for motivated irrationality, there is a hint toward the role that the reward system plays. It is textbook science that cocaine stimulates parts of the reward system directly, mimicking (by our lights) the effects of strong intrinsic desire satisfaction. It is also a commonplace that people consuming cocaine are typically convinced (while intoxicated) that they are more interesting, attractive, powerful, and so on than they would otherwise take themselves to be. Cocaine thus appears to induce something like motivated irrationality by mimicking the effects of strong intrinsic desire satisfaction, suggesting—if very tentatively—a role for the reward system in the explanation of everyday forms of motivated irrationality. At this point, the effects of the reward and punishment systems on action, feelings, and cognitions have been canvassed, and the results have been the same. In each case, the reward and punishment systems play the roles played by intrinsic desires. Further, the realizers of the reward and punishment systems are the only things in the brain that play all of these roles. This negative thesis is of course an empirically risky one. However, given the current state of neuroscientific knowledge, the negative thesis is not much riskier than the structurally equivalent thesis that the heart is the sole structure in the body that pumps blood. There are human beings with tremendously impaired reward systems who are otherwise intact—people with the most severe forms of Parkinson’s disease. These people provide scientists with an opportunity to discover other systems capable of playing the same roles as the reward system: other systems capable of driving action even when the reward system cannot, and other systems capable of producing elation even when the reward system cannot. But no such systems have come to light. There are more and less important kinds of functionality spared in people with severe Parkinson’s disease, but no coordinated package of functions that does most or all of what is expected from intrinsic desires.

3.  Intrinsic Desires are a Natural Kind Intrinsic desires play a large number of causal roles. And there is a natural psychological kind, the reward learning system, that plays these same causal roles.39 This suggests a natural-kind strategy for theorizing about intrinsic desires. Following this strategy, intrinsic desires should be theorized as being the natural kind that does, in fact, play all of the causal roles that intrinsic desires play.40 And so, following this strategy, intrinsic desires are states of the reward learning system: They are what the reward theory of desire says they are. in people with Parkinson’s disease goes back to at least Brown and Marsden (1988), another suggestive if less controlled study with the same upshot. 39 A natural kind, again, on the assumption that playing a central explanatory role in scientific psychology is a good test for being a natural kind. 40 This strategy is described and defended in Schroeder (2004); a similar strategy for theorizing desire is articulated in Morillo (1990). Compare methodology and results to the similar investigation of the emotions in Griffiths (1997).

202  Timothy Schroeder and Nomy Arpaly To treat intrinsic desire as a natural kind is to treat it as akin to, for example, water.41 Water is that substance found in rain, in lakes, in the ocean, that is liquid, transparent, able to quench thirst, make crops grow, dissolve salt crystals. Water is not identical to, or constituted by, any of these properties or powers: It is the thing that has them. These powers might have been had by other natural kinds (some subsets of them actually are), or had by no natural kind at all, but water is the natural kind found around us that happens to be the source of these properties and powers coming together so reliably.42 To treat intrinsic desire as a natural kind is to hold that intrinsic desires are the things, whatever they turn out to be in themselves, found in people and other animals, causing the actions, feelings, and cognitions that we associate with terms such as “desire” and “want” and, especially, phrases such as “wanting for its own sake” and “desired just for itself.” If they form a natural kind, then intrinsic desires need not be identical to, or constituted by, any of their most familiar effects: The natural kind can be merely the common cause of them in human beings. If intrinsic desires are a natural kind, then all of their familiar effects might in principle (and might, in practice) stem from other causes, though intrinsic desires would be the sole common cause of these effects in us. Why, though, should philosophers treat intrinsic desires as a natural kind? There are at least four arguments worth considering, and each has some weight—but in the present context, we will not claim that any of the four is more than suggestive. The first argument is that, if there is a good a priori knowable (or armchair verifiable) theory of intrinsic desire that reveals its essence to be something other than the natural kind (whatever it turns out to be) that causes what intrinsic desires cause, then that theory is likely to theorize intrinsic desires in terms of action, pleasure, or both action and pleasure. But a number of philosophers have argued against these views on armchair grounds.43 The most obvious problem with this first argument is, of course, that these debates are far from resolved. But for our purposes, it will be enough to note that it has not been resolved against the natural-kind approach, and move on. Obviously, much more

41 Putnam (1975). 42 Note that, to treat intrinsic desires as a natural kind is not necessarily to commit to the thesis that, if there is no suitable natural kind, then there are no desires (compare Churchland 1981). Consider the study of the elements: earth, air, fire, and water. Only water turned out to be a natural kind, but it was not unreasonable for early science to approach all four elements with the hypothesis that each one was a natural kind. The falsification of that hypothesis in three out of four cases did not reveal that there was no such thing as earth, air, or fire. It showed only that, whatever earth, air, and fire are, they are not the sort of thing suitable for scientific study. We take a similar approach to desire. Intrinsic desire should be approached with the hypothesis that it is the natural kind, whatever it might be, that gives rise to (at least, nearly) all the phenomena we associate with intrinsic desire. If there is such a kind, then it is what intrinsic desires are. If there is not, some other hypothesis can be formed; Elimination of the kind from our ontology would be rather premature, even if it would be eliminated from our science (and so, from the explanatory categories of scientific psychology). 43 Against the pure action-based theory, see arguments in Stampe (1986), Schroeder (2004), Schueler (1995), and Strawson (1994). Against the pure pleasure-based theory, see arguments in Schroeder (2004). And for arguments against combined theories, see Schroeder (2004), Strawson (1994).

The Reward Theory of Desire in Moral Psychology  203 needs to be said about the specific nature of intrinsic desires, but this will not be the place where we attempt that task. The second argument is an entirely general argument based on the nature of thought and speech: Every kind we think or talk about is best thought of as a natural kind, if there is a natural kind that is a reasonable candidate to be the extension of our concepts and the referent of our words (except for kinds introduced by stipulation or for technical purposes, or the like). Consider air. If air had turned out to be one specific molecule then “air” would, like “water,” be a natural kind term. The only reason “air” is not a natural kind term is that there is no natural kind that is a reasonable candidate to be the referent of “air.” The air is mostly nitrogen gas, which accounts for most of the forcefulness of wind on windy days (just by being the main gas present, by a large margin) but which does not explain why we need to breathe air. The air is also made up of a lot of oxygen gas, which does explain why we need to breathe air, but it does not do most of the explaining of the forcefulness of the wind. So, without a reasonable candidate natural kind to refer to, “air” does not refer to a natural kind. But it could have been otherwise. From our present vantage point, “air” does not feel like a natural kind term, but had the facts about the atmosphere been different, and had science learned and widely disseminated those facts, our feelings would be different. “Desire” might be more like “water” or like “air,” in that it might have a single natural kind that is a reasonable candidate for its referent, or it might not. But if there is a candidate natural kind suited to be the referent of “desire,” then desire should be treated as a natural kind. The most obvious problem with this second argument is that it relies on a controversial principle about the referential dimension of language and thought: that concepts and terms for kinds are always best understood as referring to a natural kind if there is a candidate natural kind available to be the thing conceived or referred to. And this is certainly not the place to embark on a defense of such a principle. A third argument is that the parts of the mind studied by philosophers will, ideally, be both real and well suited to feature in ideal explanations of human thought, feeling, and action. And one good test of the reality and explanatory robustness of putative parts of the mind is that they are appealed to by our best psychological sciences,44 or could be if only our psychologists were slightly better philosophers, and would see that their technical term x is really a term for familiar part of the mind y. It seems likely that the psychological sciences will continue to appeal to the reward and punishment systems and make them robust parts of larger explanations of various aspects of thought, feeling, and action, for all the reasons canvassed in the previous section and 44 In saying this, we at least partly accept the terms of the debate set out by well-known eliminative materialists: If intrinsic desires are more like air than water, then there will be little use in focusing on intrinsic desires in any explanation of important phenomena, though it will not be any more wrong to say that there are intrinsic desires than to say that there is air. If intrinsic desires are less like water and more like fire (a totally heterogeneous collection of processes, including solar fusion and high-speed highly exothermic chemical oxidation, it seems) then it will not just be sloppy or unhelpful to appeal to intrinsic desires in explanations, it will be downright embarrassing. See, e.g., Churchland (1981) for an early starting point.

204  Timothy Schroeder and Nomy Arpaly many more besides. If the reward and punishment systems are going to be a staple of psychological explanations, then either philosophers will have to be able to interpret the reward system as realizing intrinsic desires, or they will have to be able to interpret the reward system as realizing some other mental state, or they will have to cede the explanation of these aspects of thought, feeling, and action to something philosophers do not study and give no role in their theories of action, virtue, moral worth, and so on. This lattermost option would be intolerable, however. So insofar as the psychological sciences continue to rely on the reward and punishment systems as robustly independent causal explainers of various aspects of thought, feeling, and action, philosophers would do well to determine what mental states or events should be understood to be at work in these explanations. It seems to us that these mental states are best thought of as intrinsic desires. And so there is a reason from scientific practice to embrace a reward theory of desire. A problem with this third argument is that a number of philosophers, including some with substantial involvement in action theory and moral psychology, have denied that mental states need to be treated as scientifically respectable entities that enter into non-trivial, interesting causal explanations of various aspects of thought, feeling, and action. And again, while this is an interesting debate, it is not one that will be pursued in the present work. The fourth argument for treating intrinsic desires as a natural kind is aimed solely at philosophers who are sympathetic to desire-centered action theory or moral psychology. By treating intrinsic desires as a natural kind, it is possible to go a long way toward answering many of the most damning objections to desire-centered moral psychology. This claim will be defended in section 4, but it has no doubt already occurred to the reader how useful the reward theory of desire could be to a desire-centered moral psychology. And a special virtue of responding to these objections in this way is that treating intrinsic desires as a natural kind, while not uncontroversial by any means, is neither controversial nor defensible on grounds that are primarily action-theoretic or moral-psychological. As a result, it does not beg any questions to strengthen one’s desire-centered theory by holding that intrinsic desires are a natural kind. The debate over whether mental terms are, insofar as possible, natural kind terms is a debate in the philosophy of language, philosophy of science, and philosophy of mind, and premises about love, care, acting for reasons, or even compatibilist theories of free will are not generally important premises in such debates. Of course, the four arguments just presented are far from conclusive, and are intended much more to suggest our general lines of thought than to convince the skeptic. An extended argument meant to convince the skeptic is made in Schroeder (2004, especially ­chapter  6), but even that is far from the last word in our eyes. Nonetheless, we do find something independently appealing to the straightforward thesis that when we say “desire” we are talking about something real inside us that plays a complex set of causal roles but has an essence or nature that is not automatically transparent to us.

The Reward Theory of Desire in Moral Psychology  205

4.  Solutions and Promissory Notes Suppose, then, that intrinsic desires are identical to states of the reward and punishment systems. Intrinsic desires are causally involved in movements, feelings, and cognitions, but they could fail to be involved in any of these things, and these things can be, and routinely are, influenced by things other than desires. Intrinsic desires have strengths, but the strength of any given episode of motivation, emotion, or cognitive influence could be quite a bit different from the strength of the intrinsic desire being expressed. What would be the importance of this for action theory or moral psychology? If intrinsic desires are states of the reward and punishment systems, then there is room for the desire-centered moral psychologist to hold that many of the most problematic objections to her position can be answered. In this section, we show how the reward theory of desire helps to solve certain problems, and how it shows promise to solve others. Start with love and care. A number of theories put intrinsic desires at the heart of what it is to love someone or to care for her: The theories of Harry Frankfurt are perhaps the most prominent examples here.45 Such theories have been criticized, both implicitly and explicitly, by philosophers who hold that, regardless of the fine details, intrinsic desires are just the wrong sorts of things on which to ground theories of love or care. One possible objection to desire-based theories of love and care is that it is possible to imagine a person who has the theoretically demanded qualities—who intrinsically desires the well-being of the beloved, and so on—but who does not appear to really love or care. For instance, it is possible to imagine a person who is very strongly motivated to tend to the well-being of the beloved, but whose motivation to do so seems much more like a neurosis, a compulsion, or a tic than like true love or caring.46 Here the reward theory of desire can help. The first thing to do is to be careful about the case under consideration. It is no threat to a desire-based theory of love or care to say that a person might feel alienated from her impulse to care for Miguel on a given occasion, or that she might feel rueful about doing so or might do so akratically. In none of these circumstances would there be a temptation to deny that the act done out of an intrinsic desire for Miguel’s well-being genuinely counts as an act done out of love: What we do out of love, all too often, is judged not to be best, is less than fully rational, and (less often, but still occasionally) provokes feelings of alienation (especially when the beloved is a very socially inappropriate object of love or care). For it to be a problem for a desire-based theory of love or care, a case needs to be one in which there is very little sign of anything like genuine

45 See Frankfurt (1988; 1999). 46 Some prominent examples include Darwall (2002, ch. 1) and Jaworska (2007a; 2007b) on care, and Velleman (1999) on love.

206  Timothy Schroeder and Nomy Arpaly love or care, however conflicted or conducive to folly on a particular occasion. There must be very little sign of anything like genuine love or care, while there is nonetheless a powerful motivation to act so as to do what seems to improve another’s well-being, a motivation independent of its instrumentality to other things that might be intrinsically desired. The motivation must appear very much a sort of neurosis, compulsion, or tic. A feature of people who seem merely neurotic, merely compelled, or merely subject to exotic behavioral tics is that these people are not, in general, made delighted or contented by the actions they neurotically, compulsively, or in tic-like fashion perform. A person who feels merely forced to tend to Miguel’s well-being might feel relief that she is not prevented from acting on her compulsion, but not delight that Miguel is well or contentment that he is flourishing. But it follows from the reward theory of desire that, in human beings, getting what you intrinsically desire is a normal cause of pleasure and more complex positive emotions. It seems there are two possibilities. Either the apparently compelled person is acting on what the reward theory identifies as a genuine intrinsic desire for Miguel’s well-being, and is somehow being prevented from having the feelings that are appropriate to the satisfaction and frustration of her intrinsic desire, or the apparently compelled person is not acting on a genuine intrinsic desire for Miguel’s well-being. If the first possibility holds, then the person who feels compelled to tend to Miguel’s well-being is someone who, if only she were not depressed, brain-injured, associatively scarred, or otherwise damaged, would often feel delight in helping Miguel. She would often feel good about his flourishing whether or not she were the agent of it, and she would look forward to chances to help him out when he was in difficulties—at least, to the extent that people who love and care for others ordinarily have these feelings. And if this is how the compelled-feeling person would feel if only she were undamaged, then it seems that she does indeed love or care for Miguel, and she is simply unable to access the normal feelings that reveal love or care to us. Those feelings are normal given her desire, according to the reward theory of desire and the earlier claims about reward’s relation to pleasure and displeasure. And if those feelings would be normal, then it seems wrong to say that the imagined person is merely compelled, merely acting on a tic. The first possibility, in short, reveals no problem for a desire-based theory of love or care, given the reward theory of desire. But then, the second possibility does not even appeal to an intrinsic desire for Miguel’s well-being. It appeals to some other state capable of causing the behavioral impulses imagined. And such impulses are obviously not a threat to a desire-based theory of love or care, since they are understood to not stem from intrinsic desires for Miguel’s well-being. Thus, the specter of a compulsive well-being tender who counts as loving or caring on a desire-based theory of love or care is merely apparent. Another possible objection to desire-based theories of love and care is that it is possible to imagine a creature who lacks the theoretically demanded qualities—who does not intrinsically desire the well-being of the beloved—but who nonetheless appears to

The Reward Theory of Desire in Moral Psychology  207 really love or care. By turning to thought experiments, one can imagine a person who lacks any capacity for motivation, but whose feelings (and perhaps patterns of attention) convincingly attest that she loves another—perhaps, as David Velleman might have it, by being emotionally open to the other’s value as a rational agent. Does this unmotivated person nonetheless have an intrinsic desire for the well-being of the beloved? If the reward theory of desire is correct, then the answer might be “yes.” It might be the case that the unmotivated person has an intrinsic desire in virtue of having an appropriate state of her reward system; changes in the satisfaction of this state would be represented in the person’s feelings, but the state would (by the terms of the thought experiment) have no access to any mechanisms by which actions might be produced. This would still be an intrinsic desire. Furthermore, if this feeling but unmotivated being gets pleasure from the well-being of the beloved, and displeasure from the misfortunes of the beloved, then it can be asked whether these feelings are normal or not given its state. They are normal just in case the being does intrinsically desire the well-being of the beloved. So the unmotivated feeler either feels about the beloved in a way that is biologically normal or not. If the feelings are normal, then the being intrinsically desires the well-being of the beloved, and the reward theory of desire shows how this is possible in a creature incapable of motivation. But if the feelings are not normal, then it is hard to see the imagined being as genuinely loving. If one feeling but non-acting creature feels bad at the injury of another feeling but non-acting creature, but this feeling bad is not normal given its inner psychological makeup, then it can hardly be said that the first creature loves the second—since love, it seems, suffices to make such feelings normal responses to the injury of the beloved. Of course, the desire-based theorist of love who holds the reward theory of desire must grant that it is possible that a person lack all capacities to act and all capacities to feel—and even all capacities for voluntarily directed attention—and yet still love. To love, that is, without showing any of the ordinary signs of love at all. But this conclusion need not be particularly problematic. When a loving creature is stripped of its capacity for emotional warmth its love will appear cold; when a loving creature is stripped of its capacity for actions expressing concern for the well-being of the beloved, its love will appear unconcerned. But these are just appearances. As love is stripped of more and more of its common effects, it looks less and less like love, the desire-based theorist of love can hold, because all of the signs of love are disappearing. Our ordinary epistemic access to that love is disappearing—but the love itself might remain. Given the reward theory of desire and its relation to pleasure, an emotionless person who putatively loves is a person for whom feelings of joy at the salvation of a putative loved one would be normal if the person were capable of feelings of joy. This suggests that this normal sign of love—taking joy in the salvation of the beloved—is missing, but the thing that would normally lead to the sign (that is, the love itself) is not, because what is missing is the response, not the thing normally causing the response. And likewise, given the reward theory of desire, an unmotivated person who putatively loves is a person for whom trying to assist the well-being of a loved one would be very reasonable if the

208  Timothy Schroeder and Nomy Arpaly person were capable of action.47 Again, this normal sign of love might be missing, but the thing that would normally lead to it is still present: The reasonable act is missing, but the thing making the action reasonable exists. If intrinsic desires are things that are potentially hidden from the casual eye, then so too will be love, on a desire-based theory of love. But if it is true of a creature that it contains a state normally causing the feelings we associate with love (only it cannot feel) and warranting the actions we associate with love (only it cannot act), then it seems reasonable that the creature loves: that it is love itself that makes normal the feelings and makes reasonable the actions. Perverse behavioral impulses of all sorts have been thought to be problems for desire-based moral psychology. The person who has a behavioral impulse to turn on radios,48 the person who has a behavioral impulse to count the blades of grass on squared-off lawns,49 the person who has a sudden impulse to drown her screaming baby in its bathwater,50 the person who has a behavioral impulse to seek out cocaine to the exclusion of all else in life, and the person who has a behavioral impulse to twitch, jerk, and shout obscenities (as happens in cases of Tourette syndrome) all seem like people who put theoretical pressure on the desire-based moral psychologist. Do these people have rationalizations for their actions? Good rationalizations (assuming the behavioral impulses are very strong)? Do these people illustrate the claim that rational action must always be action guided by reason, not desire?51 If these people knowingly do something morally good or required as a result of their behavioral impulses, will the desire-based moral psychologist grant them moral credit? If these people knowingly do something morally bad or forbidden as a result of their behavioral impulses, will the desire-based moral psychologist call them selfish or malicious rather than victims of mental states gone awry? All of these problems loom for the desire-based moral psychologist because of the assumption that, if a person is behaviorally impelled to act in a coherent way that is responsive to the person’s instrumental beliefs, then what motivates the person must be an intrinsic desire—and if the behavioral impulse is strong, then the intrinsic desire must be a strong one. An action-based theory of desire, or some variant of it, is being presupposed. But if the reward theory of desire is correct, then there remains the possibility that in none of the above cases is the person in question one who desires the end to which the person is behaviorally impelled. On the reward theory of desire, being motivated to bring it about that p is not a sufficient condition for intrinsically desiring that p, or even desiring that q when p seems instrumental to, or a realizer of, q. Thus, the reward theory of desire allows the possibility that the people just described do not actually desire to turn on radios or count blades of grass, drown the baby, buy cocaine, or shout obscenities. The reward theory of desire leaves open the possibility



Assuming that intrinsic desires can make actions reasonable, of course. Quinn (1993, 237).    49  Rawls (1971, 432). 50 Watson (1975, 210). 51 Quinn (1993), Watson (1975), and many others have drawn such conclusions.

47

48

The Reward Theory of Desire in Moral Psychology  209 that the desire-based theorist of moral psychology need not worry about any of these cases. We will not show that all of these cases in fact come out as the desire-centered moral psychologist would wish; that is a project for another occasion or three. But the foregoing arguments have made it clear how such arguments might proceed. On the reward theory, there are many opportunities for the signs of intrinsic desires to be present or absent, strong or weak, without the underlying intrinsic desires themselves being present or absent, strong or weak. Of course, the reward theory of desire merely leaves the possibility open. Whether or not the possibility is actual depends on the psychological details. Are the states that would bring about the kinds of radio-turning-ons, grass-countings, baby-drownings, cocaine-buyings, and obscenity-shoutings of interest to moral psychology states of the reward or punishment systems or are they not? Are they of a kind with the things that we recognize as intrinsic desires, or are they of some other, distinct kind—a kind that, by being distinct from intrinsic desire, can pose no objection to the desire-based theorist of moral psychology? These questions need more evidence before being answered. Turning back to the objections raised to desire-based theories of moral psychology, there is another family of objections to be highlighted here. According to these objections, it is obvious that desires cannot be the foundation of moral psychology because it is obvious that sometimes we are moved by things other than desires in many cases of acting for reasons, acting with moral worth, and so on. The person who is tempted to skip the PTA meeting in order to read in front of the fire but who goes to the meeting anyway is a central sort of case here.52 Then there is the person who helps another out of “mere inclination” (i.e., a desire to do so) standing in contrast to the person who helps another out of “duty” (i.e., a cognition of something in the domain of reasons or values).53 Or again, there is the distinction between the person who is seduced by his desires (i.e., his reason is persuaded to make a choice it would not otherwise have made because of the influence of desire) and the person who is overwhelmed by his desires (and so his reason plays no part in what he does).54 As before, all of these problems loom for the desire-based moral psychologist because of assumptions about the nature of desire. In the first case above, the idea that one does not go to the PTA out of a desire to do so is supported principally by the idea that when one acts successfully on a desire one can expect to be pleased to be getting what one wants, along with the observation that there are many cases, such as giving up an evening by the fire for a PTA meeting, in which one is motivated without there being any reasonable expectation that one will be pleased if one succeeds. A similar line of thought might lie behind the idea that being motivated by duty is something other than (or more than) being motivated by an intrinsic desire to do the right thing

52 See Schueler (1995, 29); for similar thoughts, see also Davis (1986), Vadas (1984). 53 Obviously Immanuel Kant is the primary modern source of this idea: See Kant (1998). More recently the idea can also be found clearly expressed in, e.g., Herman (1993, ch. 1). 54 Watson (1975).

210  Timothy Schroeder and Nomy Arpaly (or something similar). And then, the idea that desires can seduce—the idea that seduction is an apt metaphor for the process by which desire can influence action in certain cases—strongly suggests that our experiences of pleasure are an important part of the evidence for the idea. A pleasure-based theory of desire, or some variant of it, seems to be what is behind these ideas.55 But again, if the reward theory of desire is correct, then there remains the possibility that in none of the above cases is the feeling of pleasure a flawless guide to the role being played by intrinsic desires. Since, on the reward theory of desire, being pleased by improved prospects of it being the case that p is not a necessary condition for desiring that p, the reward theory of desire allows there to be a possibility that the person going to the PTA does so out of an intrinsic desire, that there is no distinction in kind between apparently acting out of inclination and apparently acting out of duty, and that there is no distinction in kind between apparent motivation by reason seduced by desire and apparent motivation by desire independently of reason. Alternatively, one might use the reward theory of desire to argue that the distinction in kind is between appetitive desires, where success in attaining the desired end promotes a reward signal, and aversions, where success in avoiding the aversive end merely avoids promoting a punishment signal. The reward theory of desire leaves open the possibility that the desire-based theorist of moral psychology need not worry about any of these cases. Although the above-mentioned potential benefits are potential benefits of holding the reward theory of desire, there are other ways of achieving these benefits too. Some of our reasons for preferring the reward theory of desire are found in the defense of the theory presented in Schroeder (2004), and we showcase the virtues of the reward theory in moral psychology in Arpaly and Schroeder (2014). But none of these reasons for preferring the reward theory of desire to other theories, in moral psychology or elsewhere, will be decisive. Other desire-based theorists of moral psychology have taken other approaches to the sorts of problem cases we just listed, and this chapter will finish with a short discussion of them. These approaches have generally taken the form of placing restrictions on the kinds of desires that a desire-centered moral psychology should be built upon. Such desires should be ones that one wholeheartedly wants to have,56 or should be ones that could survive cognitive reflection,57 or should be ones generated by one’s beliefs about what one would intrinsically desire if one were rational,58 or should be properly integrated into one’s overall psyche,59 or . . . Inevitably (or so it seems to us), these extra restrictions on the sorts of desires that count create additional problems

55 Davis (1986) defends a motivation-plus-pleasure theory of desire. 56 Frankfurt comes to something like this view by his paper “Identification and wholeheartedness,” though perhaps by that time Frankfurt had also backed away from the broad moral psychological ambitions on display in “Freedom of the will and the concept of a person.” Compare Frankfurt (1971) to Frankfurt (1987). 57 Tiberius (2002).   58 Smith (1994).    59 E.g., Fischer and Ravizza (1998).

The Reward Theory of Desire in Moral Psychology  211 for the moral psychologist.60 One of us has written extensively about the fact that people can take foolish meta-attitudes toward their own desires, and so restrictions based on meta-attitudes (both cognitive and conative) face serious challenges: What shall we make of a person who disapproves of her homosexual desires, for instance?61 Are these desires “not truly her own”? What about the person who, like Huckleberry Finn, appears to have a virtuous heart but false moral beliefs? Restrictions based on the desires being properly integrated within the overall psyche face challenges from the ordinary sorts of desires, such as sexual desires, that spring themselves upon us through non-psychological processes without being lesser partners in our moral-psychological lives for all that. And restrictions based on rational acceptability are restrictions that give up too much of what one is looking for in a genuinely desire-based moral psychology. This mention of challenges faced by other desire-based theorists is not intended as a refutation of the opposing views. We merely state our reasons for staking out a different path: that of treating all intrinsic desires on a par, for purposes of action theory and moral psychology, while being more mindful than most of what intrinsic desires really are. If this path can provide solid responses to the many objections that have been leveled at desire-based theories without subdividing intrinsic desires into those we like as theorists and those we need to push aside, then that seems a mark in favor of our theory: It is unified in its treatment of intrinsic desires. But this would only be a small mark in favor. As at so many points in this chapter, our point is ultimately a modest one: to suggest that there is an approach here, inspired by our best scientific knowledge, that might repay serious philosophical attention.

