Strokes of Luck: A Study in Moral and Political Philosophy 9780198868507

Citation preview

Strokes of Luck

Strokes of Luck A Study in Moral and Political Philosophy GERALD LANG

1

1 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 © Gerald Lang 2021 The moral rights of the author have been asserted First Edition published in 2021 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2021937350 ISBN 978–0–19–886850–7 DOI: 10.1093/oso/9780198868507.001.0001 Printed and bound in the UK by TJ Books Limited 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.

In memory of Shepley Orr

Preface and Acknowledgements As an undergraduate student, I was assigned, and duly read, the two classic articles on moral luck by Thomas Nagel and Bernard Williams. I was gripped, but puzzled. I could grasp Nagel’s worries about the ramifications of luck in our moral lives, and I agreed with him that those ramifications surely needed to be controlled in some way. But how? That was much less clear to me. Once Pandora’s Box has been opened, the damage has already been done. It cannot be undone merely through the effort to close it again, or by pretending that it had never been opened in the first place. Despite a number of readings of it, I was even more puzzled by Williams’s article. Its connection to Nagel’s concerns seemed opaque to me, and I struggled to see how exactly the killer blow to ordinary morality had been administered, though I was also intrigued and stimulated, as many others have been, by Williams’s charismatic opposition to morality’s more claustrophobic aspects. The years passed, and I became interested in, and puzzled by, other matters. At some point, I began to do sustained work on distributive justice, and on luck egalitarianism in particular. It occurred to me that some of the central cases one encountered in the justice literature had a similar structure to the cases often encountered in the moral luck debate: pairs of agents, doing similar things which turn out differently, and being blamed or rewarded to different degrees. Meanwhile, something about the way in which luck egalitarians had set up their basic case was leaving me dissatisfied. The interpersonal pairwise comparisons standardly prioritized by luck egalitarians may reveal the presence of luck, but this form of luck did not strike me as one which manages to track anything worth tracking. It does not, therefore, require correction, or annulment. It is a source of distortion, not the basis for a serious complaint. I began to wonder whether there were common lessons, or else interestingly contrasted lessons, to be recovered for both debates: the debate about blameworthiness in normative ethics, and the debate about just distributions in political philosophy. The eventual result, leaving aside some other twists and turns, is the book before you. Because luck intrudes in our lives in various ways, it raises a large number of philosophical issues connected to our practical existence. Not all of them will be tackled in this book. With the exception of some strands of discussion towards the end of Chapter 1, I will be largely abstaining, for example, from examining the role of luck in moral epistemology, though this is an important topic which has been intensively studied in recent years, especially in the light of reflection on our Darwinian inheritance. I will be similarly sparing about my attention to the role

x  Preface and Acknowledgements of luck in epistemology more generally, aside from some modest stage-­setting and a small number of policy announcements in the Introduction. More seriously, and except for some brief skirmishes in the Introduction, Chapter  4, and Appendix II, this book omits any concentrated and detailed discussion of luck, agency, and free will and responsibility. Other writers interested in moral luck have had a great deal to say about luck and free will, and I am not claiming that they strayed beyond their brief, or misidentified what can be fruitfully discussed under this remit. So this particular omission does call for more concerted defence; I will try to explain myself in the Introduction. I have been thinking about these issues for a number of years, on and off, and I have accumulated many debts, both institutional and personal. First and foremost, I thank the Mind Association, which was kind enough to award me a fellowship relieving me of normal teaching and administrative duties for one semester in the 2016–17 academic year, and which therefore allowed me to complete a working draft of a decent chunk of it within that year. Without that period of leave, completion of this project would have been severely delayed. In association with this award, I was invited, in July 2017, to present some of this work at the Joint Session of the Aristotelian Society and Mind Association in Edinburgh. I thank those who were there for their good advice and heart-­ warming encouragement. The Mind Association period of leave was conjoined with a standard sabbatical period of leave provided by the School of Philosophy, Religion, and History of Science at the University of Leeds, which has been an enjoyable and stimulating environment for doing philosophy over the last fifteen years or so. I am very grateful to my colleagues, and ex-­colleagues, for their intellectual camaraderie, good advice, and friendship, and for having sat, kindly and patiently, through early-­doors efforts to articulate my various philosophical hunches about these and many other issues. I have learned, and continue to learn, a huge amount from them. I am also grateful to philosophical and philosophically interested friends elsewhere, many of them encountered every now and then at conferences, seminars, and workshops. I do not want to risk offence by mentioning some only to fail to mention others, but I do wish to thank on this score Lucy Allais, Gustaf Arrhenius, Elizabeth Barnes, Jessica Begon, Corine Besson, Thomas Brouwer, Krister Bykvist, Ross Cameron, Jennifer Carr, Sophie-­Grace Chappell, Pei-­Lung Cheng, Alix Cohen, Bill Cooper, John Divers, Jamie Dow, Ed Elliot, Daniel Elstein, Carl Fox, Helen Frowe, Mollie Gerver, Rachel Goodman, James Harris, Antony Hatzistavrou, Greville Healey, Tom Hancocks, Ulrike Heuer, Iwao Hirose, Kent Hurtig, Ward Jones, Matthew Kieran, Rob Lawlor, Gail Leckie, Kasper Lippert-­Rasmussen, Heather Logue, Brian McElwee, Andy McGonigal, Chris Megone, Aaron Meskin, Margaret Moore, Daniel Morgan, Gary Mullen, Morten Nielsen, Martin O’Neill, Alex Pelling, Oliver Pooley, Adina Preda, Massimo Renzo, Simon Robertson, Léa Salje, Paolo Santorio, Joe Saunders, Daniel Schwartz, Tasia Scrutton, Shlomi Segall, Scott Shalkowski, Matt Smith, Helen

Preface and Acknowledgements  xi Steward, Adam Swift, Victor Tadros, Georgia Testa, Cain Todd, Patrick Tomlin, Jason Turner, Pekka Väyrynen, Kristin Voigt, Alex Voorhoeve, Robbie Williams, Nicole Woodford, Richard Woodward, and Nick Zangwill. I have bent the ears of Jess Isserow and Jack Woods, in particular, about these issues a number of times over the last few years, and Jack also went to the trouble of providing helpful and instructive comments on a number of the early chapters. I am very grateful to them. A manuscript workshop was held on an earlier complete draft of this book in November 2018 in Dublin, under the auspices of the Political Studies Association of Ireland. I am very grateful to Adina Preda and Peter Stone for the invitation, and the PSAI for the financial support. In addition to Adina and Peter, I also thank John Baker, Brian Carey, Christopher Cowley, Katherine Furman, and Brian O’Connor for going to the trouble of reading and commentating on the various chapters. I learned a huge amount from them, though I worry that they will not be completely satisfied with my attempts to accommodate their concerns. For many years now, Roger Crisp and Brad Hooker have been encouraging and supportive presences in my career. I thank them warmly for everything they have done for me. In May 2015, I organized a workshop on the Dimensions of Luck, held in Leeds, where the speakers were Roger Crisp, Duncan Pritchard, Katie Steele, Daniel Statman, and Zofia Stemplowska. I enjoyed it so much that I finally talked myself into writing a book about at least some aspects of luck. I thank them all. Thanks also to the Faculty of Arts in Leeds for the funds enabling me to hold this workshop. There are also substantial and more specific debts on the various chapters here. The arguments presented here in Chapters 1 to 3 were presented, in slightly different and sometimes over-­compressed formats, at seminars in Cape Town, Leeds, Oxford (at the Moral Philosophy Seminar), St Andrews (at the Philosophy Club), Sussex, and Warwick (at the Centre for Ethics, Law, and Public Affairs). I also ran an enjoyable class as a visiting lecturer at Rhodes University in Grahamstown in 2012, in which some of the early moves were written up and tested out. I thank everyone at these events for their valuable advice and criticism. The argument now distributed between Chapter 4 and Appendix II was much benefited by the discussion of a shorter version of it in a meeting of the White Rose Early Career Ethics Researchers in 2016, organized by Richard Chappell. I thank, in addition to Richard, Jessica Begon, Daniel Elstein, Carl Fox, and Johan Gustafsson for their insightful comments. An earlier version of Chapter 5 was presented at a conference on the thirtieth anniversary of the publication of Bernard Williams’s Ethics and the Limits of Philosophy in Oxford, organized by Sophie-­Grace Chappell. I thank Sophie-­Grace for the invitation, and I am grateful to the informed and enthusiastic audience for the many extremely useful comments I received on that happy occasion. A revised version of it was then presented in Leeds; thanks to everyone for comments I

xii  Preface and Acknowledgements received there and then. For further useful exchanges and comments, I thank Victor Durà-­Villà and Jake Wojtowicz. Some of the material in this chapter has already appeared in ‘Gauguin’s Lucky Escape: Moral Luck and the Morality System’, published in Ethics Beyond the Limits: Bernard Williams’ Ethics and the Limits of Philosophy, edited by Sophie-­Grace Chappell and Marcel van Ackaren, and published by Routledge in 2019. Chapter 6 is partly constructed out of already published work on luck egalitarianism, notably ‘How Interesting is the “Boring Problem” for Luck Egalitarianism’, published by Wiley in Philosophy and Phenomenological Research in 2015. An early version of some of that material was presented at the Society of Applied Philosophy Conference in Zurich in 2013. The comments I received there were extremely useful as I worked further on this material. In addition, I thank Kasper Lippert-­Rasmussen, Adina Preda, Daniel Schwartz, Shlomi Segall, Saul Smilansky, and Kristin Voigt for vital further exchanges about these topics, and Nicola Mulkeen for detailed and instructive comments on a later draft of the chapter. An embryonic version of Chapter 7 was presented at the AHRC Workshop on Equality at the University of Exeter, organized by Keith Hyams and Robert Lamb in 2009; a couple of later versions were presented, to different audiences on different occasions, in Leeds. I am grateful for the many helpful comments I received on those occasions, and am particularly grateful to Jerry Cohen, in what sadly turned out to be the last time I saw him, for a very useful suggestion about what exactly I should be looking for in his Rescuing Justice and Equality. A yet more recent and refocused version of some of the material on Rawls was presented at the IDEA Seminar in Leeds, at the Society of Applied Philosophy conference in 2018, where I had helpful exchanges with Miranda Fricker, Kasper Lippert-­ Rasmussen, Serena Olsaretti, and Shlomi Segall, and most recently at a MANCEPT seminar in Manchester, where I received helpful insights from Steve de Wijze, Nicola Mulkeen, Miriam Ronzoni, Hillel Steiner, and others. Thanks also to Adina Preda for useful early discussion and to Pei-­Lung Cheng for comments on a later draft of the chapter. Chapter 8 draws upon a number of sources. One of them is a talk on international justice which was presented in rather different versions at the Conference on the Demandingness of Morality, the AHRC Scottish Ethics Network, at the University of Stirling, and also in Leeds, Stockholm, Sheffield, and at the Conference on Poverty, Charity, and Justice, held at the University of Witwatersrand. I offer my thanks to, respectively, Kent Hurtig, Gustaf Arrhenius, Helen Frowe, and Lucy Allais for these various invitations, and I am grateful to these different audiences for their challenging, constructive, and insightful comments. This chapter also embodies ideas and passages from some work on animal ethics and discrimination which was published in the Oxford University Press collection Luck, Value, and Commitment: Themes from the Ethics of Bernard

Preface and Acknowledgements  xiii Williams, which I co-­edited with Ulrike Heuer. Those whom I thanked there are duly thanked again. Finally, this chapter draws upon some ideas about basic equality which were put to work in my review of Jeremy Waldron’s One Another’s Equals, published by Oxford University Press in Mind in 2019. My early ideas on basic equality were presented at the Workshop on Equality: Grounds, Scope, and Value, organized by Adina Preda in Oxford in 2015. A subsequent version of that paper was presented in a panel at the Society of Applied Philosophy in Belfast in 2016, with Adina Preda, Kristin Voigt, and Ian Carter as my co-­panellists, and then at seminars in Birmingham, Leeds, the London School of Economics, and York. I am very grateful to everyone involved for their comments and insights. Adina Preda, in particular, has been an active and important interlocutor for most of these chapters, as these acknowledgements attest. I am especially grateful to her. As ever, I am grateful to Peter Momtchiloff at Oxford University Press for his patience, professionalism, and friendly encouragement. It is always a pleasure to work with him. Two anonymous reviewers for OUP made a number of very shrewd suggestions on an advanced draft that I have tried my best to incorporate, and a number of telling objections that I have tried my best, in various ways, to disarm. They may be within their rights to have expected even more, but the final version of the text is, I hope, considerably better for the attention they have paid to it, and they have my earnest gratitude. There is something about writing a monograph that can make you think you are doing philosophy for the first time. Presumably that has to do with the concentrated collection of intellectual demands, and feelings of exposure and vulnerability over the longer distances. In any case, I want to acknowledge some other, more personal debts. I thank Ronan McDonald and Sarah Montgomery for their friendship, and for always being there, at least via social media. I also thank my brother, Jim, for his friendship and company. My educational path was nourished and supported, generously and uncomplainingly, at every point by my mother, Ann Lang, and my late father, William Lang. I am forever grateful to them. My final acknowledgement combines intellectual and personal debts. My old friend Shepley Orr passed away as the first draft of this book was nearing completion. I regret that we never found the opportunity to discuss many of the arguments that would have been deepened and sharpened had he laid eyes on them, and I remember with affection all the hours, from years past, of enlightening and often hilarious conversations about philosophy and just about every other topic under the sun. I miss him keenly. This book is dedicated to his memory. Leeds July 2020

Gerald Lang

Introduction 1.  Moral Luck: Old and New Philosop­­­hy still has the capacity to surprise, but the issues it investigates tend to have been on the books for a long time. The problem of moral luck was strongly anticipated by Adam Smith: Had Caesar, instead of gaining, lost the battle of Pharsalia, his character would, at this hour, have ranked a little above that of Cataline, and the weakest man would have viewed his enterprise against the laws of his country in blacker colours, than, perhaps, even Cato, with all the animosity of a party-­man, ever viewed it at the time. His real merit . . . would have been acknowledged . . . But the insolence and injustice of his all-­grasping ambition would have darkened and extinguished the glory of all that real merit . . . Fortune has in this . . . great influence over the moral sentiments of mankind, and, according as she is either favourable or adverse, can render the same character the object, either of general love and admiration, or of universal hatred and contempt.1

It is beyond reasonable doubt, however, that the specific impetus behind contemporary discussions of moral luck was a symposium on moral luck by Bernard Williams and Thomas Nagel, published by the Aristotelian Society in 1976.2 The phrase ‘moral luck’ is also owed to Williams. Most of the recognizable contributions to that particular debate that have been made in the intervening period in Anglo-­American philosophy have proceeded, in one way or another, from the stage-­setting originally assembled by Nagel or Williams. What is perhaps less obvious, but still strongly arguable, is that the debate on moral luck that has emerged since then owes rather more to Nagel than to Williams. Though bits and pieces of Williams’s article are routinely referred to, and sometimes receive substantive discussion—one thinks here, in particular, of the Gauguin case, or the ‘agent-­regret’ experienced by the lorry-­driver in one of 1  See Smith (1759/1976), II.iii.2; for a recent useful discussion, see Hankins (2016). Nagel (1979), pp. 31–2, acknowledges Smith’s important path-­finding work in this area, and Crisp (2017) examines Smith’s views in some detail. See also Section 5. 2  See Williams (1976), and Nagel (1976). Both essays were subsequently revised and reprinted, and the literature on moral luck usually deals with these later versions: I will, accordingly, be referring mostly to Williams (1981a) and Nagel (1979). Another precursor for the contemporary moral luck debate is Feinberg (1962), as Hartman (2017), p. 2, also notes. Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0001

2 Introduction Williams’s imaginary examples—there is something elusive about Williams’s treatment of moral luck which has obstructed its comfortable or wholesale in­corp­or­ation into the moral luck literature. This is because, when examined more deeply, Nagel and Williams are chasing different philosophical scents; their quarries are different. Williams’s distinctive concerns will be revisited in detail in the course of the book. In the preliminary orientation for the moral luck debate that I provide in this Introduction, I will be chiefly concerned with Nagel. But it is not a bad idea to start even further back.

2.  What is Luck? In his article, Williams does not provide any analysis of the concept of luck, and simply adopts the policy of using the notion of luck ‘generously, undefinedly, but, I think, comprehensively’.3 Might we properly expect a greater degree of ana­lyt­ ic­al candour than this? Resistant to analysis though he might be, Williams’s rough working idea seems to be that the presence of luck denotes an absence of control over the outcomes of acts and plans. The same basic characterization of luck is also employed by Nagel: Where a significant aspect of what someone does depends on factors beyond his control, yet we continue to treat him in that respect as an object of moral judgment, it can be called moral luck.4

But Nagel himself encourages connections to be made between ethics and epis­ tem­ol­ogy.5 And, if moral luck is definitely a species of luck, then our characterization of luck might properly require more critical detail, because we will need to know what luck amounts to, across both moral and epistemological contexts, if we are going to successfully identify what a particular sub-­species of luck—moral luck—amounts to. This approach is upheld by at least some contributors to the moral luck debate.6 On such views, there are both opportunities to get ahead, and dangers to avoid. Rik Peels, for example, is convinced that ‘we need to get a firm grip on the notion of luck before we can solve the problem of moral luck’.7 And Steven Hales warns us that ‘moral luck is possible only if one assumes a specific theory of luck, one that is not a suitable account of epistemic luck’.8 In fact, because Hales thinks that all the major theories of luck face serious critical problems, and because he assumes that moral luck must be a genuine species of luck, he doubts whether 3  Williams (1981a), p. 22. 4  Nagel (1979), p. 26. 5  Nagel (1979), pp. 27, 36. 6  See, for example, Peels (2015), Hales (2015), Pritchard (2005), ch. 10, and Brogaard (2003). 7  Peels (2015), p. 77. 8  Hales (2015), p. 2385.

What is Luck?  3 there is any interestingly unified phenomenon of ‘moral luck’.9 If Hales is right, the debate on moral luck will have been floundering all this time under mistaken assumptions. So the stakes are high. Is moral luck a species of luck that can be generally applied across both moral and epistemological contexts? Before we continue, it might be noted that it is not just the areas of moral and political philosophy which have been slow to open up the concept of luck to ana­ lyt­ic­al inspection. As noted by Duncan Pritchard, a notable proponent of anti-­ luck epistemology: The difficulty that faced anti-­luck epistemology when I first tried to develop it . . . was that I found to my surprise that there was next to nothing in the philosophical literature on the nature of luck. Indeed, I think it is fair to say that luck was at this time largely treated as an undefined primitive.10

Pritchard and others have attempted, accordingly, to make good on these lacunae.11 The notion of luck seems to belong to the same conceptual neighbourhood as the notion of the accidental, or the unforeseen, or the uncontrollable, or the unchosen. But these notions do not always go together, thus necessitating a more fine-­tuned approach. Williams and Nagel propound what might be referred to as the Lack of Control Account of luck. The Lack of Control Account may have deficiencies as a general account of luck. One counterexample that has emerged in the literature is provided by the humdrum daily event of the sun’s rising. This is an event over which we lack control. It is also an event which is significant to us. We cannot dismiss it as not being a matter of luck because it has no implications for our lives and interests.12 Be that as it may, we would not normally describe the sun’s rising as a matter of luck.13 Pritchard advances an account which neatly provides for the sunrise case.14 This Modal Account analyses luck in terms of outcomes in the nearest possible worlds with the same initial conditions as the actual world. More precisely, and as Pritchard originally presented it, the Modal Account holds that an event is lucky if it satisfies two conditions: first, it is an event that occurs in the actual world but

9  See also Hales (2016). 10  Pritchard (2014), p. 595; cf. Pritchard (2005), p. 125. 11  It is a large and growing literature. For just a small sample of it, see, for example, Pritchard (2005), (2014), Lackey (2008), Coffman (2014), (2015), and Hales (2016). (I should add here that the ‘strokes of luck’ account advanced by Coffman has essentially nothing to do with the commitments defended in this book, its title notwithstanding.) 12  For the addition of this significance condition, see Pritchard (2005), p. 126. The significance condition is opposed to any simple account of luck which explicates it merely in terms of the low probability of an event’s occurrence. 13  See Latus (2000), p. 167, and Pritchard (2005), p. 127. 14  See Pritchard (2005), pp. 128–33.

4 Introduction not in a wide class of the nearest possible worlds where the relevant initial conditions for that event are the same as in the actual world; and second, it satisfies the significance condition.15 The sunrise does not count for us as lucky, because it would not fail to rise in any of the nearby possible worlds. A huge amount about our world, and about the physical laws applying to it, would need to be altered for tomorrow’s sunrise not to take place. The Modal Account also provides a decent accommodation of other common cases. If I buy a winning lottery ticket, I am considered lucky. Why am I lucky in this case, but not lucky in the sunrise case? Both of them lie beyond my control. It is because, in nearby possible worlds, I would very easily not have been in possession of a winning lottery ticket: all it would take for me to have a losing ticket, rather than a winning ticket, is for me to have purchased another ticket, or for a slightly different combination of numbers to have been chosen. By contrast, the sun would not stop rising in nearby possible worlds. Finally, and in any case, the Modal Account makes a decent fist of the central cases that engage with the particular concerns of the Control Account. It may be a matter of good luck that my negligence produces no harmful consequences, and a matter of bad luck that my negligence leads to disaster. Though I may be negligent and experience good luck, I could so easily have experienced bad luck. Similarly, I may be negligent and experience bad luck, but I might just as easily have experienced good luck. I may make a malicious attempt on someone’s life: even if this attempt is successful, I might easily have produced only an unsuccessful attempt, and if my attempt is unsuccessful, it might just as easily have been successful. As a generally serviceable notion of luck, the Modal Account may score more highly than the Lack of Control Account. But that does not reveal the Lack of Control Account to be an inadequate basis for the problem of moral luck.16 As I see it, Williams’s insouciance was justified; the various wrinkles in the current lively debate on the analysis of luck will not pay dividends if our particular aim is to understand the significance of luck in moral and political philosophy. The most salient pressures favouring the Modal Account over the Lack of Control Account of luck simply do not apply to the debate on moral luck. Why is this? The very restriction to the moral domain ensures that we will be talking about items that relate to moral appraisability, in either its evaluative or deontic guise. This means that we will be concerned with moral agents, their acts, 15  Pritchard (2005), pp. 128, 132. It should be noted that Pritchard (2014) withdraws his former commitment to the significance condition; I will not pursue the details here. 16  There are denials to consider here: see Pritchard (2005), ch. 10, and Peels (2015). Some of these will be considered as we go along: see, for example, Chapter 2, n. 29. See also Hartman (2017), pp. 25–7, for some sensible reservations about the ostensible advantages of the Modal Account for the­or­ iz­ing about moral luck.

Types of Moral Luck  5 their dispositions, their thoughts, and the practices in which they participate and the institutions they create. Counter-­examples to the Lack of Control Account such as the sunrise case do not damage the application of the Lack of Control Account to moral luck. This is because the very restriction to the moral domain ensures that we will be talking about morally appraisable items—acts, intentions, and so forth—that relate to moral agents, and what matters here is the degree of control that agents enjoy over these items. These ideas can be further enriched if we take a more detailed look at the particular problems associated with moral luck.

3.  Types of Moral Luck The problem of moral luck arises out of a tension between two deep facts about moral thought and practice. On the one hand, we are reflectively inclined to affirm that the objects of moral appraisal—agents, actions, practices, institutions, political policies, and so on—should not be the product of luck, or due to the operation of factors which lie beyond agents’ control. On the other hand, we also find upon reflection that our moral appraisals are saturated with luck-­dependent content, and that they could not be easily revised to eliminate the operation of luck without leaving those appraisals with indeterminate content, or no content at all. Nagel outlines a number of different areas of moral luck, and it is customary to enumerate them in preliminary discussions of the subject. I have no innovations to offer in this respect. These are recognizable landmarks in the moral luck landscape, and they provide helpful orientation. One of these—and still the most actively discussed in the literature—is ‘resultant’, ‘outcome’, or ‘consequential’ luck: luck in the results, outcomes, or consequences of the acts agents perform, or attempt to perform. Acts have worldly consequences which are not determined purely by the agent, even if these consequences conform to the agent’s intentions. To take a mundane and morally in­nocu­ous example, if my intention is make a cup of coffee for myself, then I may succeed in doing that—I usually do—but the world has to cooperate with me in certain ways, and I lack at any particular moment complete control over how the world is. The kettle has to be working; the tap water must be flowing; someone else must have seen to it that coffee beans, rather than something that merely looks like coffee beans, have been deposited in the packet of coffee; it must be the case that no one has stolen all my coffee cups overnight, and so on. If the world does not cooperate with me, then my intention will be frustrated, and my coffee-­ making exercise will not turn out as planned. This will be true of any acts, including acts which are much more morally consequential.

6 Introduction One familiar example of morally consequential acts concerns a pair of drunken, reckless drivers.17 One of these drivers has an unsteady but uneventful journey home, harming no one. The second drunken driver hits and kills a pedestrian. It was just a matter of luck that this pedestrian was in the wrong place at the wrong time. The results of the drunk drivers’ acts are very different, but this difference is a matter of luck. Similarly, consider two assassins: one of them succeeds in his attempt to assassinate his target, whereas the other fires a bullet which ends up in a passing pigeon. The difference between them, once again, is just a matter of luck. ‘Circumstantial luck’ is luck in respect of the degree of morally revealing or challenging circumstances which an agent encounters. Nagel gives the example of German citizens who were exposed to the emergence of the Third Reich in the 1930s.18 Though it may have fallen on each of them, morally speaking, to do their bit in resisting and subverting this political regime, and though we may blame them for failing in large measure to do that, it cannot be denied that they were placed in much more morally exposed circumstances than, say, the citizens of the typical South American country at the time. The same person who became an officer in a concentration camp in the Third Reich might have led a relatively blameless life as an émigré businessman in Argentina, if only his background circumstances had been different. Those German citizens who now, as we see it, failed that moral test, are considered blameworthy, but it was a matter of circumstantial luck that they were in those circumstances, rather than in more peaceful political conditions in which their frailties would not have been exposed and their strengths had been given a proper outlet of expression. As another example of circumstantial luck, Judith Thomson invites us to consider two judges.19 Judge Actual accepts a bribe, thus corrupting the course of justice, whereas Judge Counterfactual, who shares Judge Actual’s dispositions, does not accept a bribe, simply because he was not given the relevant opportunity. The difference between them is simply a matter of luck. Can it be correct, then, to deem Judge Actual more blameworthy than Judge Counterfactual? We need to be precise about the difference between resultant luck and circumstantial luck, because resultant luck can undeniably be described as a species of circumstantial luck. Differences in the actual outcomes of agents’ acts may be truthfully described as being due to circumstances beyond these agents’ control. Reconsider the case of the drunken drivers: the driver who collides with the pedestrian does so because his journey unfolds in circumstances where there is a pedestrian in the wrong place at the wrong time. The luck of the luckier drunken driver, by contrast, can also be described as circumstantial: it consists in the absence of any such pedestrian. 17  We will be spending a great deal of time with this case, as well as the second case about assassins, in the early chapters of this book. 18  Nagel (1979), pp. 26, 28. 19  Thomson (1993), pp. 206–7.

Types of Moral Luck  7 Collapse can be avoided. Although facts about resultant luck can be re-­described in the language of circumstantial luck, the converse does not hold. Nagel was therefore not over-­counting the relevant sources of luck; there is a distinctive hue in this further category of luck which certifies it as a genuinely distinct form of moral luck from resultant luck. On this more precise articulation of it, circumstantial luck should be understood as the luck of being exposed to circumstances in which intentions are formed, acts are performed or at least attempted, and dispositions are developed or actualized.20 ‘Constitutive luck’ is luck in respect of an agent’s character traits, constitution, dispositions, skills, and background. It is luck in respect of an agent’s ‘natural endowments’ and ‘social endowments’, as John Rawls routinely referred to them.21 We can claim no credit for these endowments; we just have them. Even if we are capable of shaping ourselves by willed effort and reflection, we have no control over our starting points, which may themselves constrain or influence the willed effort and reflection we can actually expend, and which are likely at least in part to determine how we act, and where we end up. There is no very sharp distinction between circumstantial luck and constitutive luck. Finally, there is ‘causal luck’, or luck in respect of the causal antecedents of an agent’s acts. Causal luck is a type of circumstantial luck which is concerned with the immediate causal antecedents into our acts and decision-­making. Nagel plainly has in mind the global determinism thesis, according to which everything we do, and everything that happens to us, is subsumable under the laws of nature. We do not get to choose those laws of nature, and we do not get to choose those immediate causal antecedents. To make this plain, think of an arbitrary point in time before you were born. Everything that happens afterwards—every event and action—supposedly conforms to, and falls under, those laws of nature as they proceeded from that pre-­mortem collection of factual circumstances.22 We are blamed and praised for what we do, or omit to do, but what we do or omit to do, on this way of looking at it, is downstream from deterministic currents that we had no control over. Think of how pervasive these categories of luck are in our practical lives. There would seem to be no corner of them that does not betray their influence. How we act in the circumstances we face, the decisions we make, how our acts turn out, the traits and dispositions which are mobilized in the acts we perform, in the

20  Some of these items may appear to belong to the territory mapped out by Nagel’s category of constitutive luck. The boundaries between circumstantial luck and constitutive luck are in fact somewhat porous, as we shall see when Michael Zimmerman’s category of ‘situational luck’ is under scrutiny in Chapter 4. 21  This is common terminology in Rawls (1971). 22 This is a rough description of Peter Van Inwagen’s ‘Consequence Argument’: see Van Inwagen (1975).

8 Introduction circumstances we face: these seem utterly central to our attempts to make sense of our moral lives. But these verdicts also seem riddled with luck. So can we attempt some form of purification of our moral appraisals? Can we purge these appraisals of luck, so that corrected versions of them make due allowance for, and attempt to discount or neutralize, the influence of luck? Given the pervasiveness of luck, Nagel’s worry is that we could not do so. If we made an honest attempt to implement this purging of luck-­influenced content, Nagel fears that ‘the area of genuine agency, and therefore of legitimate moral concern, [would seem] to shrink . . . to an extensionless point’.23 There is a discernible pattern to the concerns at work as we negotiate these various categories of moral luck. Robert Hartman’s observation that ‘the moral luck debate is about not luck per se but a tension in our ordinary thinking about moral responsibility’ seems perfectly correct.24 We lack control over a number of major dimensions which matter to our moral evaluation. That is puzzling, and it calls for investigation.

4.  Moral Luck and Free Will If there is an elephant in the room in discussions of moral luck, it must be free will and moral responsibility. The categories of constitutive luck and causal luck in Nagel’s taxonomy seem particularly germane to free will and moral responsibility. One could be forgiven for assuming that the debate on moral luck, as it has developed in the literature over the last forty years, is a relatively small component in the broader debate about these issues. On this view, moral luck is a minor cog in a larger wheel. Moral luck can only expect to be engulfed in these larger debates upon closer examination of the underlying issues. The debate on moral luck cannot expect to be satisfactorily conducted, far less settled, independently of the events unfolding in the never-­ending free will campaign. Moreover, if the Lack of Control Account is not being disputed as an appropriate characterization of luck in the moral domain, then the relevance of free will and responsibility to these issues will be all the more obvious. This is a serious challenge. But I do not accept that moral luck is merely a junior partner in the broader debate about free and responsibility; the debates are not structured in that way, or at least need not be structured in that way. The

23  Nagel (1979), p. 35. 24  Hartman (2017), p. 24. It should be noted that this particular division in the debate cuts across substantive commitments with respect to moral luck itself. Similar commitments are evinced, for example, by Enoch and Marmor (2007), who are staunchly opposed to moral luck, while Hartman and I accept its existence. They are anti-­luckists, and we are anti-­anti-­luckists.

Moral Luck and Free Will  9 nature of free will and responsibility is, admittedly, a large, adjacent, and sometimes threatening presence in the moral luck debate. Large-­scale treatments of moral luck can be naturally partnered with discussions of free will, and I do not claim that the inclusion of causal luck in Nagel’s taxonomy was poorly judged.25 But the free will debate does not wholly engulf the moral luck debate. There are distinctive concerns in the moral luck debate which can and should be investigated independently of wider issues about free will and moral responsibility. These distinctive concerns are chiefly relevant to resultant luck and circumstantial luck, and to a certain extent constitutive luck, and these three categories of moral luck will indeed form the substance of Part I of this book.26 Consider, first, resultant luck: one assassin hits his target, while the other assassin misses hers. The difference between them, by assumption, is due to luck. Now we assume that both assassins, in making their attempts in the first place, have satisfied the normal conditions of responsibility and control. Their satisfaction of these conditions, whatever they may amount to in detail, is necessary for each of them to qualify as a candidate for moral blameworthiness. Resultant luck now presents us with a further control-­related problem. If each of these agents lacks control over the differences in the outcomes of their respective attempts, then will we not have to agree that they are equally blameworthy? How else can we be in a position to insist on their individual satisfaction of the ordinary conditions of control and responsibility for their moral blameworthiness, but then deny the relevance of a further control-­related difference between them as we determine what each of them is blameworthy for, and the degree to which each of them is blameworthy? That is the essence of the problem of resultant luck. This problem does not ignore free will and responsibility, but assumes that those prior conditions are in place in respect of the normal sources of action for individual agents in order to raise another problem about control and responsibility. If we are going to insist that there are control conditions which must hold in order to make each of these agents morally appraisable in the first place, then we should surely be troubled by our lack of control over a further important dimension of our moral appraisal. The same broad points hold for circumstantial luck. Return to Thomson’s Judge Actual and Judge Counterfactual. Neither of them would have been a candidate for moral blameworthiness in the first place were it not for their satisfaction of the normal conditions for moral responsibility. Whatever these conditions amount to, they are, in the first instance, established separately for each of these agents in turn. But the fact that Judge Actual accepts the bribe, while Judge Counterfactual does not, reflects a difference in circumstances which neither of 25  See, for example, Levy (2011), and Mele (2006). 26  Constitutive luck will also enjoy a second life in Part II, when I turn my attention to distributive justice.

10 Introduction them has any control over. So there would appear to be a further control-­related problem which is not budgeted for within the original terms and conditions for the assignment of individual moral blameworthiness. There will be no room to investigate this further control-­related problem unless we assume business as usual for the satisfaction of the individual blameworthiness conditions. We can consider further cases along this spectrum, thereby edging into constitutive luck. Imagine that there is a third judge, Judge Possible, who does not have corrupt dispositions, but who would have developed them had her circumstances and early social conditioning been different.27 Once again, the fact that she was not placed in these early character-­forming circumstances is just a matter of luck. The differences between Judge Possible, on the one hand, and Judge Actual and Judge Counterfactual, on the other hand, lie beyond the control of any of them. Intuitively, it would appear that both Judge Actual and Judge Counterfactual are blameworthy in ways that do not apply to Judge Possible. But the differences in these intuitive verdicts are due to luck; they thus disclose a further control-­related problem. To recapitulate, there are puzzles about responsibility and control in all these cases that will not even get off the ground unless we assume something like normal service when it comes to the satisfaction of the generic blameworthiness conditions for each individual. If we are convinced that no such conditions for individual blameworthiness exist, then we will not even reach those further puzzles. We can put these points in another way. If you are a pessimist about free will  and responsibility, then it may seem at first that the problem of moral luck is less interesting for you. This is because the question of whether there should be differential blameworthiness between agents whose differences are a matter of luck will be dwarfed by the worry that there can be no justified ascriptions of individual blameworthiness in the first place, due to the absence of free will. But there are two points we can make against even pessimists about free will. First, these arguments between optimists and pessimists are complex and difficult, and we seem entitled to think in conditional terms: supposing the pes­sim­ is­tic arguments for free will were not correct, there would still be these further control-­related puzzles to consider. So what is to stop us from considering them? Second, even for error theorists about moral responsibility, appearances are usually maintained. Pessimists about free will tend to keep ersatz versions of our blaming practices alive, in order to redeem whatever social utility these practices can claim. If so, then an internal problem to negotiate further will be provided by  these categories of moral luck. We will still want to know whether the 27  Thomson does not mention Judge Possible: this addition to the cast of agents was pioneered by Zimmerman (2002).

Moral Luck, Blameworthiness, and Responsibility  11 control-­related problems associated with resultant luck, circumstantial luck, and constitutive luck can be satisfactorily dealt with. So the phenomenon of moral luck still generates a collection of problems which require investigation. Suppose, by contrast, that you are an optimist about free will and responsibility. If you fall into this camp, then you will be committed to the existence of blameworthiness conditions that can be satisfied by individuals one at a time. But that leaves us with these further control-­related problems, concerning the different types of moral luck, which have not been provided for. Again, these issues call for investigation. In short, whether you are an optimist or a pessimist about moral responsibility, there are reasons to support an investigation—a separate investigation—into moral luck. The problem of moral luck is not engulfed by free will and responsibility, even if it is entirely understandable for philosophers who are interested in one of them to take an interest in the other.

5.  Moral Luck, Blameworthiness, and Responsibility Although my policy is to avoid direct confrontations with the free will and responsibility debate in these pages, I will be making repeated references to blameworthiness and praiseworthiness, especially in Part  I.  These notions still import various suppositions about moral responsibility. So it is a good idea at this point to say something more about theories of blameworthiness, even though my aim is to keep critical interventions to a minimum. I will briefly describe this territory before commenting on its connection to moral luck. As for the free will debate, I think that there is a way of maintaining a manageable distance between these debates about blameworthiness and the moral luck debate. There are different possible types of moral appraisability or assessability, which I will use as broad umbrella terms. An important and highly visible central div­ ision in this territory has been advanced by Gary Watson.28 As Watson puts it, there are two faces of responsibility. At one end of the moral appraisability spectrum, we find aretaic or descriptive appraisals, predications, or verdicts. The flavour of this type of appraisal was expressed by John Dewey in the following way: ‘Responsibility is . . . one aspect of the identity of character and conduct. We are responsible for our conduct because that conduct is ourselves objectified in actions’.29 In recent literature, this dimension of responsibility has become known as responsibility as attributability. Other terms have also been used for it: Watson describes

28  See Watson (1996). This is complicated territory, of course, and Watson’s division does not settle all the directions one can take. See, for example, Mason (2019) for other ideas. 29  Dewey (1957), pp. 1601–1; cited at Watson (1996), p. 227; original emphases.

12 Introduction it as ‘responsibility as self-­disclosure’, and it is roughly equivalent to what Susan Wolf calls ‘real self views’ of moral responsibility.30 Responsibility as attributability is ultimately concerned with what the agent, morally speaking, is like. In the attitudes she holds and in the acts she performs, an agent discloses or expresses herself and her values. Her character and her broad orientation towards other moral agents and patients are also revealed in the moral track record she accumulates through her practical career.31 T. M. Scanlon considers someone whose conduct betrays the fact that he ‘does not place any value on other people’s lives or interests’, and asks, rhetorically: ‘what clearer grounds could one have for saying that he is a bad person and behaves wrongly?’32 Qualities of character will be captured by the terms we use to describe those acts and attitudes. These may allow us to describe an agent at different times as cruel, or kind, or cold, or fair-­minded, or generous, or sanctimonious, or vengeful, or forgiving, or lazy, or any number of other things. Her acts and attitudes embody or reflect these dispositions or traits of character.33 And, of course, we can also characterize an agent’s acts in thinner terms, and more flatly, as bad or wrong. As long as her acts and attitudes reflect what this agent is like, then she can be appraised in this aretaic way.34 If the traits are morally undesirable ones, we will, accordingly, end up with different exercises in ‘unwelcome description’.35 Attributability responsibility need not simply take an agent’s acts and attitudes at face value: acts and attitudes can, after all, be coerced, distorted, or ma­nipu­lated.36 Some further tests for screening out manipulation or coercion may have to be satisfied to arrive at a relevantly undistorted view of what the agent is like, morally speaking. Yet even if these scruples are heeded—and nothing prevents more sophisticated conceptions of attributability from taking them into con­ sid­ er­ ation—a number of writers are convinced that responsibility as attributability leaves out something important. Wolf has influenced much of this thinking: . . . when we hold an agent morally responsible for some event, we are doing more than identifying her particularly crucial role in the causal series. . . . [W]e are not

30  See Wolf (1990), esp. ch. 2. Wolf is Watson’s main critical interlocutor in his discussion. As we shall see, Wolf herself rejects real self views. 31  Zimmerman (2002) makes repeated references to the idea of a moral ‘track record’. 32  Scanlon (1998), p. 284. 33  There is a further debate, adjacent to this one, about how stable or enduring character traits are, and to what extent qualities of character should be viewed as largely a product of the circumstances we find ourselves in at the time. See Harman (1999), (2000), and (2009), Athanassoulis (2000), Doris (2002), Kupperman (2001), and Webber (2006) for instructive discussions of these particular issues. 34  See Adams (1985) for a strong defence of this form of responsibility. Scanlon (1998), esp. ch. 6, and Smith (2005) are also relevant. 35  A phrase used by Smith (2012), p. 583. 36  Wolf (1990) understands real self views to be causally structured, so these views are of course exposed to these problems. Attributability views can be less crude than this, however—a point to which we shall return.

Moral Luck, Blameworthiness, and Responsibility  13 merely judging the moral quality of the event with which the individual is so intimately associated; we are judging the moral quality of the individual herself in some focused, non-­instrumental, and seemingly more serious way. We may refer to the latter sense of responsibility as deep responsibility, and we may speak in connection with this of deep praise and blame.37

What is required for ‘deep’ responsibility, in Wolf ’s sense? One of her concerns in this passage is that the appeal to an agent’s ‘role in the causal series’ may not suffice for serious censure. Attributability theorists can agree. As I suggested above, they may wish to refine purely causal criteria in order to arrive at a less distorted view of what can be reasonably attributed to an agent. But the concerns with deep responsibility are likely to go beyond this consideration. The easiest way of proceeding here may be to enumerate some considerations that do not especially engage the concerns of responsibility as attributability. As Neil Levy explains: . . . attributability requires relatively little in the way of control. An act or omission can rightfully be attributed to me whether or not I ever exercised control over acquiring the attitude that it expresses. So long as my action is rightly taken to be expressive of my real self—so long, that is, as it is the product, in the right kind of way, of my beliefs and desires, values and commitments, and not of hypnosis, brain manipulation, mental illness or what have you—then it is properly attributable to me.38

Attributability is not especially concerned with the fair opportunity to have avoided blameworthiness. Perhaps some people really are malicious, spiteful, vindictive, or cruel: this is simply what they are like, and these qualities are faithfully expressed by the way they think and the way they act. Due to early facts about formative conditioning or upbringing, these agents may not have had any choice about the matter, and they may lack control over the expression of these personality traits in their present circumstances. And this fact may serve, in turn, to shield them from certain ascriptions of responsibility. Watson refers to this further type of responsibility as responsibility as accountability. Unless these more demanding conditions for accountability are satisfied, agents are not accountable to others for their attitudes and acts. How does accountability differ from attributability? What interest is served by this more demanding notion of responsibility? Watson’s and Wolf ’s primary

37  Wolf (1990), p. 41; passage cited by Watson (1996), p. 228.

38  Levy (2005), p. 3.

14 Introduction concern lies with agents’ vulnerability to hostile reactive attitudes: attitudes such as indignation and resentment.39 As Watson writes: In one way, to blame (morally) is to attribute something to a (moral) fault in the agent; therefore, to call conduct shoddy is to blame the agent. But judgments of moral blameworthiness are also thought to involve the idea that agents deserve adverse treatment or ‘negative attitudes’ in response to their faulty conduct.40

It is generally unpleasant to be on the receiving end of these attitudes, and other sanctions, formal and informal, may ensue from them. Agents who are exposed to these reactive attitudes may be shunned or cold-­ shouldered by others,41 or acquire a bad reputation which frustrates their future encounters and op­por­tun­ities. These reactions may form, in turn, the basis of further formal or institutional sanctions: fines, penalties, reprimands, and various forms of official censure and setbacks. And the agents on the receiving end of these attitudes may suffer feelings of guilt or shame, or a loss of self-­esteem. An important question arising immediately from these theoretical divisions is this: if we judge that attributability conditions are satisfied, then won’t negative attitudes more or less automatically ensue from them? To take an example, the description of Amy as spiteful cannot be easily combined with a commitment to moral neutrality about this aspect of her character. Spitefulness is not just another ordinary ascriptive characteristic of Amy’s, like her height or hair colour. Unlike her height or hair colour, Amy’s spitefulness is not a morally colourless fact about her. As a result, we cannot be expected to be morally indifferent to the evidence that Amy is spiteful; negative feelings towards Amy seem justified on the basis of this aretaic predication alone. In other words, exercises in aretaic predication appear to be internally connected to what Watson calls ‘reactive entitlements’.42 Our justification for reacting negatively to the evidence that Amy’s acts and attitudes have betrayed spitefulness ensues simply from her spitefulness, especially if—as Watson thinks—attributability amounts to a genuine form of responsibility. This problem threatens to scupper the division between attributability and accountability. If reactive entitlements are licensed by aretaic ascriptions alone, then it is far from clear that responsibility as accountability enjoys any distinctive domain to operate in. It can still be noted, of course, that some acts reflecting those traits were not performed in the presence of a fair opportunity to have avoided performing them, and that fact may make a difference to our responses 39  Strawson (1962) famously makes the reactive attitudes central to our understanding of moral responsibility. His discussion has generated a huge critical industry. See Todd (2016) and Hieronymi (2020) for some penetrating recent reflections on this debate. 40  Watson (1996), pp. 230–1; original emphasis. 41  Scanlon (2008), ch. 4, analyses blame in terms of the relationships-­disrupting implications of wrong action. 42  Watson (1996), p. 230.