Acknowledgments Our work here is largely drawn from c­hapter  6 of Arpaly and Schroeder (2014). We thank Justin D’Arms and Dan Jacobson for extensive comments on the present paper, and Peter Railton for challenging feedback on an earlier talk covering the same material.

References Aron, A., Fisher, H., Mashek, D., Strong, G., Li, H., and Brown, L. 2005. Reward, motivation, and emotion systems associated with early-stage intense romantic love. Journal of Neurophysiology, 94, 327–37.

60 A very early expression of some of our reservations about using a restricted set of desires (or anything else) to delineate the “real” self, the motives of which are the only ones that render one apt for moral praise or blame, is found in Arpaly and Schroeder (1999). While our positive claims in that work are quite different from our present claims, we think that the negative arguments still stand. 61 See, e.g., Arpaly (2003).

212  Timothy Schroeder and Nomy Arpaly Arpaly, N. 2003. Unprincipled Virtue. New York: Oxford University Press. Arpaly, N., and Schroeder, T. 1999. Praise, blame, and the whole self. Philosophical Studies, 93, 161–88. Arpaly, N., and Schroeder, T. 2014. In Praise of Desire. New York: Oxford University Press. Baddeley, A. 2003. Working memory:  Looking back and looking forward. Nature Reviews: Neuroscience, 4, 829–39. Brook, A. 2006. Desire, reward, feeling: Commentary on Three Faces of Desire. Dialogue, 45, 157–64. Brown, R., and Marsden, C. 1988. Internal versus external cues and the control of attention in Parkinson’s Disease. Brain, 111, 323–45. Churchland, P. 1981. Eliminative materialism and the propositional attitudes. Journal of Philosophy, 78, 67–90. Danto, A. 1963. What we can do. Journal of Philosophy, 60, 435–45. Danto, A. 1965. Basic actions. American Philosophical Quarterly, 2, 141–8. Darwall, S. 2002. Welfare and Rational Care. Princeton: Princeton University Press. Davidson, D. 1980. Essays on Actions and Events. Oxford: Oxford University Press. Davis, W. 1986. The two senses of desire. In J. Marks (ed.), Ways of Desire:  New Essays in Philosophical Psychology on the Concept of Wanting, pp. 63–82. Chicago: Precedent. Deakin, J. 1983. Role of serotonergic systems in escape, avoidance, and other behaviors. In S. Cooper (ed.), Theory in Psychopharmacology, vol. 2, pp. 149–93. New York: Academic. De Sousa, R. 2006. Dust, ashes, and vice: On Tim Schroeder’s theory of desire. Dialogue, 45, 139–50. Dretske, F. 1988. Explaining Behavior: Reasons in a World of Causes. Cambridge, MA: MIT Press. Fischer, J., and Ravizza, M. 1998. Responsibility and Control: A Theory of Moral Responsibility. New York: Cambridge University Press. Frankfurt, H. 1971. Freedom of the will and the concept of a person. Journal of Philosophy, 68, 5–20. Frankfurt, H. 1987. Identification and wholeheartedness. In Frankfurt 1988, 159–76. Frankfurt, H. 1988. The Importance of What We Care About. New York: Cambridge University Press. Frankfurt, H. 1999. Necessity, Volition, and Love. New York: Cambridge University Press. Griffiths, P. 1997. What Emotions Really Are:  The Problem of Psychological Categories. Chicago: University of Chicago Press. Herman, B. 1993. The Practice of Moral Judgment. Cambridge, MA: Harvard University Press. Jaworska, A. 2007a. Caring and full moral standing. Ethics, 117, 460–97. Jaworska, A. 2007b. Caring and internality. Philosophy and Phenomenological Research, 74, 529–68. Johnsrude, I., Owen, A., Zhao, W., and White, N. 1999. Conditioned preference in humans: A novel experimental approach. Learning and Motivation, 30, 250–64. Kant, I. 1998. Groundwork of the Metaphysics of Morals, trans. M. Gregor. New York: Cambridge University Press. Knutson, B., Adams, C., Fong, G., and Hommer, D. 2001. Anticipation of increasing monetary reward selectively recruits nucleus accumbens. Journal of Neuroscience, 21, 1–5. Langston, J., and Palfreman, J. 1995. The Case of the Frozen Addicts. New York: Pantheon. Latham, N. 2006. Three compatible theories of desire. Dialogue, 45, 131–8.

The Reward Theory of Desire in Moral Psychology  213 Ledoux, J. 2000. Emotional circuits in the brain. Annual Review of Neuroscience, 23, 155–84. Liechti, M., Saur, M., Gamma, A., Hell, D., and Vollenweider, F. 2000. Psychological and physiological effects of MDMA (‘Ecstasy’) after pretreatment with the 5-HT2 antagonist ketanserin in healthy humans. Neuropsychopharmacology, 23, 396–404. Liechti, M., and Vollenweider, F. 2000. Acute psychological and physiological effects of MDMA (‘Ecstasy’) after haloperidol pretreatment in healthy humans. European Neuropsychopharmacology, 10, 289–95. Lowry, C., Hale, M., Evans, A., Heerkens, J., Staub, D., Gasser, P., and Shekhar, A. 2008. Serotonergic systems, anxiety, and affective disorder. Annals of the New  York Academy of Sciences, 1148, 86–94. Moll, J., Krueger, F., Zahn, R., Pardini, M., Oliveira-Souza, R., and Grafman, J. 2006. Human fronto-mesolimbic networks guide decisions about charitable donation. Proceedings of the National Academy of Sciences, 103, 15623–8. Morillo, C. 1990. The reward event and motivation. Journal of Philosophy, 87, 169–86. Porrill, J., Dean, P., and Stone, J. 2004. Recurrent cerebellar architecture solves the motor-error problem. Proceedings of the Royal Society B: Biological Sciences, 271, 789–96. Putnam, H. 1975. The meaning of “meaning.” Minnesota Studies in the Philosophy of Science, 7, 131–93. Quinn, W. 1993. Morality and Action. New York: Cambridge University Press. Rawls, J. 1971. A Theory of Justice. Cambridge, MA: Harvard University Press. Rilling, J., Gutman, D., Zeh, T., Pagnoni, G., Berns, G., and Kilts, C. 2002. A neural basis for social cooperation. Neuron, 35, 395–405. Salimpoor, V., Benovoy, M., Larcher, K., Dagher, A., and Zatorre, R. 2011. Anatomically distinct dopamine release during anticipation and experience of peak emotion to music. Nature Neuroscience, 14, 257–62. Sampaio, J., Bobrowicz-Campos, E., André, R., Almeida, I., Faria, P., Januário, C., Freire, A., and Castelo-Branco, M. 2011. Specific impairment of visual spatial covert attention mechanisms in Parkinson’s Disease. Neuropsychologia, 49, 34–42. Schroeder, T. 2004. Three Faces of Desire. New York: Oxford University Press. Schroeder, T. 2005. Tourette Syndrome and moral responsibility. Philosophy and Phenomenological Research, 71, 106–23. Schroeder, T. 2006. Reply to critics. Dialogue, 45, 165–74. Schueler, G. 1995. Desire: Its Role in Practical Reason and the Explanation of Action. Cambridge, MA: MIT Press. Schultz, W., Dayan, P., and Montague, R. 1997. A neural substrate of prediction and reward. Science, 275, 1593–9. Schultz, W., Tremblay, L., and Hollerman, J. 2000. Reward processing in primate orbitofrontal cortex and basal ganglia. Cerebral Cortex, 10, 272–83. Shastri, L. 2002. Episodic memory and cortico-hippocampal interactions. Trends in Cognitive Sciences, 6, 162–8. Skinner, B. 1938. The Behavior of Organisms. New York: Appleton-Century-Crofts. Smith, M. 1994. The Moral Problem. Oxford: Blackwell. Soubrié, P. 1986. Reconciling the role of central serotonin neurons in human and animal behavior. Behavioral and Brain Sciences, 9, 319–64.

214  Timothy Schroeder and Nomy Arpaly Stampe, D. 1986. Defining desire. In J. Marks (ed.), Ways of Desire: New Essays in Philosophical Psychology on the Concept of Wanting, pp. 149–73. Chicago: Precedent. Stellar, J., and Stellar, E. 1985. The Neurobiology of Motivation and Reward. New York: Springer-Verlag. Strawson, G. 1994. Mental Reality. Cambridge, MA: MIT Press. Sutton, R., and Barto, A. 1998. Reinforcement Learning: An Introduction. Cambridge, MA: MIT Press. Thagard, P. 2006. Desires are not propositional attitudes. Dialogue, 45, 151–6. Tiberius, V. 2002. Practical reason and the stability standard. Ethical Theory and Moral Practice, 5, 339–53. Vadas, M. 1984. Affective and non-affective desire. Philosophy and Phenomenological Research, 45, 273–80. Velleman, D. 1999. Love as a moral emotion. Ethics, 109, 338–74. Watson, G. 1975. Free Agency. Journal of Philosophy, 72, 205–20.

10 Does Evolutionary Psychology Show That Normativity Is Mind-Dependent? Selim Berker

1. Introduction Suppose we grant that evolutionary forces have had a profound effect on the contours of our normative judgments and intuitions. Can we conclude anything from this about the correct metaethical theory? In particular, can we conclude anything about the ultimate grounds of normativity? I will argue that, for the most part, we cannot. I focus here on the powerful line of argument developed by Sharon Street in an important series of articles (Street 2006; 2008b; 2009a; 2011; MSa; MSb).1 Street claims that the evolutionary origins of our normative judgments and intuitions cause insuperable epistemological difficulties for a metaethical view she calls “normative realism.” I will be claiming, in reply, that there are two largely independent lines of argument in Street’s work which need to be teased apart. The first of these involves a genuine appeal to evolutionary considerations, but it can fairly easily be met by her opponents. The second line of argument is more troubling; it raises a significant problem, one of the most difficult in all of philosophy, namely how to justify our reliance on our most basic cognitive faculties without relying on those same faculties in a question-begging manner. However, evolutionary considerations add little to this old problem, and rejecting normative realism is not a way to solve it. My way of arguing for these conclusions will involve two basic strategies, deployed in tandem. First, I will be insisting that Street’s own preferred metaethical view, so-called 1 Similar evolutionary arguments have been offered by Richard Joyce (2001, ch. 6; 2006; forthcoming) and Matthew Bedke (2009; forthcoming). The original version of this chapter was a response to all three authors, under the title “The Metaethical Irrelevance of Evolutionary Theory.” However, when I returned to that piece after several years spent working on other projects, I found that I had so much to say about Street’s articles, much of it not applicable to Joyce’s and Bedke’s work, that it no longer made sense to treat all three authors as offering versions of essentially the same argument. I hope on a future occasion to give Joyce’s and Bedke’s evolutionary arguments the attention they deserve.

216  Selim Berker “Humean constructivism,” is just as threatened by her arguments as her opponents’ views are. I do so because seeing why Street is mistaken in thinking that Humean constructivism is better placed to avoid her evolutionary challenge than varieties of normative realism are will allow us to separate the two independent lines of argument in Street’s work, and will allow us to recognize which of her explanatory demands are reasonable and which unreasonable. Second, I will be paying close attention to two different dependency relations: the causation relation and the in-virtue-of relation. I do so because I believe Street’s first line of argument only looks troubling for her opponents if we ignore the second of these relations. This is ironic, because Street’s own explanation of why Humean constructivism avoids that first line of argument involves an appeal to the in-virtue-of relation. But if she is allowed to appeal to this relation, then her opponents should be allowed to appeal to it as well. By doing so, normative realists can defuse Street’s first line of argument. Of course, they must still contend with Street’s second line of argument. But so too, I will argue, must Street.

2.  Street on Realism versus Antirealism Street takes her evolutionary argument to refute a metaethical view she dubs “normative realism” and to support a metaethical view she dubs “normative antirealism.” However, Street uses these terms in a somewhat idiosyncratic manner, so it is worth pausing to get clear on what, exactly, Street takes the target of her argument to be. Following Street, let us use the expression “evaluative attitudes” to cover at least the following:  “desires,” “attitudes of approval or disapproval,” “unreflective evaluative tendencies” (such as a tendency to experience fact F as counting in favor of or as demanding ϕ-ing), and “consciously or unconsciously held normative judgments” (such as the judgment that F is a reason for agent A to ϕ in circumstance C).2 As Street sees it, the realism versus antirealism dispute in metaethics 3 turns on whether normative facts are grounded in our evaluative attitudes. More precisely, she defines realism and antirealism about normativity as follows: normative realism: There are at least some normative facts or truths that hold independently of all our evaluative attitudes. 2 See Street 2006, 110; 2008b, 226 n. 3; 2009b, 295 n. 8; and MSb, 39 n. 3. Perhaps a better term for this cluster of attitudes would be “normative attitudes,” since many of them concern what is fitting and what is required, not just what is good. (In her 2006, Street uses “evaluative” as the most general term for anything having to do with oughtness, goodness, appropriateness, and so on; in subsequent work, however, she adopts the more standard practice of using the term “normative” to play this role, with one main exception, namely continuing to call the relevant set of attitudes “evaluative attitudes.”) In the literature on the metaphysics and epistemology of intuitions, there is currently a debate over whether an intuition is a type of judgment or rather some other sort of entity, such as an intellectual seeming. (For references, see Pust 2012, §1.) Street’s category “evaluative attitude” is wide enough that, whichever way we side in this debate, normative intuitions count as evaluative attitudes. 3 In this chapter I use the term “metaethics” broadly, to cover the second order investigation of all first order norms, not just distinctively moral ones and not just those governing action.

NORMATIVITY MIND-DEPENDENT?  217 normative antirealism: There are no normative facts or truths that hold independently of all our evaluative attitudes.4

The sort of dependency at issue here is an asymmetric relation of metaphysical dependence between individual facts or truths.5 Let N be some normative fact, such as [I have reason to ϕ], and let A be some attitudinal fact, such as [I judge that I have reason to ϕ].6 To say that N depends on A, in the relevant sense, is to say any of the following: A grounds N. A makes it the case that N obtains. N obtains in virtue of A. N obtains because A obtains.

To say that N holds independently of A is to say that N does not depend, even in part, on A. This variety of dependence should be familiar from discussions of the Euthyphro dilemma.7 Socrates asked: Is an act pious because it is loved by the gods, or is it loved by the gods because it is pious? The modern, secular version of this question becomes: Do 4 See Street 2006, 110; 2008a, 207; 2008b, 208; 2009a, 214; 2009b, 274; and MSb, 2. There are two extra complications that Street occasionally adds to her definitions of realism and antirealism, but which I think we do best to ignore. First, Street sometimes defines normative realism as the view that there are at least some normative facts or truths that hold independently of all our evaluative attitudes “and what follows, as a logical or instrumental matter, from those attitudes in combination with the non-normative facts,” and qualifies her definition of normative antirealism in a similar way (Street 2009b, 274; see also Street 2009a, 214, and MSb, 2). However, there is no need to include this extra conjunct in our definitions of realism and antirealism. The fact [P follows, as a logical or instrumental matter, from our evaluative attitudes together with the non-normative facts] does not hold independently of the facts about our evaluative attitudes. Therefore the definitions of realism and antirealism without the extra conjunct are logically equivalent to the definitions with the extra conjunct. (Moreover, talk of what “follows, as a logical or instrumental matter,” from a set of evaluative attitudes plus the non-normative facts only makes sense given the truth of the metaethical view Street calls “constructivism,” so defining antirealism with the additional conjunct threatens to collapse the distinction between antirealism and constructivism, whereas Street intends constructivism to be only one variety of antirealism.) Second, Street sometimes requires normative realists to hold that at least some of the evaluativeattitude-independent normative truths concern what one has reason simpliciter to do or what one ought simpliciter to do (Street 2008b, 221–5). I think it is a mistake to build this requirement into the definition of normative realism. It is not uncommon for theorists to hold that there exist mind-independent, genuinely normative facts despite there being no facts about what one ought to do, full stop. (To give but one example: Consider the view that there are facts about what we objectively ought to do, and facts about what we subjectively ought to do, but no facts about a variety of “ought” that takes into account both objective and subjective considerations.) Even if, ultimately, such views are unsatisfactory, they should not be ruled out of court by terminological decree. 5 Henceforth I write “fact” instead of “fact or truth.” Don’t read too much into my use of the word “metaphysical” in the phrase “metaphysical dependence”; I am simply using it to flag that the sort of dependence being discussed is not of a causal, conceptual, semantic, or epistemic variety. If you prefer talk of normative dependence, feel free to use that terminology instead. 6 Throughout, I follow Gideon Rosen (2010) in using “[p]” as shorthand for “the fact that p,” and I follow Matthew Evans and Nishi Shah (2012) in using “attitudinal fact” as shorthand for “fact about evaluative attitudes.” 7 Street identifies the type of dependence at issue in the normative realism/antirealism debate with the type of dependence at issue in the Euthyphro dilemma in her 2009a, 213; 2009b, 274; 2010, 370; 2012, 41; and MSb, 40 n. 8.

218  Selim Berker I have reason to perform some act because my evaluative attitudes favor it, or do my evaluative attitudes favor that act because I have reason to perform it? Or, in other words, is normativity mind-dependent? Are reasons, values, and duties found or created? There are two main varieties of normative realism:8 naturalist normative realism: There are at least some normative facts that hold independently of all our evaluative attitudes, and all of these normative facts are either identical to or entirely grounded in natural facts.9 non-naturalist normative realism: There are at least some normative facts that hold independently of all our evaluative attitudes, and at least some of these normative facts are non-natural and ungrounded.

There are two varieties of normative antirealism: nihilist normative antirealism: There are no normative facts. non-nihilist normative antirealism: There are at least some normative facts, and all of these normative facts are at least partially grounded in facts about our evaluative attitudes.

According to Street, non-nihilist versions of normative antirealism include Bernard Williams’s account of internal reasons, David Lewis’s dispositional theory of value, Christine Korsgaard’s Kantian constructivism, and her own Humean constructivism.10 Using the labels “realism” and “antirealism” to pick out the two sides of this dispute over the mind-dependence of normativity leads—or, at least, threatens to lead—to some surprising taxonomic consequences. In particular, it seems that preference utilitarianism and the combination of ethical egoism and an actual-desire-based theory of well-being both count as antirealist on Street’s definition. Consider the following versions of such views (similar points hold for other versions): preference utilitarianism: Agent A ought to ϕ in circumstance C if and only if, and because, A’s ϕ-ing in C better serves everyone’s preferences than any alternative available to A in C. actual-desire-based ethical egoism: Agent A ought to ϕ in circumstances C if and only if, and because, A’s ϕ-ing in C better promotes A’s desires than any alternative available to A in C.

The “because” in these formulations picks out the in-virtue-of relation. Thus, according to the former view, [Agent A ought to ϕ in circumstance C] is grounded in [A’s ϕ-ing in C better serves everyone’s preferences than any alternative available to 8 I say “main” because there are other forms of normative realism beyond these two. For example, a realist view on which every normative fact is at least partially grounded in another normative fact does not count as either naturalist or non-naturalist, by these definitions. 9 In what follows I shall, for ease of exposition, ignore the varieties of naturalist normative realism that take normative facts to be identical to natural facts. Everything I go on to say about versions of naturalism formulated in terms of grounding will apply, mutatis mutandis, to versions of naturalism formulated in terms of identity. 10 Street cites Williams as an antirealist in her 2009b, 295; 2012, 42 n. 6; and MSa, 11–12; she cites Lewis as an antirealist in her 2006, 163 n. 57; 2011, 16, 31; and 2012, 42 n. 6; and she cites Korsgaard as an antirealist in nearly every article she has written.

NORMATIVITY MIND-DEPENDENT?  219 A in C], which itself is grounded in facts about everyone’s evaluative attitudes (preferences being one form of evaluative attitude). And according to the latter view, [Agent A ought to ϕ in circumstances C] is grounded in [A’s ϕ-ing in C better promotes A’s desires than any alternative available to A in C], which itself is grounded in facts about A’s evaluative attitudes (desires being one form of evaluative attitude). It follows that both of these theories qualify as forms of antirealism.11 But shouldn’t preference utilitarianism and ethical egoism be compatible with realism? In one way, it is not particularly troubling if Street’s use of the labels “realism” and “antirealism” forces us to deem preference utilitarians and ethical egoists to be antirealists. The reason these examples seem to cause trouble for Street’s taxonomy is due to the widespread belief that (i) utilitarianism and egoism are positions in normative (i.e. first order) ethics, (ii) realism and antirealism are positions in meta- (i.e. second order) ethics, and (iii) any position in normative ethics is compatible with just about any position in metaethics (except, perhaps, for metaethical views such as nihilism which entail that there is no such subject as normative ethics). But maybe we should give up on some or all of these assumptions. Maybe some normative ethical views have direct metaethical implications. Maybe mind-dependence is an issue in normative ethics, not metaethics.12 And maybe utilitarianism and egoism are best thought of as meta­ ethical positions, or as positions both in metaethics and in normative ethics. In the end, does it matter too much whether a philosophical position falls on one or the other side of the metaethics versus normative ethics divide? Indeed, does it even matter that there be a coherent metaethics versus normative ethics divide? Subdisciplinary taxonomy is not an end in itself. But there is another, more pressing reason why deeming preference utilitarianism and ethical egoism to be forms of antirealism should trouble Street. Street wants her evolutionary argument for antirealism to have a surprising conclusion. However, if preference utilitarianism and ethical egoism end up being forms of antirealism, in Street’s sense, then it will turn out that many naturalists can accept Street’s evolutionary argument without worry. Indeed, given that, according to Street (2006, 146–52), experiences of pleasure and pain are best thought of as being constituted by evaluative attitudes, it will turn out that almost all naturalists are untouched by Street’s argument, since there are very few naturalists who ultimately ground normative facts in something other than conative states such as desires and experiential states such as pleasure

11 I assume here that the grounding relation is transitive. I also assume that these theories either deny that there are any other normative facts beyond the ones within their scope, or ground those other normative facts in the same sorts of considerations that ground facts about what one ought to do (for example, grounding [A is epistemically justified in believing that p] in facts about everyone’s preferences or facts about A’s desires; see Kornblith 1993 and Petersen 2013 for views of this sort). 12 As Simon Blackburn, Ronald Dworkin, and Allan Gibbard have all urged. (Street expresses a willingness to concede this point to Blackburn, Dworkin, and Gibbard for the sake of argument in multiple places; see her 2008b, 227 n. 23; 2009a, 215; 2009b, 295 n. 9; 2010, 378; and MSb, 9, 40 n. 8.)

220  Selim Berker and pain.13 Naturalist normative realism will be a position in logical space, but one not occupied by any practicing philosophers. It is for this reason, I suspect, that Street in effect embraces a second way of responding to the taxonomic puzzle I have raised.14 On this approach, we insist that facts about what grounds normative facts themselves count as normative facts. Then whether a given form of preference utilitarianism is a realist or antirealist position will depend on whether the grounding fact [[A ought to ϕ in C] is grounded in [A’s ϕ-ing in C better serves everyone’s preferences than any alternative available to A in C]] is itself grounded in facts about evaluative attitudes. And similarly for forms of ethical egoism that embrace actual-desire-based theories of well-being: They will be compatible with both realism and antirealism, depending on whether a certain grounding fact is itself grounded in facts about evaluative attitudes. This move saves the debate, but it is not without its costs. I mention here just two. First, it commits the antirealist to an infinite hierarchy of grounding facts. For every normative fact N, not only must N be grounded in at least one attitudinal fact, but that first order grounding fact must itself be grounded in at least one attitudinal fact, that second order grounding fact must itself be grounded in at least one attitudinal fact, and so on, ad infinitum. Thus if Ai are all attitudinal facts, we have the following:15 [N ← A1], [[N ← A1] ← A2], [[[N ← A1] ← A2] ← A3], . . . .

The existence of such an infinite hierarchy of grounding facts—not groundings “all the way down,”16 but rather groundings “all the way out”—might make some queasy. For

13 Knut Skarsaune makes a related point in his 2011, 240–1. 14 My evidence that Street embraces this reply is threefold. First, she explicitly states that whether varieties of naturalism formulated in terms of natural–normative identities count as realist or antirealist depends on whether those identities are themselves grounded in attitudinal facts (Street 2006, 135–41; 2008b, 223). The natural extension of this proposal to varieties of naturalism formulated in terms of natural–normative grounding claims is to say that whether such views count as realist or antirealist depends on whether the relevant natural–normative grounding claims are themselves grounded in attitudinal facts. Second, one of Street’s ways of replying to a common objection to her evolutionary argument crucially relies on the claim that facts about the grounds of normative facts are substantive normative facts (see §5 in this chapter). Third, Street’s own metaethical view, Humean constructivism, is a proposal about the ultimate grounds of all normative truths (see §6), and she regularly treats that view as itself a normative truth grounded in the very thing which, according to that view, grounds all normative truths (see Street 2009a, 216 n. 7; 2010, 378, 382 n. 16; MSa, 14–17; and MSb, 36–8). 15 Throughout, I use “F1 ← F2” as shorthand for “F1 is at least partially grounded in F2.” (This is a slight variant of Rosen’s conventions in his 2010.) 16 Groundings “all the way down” would have the following form: [N ← A1], [A1 ← A2], [A2 ← A3], . . . .