Moral Luck, Blameworthiness, and Responsibility  15 to the agents who perform such acts. We can reserve some additional discredit for agents who had a reasonable opportunity to avoid performing blameworthy acts, but performed them nonetheless. Watson’s pluralist stance offers a way of providing for this approach: Moral accountability is only part, and not necessarily the most important part, of our idea of responsibility. The self-­disclosure view describes a core notion of responsibility that is central to ethical life and ethical appraisal.43

On Watson’s view, attributability and accountability both have roles to play in the determination of an agent’s responsibility. There is no need to decide on a victor between them. We can instead appeal to a division of labour as we construct an overall picture of agents’ responsibility: there are simply the ways these agents are (­covered by attributability responsibility), and there are the objectionable things they do which they had the opportunity to avoid (covered by accountability responsibility). Others, however, wish to divide this landscape differently. On Neil Levy’s rival approach, attributability verdicts do not furnish us with genuine ascriptions of responsibility.44 We should instead divide the territory into ascriptions of responsibility as governed by a strong control condition—Levy calls this theory volitionism—and ascriptions of different species of badness, which do not pretend to function as ascriptions of responsibility as well. There are a number of ingredients in Levy’s argument. He thinks, first, that any notion of moral responsibility is defined in terms of vulnerability to reactive attitudes: To say that an agent is morally responsible (for an act, omission or attitude) is to say that the Strawsonian reactive attitudes are justified in relation to her with regard to that act, omission or attitude . . .45

This immediately raises the stakes: if attributability is a genuine form of responsibility, then reactive entitlements are going to be generated. Levy’s next point is that friends of attributability have already budgeted for a  distinction between blameworthiness and badness as a result of the earlier concern that unrefined causal criteria were always too crude for delivering ­ ­attributability in a reliable way:

43  Watson (1996), p. 229. 44  See, again, Levy (2005). Levy (2007a) and (2007b), and Levy and McKenna (2009) also offer rele­vant discussions. 45  Levy (2005), p. 2.

16 Introduction . . . volitionists might point out [that] even attributionists find fault with agents without holding them to be at fault: when these faults are the product of transient mental illness, for instance. It is, therefore, false that there is no conceptual room for a distinction between the bad and the blameworthy, and attributionists owe us an argument for closing the gap between the two.46

Finally, Levy thinks that, if a strong control condition is not in place, blame­ worthi­ness seems inappropriate. Levy provides several arguments to this effect, but I shall concentrate only on one of them, whose flavour is captured in the following passage: If mental illness or brain damage results in transient change of character, Scanlon argues, then we cannot attribute the attitudes it produces to the agent. However, if the alteration is permanent and the agent herself does not reject her new attitudes, then we can attribute them to her. She now has a new character, and is responsible for it.47

Levy finds this implausible: responsibility cannot be assigned to agents simply because the burdens they labour under are long-­lasting rather than transitory. The crucial point, for Levy, is that these agents lacked control over these changes in character, and are not responsible for them. Moreover, these changes in character may render them unable to grasp the justification for the requirements which they have supposedly flouted: If psychopaths do not grasp the moral demand implicit in asking for a justification of an action, they cannot challenge it (except in an externalist sense irrele­ vant to moral responsibility . . .). In general, ignorance of a moral concern excuses someone of responsibility for failing to consider it.48

Such agents can still be described in moral terms: they continue to be cruel, or vindictive, or spiteful. But badness is not a sufficient condition for responsibility, and it does not necessarily warrant blameworthiness. Other writers, such as Angela Smith, also wish to compress the spectrum of possibilities, but in a different place.49 Smith thinks that the notion of answerability ought to take the central place in our thinking about moral responsibility. Agents are responsible if and only if they are in a position to answer the charges

46  Levy (2005), p. 6. 47  Levy (2005), p. 8. 48  Levy (2005), p. 9. 49 See, for example, Smith (2005), (2008), (2012), and (2015). Smith (2012) is a response to Shoemaker (2011).

Moral Luck, Blameworthiness, and Responsibility  17 levied upon them. Answerability is possible, in turn, if what agents are held responsible for admits of judgement-­sensitive reasoning.50 As Smith explains: In order for an agent to be answerable for something, it seems that thing must be connected to her in such a way that it makes sense to ask her to rationally defend or justify it. This, in turn, suggests that the thing in question must in some way reflect her own judgment or assessment of reasons.51

The aspects of an agent’s moral track record that can be fruitfully harnessed with answerability, in turn, are largely coeval with attributability. If agents perform acts or hold attitudes that are morally significant, then there is something these agents can say on behalf of their acts and attitudes. Those acts and attitudes can be justified, or not. This is the key feature for Smith: attributable features are also answerable features, because the moral significance of these features places the bearers of them in a possible justificatory relationship with potential interlocutors. What about the relationship between answerability and accountability? On Smith’s view, there may be variation in the responses that are appropriate to someone who has erred, depending on whether this agent has violated specific moral obligations to others, or disappointed her friend or partner, or merely let herself down in some way. But the basic story will always be the same: she has behaved in ways which at least in principle make her answerable to other agents. For Smith, there are no interestingly deep differences between attributability and accountability. Though my instincts are most sympathetic to Smith’s broad approach, I do not have to endorse her arguments and rebut those of her opponents in order to be able to investigate the relevance of moral luck. This is because the criteria that need to be satisfied for an agent to be eligible for moral blameworthiness in the first place can be detached from the difference made by luck in this agent’s final appraisal. Consider resultant luck. You can be as fussy as you like about the conditions that need to be met for an agent to be appraisable in these cases, which involve malicious attempts or recklessness or negligence. You might favour a less demanding attributability standard, or a more demanding accountability standard, or you may embrace the answerability account. My claim is that, having qualified for blameworthiness in the first place, luck then makes a further difference to what these agents can be blameworthy for. If the more demanding standards set by accountability are insisted upon, then some agents will fail that prior eligibility condition, and will not feature in the subsequent discussion about moral luck. But if these agents do satisfy the prior eligibility tests, then luck will 50  Judgement-­sensitivity is especially emphasized by Scanlon (1998), and Scanlon’s argument is an avowed influence on Smith’s account. 51  Smith (2015), p. 103.

18 Introduction make a difference to their final degree of blameworthiness. Or so I maintain. Luck makes a similar difference to accounts that require less for appraisability, such as attributability accounts. In general, then, we consult our favoured theory of responsibility to certify an agent as eligible for moral blameworthiness in the first place, but—as I see it— luck then makes a further difference to how blameworthy this agent is. Proponents of each of the accounts I have briefly described here—attributability, accountability, answerability—are free either to embrace or to repudiate these anti-­anti-­luckist arguments. Their enthusiasm for these arguments, or their aversion to them, will not be determined by whichever of these accounts of responsibility they embraced in the first place. Once again, the argument about moral luck is not subsumed by these other arguments about the conditions for moral responsibility. The final point I want to make here concerns the connection between blame­ worthi­ness and blame. To be blameworthy is to merit blame, or to be worthy of blame. But blameworthiness and blame do not always go together. If Kristin’s wrongdoing is unwitnessed, there will be no one around to blame her. It does not follow that Kristin has ceased to be blameworthy. Performing acts of wrongdoing in secret is obviously not enough to cleanse Kristin’s acts of wrongness. Her acts can still be appraised as wrong, and as blameworthy, which means that she can be blamed if someone gets to know about her wrongdoing. This is a highly obvious instance of a gap between blameworthiness and blame, explained by an epistemic deficit, but there may also be other ways of driving a possible wedge between them. Some think that there are fairly stringent qualifications for blaming, even when we are dealing with blaming someone who has comfortably satisfied blame­ worthi­ness requirements. In other words, the fittingness of blame can come apart from any particular agent’s standing to blame. More, then, may be required than simply accuracy in the blameworthiness judgement if the agent who attempts to blame is to enjoy the genuine standing to blame. The standing to blame can be lost, or not acquired in the first place, for different reasons.52 We may think that it is not always someone’s business to offer moral criticism. Imagine that Emma has cheated on Charles. This is Charles’s concern, but, intuitively, it may be none of my concern if Emma and Charles are strangers to me. If I have in some way provoked an offence, or acted in such a way as to amount to complicity with the targeted agent’s offence, I may also lack the standing to blame. And, if I am guilty of the same type of offence as that of my critical target—that is, if I am a hypocrite—I may similarly lack the standing to blame. Since the standing to blame may come apart from an agent’s blameworthiness, I will tend to refer in the subsequent discussion to blameworthiness rather than blame. When I do refer to blame, I will typically be assuming that there is no

52  See Smith (2007), Coates and Tognazzini (2013b), Bell (2013), Cohen (2012a) and (2012b), Lippert-­Rasmussen (2018), ch. 4, and Todd (2019) for useful explorations of these arguments.

Moral Luck as a Distributive Problem  19 interesting gap in this case between blameworthiness and blame. My main critical quarry, in any case, is blameworthiness.53

6.  Moral Luck as a Distributive Problem How should we respond to Nagel’s dilemma, where we must settle either for unsatisfactorily impure moral appraisals, or for pure but empty moral appraisals? It seems vital to me that we resist the first horn of this dilemma. The worries which congregate under it are overstated. Moral life is exposed to moral luck, but that is nothing to be afraid of, or to recoil from. It would be morally myopic to attempt to expel these types of luck from our moral appraisals, and—even if that exercise left us with appraisals that were satisfactorily determinate—we would be excluding morally significant information. I will be arguing both that there is moral luck, and also that the sort of luck which matters to moral evaluation is not the alien interloper to moral appraisability which anti-­luckists depict it as being. Moral luck is much more fitting to serve as a faithful guide to what matters in moral appraisal than anti-­luckists think. Many sources of luck to which anti-­ luckists proclaim hostility encode information which any sensible theory of moral appraisability should wish to preserve. To maintain a reasonably unified trajectory of argument, I want to focus the argument on roughly common ground between normative ethics and political philosophy, which is why I restrict my attention largely to the allocation of blame­ worthi­ness in Chapters 1 to 5, and to the role of luck and its conceptual cousin, arbitrariness, in our theorizing about distributive justice in Chapters 6 to 8. By restricting the focus in this way, we can make progress in both of these areas. Or so I will contend. What I have to offer here, in summary, is a fundamentally distributive diagnosis of the problem of luck in moral and political philosophy. That description of my underlying approach will not raise eyebrows in the discussions of luck egalitarianism and the other responsibility-­sensitive accounts of distributive justice which we find in political philosophy. Those debates cannot fail to have a fundamentally distributive character. They are ineluctably relational. But it may require some more commentary in terms of its application to how blameworthiness is allocated in normative ethics. My contention will be that the sharpest point of engagement with the anti-­luckist tendency is to see its concerns as fundamentally distributive, and as being centred on the pairwise comparison: here is something—blameworthiness—which is distributed unequally between pairs of agents, or between an agent in one set of circumstances and the same agent in 53  I have said little here about praiseworthiness, as opposed to blameworthiness, but I re-­engage with it in Chapter 3. Unlike others, I favour an asymmetrical account of blameworthiness and praiseworthiness in moral luck.

20 Introduction different circumstances, despite the relative lack of control from either of them about these facts; and that is morally problematic. It is no coincidence that the cases I have so far offered as evidence for the problem of moral luck, and will continue to explore, involve comparison between one agent and another agent. This is true of the cases involving drunken drivers and assassins; it is true of Thomson’s Judge Actual and Judge Counterfactual; and it is true of citizens who lived under the Third Reich, and citizens who lived under less morally chal­len­ging political regimes. These are squarely cases of pairwise comparison. My contention is that, although the luck which emerges from these pairwise comparisons is a genuine form of luck, it is not a morally significant form of luck. To endow it with any importance is to fall into error. Because it is the pairwise comparison which generates problems of moral luck in the first place, anti-­ luckists suppose that the solution to these problems must also be distributive: we must redistribute blameworthiness in order to tackle the underlying distributive problem. In my view, that is the wrong approach. The solution to the problem is not to rebalance a distributive injustice or irregularity due to the luck that emerges from the pairwise comparison, but to deny the existence of the distributive in­just­ ice in the first place. The very generation of the data which so troubles anti-­ luckists is fundamentally flawed. Pairwise comparisons lead us astray. The main problem with the anti-­luckist rebalancing exercise is that it expels morally essential information. When anti-­luckists compare one agent’s situation with another agent’s situation, and note the presence of luck in the comparison between them, they are forced to make adjustments to the appraisal of each of these agents in order to expel the influence of luck and thereby equalize blame­ worthi­ness. They will be forced to revise our intuitions about the blame­worthi­ ness of each agent in the pair of agents in order to bring them into alignment with each other. Adam Smith calls this anti-­ luckist commitment the ‘equitable maxim’:54 as an instance of this maxim, we should treat the two drunken drivers on a moral par, by deeming them to be equally blameworthy. The comprehensive application of the equitable maxim in an anti-­luckist approach, as Roger Crisp explains, will involve a re-­ordering of what we think matters to moral thought: In cases of resultant luck . . . harmony between the equitable maxim and the views of responsibility implied by our result-­sensitive sentiments is impossible, since they are contradictory.55

On Crisp’s view, if the drunken drivers are equally blameworthy, then responsibility for the fatality resulting from one of their drunken journeys has to be removed

54  See Smith (1759/1976), II.iii.Intro. 1–5.

55  Crisp (2017), p. 11.

Moral Luck as a Distributive Problem  21 from the moral ledger of that driver. This is what the suppression of our result-­ sensitive sentiments will require. The fact that a pedestrian has died, as a result of activity for which the driver lacked any moral justification or excuse, has to be somehow subtracted from the list of items for which we can find the driver blameworthy. Or perhaps—at best—the fact that the driver caused the pedestrian to die is still retained on that list, but it does not elicit any blameworthiness in its own right. This is odd. Such a fatality is a foreseeable consequence of the sort of behaviour that this drunken driver was engaging in; and it is the imagined state of affairs consisting in harm or death to pedestrians as a result of drunken driving which explains why drunken driving is so strongly condemnable in the first place. We condemn drunken driving because of the high risk of death and injury. So if the worst actually happens and those risks turn bad, why shouldn’t we hold the driver responsible for the pedestrian’s death, and why doesn’t the pedestrian’s death help to determine what the driver is blameworthy for, as well as the degree to which the driver is blameworthy? Somehow the equitable maxim has intruded deeply— too deeply—into our grasp of what is relevant and what is irrelevant to blameworthiness. Because the pairwise comparison between the two drivers seems relevant, we feel forced to reassess the moral relevance of what each of these drivers does, even if we might initially have thought it possible to achieve a purchase on the moral significance of their actions in the absence of that pairwise comparison. We should guard against this anti-­luckist line of thought. The apparatus of the pairwise comparison encourages us to redact morally vital information about the agents being compared to each other in those pairwise comparisons. I will argue that pairwise comparisons provide a faulty and misleading guide to what matters in moral evaluation. This contention will be pursued across the various categories of moral luck. Chapters 1 to 3 will be concerned with aspects of resultant luck. The appeal of what I call the ‘Anti-­Luck Account’ will be traced to the influence of two central intuitions: the ‘Irrelevance Intuition’ and the ‘Fairness Intuition’. The Irrelevance Intuition will be treated in detail in Chapter  2, while the Fairness Intuition will be examined in Chapter 3. When these two intuitions are examined more closely, their appeal wears off, and my favoured anti-­anti-­luckist account—the ‘Restricted Luck-­Sensitive Account’, or the Restricted Account, for short—will become more appealing. At least that is the hope. The trail continues in Chapter 4, where I consider the expansion of anti-­luckist tendencies into the area of ­circumstantial luck and ‘situational luck’, which is a category of moral luck combining circumstantial luck and constitutive luck. My responses to these further categories of luck take their cue in part from the position I have worked out for resultant luck. (Appendix II extends certain aspects of the arguments I explore in Chapter 4.)

22 Introduction Chapter 5 completes Part I of the book by examining Williams’s treatment of moral luck. My interpretation of Williams makes his fundamental interests rather different from Nagel’s, though there are certainly points of affinity between them, and despite the fact that they have some common enemies. For the most part, Williams uses the existence of luck as a way of exploring the grip that morality has over us; certain types of luck, for him, can serve to remove someone from the force-­field of moral assessment on one particular but influential understanding of what morality is. I explain in Chapter 5 why I do not think Williams’s argument succeeds, and why his insights do not impugn the Restricted Account. The management of constitutive luck is further explored in contemporary debates about distributive justice, and these issues will be the focus of Part II. In Chapter 6, I address luck egalitarianism, an influential family of theories in the distributive literature which is committed to neutralizing the influence of relative luck between agents. As in cases of moral luck, the usual focus for these debates is the pairwise comparison. In one sense, this interpersonal focus should not be surprising when it is encountered in the domain of distributive justice. After all, justice is an essentially relational political value; it cannot fail to involve comparisons between different agents. But in another sense, the emphasis on the neu­tral­ iza­tion of luck in luck egalitarianism is destined to struggle. As we shall see, the very existence of that interpersonal focus in pairwise comparisons makes it difficult to see how luck egalitarianism can deliver an adequate account of individual luck which can properly regulate distributive relations. Either everything affecting an individual’s distributive holdings is down to luck, regardless of what principles of distribution we employ—which is close to self-­defeating—or we have to distinguish between different types of luck in order to tell a sensible distributive story. If we are forced to distinguish between different types of luck in order to arrive at sensible distributive criteria, we are probably tracking factors other than luck. And that is what I think we are doing. How does the role of pairwise comparisons in anti-­luckist normative ethics compare to the role of pairwise comparisons in anti-­luckist theories of justice? In normative ethics, when we are considering the distribution of blameworthiness, anti-­luckists think that a disruptive influence on the ascription of responsibility at the individual level—the involvement of luck—must have a similarly disruptive application to the ascription of responsibility at an interpersonal level. This conviction is then used to press Smith’s equitable maxim into service. The antidote in normative ethics for this sort of anti-­luckist theorizing is to accept that we should never have ascended to that interpersonal level of comparison in the first place. Now consider the situation in political philosophy. It would be understandable to hold that interpersonal comparisons are unavoidable in theorizing about just­ ice. That truth, in turn, might encourage anti-­luckists to employ the very same theoretical apparatus in their theorizing about justice as that employed in normative ethics: the pairwise comparison. Perhaps the fact that it would be perfectly

Moral Luck as a Distributive Problem  23 natural to reach for pairwise comparisons in normative ethics if one were convinced that interpersonal comparisons are a legitimate concern also fuels the conviction that pairwise comparisons must be central to the interpersonal comparisons pursued in political philosophy. But in this particular respect, political philosophy is in danger of being misled by normative ethics. What is unavoidable in theorizing about justice is the focus on a common standard or baseline that can permit interpersonal comparisons. But pairwise comparisons need not be the focus of interpersonal comparisons—and they should not be, on pain of incoherence or an incomplete attention to what matters in justice. I pursue this argument in detail in Chapter 6. Someone who is frequently cited as a progenitor for luck egalitarianism is John Rawls, whose theory of ‘justice as fairness’, as outlined in his A Theory of Justice, displays hostility towards the influence of factors that Rawls frequently describes as being ‘arbitrary from a moral point of view’.56 For many commentators, moral arbitrariness is effectively synonymous with luck. On these views, when Rawls sets out to neutralize morally arbitrary influences, he should be understood to be neutralizing differential luck. This may be the standard interpretation of Rawls, but I shall suggest in Chapter 7 that there is another way of interpreting his main arguments, which supports a different understanding of the role and significance of moral arbitrariness. Since this interpretation makes better sense of why Rawls employs his contractarian apparatus, and is also strongly anchored in at least some of his own texts, I think it is preferable to the standard luck-­neutralizing interpretation. It is also largely free of the problems that infect other responsibility-­ sensitive theories of justice. The final chapter, Chapter  8, also uses broadly Rawlsian commitments to explore a trio of problems associated with arbitrary moral boundaries. These problems are connected to luck in a slightly different sense, though they are significantly related to the other issues I explore in this book. The problems are all connected to the grounds of discrimination in our treatment of individuals, both human and non-­human. Some of the grounds on which we habitually rely in our ordinary moralizing in these areas seem due to luck. Those whose treatment is governed by these standards might be reasonably thought to be the beneficiaries of good luck, while those who fall outside the scope of these theories might be reasonably suspected to be the victims of bad luck. Is that problematic? One of these problems is international justice. Cosmopolitans and non-­ cosmopolitans disagree about the distributive relevance of state boundaries. For cosmopolitans, these boundaries are morally shallow, and cannot justify the stark differences in life chances for individuals who happen, through no fault of their own, to be born in different places. For non-­cosmopolitans, these boundaries

56  Rawls (1971), Part One.

24 Introduction may be morally relevant, given differences in the normative environment facing insiders and outsiders to any given community or scheme of justice. Although there is progress to be made on these non-­cosmopolitan fronts, it can seem awkward for non-­cosmopolitans to affirm the relevance of factors that cannot be anything other than morally arbitrary, or the product of brute luck, for each individual affected by them, if these theorists also endorse a luck-­sensitive theory of justice for those subjects of justice who qualify for membership in a particular scheme of justice. These non-­cosmopolitans even risk the charge of incoherence: how can they be committed to a theory whose central prescriptions are committed to the neutralization of arbitrary disadvantage, while accepting the relevance of boundaries to a scheme of justice that are themselves fixed in an arbitrary way? I do not offer a defence of non-­cosmopolitanism in this book. But I do think that this central charge of incoherence can be dispersed, if our interpretation of moral arbitrariness conforms to the one advanced in Chapter  7. That is an important conclusion if we are to succeed in identifying what keeps cosmopolitans and non-­ cosmopolitans apart in these debates. A second problem of arbitrary boundaries concerns human moral equality. Human beings come in all sorts of shapes and sizes. They seem morally heterogeneous as well, whether by appeal to their moral track records, or their characters and dispositions, or their capacities for well-­being. Rawls treats all normal human adult persons as potential subjects of justice, if they have the wherewithal for what he calls moral personality: the capacity to be guided by a sense of justice, and to form and pursue a conception of the good. This assumption of basic moral equality among human subjects also risks arbitrariness, if there are relevant differences among these subjects of justice. If some individuals have a greater ability than others to be guided by justice and to realize a conception of the good, then why shouldn’t they enjoy a higher status? A relaxed attitude to differences above the threshold for those individuals who meet the moral personality standard also appears to combine uneasily with the dismissal of those individuals—most of them being non-­human animals—who are deemed not to qualify for justice. These questions create pressure to seek a more rigorous account of human moral equality that is anchored in a single characteristic, or else collection of characteristics, that is equally distributed in an all-­or-­nothing way. One problem with this search for ‘basic equality’ is that the candidates that could successfully realize it seem to be in vanishingly small supply. I argue that the basic equality project is wrong-­headed, and that Rawls’s relaxed view of the matter is precisely the one to follow. Yet another hotly contested site of dispute—and the third problem considered in this chapter—concerns interspecies relations. In the philosophical literature, the charge of ‘speciesism’ has acquired a reputation which rivals obviously invidious forms of discrimination such as racism and sexism. The problem with the

Moral Luck as a Distributive Problem  25 species barrier is that it seems arbitrary, particularly when we admit that some ‘marginal’ human cases instantiate lives of little mental complexity or accomplishment. Again, I argue that the arbitrariness of the species boundary is not a barrier to policies and principles which reserve a moral status for human beings that is not on offer to non-­humans. This is because many of the anti-­speciesism arguments are committed to an ‘individualist’ approach to moral status: an approach which ignores relational and extrinsic properties of individuals and just appeals to the properties that are in some recognizable sense intrinsic to them. The challenges presented by these cases of arbitrary boundaries can, I believe, be met. Arbitrary boundaries do not present an insuperable problem for our moral practice, even if they are based on certain forms of luck. My conclusions here add to the corpus of arguments advanced elsewhere in this book. Their joint upshot is that luck is not the interloper or alien moral presence it is often reputed to be in our moral and political lives. Luck is ineliminable from our lives and practices, but in many cases that does not matter. And that, as it happens, deserves to count as a stroke of luck for us.

1

What Results from Resultant Luck? 1.1.  Two Cases The debate on moral luck has made us familiar with cases with the following type of interpersonal structure:1 Drivers: In World 1, Fred drives home drunk, and runs over and kills a pedestrian whom he would not have run over had he been sober. In World 2, Thomas drives home drunk, encounters only empty roads, and arrives safely. Assassins: In World 1, Imogen, a hired assassin, takes aim and shoots the President, killing him. In World 2, Phoebe, a hired assassin, takes aim and shoots, but inadvertently misses the President, who unexpectedly and unforeseeably moves just as she pulls the trigger.2

We will be spending a great deal of time with these cases. They exemplify what Thomas Nagel calls ‘resultant luck’, or ‘outcome luck’: luck in respect of the results or outcomes of actions.3 These cases also raise slightly different issues, as we shall see. In both of them, however, the difference between what befalls these agents seems, in part, to be a matter of luck. In Drivers, the fact that no pedestrian appeared in front of Thomas on his drunken journey home is beyond his control; a mere matter of luck. Similarly, in Assassins, the fact that the President unexpectedly moves, thus rendering her attempt unsuccessful, is also beyond Phoebe’s control; again, it is a mere matter of luck. So isn’t Thomas as blameworthy as Fred, and isn’t Phoebe as blameworthy as Imogen? I think the answer to these questions, in both cases, is ‘No’. Certainly, there is a sense in which luck is at play in both cases, as we move from World 1 to World 2, but we should not recoil from the existence of such luck, or dispute its relevance for our moral evaluation of these agents and what they have done. This kind of luck will not successfully sustain a complaint from Fred about his being blameworthy to a greater degree than Thomas, or a complaint from Imogen about her being blameworthy to a greater degree than Phoebe. As I see it, morality in general, and our assignments of blameworthiness in particular, can and must 1  See Nagel (1979). 2  Assume that the President is by no means a legitimate target of political assassination. 3  Zimmerman (1987) appears to have coined the term ‘resultant luck’.

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0002

30  What Results from Resultant Luck? provide for the existence and moral significance of such luck. To deliver this conclusion, I will provide a much more detailed analysis of the appeal of the anti-­ luckist tendency, and hope by so doing to discredit it. The basic purpose of this chapter is to set up the various pieces for that debate. The argument will be structured as follows. In the scene-­setting Section 1.2, I set out three distinct approaches to moral luck: the ‘Strict Liability Account’, the ‘Anti-­ Luck Account’, and the ‘Restricted Luck-­ Sensitive Account’. (As I will explain, the Strict Liability Account seems to me a non-­starter, but its inclusion permits helpful structural contrasts with the other two accounts.) In Section 1.3, I re-­describe the Anti-­Luck Account as the product of two requirements which reflect the influence of two intuitions, the ‘Irrelevance Intuition’ and ‘Fairness Intuition’. (These two intuitions are explored in detail in Chapters 2 and 3, respectively.) Section  1.4 enumerates, in more fine-­grained detail, the various types of luck which are manifested in Drivers and Assassins, and assesses their potential relevance to the concerns of the Anti-­ Luck Account. Section  1.5 examines various attempts by anti-­ luckists to accommodate the ostensible relevance, in some cases, of resultant luck. I shall suggest that these attempts are unsuccessful. If relevant luck is ostensibly relevant, it cannot be domesticated as easily as this collection of proposals pretends. A caveat before I go any further: I am not going to pay any special attention here to the status of resultant luck and omissions, or ‘negative’ responsibility.4 My aim is to get clearer on the central cases represented by Drivers and Assassins, without having to worry about these further complexities. Even these central and familiar cases, as we shall see, provide us with many complications to untangle.

1.2.  Three Rival Approaches I shall open the argumentative bidding by distinguishing between three rival approaches to moral luck, indicating prima facie strengths and weaknesses in each of them. (I do not claim that these approaches exhaust the relevant options.)

1.2.1.  The Strict Liability Account According to the Strict Liability Account, moral blameworthiness is settled by whatever actual harmful consequences ensue from one’s acts. As applied to our two cases, the Strict Liability Account will therefore find Fred and Imogen to be

4  But see Raz (2010) and Sartorio (2012), esp. pp. 170–7, for instructive discussions.

Three Rival Approaches  31 much more blameworthy than their respective counterparts. Since Thomas does not, in fact, cause any injury or harm, he will not be held blameworthy. The same goes for Phoebe, who does not succeed in assassinating the President. The Strict Liability Account may pay due attention to the significance of outcomes to our appraisals of agents, but it will nonetheless strike most people as hopeless in other respects, at least outside very tightly specified domains: the verdicts it licenses would allow good people to be unjustly held blameworthy, and bad people to unjustly escape blameworthiness. To explain further, consider four more specific problems with it. First, the fact that Thomas and Phoebe are not blameworthy at all will strike many as unsatisfactory. Second, there are other cases in which the Strict Liability Account will hold that intuitively innocent and non-­blameworthy agents are blameworthy after all. Consider: Sober Driver: Noah, who never drinks and drives, drives home with due care and attention and accidentally, through no fault of his own, runs over and kills a pedestrian.5

Noah is not guilty of recklessness, yet the Strict Liability Account will find him just as blameworthy as Fred. This result seems unacceptable. Third, and leaving aside Sober Driver, the fact that Fred is much more blameworthy than Thomas, and that Imogen is much more blameworthy than Phoebe, will also be resisted by those with strong anti-­luckist tendencies. Fourth, and more generally, the Strict Liability risks distancing moral assessment from practical action-­guidance to a frankly uncomfortable extent. If the lion’s share of moral evaluation turns on what actually happens in the world, and if the connections between what agents aim to do and what happens in the world as a result of what they do are systematically skewed by luck, then, to borrow a phrase from Christopher Kutz, ‘moral theory looks more like the weather forecast than the Ten Commandments’.6 That distance has to be reduced; moral theory must have something to do with the evaluation of how agents decide to act. Normative moral theory is primarily prescriptive, not predictive.7

5  The notion of ‘agent-­regret’ explored at length by Williams (1981a) might be thought to put pressure on Noah’s sense of non-­involvement with the cost of his action. See Chapter 5 for a fuller discussion, and especially Section 5.7. But briefly: the notion of agent-­regret does not actually imply that Noah should be deemed blameworthy, even if we, along with Williams, would expect morally sensitive agents to find it difficult to achieve an emotional detachment from what they have caused to happen. 6  Kutz (2009), p. 244. 7  Not always, or exclusively: see, for example, Boyd (1988) and Railton (1986). But the region of ethics I am exploring is primarily prescriptive, rather than predictive.

32  What Results from Resultant Luck?

1.2.2.  The Anti-­Luck Account As a stark corrective to the Strict Liability Account, the Anti-­Luck Account denies that luck in harmful consequences can make any difference to the degree of one’s fundamental blameworthiness. On the Anti-­Luck Account, Fred will be no more blameworthy than Thomas, and Imogen will be no more blameworthy than Phoebe. And, since he acted without fault and was therefore entirely unlucky, Noah will not be blameworthy at all. Moreover, the Anti-­Luck Account guarantees that action-­guidance and moral evaluation are stationed together. The Anti-­Luck Account thus appears to take immediate care of the four disadvantages of the Strict Liability Account. This is an impressive sweep. I will be taking a more detailed look at the structure and motivation of the Anti-­Luck Account below, but even as it stands, and despite its superiority to the Strict Liability Account, it is likely to leave some dissatisfied. Many will applaud the Anti-­Luck Account for sparing Noah from blame, and for ensuring that Thomas and Phoebe are blameworthy to some extent, but they may still be convinced that Fred is more blameworthy than Thomas, and that Imogen is more blameworthy than Phoebe. After all, Fred’s reckless behaviour has caused the death of a child, whereas Thomas is guilty of no such offence; and Imogen’s act has succeeded in killing someone who was not liable to be killed, whereas Phoebe’s act has not. Acts have worldly consequences, and we may be inclined to accept that we are often blameworthy for those consequences. These thoughts can be pursued from a different direction. Is ethical assessment exclusively concerned with the realm of an internal realm, free from all worldly luck? Does the Anti-­Luck Account omit too much? Think about it from the first-­ person perspective. If you were unwise enough to get into the car when you were drunk, would that decision strike you as being the end of the moral story? Wouldn’t it morally matter to you that you did not run anyone over when you were careering home? Imagine your morning-­after regrets: ‘That was really stupid of me. I could have killed someone!’ You would be not just expressing contrition at the fact you had behaved recklessly, but also expressing relief at the fact that you had not, as a result of acting recklessly, caused someone to die. Your relief at the fact that you did not kill someone appears to carry the implication that things would have been worse for you—morally worse, not just prudentially worse—had you done so. Had you killed someone, your moral track record would be worse than it would otherwise have been, and that would have been your fault. To put it another way, the fault in your action—acting recklessly—would have been manifested in a more egregious way. Not everyone will accept this intuition, and the argument to follow does not assume that you were convinced by it. But I, for one, find it compelling. My aim is to find a way of giving it a more full-­bodied expression.

Three Rival Approaches  33

1.2.3.  The Restricted Luck-­Sensitive Account These thoughts, preliminary and under-­defended though they may be, help to motivate the generation of an intermediate position between the Strict Liability Account and the Anti-­Luck Account. This brings us, accordingly, to our third approach. The Restricted Luck-­Sensitive Account, or the Restricted Account for short, consists of two claims. According to the first claim, an eligibility condition for blameworthiness is that agents’ acts have to reflect morally objectionable mental (or internal) states. Moreover, the moral objectionableness of these mental states must be fixed independently of the outcomes of these acts, on pain of an immediate collapse back into the Strict Liability Account. The mental states which will thus qualify are primarily malicious intentions, and internal states which betray recklessness or negligence. We can call this claim the Internal Claim, since it takes an interest in the moral quality of mental states that are internal to the agent. I shall allow myself to use the term ‘culpability’ and its cognates to gloss the internal states which satisfy the Internal Claim: in this sense, Imogen, Phoebe, Fred, and Thomas are all culpable, while Noah is non-­culpable. Imogen and Phoebe are malicious, while Fred and Thomas are reckless. These are simply different ways of being culpable, according to this sense of culpability. The second leading claim of the Restricted Account is that, having satisfied the Internal Claim, agents will then be blameworthy in rough proportion8 to the badness of the actual harmful luck-­dependent consequences which proceed from these acts. Call this the External Claim. The Restricted Account will not hold Noah in Sober Driver to be blameworthy, as his act does not satisfy the Internal Claim. The External Claim is not activated if the Internal Claim has not been satisfied. By contrast, the acts performed by Fred and Thomas, and Imogen and Phoebe, do satisfy the Internal Claim, since these agents acted either recklessly or maliciously. Having satisfied the Internal Claim, they are then eligible for evaluation by the External Claim; and, according to the External Claim, Fred will be more blameworthy than Thomas, and Imogen will be more blameworthy than Phoebe. This is because Fred’s act carried actual harmful consequences which render him more blameworthy than Thomas, and Imogen’s act carried actual harmful consequences which render her more blameworthy than Phoebe. The Restricted Account may appear, at least at first, to offer a successful correction of excesses and faults which can be reasonably suspected to mar the Strict Liability Account and the Anti-­Luck Account. There are four points to make. 8  The qualifications which explain the presence of the word ‘roughly’ in the External Claim will be advanced in Chapter 2, Section 2.5.

34  What Results from Resultant Luck? First, the Restricted Account permits Noah to be excluded from the scope of blameworthiness. Second, it holds that Thomas and Phoebe should be included within the scope of blameworthiness. Third, it deems Fred to be more blameworthy than Thomas, and Imogen to be more blameworthy than Phoebe. Fourth, and more generally, although the Restricted Account accepts that worldly consequences can affect the moral evaluation of agents, it does not turn moral theory into the sort of weather forecasting that Kutz was worried about. Agents get to be morally evaluated only given the presence of certain types of internal states, over which they possess a normal degree of control, in the sources of their actions. But worries remain: the Restricted Account can easily appear to be a botched compromise, or an unstable half-­ way house. These worries can be approached from two sides. Consider, first, the Internal Claim. If the Restricted Account is correct to provide for this condition at all, then it is answerable to the complaint that a more thorough treatment of it is available: one which makes blameworthiness precisely proportionate to the degree to which agents initially satisfied the Internal Claim. Since it is already clear that the difference in the worldly outcomes between Fred’s and Thomas’s car journeys was a matter of luck, rather than anything that either of them had control over, then a stricter form of moral book-­keeping will insist that Fred is held to be no more blameworthy than Thomas. We still need an explanation of how any difference in the final degree of blameworthiness can exist between agents whose mental states at the time of acting were, or seemed to be, equally objectionable. We can make much the same points from the other direction, by fastening on actual consequences. This brings us to the External Claim. If it is the actual harmful consequences that matter to blameworthiness, as the External Claim proclaims, then why does the Restricted Account’s refusal to embrace the uncompromising picture of the Strict Liability Account amount to anything other than a failure of nerve? The Restricted Account attempts to pay tribute to both actual consequences and objectionable motivation in an overall moral evaluation of these agents, but these forces threaten to push in opposite directions. Acts embodying culpable motivations are not perfectly matched by how things actually turn out in the world. Even if a malicious intention is successfully manifested in the world, that outcome will also reflect luck. To be successfully manifested in the world, it must be the case that a malicious intention has not been obstructed, and the fact that no such obstruction takes place is plainly a matter of luck. There is, according to this sceptical line of thought, no principled case for setting up camp on the fault-­line which lies on the interstices of these respective areas. That, at least, is how the Restricted Account may strike you when it is first encountered in this dialectic. Despite these initial reservations, it seems to me to

Three Rival Approaches  35 be not just conceptually coherent, but morally defensible. To make further progress, however, we need to undertake a more detailed examination of the Anti-­Luck Account. The faults we will uncover in it help us to see how there might be satisfactory accommodation of both culpable internal states and worldly outcomes; how one and the same account can provide satisfactorily for the Internal Claim and the External Claim as well. There is one more challenge I want to flag at this stage. Some may worry that this preparatory reasoning for the Restricted Account has been obtuse, because it neglects the force of the distinction between the scope of blameworthiness and the degree to which someone is blameworthy. This distinction, which has been forcefully presented by Michael Zimmerman,9 will allow us to accept, in Drivers, that Fred is responsible for a greater number of things than Thomas is, since Fred’s driving was not just reckless, but deadly: a pedestrian died. Similarly, in Assassins, Imogen is responsible for a greater number of things than Phoebe is, since she has not just a malicious attempt, but also a murder, to answer for, whereas Phoebe is answerable for just a malicious attempt. The responsibility of Fred and Imogen enjoys, then, greater scope than that of Thomas and Phoebe, respectively. But the degree to which they are blameworthy might still be the same: Fred might still be no more blameworthy than Thomas, and Imogen might still be no more blameworthy than Phoebe. I am not convinced of the uses to which anti-­luckists think they can put this distinction. My argument against the significance of Zimmerman’s distinction will be given later in this book.10 For the time being, I am trading on basic intuitions which are not immediately scuppered by the distinction between the scope and the degree of responsibility. The Restricted Account holds that Imogen is responsible for a murder as well as a malicious attempt, and that, therefore, she is more blameworthy than Phoebe, who is responsible only for a malicious attempt. Similar remarks can be made about Fred and Thomas. In each case, it is assumed that degrees of responsibility rise along with expansions in the scope of responsibility. We need arguments, of course, to vindicate such intuitions—to show that enlargements in the scope of responsibility are partnered by enlargements in the degree of responsibility. But these are nonetheless the basic intuitions the Restricted Account rests upon. It is not inappropriate to appeal to them in order to make some divisions among the theoretical options available to us, even if they can enjoy only limited authority at this stage.

9  See Zimmerman (2002). 10  It might be mentioned even at this stage, however, that Zimmerman’s distinction is not always accepted, at least in the form in which he presents it, even by those who reject moral luck either in whole or in part: see, for example, Khoury (2018), pp. 1361–3, and Peels (2015).

36  What Results from Resultant Luck?

1.3.  The Structure of the Anti-­Luck Account 1.3.1.  Two Principles As others in these debates have noted, the Anti-­Luck Account consists of two main theses: I call them the Control Requirement and the Parity Requirement.11 They can be defined as follows: Control Requirement: We are morally assessable only to the extent that what we are morally assessed for depends on factors under our control. Parity Requirement: Two agents ought not to be morally assessed differently if the only differences between them are due to factors beyond their control.

The Control Requirement will already seem familiar: the Control Requirement states the central reason why the Anti-­Luck Account recoils from the presence of luck in moral appraisals. I turn to the Parity Requirement shortly, but it is the Control Requirement which remains the more important normative building block in the Anti-­Luck Account, so I will start with that.

1.3.2.  The Control Requirement and the Irrelevance Intuition The Anti-­Luck Account’s insistence on the Control Requirement invites more commentary. What explains the hostility to appraisals which stray beyond agents’ control? The Control Requirement appears to rest on two ideas working in combination. The first of them will be familiar: it is the principle of ‘Ought Implies Can’, according to which, if you ought to T, then you must be capable of acting so as to T. If agents ought to act in non-­blameworthy ways, then they must be capable of avoiding blameworthiness. But, if blameworthiness is sensitive to luck, then we have an ‘ought’ which is not partnered, in the relevant sense, by a ‘can’. Thus the idea that there is any such luck-­sensitive ‘ought’ must be chimerical. The second idea is that our moral appraisals are ultimately concerned with the quality of will displayed by agents, because it is our quality of will which reveals us most fully in our role of moral agents.12 Though there need not be any 11  For this helpful distinction, see Nelkin (2008). The name of the second requirement is mine. 12  Compare Andre (1983), pp. 204–5; progressively broader discussions are provided by Arpaly (2002), Scanlon (1988), and Strawson (1962). Clearly, it can be difficult to interpret the Control Requirement in a way that avoids total immersion in the free will and responsibility debate. See the Introduction, Section 4, for more on this. As I explain there, it is necessary to settle for a placeholder-­ type characterization of control precisely in order to prevent the moral luck issue from being engulfed by that larger debate.

The Structure of the Anti-­Luck Account  37 antecedent denial that we can appraise agents for the consequences of what they do, the fact remains that when we permit our moral appraisals of agents to be swayed by factors operating beyond agents’ control, including those which concern the actual consequences of our acts, we dilute the quality of the evidence shedding light on those aspects of action which most fully disclose what agents are like, morally speaking. As a result, luck-­ sensitive appraisals involve an investment in information which is, variously, uninteresting, unrevealing, or distracting. Such factors cannot disclose anything of moral significance, and merely muddy the evidential waters. They cannot tell us anything about the things that properly occupy our attention when we appraise agents; they do not tell us anything about the agent’s character, values, motivations, and dispositions, which are the principal matters on which we properly fasten when we make moral judgements about other agents. I shall refer to this line of thought as the Irrelevance Intuition.

1.3.3.  The Parity Requirement and the Fairness Intuition What about the Parity Requirement? It is often treated as nothing more than a straightforward corollary of the Control Requirement.13 To return to Assassins, we suppose that Imogen and Phoebe have entirely similar aims, dispositions, and motivations, and that both of them attempt to kill the President. For reasons beyond the control of either of them, Imogen succeeds in killing him, whereas Phoebe fails to kill him. Since this difference in outcomes between their respective attempts to kill him is not due to anything they have any control over, the Parity Requirement will insist that they collect the same moral appraisal. But we could of course have arrived at that verdict, for each of these agents, by tracing out the application of the Control Requirement alone. The Parity Requirement does not seem to add anything to what the Control Requirement entails when this latter requirement is applied at an interpersonal level. There is, however, a larger significance to the Parity Requirement, for there is another important intuition nourishing the Anti-­Luck Account which applies principally to the Parity Requirement, rather than to the Control Requirement. If this intuition is in good order, it confirms that the Parity Requirement may state more than a simple corollary of the Control Requirement. At least, the possibility that the Parity Requirement states more than just a simple corollary of the Control Requirement cannot be dismissed out of hand. The intuition is that it is unfair to blame agents for outcomes which are owed, in part, to factors operating beyond

13 See, for example, Nelkin (2008), Enoch and Marmor (2007), p. 417, and Zimmerman (2006), p. 590.

38  What Results from Resultant Luck? these agents’ control. Call this the Fairness Intuition. There are two points that I wish to make about the status of the Fairness Intuition.14 The first and most modest point is that the Fairness Intuition can claim to boast some independent critical life from the Irrelevance Intuition, even if, as it turns out, its verdicts are necessarily co-­extensive with those of the Irrelevance Intuition. This is because the Fairness Intuition pays explicit attention to the mere difference in moral verdicts on different agents. The mere appearance of the interpersonal dimension is sufficient to engage the distinctive concerns of fairness. Taking Assassins as our example, the fact that Imogen is more blameworthy than Phoebe may invite Imogen’s complaint that she is treated unfairly by comparison with Phoebe. The fact that Imogen killed the President, whereas Phoebe failed to kill him, is due, once more, to factors that lie beyond the control of either of them. It is therefore unfair, Imogen might argue, that Phoebe is less blameworthy than she is. This point concedes that the grounds of Imogen’s complaint are ultimately rooted in a worry about the limits of control, applying to each of them sep­ar­ ate­ly in the first instance, and that this worry can be explained by the application of the Control Requirement to each of them in turn. However, the mere fact that we are dealing with an interpersonal case, rather than an intrapersonal case, is then enough to sustain a further complaint about the value of fairness in particular. To put the point another way, the same facts may sustain concerns from both the Irrelevance Intuition and the Fairness Intuition. The worry about fairness may track the worry about a lack of control, and fail to apply in the absence of a worry about a lack of control, but it does not follow that these cannot be represented as two distinct worries. It should not be surprising that a single solution—the neutralization of luck in our ­

14  Hanna (2014), pp. 690–2, provides a powerful attack on the putative entailment between the Control Requirement and the Parity Requirement. His particular focus is on the disruptive influence of circumstantial luck. As Hanna points out, the claim that agents are assessable only for factors under their control is actually consistent with the claim that they can be assessed differently, since, as he puts it, ‘[w]e’re assessable for what we do with the control we have, not for what we would do if things were different’ (Hanna 2014, p. 691). What the Parity Requirement probably takes exception to, as Hanna suggests, is the different thought—a thought not captured by the Control Requirement as officially stated—that ‘it’s only because of factors beyond [these agents’] control that they’re not assessable for the same things’ (2014, p. 691). I agree with Hanna that circumstantial luck disrupts the entailment between the Control Requirement and the Parity Requirement. As a result, this may permit the Fairness Intuition to make some independent substantive contribution to the Anti-­Luck Account. Nonetheless, the entailment between the Control Requirement and the Parity Requirement does seem more secure when we are focusing on resultant luck alone. This is because the distinction between what Fred is assessable for with the control he has and what Thomas would be assessable for had factors beyond his control behaved differently is harder to pin down. The fact that this is a case of resultant luck implies that Thomas has already performed an act, and that the worldly consequences of it are not fully under his control. That puts him in the same situation as Fred. In Drivers and in other cases of purely resultant luck such as Assassins, there is no obvious disruption of the entailment between the Control Requirement and the Parity Requirement.