NORMATIVITY MIND-DEPENDENT?  221 instance, we might wonder whether there are enough evaluative attitudes to ground everything in this infinite hierarchy.17 A second cost of this move is that Lewis and Williams, two of Street’s canonical examples of antirealists, now no longer count as antirealists, in her sense. Lewis can plausibly be read as grounding facts about value in attitudinal facts, and Williams can plausibly be read as grounding facts about reasons for action in attitudinal facts.18 But neither Lewis nor Williams can plausibly be read as holding that the fact that value or reason facts are grounded in attitudinal facts is itself grounded in attitudinal facts.19 I have gone through this excursus on Street’s definitions of realism and antirealism for two reasons. First, it shows just how strong a view Streetian antirealism is. Indeed, I doubt that anyone other than Street has ever defended a non-nihilist version of normative antirealism, in her sense. This makes it all the more impressive if 17 A partial reply to this worry: Because we are dealing with groundings “all the way out” and not “all the way down,” allowing that Ai = Aj when i ≠ j does not commit us, via transitivity, to a fact partially grounding itself. Thus it is open to the antirealist to ground this infinite hierarchy of grounding facts in a finite number of attitudinal facts. (And, as we shall see, this is precisely what Street proposes: for Street, the same attitudinal facts which ground [N ← A1] also ground each subsequent fact in our regress. See n. 54.) 18 For a contrary interpretation of Williams, see Manne 2014, 106. 19 Given these two costs, some might wonder whether the medicine I have offered Street, and which I believe she herself takes (see n. 14), is worse than the disease. Maybe Street should pursue one of the following strategies instead:  Insist that facts about the grounds of normative facts are not themselves normative, and try to find a way of rephrasing the definitions of realism and antirealism so that preference utilitarianism and ethical egoism are deemed to be compatible with either realism or antirealism, whereas Lewis’s and Williams’s views are automatically categorized as antirealist. •  Insist that the because/grounding relation featured in first order ethical theories is distinct from the because/grounding relation employed in metaethics (and hence in Street’s definitions of realism and antirealism), and hold that only facts concerning the extension of the former relation count as normative facts. •

I am dubious of the assumptions that underlie both of these proposals. (I doubt that facts about the grounds of normative facts are somehow not normative despite entailing the normative facts being grounded, and I doubt that we have two sorts of grounding relation here, one “internal” to first order ethical theory and the other “external” to it; for more discussion of this second issue, see Berker MSc.) But even if we put my reservations to one side, these two strategies will not succeed in what is presumably their primary aim, which is to avoid saddling Street with a commitment to an infinite hierarchy of groundings “all the way out.” For as we shall see, even if facts about the grounds (or groundsmetaethics) of normative facts are not themselves normative, facts of this sort—because they lack causal powers—will be just as susceptible to Street’s Darwinian challenge as normative facts are. Hence if Street’s argument for the attitude-dependence of normative facts is sound, a similar argument can establish that facts about the grounds (or groundsmetaethics) of normative facts are attitude-dependent, whether or not those facts count as normative. Thus even if antirealism as such is not committed to an infinite hierarchy of groundings “all the way out,” the sort of antirealism motivated by Street’s evolutionary argument is so committed. (For similar reasons, even if we can find a way of reformulating antirealism so that Williams and Lewis count as antirealists, it will still turn out that Williams’s and Lewis’s views fall within the target of Street’s Darwinian argument, since Williams and Lewis do not take their claims about what certain normative facts depend on to themselves be attitude-dependent claims.) In order to avoid having to distinguish between antirealism as such and the sort of antirealism motivated by Street’s evolutionary argument, I will, in what follows, work under the assumption that facts about the grounds of normative facts qualify as normative facts, and under the assumption that the sort of grounding or in-virtue-of talk at issue in metaethics is not distinct from the sort of grounding or in-virtue-of talk at issue in normative ethics.

222  Selim Berker Street’s evolutionary argument can show that normative antirealism is true. Second, this excursus establishes what will become a common theme in this chapter: that when evaluating Street’s argument, what grounds grounding facts is where most of the philosophical action is.

3.  Street’s Darwinian Dilemma for Normative Realists Street’s evolutionary argument against normative realism starts from the following premise: the Darwinian hypothesis: Natural selection and other evolutionary factors have had a tremendous influence on the content of our evaluative attitudes.

Given this hypothesis, Street holds that realists must take a stand on the relation between the evolutionary forces that have influenced the content of our evaluative attitudes and the attitude-independent normative truths posited by the realist. This leads to a dilemma for the realist: first horn (pushing-toward horn): Hold that evolutionary forces have tended to push our normative judgments (and other evaluative attitudes) toward the attitude-independent normative truth. second horn (at-best-random horn): Hold that evolutionary forces have tended to push our normative judgments (and other evaluative attitudes) either away from or neither away from nor toward the attitude-independent normative truth.20

Talk of pushing here is of course metaphorical:  The crucial issue is whether evolutionary forces have tended to influence our judgments about reasons, values, duties, and other normative matters in such a way as to make them line up with the attitude-independent facts about such matters. The first horn holds that this is the case; the second horn holds that it is not.21

20 In what follows, I drop the “and other evaluative attitudes” qualifiers in these horns and restrict my discussion to evolution’s impact on our normative judgments; parallel issues arise for other types of evaluative attitudes. 21 Let us use “EF” as shorthand for “the evolutionary forces that have influenced the content of our evaluative attitudes” and “NT” as shorthand for “the attitude-independent normative truths posited by the realist.” In her early work, Street characterizes her dilemma by saying that, on the first horn, the realist “asserts [or affirms] a relation” between EF and NT, whereas, on the second horn, the realist “denies a relation” between EF and NT (Street 2006, 121; 2008, 208–9). I find this way of describing the horns of Street’s dilemma unhelpful. In the logician’s sense of “relation,” there is always a relation between EF and NT: There always exist infinitely many two-place relations that hold between EF and NT. But if we use “relation” in a more restricted sense, then it is far from clear what relation Street means to be picking out. (David Copp makes a similar point in his 2008, 205 n. 9.) Luckily, in her later work, Street clarifies which relation she has in mind, namely the pushing-toward relation I have invoked in my formulation of her dilemma (Street 2008b, 208, 226 n. 4; 2011, 12).

NORMATIVITY MIND-DEPENDENT?  223 According to Street, the problem with the first horn is empirical (Street 2006, 125– 35; 2008b, 209; 2009a, 234–6; 2011, 12–13). Street claims that realists who embrace this horn are forced to endorse the following explanation: the tracking account: Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth because it promoted our ancestors’ reproductive success to make true normative judgments (or to make proto versions of them).

But Street thinks the tracking account is bad science; she insists that a far more scientifically respectable account—in terms of parsimony, clarity, and degree of illumination—is the following: the adaptive-link account: Evolutionary forces have pushed us toward making certain normative judgments because (i) making (proto versions of) those judgments made our ancestors more likely to act in accordance with them, and (ii) it promoted reproductive success to act in those ways.

Suppose evolutionary factors are partially responsible for my judging the proposition to be true.22 Figure 10.1 shows the sort of dependency structure put forward by the tracking and adaptive-link accounts for the fact that I make this judgment.23 The tracking account postulates the existence of normative facts, whereas the adaptive-link account does not, so the tracking account is less parsimonious than the adaptive-link account (Street 2006, 129). The tracking account leaves it mysterious how the truth of certain normative facts could make judgments about the obtaining of those facts reproductively advantageous, whereas the adaptive-link account contains no such obscurities, so the tracking account is less clear than the adaptive-link account (Street 2006, 129–32). And, when we look at the full pattern of normative judgments influenced by evolutionary factors, the adaptive-link account reveals an illuminating unity to these judgments which the tracking account is unable to detect (Street 2006, 132–4). All told, Street argues, the tracking account does not fare well in comparison to the adaptive-link account, as a purely empirical matter. So perhaps the realist is better off embracing the second horn of Street’s dilemma, according to which evolutionary forces have tended to push our normative judgments in ways that are at best random with respect to the attitude-independent normative truth. Here Street thinks the problem is not empirical but epistemological (Street 2006, 121–5; 2008b, 208–9; 2009a, 233–4; 2011, 13–14). She writes: As a purely conceptual matter, the independent normative truth could be anything. . . . But if there are innumerable things such that it’s conceptually possible they’re ultimately worth pursuing, and yet our [normative judgments] have been shaped from the outset by forces that are as good as random with respect to the normative truth, then what are the odds that our [normative

22 Throughout, I use “

” as shorthand for “the proposition that p.” 23 In all figures in this chapter, dashed arrows represent the “tends to (at least partially) cause” relation, and solid arrows represent the “tends to (at least partially) ground” relation.

224  Selim Berker Tracking account:

Adaptive-link account:

[I judge ]

[I judge ]

[Judging promoted reproductive success in my ancestors’ environment]

[Judging promoted reproductive success in my ancestors’ environment]

[My ancestors had conclusive reason to φ]

[Judging [φ -ing promoted reproductive success made my ancestors more likely to φ] in my ancestors’ environment]

Figure 10.1 

judgments] will have hit, as a matter of sheer coincidence, on those things which are independently really worth pursuing? (Street 2011, 14)

Thus on the second horn the realist is forced to embrace the “skeptical conclusion” that “our normative judgments are in all likelihood hopelessly off track” (Street 2008b, 208). Since this horn is just as unpalatable as the first one, and since the realist has no option but to choose one of them, Street concludes that normative realism is false. Street presents her argument as if it is an argument for normative antirealism. But she does not actually argue against the skeptical conclusion which threatens the realist on the second horn of her dilemma: She simply dismisses this possibility as “implausible” (Street 2006, 109, 122; 2008b, 209; 2011, 14; MSb, 18, 21, 35) or “unacceptable” (Street 2006, 135; 2008b, 211; 2009a, 228, 238; MSa, 11; MSb, 37). This is too quick. Skepticism with regard to normative matters—much like skepticism about other matters (the external world, induction, and so on)—is a legitimate theoretical possibility that must be reckoned with, not casually brushed aside. So really Street’s evolutionary argument is for a disjunctive conclusion: Either normative antirealism is true, or normative skepticism is true (where the former is a metaphysical thesis about the grounds of normative truths, the latter an epistemological thesis about our inability to know such truths). No matter, though: It would still be an incredibly significant contribution if even this disjunctive conclusion could be established.24

24 At times Street gestures toward what might seem to be an argument against normative skepticism, when she claims that “one must reject [normative skepticism] if one is to go on making normative judgments at all” (Street MSb, 35; see also Street 2010, 383 n. 60, and 2011, 16). However, this is not an argument against the truth of normative skepticism; rather, it is an argument against the rational coherence of accepting normative skepticism while continuing to make normative judgments.

NORMATIVITY MIND-DEPENDENT?  225

4.  Complications with the First Horn The general thrust of Street’s argument is clear, but once we look under the surface, complications arise. In particular, there are two problems with the pushing-toward horn of Street’s dilemma.25 First problem: the adaptive-link account is inadequate as it stands. As applied to the proposition , the adaptive-link account starts by assuming a version of motivational internalism (Street 2006, 157 n. 13; 2008a, 228 n. 37, 230; 2010, 376): (MI) Necessarily, if an agent judges , then she is at least somewhat motivated to ϕ.26

Then the thought is that if ϕ-ing promoted reproductive success in our ancestors’ environment, those of our ancestors who judged did better at propagating their genes than those who didn’t. However, this explanation is not fully satisfactory. First, to apply the adaptive-link account across the board, we need a version of motivational internalism to hold with regard to every normative proposition. But it is far from clear that all normative claims have a distinct necessarily accompanying motivational shadow. For example, consider the following propositions: ; ; .27 25 There are also problems with the at-best-random horn, but I pass them over on this occasion. First issue: Does Street think our normative judgments fail to be justified or to constitute knowledge if they in fact do not track the normative truth, or do they only fail to be justified or to constitute knowledge if we realize that our normative judgments do not track the normative truth? Second issue: What exactly is this tracking relation? The textual evidence suggests that Street takes the relevant tracking relation to be a Nozick-style sensitivity condition (if it were false that p, S would not judge that p), but where instead of evaluating the relevant subjunctive conditional in terms of metaphysically possible worlds, as is customary, we instead evaluate it in terms of conceptually possible worlds. (Presumably this last bit is to avoid the consequence that judgments about metaphysically necessary propositions automatically satisfy the tracking requirement, since subjunctive conditionals with impossible antecedents are trivially true.) But requiring our judgments to satisfy such a constraint in order to be justified or in order to constitute knowledge leads to deeply counterintuitive consequences: Some of these consequences are the familiar ones that bedevil all sensitivity views, and some of these consequences are new ones arising from the appeal to conceptual possibility. I hope to discuss these matters in greater detail in future work. 26 Street stresses the importance of motivational internalism for the adaptive-link account, but in fact even motivational externalists can endorse a version of the adaptive-link account, as long as they hold that judging is highly correlated in the actual world with being at least somewhat motivated to ϕ. 27 Why the first of these is problematic: It’s one thing to claim that an agent’s motivations must reflect her all-things-considered verdicts about her reasons, but quite another to claim that every single judgment about an overridden pro tanto reason must have a motivational residue. Why the second of these is problematic: Motivation’s scalar structure does not map well onto non-scalar categories such as the merely permitted. Why the third of these is problematic: If the adaptive-link account explains why we evolved to have a tendency to judge in addition to having a tendency to judge , then the motivational profile of the former must be different from the motivational profile of the latter. But it is not clear that they are. (Note that parallel problems arises even

226  Selim Berker Second, the adaptive-link account leaves it mysterious why we didn’t evolve merely to have the relevant motivations on their own, without any accompanying normative judgments. If the only evolutionary value of normative judgments are their motivational effects, wouldn’t it have been much cheaper for evolutionary forces to instill in us those motivational effects by themselves without any accompanying judgments (or proto-judgments)?28 I am not denying that there is an acceptable evolutionary explanation of our tendency to make certain normative judgments. I am not even denying that the mechanisms highlighted by the adaptive-link account could play an important role in this more complete explanation.29 My point, rather, is that the more complete evolutionary explanation will lack the beguiling simplicity of the adaptive-link account. The second problem with the pushing-toward horn of Street’s dilemma is more serious:  Street’s argument for why realists who embrace this horn must accept the tracking account rests on an equivocation. Here is what Street says: The only way for realism both to accept that [our evaluative] attitudes have been deeply influenced by evolutionary causes and to avoid seeing these causes as distorting is for it to claim that these causes actually in some way tracked the alleged independent truths. There is no other way to go. To abandon the tracking account . . . is just to adopt the view that selective pressures either pushed us away from or pushed us in ways that bear no relation to these [normative] truths. (Street 2006, 134–5)

But this passage uses the label “tracking account” in a broader way than Street uses it elsewhere. In these sentences Street is understanding the tracking account as follows: tracking account (in the broad sense): Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth.

In contrast, when she first introduces the tracking account (Street 2006, 125–6), and when she argues that it is scientifically unacceptable (Street 2006, 126–34), Street understands the tracking account more narrowly: tracking account (in the narrow sense): Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth because it promoted our ancestors’ reproductive success to make true normative (proto) judgments.30 for versions of the adaptive-link account formulated in terms of motivational externalism: The motivational profiles here need not be necessary ones.) 28 Michael J. Deem makes a similar point in his 2012. See also Parfit 2011, 2: 527–8. 29 For example, maybe we can make Street’s adaptive-link account more empirically adequate by appending to it an account which appeals to Allan Gibbard’s idea of normative governance, according to which our linguistically infused normative judgments have a tendency to influence each other’s normative judgments through discussion and avowal. See Gibbard 1990, ch. 4. 30 In fact, sometimes Street understands the tracking account more narrowly still: tracking account (in the even more narrow sense): Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth because it promoted our ancestors’ reproductive success to make true attitude-independent normative (proto) judgments.

NORMATIVITY MIND-DEPENDENT?  227 According to the first horn of the Darwinian dilemma, evolutionary forces have tended to push our normative judgments toward the attitude-independent normative truth. If we take “pushing our normative judgments toward the normative truth” to be equivalent to “making our normative judgments track the normative truth,” then this horn does indeed entail the broad tracking account—does indeed entail that evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth. But this entailed fact is compatible with any number of stories concerning in virtue of what evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth. The narrow tracking account is only one of those many stories.31 Therefore Street has not successfully shown that the first horn forces realists to accept a tracking account in the narrow sense, which is what she needs for her argument to go through. We can formulate this problem as a dilemma: Either we understand the tracking account in the broad sense, in which case Street has not provided an argument that the tracking account is scientifically unacceptable, or we understand the tracking account in the narrow sense, in which case Street has not established that realists who take the first horn of her dilemma must endorse the tracking account. Either way, Street’s first horn requires sharpening before it can impale the realist.32

5.  The Third-Factor Response One popular way of resisting Street’s Darwinian dilemma is to exploit the opening left by the second problem I have just mentioned for its pushing-toward horn. The typical way of doing this is to offer, instead of a narrow tracking account, an explanation of the following form: a third-factor account: Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth because, for each normative judgment influenced by evolution in this way, there is some third factor, F, such that (i) F tends to causally (help) make it the case that (proto) judging in that way promotes reproductive success (when in our ancestors’ environment), and (ii) F tends to metaphysically (help) make it the case that the content of that judgment is true.33 Street uses “tracking account” to pick out the narrow tracking account in her 2006, 109, 125–8, 129–34, 151; 2009a, 240, 242; and 2011, 13; and she uses “tracking account” to pick out the even-more-narrow tracking account in her 2006, 129, 141, 151, 154; 2008b, 209–10; 2009a, 234, 241; and 2011, 12, 13. The only place where Street uses “tracking account” to pick out the broad tracking account is in the one place where she explicitly argues that realists who embrace the first horn of her Darwinian dilemma must endorse the tracking account, in her 2006, 134–5. 31 Copp makes a similar point, when he distinguishes between what he calls “the tracking thesis” and “the tracking account” (Copp 2008, 194–5). 32 In her 2008b, 209, 211, Street seems to recognize that a narrow tracking account is not the only option on the first horn of her Darwinian dilemma. However, in her 2009a, 234, and 2011, 12–13, she goes back to saying that a narrow tracking account is the only option on that horn. 33 I assume here, for convenience, that facts are among the causal relata. If you think that, say, only events can be causes and effects, feel free to reformulate everything I say accordingly.

228  Selim Berker The first person to offer a third-factor account in response to an argument much like Street’s was Robert Nozick, who in his 1981 book Philosophical Explanations wrote: The ethical behavior will serve inclusive fitness through serving or not harming others, through helping one’s children and relatives, through acts that aid them in escaping predators, and so forth; that this behavior is helpful and not harmful is not unconnected to why (on most theorist[s’] views) it is ethical. The ethical behavior will increase inclusive fitness through the very aspects that make it ethical, not as a side effect through features that only accidentally are connected with ethicality. (Nozick 1981, 346)

Third-factor accounts of various forms have also been offered by Kevin Brosnan (2011), David Copp (2008), David Enoch (2010), Knut Skarsaune (2011), and Erik Wielenberg (2010).34 As I see it, there are two basic thoughts behind third-factor accounts. The first is that we don’t, in general, need to posit a direct dependency relation between a judgment that p and the fact that p in order to explain why it is not a cosmic coincidence that the former tracks the latter. Certainly a direct dependency relation will do: If my judgment that p is caused by the fact that p, or if my judgment that p causes the fact that p, then it is no mystery why judgment tracks fact. But this is not the only option. For example, a common-cause structure will suffice as well. Suppose you intend to stay home sick tomorrow and tell me so; as a result, I judge that you will stay home sick tomorrow. When tomorrow you do in fact stay home, it is no mystery why my judgment that you would tracked the truth, but in this case we do not have a direct causal link: My judgment did not cause you to stay home (I don’t have this much influence on you, alas), and your staying home did not cause my judgment the day before (no backwards causation here). Rather, a common-cause structure explains the tracking relation between judgment and fact: Your intention to stay home both caused my judgment and caused the content of that judgment to be true.35 The second thought behind third-factor accounts is that, when explaining why a given judgment tracks a given fact, any sort of a dependency relation is enough: We need not restrict ourselves to the causation relation. In particular, the grounding relation (and its converse, the in-virtue-of relation) will do. Thus in order to explain why my judgment that p tracks the fact that p, we can take a common-cause structure and swap in the grounding relation for one of the two causal relations. What results, when

is normative, is a third-factor account (see Figure  10.2):  We posit some non-normative third fact on which my judgment that p causally depends and on which the fact that p metaphysically depends. Doing so allows us to explain tracking relations

34 These authors all offer third-factor accounts as ways of defending realism against Street’s evolutionary challenge. Simon Blackburn (MS) and Jamie Dreier (2012; see esp. p. 283) in effect offer third-factor accounts as ways of defending quasi-realism against Street’s challenge. The label “third-factor account” originates, I believe, in Enoch 2010. 35 I assume here, for illustrative purposes, that an intention causes (rather than partially constitutes) its corresponding action.

NORMATIVITY MIND-DEPENDENT?  229 A third-factor account: [I judge ]

[Judging promoted reproductive success in my ancestors’ environment]

[I have conclusive reason to φ ]

F

Figure 10.2 

between normative facts and normative judgments, without taking normative facts to have causal powers. Several comments about third-factor accounts are in order. First, there is no need, in a third-factor account, to hold that the third factor entirely grounds the relevant normative fact; a relation of partial grounding is enough. After all, in more mundane cases in which one explains why a non-normative judgment tracks a non-normative fact by appealing to a direct causal relation between judgment and fact, there is no need for the fact to cause the judgment entirely on its own, or for the judgment to cause the fact entirely on its own; partial causation is, in most cases, enough. (Hence the “help” qualifiers in my formulation of third-factor accounts.) Second, given that the sort of tracking at issue here is not perfect tracking, but rather fairly good tracking, we need not postulate that our third factor, when present, always causes the judgment in question and always grounds the fact judged to obtain; a strong enough tendency to cause the judgment and to ground the fact will do. (Hence the appeal to tendencies in my formulation of third-factor accounts.) Third, once we grasp the general idea behind third-factor accounts, we can see that really they are a template for a whole host of different accounts that posit a complex dependency structure involving both causal and grounding links. For example, depending on how the evolutionary facts pan out, it may well be more plausible to posit the following instead of a strict third-factor account: A fourth-factor account: Evolutionary forces have tended to make our normative judgments track the attitude-independent normative truth because, for each normative judgment influenced by evolution in this way, there is some factor, F, and some factor, F*, such that (i) F tends to causally (help) make it the case that (proto) judging in that way promotes reproductive success (when in our ancestors’ environment), (ii) F* tends to metaphysically (help) make it the case that the content of that judgment is true, and (iii) F and F* stand in a suitable causal relation with one another (where identity is treated as the limiting case of a causal relation).

230  Selim Berker This proposal involves adding intricacies to the causal side of a third-factor account. We might also, in addition, complicate the normative end of things. For instance, we might posit that our fourth factor, F*, (tends to) partially ground a normative fact which (tends to) stand in various partial grounding relations—either upwards or downwards—with the normative fact being judged to obtain.36 (A fifth-factor account?) And so on: The possibilities are legion.37 Thus third-factor accounts, and others of their ilk, represent an extremely versatile way of embracing the first horn of Street’s dilemma without resorting to a narrow tracking account. This versatility explains, I believe, why so many authors have—apparently independently—hit upon this way of replying to Street. Indeed, if you don’t think that normative facts have causal powers, a third- (or fourth-, or . . .) factor account of how our normative judgments track the normative truth despite being heavily influenced by evolutionary forces is almost inevitable. But this is not the end of the story, for we can find in Street’s work two very powerful objections to third-factor accounts.38 According to the first of these objections, Street can simply rerun her Darwinian dilemma “one level up.”39 The third-factor theorist relies on at least one claim of the following form: (G) Non-normative fact F (at least partially) grounds normative fact N.

Being a realist, the third-factor theorist presumably thinks that (G)’s truth does not depend on any facts about our evaluative attitudes.40 But now we can ask: How does

36 Enoch (2010, 431)  also broaches the possibility of a broadly third-factor account with this sort of structure. 37 There are difficult questions here about exactly which dependency structures of the sort I have sketched are sufficient to underwrite a tracking relation between judgment and fact. How complicated can we make the overall dependency structure, and how tight must each link in that dependency structure be? Taking a stand on these issues, however, would require us to give a substantive account of what tracking comes to, and that is not a task I have undertaken in this chapter. (See n. 25.) Note, also, that these difficult questions arise even when we have a purely causal dependency structure underwriting a tracking relation; they are not unique to dependency structures involving both causal and grounding links. 38 Henceforth I  use “third-factor account” as a general term encompassing third-factor accounts, fourth-factor accounts, and all others of that sort. 39 See Street 2006, 135–41. Street explicitly formulates the objection to be considered as an objection to versions of naturalist normative realism that are framed in terms of natural–normative identities. However, a similar objection can be offered against versions of naturalist normative realism framed in terms of grounding relations between natural and normative facts, and this latter objection is easily extended into an objection against all third-factor accounts. 40 Actually, there is a problem for Street here. Strictly speaking, a normative realist might hold that N is not grounded in any attitudinal facts, even though (G)—a fact about what grounds N—is itself grounded in attitudinal facts. (In general, it is not true that if F1 partially grounds the fact [F2 partially grounds F3], then F1 partially grounds F3.) So Street is not entitled to the assumption she needs to get this objection running, namely that the realist must deem (G) to be attitude-independent. (A similar problem arises for the version of this objection directed against varieties of naturalist realism formulated in terms of identity, since in general it is not true that if F1 partially grounds the fact [F2 = F3], then F1 partially grounds F3.) But even though I believe there is a problem here, I shall ignore it in what follows. The form of normative realism that Street overlooks when offering her objection is extremely recherché, and it would be better if third-factor theorists could provide their reply to Street without committing themselves to such an unusual view.