The Structure of the Anti-­Luck Account  39 blameworthiness assignments, as commended by the Anti-­Luck Account—is capable of dispersing more than one problem. The second and stronger point enrols Joel Feinberg’s distinction between non-­ comparative justice and comparative justice.15 Cases of comparative justice are essentially relational: the appropriate treatment of an agent’s comparative claim depends on how its treatment compares to the treatment of other agents’ comparative claims. Cases of non-­comparative justice, by contrast, take the form of giving agents their due, where what they are due—the content of their non-­ comparative claims—is determined independently of what other agents are due. As a corollary of the Control Requirement, the Parity Requirement is concerned with agents’ non-­comparative claims. But as a sponsor for the Fairness Intuition, the Parity Requirement may be getting at another idea, which is that agents have comparative claims not to be blamed more harshly than other agents, if the difference between the levels of blameworthiness assigned to them is due to factors beyond their control. So, whatever degree of blameworthiness is assigned to Fred, the Parity Requirement entails that it will be unfair if Thomas is not blameworthy to an equal degree. If Fred is assigned a very high degree of blameworthiness, there will be unfairness if Thomas is any less blameworthy than Fred. Similarly, if Fred is assigned a more moderate degree of blameworthiness, there will be unfairness if Thomas is any less blameworthy than Fred. So the offence against fairness is created, specifically, by the difference in treatment between the two agents, not the gap between the treatment meted out to each of these agents and the non-­relative treatment of them that would be demanded by the Control Requirement. Of course, if both Fred and Thomas were blamed to what we regarded as an excessive degree, or in relation to outcomes beyond their control, it would not strain the ordinary sense of fairness to say that both of them had been treated unfairly. But there is a more strictly comparative sense of fairness according to which, if they are both treated the same, then the charge of unfairness cannot arise. It is this second sense of fairness, comparative and interpersonal in character, which matters to the Fairness Intuition.16 The Irrelevance Intuition and the Fairness Intuition will be discussed in detail in Chapters 2 and 3, respectively. Before those discussions are launched, however, we need to take a more detailed look at the types of luck manifested by Drivers and Assassins. When these are grasped more clearly, it will be easier to see what the Irrelevance Intuition and Fairness Intuition are trading upon, and which species of luck the Anti-­Luck Account has invested in.

15  See Feinberg (1974). See also Temkin (2017), pp. 54–6. 16  It is this sense of fairness which matters to standard luck egalitarian theories, which will be under the spotlight in Chapter 6.

40  What Results from Resultant Luck?

1.4.  Three Forms of Resultant Luck It has already been established that Drivers and Assassins exemplify resultant luck: the outcomes of the acts performed by the agents in these cases are due, in part, to factors that are not fully under their control. That bare characterization of resultant luck need not, and will not, be questioned. So what further investigation is there left to conduct? The problem here is that resultant luck has an internal structure that merits further investigation. For one thing, we might observe that luck normally divides into good luck and bad luck. What kind of luck is in play in these cases? Consider Assassins. Imogen achieves her aim: she sets out to kill the President, and she succeeds in doing so. Is that good luck? It looks like good luck: from her point of view (at least until she gets caught), it is good luck. But she is also going to be more blameworthy than Phoebe unless anti-­luckists have their way: is that then bad luck? Or does Imogen experience good luck and bad luck? And what about Phoebe? Does she also experience a mixture of good luck and bad luck, or only one of these forms of luck? Does it matter? The same set of questions can be raised about Fred and Thomas in Drivers. This section will investigate these issues further. The guiding purpose here is to identify that form of luck which is eligible to be pressed into service by the Anti-­Luck Account. If the Anti-­Luck Account is to dispute the tenability of the unequal levels of blameworthiness between Imogen and Phoebe, and between Fred and Thomas, then it will have to appeal to the forms of luck which are exemplified by these cases. Now the fuller story about the Anti-­Luck Account’s shortcomings will not be pursued in this chapter; that business will be conducted in Chapters 2 and 3. But we are in a position even at this stage to carry out a preliminary inspection of the luck-­based materials available to the Anti-­Luck Account. We will consider three such forms of luck.

1.4.1.  Aim-­Sensitive Luck The first form of luck is aim-­sensitive luck, or luck in respect of the successful realization of the agent’s aims. Starting with Assassins, Imogen clearly experiences good aim-­sensitive luck, in the sense that the successful realization of her aim was not thwarted by bad luck: the President did not unexpectedly move, a pigeon did not inadvertently stray into the gun-­to-­President’s-­head trajectory traced by the bullet,17 and so on. Phoebe also experiences aim-­ sensitive luck, but the 17  Cf. Nagel (1979), p. 31.

Three Forms of Resultant Luck  41 aim-­sensitive luck she experiences is bad luck. She fails in her aim: she set out to kill the President, but she fails to do so due to factors over which she has no control. Now for Drivers: in this case, it is Fred who experiences bad aim-­sensitive luck, whereas Thomas experiences good aim-­sensitive luck. By assumption, both these drivers aim to get home safely, without causing injury to anyone else, but Fred is disappointed in his aim, whereas Thomas satisfies his aim. If aim-­sensitive luck is going to be employed by the Anti-­Luck Account in a protest against resultant luck, anti-­luckists will have to acknowledge a striking asymmetry between Assassins and Drivers. In Drivers, Fred has worse aim-­ sensitive luck than Thomas, and he is also more blameworthy than Thomas. For Fred, there is more blame, and worse luck; for Thomas, there is less blame, and better luck. It follows that aim-­sensitive luck is the right sort of resource to invest in for the Anti-­Luck Account’s opposition to the inequality of blameworthiness between Fred and Thomas.18 The structure of the connections between blameworthiness and luck is just what anti-­luckists would expect to find. These connections are differently structured in Assassins. For Imogen, there is more blame, and better aim-­sensitive luck; for Phoebe, there is less blame, and worse aim-­sensitive luck. It follows that the structure of the connections between blameworthiness and aim-­sensitive luck in Assassins cannot be what anti-­luckists were after. Imogen cannot complain that she is unlucky to be deemed more blameworthy than Phoebe by appealing to aim-­sensitive luck, if she has already enjoyed better aim-­sensitive luck than Phoebe did. Correlatively, Phoebe can hardly complain, on the grounds that she has suffered worse aim-­sensitive luck than Imogen, that she is then deemed to be less blameworthy than Imogen. The way that aim-­sensitive luck has fallen out between them is incapable of sustaining a complaint from either of them that blameworthiness between them has not been equalized. For the analysis of Assassins, aim-­sensitive luck cannot give anti-­ luckists what they were looking for. Now perhaps the Anti-­ Luck Account will turn out to favour different approaches in Drivers and in Assassins. This possibility cannot be ruled out at this stage, since there may, after all, be other forms of luck that anti-­luckists can enrol into their analysis. That would leave anti-­luckists free to apply the category of aim-­sensitive luck to Drivers and to reserve some other, more suitable, category of luck for application to Assassins. But it would be surprising, even at this early stage, if the Anti-­ Luck Account’s responses to Drivers and Assassins were dependent on very different sources of luck in each case. On initial inspection, these two cases seem to present the Anti-­Luck Account with a common cause of 18  Again, whether the Anti-­Luck Account would be wise to award a central place to aim-­sensitive luck is a different matter. Here we are exploring the options; we are considering questions of bare eligibility.

42  What Results from Resultant Luck? concern. There are luck-­based differences between the agents in both Drivers and Assassins. Moreover, we know that the Anti-­Luck Account opposes the influence of luck in the assignment of blameworthiness to the agents in both these cases. But these facts appear to imply, at least indirectly, that both cases exemplify a single form of lower-­level luck, hitherto unrecognized, to which the Anti-­Luck Account can appeal in order to press its complaints. If so, what is that form of luck? We must do more to pin it down.

1.4.2.  Pure Interpersonal Luck Imogen enjoys better aim-­sensitive luck than Phoebe, and Fred enjoys worse aim-­ sensitive luck than Thomas. But Imogen is more blameworthy than Phoebe, and Fred is more blameworthy than Thomas. It thus appears that aim-­sensitive luck cannot be the key to the common treatment of Assassins and Drivers. However, the real problem in each case appears to concern the inequality of aim-­sensitive luck. The inequality in aim-­sensitive luck between Imogen and Phoebe explains why Imogen is more blameworthy than Phoebe is in Assassins, and the inequality in aim-­sensitive luck between Fred and Thomas also explains why Fred is more blameworthy than Thomas is in Drivers. The heart of the complaint advanced by the Anti-­Luck Account seems to be that Imogen is more blameworthy than Phoebe in Assassins, and that Fred is more blameworthy than Thomas in Drivers, as a result of factors which lie beyond the control of either of these agents. These factors appear to boil down to the existence of unchosen differences between these pairs of agents. Now it is simply bad faith for Imogen to complain that her attempt to kill the President was successful, and it is not open to her to object to the sanctions that come her way for doing so, simply because she would prefer things to be otherwise. But it is not bad faith—or at least not immediately so—for Imogen to be able to point to an unchosen difference between the verdict befalling her and the verdict befalling Phoebe. The fact that Phoebe is deemed less blameworthy than Imogen is due to factors lying beyond the control of either of them. Similarly, it may be bad faith for Fred to complain that his reckless behaviour was the cause of the pedestrian’s death. His aims may not have been malicious, like Imogen’s, but he still acted recklessly, and exposed others to the unnecessary risk of death and injury. If there is a gamble that you are morally required not to make, then it is bad faith for you to complain when you are blamed in cases where that risk turns out badly. However, Fred is not debarred from pointing to a troubling contrast between what befalls him and what befalls Thomas. Again, the fact that Thomas is deemed less blameworthy than Fred is due to factors beyond the control of either of them. It is the unchosen differences between them that appear to be the source of Fred’s complaint.

Three Forms of Resultant Luck  43 In both Drivers and Assassins, this luck has an essentially interpersonal character. We may call it pure interpersonal luck, or interpersonal luck for short.19 This is the key sub-­species of luck lying behind the worries about resultant luck. To provide some intuitive evidence for this claim, consider the following variants of Drivers and Assassins: Single Driver: Fred drives home drunk, and runs over and kills a pedestrian whom he would not have run over had he been sober. Single Assassin: Imogen, a hired assassin, takes aim and shoots the President, killing him.

Single Driver and Single Assassin lack the interpersonal dimensions of Drivers and Assassins. In these cases, I suspect that there will be less intuitive resistance both to the claim that Fred is blameworthy for recklessly killing the pedestrian, and to the claim that Imogen is blameworthy for maliciously killing the President. These modifications of the cases are free of the luck-­reflecting comparisons between Fred and Thomas, in Drivers, and between Imogen and Phoebe, in Assassins, and as a result we are not distracted by those comparisons. If those comparisons matter, then intuitive responses to Drivers and Assassins will be shaped by the presence of interpersonal luck. It is because Thomas is less blameworthy than Fred, though the difference between them lies beyond their control, that Fred’s level of blameworthiness becomes problematic. The same point can be made about Imogen and Phoebe in Assassins.

1.4.3.  Pure Comparative Luck Of course, the awareness of interpersonal luck will then sharpen our awareness of the counterfactual possibilities which apply to these agents, taken one by one. Imogen did not miss the President, but she easily might have done. The difference between hitting her target and not hitting it did, after all, lie beyond her control. The world had to cooperate with her in order to produce this outcome, and she lacked decisive control over whether it would do so. What is the significance of this unchosen difference? Similarly, Fred ran over a pedestrian, but he easily might not have done. He might, like Thomas, have encountered only empty roads. The difference between killing and not killing was in this sense beyond his control. What should we say about that truth? These comparisons are not interpersonal, since they are

19  This form of luck is centrally relevant to debates about responsibility-­sensitive justice, which will be examined in Chapters 6 and 7.

44  What Results from Resultant Luck? concerned only with counterfactual possibilities which apply to each agent, taken in turn. Nonetheless, these intrapersonal cases do trade on the same structure of luck which is present in the cases of pure interpersonal luck: they involve a comparison between different possible outcomes. This broader form of luck, of which interpersonal luck is a sub-­species, may be called pure comparative luck, or comparative luck for short. It is this form of luck, I believe, in which the Anti-­Luck Account needs to invest. Comparative luck handles both the intrapersonal and the interpersonal cases, and it applies equally to Drivers and Assassins. We shall return to comparative luck in Chapter 2. It will have a large role to play in this book.

1.5.  Why the Accommodation Strategies Fail Resultant luck can sometimes seem relevant in ways which would appear to defy the strictures of the Anti-­Luck Account. I will be arguing in Chapters 2 and 3 that this intuitive data should motivate us to find fault with the Anti-­Luck Account. In the meantime, it should be noted that friends of the Anti-­Luck Account have evolved certain strategies for the accommodation of this intuitive data without making any official concessions to critics of the Anti-­Luck Account. I shall call these the ‘accommodation strategies’. The aim of this section is to examine a few of them. (Appendix I will extend this line of argument a bit further: its topic will be a further argument from anti-­luckist quarters that is concerned specifically with empirical flaws in our responses and judgements.) My argument will be that these accommodation strategies are unsuccessful.20 If we wish to preserve appearances for Drivers and Assassins—that is, if we wish to hold on to the verdicts that Fred is more blameworthy than Thomas, and that Imogen is more blameworthy than Phoebe—then a more full-­frontal or principled assault on the Anti-­Luck Account is necessary. That conclusion does not, in and by itself, deliver any decisive argument for the Restricted Account, but it does help to suggest that the intuitive data arising from differences in our verdicts about these agents can at least be straightforwardly donated to theories opposed to the Anti-­Luck Account. It remains to be seen whether any such theory—the Restricted Account in particular—is in a position to make productive use of this data.21 20  In taking on the accommodation strategies, I have been helped by the discussions in Domsky (2004), esp. pp. 448–53, and Domsky (2005). I am considerably less sympathetic, however, to Domsky’s enrolment of the literature on empirical psychological biases to undermine opposition to the Anti-­Luck Account: again, see Appendix I. 21  My treatment here does not pretend to be exhaustive. One putative form of accommodation strategy turns on different senses of ‘blame’: see Thomson (1993) for the proposal, and Domsky (2004), pp. 449–50 for discussion of it. Thomson’s argument is discussed more fully in Chapter  2, Section 2.1.4.

Why the Accommodation Strategies Fail  45

1.5.1.  The Epistemic Strategy One important accommodation strategy is the Epistemic Strategy. The Epistemic Strategy explains the verdicts that Fred and Imogen are more blameworthy than their respective counterparts by appealing to epistemic factors. More specifically, our accommodation of resultant luck is due, or can be due, to our efforts to compensate for our epistemic shortcomings. The judgement that there is differential blameworthiness between the agents in each pair reflects only an ‘epistemological’ rather than ‘ontological’ interpretation of their blameworthiness.22 If these ­epistemic shortcomings were corrected, the Anti-­Luck Account would find itself unchallenged. This is how the Epistemic Strategy goes. We may favour harsher sanctions for Fred (relative to Thomas) and Imogen (relative to Phoebe) because the deaths they cause offer indispensable evidence that they were acting from morally flawed mental states. Though the mental states which compose or contribute to each of Phoebe’s and Thomas’s acts may be indistinguishable from those composing or contributing to Fred’s and Imogen’s, failed attempts and near misses tend to leave fewer evidential traces in the world than successful attempts and collisions. So, when we envisage the central candidates for the assignment of blameworthiness, we will tend to summon to mind these successful malicious attempts and harm-­ causing outcomes from negligent and reckless acts. These outcomes give us a more reliable basis for casting blame at the agents who are involved in their production. For agents who have made unsuccessful attempts, and for agents whose negligent or reckless actions have resulted in no harm, either there is room for doubt about their culpability, or else we are less disposed to find them blameworthy because unsuccessful attempts and harmless negligent acts provide less vivid or stable evidence of agents’ badness than successful attempts and harmful negligent acts.23 It must, of course, be conceded to the Epistemic Strategy that those agents whose attempts have been unsuccessful, or whose negligent or reckless activity has resulted in no harm, will often go under the radar. We simply do not get to know about them. These are, variously, cases of lucky misses, and lucky escapes. But the Epistemic Strategy cannot obviously account for the intuition, held by holders of the Restricted Account, that there is differential blameworthiness between the pairs of agents in Drivers and Assassins. As these cases are described, Thomas and Phoebe have satisfied the Internal Claim just as fully as Fred and Imogen. The cases have been carefully set up so that it is only the luck-­based outcome that differs. Every part of the background psychological story in Drivers is shared by Fred and Thomas. The same goes for Assassins: in terms of 22  See Richards (1986) for this language. 23  For such arguments, see, for example, Rescher (1993).

46  What Results from Resultant Luck? psychological activity and contribution, we cannot distinguish between Imogen and Phoebe. So our persistence in the verdict that Thomas is less blameworthy than Fred, and that Phoebe is less blameworthy than Imogen, is not obviously due to any epistemic deficit. Proponents of the Epistemic Strategy are likely to make two replies at this point. The first is that, when the various parameters of these cases have been carefully emphasized, so that the only differences between the pairs of agents are genuinely grasped as being luck-­based, some people might switch to anti-­luckist verdicts after all. This is true, but it will not account for those whose verdicts remain anti-­anti-­luckist because they are unfazed by resultant luck. A second reply is that although cases such as Drivers and Assassins may have been carefully controlled for such epistemic factors, our responses to them nonetheless reveal the influence of habits that have already been formed and are hard to break. Our exposure to cases with epistemic controls built into them cannot overcome our habits of thought that have been nourished through exposure to cases lacking such controls. As an example of this style of response, Neil Levy has defended a form of the Epistemic Strategy in hard-­wired, evolutionary terms.24 His leading claim is that our habit of making outcomes relevant to blameworthiness is ultimately in the service of ascertaining which mental states our fellow moral agents are acting from. Levy offers a broader story to make good on this proposal. As creatures who value cooperation, our primary interest lies in the quality of will displayed by other agents. It is these agents’ quality of will that best attests to the reliability of their future cooperation and their ongoing commitment to moral rules. Now if we were to dismiss evidence of actual outcomes as irrelevant to the quality of will displayed by these agents, we would make it too easy for such an agent to pretend that the quality of her will was non-­objectionable, and that the harmful outcome of what she did was purely accidental. So we play safe by tying our judgements about quality of will to bad worldly outcomes. Yet our ultimate interest in allocating blameworthiness lies in agents’ quality of will, not in the outcomes they produce. The outcomes are consulted only as a kind of heuristic. The problem with Levy’s suggestion is twofold. First, if we need to appeal to bad outcomes to reassure ourselves that the quality of agents’ wills is deficient, then we would lack an explanation of why poorly willed agents such as Thomas and Phoebe are blameworthy. If, by contrast, we are already satisfied that the quality of will manifested by Thomas and Phoebe is morally objectionable, then we are surely bound to reach exactly the same verdicts about Fred and Imogen. It is hard to see how the harmful outcomes produced by Fred and Imogen could  then possibly enhance their blameworthiness, since—by common

24  See Levy (2016).

Why the Accommodation Strategies Fail  47 agreement—those outcomes reflect only resultant luck. Second, the evidence-­ gathering strategy is ineffective in any case: if a doubt has arisen as to whether an agent’s act betrays a morally objectionable quality of will, the appeal to harmful outcomes will not resolve it. This is because the mere fact that an act causes a harmful outcome does not demonstrate that the agent’s quality of will was morally objectionable.25 An act’s harmfulness does not, in and by itself, favour the imputation to that agent of a morally objectionable quality of will rather than a morally non-­objectionable quality of will. Of course, our evolutionary development may have instilled bad habits in us, and coaxed us into commitments that in no way advance our aims, but it seems like an unconvincing piece of theory-­saving to insist that these are the unshakeable habits we have acquired, and moreover that the appeal to their existence quickly dissolves the problem of moral luck in favour of the Anti-­Luck Account. It is surely more natural to think that, if we are convinced that Fred is more blameworthy than Thomas, and that Imogen is more blameworthy than Phoebe, these verdicts simply reflect the belief that outcomes matter, and that these agents’ resultant luck is no barrier to these assignments of blameworthiness. The content of that belief cannot be explained, or explained away, by the Epistemic Strategy. Now the Anti-­Luck Account may yet show that belief to be false, but it cannot be pretended that the verdicts of differential blameworthiness are, at a deeper or purely ‘ontological’ level, in accord with the Anti-­Luck Account. They are not. They pose a challenge to the Anti-­Luck Account.

1.5.2.  The Deterrence Strategy The Deterrence Strategy makes harsher sanctions for Fred and Imogen the product, not of their greater blameworthiness, but of policies which are meant to alter the incidence of crime and wrongdoing.26 The central idea behind the Deterrence Strategy is that, by punishing offenders who produce bad outcomes in the world, we improve those outcomes. That policy can be endorsed without capitulating to the claim that culpable agents with bad luck are more blameworthy than culpable agents with good luck. 25  Kumar (2019) appeals to a similar argument as part of an anti-­anti-­luckist strategy. He then embeds this reasoning within a broadly consequentialist framework, owing something to the hybrid rule-­consequentialist theories of punishment advanced by Rawls (1955) and Hart (1968), which allows him to account for the relevance of outcomes, and thus the existence of resultant luck. Kumar’s account is interesting, but these building blocks offer hostages to fortune, and I am not convinced that they are fully reliable. My aim, in any case, is to offer a different way of accommodating the relevance of outcomes. 26  The tenor of the argument given by Browne (1992) is notably hostile to the Deterrence Strategy. Browne’s central contention is that the typical emphasis on punishment in this literature actually helps to make anti-­luckism seem more attractive than it really is.

48  What Results from Resultant Luck? The Deterrence Strategy is not convincing, upon further consideration. First, since agents do not make separate decisions about whether their malicious attempts will be successful, or whether their negligence will be harmful—all of that is a matter of luck—no policy can incentivize offenders to make unsuccessful rather than successful attempts, or to engage in harmless rather than harmful negligent behaviour.27 Second, and in line with David Lewis’s suggestion, although it might be sensible to dis-­incentivize the making of further attempts if the original attempt is unsuccessful, it does not follow that we should punish successful attempts more harshly than unsuccessful attempts; we can simply ensure that each fresh attempt is punished, so that agents have a genuine incentive not to renew the making of a malicious attempt if at first they are unsuccessful.28 The basic rationale behind the Deterrence Strategy, when all is said and done, must be that deterrence is better achieved if the global levels of blameworthiness are increased. The more hostile the environment is for potential wrongdoers, the better. The underlying idea here will be that, rather than punishing each wrongful attempt or negligent act to the same degree, regardless of whether those acts produce harm, we raise the degree of blameworthiness for harmful acts, so that these acts collect some additional blameworthiness. That additional injection of blameworthiness can also deter offenders who have not yet made any attempt, whether successful or unsuccessful. But if the justification for the Deterrence Strategy consists simply in having more blameworthiness swilling around in the system, or in producing a more hostile environment for would-­be offenders to deliberate in, then the aims of the Anti-­Luck Account would be more consistently achieved by adopting other measures. We might attach, most obviously, a higher level of penalization to negligent or malicious acts, regardless of whether these acts actually produce harm. Again, verdicts which suggest differential levels of blameworthiness between these agents do not naturally derive from any commitment to the Deterrence Strategy.

1.5.3.  The Taking Responsibility Strategy The central idea behind David Enoch’s favoured form of accommodation strategy is indebted to Susan Wolf and to some extent Margaret Urban Walker. I call it the Taking Responsibility Strategy.29 To this end, Enoch presses hard on the distinction between being responsible for an outcome, and taking responsibility for that outcome. I might, for example, take responsibility for the actions of my

27  Lewis (1989), p. 54. 28  Lewis (1989), p. 55. 29  See Enoch (2012), Wolf (2001), and Walker (1993). Strands of this material will be revisited in Chapter  5, Section  5.7. Another treatment of this approach to responsibility is offered by Mason (2019), ch. 8.

Why the Accommodation Strategies Fail  49 misbehaving teenage child, or my country’s conduct in international affairs. Or I might, if I am a lorry-­driver, take responsibility for the harm that I faultlessly inflict on a pedestrian in a traffic accident.30 Because it is assumed that the harms in these cases are not the upshots of my core agency, I am not responsible for them. Enoch is an ardent anti-­luckist who takes the Control Principle with the utmost seriousness.31 But because these cases all fall within the penumbra of my agency, I might nonetheless be under a duty to take responsibility for them. Enoch compares the structure of the Taking Responsibility Strategy to promise-­making. I can be under a duty to make a promise to T, where in the absence of that promise I would not be under a duty to T. As an example, I might be under a duty to promise my gravely ill friend to look after his daughter if he dies. I would not be under a duty to assume full responsibility for her without that promise, but it might be morally obligatory for me to make this promise to him because, in the absence of it, he will be tormented by anxiety over her future prospects.32 I may have a duty, in these circumstances, to allay his anxiety by making him the promise, which will then generate the further duty to take care of his daughter if the worst happens and she is actually orphaned. In other cases, we will also have to invoke additional considerations to explain why the duty to undertake a duty exists. One productive way of thinking about this issue is to ask why it is good for us to possess the normative powers to take responsibility in these cases of penumbral agency.33 For the case in which I take responsibility for my delinquent teenage child, I may appeal to the value of the parent-­child relationship.34 That relationship would have a different and arguably inferior character if parents lacked the power to share certain burdens arising from their children’s misdemeanours. More generally: . . . although there may be value in sometimes thinking of ourselves and others as pure agents, still it seems clear that there is value in not constantly thinking of ourselves in this way. The power to take responsibility for the consequences of our actions seems necessary for our ability to think of ourselves as such impure agents.35

Enoch also suggests that the Taking Responsibility Strategy can be extended to a case such as Drivers.36 But we need to be cautious here. In particular, we need to 30  Enoch (2012), pp. 97–9. Enoch takes the example of the lorry-­driver from Williams (1981a). This lorry-­driver is more akin to Noah in Sober Driver, whom we encountered in Section 1.2.1. Noah is not impugned by the Restricted Account, so the example seems anomalous for my purposes. As we shall see, though, Enoch proposes an extension to deal with cases of resultant luck such as Drivers. 31  See Enoch and Marmor (2007), and Enoch (2008). 32  Enoch (2012), pp. 105–6. 33  See Owens (2012) for a more general exploration of this territory. I briefly express some reservations about Owens’s project in Lang (2015c). 34  Enoch (2012), pp. 123–4. 35  Enoch (2012), p. 124; original emphasis. 36  Enoch (2012), p. 130.

50  What Results from Resultant Luck? be told which considerations ground this particular extension of his account. What are the advantages of operating with an impure agency model rather than a pure agency model in Drivers? What is in it for Fred? What is in it for the rest of us? In Enoch’s other cases, we can appeal to the value of the relationships involved. This consideration covers the parent-­child case, and also the citizen-­country case. In the lorry-­driver case, we might also appeal to the value of social relationships where those who faultlessly cause harms feel obliged to make amends and to express regret for what they have done. (Perhaps, if we were under a standing invitation to accept responsibility for only ‘pure agency’ while being able to dismiss talk of responsibility for any manifestations of ‘impure agency’, we would find it all too easy to persuade ourselves that the outcomes were hardly ever our fault.) Drivers seems different from these other cases. Fred has acted wrongly, after all, for which we are already blaming him. The question now is how far that fault extends. Opponents of the Anti-­Luck Account will make the pedestrian’s death the direct upshot of Fred’s reckless behaviour, but Enoch requires another explanation. Since the Anti-­Luck Account holds Fred to be no more blameworthy than Thomas, it would seem that his moral responsibility must stop at reckless driving, before the pedestrian’s death can be even taken into account.37 Meanwhile, the supplementary considerations which show that Fred is nonetheless under a duty to take responsibility for the pedestrian’s death are elusive. Perhaps they consist simply in deference to the common-­sense conviction that Fred is responsible for the pedestrian’s death. But that does not look like a form of accommodation strategy; it looks like a capitulation to the sort of anti-­ anti-­luckism that is represented by the Restricted Account. Enoch’s task is a delicate one: he needs to explain why it would be inappropriate for Fred not to take responsibility for the pedestrian’s death while releasing him for responsibility for her death in the absence of that undertaking. In achieving this goal, there is less room for manoeuvre than Enoch thinks. On the one hand, if we are agreed that Fred is not responsible for the pedestrian’s death, due to the fact that it was just a matter of luck that his reckless driving caused her to die, then what is stopping Fred from saying just that? His refusal to take responsibility for her actual death can be explained precisely by the fact that he is not responsible for it. On that matter, he can expect Enoch’s backing. On the other hand, if we think that Fred should take responsibility for her death, then the most straightforward explanation of this duty is that he is responsible for it. In fairness to him, Enoch does not claim to be offering much more than a schema for further development—his aim is only to establish ‘a possibility-­result 37  Enoch gives us little to work with here, but this issue is complex. It is also relevant to the scope/ degree distinction, which I have already flagged up in Section 1.2.3. See also Chapter 3, Section 3.2.2, and Chapter 4, Section 4.4.2.

Why the Accommodation Strategies Fail  51 rather than an outright conclusion’38—but his account, just as it stands, does lack obvious answers to these puzzles.39

1.5.4.  The Next Steps The various accommodation strategies have failed to deliver any decisive victories for the Anti-­Luck Account. There is genuine rivalry between the Anti-­Luck Account and the Restricted Account, and the reflective intuitions that there are morally significant differences between Thomas and Fred, or between Phoebe and Imogen, cannot be handled in these particular ways. They cannot be explained away by the accommodation strategies. Of course, it does not follow at this stage that these intuitions are better negotiated by the Restricted Account than by the Anti-­Luck Account, since their dependability has not yet been demonstrated. That task comes next. In Chapter 2, I will examine the Irrelevance Intuition, and in Chapter 3, my attention will turn to the Fairness Intuition.

38  Enoch (2012), p. 100. 39  Here is another possible problem for Enoch: even if he right to think that Fred is under a duty to take responsibility while Thomas is not, mustn’t that difference be due to circumstantial luck? Enoch accepts that such a form of luck is involved, but insists that this form of luck is ineliminable: we cannot slough off our moral duties just because we lack control over the circumstances under which they apply to us (2012, pp. 130–1). As a substantive verdict, that stance strikes me as a wise one, but I am not convinced that it fully satisfies the spirit of the Anti-­Luck Account to which Enoch has pledged his allegiance. Chapter 4 offers a more general discussion of circumstantial luck.

2

Resultant Luck and the Irrelevance Intuition It is time to take a more concerted look at the Irrelevance Intuition. I will tackle the Irrelevance Intuition in two distinct phases. First, in Section 2.1, I shall outline a new case—the Two Buttons case—that furnishes a number of important lessons in a particularly vivid way. Second, in Sections 2.2 and 2.3, I return to our original cases, Drivers and Assassins, to suggest that the lessons yielded by Two Buttons can be applied, with suitably adjusted commentary, to the agents in these cases as well. If my argument succeeds, the Irrelevance Intuition will have been shown to be untenable. A further challenge, concerning a division between primary and secondary duties, is tackled in Section  2.4, and Section  2.5 concludes with a couple of important caveats.

2.1. The Two Buttons Case I am going to introduce and then dissect a new case, in order to allow a number of connected principles to be advanced. These principles will permit the Restricted Account to take on more shape and structure, and to illuminate its normative grounds. Consider: Two Buttons: There are two agents, Akira and Dylan, and there is a devilish machine, with two buttons. Pressing Button 1 will cause ten innocent ­strangers to be killed. Pressing Button 2 will cause an eleventh innocent stranger to have her arm broken, but to leave the ten unharmed. The buttons do not carry this precise information. Both Button 1 and Button 2 are adorned with the same sign, in large lettering: ‘DO NOT PRESS! OTHER PEOPLE, PERHAPS SEVERAL OF THEM, WILL BE KILLED OR SEVERELY HARMED’. Neither Akira nor Dylan is forced to press either button, but each of them is disposed to do so, because each of them has the malicious aim of causing some degree of harm to others. Dylan presses Button 1, and Akira presses Button 2.

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0003

The Two Buttons Case  53 As stylized as it may be, we can recover three significant lessons from Two Buttons. I shall now present them, before applying these lessons to Assassins and Drivers.

2.1.1.  The Relevance Claim The first lesson concerns the presumptive relevance of the actual outcome to the level of blameworthiness it is appropriate to extend to the agent who caused it. Our more particular and focused question is this: is it proper to hold Dylan to be more blameworthy than Akira, in light of the fact that he is causally responsible for the deaths of ten individuals, while Akira is causally responsible only for another individual’s broken arm? Their intention was similar: it was to cause some degree of harm to other people. Both of them are, not just causally responsible, but plainly culpable. Their motivations were malicious. But Dylan causes more harm than Akira; the outcome he causes is therefore morally worse than the outcome Akira causes. Now it may be a matter of luck that Dylan causes more harm than Akira, since it is just a matter of luck that he presses Button 1 rather than Button 2; by assumption, he did not have any way of knowing what the results of pressing it would be. However, the actual harm which ensues from his action cannot be easily detached from the overall level of blameworthiness which is properly assigned to him. Because he caused more harm—because the outcome of pressing Button 1 is morally worse than the outcome of pressing Button 2—we can presumptively say that Dylan is more blameworthy than Akira. This line of thought suggests the following thesis: Relevance of Outcomes to Blameworthiness Claim: The badness of an outcome is, roughly, proportionally relevant to the level of blameworthiness it is appropriate to extend to moral agents who are at fault for causing the outcome to occur.

I will abbreviate this thesis to the Relevance Claim in what follows.1 One rival approach in this case would be to split the difference between the Button 1 outcome and the Button 2 outcome, thus making Dylan and Akira equally blameworthy. If the connection to the badness of outcomes is to be preserved, the average harm caused in Two Buttons will be slightly under the 1  The ‘roughly’ qualification will be addressed in Section  2.5. For those who are eager to apply Zimmerman’s distinction between the scope and degree of responsibility, first mentioned in Chapter 1, Section 1.2.3, to the agents in Two Buttons, see Chapter 3, Section 3.2, when it is tackled in connection to the Fairness Intuition. For the time being, I am assuming that scope and degree go together: since Dylan is responsible for a larger amount of harm than Akira, he is also responsible to a higher degree than Akira is.

54  Resultant Luck and the Irrelevance Intuition badness of five killings (if we assume that the badness of a broken arm is less than the badness of a killing, and then split the difference between the badness of ten deaths produced by pressing Button 1 and the badness of one broken arm produced by pressing Button 2). But this option seems artificial and under-­ motivated. Dylan presses a button which causes ten innocent individuals to be killed. If his blameworthiness for doing this is to accommodate the relevance of worldly outcomes, then it seems arbitrary to withhold blame from him for anything less than the killing of the ten individuals. Why should he be blameworthy, in effect, for only some of these killings, but not all of them, since he did kill all of them, and is blameworthy for doing so? It distorts the connections between action and outcome to affirm that the level of blameworthiness should be tied to an outcome which we know in advance will not be actually instantiated. The more straightforward approach is to use the facts about the badness of actual outcomes to shape the facts about blameworthiness. It then follows that Dylan will be blamed for the deaths of ten individuals, while Akira will be blamed for the broken arm of an individual. Both outcomes are bad, but the death of ten individuals is worse than the breaking of an individual’s arm. So Dylan is more blameworthy than Akira.

2.1.2.  The Denial Claim The second lesson concerns the weakness of a line of defence against the application of the Relevance Claim. Imagine that Dylan now issues the following complaint: Dylan’s Complaint: Akira and I may have acted on objectionable intentions, and we are both blameworthy for that. But the fact that I produced a worse outcome by pressing Button 1 cannot reflect especially badly on me, since it did not relevantly lie within my control that I pressed Button 1, in particular, while Akira pressed Button 2. The fact that I pressed Button 1 rather than Button 2 was just a matter of luck, and luck cannot make me any more blameworthy than she is.

Notice what Dylan is demanding: he is demanding that, however things turn out, he can be deemed no more blameworthy than Akira. This suggestion is reinforced if we imagine that things do in fact turn out differently: in this alternative scenario, in world W*, Akira presses Button 1, and Dylan presses Button 2. In W*, we might expect a structurally identical complaint from Akira. If Dylan’s complaint is worth taking seriously, then Akira’s complaint would have to be taken just as seriously. These complaints bottom out at the same place. They jointly demand that, regardless of who presses which button, neither of

The Two Buttons Case  55 them can be deemed more blameworthy than the other. But that is just to insist that blameworthiness be equalized for acts which cause outcomes that differ in their degrees of badness. What these agents are demanding, in short, is that the badness of the wrongful acts they perform should be detached from the blameworthiness which it is appropriate to ascribe to them for performing those wrongful acts. It is difficult to see how the appeal to lack of control can uphold Dylan’s complaint. True, Dylan does not have control, in the relevant sense, over which of the two buttons he presses, due to his epistemic limitations. He does, however, have control over the fact that there is one of them he presses.2 He might have declined to press either of them, and this is what he was morally required to do. In summary, then, Dylan’s appeal to his relative lack of control over which button he presses as a reason for evading blameworthiness seems unconvincing in virtue of two facts. First, it lay within Dylan’s control to decline to press either button. Second, he was under a moral obligation to decline to press either button. Dylan therefore cannot credibly appeal to the tribunal of morality to correct one type of putative injustice if he was already under moral instruction to do something else. In particular, he cannot appeal to the authority of the Ought Implies Can principle if he has ignored the ‘ought’ which instructed him to refrain from pressing either of these buttons in the first place. The insistence that Dylan is more blameworthy than Akira does not, therefore, involve any suppression of the force of Ought Implies Can. It simply involves the refusal to re-­accommodate Ought Implies Can at the expense of the Relevance Claim. Since outcomes matter, and since it is possible to accommodate Ought Implies Can in another way, Dylan lacks a compelling defence against his greater blameworthiness. These reflections now permit us to recover the second lesson: Denial of a Defence Based on Lack of Control Claim: Moral agents lack an adequate defence against the blameworthiness-­allocating implications of the Relevance Claim if their decision to perform a wrongful act with an outcome which is partly due to luck lies within their control.

I will abbreviate this to the Denial Claim in the following discussion. The fact that Dylan lacks control over the badness of the outcome he produces by pressing one of the buttons does not exempt him from being deemed blameworthy for that outcome, if he fails to perform an alternative, morally 2  More precisely: in a certain sense, Dylan does have control over his pressing of Button 1. He does, after all, intentionally press it, and he exercises agency in doing so. But in another sense, he lacks full control over the circumstances in which he comes to press Button 1 rather than Button 2, since he lacks epistemic control. Two Buttons thus illustrates the difference between deliberate control and coincidental control: due to his epistemic limitations, Dylan enjoys coincidental control but not deliberate control over which button he presses. See Zimmerman (2006), p. 592, for these useful terms.

56  Resultant Luck and the Irrelevance Intuition required, act that did lie within his control. He cannot credibly appeal to a lack of control over the outcome of the wrongful act if he has already freely declined to do what morality instructed him to do.

2.1.3.  Combining the Internal Claim with the External Claim So far we have been squarely concerned with the External Claim. To recapitulate, this is the claim that the degree of an agent’s blameworthiness is roughly proportioned to the badness of the actual harmful luck-­dependent consequences which proceed from her acts.3 The Relevance Claim and the Denial Claim are intended to secure the External Claim. This will not be surprising, since the Restricted Account is committed to taking steps to uphold the External Claim. But what about the Internal Claim? To recapitulate, this is the claim that agents are eligible for blameworthiness only if they have acted in ways that reflect objectionable mental states such as malicious intentions, or recklessness, or negligence. Aren’t we still left with a stubborn gulf separating the Internal Claim and the External Claim? I believe that Two Buttons closes that gap to some extent. There are conceptual links between the Internal Claim and the External Claim, if we are patient enough to flush them to the surface. When we have done so, we will be in possession of the third lesson from Two Buttons. But we must proceed slowly here. If the Relevance Claim and Denial Claim are true, then Akira and Dylan are blameworthy for what each of them respectively caused to happen. The outcomes of their acts help to fix their blameworthiness. But we should now, I believe, be prepared to go further, and endorse the suggestion that these ‘external’ facts also morally shape the ‘internal’ facts, concerning the relative moral badness of the internal states on which they acted. The rough shape of the proposal is this. In Two Buttons, what Akira and Dylan cause to happen is non-­accidentally related to what they aim to do, and what they in fact do; they act, after all, in order to produce these outcomes. Dylan satisfies his aim of causing harm to others by pressing Button 1, which causes ten individuals to be killed, and Akira satisfies her aim of causing harm to others by pressing Button 2, which causes another individual severe injury. Just as the outcomes caused by their acts vary in terms of badness, I now want to suggest that the internal states on which they acted can also vary in terms of badness.4 Dylan’s internal states are morally worse than Akira’s internal states, because the 3  The terms ‘External Claim’ and ‘Internal Claim’ were first encountered in Chapter 1, Section 1.2.3. 4  A brief clarification: clearly, these acts are both wrong, but ordinary English makes wrongness a binary property—either an act is wrong or it is not. However, wrong acts can vary in terms of badness, which is a scalar property.

The Two Buttons Case  57 act he performs to which those internal states are significantly motivationally related causes ten people to be killed, whereas the act Akira performs to which her internal states are significantly motivationally related causes only one person to have her arm broken. More generally: if the result of an act is morally worse, then the internal states which are embodied in or motivationally related to that act are also morally worse. Dylan’s wanting to harm others is worse than Akira’s wanting to harm others, because Dylan’s wanting leads to more harm, while Akira’s wanting leads to less harm. This is a recognizably externalist line of thought. It may appear that this externalist suggestion is suspiciously easy to secure, and therefore little better than a trick, since there is undeniably a way in which an agent’s mental states can be evaluated in terms of the outcomes they produce via agential activity. It is not a demanding task to concoct a dimension of assessment according to which Dylan’s mental states are worse than Akira’s mental states if and only if they eventuate in more harm. There are many conceivable dimensions of assessment. Why should this one command any special interest? But my suggestion, as I see it, is not the product of idle or mischievous stipulation. Its underlying point is that we have no genuine purchase on the objectionableness of these mental states unless we import information about how the acts that embody them turn out in the world. This externalist line of thought may be a surprising one. It requires, and will receive, a much fuller defence, but it should be acknowledged in the meantime that there is some strong independent critical pressure on the Restricted Account to embrace it. Embracing the thought that external outcomes can help to fix the moral significance of the internal mental states is not just an eye-­catching but strictly unnecessary expansion of the Restricted Account’s commitments. In fact, eschewal of this line of thought carries certain strategic costs. On an anti-­luckist view, the claim that Akira and Dylan should not be blamed to a differential extent ultimately reflects the conviction that they are equally bad. This is precisely Judith Thomson’s strategy, which we will now examine in some detail.5

2.1.4.  Acts, Agents, and Blame Some brief scene-­setting to kick things off: Thomson distinguishes between two senses of ‘blame’, and it will be useful to have them in mind as we examine the uses to which Thomson puts them. The first sense of blame is concerned with questions of attribution, and is non-­scalar, or all-­or-­nothing:

5  See Thomson (1993).

58  Resultant Luck and the Irrelevance Intuition A person P is to blame for an unwelcome event or state of affairs just in case P caused it by some wrongful act or omission for which P had no adequate excuse.6

The second sense of blame is concerned with the connection between character and act (or omission), and it also comes in degrees, which makes it scalar: A person P is to greater or lesser blame for doing (or being) such and such, where his doing (or being) such and such is unwelcome, just in case P’s doing (or being) the such and such is stronger or weaker reason to consider P a bad person.7

Thomson thinks that it is only the second sense of blame which can count towards agents’ ‘demerit’, or their being such as to justify our looking askance at them, or thinking ill of them.8 She considers a negligence case, involving two agents, Bert and Carol, in order to drum up support for this point: Suppose (i) we hear that Bert was negligent and caused a death, whereas though Carol was also negligent she caused nothing untoward . . . But suppose (ii) we hear in addition that Bert was no more negligent than Carol, and that the difference between them lies in the bad luck he had and she did not. Now do we believe Bert is more to blame than Carol? I am sure we do not. Bert of course is to blame for a death (first notion ‘blame’), and Carol is not; but Bert is no more to blame than Carol (second notion ‘blame’), for given (ii), it is plain that Bert’s history gives no more reason to think him a bad person than Carol’s gives to think her a bad person.9

Applied to Two Buttons, Thomson’s argument would invite us to agree that Akira’s and Dylan’s characters are equally bad, since the fact that Dylan presses Button 1 and Akira presses Button 2 is just a matter of luck, and luck cannot make any justifiable difference to how bad we think they are. Even if one has every intention of resisting Thomson’s anti-­luckist views, it is easy enough to be talked into conceding to her that Dylan and Akira are equally bad. That is because the concession still appears to leave room for different sorts of subsequent evaluation. This concession among anti-­anti-­luckists implies the hope that they may be able to cut off the Thomson-­style anti-­luckism at a later 6  Thomson (1993), p. 200. 7  Thomson (1993), p. 202. The ‘being such and such’ part of this proposal will be attended to, in effect, in Chapter 3. 8  Thomson (1993), p. 202. 9  Thomson (1993), p. 205; original emphasis.