NORMATIVITY MIND-DEPENDENT?  231 the third-factor theorist know that (G) is true? Whatever method we use to judge that (G) is true (whether it be reflective equilibrium or some other method) is, no doubt, greatly saturated with evolutionary influences. And this raises the familiar question: What, according to the realist who resorts to a third-factor response, is the relation between the evolutionary forces that have influenced our judgment that (G) is true and the evaluative-attitude-independent fact that (G) is true? On the one hand, if evolutionary forces have tended to push us toward making a correct verdict about (G)’s truth, then—Street would insist—the only explanation open to the realist of why this is so is a tracking account, which loses out to the more scientifically acceptable adaptive-link account. On the other hand, if evolutionary forces have tended to push us in ways that are at best random with respect to the truth of (G), then—Street would insist—we are in all likelihood wrong in judging (G) to be true. In other words, the third-factor theorist is caught in a Darwinian dilemma, one level up.41 Street’s second powerful objection to third-factor accounts rests on the charge that such accounts are “trivially question-begging” (Street 2008b, 214–17; 2011, 17–19; MSb, 17–30). Here is how Street puts the point, in response to Copp’s version of a third-factor account: It is no answer to [the Darwinian] challenge simply to assume a large swath of substantive views on how we have reason to live . . . and then note that these are the very views evolutionary forces pushed us toward. Such an account merely trivially reasserts the coincidence between the independent normative truth and what the evolutionary causes pushed us to think; it does nothing to explain the coincidence. (Street 2008b, 14)42

Does Street’s objection here mean that, when it comes to truths of other sorts, we are also prohibited from appealing to substantive truths of that sort when explaining how we were selected, either directly or indirectly, to track those truths? And wouldn’t such a ban lead to universal skepticism—not just skepticism about our ability to track facts about reasons (if normative realism is true), but also skepticism about our ability to track tables, chairs, and all other mind-independent entities? Not so, says Street. She distinguishes two sorts of explanations we might give as to why evolutionary 41 Note that although it is natural to assume that (G) is normative when offering a “one level up” objection against it, the objection in question does not actually depend on the assumption that (G) is normative. Rather, all it relies on (in addition to the assumption that (G) is attitude-independent; see n. 40) is the assumption that (G) does not stand in causal relations with other facts, which is independently plausible even if (G) is non-normative. Similarly, even though third-factor theorists usually appeal to a version of (G) in which the relevant grounding relation is the sort employed in first order ethical theories, Street’s “one level up” objection applies just as much to versions of (G) in which the grounding relation is of a distinctively metaethical sort, since—if there is such a relation—facts about its extension do not have causal powers. (This is why I said in n. 19 that neither the strategy of denying that facts about the grounds of normative facts are normative nor the strategy of distinguishing between an “internal” normative-ethical grounding relation and an “external” metaethical grounding relation will avoid committing Street to an infinite hierarchy of groundings “all the way out.”) 42 Note that, in offering this reply to third-factor theorists, Street is assuming (G) to be a substantive normative view. This is my second piece of evidence, mentioned in n. 14, that Street takes facts about the grounds of normative facts to be normative.

232  Selim Berker forces have made us track truths about the presence of midsized objects in our immediate environment: Account A: There are six chairs, a laptop, and a table in my immediate environment. But evolutionary forces gave rise to the capacity I used to make this very judgment. This gives me reason to think my capacity about midsized objects in my immediate environment is reliable. Account B: Midsized objects in our immediate environment are the kinds of things one can run into, be injured by, eat, and be eaten by. Other things being equal, then, creatures with an ability accurately to detect midsized objects in their immediate environment tended to survive and reproduce in greater numbers than creatures who lacked this ability. I am a product of this evolutionary process. This gives me reason to think my capacity to make judgments about midsized objects in my immediate environment is reliable. (Street 2008b, 216–17; cf. Street 2011, 19, and MSb, 25)

Street concedes that there is a sense in which Account B is question-begging, since many of its claims (that midsized objects in our immediate environment can injure and eat us, that evolutionary processes occur in the world around us, and so on) are ones we came to believe partly through use of our faculties for making judgments about the presence of midsized objects in our immediate environment. Nevertheless, Street maintains, there is an epistemic contrast between A and B: Account B is, according to Street, ultimately question-begging but still gives us good reason to think we’re reliable on these matters, whereas Account A is trivially question-begging and gives us no reason to think we’re reliable. Moreover, Street insists that third-factor accounts are of the same form as Account A. Thus things do not look good for the normative realist. At first it seemed that third-factor accounts would present the realist with a promising way of grasping the pushing-toward horn of Street’s dilemma without committing to a scientifically indefensible (narrow) tracking account of the relation between evolutionary forces and the attitude-independent normative truth. But now we have seen two formidable objections to third-factor views: They appear to be susceptible to a version of Street’s dilemma one level up, and they appear to question-beggingly assume the very thing they aim to show. In fact I do not think that matters are nearly so dire for the realist. But before I argue why this is so, it will help to take a detour through Street’s explanation of why her own metaethical view avoids the Darwinian dilemma.

6.  Street’s Solution to the Darwinian Dilemma According to Humean constructivism, Street’s preferred variety of antirealism, (G′) The non-normative fact [A judges , and her ϕ-ing does not conflict with anything else she more deeply judges that she has reason to do] grounds the normative fact [A has conclusive reason to ϕ].43 43 See Street 2008a, esp. 234–6. (G′) follows, by the transitivity of the grounding relation, from two more specific claims that Street makes: (G′′) [A judges , and her ϕ-ing does not conflict with anything else she more deeply judges that she has reason to do] grounds [A’s judgment

NORMATIVITY MIND-DEPENDENT?  233 How does such a view avoid the Darwinian dilemma? In fact, we find in Street’s work two different explanations of why antirealists such as the Humean constructivist can meet her evolutionary challenge. The first of these I call the grounding account, and the second I call the theoretical-reasoning account.44 According to the grounding account, our judgments about our reasons tend to track the truth about what reasons we have because the former is what grounds the latter.45 Suppose (G′) is true, and suppose A’s judgments about what she has conclusive reason to do don’t conflict too greatly with one another. It follows that A’s judgments about what she has conclusive reason to do are mostly true, and hence she is not “hopelessly off track” in making such judgments. Moreover, this will be so regardless of the evolutionary story we tell about A’s tendency to make judgments of this sort: Given any causal story “upstream” in the order of explanation from the judgments themselves, A’s judgments about what she has conclusive reason to do will—cases of conflict aside—make themselves true “downstream” in the order of explanation. In other words, if Humean constructivism is correct, then our judgments about our own reasons mostly track the truth, because they serve as the ground of their own truth. Thus the Darwinian dilemma is avoided, at least with respect to such normative judgments.

withstands scrutiny from the standpoint of A’s other normative judgments]. (G′′′) [A’s judgment withstands scrutiny from the standpoint of A’s other normative judgments] grounds [A has conclusive reason to ϕ]. Note that (G′) is not the only way in which, for a Humean constructivist, non-normative facts about evaluative attitudes can ground normative facts about reasons. For example, Street is also committed to (G*) [A judges , ψ-ing is a necessary means to ϕ-ing, and A’s ψ-ing does not conflict with anything else she more deeply judges that she has reason to do] grounds [A has conclusive reason to ψ]. According to the Humean constructivist, the ultimate grounds of a person’s reasons are always of the sort found in (G′) and (G*): they concern that person’s judgments about her reasons, how deeply those judgments are held, and the instrumental connections between the actions (or attitudes) which the person judges that she has reason to perform (or hold). So for illustrative purposes I will simply work with (G′) in the body of the text. 44 In her early work, Street tends to stick to the grounding account: See, in particular, Street 2006, 153–4, where she explicitly states that the grounding account is the official explanation of why antirealists escape the Darwinian dilemma. In her more recent work, however, Street slides back and forth between the grounding and theoretical-reasoning accounts (sometimes on the very same page): See Street 2011, 22 n. 41, and MSb, 12, for endorsements of the grounding account, and see Street 2011, 16, and MSb, 12, 30, for endorsements of the theoretical-reasoning account. 45 A qualification here is in order. Sometimes Street says that the normative truth obtains in virtue of (i.e. is grounded in) the facts about our judgments about reasons (as she does throughout Street MSa), and other times she says that the normative truth is constituted by the facts about our judgments about reasons (as she does throughout Street 2008a). I think it is a live issue whether the grounding relation is identical to the constitution relation. But even if they are distinct relations, this has no bearing on any of the claims I make in this chapter: At any point in my argument, we can replace mention of the grounding relation with mention of the constitution relation, whether in my discussion of Street’s own view or in my discussion of her third-factor opponents. Constitution, like grounding, is a non-causal form of dependence, so it can play the same role that grounding does in any of these accounts. (Similar comments apply to other non-causal forms of dependence, such as reduction, if that is something distinct from grounding, constitution, and identity [whether synthetic or analytic].)

234  Selim Berker There is, however, a major problem with the grounding account. Focus on our hypothetical character A again. Humean constructivism plus the grounding account provides a nice explanation of why A’s judgments about her own reasons mostly track the truth. However, A’s judgments about her own reasons are just a tiny subset of her judgments about reasons: Presumably she also makes a vast number of judgments about other people’s reasons. But the same story cannot be told about why these other judgments tend to track the truth, since the truth of A’s judgments of these other sorts depend, according to Street, on other people’s normative judgments. From A’s perspective, the truth about what other people have reason to do is a mind-independent matter—independent, that is, of A’s mind. However, the grounding account only applies to those of A’s judgments whose truth is dependent on A’s mind. So the grounding account, on its own, is unable to explain why A’s judgments about other people’s reasons tend to track the truth.46 Here is another way of putting the point. Humean constructivism posits a fundamental disunity among A’s normative judgments. When it comes to A’s normative judgments about her own reasons, thinking makes it so (cases of conflict aside). But when it comes to A’s judgments about other people’s reasons, thinking does not make it so (even when A is unconflicted)—judgments of that sort do not possess the magic property of ensuring their own truth. And it is precisely this magic property which plays a central role in the grounding account. Thus, from the perspective of the grounding account, the Humean constructivist is no better off than the normative realist when it comes to explaining why A’s judgments about other people’s reasons tend to track the truth.47 This problem was obscured by Street’s talk of “our” evaluative attitudes in her formulation of antirealism. Once we see that this pronoun must be used in a distributive (rather than a collective) sense for the grounding account to work in the case of one’s judgments about one’s own reasons, we can see that this problem is perfectly general, applying to all antirealist theories that attempt to make use of the grounding account in order to avoid the Darwinian dilemma.48 An obvious reply suggests itself. Perhaps the evolutionary factors in virtue of which I tend to make the judgments I do about your reasons also make it the case that you tend to make similar judgments about your own reasons. This causal hypothesis, together with (G′), could then be used to explain why my judgments about your reasons mostly track the truth. However, this reply is unconvincing. First, it requires the

46 In fact, the same point applies to A’s judgments about her own past and future reasons, since Humean constructivism ties the truth about one’s reasons at a given time to one’s judgments at that time about one’s reasons at that time. 47 After writing this chapter I discovered that Matthew Chrisman makes a somewhat similar point in his 2010, 338–9, although here Chrisman is discussing whether Street’s constructivism can account for ethical knowledge, not whether it can avoid the Darwinian dilemma via the grounding account. 48 The objection I have just pressed against the grounding account reveals another class of normative propositions, beyond the ones I mentioned in §4, that Street’s adaptive-link account is ill-equipped to handle on its own, namely propositions about other people’s reasons. My judging that you have conclusive reason to ϕ does not entail that I am at least somewhat motivated to ϕ. (It might, however, entail that I am at least somewhat motivated to help you ϕ, to praise you for ϕ-ing, and so on.)

NORMATIVITY MIND-DEPENDENT?  235 Humean constructivist to posit a level of convergence in our judgments about reasons that seems incompatible with the existence of persistent normative disagreements. Second, and more seriously, this response does not help explain why our judgments about the reasons of people who do not share our evolutionary history are generally reliable (including many of the judgments that Street herself makes about the reasons of various hypothetical figures during the course of her arguments both against normative realism and for Humean constructivism). In the end, I think there is no way around this problem for the grounding account.49 It is perhaps because she recognizes this problem for the grounding account that, in more recent work, Street sometimes provides a quite different account of how antirealists such as the Humean constructivist are able to avoid the Darwinian dilemma. On this account, rather than us, as theorists, appealing to the truth of (G′) in order to explain why a given agent’s judgments about reasons mostly track the truth, instead we imagine that our agent knows the truth of (G′) and uses it in her deliberations to reason her way to the correct theoretical verdict about what both she and other people have reason to do. Since we are assuming that facts about people’s normative judgments are purely non-normative facts, our agent’s judgment is not the sort of judgment that, according to Street, is susceptible to a Darwinian dilemma. So our agent can use her knowledge of this proposition and her knowledge of (G′) to reason her way to judgments about the normative proposition which track the truth, regardless of whether A is the deliberator herself or someone else. Thus the theoretical-reasoning account avoids the previous problem for the grounding account, and perhaps for this reason is to be preferred. However, the theoretical-reasoning account faces a serious problem of its own.50 This is not so much 49 There is an additional problem for the grounding account beyond the main one mentioned in the text. Presumably in order to address the epistemological threat that evolutionary considerations present, it is not enough to show that our normative judgments track the truth: We also want those judgments to constitute knowledge. However, the sort of tracking relation that we get from the grounding account seems irrelevant to knowledge. Consider the proposition . Whenever I believe this proposition, I make it true, and hence my belief in this proposition tracks the truth in the sense we have been considering here. However, it does not follow that whenever I believe that there are beliefs, I know that there are beliefs, for I might believe that there are beliefs for utterly idiotic reasons. State-given tracking of the sort featured both in this example and in the grounding account does not seem to be the right sort of tracking relation to underwrite knowledge. 50 It also faces less serious problems. In particular, its story about how someone apprised of the truth of Humean constructivism is to gain knowledge of her own reasons is rather awkward. In such a case, our deliberator is supposed to run through the following train of thought: “I judge that I have conclusive reason to ϕ, and my ϕ-ing does not conflict with anything else I more deeply judge myself to have reason to do. By (G′) it follows that I have conclusive reason to ϕ. So I have conclusive reason to ϕ.” Thus our deliberator comes to judge that she has conclusive reason to ϕ on the basis of her own judgment that she has conclusive reason to ϕ, and only the second time around does this judgment constitute knowledge. This is a truly bizarre picture of how knowledge of one’s own reasons is possible. (Thus perhaps the best option for Street is to combine a grounding account with respect to one’s judgments about one’s own reasons with a theoretical-reasoning account with respect to one’s judgments about the reasons of others.)

236  Selim Berker a problem for the account itself, as it is a problem for Street’s claim that only antirealists can make use of this account as a way of avoiding her Darwinian dilemma. What makes (G′) a distinctively antirealist claim is its grounding of a normative fact in the truth of some fact about our evaluative attitudes—in this case, the evaluative attitude of making a judgment about one’s own reasons. But the theoretical-reasoning account doesn’t in fact rely on this feature of (G′). All that matters for the theoretical-reasoning account are two things: that (G′) itself is knowable, and that (G′) grounds a normative fact in some fact which we can use normal empirical means to ascertain. As far as the theoretical-reasoning account goes, the crucial issue isn’t whether (G′) makes [A has conclusive reason to ϕ] mind-dependent, but rather whether it makes that fact knowably dependent on something empirically knowable. Whence, then, Street’s claim that antirealism is the only non-skeptical way of avoiding her dilemma? Here a comparison with third-factor accounts is particularly revealing. Recall that the most basic version of a third-factor account involves us, as theorists, appealing to the truth of (G) Non-normative fact F (at least partially) grounds normative fact N

in order to explain why a given agent’s judgment about N tends to track the truth. Thus the realist’s third-factor account is very close to the antirealist’s grounding account:  Third-factor accounts posit a slightly more complicated structure, but their basic idea is the same one found in Street’s grounding account, namely that we can bridge the gap between the normative and non-normative realms by appealing to the grounding relation.51 Moreover, once we see this parallel, it is easy enough to formulate a theoretical-reasoning version of a third-factor account in which we imagine a deliberator who knows (G) making use of that principle, together with her knowledge that F obtains, to reason her way to knowledge that N obtains. All of which suggests the following conclusion: If the antirealist can make use of the theoretical-reasoning account to explain how our normative judgments are able to track the normative truth, then the realist can make use of this account as well. The theoretical-reasoning account is not a distinctively antirealist way of avoiding the Darwinian dilemma.52

7.  Hoist by Her Own Petard? I have just considered the two different accounts that we find in Street’s work of how antirealists in general and Humean constructivists in particular can fend off

51 Or by appealing to some other non-causal form of dependence, such as constitution or reduction. (See n. 45.) 52 A referee offered the following objection to my claim of parity here: Maybe the antirealist can avail herself of the theoretical-reasoning strategy whereas the realist cannot because although the realist’s dependence claim, (G), is beyond our ken, the antirealist’s dependence claim, (G′), is not beyond our ken. However, this response assumes the very thing which, at this stage in the dialectic, we are aiming to show. The whole

NORMATIVITY MIND-DEPENDENT?  237 her evolutionary challenge. I have argued that the first of these—the grounding account—is at best a defense of the epistemic standing of a small subset of our normative judgments, whereas the second—the theoretical-reasoning account—is just as available to the realist as it is available to the antirealist. We have also seen how the grounding account is very similar in spirit to third-factor accounts, and how it is possible to construct a variant of a given third-factor account which makes it into an instance of a theoretical-reasoning account. Given these close parallels between Street’s antirealist strategies for defusing the Darwinian dilemma and third-factor ways of defusing that dilemma, it is only natural to revisit Street’s “one level up” and “trivially question-begging” objections to third-factor accounts, to see whether they apply to her antirealist strategies as well. And, in fact, I think it is quite clear that they do.53 Recall that according to Street’s first objection, third-factor accounts are not a response to the Darwinian dilemma because they rely on grounding claims such as (G), and we can rerun the Darwinian dilemma “one level up” with respect to the third-factor theorist’s judgment that (G) is true. A version of this same objection applies to both of Street’s antirealist strategies. Both strategies rely on a grounding claim such as (G′) being judged to be true, either by us as theorists (in the case of the grounding account) or by our imagined deliberator (in the case of the theoretical-reasoning account). Thus we can rerun the Darwinian dilemma “one level up” with respect to either our or the deliberator’s judgment that (G′) is true. According to Street’s second objection, third-factor accounts are “trivially question-begging” since they appeal to substantive normative truths such as (G) when trying to explain how our normative judgments track the truth. However, (G′) is just as much a substantive normative truth as (G) is, so Street’s two antirealist strategies also involve a question-begging appeal to normative truths, whether on the part of us as theorists (in the grounding account) or on the part of our imagined deliberator (in the theoretical-reasoning account). Conclusion: If third-factor accounts fall prey to Street’s “one level up” and “trivially question-begging” objections, then so too do her two antirealist strategies. Here is another way of reaching that same conclusion. Focus on the grounding account for the moment. (Parallel comments apply to the theoretical-reasoning account.) The grounding account is designed so that it can be appended to Street’s adaptive-link account of the evolutionary origins of our normative judgments, as in the left half of Figure 10.3. But third-factor accounts are also perfectly compatible with the adaptive-link account. The right half of Figure 10.3 shows one way of embedding the adaptive-link account within a third-factor account. Now stare hard at the dependency structures in the left and right halves of Figure 10.3. Both dependency structures posit essentially the same story about the causal facts, and both tether that causal story point of Street’s evolutionary argument is to give us a reason to think that realism faces epistemological difficulties which antirealism does not; we cannot simply assume that conclusion at the outset. 53 Russ Shafer-Landau makes a similar point in his 2012, 13–14.

Figure 10.3 

[φ -ing promoted reproductive success in my ancestors’ environment]

[Judging made my ancestors more likely to φ ]

(G)

[I have conclusive reason to φ ]

F = [φ -ing has property P]

[φ -ing promoted reproductive success in my ancestors’ environment]

[Judging promoted reproductive success in my ancestors’ environment]

[Judging promoted reproductive success in my ancestors’ environment]

[Judging made my ancestors more likely to φ ]

[I judge ]

Third-factor + adaptive-link account:

[I judge ]

(G’)

[I have conclusive reason to φ]

Grounding + adaptive-link account:

NORMATIVITY MIND-DEPENDENT?  239 to the normative realm via a grounding link, the only difference being where in the causal story this tether occurs. But why is this a distinction that makes a difference? If it is “trivially question-begging” for the third-factor theorist to tie her tether near the bottom of the causal structure common to the left and right halves of Figure 10.3, it should also be “trivially question-begging” for the antirealist to tie her tether near the top of that structure. And if we can rerun the Darwinian dilemma on the third-factor theorist’s lower-down tether, we should also be able be to rerun that dilemma on the antirealist’s higher-up tether. The two dependency structures in Figure 10.3 are equally question-begging, and they are equally susceptible to additional Darwinian dilemmas “one level up.” Street anticipates that her “one level up” and “trivially question-begging” objections might be thought to apply to the antirealist’s position as well as the realist’s, and in both cases she replies in the same way. She insists that there is no room to run the Darwinian dilemma “one level up” against the antirealist because “in arriving at his or her metaethical view, the antirealist does not need to rely on our substantive [normative] judgments.” She continues: This may be seen by imagining an alien investigator who (1) quite recognizably possesses [normative] concepts; (2) accepts [normative] judgments . . . with entirely different substantive content than our own; and who nevertheless (3) arrives at the same metaethical view as the human antirealist; and (4) does so based on the exact same considerations. Examples of such considerations might include the Darwinian dilemma itself . . . . (Street 2006, 163 n. 57)

Street’s thought here is that if no substantive normative judgments or intuitions are relied upon to reach the truth of antirealism, then we cannot rerun a version of the Darwinian dilemma on the judgments and intuitions which the antirealist uses to support her antirealism, and the antirealist is begging no questions by appealing to the truth of antirealism in the grounding or theoretical-reasoning accounts. Elsewhere, Street makes the same point for the specific case of constructivism (rather than for antirealism in general). She insists that all agents are committed to the truth of constructivism: “. . . no matter what one’s starting set of normative judgments [is], constructivism follows from within the standpoint constituted by those judgments” (MSb, 37).54 As no normative judgments are used to derive constructivist principles such as (G′), there are no materials on which to rerun the Darwinian dilemma one level up, and no questions are begged when the antirealist appeals to (G′) during her

54 Other places where Street asserts that the truth of constructivism follows from any arbitrary set of normative judgments together with the non-normative facts include Street 2009a, 216 n. 7; 2010, 378, 382 n. 16; and MSa, 16–17. Thus Street holds that if NJ is a fact summarizing a given person’s complete set of normative judgments at a given time, NJ grounds the truth of every constructivist claim about the grounds of normative truths. One such claim is (G′). But another is the claim that NJ grounds the truth of (G′). Therefore we have [(G′) ← NJ], [[(G′) ← NJ] ← NJ], and so on. This is how Street accounts for the infinite hierarchy of groundings “all the way out” to which she is committed by her definition of antirealism.

240  Selim Berker explanation of how our normative judgments are able to track the attitude-dependent normative truth. I am deeply suspicious of this reply on Street’s part. I mention here three objections. First: Her reply doesn’t actually address the “trivially question-begging” issue. Suppose that—given Street’s understanding of what these expressions come to—it does “follow” from “within” everyone’s “practical standpoint” that Humean constructivism is true. This fact is only relevant if we already assume the truth of Humean constructivism, and hence beg the very question at issue. Showing that a theory is by its own lights not question-begging is not the same as showing that a theory is not question-begging.55 Second objection: Despite her claims to the contrary, Street’s argument for antirealism in general does, in fact, rely on substantive normative judgments and intuitions. Her Darwinian dilemma against realism relies on substantive normative intuitions of the following sort, among others: intuitions about the epistemic relevance of tracking the truth, intuitions about the good- and bad-making features of explanations (scientific and otherwise), intuitions about the nature of the in-virtue-of relation when it takes normative relata, intuitions about when pain does and does not provide reasons (Street 2006, 149), and intuitions about whether belief attributions qualify as normative judgments (Street 2009a, 224–5). Moreover, Street acknowledges (in her MSb, 31–6) that the Darwinian dilemma on its own is not enough to conclusively support antirealism; one also has to make a case that the force of the Darwinian dilemma in favor of antirealism is not outweighed by the counterintuitiveness of antirealism’s apparent consequence that (for example) an ideally coherent Caligula has conclusive reason to torture others for fun. Street devotes an entire article to arguing that this consequence is not as counterintuitive as it might seem (Street 2009b). This, in effect, is to concede that an alien investigator who accepts more deeply than any other normative proposition is not committed to the truth of antirealism. Third objection: Street’s argument for Humean constructivism in particular also relies on substantive normative judgments and intuitions. An argument for antirealism is not yet an argument for Humean constructivism. And when we turn to Street’s arguments for Humean constructivism, we find a variety of implicit appeals to normative intuitions of a substantive sort. I focus here on two particularly striking examples. The first of these occurs during Street’s central defense of Humean constructivism

55 Suppose I hold a theory on which there is no such thing as the fallacy of begging the question. When you ask me why this theory is true, I say, “My theory is true because it’s true. Moreover, I’m not begging the question in asserting this, because there’s no fallacy of begging the question.” Within the confines of my theory, my reasoning in impeccable. But I nonetheless count as begging the question.

NORMATIVITY MIND-DEPENDENT?  241 in the case of instrumental normativity. She starts by insisting that the following two claims hold as a matter of conceptual necessity (Street 2008a, 227–9): (C1) If A judges , judges , and attends to the matter in full awareness, then A judges .56 (C2) If A judges and it is true that A’s ψ-ing is a necessary means to A’s ϕ-ing (whether or not A is aware of this), then by A’s own lights A has a reason to ψ.

In the next section (Street 2008a, §8) Street then slides, without argument, from this last claim to the following one: (C3) If A judges and it is true that A’s ψ-ing is a necessary means to A’s ϕ-ing (whether or not A is aware of this), then A has a reason to ψ.

However, this last move requires the backing of substantive intuitions, and it is doubtful that (C3) is true as a matter of conceptual necessity (if indeed it is even true). We can see this by considering analogues of these three claims in which A’s normative judgments are in the third-person:57 (C1*) If A judges , judges , and attends to the matter in full awareness, then A judges . (C2*) If A judges and it is true that B’s ψ-ing is a necessary means to B’s ϕ-ing (whether or not A is aware of this), then by A’s lights B has a reason to ψ. (C3*) If A judges and it is true that B’s ψ-ing is a necessary means to B’s ϕ-ing (whether or not A is aware of this), then B has a reason to ψ.

(C1*) and (C2*) may well be true as a matter of conceptual necessity, but (C3*) is definitely false. Thus there is no easy inference from to , and as a result bridging the gap between (C2) and (C3) requires substantive judgment. A second place where Street relies on substantive normative judgments while defending Humean constructivism occurs during her discussion of how Humean constructivists are to handle cases in which an agent’s normative judgments conflict with one another. (C3) only concerns whether A has a single pro tanto reason to ψ; what, though, might we say about the conditions under which instrumental considerations

56 Actually, Street wavers between two different versions of this principle. Sometimes (Street 2008a, 228; 2012, 43–4) she endorses (C1), and other times (Street 2008a, 227; 2010, 374; 2012, 43, 46) she endorses a weaker principle, namely: (C0) If A judges , judges , and attends to the matter in full awareness, then A doesn’t judge . (C0) is weaker than (C1) since does not entail , even under conditions of full awareness (suspension of judgment is also an option). 57 I learned this trick from Scanlon 1998, 28–9.

242  Selim Berker make it the case that A has conclusive overall reason to ψ? It is tempting to think that the Humean constructivist should hold (C4) If A judges and it is true that A’s ψ-ing is a necessary means to A’s ϕ-ing (whether or not A is aware of this), then A has conclusive reason to ψ.