The Two Buttons Case  59 pass. In particular, and noting Thomson’s own division between the two senses of blame, they might distinguish between evaluation of the agent, and evaluation of the agent’s act. Even if Akira and Dylan are equally bad, their acts, or the blame­ worthy results of what their acts, undeniably differ in badness. The badness of ten deaths exceeds the badness of one broken arm. That leaves conceptual room for the blameworthiness to attach to the act, rather than the agent, which will allow the Relevance Claim to be upheld. Following Robert Hartman, we can label this line of thought the Two-­ Judgement Strategy.10 Applied to Two Buttons, the Two-­Judgement Strategy concedes that the lucky difference between Akira and Dylan may not establish any difference in our verdicts about them, or about their characters. But it insists that this still gives us room to maintain different verdicts about their acts. This concessive strategy is also adopted, in effect, by Michael Otsuka. He thinks that blameworthiness should be allowed to vary, even if the variation in outcomes reflects luck, just as long as the agents are at fault for performing the act in the first place. These commitments are congenial to the Restricted Account. But Otsuka takes a more concessive line in assessing the badness of character: How morally bad a person is might depend only on factors internal to him such as his intentions, motives, beliefs, and volitions, rather than on any consequences of his actions that lie beyond his skin.11

This line of thought goes deep, and it may appear impregnable. Arguing from an opposed anti-­luckist perspective, Enoch and Marmor take a similar view about a pair of agents similar to Fred and Thomas in Drivers: Surely, no difference is warranted with respect to the judgment about the moral character of these two agents. By drinking and driving they both exemplified a similar flaw of character, and to a similar degree of severity. Generally speaking, and barring backward causation, it is clear that there is nothing in the future consequences of an agent’s actions that can affect his moral character retrospectively . . .12

10  See Hartman (2017), pp. 124–5. Hartman is avowedly following the example originally set by Greco (1995); his aim is to refine Greco’s original proposal and defend it against objections. 11  Otsuka (2009), p. 381. 12  Enoch and Marmor (2007), p. 409. In addition to Hartman (2017) and Greco (1995), Lewis (1989), p. 54, and Moore (1994) also seem tempted by this move. It appeals to writers on both sides of the substantive debate about resultant luck.

60  Resultant Luck and the Irrelevance Intuition

2.1.5.  Doubting the Two-­Judgement Strategy The Otsuka-­style concession strikes me as unwise, because it is difficult to maintain.13 The problem, more particularly, concerns the nature of the connection between what Akira and Dylan are like, morally speaking, and the badness of what they do. In Two Buttons, by assumption, what Akira and Dylan do is the only admissible evidence for what they are like.14 Now Two Buttons is supposed to furnish us with iron-­clad evidence that Akira and Dylan are equally bad. This is Thomson’s view, and it is what the anti-­anti-­luckist critic is supposed to have conceded to her. But the only evidence for their equal badness consists in the acts they have performed, together with the psychological background to these acts. If the only evidence for their equal badness consists in the acts they performed, together with the relevant psychological backgrounds, it is difficult to see how the verdict on the agents can get out of alignment with the verdict on their acts. If we are agreed that Dylan and Akira are equally bad—if that is deemed a secure datum—then there is also going to be some pressure to agree that what they do is equally blameworthy. That is the thought which Thomson appears to be relying upon. Once it is granted, it is much more difficult to escape from her anti-­luckist orbit than Otsuka and others may think. Concessive anti-­luckist critics might think that this reply is wide of the mark. The evidence is not restricted to the acts performed by the agents, together with the results of those acts, they might insist: for we can easily abstract from the result of the acts and focus only on the internal states which fed, in each case, into the acts performed by these agents. This is precisely what Otsuka is suggesting we do. The trouble with this strategy is that, by retreating in this manner to a more purely internal luck-­ free realm, we risk being unable to accommodate the Relevance Claim and the Denial Claim. If Akira and Dylan are bad, and moreover equally bad, then the results of what they do have simply not been consulted in order to determine their badness. And, if that is so, then we will be paying only lip service to the relevance of those outcomes. They will not have any effect on our calculations as to these agents’ demerit. The evidence for Akira’s and Dylan’s badness should not be restricted to their internal states, free of any connection to the outcomes caused by the acts of which those states were a constituent. In summary, once the concession to badness of character has been made, then Thomson’s anti-­luckist resolution of the issues seems powerful. After all, if there is a choice about what blameworthiness attaches to—agent or their acts—then blameworthiness would seem to attach fundamentally to agents. While there is no 13  I do not insist that the Two-­Judgment Strategy is uniformly unusable, though I have elsewhere presented doubts about its applicability to other contexts: see Lang (2004). 14  The same goes for Thomson’s Bert and Carol case, and for all the standard cases found in this philosophical sub-­genre.

The Two Buttons Case  61 outright inconsistency between insisting that Akira and Dylan are equally bad, and also insisting that what they do nonetheless differs in blameworthiness—the Two-­ Judgement Strategy does not suffer from outright incoherence—the connections are not clean. It is hard to see how the worldly outcomes can make their presence felt, if they do not have to be consulted in order to arrive at a final reckoning of the badness of the agents who caused them to occur. Akira’s act is her act, and Dylan’s act is his act. If these acts are blameworthy to an unequal degree, then Akira and Dylan should be blameworthy to an unequal degree. If Akira and Dylan are nonetheless blameworthy to the same degree, it is difficult to see how the differential blameworthiness of their acts has genuinely been provided for. Is there another way for anti-­anti-­luckists to go? I believe there is.

2.1.6.  The Determination Claim Otsuka’s claim succinctly states a familiar thought, which plainly nourishes the Irrelevance Intuition. The main items for the evaluation of agents are psychological in nature: they are the medley of the agent’s intentions, motives, beliefs, and volitions. It is only when we fasten on these agential psychological features that we have secure data for the moral evaluation of the agent. Now these items cannot be disregarded. They do indeed loom large in any estimation of the agent’s badness, and degree of fault. I do not propose that they be left aside. But the moral character of an agent’s internal mental items should not be assumed to hold independently of their connections to the world. My proposal, in short, is that we should be prepared to externalize these internal items. This will allow us, in Two Buttons, to maintain the close fit between agent-­evaluation and act-­evaluation. Dylan’s act in Two Buttons is morally worse than Akira’s act, and his internal items are also morally worse than hers. In this resolution of the issues, we allow the act-­ evaluations to shape the agent-­ evaluations. Insofar as Two Buttons provides evidence of how bad Akira and Dylan are, we should conclude, accordingly, that Dylan is morally worse than Akira. The internal items over which he relevantly enjoyed control, and which are embodied in his act, have a significance that is not fully disclosed until we discover what they cause to happen. So, because what Dylan caused to happen is worse than what Akira caused to happen, we should conclude that Dylan’s internal states are morally worse than hers. And that, in turn, makes Dylan morally worse than Akira. This claim may, at first, seem simply perverse. Imagine that, in the nearby possible world W*, Dylan presses Button 2, while Akira presses Button 1. Is there really a convincing story we can tell as to how Dylan is a better person in W* than

62  Resultant Luck and the Irrelevance Intuition he is in the actual world, where he pressed Button 1, whereas Akira is a worse person in W* compared to how she is in the actual world? Aren’t both agents the same, morally speaking, in both worlds? In pointing to different outcomes based on exercises in random button-­pressing, we are not providing the materials for an intelligible story of moral deterioration, on the one hand, and of moral improvement, on the other hand. This is, of course, an understandable response. When we discuss character, we take an interest both in individuals’ moral track records and in the medley of dispositional states that can be significantly related to those track records.15 The events reported in Two Buttons are a small fragment of those broader histories. They may not be able to tell us all that much by themselves. We would need to enrich the story considerably to arrive at a comprehensive appraisal of Akira’s and Dylan’s characters. Though this point is of course true, it cuts both ways. If the acts performed in Two Buttons fail to provide us with sufficient evidence of the badness of Akira’s and Dylan’s respective characters, then they also fail to provide us with the evidence that Akira and Dylan are equally bad. Perhaps we should simply adopt a policy of agnosticism about their characters, and insist on our exposure to a greater range of evidence before coming to any conclusions at all about them. But that will not give Thomson what she was after. Thomson is banking on the fact that in these precisely demarcated, luck-­reflecting cases of interpersonal wrongful action, we can readily agree that the characters are equally bad. Their equal badness is then, in turn, supposed to impose pressure on us to agree that they are equally blameworthy. If Thomson wants to use this case to establish the equal badness of Akira and Dylan, then the attempt to use it to point to their differential badness does not risk embarrassment simply because the evidential base is relatively narrow and because we can always find more to say about character and agential history. The important question is this: insofar as the acts performed in Two Buttons tell us anything about Akira’s and Dylan’s respective characters, as one part of the overall moral track records that can be attributed to them, what conclusions are to be drawn? My answer is simply that Dylan is worse than Akira, because his enacted dispositions make him worse than Akira’s enacted dispositions. We are now in a position to state the third lesson: Moral Determination of Internal States by Actual Outcomes Claim: The badness of an actual outcome caused by moral agents who act on morally flawed internal states can help to fix the degree of badness of the mental states which were embodied in these agents’ acts.

15  See Chapter 4 and Appendix II for a differently focused discussion of these issues.

The Two Buttons Case  63 This will be abbreviated to the Determination Claim in the following discussion. Though Akira and Dylan both act from the intention to cause some degree of harm to others, the badness of these internal states is not fully determined until we have the information about the actual outcomes of caused by their acts. In the absence of that information about the actual outcomes, the badness of these internal states remains partly under-­determined. We may be able to tell, before we know what the outcomes are, that the internal mental states on which these agents act are culpable, and thus morally bad, but we do not yet know how bad these internal states are. The final degree of badness of these internal states is, at least to some extent, porous, and it is fixed by actual outcomes. This means that the badness of the internal states which are embodied in these agents’ acts is fixed retrospectively, through importing information about how these acts turn out. This may be an unusual proposal, but it does not involve, as Enoch and Marmor suggest, any mysterious species of backward causation.

2.1.7.  Action and Externalism What is the status or ambition of the Determination Claim? I do not regard it as a contribution to the theory of action or philosophical psychology. It is not genuinely competitive with accounts in the theory of action which insist upon narrower resolutions. Donald Davidson, for example, tells us the following: We must conclude, perhaps with a shock of surprise, that our primitive actions, the ones we do not do by doing something else, mere movements of the body— these are all the actions there are. We never do more than move our bodies: the rest is up to nature.16

My proposal does not issue a challenge to Davidson. These are different debates. Why this insouciance? I can state a couple of reasons. First, and most importantly, it cannot be resolved within action theory whether it is only primitive actions that are properly evaluated by moral standards. That particular question is a moral question, and it can only be resolved within normative ethics. The notion of a human action, and the distinction between a human action and what falls outside a human action, strictly considered, does not carry special significance to this particular debate in moral philosophy. At least, I can see no reason to suppose that it does. To maintain otherwise would be to let the tail wag the dog. Thus the metaphysical question of what constitutes an action,

16  See Davidson (1980), p. 59.

64  Resultant Luck and the Irrelevance Intuition and how the division between internal and external factors in the individuation of action should be understood, is not my subject. Second, the ultimate destination for rival internalist views about blameworthiness may take us beyond Davidsonian primitive actions in any case. Andrew Khoury wishes us to imagine different versions of a single-­shooter case much like a non-­interpersonal version of Assassins. He outlines three cases. In Case 1, Lee, an assassin, makes a successful attempt on his victim. That makes Lee like Imogen in Assassins. In Case 2, Lee makes an unsuccessful attempt on his victim. That puts Lee on a par with Phoebe in Assassins.17 In Case 3, Lee takes himself to have made an attempt on a victim, but fails due to the fact that, ‘unbeknownst to him, he has just recently had his brain removed from his skull, placed in a vat of liquid, and hooked up a computer which stimulates the external environment’.18 In Case 3, Lee fails to make any genuine attempt: he handles no rifle, he squeezes no trigger, and there is no actual victim for him to hit or miss. Everything that unfolds is simply going on in his head, and nowhere else. Khoury assumes, more generally, that ‘[f]or any bit of behaviour that any agent engages in, there will be a hypothetical mental twin who has been recently envatted’.19 Due to the conceptual possibility of this envatted counterpart, and because Khoury holds that there are no interesting moral differences in Lee’s responsibility across the three cases—because there is ‘no difference from the inside’20—he embraces the conclusion that ‘[w]e are responsible only for elements of our mental life’. ‘Trying, volitions, and willings’, more particularly, are the elements of our mental life for which we can be held responsible.21 The logic of a strongly internalist position, then, when pushed to its limit, arguably goes beyond Davidsonian primitive actions to reach mental states, and mental states only.22 The commitments evinced by the Determination Claim parallels externalist positions with respect to mental or semantic content.23 But these comparisons are bound to be complicated, and I am not going to labour them here. If an externalist account of blameworthiness is to earn its stripes, it must do so by appealing to considerations which are proprietary to the particular concerns of this debate, rather than through off-­the-­p eg ap­pro­pri­ ation from debates in the philosophy of mind and language. We must find

17  Khoury (2018), pp. 1358–9. 18  Khoury (2018), p. 1360. 19  Khoury (2018), p. 1361. 20  Khoury (2018), p. 1359. 21  Khoury (2018), p. 1364. 22  Khoury is not hostile to the claim that there are different ways of describing what Lee does across the three cases. In Case 1, one description of what Lee does is to commit an act of homicide. The victim’s being killed may be merely a contingent property of the event of Lee’s willing, but the fact that this property is contingent does not rule out the truth of that homicide-­involving description. ‘Thus,’ Khoury concludes, ‘the implications of the view that we are only responsible for our willings may be less revisionary than it may initially appear’ (2018, p. 1366). I myself have little doubt that Khoury’s internalist commitments have revisionary implications: see my discussion of the ‘Division Objection’ in Chapter 3, Section 3.2. 23  For the classic expressions of some of these arguments, see Putnam (1975), McGinn (1977), Burge (1979) and (1986).

The Two Buttons Case  65 what we are looking for in a theory of substantive responsibility, and then follow the argument to where it leads.

2.1.8.  Does the Determination Claim Cheat? There are other immediate concerns with the Determination Claim. Is the Determination Claim guilty of cheating? It may seem suspiciously made-­to-­order: first, the full force of the External Claim is brought to bear on Akira and Dylan, via the Relevance Claim and the Denial Claim, and then the Internal Claim is made to adapt to that pressure. That would be an uncharitable verdict. Consider how porous the internal states are in Two Buttons. Akira and Dylan act with the intention of causing harm to others, but they do not know how exactly that intention is to be satisfied at the time of acting. The outcomes of the button-­pressing then plausibly fix the badness of these intentions, because they fix the particular ways in which these intentions come to be satisfied. Dylan’s intention to cause harm to others is satisfied, more particularly, by the killing of the ten which ensues from his pressing Button 1, and Akira’s intention to cause harm to others is satisfied, more particularly, by the broken arm which ensues from her pressing Button 2. Given this structuring role between intention and outcome, it seems plausible that the badness of the internal states embodied in Akira’s and Dylan’s respective acts is fixed by the respective outcomes of these acts. Dylan succeeds in what he set out to do: he succeeds by causing ten individuals to die. The same goes for Akira: she succeeds by causing an individual to have her arm broken. The way in which their respective aims are satisfied requires that we consult the information about the actual consequences of what they do.

2.1.9.  Two Buttons and Comparative Luck We have used Two Buttons to yield three claims: the Relevance Claim, the Denial Claim, and the Determination Claim. In the sections to follow these lessons will be applied to Assassins and Drivers. But before returning to these cases, it may be worth trying to get clearer on the precise intuitions that are mobilized in Two Buttons. Now I have argued for an ascription of differential blameworthiness between Akira and Dylan. But anti-­luckists will resist this argument. What might be driving their resistance? To get clearer about this question, imagine the following variants of Two Buttons: First Single Two Buttons: Like Two Buttons, but there is only Dylan. Dylan is not forced to press either button, but he is disposed to do so, because he has the malicious aim of causing some degree of harm to others. Dylan presses Button 1.

66  Resultant Luck and the Irrelevance Intuition Second Single Two Buttons: Like Two Buttons, but there is only Akira. Akira is not forced to press either button, but she is disposed to do so, because she has the malicious aim of causing some degree of harm to others. Akira presses Button 2.

These variants of Two Buttons shadow the variants of Drivers and Assassins outlined in Section 1.4.2. Will anti-­luckists wish to deny that Dylan should be blamed for the ten dead individuals in First Single Two Buttons? Will they wish to deny that Akira should be blamed for the individual’s broken arm in Second Single Two Buttons? My suspicion, once again, is that the inclination to accept these implications may be stronger in them when the interpersonal structure of the original case has been neutralized. You might resist this, on the grounds that what really matters is that Dylan might just as easily have pressed Button 2 in First Single Two Buttons, and that Akira might just as easily have pressed Button 1 in Second Single Two Buttons. It is these counterfactual possibilities which matter, rather than the fact that they are realized, in Two Buttons, in an interpersonal way. We should concede the point. What the point establishes is that it is comparative luck, rather than interpersonal luck, which is really pulling the strings here. Can comparative luck show that the Relevance Claim, the Denial Claim, and the Determination Claim are misapplied to Two Buttons? No. Comparative luck matters when the gamble is benign.24 In Two Buttons, the act of button-­pressing is not benign. Dylan lacks the right to insist that he can be judged to be no worse than he would be if he pressed Button 2, because he lacks the right to insist that he is not blameworthy for the particular consequences of pressing Button 1. He ought to have declined to press either button. Neither does Dylan have the right to deny that the internal items embodied in his act have a significance which is then disclosed by the result of pressing Button 1.

2.2.  Assassins and Two Buttons With these lessons under our belt, I shall now return to Assassins. A briefer treatment of Drivers will follow, in Section  2.3. This part of the argument needs to proceed slowly, in order to fully unfurl the implications of the Relevance Claim, Denial Claim, and Determination Claim for these standard cases of resultant luck.

24  See Chapter 3, Section 3.4.1.

Assassins and Two Buttons  67

2.2.1.  Assassins and the Relevance Claim The immediate task for the Restricted Account is to establish that Imogen can be rightly blamed for what she does, taking the actual outcome into consideration. Does Imogen suffer from a form of luck which demonstrates that the high level of blameworthiness assigned to her is somehow misplaced? Is the fact that Imogen killed the President really irrelevant to how we should morally appraise her? I suggest not. The outcome of Imogen’s act may have involved luck, but the crucial fact is that the world conformed to her intention. The luck she experiences actually consists in the absence of factors which would have subverted her intention.25 We already know that Imogen experiences good aim-­sensitive luck.26 In other words, the luck she experiences assists the realization of her intention, and the moral quality of her will is faithfully reflected in the state of affairs produced by her act. Since Imogen set out to change the world in such a way as to conform to the content of her desire, and actually succeeded in doing so, she can be blamed for the actual outcome. The luck involved in the actual outcome does not disrupt any morally relevant connections among the unfolding sequence of intention, motivation, act, and outcome which can be imputed to her. Our moral appraisal of her does not, therefore, naturally fall short of the successful realization of the intention on which she acted. In pointing to the outcome she caused, we manage to say something morally informative about the act she performed, which allows us, in turn, to say something morally informative about her. So the Relevance Claim applies to Imogen in full force. By extension, the Relevance Claim also applies to Phoebe. Phoebe may have acted on exactly the same intention as Imogen, but her attempt was unsuccessful. It follows that there is a difference between Imogen’s blameworthiness and Phoebe’s blameworthiness; Imogen is more blameworthy than Phoebe.

2.2.2.  Assassins and the Denial Claim Moreover, the Denial Claim still applies in full force. It is not open to Imogen to insist that her blameworthiness cannot exceed Phoebe’s blameworthiness on the grounds that it was not under her control whether her attempt would be successful. It also lay within her control to decline to make the attempt on the President’s life. Imogen cannot appeal to the tribunal of morality to concede that 25  Our awareness of this form of luck is of course considerably sharpened by the presence of such factors in Phoebe’s case. (We would ordinarily not pay much or indeed any attention to this form of luck.) 26  See Chapter 1, Section 1.4.1.

68  Resultant Luck and the Irrelevance Intuition successful attempts are no more blameworthy than failed attempts on the grounds that she did not have any control over whether her attempt was going to be successful, when morality had already instructed her not to make the malicious attempt. As far as the Denial Claim is concerned, our view of Assassins should take our cue from Two Buttons.

2.2.3.  Assassins and the Determination Claim What about the Determination Claim? This may seem more problematic. The intentions at play in Two Buttons had a porous, partly under-­ determined character, and this feature arguably helped to make Two Buttons susceptible to the Determination Claim. Akira and Dylan both act on the intention to cause some degree of harm to others. But the way in which they satisfy this intention is different: Dylan satisfies it by killing ten, while Akira satisfies it by causing severe injury to one. They do not know how their aims will be satisfied until they discover what results ensue from pressing these buttons. The intentions at play in Assassins present us with a harder case, because Imogen’s and Phoebe’s intentions seem both identical and fully determinate. Each of them aims at nothing more and nothing less than the killing of the President. Now if the Relevance Claim enjoys application, it will follow that Imogen is more blameworthy than Phoebe. If the Denial Claim enjoys application, it will follow that Imogen lacks a defence against her greater degree of  blameworthiness. If the Determination Claim also enjoys application to Assassins, it will follow that Imogen’s act was worse than Phoebe’s, because the badness of the act which embodies this intention turned out to exceed the badness of the act which ­embodies Phoebe’s intention. But how can that be, if they acted on exactly the same intentions? The content of their respective intentions was exactly the same. One can, in fact, imagine two lines of complaint being issued at this point, from Imogen and Phoebe, respectively. Let us state them in turn, starting with Imogen: Imogen’s Complaint: Phoebe and I may have acted on objectionable intentions, but if having an objectionable intention is what makes us eligible for blame in the first place, then we must surely be held equally blameworthy. The fact that my attempt was successful, whereas Phoebe’s attempt was unsuccessful, was just a matter of luck, and luck cannot make me any more blameworthy than she is. It surely cannot make my intention any morally worse than hers was. To remind you, we acted on exactly the same intention!