However, (C4) quickly leads to paradox. Suppose—to use an example of Street’s (2008a, 235)—Beth judges that she has conclusive reason to eat a bowl of chili in front of her, and she also judges that she has conclusive reason to live a long, healthy life. Suppose, moreover, that Beth has a fatal58 allergy to peanuts, which unbeknownst to her the bowl of chili in front of her contains. Applying (C4) to Beth’s first judgment, we get the result that she has conclusive reason to eat the bowl of chili. (A’s ϕ-ing is trivially a necessary means to A’s ϕ-ing.) Applying (C4) to Beth’s second judgment, we get the result that she has conclusive reason to not eat the bowl of chili. But one can never, at the same time, both have conclusive reason to ϕ and have conclusive reason to not-ϕ. So something has gone wrong.59 Street’s way out of this problem is to insist that whichever of Beth’s two normative judgments is more deeply hers is the one that carries the day. In other words, Street proposes that the Humean constructivist endorse the following instead of (C4): (C5) If A judges , A’s ψ-ing is a necessary means to A’s ϕ-ing, and A’s ψ-ing does not conflict with anything else she more deeply judges that she has reason to do, then A has conclusive reason to ψ.

However, this move requires the backing of substantive normative judgment: (C5) does not follow from (C1), (C2), and (C3) by means of conceptual truths alone. A way of bringing this out is to consider Street’s proposal for how to determine which of A’s two competing normative judgments is more deeply hers: “this is a function,” Street writes, “of how strongly [A]‌holds the normative judgments in question and how close to the center of [A’s] total web of normative judgments they lie” (Street 2008a, 235). But exactly what function is this? Do we have some way of measuring the strength of a normative judgment, and another way of measuring a judgment’s nearness to the center of our agent’s total web of normative judgments, and finally a way of combining these two factors into an overall “depth score” that can be compared to the “depth score” for any other normative judgment made by that agent? This suggestion is, of course, absurd. But it brings out the degree to which figuring out the nature of Street’s alluded-to function is a highly non-trivial matter. Moreover, it is a matter over which substantive disagreement can occur: If I hold that comparative depth among normative judgments 58 Street says “life-threatening” allergy (2008a, 235), but this is not strong enough to ensure that Beth’s not eating the chili is a necessary means to living a long, healthy life. (Another complication: Maybe, even though Beth’s not eating the chili is necessary for her continuing to live, it is not a means to her continuing to live. But if so, then this is merely a problem for this particular example, and we can easily find another example with the desired structure.) 59 Note that a similar problem does not arise for (C3): There is nothing paradoxical about asserting that Beth both has a pro tanto reason to eat the bowl of chili and has a pro tanto reason to not eat the bowl of chili.

NORMATIVITY MIND-DEPENDENT?  243 is determined via function F1, and you hold that it is determined via slightly different function F2, it seems a stretch to think that one of us must be conceptually confused. We are in the realm of substantive judgments here, not conceptual truths.60 In fact, matters are even worse than this. The sorts of conflict cases that a Humean constructivist must contend with when filling in the details of her theory are not restricted to cases involving an instrumental conflict between a single judgment about what one has conclusive reason to do and another single judgment about what one has conclusive reason to do. The Humean constructivist also owes us a story about what happens in more complex instrumental conflict cases—for example, cases in which A’s judgment that she has conclusive reason to ϕ conflicts with several of her normative judgments about what she has conclusive reason to do. In such cases do we “add up” the depth score of each of these competing judgments to get a total depth score to compare to that of A’s judgment that she has conclusive reason to ϕ, or is some more complicated aggregation function at work? There are also higher order conflict cases, such as a case in which A judges that she has conclusive reason to ϕ, but also judges that she has conclusive reason not to judge that she has conclusive reason to ϕ.61 Does the higher order judgment always take precedence in such a case? Or does it matter how strongly held and near to the center of one’s web of belief the higher and lower order judgments are? Finally, the Humean constructivist must say something about what we might call cross-categorical conflict cases, in which an agent’s judgments using one normative category conflict with her judgments using another normative category (together, possibly, with her judgments about how those categories connect). What if A’s judgment about her reasons for and against ϕ-ing are in tension with her judgment that she ought not ϕ (together, possibly, with her judgments about the connection between individual reasons and overall oughts)?62 Does whichever of these normative

60 In a footnote, Street addresses the worry that she is appealing to substantive assumptions in moving from (C4) to (C5), but what she says is unconvincing. She writes: One might worry that in according priority to those normative judgments which are more strongly held and which lie closer to the core of a person’s interlocking web of normative judgments, the account smuggles in a substantive value. My reply is that the priority accorded these normative judgments doesn’t reflect a substantive value, but rather reflects the fact that we are asking about agent A’s reasons, not someone else’s reasons, and agent A is, in an important sense, to be identified with her most strongly and centrally held values. (Street 2008a, 235 n. 45) There are two problems here. First, the issue of whether (C5) relies on substantive normative assumptions is not the same as the issue of whether (C5) smuggles in substantive values: Not all normative claims are claims about whether something is of value; in particular, some concern how values play off against each other. Second, and more importantly, Street does not avoid appealing to substantive normative assumptions in her reply: The claim is a non-trivial claim that requires the backing of substantive normative judgment/intuition to defend. 61 If epistemic akrasia is not possible, then this specific example of a higher order conflict will never arise. But the non-existence of epistemic akrasia is compatible with other, less extreme varieties of higher order conflict. 62 Or maybe the mere truth of how individual reasons connect with overall oughts is enough to generate a tension between A’s reasons judgments and her ought judgment.

244  Selim Berker categories is in fact more basic determine which judgment gets priority? Or does it matter which normative category A takes to be more basic? What if the judgments employing the less-basic (or taken-to-be-less-basic) category are more deeply A’s than the judgments employing the other category? I bring up these issues to demonstrate just how complicated filling in the details of Humean constructivism becomes, once we realize the full range of conflicts that can arise amongst a person’s normative judgments. In one way, the lack of specifics here is not a problem: Sorting out those specifics is a research project waiting to happen, a dissertation begging to be written. However—and this is the central point—that research project does not seem to be one which can be conducted purely through an appeal to conceptual truths. Substantive disagreement over the best way to formulate Humean constructivism is possible, and resolving those disagreements requires making substantive normative judgments and appealing to substantive normative intuitions. Thus it is simply a mistake to say, as Street does, that arguing for Humean constructivism is merely a matter of “descriptive philosophical analysis” in which we figure out “what is constitutive of the . . . attitude of normative judgment” without relying on any substantive normative intuitions (Street 2010, 374). As a result, Street’s way of deflecting her own “one level up” and “trivially question-begging” objections fails.

8.  Our Epistemic Predicament I have just argued that Street’s own metaethical view is just as vulnerable to her evolutionary challenge as realist views are. Where does this leave us? Does this mean that Street’s Darwinian dilemma is in fact an argument for normative skepticism? Not necessarily. Three features of Street’s reply to her own “one level up” and “trivially question-begging” objections are particularly revealing. First, Street’s appeal to an alien investigator shows that evolutionary theory is really beside the point in generating her challenge. After all, our alien investigator might not be the outcome of evolutionary forces. Yet our alien investigator is supposed to be able to use a version of the Darwinian dilemma to realize the truth of antirealism. How could this be? Answer: A problem akin to the Darwinian dilemma arises even if the alien investigator was the outcome of non-evolutionary causal forces, or even if the alien investigator popped into existence five minutes ago.63 All we need to generate this problem is the fact that our normative judgments have causal origins, regardless of what those origins are. Evolutionary theory makes the causal origins of our normative judgments particularly vivid, but it is not essential for generating the puzzle.64

63 Or even, I would argue (though I shan’t argue for it here), if the alien investigator was created by God. 64 As Street herself admits in the final section of her 2006. Bedke (2009; forthcoming) also stresses that evolutionary influences are just one causal force among many when developing his version of a debunking argument.

NORMATIVITY MIND-DEPENDENT?  245 Second, Street’s way of trying to fend off her own “one level up” and “trivially question-begging” objections does not actually rely on the antirealist nature of Humean constructivism. The crucial move in Street’s reply to those objections is her insistence that the central tenets of Humean constructivism, such as grounding claim (G′), can be established without appeal to substantive normative judgments and intuitions. The fact that (G′) grounds a normative fact in facts about our evaluative attitudes plays no role in her response. Swap a different non-normative fact into (G′) and Street’s response would work just as well, provided that the new version of (G′) can be defended without reliance on substantive normative judgments and intuitions. But why does Street think that (G′) and the other grounding claims made by the Humean constructivist can be defended without recourse to substantive normative judgments and intuitions? It is because she—mistakenly, I have argued—thinks that (G′) and the like are conceptual truths.65 This leads to my third observation. Really what is doing all of the work in Street’s response to her own “one level up” and “trivially question-begging” objections is her contention that the Humean constructivist’s central claims about what grounds normative facts are conceptual truths. However, no claims about which non-normative facts ground normative facts are conceptual truths: Moore’s open-question argument blocks all roads here. So what Street is hankering after is impossible. Thus we should be suspicious of Street’s demand that we show, of all the conceptually possible normative truths, that evolution has allowed us to land on the correct one. If the only way of replying to Street’s “one level up” and “trivially question-begging” objections is to appeal to normative grounding claims which are conceptual truths, then there is no way to respond to those objections, whether they are leveled against third-factor accounts or against antirealist proposals. Just as Street can ask the third-factor theorist, “Of all the conceptually possible ways in which the normative facts could be partially grounded in the non-normative facts, what are the odds that you’ve hit on the right one when you judge (G) to be true?” so too we can ask Street, “Of all the conceptually possible ways in which the attitudinal facts might determine the normative facts, what are the odds that you’ve hit on the right one when you judge (G′) to be true?” Indeed, if conceptual connections are required between successive layers of facts in one’s explanation of why a given variety of judgment tracks the truth despite being influenced by evolutionary factors, then Street’s “one level up” and “trivially question-begging” objections also apply to explanations of our ability to make judgments about the presence of midsized objects in our immediate environment: “Of all the conceptually possible ways in which the causation relation could work, what are the odds that we’ve hit on the right one when, in offering Darwinian explanations, we make judgments about the causal properties of predators, progeny, and the like?”66 In 65 Street often says things like “It is constitutive of normative judgment that p,” where

is some claim about normative judgments, but for her this is just a fancy way of saying, “It is a conceptual truth that p.” For evidence that she uses these two expressions interchangeably, see Street 2008a, 228; Street 2009a, 226, 228–9, 231–2, 234, 236, 242; and Street 2012, 43, 46, 51, 55. 66 See Clarke-Doane 2012, 323, and 2014, 251–2, for a somewhat similar point.

246  Selim Berker short, Street’s conceptual demand threatens to make all synthetic knowledge impossible. But it’s one thing to voice the old empiricist worry over how synthetic a priori knowledge is possible, and quite another to call into question synthetic knowledge of any sort.67 As a result, I think we should simply ignore Street’s insistence that the sort of tracking at issue is tracking with regard to all of the conceptually possible normative truths. The only reasonable demand in the vicinity is one formulated with regard to metaphysically possible truths. This removes one important obstacle to be overcome when answering Street’s Darwinian challenge.68 What about the remaining obstacles? Let us distinguish between Street’s original Darwinian dilemma (as presented in Street 2006) and her amped-up Darwinian dilemma (once considerations about whether an explanation is question-begging are on the table). In response to the original Darwinian dilemma, I think that the following “divide and conquer” strategy is a perfectly adequate response on the part of the realist:69 • Argue that our tendency to make some normative judgments (such as, perhaps, our tendency to judge that incest is inherently wrong) is the outcome of evolutionary forces pushing us away from the truth. • Argue that our tendency to make other normative judgments (such as, perhaps, our tendency to judge that our own pain is, other things being equal, 67 For similar reasons we should be suspicious of Bedke’s appeal to conceptual possibility in his 2009 and forthcoming. Bedke only considers conceptually possible worlds in which all of the natural facts are the same as they are in the actual world but the normative facts are different. However, if we consider conceptually possible worlds in which some of the natural facts are fixed while others vary, Bedke’s debunking argument threatens to undermine any justification we might have on the basis for perception for believing synthetic non-normative truths. 68 I have just interpreted Street as holding that the central claims of Humean constructivism are conceptual truths. I do so because I think this is the best way to make sense of why she holds that Humean constructivism can be defended without recourse to substantive normative judgments and intuitions, and the best way to make sense of why she thinks Humean constructivism is better placed than realism to avoid her Darwinian dilemma when that dilemma is formulated in terms of conceptual possibilities. However, more recently (in her MSa, p. 9 of the draft of 20 October 2011), Street tells us that although “there once was a time—very long ago” when she thought that her constructivist account of the grounds of normative reasons was a conceptual truth, she now no longer believes this. (She does not tell us when this change in her thinking occurred.) In the end, this interpretive issue has no bearing on my argument from the past few paragraphs, since I can phrase that argument as a dilemma. Either Humean constructivism is being put forward as a conceptual truth, or it is not. If it is being put forward as a conceptual truth, then Street is right that the theory she puts forward would, if true, be especially well placed to meet a conceptual-possibility-based version of her evolutionary challenge, but this is no help, since that theory is demonstrably false. On the other hand, if Humean constructivism is not being put forward as a conceptual truth, then Street has just as much trouble with the conceptual-possibility-based version of her evolutionary challenge as realists do. Conclusion: it is unreasonable to formulate that challenge in terms of conceptual possibilities. 69 It is also open to the realist to insist that some of our normative judgments are not influenced by evolutionary factors. Derek Parfit (2011, chs. 32–3), Katarzyna de Lazari-Radek and Peter Singer (2012), and Russ Shafer-Landau (2012, 5–8) pursue this strategy. I have been setting that strategy aside in the current chapter, since the primary question I am asking is: If we concede that our normative judgments are heavily shaped by Darwinian forces, what follows?

NORMATIVITY MIND-DEPENDENT?  247 to-be-avoided) is the outcome of evolutionary forces pushing us toward the truth, by giving a third-factor account of how this is possible. • Argue that our tendency to make yet other normative judgments (such as, perhaps, our tendency to make grounding claims with normative relata) is a by-product of our having a selected-for general faculty that allows us to make judgments of a given sort, both normative and non-normative (such as a general faculty for reasoning our way to grounding claims).70 In essence, what I am urging is that the realist survey those of our normative judgments which evolutionary factors push us toward making, and then think hard about the truth of each of those judgments. If the judgment does not strike the realist as being plausible, all things considered, then she should take the first option. If the judgment does strike her as plausible, all things considered, and there is a direct evolutionary story for our tendency to make it, the second option is the way to go. If the judgment strikes the realist as all-things-considered plausible but the evolutionary story for our tendency to make it is an indirect one, then the third option is a possibility. There is no need to adopt a uniform account with regard to all normative judgments. If this response to the original Darwinian dilemma is successful, then it is also successful against any Darwinian dilemmas that might be offered “one level up.” In particular, there is nothing to stop a realist who invokes a third-factor explanation of our ability to reliably make some first order normative judgment from offering a third-factor account of our ability to make the grounding judgment (G) featured in the original first order explanation. After all, this in effect is what Street urges when the “one level up” objection is levied against her own explanation of how our first order normative judgments are able to track the truth. At the first order level she invokes a Humean constructivist grounding claim. When the Darwinian dilemma is pressed against our ability to judge this grounding claim to be truth, Street appeals to another Humean constructivist grounding claim, namely the claim that the truth of the original Humean constructivist grounding claim is grounded in our evaluative attitudes, since the truth of Humean constructivist follows from within everyone’s practical standpoint. When the Darwinian dilemma is pressed against this second grounding claim, Street responds in a similar way, and so on, ad infinitum. But if reiterating the 70 At one point (Street 2006, 142–4), Street tries to argue against by-product hypotheses by rerunning her Darwinian dilemma on the relation between the evolutionary forces that shaped the directly selected-for faculty and the attitude-independent normative truths posited by the realist. I lack the space to fully address Street’s argument here, but I will offer two comments. First, the new dilemma she presses against by-product hypotheses is actually quite different in both content and form from her original Darwinian dilemma. (In particular, the relation at issue is not the pushing-toward relation featured in the original Darwinian dilemma. That is why her argument against the “deny a relation” horn of the new dilemma does not appeal to the dire epistemic consequences of embracing that horn, as it did in the original dilemma, and why her argument against the “assert a relation” horn relies on extra steps not present in the original dilemma.) Second, Street’s new dilemma nonetheless overlooks many of the same possibilities she neglected when offering her original Darwinian dilemma. (In particular, she continues to assume that the only sort of tracking account available to the realist is a narrow tracking account.)

248  Selim Berker same strategy each level up works for the Humean constructivist, then reiterating the same strategy one level up should work for the realist as well. And, anyway, we should have been wary of Street’s claim that the Darwinian dilemma applies “one level up” with the same force that it applies at the original level. After all, we saw that the adaptive-link account is particularly shaky when it is applied to our tendency to make judgments about in-virtue-of-what a given normative claim is true. This problem only becomes worse as we keep going up successive levels: With each stage the force of the first horn of Street’s dilemma diminishes, since it becomes less and less clear that we know even the vague outlines of a version of the adaptive-link account that can then be compared in terms of empirical adequacy to a given tracking or third-factor account of the phenomena at that level. Moreover, the force of the second horn also diminishes as we go up levels: It’s one thing to be forced to say that we do not know any first order normative truths, but quite another to say that we lack knowledge of the increasingly complex normative grounding claims invoked in these successive explanations. Indeed, given the current state of play in the field of normative ethics, it seems quite plausible to hold that we lack knowledge of the precise grounding structure both “all the way down” and “all the way out” (if it keeps going out) with respect to even the most obvious of first order normative truths.71 Thus I  believe that my “divide and conquer” strategy allows the realist to fend off both Street’s original Darwinian dilemma and her “one level up” objection to third-factor accounts. However, that strategy, on its own, does not constitute a reply to Street’s “trivially question-begging” worry, and hence does not address her amped-up Darwinian dilemma. After all, a vital part of my “divide and conquer” strategy makes use of third-factor explanations, which invoke normative grounding claims, and we might worry that in making these claims, we are relying on normative judgments and intuitions that might themselves be “tainted” by evolutionary forces. But now the thing to do is to note that the skeptical worry here is perfectly general: It is just an instance of the general epistemological problem of how we can show that our most fundamental cognitive faculties (perception, introspection, induction, deduction, intuition—what have you) are reliable without relying on those very faculties when attempting to show this. The Cartesian circle, Chisholm’s problem of the criterion, Hume’s problem of induction, attempts to justify modus ponens by appealing to modus ponens, the recent literature on bootstrapping and easy knowledge—these are all, I believe, manifestations of the same fundamental epistemological unease. There seems to be something viciously circular about appealing to a given cognitive faculty when attempting to vindicate the epistemic standing of that very faculty. But, with our most basic cognitive faculties, what recourse do we have except to appeal to those faculties during their vindication? 71 Ignorance of what grounds a truth is compatible with knowledge of that truth. Aristotle knew many truths about water, even though he didn’t know that some of those truths hold in virtue of certain truths about hydrogen and oxygen atoms.

NORMATIVITY MIND-DEPENDENT?  249 This is a formidable problem—so formidable, in fact, that I think we should countenance the possibility that it has no satisfactory solution.72 However, I don’t believe that this is a special problem for our normative cognitive faculties. Moreover, it is not a problem that we need to appeal to evolutionary considerations to generate. If we discover tomorrow that all of the evidence for evolutionary theory is an elaborate hoax, we are left with a version of this problem. If we discover (somehow) that we were all created five minutes ago, the problem persists. To generate the problem, we need only start thinking about what makes it the case that our most basic normative cognitive faculties track the truth. Since the truths in question are normative truths, there is no way to approach this problem except through appeal, in part, to our normative cognitive faculties. To think otherwise is to think that it is possible to derive an ought from an is (or, at least, that it is possible to derive a substantive ought from an is plus a conceptual ought). But what about Street’s claim, via Accounts A and B, that the problem here is more pressing for our judgments about normative matters than it is for our judgments about the presence of midsized objects in our immediate environment? Recall the details of Accounts A and B: Account A: There are six chairs, a laptop, and a table in my immediate environment. But evolutionary forces gave rise to the capacity I used to make this very judgment. This gives me reason to think my capacity about midsized objects in my immediate environment is reliable. Account B: Midsized objects in our immediate environment are the kinds of things one can run into, be injured by, eat, and be eaten by. Other things being equal, then, creatures with an ability accurately to detect midsized objects in their immediate environment tended to survive and reproduce in greater numbers than creatures who lacked this ability. I am a product of this evolutionary process. This gives me reason to think my capacity to make judgments about midsized objects in my immediate environment is reliable. (Street 2008b, 216–17)

Street claims that Account A is trivially question-begging and gives us no reason to think we’re reliable on these matters, whereas Account B is ultimately question-begging but nonetheless gives us good reason to think we’re reliable. She also claims that third-factor accounts are of the same form as Account A. However, neither of these claims is convincing. First, third-factor accounts are actually more like Account B than they are like Account A.73 Account B features a variety of specific causal claims which are put together to give us a more complicated causal claim that underwrites the reliability of the very perceptual faculties used to support those original causal claims. Third-factor accounts often feature a variety of specific claims about individual reasons which are put together to give us a more complicated normative claim that underwrites the reliability of the very normative faculties used to support those original claims about reasons.



For my own attempt at a solution, see Berker MSa and MSb. Erik Wielenberg (2010, 459 n. 61) and Ronald Dworkin (2011, 447 n. 9) make similar points.

72

73

250  Selim Berker Second, Account B is, to my mind, just as epistemically problematic as Account A is. Suppose we discover a book of unknown origin that makes various claims about a hitherto undocumented era of the historical past. We begin to wonder whether this book’s claims track the truth. Then we find, halfway through the book, an elaborate story about how books of this sort were carefully screened for their accuracy, the unreliable ones being destroyed. (The book contains a story of, as it were, unnatural selection that applies to itself.) Does this story give us any reason to think that our book tracks the truth? I say: No, it does not. So we are left with a depressingly difficult epistemic problem, one that may well have no satisfying solution. But this is a problem that all metaethical theorists who are not normative skeptics face, regardless of their stance on the mind-dependence of normativity. And it is not a problem unique to the normative realm. The evolutionary origins of our normative faculties do not raise a special problem which only antirealists are in a position to solve. Rather, there is a problem here that afflicts all of our faculties, regardless of their origins.

Acknowledgments Versions of this chapter were presented at the 2012 Moral Psychology and Human Agency Workshop at the University of Michigan, as talks at the University of Toronto and the University of Montreal, and in my spring 2013 graduate seminar at Harvard University. I  thank the audiences on all of these occasions for their questions and feedback. In particular, conversations with James Bondarchuk, Sarah Buss, Tom Donaldson, Matt Evans, Jeremy David Fix, Thomas Hurka, Jim Joyce, Paul Julian, Guy Kahane, Leonard Katz, Douglas Kremm, Kate Manne, Paul Marcucilli, Jennifer Nagel, Mark T.  Nelson, Derek Parfit, Peter Railton, Said Saillant, Andrew Sepielli, Walter Sinnott-Armstrong, Zeynep Soysal, Sarah Stroud, Christine Tappolet, and Sergio Tenenbaum led to changes in the chapter. I also received immensely useful written comments from Fix, Katz, Kremm, and Parfit. Finally, I owe a special debt of gratitude to Justin D’Arms, Daniel Jacobson, and an external referee (since revealed to be Matthew Bedke) for their thoughtful and probing comments on an earlier draft.

References Bedke, M. 2009. Intuitive non-naturalism meets cosmic coincidence. Pacific Philosophical Quarterly, 90, 188–209. Bedke, M. Forthcoming. No coincidence? To appear in Oxford Studies in Metaethics. Berker, S. MSa. Graphic coherence. Unpublished manuscript. Berker, S. MSb. A graph-theoretic account of epistemic structure. Unpublished manuscript. Berker, S. MSc. The unity of grounding. Unpublished manuscript. Blackburn, S. MS. Sharon Street on the independent normative truth as such. Unpublished manuscript. .

NORMATIVITY MIND-DEPENDENT?  251 Brosnan, K. 2011. Do the evolutionary origins of our moral beliefs undermine moral knowledge? Biology and Philosophy, 26, 51–64. Chrisman, M. 2010. Constructivism, expressivism, and ethical knowledge. International Journal of Philosophical Studies, 18, 331–53. Clarke-Doane, J. 2012. Morality and mathematics:  The evolutionary challenge. Ethics, 122, 313–40. Clarke-Doane, J. 2014. Moral epistemology: The mathematics analogy. Noûs, 48, 238–55. Copp, D. 2008. Darwinian skepticism about moral realism. Philosophical Issues, 18, 186–206. Deem, M. J. 2012. Dehorning the Darwinian dilemma for realist theories of value. Paper presented at the 2012 Harvard–MIT Graduate Philosophy Conference. De Lazari-Radek, K., and Singer, P. 2012. The objectivity of ethics and the unity of practical reason. Ethics, 123, 9–31. Dreier, J. 2012. Quasi-realism and the problem of unexplained coincidence. Analytic Philosophy, 53, 269–87. Dworkin, R. 2011. Justice for Hedgehogs. Cambridge, MA: Harvard University Press. Enoch, D. 2010. The epistemological challenge to metanormative realism: How best to understand it, and how to cope with it. Philosophical Studies, 148, 413–38. Evans, M., and Shah, N. 2012. Mental agency and metaethics. Oxford Studies in Metaethics, 7, 80–109. Gibbard, A. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press. Joyce, R. 2001. The Myth of Morality. Cambridge: Cambridge University Press. Joyce, R. 2006. The Evolution of Morality. Cambridge, MA: MIT Press. Joyce, R. Forthcoming. Evolution, truth-tracking, and moral skepticism. To appear in B. Reichardt (ed.), Problems of Goodness: New Essays on Metaethics. Bonn: Bernstein Verlag. Kornblith, H. 1993. Epistemic normativity. Synthese, 94, 357–76. Manne, K. 2014. Internalism about reasons: Sad but true? Philosophical Studies, 167, 89–117. Nozick, R. 1981. Philosophical Explanations. Cambridge, MA: Harvard University Press. Parfit, D. 2011. On What Matters. 2 vols. Oxford: Oxford University Press. Petersen, S. 2013. Utilitarian epistemology. Synthese, 190, 1173–84. Pust, J. 2012. Intuition. In E. N. Zalta (ed.), Stanford Encyclopedia of Philosophy (Winter 2012 edn.). . Rosen, G. 2010. Metaphysical dependence:  Grounding and reduction. In B. Hale and A. Hoffmann (eds.), Modality: Metaphysics, Logic, and Epistemology, pp. 109–35. Oxford: Oxford University Press. Scanlon, T. M. 1998. What We Owe to Each Other. Cambridge, MA: Harvard University Press. Shafer-Landau, R. 2012. Evolutionary debunking, moral realism, and moral knowledge. Journal of Ethics and Social Philosophy, 7(1), 1–37. Skarsaune, K. O. 2011. Darwin and moral realism: Survival of the iffiest. Philosophical Studies, 152, 229–43. Street, S. 2006. A Darwinian dilemma for realist theories of value. Philosophical Studies, 127, 109–66. Street, S. 2008a. Constructivism about reasons. Oxford Studies in Metaethics, 3, 207–45. Street, S. 2008b. Reply to Copp: Naturalism, normativity, and the varieties of realism worth worrying about. Philosophical Issues, 18, 207–28.

252  Selim Berker Street, S. 2009a. Evolution and the normativity of epistemic reasons. Canadian Journal of Philosophy, Supplement 35, 213–48. Street, S. 2009b. In defense of future Tuesday indifference: Ideally coherent eccentrics and the contingency of what matters. Philosophical Issues, 19, 273–98. Street, S. 2010. What is constructivism in ethics and metaethics? Philosophy Compass, 5(5), 363–84. Street, S. 2011. Mind-independence without the mystery: Why quasi-realists can’t have it both ways. Oxford Studies in Metaethics, 6, 1–32. Street, S. 2012. Coming to terms with contingency:  Humean constructivism about practical reason. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, pp. 40– 59. Oxford: Oxford University Press. Street, S. MSa. How to be a relativist about normativity. Unpublished manuscript. . Street, S. MSb. Objectivity and truth: You’d better rethink it. Unpublished manuscript. . Wielenberg, E. 2010. On the evolutionary debunking of morality. Ethics, 120, 441–64.

11 Sentimentalism and Scientism Justin D’Arms and Daniel Jacobson

Perhaps the central thesis of the bourgeoning empirical ethics movement is what we will term pessimism about the role of reasons and reasoning in moral psychology. This is the claim that the reasons people offer in support of their evaluative judgments are merely post hoc rationalizations of conclusions driven by unthinking, “alarm-like” emotional responses.1 Some scientifically motivated pessimists conclude that most ordinary moral judgments must be discarded in favor of claims that are not tainted by the influence of emotion. Others embrace the vagaries of human sentiment by adopting forms of relativism that reduce moral judgments to empirical facts about the psychology of individuals or the sociology of groups. Despite this stark difference, however, the champions of empirical ethics are united in holding that the emotional basis of morality systematically undermines its pretentions to rational justification. We will argue that this pessimism is significantly overstated. There can be no doubt that people often adopt views for reasons other than the evidence, and hold them without ample justification. Yet any pessimism strong enough to call into question the possibility of moral justification, or to motivate the rejection of any contingently human influence on ethics, will inevitably neglect an important class of anthropocentric but nonetheless legitimate reasons. Moreover, no such strong pessimism is mandated by science. Although we grant summarily that any account of ethics that renders it incompatible with science must yield skepticism, we deny that empirical methods suffice to settle fundamental normative questions. Pessimists would either reduce moral claims to empirical facts or reject them as systematically unjustified, but this approach is not a deliverance of science but a dogma of scientism: the view that empirical methods are adequate to all areas of knowledge, including knowledge of values and practical reasons.

1 We will focus here on Greene (2008a, 2008b), Haidt (2001), Haidt and Bjorklund (2008), Prinz (2007), Rozin, Millman, and Nemeroff (1986), and Singer (2005); but we take the pessimist thesis, and scientism more generally, to be widely accepted in the increasingly influential empirical ethics movement.

254  Justin D’Arms and Daniel Jacobson We contend to the contrary that an overtly anthropocentric theory in the sentimentalist tradition can vindicate many ordinary moral claims, and reveal reasons that derive from deeply seated human concerns, in an intellectually respectable manner. The specific view we favor, rational sentimentalism, accepts most of the empirical claims motivating pessimism.2 As we understand sentimentalism in general, the theory gives the emotions a constitutive role in evaluative judgment, not just an important epistemological and motivational role in moral psychology.3 Unlike some other forms of sentimentalism, though, our theory holds that the emotions are amenable to correction and regulation by reason. Rational sentimentalism thus conflicts with the relativistic forms of sentimentalism that some champions of empirical ethics advocate, which uncritically accept emotional responses, and judgments made on their basis, without the possibility of rational correction. But it also opposes the other approach characteristic of empirical ethics, which would purify ethics of everything contingently human, especially the emotions. The scientism implicit in much empirical ethics is evident in the following dilemma, put forward by Peter Singer as a consequence of the pessimistic account of moral psychology, which sees moral reasoning as the rationalization of emotional responses. Singer (2005, 351) writes: In the light of the best scientific understanding of ethics, we face a choice. We can take the view that our moral intuitions and judgments are and always will be emotionally based intuitive responses, and reason can do no more than build the best possible case for a decision already made on nonrational grounds. That approach leads to a form of moral skepticism . . . Alternatively, we might attempt the ambitious task of separating those moral judgments that we owe to our evolutionary and cultural history, from those that have a rational basis.

Those who choose the skeptical horn of Singer’s dilemma by advocating relativism reject the possibility that emotions can be assessed for their rational justification but think that morality does not require it, whereas those who attempt to give ethics a purely rational basis deny that emotionally based judgments are ever rationally justified.4 Although the champions of empirical ethics differ as to which horn of the dilemma they find more congenial, they share this underlying pessimism about most ordinary moral intuitions. The advocates of empirical ethics thus agree that pessimism yields Singer’s dilemma but differ in their response to it. Most of the movement’s psychologists accept

2 But cf. Kahane (this volume) and Jacobson (2012) for some doubts about specific results. For the most part, though, we will be calling into question the normative implications of findings we accept or grant ex hypothesi. 3 This definition is somewhat stipulative, as a diverse range of views have been called sentimentalist. In our taxonomy, theories that merely give the emotions an important role in evaluative judgment are not yet sentimentalist. 4 There are other forms of justification possible, in terms of the advantageousness of emotions or their conformity with cultural standards, but these are fundamentally different notions than rational justification.

Sentimentalism and Scientism  255 emotionally driven intuitions as self-justifying, which leads them to some form of relativism. Jonathan Haidt and Fredrik Bjorklund (2008) put forward a view they call social intuitionism, according to which all ethical statements refer implicitly to the code embraced by the speaker’s culture.5 Although Jesse Prinz (2007) also adopts a form of relativism, most of the movement’s philosophers either overtly endorse skepticism or else follow Singer in aspiring to give a rational basis for morality that is expressly denatured of anything specifically human. Joshua Greene (2008a; 2008b) develops this hyper-rationalist approach most forcefully. According to Singer and the hyper-rationalists, any attempt to defend commonplace moral intuitions and judgments leads inevitably to skepticism. The trouble with ordinary intuitions is that they are products of contingent human evolutionary and social history, tainted by the influence of emotion. By contrast, Haidt and the social intuitionists embrace the influence of emotion by holding that moral intuitions can be given a non-rational justification sufficient to vindicate ethical knowledge. However, these social intuitionists thereby adopt forms of relativism that the hyper-rationalists consider tantamount to skepticism.6 Rational sentimentalism will serve as our example of the sort of theory that can navigate Singer’s dilemma while avoiding both its horns. The theory preserves a role for reasons and reasoning in moral judgment, vindicating ethics as essentially human but no less legitimate for this anthropocentrism. Although we cannot hope to mount an adequate argument for rational sentimentalism here, our more modest goals are significant enough. We will show that “the best scientific understanding of ethics” does not support any pessimism strong enough to support Singer’s dilemma. We then use rational sentimentalism to illustrate both the falsity of that dilemma and the extreme cost of its relativist and hyper-rationalist alternatives. The relativist theories foreclose the possibility of rational criticism of sufficiently entrenched sentiments and intuitions, just as Greene and Singer charge. But their hyper-rationalist alternative mistakenly denies that the emotions illuminate genuine reasons for human agents. It thus neglects the anthropocentric reasons arising from the symbolic and expressive aspects of action, which its advocates characterize as magical thinking and irrational taboo. This contention is not science but scientism—a theoretical prejudice that leads to significant moral errors.

1.  Singer’s Dilemma and the Case for Pessimism The strongest evidence for pessimism comes from the well-documented human tendency to confabulate: to invent and believe reasonable-sounding but false explanations 5 Although Haidt denies that his view is relativist, he does so only by adopting an idiosyncratic conception of relativism. Prinz grants that his view is a form of relativism. Both views make moral justification into an essentially psychological or sociological matter. 6 Nothing important hangs on how broadly we characterize skepticism, since the dilemma can be restated as between relativism and hyper-rationalism.

256  Justin D’Arms and Daniel Jacobson of one’s own behavior and choice.7 One commonly cited phenomenon arises from studies where incidental disgust, introduced in experimental subjects through various artificial means—such as foul odors and bitter tastes, filthy environments, and hypnosis—leads subjects to make harsh moral judgments about mostly innocuous, and sometimes wholly innocent, behavior they read about while disgusted.8 It should come as no surprise that people are led to exaggerate moral offenses when disgusted or angry; the more noteworthy empirical result is that subjects primed with incidental disgust sometimes produce rationales for their judgments based on features of the vignettes that neither cause nor support their judgment. Advocates of the empirical ethics movement have been impressed by these studies of incidental emotion and confabulation, and both Haidt and Greene cite them in support of pessimism. It is clear why a pessimist might take this phenomenon as corroborating evidence. In these cases, people are unaware of a factor that influences their moral judgments but has no justificatory force; moreover, the irrelevant feature distorting moral judgment is an emotional response. It is much less clear why the data on incidental emotions should be taken to show anything general about ordinary moral and evaluative judgments. In the first place, such incidental disgust must be contrasted with any integral disgust caused by the objects of the judgment: things that are disgusting, at least by the lights of the person affected. While incidental disgust taints the moral judgments it affects, integral disgust can be sensitive to features of the object evaluated: its disgustingness. Indeed, it is difficult to see how we can be sensitive to such sentimental values as the disgusting, the funny, and the shameful except through the emotions.9 Although incidental emotions do occur in the wild, as it were, not just in laboratory settings, the normal way in which disgust, anger, and the like influence judgment is as integral responses to the object judged: People are disgusted by something they deem disgusting, or angry over some putative outrage. These cases lack the problematic features associated with incidental emotions. The fact that one is disgusted by the prospect of cleaning a public toilet surely does not count against one’s judgment that the task is indeed disgusting. By contrast, incidental emotions can only taint judgment, whether they are primed in a psychological study or induced by some natural cause. If hypnotically induced disgust leads subjects gratuitously to suspect that a character in a scenario is “up to no good” (Wheatley and Haidt 2005), for instance, we can confidently expect that their judgments would change, were they given the same prompt while not under hypnotic suggestion, so as to correct this distorting influence. 7 In a classic paper, Nisbett and Wilson (1977) summarize an already lengthy literature on such confabulation. Haidt (2008, 189) expressly recalls their work on causal explanation in advancing his own theory of moral judgment. 8 See esp. Wheatley and Haidt (2005). 9 The contributions of integral guilt and anger to moral judgment involve more complex matters, some of which we will speak to shortly. In any case, advocates of empirical ethics typically want to indict the influence of emotion not just on judgments of right and wrong but evaluative judgment in general.

Sentimentalism and Scientism  257 What is more, some of the empirical work on incidental emotions is not merely compatible with rational sentimentalism but actually lends support to it by highlighting ways in which reasoning mediates emotional influence. Even in experimental circumstances, which reveal immediate responses rather than longer-term conclusions, people do not trust judgments made under the influence of incidental emotions. For instance, it turns out that when subjects know beforehand that they will have to justify their judgments about how much punishment is deserved, the effects of incidental anger are significantly reduced (Lerner, Goldberg, and Tetlock 1998).10 This suggests that people are capable of regulating the influence of emotional responses on their evaluative judgments. Although the confabulation data are striking and worrisome, the fact that people confabulate in circumstances where their judgments are being unwittingly driven by incidental emotions does not show that emotional influence is always suspect. Here and elsewhere, the evidence for pessimism seizes on something true and important but exaggerates its implications and ubiquity. Even if it is granted that incidental disgust obscures moral judgment, this is less significant than it seems, in light of the evidence about regulation; and it does nothing to show that disgust cannot reveal reasons to avoid disgusting things in particular. Much of the most frequently cited evidence for pessimism displays a similar tendency to overstatement. Haidt and his collaborators (Haidt, Bjorklund, and Murphy 2000) report finding that some experimental subjects stuck doggedly to their evaluative judgments about putatively “offensive yet harmless” actions despite being unable to articulate reasons for those judgments. Eventually they fall back on the claim that some type of action is just wrong, or wrong for inexplicable reasons. Haidt refers to such cases as moral dumbfounding, because people maintain condemnatory judgments—which he hypothesizes are driven by strong emotional responses, such as disgust at scenarios involving incest and cannibalism—without any supporting reasons. Dumbfounding thus seems like the quintessential case for pessimism. Haidt and others impressed with dumbfounding suggest that the phenomenon illustrates something about ordinary cases where people are not dumbfounded: that evaluative judgment simply recapitulates emotional response. Thus Haidt (2005) compares the role of reasoning in moral judgment to the spin of a presidential press secretary whose job is to tell credible lies that make the president look as good as possible. The difference between the ordinary case, where people support their moral judgments with reasons, and the dumbfounding cases amounts merely to the difference between a competent and an incompetent press secretary. Both lie by nature and job description—it’s just

10 Similarly, Schwarz and Clore (1983) found that mood effects due to fair and foul weather influenced people’s judgments of life satisfaction; but when respondents’ attention was drawn to the weather, no mood effects on judgment were observed. Schnall et al. (2008, 1106) take this finding to show that “rather than being obligatory, affective influences on judgment can often be eliminated by making salient an irrelevant but plausible cause for the feelings.” It seems that when people recognize the presence of a factor that obscures judgment, such as a mood effect caused by the weather, they correct for it.

258  Justin D’Arms and Daniel Jacobson that when you are dumbfounded, reason has failed at its task of constructing bogus rationalizations of your judgments. But the evidence of dumbfounding does not offer much support for this general pessimism. In the first place, the experimental evidence is quite weak. The paper giving the experimental findings has never been published; its results are barely statistically significant, and it uses a small and unrepresentative sample. The more interesting point concerns how dubious are the premises of the experiment. The experimenters assume a narrow conception of what can count as a good reason to condemn actions, limited almost entirely to harm. Even so, although Haidt (2000, 7) claims that the moral intuition scenarios were “carefully written to be harmless” and, hence, that it is “extremely difficult to find strong arguments” to justify censorious judgments, this is not the case. There are good and obvious harm-based reasons that support these judgments about which subjects were supposedly dumbfounded. Even so, one might think that the existence of good reasons does not change the fact that the subjects were unable to come up with them. But this claim is undermined by the experimental protocol, in which a “devil’s advocate” gave dubious counterarguments to those who offered reasons in support of their judgments. Moreover, since there were good reasons available in these scenarios, notwithstanding Haidt’s claims to the contrary, the experiment cannot discriminate between dumbfounding and inarticulateness. The subjects may have been responsive to reasons they could not articulate.11 None of this is to deny that people sometimes confabulate, especially in cases where they are forced to choose between identical items—where there are no grounds for choice—and in other unusual circumstances (such as hypnosis and split-brain patients). But although confabulation happens in various contexts, many of which have nothing to do with value judgment, in a host of mundane cases it seems we do act for reasons and know the reasons on which we act. Sometimes a person eats because he is hungry; that fact both explains why he ate and justifies it (by his own lights). Even when people confabulate a false causal story, sometimes they are nonetheless sensitive to evidence that they cannot articulate.12 In the evaluative domain, however, Haidt (2001) and other leading figures in empirical ethics go so far as to doubt whether we ever reason our way to moral judgment. Influential work by Haidt (2008, 187) and his collaborators advances the view that the “great majority” of our evaluative judgments are not sensitive to reasons but are mere rationalizations: bogus justifications for decisions already made on non-rational grounds, specifically on the basis of emotional responses. Greene (2008) argues that justifications offered for the moral judgments characteristic of commonsense morality 11 These criticisms are developed in more detail in Jacobson (2012). 12 For instance, in Maier’s (1931) famous hanging cord task experiment, subjects were unwittingly clued in to the solution to the task by the experimenter “accidentally” knocking into one of the cords, setting it in motion. Although they could not identify how they solved the problem but offered confabulated stories picking up on adventitious details, nevertheless they were demonstrably responsive to evidence—the hint—given to them by the experimenters.

Sentimentalism and Scientism  259 are similarly confabulated, though he thinks that certain other judgments are immune to this charge. And Singer too rejects the vast majority of moral judgments while holding out hope that ethics may yet be given an adequately rational basis. The strongest argument from pessimism to Singer’s dilemma comes from Greene. In his view, only direct appeals to the consequences of action, and more specifically to harm and benefit, offer legitimate grounds for ethical judgment. He calls such considerations “consequentialist” despite acknowledging that this is an idiosyncratic use of the term, since many of the theories philosophers deem consequentialist are not nearly so restrictive.13 We will follow his usage here by referring to this narrow range of considerations as consequentialist reasons, but nothing in our argument will tell against the best forms of consequentialism, which can acknowledge reasons that count as non-consequentialist in Greene’s sense. This exclusive focus is central to his view, as Greene (2008, 40) explicitly requires that the concepts appealed to by such reasons are “inherently neutral representations:” empirical concepts. While others are less explicit about this, we will see that (early) Haidt and Paul Rozin also tacitly presuppose that the only good reasons for moral judgments—and indeed for evaluative judgments and rational choice more generally—must be grounded in facts about the number of lives saved, or about harm and benefit construed narrowly so as to be empirical concepts. Greene’s argument for pessimism about moral reasoning that appeals to anything but these consequentialist reasons has two stages. First, he claims that all non-consequentialist reasons put forward in support of moral judgment are mere rationalizations. By this he means that the rationales offered for these verdicts, which appeal to considerations such as justice or cruelty, do not really explain the agent’s judgments. The true explanation is that they recapitulate the agent’s strong and decisive (“alarm-like”) emotional responses. Next, Greene (2008, 72) offers what we will call the argument from coincidence against all such judgments. In short: It would be an unlikely coincidence for those emotions, shaped as they are by evolutionary and cultural forces, to correspond to an “independent, rationally discoverable moral truth”— that is, a moral truth independent of anything contingently human. The rationalization charge seems, on its face, to be tantamount to the psychological claim that the rationale an agent offers in support of his judgment does not really explain why he so judges. In order for a consideration to be an agent’s reason for acting, it must seem to him to count in favor of the action, and thereby contribute to causing him to act. When people confabulate, their expressed rationale does not cause them to act and so cannot be their reason for acting. Hence confabulated rationales are merely rationalizations of whatever actually caused the judgment. But the charge

13 Greene adopts a narrow welfarist theory of value, which excludes the sort of considerations that moved philosophers away from utilitarian axiology. Moreover, his conception ignores indirect and otherwise sophisticated forms of the theory. It is widely recognized that consequentialist theories are not limited to adducing such “consequentialist” reasons, and they are well advised not to so limit themselves.

260  Justin D’Arms and Daniel Jacobson of rationalization, as it figures in Greene’s argument against non-consequentialist reasons, turns out to be more complex than it initially appears. Greene illustrates rationalization by imagining a character, Alice, who sincerely but misguidedly explains her preferences among the men she dates by citing their various good and bad features such as intelligence, humor, and self-absorption. In fact her preferences are simpler; she is attracted only to tall men. As Greene (2008, 67) tells the story: “Alice, of course, believes that her romantic judgments are based on a variety of complicated factors. But, if the numbers are to be believed, she basically has a height fetish, and all her talk about wit and charm and kindness is mere rationalization.” Alice’s putative reasons are confabulated, because they do not actually explain her date choices. Greene holds that people who make moral judgments based on non-consequentialist reasons similarly fail to grasp the true explanation of their judgments, namely their emotional responses. But there are two distinct features of Alice, which Greene does not differentiate. First, the rationales she offers for her choices— such as wit and charm and kindness—are not really her reasons. Second, it seems tacitly implied that were Alice made aware that height perfectly predicts her preferences among dates, she would disavow it as a reason. She would not endorse her actual criterion. Thus the Alice example involves both confabulation (of her putative reasons) and alienation (from her genuine reasons). Consider by contrast Greene’s most developed and realistic example of supposed rationalization in moral judgment: his argument against retributivism. According to Greene’s (2008, 50) official characterization of the view, to engage in retributive punishment is “to give wrongdoers what they deserve based on what they have done, regardless of whether such retribution will prevent future wrongdoing.” This fairly characterizes two core retributivist doctrines, both of which concern desert. The worse the transgression and the fewer mitigating circumstances, the more punishment is deserved; and the fact that someone deserves punishment counts as a reason— though not always as sufficient reason—to punish him. These two claims constitute what we will call core retributivism, which is the most defensible form of the view. Greene’s argument commits him to rejecting this modest version, because it too offers non-consequentialist reasons based in desert. Note that core retributivism does not entail the Kantian claim that it is intrinsically good for the wicked to suffer, much less that we should always give people what they deserve regardless of the cost—let the heavens fall, as it were. Surely even deserved punishment should be foregone when it would be catastrophically costly to punish. Desert is one consideration in favor of punishment but it can be overridden by other considerations, including sufficiently weighty consequentialist reasons. We will consider only core retributivism here, not either Kantian addition. This core of retributivism is especially important for our purposes, because it manifests the fundamental concern and action tendency of anger: It is sensitive to slights and wrongs, and it motivates retaliation or retribution. In defending core retributivism—which we will refer to simply as retributivism in what follows—we

Sentimentalism and Scientism  261 are in a sense (to be elaborated) defending the rational significance of the human propensity to anger. Greene’s discussion of retributivism purports to show that those who put forward abstract theories of desert or rights, in order to justify punishing wrongdoers, are guilty of rationalization. The retributivist supposedly offers a confabulated story about judgments that are really made because he is angry, and anger motivates him to retaliate. But this argument seems to misconstrue the role of anger in moral judgment, in two related respects. It posits too simple and direct a connection between sentiment and value, and it runs together two distinct points about confabulation: It is unconscious and unendorsed. First, the account is too simple. The relationship between anger and moral judgment is not as direct as Greene implies. Surely one need not have a bout of anger on every occasion when one judges that punishment is justified.14 Most obviously, one can believe that wrongdoing deserves punishment without focusing upon any particular case, much less getting angry about it. People also seem capable of making such judgments dispassionately even about particular cases, especially (but not exclusively) when the wrong does not affect their interests and lacks sympathetic victims. Thus individual retributivist judgments are not inevitably rationalizations of alarm-like emotional responses. It is more plausible that anger contributes indirectly to retributivist intuitions, in that people develop standards of wrongness and come to the conclusion that wrongdoing deserves punishment partly through their sentiments:  specifically the disposition to anger. While this suggestion does not support Greene’s charge of rationalization, it is worth elaborating briefly as an example of how rational sentimentalism offers an alternative conception of the contribution of emotion to evaluative judgment, which does not entail pessimism. It may well be that some of the plausibility of retributivism to humans arises from our predilection to embrace the internal logic of anger, which involves taking the transgressions that anger us to provide reasons to retaliate. But it is crucial to recognize how the critical assessment of our emotions, as well as consistency pressure on norms for when emotions are fitting, make a complementary contribution to a defensible retributivism. Because people are capable of such critical assessment, they do not simply conclude that whatever angers them deserves punishment. For instance, when you realize reflectively that the harm I caused you was unintentional and unforeseeable, you can recognize that it does not merit your anger. Then even if you have a lingering impulse to retaliate, given by your anger, you should conclude that it is unjustified and ought to be resisted. An equally important point is that while one’s bouts of anger contribute to the plausibility of the idea that transgressions deserve punishment, one’s ability to reason abstractly can lead to the realization that one’s similar transgressions against others are similarly blameworthy. This conclusion is not generated by 14 This is of course an empirical claim, but it seems to us obviously true, and Greene offers no evidence against it.

262  Justin D’Arms and Daniel Jacobson anger at oneself. It is rather an application of consistency pressure that helps refine anger and its characteristic motivation in ways that are more sensitive to reasons. This is not to say that our emotional responses are always so sensitive, or that a more modest form of pessimism focusing on specific human biases should be discounted. But it would be hasty to deny that transgressors deserve punishment simply because that intuition echoes the internal logic of anger. The second problem with Greene’s claim that retributivism is a rationalization of anger is even more significant for present purposes. It stems from its conflation of confabulation with alienation, both of which, one imagines, are present in the case of Alice. The role of anger in retributive thinking does not undermine retributivism unless one denies that anger—which is to say, integral rather than incidental anger—can be sensitive to good reasons to punish. (Recall again that they need not be sufficient reasons.) This is exactly what is at issue. When you are outraged at some heinous crime, in which the wrongdoer gratuitously harmed an innocent person, your anger purports to be responsive to considerations about what has been done that justify the claim that the criminal deserves punishment. Then although you recognize that your anger at wrongdoing motivates you to “give wrongdoers the punishment they deserve based on what they have done,” you will likely endorse this tendency. Compare this to Alice’s case, supposing that she were to come to realize that her overwhelming attraction to tall men fully determined her dating choices. But now imagine that she endorsed these choices on reflection rather than being alienated from them: She decides that attractiveness is all that matters to her, that her sole criterion is height, and that she is fine with it. Although science can be said to have undermined her previous, confabulated rationales about wit and kindness, it cannot show her preferences (which violate no canon of rationality) to be incorrect. Science has nothing to say on that score, although aesthetics allows us to criticize her taste as failing to be sensitive to all of the attractive and unattractive qualities in her dates, to say the least. Other normative disciplines provide grounds to criticize anyone who chooses dates exclusively on the basis of their attractiveness—even with impeccable taste. Greene aspires to undermine claims about desert, rights, and justice by making a psychological claim that retributivists engage in rationalization, which purports to obviate the need to give a substantive argument over the merits of the case for punishment. But this argument fails, because anger does not always cause judgments of wrongness and, even when it does, people often endorse their anger at heinous wrongdoing as being sensitive to genuine reasons to punish. That is, they think that what the culprit did both explains their anger at him and justifies it, as well as motivating and justifying his punishment. Although they may be mistaken, they are not like Alice and, hence, not susceptible to any purely psychological argument that they are guilty of rationalization. The Alice case gives illicit support to the rationalization argument by conflating confabulation and alienation, thereby making it seem like because Alice confabulates, she must be alienated from the true cause of her choices. In the case of anger at wrongdoing, however, there may be no confabulation even if anger is

Sentimentalism and Scientism  263 implicated in the causal story. Moreover, the judgment will likely be endorsed, notwithstanding the anger; if the agent is alienated from his anger, thinking it unjustified, then he will not judge on its basis. (This was the lesson of the first disanalogy between Alice and the retributivist.) Since Greene’s argument cannot make good on its grandiose ambition to demonstrate that retributivists are inevitably self-confuting, as Alice is tacitly assumed to be, his own theory of punishment must be compared with retributivism on the merits. The so-called utilitarian theory of punishment that Greene adopts offers an alternative to retributivism.15 It is not a conclusion of any empirical argument, however, and hence— in the absence of any reason to think that retributivists must be confused by their own lights—it is no more or less compatible with science. Instead the argument has to be made on overtly normative grounds. Since we reject pessimism and accept the possibility of rationally justifying normative claims, we think this possible. Indeed, we think normative arguments—or at any rate, compelling normative claims that few will want to deny—suffice to convict Alice of having bad taste in men, in addition to being highly imprudent in her criteria for choosing romantic partners. According to the theory of punishment Greene (2008, 70) endorses, considerations of desert do not provide any reason to punish: “consequences are ultimately the only things that should matter to decision makers,” he claims, when it comes to punishment or anything else. But this is a very difficult position to defend. Its worst implication is that we should punish the innocent whenever that would have the best consequences. In the philosophers’ standard toy case, lynching an innocent will save two lives that would otherwise be lost in a riot. In Greene’s view, the only legitimate consideration about this case is that the loss of two lives is worse than the loss of one. Greene must reject the claim that it would be unjust to violate the rights of the innocent person by lynching her—indeed, even that this injustice provides a reason not to do so—as these are just the sort of non-consequentialist considerations he considers mere rationalization. This is not just the familiar point that consequentialists cannot rule out any type of action, including lynching, a priori. Perhaps catastrophic cases suffice to justify that conclusion. The point is that even in ordinary and realistic cases, Greene’s theory commits him to holding that someone’s innocence does not provide any reason not to punish her.16 Retributivism avoids these consequences by recognizing desert as a source of practical reasons, which renders it considerably more plausible than the purely forward-looking alternative. We have already granted that some of this intuitive 15 “So-called” because the classical utilitarians did not, and modern consequentialists should not, adopt it. 16 The reader may be excused for suspecting that we must be reading Greene uncharitably here. In fact he embraces this result, albeit in the highly unrealistic (and hence less horrific) trolley cases, where the bullet he has to bite can seem more palatable. “As long as starving children get helped and people get shoved in front of speeding trolleys, that’s all I care about,” Greene (2008, 117) writes, his rhetoric leavening with jocularity a position he is committed to in all seriousness. Similarly, all his theory cares about is that one more life was saved by preventing the riot than was lost by lynching the innocent.

264  Justin D’Arms and Daniel Jacobson support comes from the human propensity to anger. While critics of retributivism are on their most favorable ground with anger, an emotion that many view with suspicion, it should be noted that when anger is directed against acknowledged wrongdoing it tends to go by such honorifics as righteous indignation. In any case, anger is hardly the only sentiment that is concerned with desert and thus conflicts with Greene’s insistence that only consequentialist reasons can justify action. More attractive attitudes such as sympathy and compassion also focus on the victims of unjust punishment and other undeserved suffering. Gratitude, which motivates us to respond positively towards those who have been kind to us, rather than bestowing our largesse wherever it would do a little more good, also manifests a concern for desert. Thus the sentiments are shot through with concerns that people be treated in ways that reflect their moral record. Furthermore, the very idea of merit, that achievement and effort deserve reward, makes essential appeal to desert. Thus the human concern for desert, which is central to anger’s internal logic, extends across a wide array of our attitudes and practices. These sentiments and other evaluative attitudes cohere with the idea that desert matters, as core retributivism claims and hyper-rationalism denies. Moreover, work in evolutionary psychology and behavioral economics also strongly suggests that concern for desert pervades human nature, for instance by making us keenly aware of cheating and committed to norms of fairness.17 Whereas sentimentalist theories embrace the idea that some such deeply seated psychological tendencies give rise to genuine reasons for humans, even if not for dispassionate aliens, Greene holds that emotions are “garbage” that should be ignored. This is the principal difference between an anthropocentric account of ethics and a scientistic view that is alienated from human nature and sentiment. To his credit, Greene confronts the fact that his view is unlivable for humans. He (2008, 76) writes: How far can the empirical debunking of human moral nature go? If science tells me that I love my children more than other children only because they share my genes, should I feel uneasy about that? . . . It seems that one who is unwilling to act on human tendencies that have amoral evolutionary causes is ultimately unwilling to be human.

We have no choice but to be human, of course, and hence for our deepest feelings to be in tension with full impartiality, sensitive to considerations of desert, and so forth. So this is not a genuine choice, and the hyper-rationalist does not offer a tenable alternative for how to live. Although Greene admits that his hyper-rationalism is inhuman, he thinks that the only other option is to embrace some form of relativism that cannot call our ethical intuitions into question. But that follows only if Singer’s dilemma exhausts the theoretical options, which is precisely what we deny.



17

See e.g. Cosmides (1989), Roth (1995).

Sentimentalism and Scientism  265 This debate over retributivism illustrates the general pattern of Greene’s argument from coincidence, which motivates pessimism about all moral judgments not based on narrowly consequentialist reasons—which is to say, any reason concerning rights, justice, desert, intention, and the distinction between action and omission (among much else). Greene argues that powerful emotional responses favor these conclusions, and he contends that it would be a remarkable coincidence if our intuitions, which correspond to these evolved emotional responses, just so happen to track moral reasons independent of human moral psychology. As he (2008, 72) puts it, “it is unlikely that inclinations that evolved as evolutionary byproducts correspond to some independent, rationally discoverable moral truth;” rather, it is much more plausible that when we feel the pull of non-consequentialist intuitions, “we are merely gravitating toward our evolved emotional inclinations.” In section 2 we will show that rational sentimentalism offers a theoretical option that answers Greene’s argument from coincidence. The route through the dilemma that Singer and Greene neglect is that our emotional responses can be sensitive to anthropocentric but nonetheless genuine reasons to act. There need be no cosmic coincidence when our emotions get it right because, although our evaluative judgments then track the moral truth, it is not an independent moral truth but one that is partly shaped by the contours of the sentiments, corrected and augmented by reasoning. Hence Singer’s dilemma, according to which science presents a forced choice between skepticism and hyper-rationalism, is a false dilemma—at least, if rational sentimentalism is coherent and defensible. Yet the idea that anthropocentrism leads to skepticism, which both Singer and Greene endorse, holds true of the forms of sentimentalism offered by their compatriots in the empirical ethics movement. As Greene notes, some of his fellow pessimists—most notably Haidt and Prinz—grant authority to emotions despite thinking that they cannot be given any rational justification. These views advocate an anthropocentric approach to morality, which, as Greene (2008, 74) puts it, settles for “a morality that is contingently human” rather than “deriving moral truths from first principles.” This description of anthropocentrism is accurate enough and describes our own view fairly. But Greene (2008, 74) also characterizes anthropocentrism in a more tendentious manner: “Rather than standing by our moral intuitions on the assumption that they can be justified by a rational theory,” he writes, “we might stand by them just because they are ours.” He then rejects anthropocentrism on the grounds that it cannot criticize entrenched intuitions but must accept them as justified. Although this characterization aptly describes the relativist positions of Haidt and Prinz, which disavow the possibility of rational criticism of accepted practices and entrenched feelings, it misconstrues those anthropocentric approaches that are not relativist or otherwise skeptical. Our contention that the hyper-rationalists set up a false dilemma thus rests on the claim that anthropocentrism—and, more specifically, sentimentalism—need not be relativist. In section 2 we illustrate and defend the (commonplace) practice of

266  Justin D’Arms and Daniel Jacobson appraising emotional responses for their rational justification by describing a sentimentalist theory on which evaluative judgments are not emotional reactions but assessments of the grounds for such reactions. Such grounds can be better and worse, and they are not simply about the instrumental advantages of responses or their congruence with societal tendencies. Hence commonplace evaluative judgments are not justified merely by consensus, or by coherence with one’s own patterns of emotional response, contrary to the claims of both forms of scientism championed by the empirical ethics movement. The strong pessimism adopted by both relativists and hyper-rationalists depends on their denial that ordinary evaluative judgments, grounded in the sentiments and other contingent features of human nature, are amenable to rational criticism and justification.

2.  Sentimentalism without Relativism The fundamental thesis of sentimentalism, as we understand it, is that evaluative concepts or properties depend essentially upon the emotions. This definition is expansive enough to include a variety of different theories, but not so broad as to include every ethical theory that gives a central role to the emotions either in moral motivation or the phenomenology of value. Sentimentalists claim that the emotions do not just detect values but partly serve to constitute them—as the funny is not just detected by our amusement but shaped by the human sense of humor.18 Consider by contrast a view that holds that values are primary qualities, independent of human nature, and then explains the correlation between emotional response and evaluative judgment by supposing that emotions track those independent values. Although this position flatters the emotions where hyper-rationalism disparages them, it does not count as sentimentalist. Such a view is particularly vulnerable to Greene’s argument from coincidence, because it offers no explanation of why our emotional responses “just so happen” to track human-independent moral facts. Sentimentalism can circumvent the argument from coincidence, however, since it does not claim that the values revealed in emotional responses are independent of human sentiments. This dependence is the essence of sentimentalism, though different versions of the theory differ crucially over how to understand it. Two of the simpler versions of sentimentalism, emotivism and social intuitionism, are pessimistic about the possibility of rational justification in ethics. Emotivism focuses on evaluative judgment, which it understands as the expression of emotional states (along with an imperative to feel similarly). In this view, to judge something good or right is to express one’s approval of it and to attempt to persuade others to take up a similar attitude. In 18 This means that the incongruity theory of humor, for instance, which identifies the funny with the incongruous—understood as an empirical concept rather than as response-dependent—is not sentimentalist, even if it regards amusement as the source of human concern for humor and our main epistemological route to the funny.

Sentimentalism and Scientism  267 effect it is to say: “I approve of this; do so as well” (Stevenson 1937). Social intuitionism focuses directly on values rather than value judgments, holding that the property of being valuable is relative to a culture, and to be valuable (in some culture) is just to be approved of by the relevant group. These simple forms of sentimentalism agree that approval and disapproval are not amenable to rational justification. We are not developing rational sentimentalism in any detail here, only broaching two of its central commitments that are relevant to rebutting pessimism. First and most important for present purposes, our view of the relation between sentiment and value is that value corresponds not with whatever people actually feel but with what feelings are fitting, and value judgments are not simply emotional responses but verdicts about what merits those responses. Thus value judgments are not justified by facts about what people are disposed to feel; rather, their defensibility hangs on the adequacy of the reasons supporting them. Precursors of this suggestion can be found in the early sentimentalists, such as Hume and Adam Smith, and it is now widely embraced by sentimentalists with very different views about the metaphysics of value (about which we remain neutral here). For instance, David Wiggins (1998, 187) claims that: “x is good/ right/beautiful if and only if x is such as to make a certain sentiment of approbation appropriate.” Contemporary expressivists, the intellectual heirs of emotivism, similarly suggest that evaluative judgment should be identified not with emotions themselves but with higher-order endorsements of such responses as appropriate (or fitting, merited, rational). As Allan Gibbard (1990, 51) puts it: “An action is morally admirable, we can say, if on the part both of the agent and of others it makes sense to feel moral approbation toward the agent for having done it.” Second, rational sentimentalism adopts a version of anthropocentrism according to which certain contingent human concerns provide reasons for action that we (humans) would not have in the absence of those concerns. But the concerns in question must be both deeply ingrained in human nature and widely supported by a network of motivational, affective, and cognitive sources in human psychology. We utilized this idea in the argument for the significance of desert above, and it will be further developed in section 3. Were pessimism correct, the distinction between fitting and unfitting sentiments could not be sustained, because emotions would only be amenable to empirical forms of assessment, for instance as in conformity with the norms of an agent’s culture or as coherent with his overall sensibility. Stevenson (1944, 138) anticipated this pessimistic thesis about the impossibility of rational justification of evaluative judgment: “if any ethical dispute is not rooted in disagreement in belief,” he claimed, “then no reasoned solution of any sort is possible.” Like emotivism, social intuitionism turns moral reasoning into nothing more than persuasion.19 As we shall see, Haidt embraces exactly this view of justification, and he does so in the context of a tellingly abhorrent argument that illustrates the impoverishment of this view.

This point was initially brought out by Brandt (1950).

19

268  Justin D’Arms and Daniel Jacobson For social intuitionists and other relativists, the only sensible notion of moral justification is that of conformity, whether to the culture or the individual’s sensibility. Haidt (Haidt and Bjorklund 2008, 216) endorses a cultural relativism according to which: “A well-formed moral system is one that is endorsed by the great majority of its members, even those who appear, from the outside, to be its victims.” Although Prinz’s view is more complex, he ultimately holds a version of individual relativism. As Prinz (2007, 177) describes it: “On the form of relativism that I have been endorsing, a speaker [who] says ‘you ought to Φ’ expresses the fact that the speaker endorses values that require Φ-ing,” where the values in question are fixed by the speaker’s sentiments.20 These relativist versions of sentimentalism coopt such normative notions as ought by replacing them with something empirically tractable: conformity to one’s culture or one’s own actual sentiments. But this undermines the normative force of these notions, since (for instance) the fact that one’s culture endorses some moral norm does not justify accepting it. The pessimism embraced by relativism and emotivism concedes the resources needed to defend classically liberal ideals of rights and justice. And relativism in particular lends itself to the defense of oppressive social practices that run counter to the progress of civilization towards institutions that better promote human flourishing. This is most evident of social intuitionism, which implies a moral equivalence between (viable) cultures. That tendency reflects a loss of confidence in what we hold to be genuinely worthy and rationally defensible ideals of Western culture: respect for individual rights and responsibilities, the goal of progress in human well-being, and the struggle against unjust and oppressive social practices. A similar point applies to the hyper-rationalism of Singer and Greene. Though that view is concerned with welfare, or at least with pain and pleasure, it rejects considerations of justice and rights as mere rationalization. It thus neglects the historical lesson that progress in human well-being is fostered by representative government, rule of law, and the other institutions of a free society—all of which demand more sophisticated forms of decision making than crude consequentialism. The rest of this section illustrates the theoretical and moral deficiencies of relativism, while the final section displays a wide range of genuine reasons that elude the scientistic approach characteristic of hyper-rationalism. The following example is intended both as a theoretical argument for the possibility of moral reasoning and justification, against pessimism, and as a normative argument in defense of the liberal ideals disparaged by social intuitionism, which if successful justifies certain patterns of approval and disapproval against others. Both points can be illustrated with a case both Haidt and Prinz discuss at some length, namely clitoridectomy—which gets called female circumcision by its apologists and female genital mutilation by its critics. Although as critics we consider the censorious term more apt, 20 Although this sounds like expressivism, it is not. Prinz (2007, 177) explains that this claim is “true if and only if the speaker has moral values that prescribe Φ-ing.” There are some adventitious complications to the view, but for a similar statement put in terms of truth conditions, cf. Prinz (2007, 179).

Sentimentalism and Scientism  269 we will avoid prejudicing our case by choosing the most neutral term available to prosecute this argument. Were Haidt’s cultural relativism correct, then the fact that a great majority of the members of a culture (whatever that amounts to) support the practice, including women, would suffice to make it part of a well-formed moral system.21 Indeed, since Haidt (2008, 216) identifies being virtuous with being “fully enculturated,” it is not merely that participation in this practice is virtuous; worse yet, opposition to it must count as vicious insofar as it precludes enculturation. The paradigm of moral argument offered by Haidt (2008, 191) comes from a debate over clitoridectomy, in which he follows the misguided tendency of many social scientists to conflate clitoridectomy with circumcision, which allows him to say, misleadingly, that the practice is “common in many cultures.” He then quotes a passionate condemnation of ritualistic female genital mutilation that he calls an argument—or rather seven arguments, one for each emotionally loaded term—so as to treat the passage as paradigmatic not just of the case against clitoridectomy but of moral argument in general. Haidt (2008, 192) asks the reader to “note that each argument is really an attempt to frame the issue so as to push an emotional button, triggering seven different flashes of intuition in the listener.” But one cherry-picked example makes no general case for pessimism. Haidt’s description accurately describes his own attempt to frame the practice as just another variety of genital alteration, however, in order to suggest that any distress you may have over it can only be prejudice. Despite the fact that there are compelling (and obvious) reasons to draw a moral distinction between circumcision and clitoridectomy, Haidt cheats by choosing a polemical statement as his paradigm of moral argumentation. In order to illustrate the implications of these points, we first need to make clear the terms of the disagreement and the burden of his argument. If pessimism about moral reasoning were correct, our disagreement over the permissibility of clitoridectomy could only be an effort to exert non-rational influence or an anthropological dispute about how much agreement over the practice exists. But we can grant Haidt the anthropology, for the sake of argument, without withdrawing our moral judgment. Let us suppose then that Prinz (2007, 209) is correct when he claims, in his own apologia for clitoridectomy, that in “most cultures where female circumcision is performed, women evidently support and promote the practice.” We will stipulate that this is true about the case at hand, so as to make the argument as difficult as possible for us critics and anti-relativists. Since we dispute that widespread agreement suffices to make a practice permissible, we posit such agreement. But we demand the same consideration in return. Since in their view it makes no difference how harmful is the practice so long as it has the requisite support, they cannot shrink from the most heinous cases, where—among many other oppressive features of the practice as it actually exists—it grievously harms young girls without their consent (notwithstanding 21 It is unclear why Haidt gets to make even this caveat (about victims), except for the contingency that women are not a minority group. When a minority is small enough that a “great majority” of the culture need not include it, then their lack of consent must in Haidt’s view be considered morally irrelevant.

270  Justin D’Arms and Daniel Jacobson the views of other women). Hence we will focus on these cases, where the practice is most harmful and yet sufficiently accepted to count as right according to relativism. Our dispute with Haidt over this case is not a matter of anthropological fact, and our dispute with Prinz is not a matter of psychological fact. We are not denying that clitoridectomy has widespread support in this culture, nor are we merely expressing our own disapproval of it. We are rather claiming that there are conclusive reasons to disapprove of the practice, which exist even if it is approved of by the majority in that culture. But our disapproval does not make it wrong, even for us; it would be wrong even if we did not disapprove, in virtue of the good reasons to disapprove of it. The clearest reasons are that the practice harms young women, without their consent, in a manner that inhibits their flourishing. This has to be granted to be a powerful prima facie argument against the practice, which could only be overcome by a very compelling counterargument. Yet the counterarguments proffered by Haidt and Prinz—and the apologists they cite approvingly—are lame, because they offer bad reasons in support of the permissibility of the practice. Their analogies to male circumcision and body piercing are very weak, since those practices are not seriously harmful (and the latter case is consensual). Whatever one thinks of those practices, the strongest reasons to disapprove of clitoridectomy simply do not apply to them. Both Prinz and Haidt allow themselves to argue against the most favorable cases: not only where the practice is supported by most women but also where it does not seriously harm women by (for one thing) permanently precluding their ability to enjoy sex—even though this is the overt goal of most extant forms of the practice. Moreover, the good reasons to disapprove of clitoridectomy are deliberately obscured by its apologists. Consider their assimilation of the practice with less objectionable practices that can also be framed as the alteration of an infant’s genitals. Castration too counts as genital alteration, but that doesn’t place it on a par with circumcision; we are confident that few men are indifferent between them. Hence there are strong reasons to condemn clitoridectomy and only poor reasons to tolerate it under circumstances where relativists must not merely tolerate but endorse the practice. While this argument adduces reasons, it also vindicates moral reasoning as being more reputable than post hoc rationalization. Anyone convinced by our argument against assimilating clitoridectomy with circumcision as simply alternative forms of genital alteration, for instance, has come to grasp better reasons through reasoning about the crucial disanalogies between these cases. This argument has several implications for our discussion. First, the most fine-grained conclusion is that clitoridectomy is wrong even where it is socially approved. That point suffices to belie relativism. Second, there are good reasons to disapprove of the practice, and the countermanding reasons offered in its favor are weak—they depend on faulty analogies and rhetorical sleight of hand. This point shows that pessimism, which denies or minimizes the role of reasons and reasoning in moral judgment, is vastly overstated. Both these arguments are compatible with various ethical theories, since they invoke harm and flourishing; they are not distinctively

Sentimentalism and Scientism  271 sentimentalist. The facts about human nature to which they appeal might, for all we’ve said, be reducible to empirical concepts compatible with hyper-rationalism. The point at hand is that because rational sentimentalism focuses on reasons to approve or disapprove of actions and practices—not on empirical facts about approval—it too can justify moral condemnation based on the good reasons to disapprove. Sentimentalism is thus compatible with rational justification, even though the simplest forms of the theory eschew it in favor of pessimism. This discussion of relativism reveals that Greene was correct to this extent: The form of sentimentalism that his compatriots in the empirical ethics movement embrace is tantamount to skepticism. But the problem lies with their relativism rather than with anthropocentrism. We do not face a forced choice between an inhuman morality and an uncritical acceptance of moral feelings and intuitions. Instead, one can recognize reasons that are neither revealed by science nor in conflict with it. Since the reasons adduced by the previous example concern harm and flourishing, however, a hyper-rationalist theory can claim to capture them as well. In section 3, we argue that the scientism manifested in the aspiration of purifying ethics of everything human misses a whole range of important reasons revealed by the sentiments.

3.  The Reasons that Elude Scientism We have thus far argued that the scientistic approaches characteristic of the empirical ethics movement are inadequate because they embrace too strong versions of pessimism and, as a result, are left with untenable normative commitments. Since rational sentimentalism can question entrenched moral intuitions, unlike relativist forms of sentimentalism, it need not endorse oppressive cultural practices that enjoy widespread acceptance. And since it is not committed to the narrow conception of reasons that hyper-rationalism can accommodate, rational sentimentalism avoids having to accept whatever turns out to be optimal according to the sole criterion of crude consequentialism. We now turn to illustrating some scientistic errors that mar the work of eminent moral psychologists who disparage a whole class of good but anthropocentric reasons, which they claim to be tainted by magical thinking and irrational taboo. These reasons arise primarily from the symbolic and expressive aspects of action; they are a particularly significant source of reasons to which the sentiments respond. Our aim is less to vindicate any particular instance of these reasons than to show that, as a class, they are not incompatible but discontinuous with science. The psychologist Paul Rozin and his collaborators (1986) claim that magical thinking pervades modern Western culture, and Haidt (2001) argues similarly that cosmopolitans too labor pervasively under irrational taboos. We do not suppose that Westerners are devoid of superstition and belief in magic—that would be absurd. But the examples that Rozin and Haidt use to demonstrate the ubiquity of magical thinking and taboo, and most importantly its influence on ordinary moral judgment, show nothing

272  Justin D’Arms and Daniel Jacobson of the kind. The actions and beliefs that they ignore or disparage as irrational, and gratuitously attribute to supernatural beliefs, can in fact be supported by good reasons. Although Rozin grants that there are other possible explanations of the data, he rejects them. Since belief in the efficacy of magic “clearly occurs in traditional societies” where those alternative explanations fail, he writes, “it seems uneconomical to invoke different accounts for the subset of these behaviors that we have documented in American culture” (1986, 711). We suggest to the contrary that interpretive charity is a bargain as compared to the cost of equating the metaphysical assumptions of traditional (i.e. pre-scientific) societies with those of modern, cosmopolitan ones. It is ironic that scientism would draw this equivalence, but this reflects an ideological presupposition rather than following from the data. What leads Rozin and Haidt to conclude that Western cultures remain in the grip of irrational taboo? Haidt (2001, 817) offers a catalogue of what he deems harmless taboo-violation tasks and scenarios, where he thinks people make condemnatory moral judgments and refuse to perform supposedly innocuous actions without reason. These actions include, among other things, eating one’s dead pet dog and cleaning the toilet with the national flag. In one of his taboo-violation tasks, called Soul, an experimenter offers the subject two dollars for signing an explicitly non-binding “contract” that grants possession of the subject’s soul after death to the experimenter. It should come as no surprise that some people turn down the money. In another experiment, Roach, many people decline to drink a cup of juice into which a sterilized roach has been dipped, despite having witnessed its sterilization. Since the “roached” juice—as Rozin charmingly calls it—has no more germs than the regular variety, Rozin and Haidt conclude that there is no reason not to drink it, merely a widespread but irrational aversion. Rozin thinks that aversions like these manifest magical thinking. What he calls the first law of sympathetic magic is the law of contagion: “things that once have been in contact with each other may influence each other through transfer of some of their properties via an ‘essence’ ” (Rozin, Millman, and Nemeroff 1986, 703). Since this essence remains “in some form of nonphysical contact with its source,” this allows for the possibility that action taken on a clipping of someone’s hair, for instance, can affect that person—which, as Rozin (1986, 703) notes, “is the basis for a major form of sorcery.” The second magical law is the law of similarity, according to which “things that resemble one another share fundamental properties” that allow actions taken on the simulacrum to affect the object it resembles (1986, 703). Thus a voodoo doll constructed to look roughly like its intended victim, and which incorporates some sort of physical residue of him, utilizes both these magical laws. People who perform voodoo rituals can reasonably be ascribed belief in these principles of magic—beliefs they would avow in uninhibited discourse. But the Americans surveyed and tested do not practice voodoo and, presumably, disavow any such belief. The magic principles are ascribed to them nonetheless, because they behave in ways that Rozin and Haidt find so irrational that their best explanation adverts to such magical thinking.

Sentimentalism and Scientism  273 Consider some of the cases that drive them to this view. Rozin contends that it is irrational to prefer putting a (new) rubber sink stopper in one’s mouth to doing the same with a rubber mold fabricated to look like vomit. Similarly, he thinks there is no reason to be averse to eating fudge shaped to resemble feces, or to be repelled by the prospect of wearing a lab coat used by Joseph Mengele, the infamous Nazi doctor.22 As long as the coat has been thoroughly cleaned, what reason could you have to resist—unless you secretly think that Mengele’s essence has somehow been non-physically imparted to the garment? That really would be magical thinking. But although there is a science of disgust, and Rozin is perhaps its most eminent figure, there is no science of the disgusting—the evaluative property of meriting disgust—and Rozin’s views about that are not authoritative or even plausible. We can agree on all the (empirical) facts about these artefacts and yet disagree about whether there is reason to be disgusted by them. Note that good reasons to be disgusted are good reasons to be in a state that, by its nature, inclines one to avoid intercourse with the objects of one’s disgust. While the weight of those reasons is open to debate, even slightly compelling reasons are strong enough to justify refusing to perform acts that you have no reason to do other than the request of some experimenter; we will take up this point in more detail presently. But first we need to support the claim that there are reasons to be disgusted by such things. We have argued elsewhere (D’Arms and Jacobson 2005)  that sentimental values such as the disgusting (and the funny, the shameful, etc.) provide reasons to normal humans in part because they embody psychologically deep and wide human concerns. Deep concerns are those that are so firmly entrenched in human psychology that they would be either impossible or very costly to extirpate; wide concerns are those that play various roles in the moral psychology of their possessor. When the object of a concern resonates throughout a person’s evaluative responses—including emotions and other feelings, desires, intentions, plans, and evaluative judgments—this is indicative of its width. Greene’s example of his bias in favor of his own children is one case of a deep and wide human concern; so too are the conviction that desert matters and the aversion to disgusting things. According to rational sentimentalism, concerns that are both deep and wide in human psychology ground anthropocentric reasons that people would not have were they differently embodied. Our discussion here begins with reasons to be disgusted, and then turns to some other deep and wide human concerns that manifest themselves in various feelings and motives common to humanity. In our view and according to common sense, the disgusting is partly a perceptual property—which is to say that things can merit disgust in virtue of the way they look, taste, or smell. Thus what Rozin explains by appeal to a magic law of similarity, according to which things that resemble one another share fundamental properties with causal powers, can instead be explained by the truism that things that look like excrement or vomit look disgusting. In these simple perceptual cases, the genuine visual similarity to something disgusting suffices to render things visually disgusting, which

22

Rozin uses Hitler’s sweater as his example, but this emendation makes the same point more vividly.

274  Justin D’Arms and Daniel Jacobson thereby justifies our aversion to the fudge and the rubber mold. One need not believe in essences or magical causal powers to hold that things that look like that are disgusting: They do not just cause but merit disgust. This explanation is simple but sufficient for these perceptual cases, and it gets the right result on properly mundane grounds. We humans care not only about what things are made of but about what they look (taste and smell) like—which is the source of a wide range of aesthetic concerns. Rational agents need not care about aesthetics, but that hardly undermines the concern for beauty and other aesthetic qualities, let alone shows them to be magical or otherwise irrational. The Roach case is somewhat more complicated, because it is not perceptual: Roached juice does not look or taste any different than ordinary juice. Rozin posits that his subjects are in the grip of a magical law of contagion, according to which the roach passes on its essence to the juice. We infer that he would not ascribe magical thinking to someone reluctant to drink if the roach had not been sterilized; in that case, Rozin and Haidt imply that disgust would be justified. But why should sterilization remove the disgustingness of having a roach dipped in your beverage? The obvious answer is that disgust is about contamination. Its function is to serve as a fast and frugal germ detector, which gets a false positive in this case. Rozin and Haidt implicitly assume what might be called the germ theory of contamination, which holds that if something is not germy (or otherwise toxic) then it is not disgusting. But it is extremely hard to accept the implications of this theory, on which nothing hygienic is disgusting to look at, eat, “kiss,” and so forth. A surfeit of obvious cases—including such things as vomit-shaped rubber molds, sterilized excretions, and well-embalmed corpses—stand as a reductio to this theory. Rational sentimentalism allows for the criticism of actual responses as unfitting, as we’ve insisted, so it is open to these psychologists to argue that the predictable disgust response to vermin and other common elicitors is mistaken whenever they are not germ-ridden. Yet such widespread patterns of emotional response to objects like these can only be overcome by countermanding reasons. And the reason on offer—the fact that the objects are not germy—is true but unconvincing in light of the aesthetic points raised previously. Instead one should reject the scientistic assumption that the evolutionary function of a mechanism like disgust determines when we have reason to be disgusted. The germ theory of contamination cannot handle perceptual cases, and it entails absurd conclusions about hygienic corpses and sterilized excrement. Moreover, it cannot make sense of another broad class of concerns, of which Mengele’s lab coat is an instance. While we suppose it is possible that someone might not want to wear the coat because she was worried about Mengele’s essence being transferred to her through some sort of non-physical contact—as Rozin would have it—surely it is more likely that her aversion stems from a broader human concern for the histories of people and things, which is reflected in a plethora of ordinary phenomena. People collect autographs. The pen Lincoln used to sign the Emancipation Proclamation has more value than

Sentimentalism and Scientism  275 an authentic pen of that era with no historical import. It is awesome to stand in the Roman Forum or the Old City of Jerusalem and reflect on who must have walked on the same flagstones. One of the crucial differences between an Old Master painting and a forgery, however well executed, is that only the former was actually painted by Vermeer. For someone with this very normal pattern of attitudes, who is appalled by Mengele’s atrocities, the fact that this garment was his lab coat will seem a very good reason not to wear it. Nothing about this pattern of concern requires the attribution of magical thinking to those who care about such things. Again, any evaluative attitude can be criticized, no matter how widespread, and it is open to Rozin simply to insist that it makes no difference who wore this coat. Our argument appeals not simply to a particular human concern, however, but to what we take to be a deep and wide pattern. The examples above exhibit the degree to which an interest in the histories of people and things is a wide human concern: It runs throughout human moral psychology, and is reaffirmed in various attitudes toward agents and the objects with which they interact. These concerns are surely deep as well: It would be very difficult, to say the least, to extirpate our interest in the past and its relation to present day things. It is the psychological depth and width of these commitments— and some others we will be raising in what follows—that allow them to figure in anthropocentric justifications that do not rely on an appeal to mere conformity. We grant of course that this suite of concerns is fundamentally human, and that a different sort of rational creature need not care about them in the least. Indeed, not everyone sees the point of autographs and memorabilia—though we doubt that many otherwise normal people are completely immune to feelings about the genealogy of places and things. But the point of these examples is to illustrate the existence of anthropocentric reasons which, though not binding on all rational agents, are nonetheless genuine and in no way incompatible with science. The temptation to deny their reality, and to attribute magical beliefs to people who are sensitive to such reasons, is a symptom of scientism. It is also grossly uncharitable. Against this scientistic prejudice, sentimentalists should insist that in order to be convicted of magical thinking, you have to believe in magic—that is, you must have some false causal beliefs. Voodoo is genuinely incompatible with science. The fact that you treasure a painting because it was made by your child, or that you do not want anything to do with the lab coat of an infamous Nazi doctor, is not. Neither is the claim that these genealogical facts give us reasons to be attracted or averse to such objects. Haidt’s (2000, 6) notion that there is no reason not to engage in the taboo-violation tasks, or to condemn the harmless but offensive actions, is similarly blind to another source of genuine but anthropocentric reasons. Both these psychologists miss the expressive and symbolic features of action, which provide another class of perfectly good reasons that elude scientism. Consider Haidt’s (2000, 7) claim that there are no good reasons not to sign the faux contract in Soul, because doing so is harmless. In some narrow sense of harm, this is true. But there are perfectly good reasons to decline the offer nevertheless. In the first place, most people are justifiably suspicious about

276  Justin D’Arms and Daniel Jacobson signing documents at the behest of strangers with unknown motives. In an environment full of dubious solicitations and outright scams, this attitude is well founded. Even if the danger in this case seems minimal, the reward is vacuous: “Someone asked me to do it, for reasons he didn’t explain” is a negligible reason, and that is enough to justify inaction. But there is another type of reason at play in these examples, the neglect of which has deeper ramifications. Merely to pretend to sell your soul to a manipulative experimenter is a symbolic act of subjection, and the fact that the contract isn’t binding does not erase this symbolism. Haidt’s failure to acknowledge the symbolic and expressive aspect of action, and its role in generating reasons to act, is strangely myopic. Again the best diagnosis of this blindness is scientism. To be sure, science has disabused us of certain forms of magical thinking involving symbolism. But the actions that Haidt and Rozin consider mere taboo-violation can be much more charitably—and insightfully—understood. Imagine wiping your dirty shoes on a doormat illustrated with the picture of a beloved. (This resembles one of Rozin’s cases, where he has people throw darts at various portraits.) That action expresses contempt, regardless of one’s actual attitude or whether anyone else sees it. This example could easily be made more graphic and obvious, but we will leave the elaboration to any unconvinced reader’s imagination. Claims about the attitude an action expresses do not require the belief that you are hurting the beloved, or even just hurting her feelings. The point is simply that there are reasons not to do things that would express attitudes repugnant to you. Notice that we are not claiming that these reasons are terribly important, let alone that they override any countervailing reasons. If you offered us a roached glass of Chateau d’Yquem or made a decent offer on the faux contract for our souls, we would surely accept—but juice is not appetizing enough to outweigh the disgustingness of drinking traces of sterile bug, and a $2 offer is insulting. Rozin and Haidt presuppose the scientistic position that there is no reason whatsoever to have these aversions and attractions, or to care about symbolism of actions such as wiping one’s feet or cleaning the toilet with something. This is not a scientific claim but a normative one, which follows from an impoverished theory of reasons accepted without argument by some scientists and philosophers, not from any discovery in empirical moral psychology. Human beings are symbol-making creatures who imbue these symbols with meaning. Of course, humans are also prone to engage in magical thinking and even to believe in magic, but those tendencies involve holding false causal claims. No such factual error need be implicated in our reluctance to debase a picture of a beloved or the symbol of a dearly held cause. To consider all such tendencies to be mere prejudice and superstition—that is, taboo—is to alienate oneself from a significant aspect of human life. Moral psychologists should be wary of coming to this conclusion too quickly or without vivid awareness of what they thereby disparage or ignore. Haidt’s scenarios of eating the family pet and cleaning the toilet with the flag suffer from this same bizarre myopia about what can be counted as a good reason. If you love something, this ramifies throughout your plans and attitudes, and you will not want to perform actions

Sentimentalism and Scientism  277 that express indifference or contempt toward it. Admittedly this is symbolic, but there is nothing inherently irrational about being attracted to good symbolism and averse to bad symbolism. On the contrary, this deeply seated aspect of human nature seems, more plausibly, not to stand in need of further justification. We cannot aspire here to mount an adequate defense of the importance of the sentimental values or to conclusively establish the rational significance of desert and the expressive aspects of action. But we take our argument to have demonstrated that these claims are considerably more plausible than the scientistic alternatives offered by our opponents. Moreover, we have shown that the influential work in the empirical ethics movement we have canvassed offers no scientific basis for rejecting our normative claims. Neither relativism nor the strictly forward-looking theory of reasons that these authors favor is mandated by their experiments or hypotheses about the origins of emotional intuitions. Hence these rival views must be contested on their normative merits. In doing so, our strategy has been to remind readers how much of what matters to them would be disallowed by the exaggerated pessimism about rational justification that these views have in common. Just as sentimental values like the disgusting are supported by deep and wide patterns of human concern, so too are concerns with genealogy, and with the expressive and symbolic meanings of actions and objects. In the absence of compelling philosophical argument for such a drastic revision of ordinary concerns and sympathies, there are good grounds to pursue an alternative approach. We have argued in this chapter that science does not present us with a forced choice between an uncritical relativism and an inhuman rationalism. An anthropocentric theory in the sentimentalist tradition—which takes seriously the idea that widespread human concerns can underwrite genuine values while allowing for critical purchase on actual emotional responses—can navigate between the horns of Singer’s dilemma. The view we favor, rational sentimentalism, avoids the dismal moral consequences of relativism (such as its apologia for oppressive social practices) and hyper-rationalism (including its dismissal of rights and desert). It captures a wide range of anthropocentric reasons that issue from the expressive aspects of action and are registered by the emotions, but which are neglected or denied by the various forms of scientism.

References Brandt, R. 1950. The emotive theory of ethics. Philosophical Review, 59, 305–18. Cosmides, L. 1989. The logic of social exchange:  Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31, 187–276. D’Arms, J., and Jacobson, D. 2005. Anthropocentric constraints on human value. In R. Shafer-Landau (ed.), Oxford Studies in Metaethics 1, pp. 99–126. Oxford: Oxford University Press. Gibbard, A. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press. Greene, J. 2008a. The secret joke of Kant’s soul. In W. Sinnott-Armstrong (ed.), Moral Psychology, vol. 3, pp. 35–91. Cambridge, MA: MIT Press.

278  Justin D’Arms and Daniel Jacobson Greene, J. 2008b. Reply to Mikhail and Timmons. In W. Sinnott-Armstrong (ed.), Moral Psychology, vol. 3, pp. 105–17. Cambridge, Mass: MIT Press. Haidt, J. 2001. The emotional dog and its rational tail. Psychological Review, 108, 814–34. Haidt, J. 2005. Interview in The Believer, August 2005. Haidt, J., and Bjorklund, F. 2008. Social intuitionists answer six questions about moral psychology. In W. Sinnott-Armstrong (ed.), Moral Psychology, vol. 2, pp. 181–217. Cambridge, MA: MIT Press. Haidt, J., Bjorklund, F., and Murphy, S. 2000. Moral dumbfounding: Where intuition finds no reason. Unpublished manuscript. Jacobson, D. 2012. Moral dumbfounding and moral stupefaction. In M. Timmons (ed.), Oxford Studies in Normative Ethics 2, pp. 289–315. Oxford: Oxford University Press. Lerner, J., Goldberg, J., and Tetlock. P. 1998. Sober second thought: The effects of accountability, anger and authoritarianism on attributions of responsibility. Personality and Social Psychology Bulletin, 24, 563–74. Maier, N. R. F. 1931. Reasoning in humans: II. The solution of a problem and its appearance in consciousness. Journal of Comparative Psychology, 12, 181–94. Nisbett, R., and DeCamp Wilson, T. 1977. Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231–59. Prinz, J. 2007. The Emotional Construction of Morals. Oxford: Oxford University Press. Roth, A. 1995. Bargaining experiments. In J. Kagel and A. Roth (ed.), Handbook of Experimental Economics, pp. 253–348. Princeton: Princeton University Press. Rozin, P., Millman, L., and Nemeroff, C. 1986. Operation of the laws of sympathetic magic in disgust and other domains. Journal of Personality and Social Psychology, 50(4), 703–12. Schnall, S., Haidt, J., Clore, G., and Jordan, A. 2008. Disgust as embodied moral judgment. Personality and Social Psychology Bulletin, 34, 1096–109. Schwarz, N., and Clore, G.  L. 1983. Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45, 513–23. Singer, P. 2005. Ethics and intuitions. Journal of Ethics, 9, 331–52. Stevenson, C. 1937. The emotive meaning of ethical terms. Mind, 46, 14–31. Stevenson, C. 1944. Ethics and Language. New Haven: Yale University Press. Wheatley, T., and Haidt, J. 2005. Hypnotically induced disgust makes moral judgments more severe. Psychological Science, 16, 780–4. Wiggins, D. 1998. Needs, Values, Truth. 3rd edn. Oxford: Oxford University Press.

Index accountability  3, 41, 43–6, 50–2, 54–6, 89–91, 96–7, 106–7, 125–6 accountability-seeking attitudes  69–71 memory 91–2 morality as accountability  3, 41, 46, 50–2, 54–6, 59–60, 63–8 sense of agential identity  96–7, 106–7 action ownership  90–3, 116 action production  197–8 action readiness states  169–71, 173 action tendencies  168–82 extinction  174, 179 focused versus unfocused  171 inaction tendencies  171–2, 178–9 precedence 170–1 preparation  170–1, 173 prioritization 170 advance directives  105–6 affect  14–15, 51, 127–31 affective deficit  26 affective disorders  128–35 brain damage  130–31 depression  131–2, 134 affective forecasting  51, 133–5 agency  4–5, 84–7, 109, 119, 123–8, 133–5 agential identity  96–8, 105–7 altruism  53, 56 Alzheimer's disease  88, 90, 96, 100, 105–6, 139 anger  53–4, 129, 168–9, 261–4 anhedonia 131–4 animalist account of identity  144–5 anthropocentrism  7–8, 254, 267, 273–7 Appiah, Kwame Anthony  62–5 apologies  51–2, 66–7 appetites  188, 196 appraisal theory  146–7 approval motive  54–9, 66–7, 150 aretaic emotional responses  98, 106 Asch, Solomon  61, 113 associative expectation system  195 attention  59–60, 170–1, 200 attributability  92, 97–100, 104–6 aversions  188–9, 196, 210, 272–4 to violation of moral norms  28–9 autism  127, 135 automatic processes  9–12, 109, 127, 165 autonomy 106 awareness  109–10, 116–7 Batson, C. Daniel  52–4, 56–9

behaviorism  168–9, 189 belief and desire pairs  6, 160–4 blame  43–5, 53, 55–6, 63, 69–71, 110, 117–8 perceptions of blameworthiness  48–9 Butler, Joseph  91 bystander effect  111–8 care: care and commitments  99–104 care for others  127–8, 133–5 care-commitment clusters  102–4 love and care  205–7 Carruthers, Peter  127 central motive states  172 character  98–100, 105, 115–6, 135 character traits  102–4 communication: of blameworthiness  51–2 of moral demands  89, 92–7 compatibility control  172–5 compulsions  85, 205–6 condemnation  44–6, 52–3, 69–70, 257–8, 271 accountability theory  46, 50–2 deterrence theory  48 egoistic theory  47–8 retributive theory  48–9 confabulation 255–62 conscience  3, 45–6, 54–68 accountability theory  54–6, 59–60, 63–8 approval theory  54–60 conscious access  5, 110–1, 116–20, 126–8 consciousness of past action  90–2 contempt  43, 69–70, 276–7 control precedence  169–70, 175–6, 178–80 core relational theme  161, 166, 168, 177–82 counterintuitive judgments  3, 14–18, 28, 30, 34 Damasio, Antonio  128–31, 164–5 Darwall, Stephen  41, 43 Darwinian dilemma  222–4, 230–1, 233–7, 239–40, 244, 246–8 deep self views  114–5 deliberation  3, 10–15, 17, 19–25, 28, 34–5, 95–6, 133, 193–4 deliberative processing, see deliberation dementia  4, 26, 86–90, 95–105 deontological judgments  3, 10–12, 14–16, 20–1, 25–9 depression  131–2, 171–2

280 Index descriptions of actions  109–12, 116–20 desert  260–4, 277 desires  6–7, 63, 100, 124, 132, 160–4, 186–9, 196–211 as a natural kind  6, 187–8, 202–4 motivational theory  186–7 reward theory  187–9, 197, 204–10 standing desires  6, 132 direction of fit  160–1, 177 disgust  70–1, 169, 182, 256–7, 272–4 displeasure 199–200 dispositions  4, 6, 100–1, 103–4, 186, 190 Doris, John  115 dual process model  3, 10, 13–17, 26, 34–5 dumbfounding 257–8

grounding account  233–7 theoretical-reasoning account  233, 235–7 third-factor account  227–33, 236–9, 245, 247–9 tracking account  223–7, 231–2 executive capacities  96, 100–1, 103, 172–3 expressive aspects of action, see symbolic aspects of action

egoism  31–2, 218–220 embodied appraisals  166–7, 177 emotions  43–4, 125, 127–31, 135, 146–51, 156–83 backward-looking 45 basis of morality  253–7; see also evaluative judgments as rationalizations, rational sentimentalism, sentimentalism bodily changes  159, 162–9, 173, 181, 200 cognitive and motivational components  43–4 cognitivism  160–4, 177 differentiation 178–82 distancing 152 flexibility  158–9, 163, 165, 167, 172–3 incidental versus integral  256–7, 262–9 informational-cum-motivational function  178–9, 181–2 intentionality 177–8 motivational theory  5–6, 156, 168–71, 176–83 perceptualism  164–8, 177 reflexive perceptualism  164–5 valenced perceptualism  165–7 pre-potent 21–2 reactive emotions  89, 92–4 responses to counterintuitive judgements  28 salience 53 self-conscious emotions  146–7, 150–1 see also aretaic emotional responses, moral emotions emotivism 266–7 empathy  27–8, 31, 53, 56, 125, 127 episodic foresight  138, 144 episodic memory  138–46, 148–51 episodic sense of identity  141, 144–9, 151 etiological sense of function  178 evaluative judgments: as rationalizations  7, 253–4, 258–62; see also confabulation evolutionary argument  2, 7, 215–6, 222–4, 233 evolutionary origins of morality  7, 222–3, 226–7, 231–2, 234–5, 244–50 adaptive-link account  223–6, 231, 237, 248

Gallistel, Charles  172 gratitude  98, 264 Greene, Joshua  3, 8, 9–26, 33–4, 255, 258–65 grief  129–30, 171–2 grounding  217–22, 228–30, 245, 247–8 guilt  4, 28–9, 43–5, 66–71, 146–8, 150–2, 175–6, 181–2

fear  6, 158–9, 161–3, 166–7, 173, 176–8, 180, 200 feelings  159, 180–1, 191, 199–200, 207–8 fission 148 footbridge dilemma, see trolley problem forgiveness 50–1 forward-looking intentions  174–5, 179 Frijda, Nico  156–8, 169–71, 176–7

Haidt, Jonathan  55, 68–71, 255, 257–8, 268–72, 275–6 Hume, David  5, 123, 125, 127, 129–30, 160 Humean constructivism  216, 218, 232–6, 240–8 hyper-rationalism  255, 264, 268 impulsivity  6–7, 157–8, 162–3, 165, 167, 173, 176, 179, 206, 208 in-virtue-of relation  218, 228, 240 indignation  53, 61, 94 inferential reasoning  23, 118–20 information processing biases  158, 167 inner state goals  166 intentionality of emotions  177–8 intentions-in-action  162, 173 intuition (faculty)  9–11, 19–21, 23, 34–5 moral intuitions  20–1, 24–30, 215, 239–40, 245–6, 248, 254–5, 264–5, 271 James, William  138, 164–5 Jaworska, Angieszka  105–6 joy  171, 207 Kant, Immanuel  14–5, 46, 137, 148 Lazarus, Richard  146–7, 161 learning, see reward learning Locke, John  90–2 love  6–7, 205–8 magical thinking  271–4 memory  2, 4, 5, 87, 90–2, 95–7, 138–46, 193 episodic  138–46, 148–51

Index  281 false episodic memories  145–6 field versus observer perspective  141–2 memory and action-ownership  90–2 memory impairments  4, 87, 88, 95–7, 139–41 semantic  138, 143 mental actions  169–70, 175, 181, 199 meta-cognition  5, 124, 126 metaphysical dependency, see grounding Milgram, Stanley  62–5 mind-reading  124, 127 moral agency  4–5, 119–20, 123–5, 127–8, 133 moral condemnation, see condemnation moral deliberation, see deliberation moral emotions  43, 68, 146–7, 149–50 moral hypocrisy effect  57–60 moral licensing effect  57–60 moral obligations  42–4, 54–7, 63–5, 67–8 moral realism  215–20 moral responsibility, see responsibility moral systems  68–70, 268 morality: distinctiveness 68–71 motivated irrationality  200–1 motivation  1–2, 128, 132–4, 205–10 belief and desire pairs  160–3 emotional motivation  43–5, 165–7, 178–9 Humean theory  160 moral motivation  46, 54–62, 68 motivational internalism  225–6 motivations versus motives  166 motivational theory of desire, see desires: motivational theory motivational theory of emotions, see emotions: motivational theory Nagel, Thomas  148 natural kinds  6, 187–8, 202–4 neglect  110–1, 117–8 net reward calculation  194–6 Nietzsche, Friedrich  69 non-inferential judgment, see intuition normative antirealism  216–22, 232–4, 236–7, 239–40, 245 neurocognitive disorder  88 neuroimaging studies  11–12, 28–9 obligations, see moral obligations omission 110 open question argument  42 operant conditioning  189 Parkinson’s disease  198, 200–1 past lives  145–6 personal identity  90–2, 95–8, 137, 144–9 perspective-taking  117–20, 135 pessimism, see evaluative judgments as rationalizations

pleasure  134, 199–200, 202, 206–7, 210 pre-potent response  21, 24 preference utilitarianism  218–20 pride  146, 150 Prinz, Jesse  125, 166–8, 177–8, 255, 268–70 pro tanto reasons 23, 241 post-traumatic stress disorder  141–3 psychoticism 28–9 psychopathy: lack of affect  127–8, 130, 132 utilitarian judgments  27, 29, 31 public goods game  47, 51, 61, 63 punishment  44–5, 47–52, 61, 92, 260–3 punishment system  188–9, 191–7, 200–1 quality of will  84–6, 89, 105 rational egoism  31–2 rational sentimentalism  7–8, 149, 254–5, 267, 271, 273–4 reactive attitudes  4, 41, 43, 84, 89, 92–7 reasons for action  6, 166, 209, 216, 243, 259, 267 reflex actions  157, 176–82 relational goal of emotion  169, 171–8, 181–2 relativism  254–5, 268–71 resentment  84, 92–7 responsibility  4–5, 70, 84–9, 94, 106, 108–10, 113–6, 118–20, 151 attributativity-responsibility  98–9, 100, 105 marginal cases  85–7 responsibility responses  97 taking responsibility  44, 51–2, 65–6 retributivism  48–52, 260–5 reward learning  189–97, 201 reward system  6, 187–8, 191, 194–201, 204, 207 Ross, W. D. 23 Rozin, Paul  271–4, 276 sadness 171 scientism  253–5, 271–7 self (psychological concept)  137–43, 151 self-appraisal  5, 126 self-approval 59 self-conscious emotions  146–7, 150–1 social function  150 self-reflection  4, 90, 118–20, 146–7 self-regulation  5, 123–5, 127–9, 133–5 self-regulation failure  128 sentimentalism  123–7, 264, 266–7; see also rational sentimentalism shame  66–70, 100, 103–4, 146–7, 150–1, 181–2 Sher, George  110, 116 shortsightedness 158 situational influences  109–12, 114–5; see also bystander effect, stereotyping social approval  60, 63, 66 social intuitionism  255, 267–8

282 Index social norms  60–5 skepticism  224, 231, 254–5 Singer, Peter  7, 32, 254–5, 259 stereotyping  108–9, 111–2 Strawson, P. F. 41, 43, 84–6, 93–4, 104, 126 Street, Sharon  7, 215–27, 231–49 Stroop paradigm  20–1 symbolic aspects of action  275–7 sympathetic engagement  127–8, 133–5 taboo  70, 271–2, 275–6 trait self-knowledge  139–41

trolley problem  3, 11, 13–22, 31, 34 utilitarian judgments  3, 10–22, 24–5 utilitarian reasoning  18–19, 21–34 utilitarianism 42–3 preference utilitarianism  218–20 valence markers  166–7, 177 volitional control  103–4 Watson, Gary  85–6, 93, 97–8 weighing reasons  22–4, 28–9 Williams, Bernard  24–5, 218, 221