Assassins and Two Buttons  69 Now here is Phoebe: Phoebe’s Complaint: You cannot have it both ways. If you want to maintain that Imogen merits a level of blameworthiness which fully takes into account the actual outcome she causes, and you also want to maintain that the badness of this actual outcome determines the badness of the intention on which she acts, then it is no longer obvious that I am to blame any more than Shona, an entirely uninvolved bystander to the whole saga. Perhaps it was just a matter of luck that my attempt was unsuccessful, but you have already denied that the luck Imogen experiences makes any difference to the relative degree of blameworthiness which is due to her. So why not apply consistent standards and concede that, on this occasion at least, I am not blameworthy to any degree? In other circumstances, my intention might have been lethal. I grant you that. But it was harmless in this particular case, so the Determination Claim must surely imply, in this case, that I am blameless, just like Shona.

These complaints impose different types of critical pressure on the Restricted Account. Imogen’s complaint resists the Determination Claim by pointing to the exact qualitative similarity between the content of her intention and the content of Phoebe’s intention at the time of acting. As Imogen sees it, the luck which was involved in the production of the actual outcome cannot disturb this parity of badness between their intentions, and thus their acts. In other words, Imogen uses the resources of the Internal Claim to resist the encroachment of the External Claim. Phoebe’s complaint, by contrast, goes in the opposite direction. There are actually two readings of Phoebe’s complaint. On one reading of it, she uses the resources of the External Claim to block application of the Internal Claim. On another, more sophisticated reading of it, she is prepared to accept that the Determination Claim establishes that the moral quality of the internal states can be externalized, but then takes advantage of the harmlessness of what she did as a way of arguing for the moral innocence of the intention upon which she acted. On this reading, Phoebe’s complaint, in effect, raises the question of why the Restricted Account does not collapse, when all is said and done, into the Strict Liability Account. If outcomes are relevant to blameworthiness, as the Relevance Claim insists, and if these outcomes are also supposed to help fix the badness of the internal states on which they acted, as the Determination Claim contends, then a malicious intention which fails to produce an objectionable outcome must be off the moral hook. There is nothing left, in the Restricted Account, to make Phoebe’s intention objectionable. Her act was harmless. These pressures on the Restricted Account can be resisted. We can start with Phoebe’s complaint. According to the Determination Claim, the badness of a

70  Resultant Luck and the Irrelevance Intuition malicious intention is not fully settled before its actual consequences are enumerated. Alternatively put: the actual outcome partly, but only partly, determines the badness of the intention. The implication is clear: the fact that such an intention is bad, and therefore blameworthy to some degree, can be established before the results are in. Neither is this an ad hoc fix, designed simply to accommodate the competing pressures of the Internal Claim and the External Claim. The very ambition of the Restricted Account is to assign blameworthiness to acts which reflect culpable mental states. Malicious intentions clearly qualify as culpable, and that is basically a matter of attitudinal orientation. It is a descriptive property of her internal states, which can be registered even before the results of acting upon them have arrived. Phoebe’s intention is clearly culpable, just as Imogen’s intention is clearly culpable, and we do not have to await the outcome of her attempt before we are in a position to say so. These agents’ orientation towards other moral agents and patients is not what it should be, and so both of them qualify for some level of blameworthiness. That is the first major claim, or opening bid, of the Restricted Account. But the badness of malicious intentions thus acted upon cannot be fully understood in the absence of information about the states of affairs ensuing from acts which embody those malicious intentions. That is the second major claim in the Restricted Account, and it provides the substance of the reply to Imogen’s complaint. To be clear, it is not the propositional content of Imogen’s intention that is targeted by the Determination Claim; the critical target is Imogen’s having this intention and acting on it. Both Imogen and Phoebe have the same intention. The state of affairs they wish to bring about is the same, so the content of the intention on which they act is also the same. That truth is not challenged by the Determination Claim. The Determination Claim is concerned instead with the badness of the act which embodies this intention. As I see matters, the badness of this act is partly up for grabs; it depends on what happens in the world. It follows that the badness of acting on malicious intentions is partly fixed retrospectively, depending on whether the aim embodied in the malicious agent’s act was successfully realized. As a result, and in conclusion, there is variation in what agents acting on objectionable intentions can be blamed for, depending on how things turn out. The application of the Determination Claim to Assassins and Two Buttons invites a little more commentary. In Two Buttons, the agents know in advance that they are either going to harm others, but they do not know how harmful their acts will be. That feature may help to make it plausible to apply the Determination Claim to Two Buttons: whichever outcome eventuates, it seems reasonable to hold them responsible for it, and to insist that the outcome is blameworthy because it faithfully conforms to their intention to have caused some degree of harm to others. The outcomes do not mislead. They simply add determinacy to the

Drivers and Two Buttons  71 intentions which can only be described, prior to the outcomes, with a certain degree of indeterminacy. Now the intentions acted upon by Imogen and Phoebe lack the porousness of the intentions acted upon by Akira and Dylan. They are fully determinate. But this difference in structure does not defeat the application of the lessons from Two Buttons. The moral quality of the intentions acted upon by Akira and Dylan in Two Buttons only exemplify, in sharper or more vivid form, what is also true in Assassins. Akira and Dylan do not know what is going to happen when they press the buttons. By comparison, Imogen and Phoebe do not know whether their attempts are going to turn out successfully: their attempts could be successful, or they could fail. Since the actual outcome helps to determine how wrongly they have acted, and thus how blameworthy they are, it will also determine how bad the internal states on which they acted are. Because Imogen’s attempt was successful and Phoebe’s attempt was unsuccessful, it was worse for Imogen to have acted on this intention than it was for Phoebe to have acted on her intention. Hence Imogen is more blameworthy than Phoebe.

2.2.4.  Assassins and Comparative Luck So far, there is no reason to think that the Relevance Claim, Denial Claim, or Determination Claim lack application to Assassins. The differences between Two Buttons and Assassins do not block or undermine the application of these claims to Assassins. What about comparative luck, though? Can this make an eleventh hour difference? It might be complained that, even if the three claims can be applied to Imogen and Phoebe one by one, there is still a problematic difference between these two agents, which is due to luck. Imogen is more blameworthy than Phoebe, but that difference lies beyond the control of either of them. This source of luck definitely exists. The question is whether it matters. My view is that it cannot make a difference to the fate of the Irrelevance Intuition. If we are free to apply the Relevance Claim, the Denial Claim, and the Determination Claim to both Imogen and Phoebe, then the results of those applications should stand. It follows that comparative luck is not tracking anything worth tracking. The only unexamined possibility for the relevance of comparative luck lies with the Fairness Intuition. That will be the business of Chapter 3.

2.3.  Drivers and Two Buttons Now we turn to Drivers. Most of the lessons that were drawn from Two Buttons, and then applied to Assassins, can also be applied to this case as well, so the

72  Resultant Luck and the Irrelevance Intuition commentary here can afford to be briefer. But there are descriptive differences between Drivers and Assassins, as we have already acknowledged, and we need to consider whether these make any further normative difference.

2.3.1.  Drivers, Recklessness, and Aim-­Sensitive Luck We should remind ourselves of two facts about Drivers. Drivers involves recklessness, rather than the malicious intentions which were directing Imogen’s and Phoebe’s acts. Fred and Thomas are aware of the fact that they are imposing an unjustified risk on pedestrian safety, but they get into their cars and begin their journeys home anyway. Neither Fred nor Thomas sets out to harm anyone, but— we assume—each of them aims simply to get home in one piece. Though they act recklessly, they do not seek anyone’s harm or injury. Second, and as a result, the distribution of aim-­sensitive luck differs in Drivers from the distribution of aim-­sensitive luck in Assassins. In Drivers, Fred has had bad aim-­sensitive luck, whereas Thomas has had good aim-­sensitive luck. Yet, if resultant luck is taken into account, Fred is more blameworthy than Thomas. In this case, then, Fred is on the losing side of an inequality in aim-­sensitive luck between him and Thomas, and he is also held to be more blameworthy than Thomas. Does this make Fred’s moral position brighter than Imogen’s, in such a way as to jeopardize the unequal assignment of blameworthiness recommended by the Restricted Account? It was suggested earlier that aim-­sensitive luck will not give the Anti-­Luck Account what it was looking for across the two central cases, Assassins and Drivers,27 but it does not follow that Fred should be stymied at the very outset from presenting a complaint about his treatment which is based on his worse aim-­sensitive luck. Perhaps this difference in the aim-­sensitive luck between Fred and Thomas might soften our attitude to Fred. I deny that the appeal to the difference in aim-­sensitive luck can successfully uphold Fred’s complaint. Aims are not the only psychological items to which we should be paying attention when we consider recklessness. Reckless behaviour exposes others to severe risk of danger and injury, in ways which qualify reckless agents as blameworthy, however their acts turn out.28 Again, reckless agents have an attitudinal orientation towards others which is not what it should be, though this orientation is manifested differently in reckless agents than it is in malicious agents.29 For this reason, reckless behaviour is the sort of behaviour which engages the Internal Claim; these acts are culpable.

27  See Chapter 1, Section 1.4.1. 28  See Steinbock (1985), and Husak (1994). 29  Pritchard (2005), pp. 255–72, is attracted to an anti-­luckist position, though he exploits the Modal Account of luck (see the Introduction, Section 2) to point to useful distinctions we can make

Drivers and Two Buttons  73 Though I have stipulated that Fred and Thomas are reckless, they might have been negligent instead: that is, although they may not have been consciously aware of the risk they are imposing on pedestrians and other drivers, they may have fallen short of the awareness of risk which it is incumbent upon us to display. If we were assuming that it is negligence rather than recklessness of which Fred and Thomas are guilty, the facts about aim-­sensitive luck would be the same, and the commentary we have already offered could be straightforwardly re-­applied.

2.3.2.  Drivers and the Relevance Claim If reckless behaviour engages the Internal Claim, then we cannot deny the relevance of the actual outcomes caused by reckless behaviour in Drivers, any more than we can deny the relevance of the actual outcomes caused by malicious behaviour in Assassins. The story, at this point, will go through in a broadly similar way. The disvalue of reckless behaviour is reliably manifested by Fred’s actions, since the fact that he acts recklessly causes the pedestrian to be killed. Morality prohibits individuals from acting recklessly precisely because of the unacceptable risk that things will turn out the way they do in Fred’s case. There is a direct connection between his recklessness and the pedestrian’s death; by assumption, if he had refrained from driving recklessly while drunk, the pedestrian would have not died. In a way which parallels Assassins, the wrongness of what Fred did reaches out into the world; it reaches as far as the death of the pedestrian. The harm caused by his recklessness was the result of his decision to act recklessly. If the wrongness of what Fred did reaches out into the world, we should expect the Relevance Claim to be applicable to Drivers. Thomas’s act is also blameworthy, but the fact that he did not cause anyone to be killed or injured will carry the implication that he is less blameworthy than Fred. Though he cannot take any credit in this lucky difference between him and Fred, he did less harm than Fred did, and so he is less blameworthy than Fred is.

among different manifestations of recklessness. In particular, Pritchard suggests that it is worse to drive drunk through a crowded city, for example, than a largely deserted rural area. I agree with him. We may indeed wish to provide for different degrees of recklessness, and factual differences in the social environment, such as those which concern local population density, can certainly contribute to our account of these differences. However, I do not accept Prichard’s contention that these points usefully contribute to a debunking explanation of why we think that Fred and Thomas are blame­ worthy to an unequal degree. We are committed to this difference in blameworthiness because we think that outcomes matter, particularly when they are foreseeably related to the acts which caused them to happen. Our commitment to the Relevance Claim goes very deep. It seems to me that this commitment is not explained away or even really addressed by Pritchard.

74  Resultant Luck and the Irrelevance Intuition

2.3.3.  Drivers and the Denial Claim Along with the Relevance Claim, we should also expect the Denial Claim to be satisfied. Fred lacks a defence against the higher degree of blameworthiness he collects, due to the fact that he has already ignored morality’s demand to refrain from driving while drunk. Because Thomas also ignored the moral demand to refrain from driving while drunk, he is also blameworthy, even though his reckless behaviour did not produce death or injury. The Denial Claim is the claim that agents who behave wrongly cannot seek refuge in Ought Implies Can. They could, and should, have satisfied the ‘ought’ of not behaving recklessly in the first place. If there is already a case for accommodating the worldly consequences of what they do, as the Relevance Claim contends, then the Denial Claim ensures that this case cannot be scuppered through appeal to Ought Implies Can.

2.3.4.  Drivers and the Determination Claim There is also no reason to think that the Determination Claim remains unsatisfied in Drivers, though the route to its satisfaction differs from that which is involved in Two Buttons or Assassins. In Two Buttons, Akira and Dylan do not know which button they are pressing: for each of them, it could be Button 1 or Button 2. Assassins is different: in this case, Imogen and Phoebe do not know whether their malicious attempt is going to turn out successfully: their attempt might be successful, or it might fail. And Drivers is different yet again. In this case, Fred and Thomas do not know whether they are going to get away with acting recklessly: they might cause death or injury, or they might not. Despite these differences in structure, these cases have an important similarity: they all involve acts which are culpable. This feature sustains an important implication. Like Two Buttons and Assassins, it is not fully determined in Drivers how bad Fred’s internal states were until the particular outcome eventuates. Because Fred is more blameworthy than Thomas, due to the badness of the outcome he caused, the culpability of his reckless internal states is also augmented. As it turned out, Thomas’s internal states were only moderately bad, because they did not result in severe harm. As things turned out for Fred, however, his internal states were extremely bad, because they resulted in the avoidable death of an innocent pedestrian. Now it is true that Fred did not have full control over the conditions under which his internal states collected this extreme level of badness. But we should not be complaining about a lack of volition or control at this point in the story. We do not create or design the world, and what we do combines with unchosen facts about physical laws and the physical environment to produce certain outcomes. Yet we are not captives to the unchosen nature of this environment. We can refrain from acting wrongly, and thus avoid exposure to the unchosen

Primary versus Secondary Duties  75 combination of wrongful internal states and the world in which those states are manifested.

2.3.5.  Drivers and Comparative Luck There is no reason to think that the existence of comparative luck makes any more of a difference to Drivers than it did to Assassins. As for Assassins, there are differences between Two Buttons and Drivers, but these do not block or undermine the application of the Relevance Claim, the Denial Claim, and the Determination Claim to Drivers. Once these claims can be demonstrated to enjoy application, it is too late for comparative luck to make any difference. Comparative luck exists, but it seems irrelevant. If it was relevant, then we would have to adjust the commentary on what Fred and Thomas do. But there is no reason to adjust the commentary. We should conclude instead that comparative luck does not manage to track anything important. Once again, there may be an eleventh hour appeal from the Fairness Intuition. Those issues will be explored in depth in Chapter 3.

2.4.  Primary versus Secondary Duties In this section, I want to consider another possible challenge, which purports to undermine the externalist foundations of the Restricted Account.30 I apply this challenge to Drivers and Assassins in turn.

2.4.1.  The Separate Duties Challenge According to this challenge, we can divide the duties pertaining to Drivers into two types, which can be stated in the following imperatival form. The primary duty is this: ‘Refrain from driving while drunk’. The secondary duty is this: ‘If you drive drunk, do not harm anyone’. The secondary duties are fall-­back duties, which kick in when primary duties have been violated. Once Fred and Thomas embark on their drunken journeys, they have already flouted their primary duties. The fact that they have both violated their primary duties has nothing to do with resultant luck. It is due squarely to recklessness instead, which is manifested to an equal degree in the violation of their primary duties. As far as their primary duties are concerned, then, they are equally culpable. But, having flouted their primary duties, Fred and Thomas still have a secondary duty which 30 This distinction between primary and secondary duties turns up in other contexts: see, for example, Gardner (2014), for its application to tort law.

76  Resultant Luck and the Irrelevance Intuition may or may not be flouted. Fred flouts his secondary duty, whereas Thomas satisfies his secondary duty. But the difference between them with respect to the violation or non-­violation of their secondary duties can only be a matter of luck. It is not due to recklessness, which was at play only in respect of the violation of their primary duty. As a consequence, Fred’s bad luck does not exacerbate the disvalue of his recklessness. So this blocks the Determination Claim. Furthermore, and as a result, it reinstates the problem of resultant moral luck. The same line of analysis can be extended to Assassins. The primary duty in Assassins is this: ‘Refrain from attempting to assassinate someone who is not liable to be assassinated’. The secondary duty is this: ‘If you attempt to assassinate someone, do not be successful’. Again, like Drivers, the violation of Imogen’s and Phoebe’s primary duties in Assassins has nothing to do with resultant luck. Their respective violations of the primary duty are explained by their malicious intentions. Having violated their primary duties, there are still secondary duties in play: Imogen violates her secondary duty, while Phoebe does not. But whether or not the secondary duties are violated is due solely to luck, not to the malicious intentions which explained how the primary duties came to be violated. The malice which explains the violation of the duty not to attempt to kill the President has nothing to do with the successful completion of the attempt, and so has no bearing on the fact that Imogen succeeds in assassinating the President. The same challenge duly arises: if Imogen’s violation of the secondary duty is due solely to luck, and Phoebe’s non-­violation of the secondary duty is also due solely to luck, then the Determination Claim is blocked, and the problem of resultant luck is reinstated. Because it rests upon this way of dividing duties, I shall call this challenge the Separate Duties Challenge. The Separate Duties Challenge is supposed to undermine the ambition of the Restricted Account to show that the Internal Claim can be reconciled with the External Claim through the adoption of an externalist account of the internal states. The faulty internal states which matter to the violation of the primary duty do not explain the subsequent violation of the secondary duty. As I see it, the division between the primary and secondary duties is contrived, at least in connection to resultant luck, and the Separate Duties Challenge is wrongheaded. I discuss its application to Drivers and Assassins in turn.

2.4.2.  The Separate Duties Challenge and Drivers In connection with Drivers, the strategic purpose of making the division between primary and secondary duties is to contract the domain to which recklessness can be deemed relevant: recklessness supposedly matters only to the violation of the primary duty, not to the violation of the secondary duty. But this cannot be

Primary versus Secondary Duties  77 correct, since the malign influence of recklessness endures even after Fred and Thomas unwisely step into their vehicles. By assumption, the sudden appearance of the unfortunate pedestrian during Fred’s journey home is one for which Fred, in virtue of his drunkenness, is unable to properly deal with. But then the explanation of why Fred kills the pedestrian cannot fail to invoke his drunkenness, or the recklessness which explains why, though drunk, he is behind the wheel in the first place. Had he not been drunk, he would have been able to avoid killing her. (That is an explicit assumption of Drivers.) So Fred’s drunkenness surely belongs to the explanation of why he kills her. But then the explanation can afford to go deeper: had Fred not been reckless, he would not be driving while drunk. These truths demonstrate that a non-­elliptical explanation of why Fred kills the pedestrian cannot fail to invoke his recklessness. Recklessness makes an indispensable contribution to the explanation of why Fred violates his secondary duty. So the Separate Duties Challenge wrongly supposes that the bad-­making internal states in Drivers are relevant to violation of the primary duty only, rather than to violation of the primary duty and the secondary duty. Two points might be made in response. First, and less promisingly, it might be maintained that Fred’s drunkenness while driving robs him of the ability to avoid pedestrians who are unlucky to find themselves in his path, and therefore annuls the claim that he has any further duty at this point, the violation of which can then be explained by his recklessness. But this reply is, of course, not one to which proponents of the Separate Duties Challenge can appeal, since it effectively implies that there was only ever one duty operating here: to refrain from driving while drunk. We already know from the Relevance Claim that the duty to refrain from driving while drunk cannot exhaust what is morally relevant; the death and injury which ensues from drunken driving cannot be irrelevant to the moral evaluation of drunken drivers whose acts cause death and injury. The Separate Duties Challenge was meant to concur with this broad approach to moral appraisal. The second point is that recklessness is supposed not to belong to the ex­plan­ ation of why Fred violates his secondary duty, since Thomas was also reckless, yet Thomas managed not to violate his secondary duty. If so, we should invoke only those explanatory materials which can explain the difference between Fred and Thomas. That way of sharpening the task will leave us with luck, but not recklessness. This point is unsatisfactory, because it over-­generates implications. Both Fred and Thomas were drunk. Thomas did not run anyone over, but Fred did. So does it now follow that Fred’s drunkenness makes no contribution to the ex­plan­ ation of why he violates his secondary duty? No, of course not! By assumption, Fred would have been able to avoid killing the pedestrian had he not been drunk. So his drunkenness cannot be omitted from any satisfactorily complete ex­plan­ ation of why he kills the pedestrian. Now it is true that Fred’s recklessness did not necessitate the killing of the pedestrian. The pedestrian’s death occurred because Fred was driving while

78  Resultant Luck and the Irrelevance Intuition drunk, and because the pedestrian was in the wrong place at the wrong time. The fact that the pedestrian was in the wrong place at the wrong time is not fully entailed by any facts about Fred.31 Nonetheless, and again, the explanation of the role Fred played in the pedestrian’s death cannot fail to invoke his drunkenness, because the fact that he was drunk explains why, if, and when the pedestrian appears, he is unable to prevent a deadly collision with her. Thomas was also drunk, but managed not to violate his secondary duty. Should we say that this fact manages to expel drunkenness from the explanation of why, having recklessly decided to drive while drunk, Fred runs over and kills the pedestrian? No, it does not. What holds for the explanatory relevance of drunkenness also holds for the explanatory relevance of recklessness. The point that Thomas was also reckless, though he managed not to violate his secondary duty, does not displace the role of recklessness from the explanatory picture which applies to Fred and the unfortunate pedestrian.

2.4.3.  The Separate Duties Challenge and Assassins The problems for the Separate Duties Challenge assume an even sharper form for Assassins. This is because the secondary duty in this case seems close to incoherent: though it may be a matter of luck whether Imogen’s and Phoebe’s attempts are successful, there is by assumption nothing they can do after they have squeezed the trigger. It follows that there is no further course of action to which the secondary duty applies, when the primary duty has already been flouted. If we are going to make the secondary duty non-­empty, it can only be interpreted in a way which relates to the making of the initial attempt. The only way to avoid violating the secondary duty in Assassins is to refrain from making the attempt in the first place. This means that the malicious internal states which explain the violation of the primary duty cannot fail to be relevant to the violation of the secondary duty as well. But the initial attempt was supposed to be the subject-­matter of only the primary duty, not the secondary duty. Of course, one might imagine other types of attempts, in which there would be scope to apply corrective measures to attempts that were already under way. Imagine a variant of Assassins, involving the administering of poison: Imogen and Phoebe both administer slow-­acting poison to the President. If unchecked, their acts will result in the President’s death. Call this case Poison Assassins. Various secondary duties can indeed be enumerated in Poison Assassins. Both Imogen and Phoebe are under duties not to administer poison to the President in the first place, but they are also under additional duties, the performance of which 31  It is partly explained by Fred-­related facts, since being in the wrong place at the wrong time is plainly a relational or extrinsic fact about the proximity of the unlucky pedestrian to Fred’s reckless progress on the roads.

Two Caveats  79 may require them to bear severe costs,32 to ensure that their attempts on the President’s life not be fatal, should they have already violated that prior duty. They may be under an extremely stringent duty to provide an antidote, for example, or to provide medical assistance. These are separate duties. Some of these duties become relevant only when preceding duties have been flouted. If no poison has been administered, then no antidote has to be provided, and no emergency medical help needs to be called in. The secondary duties remain conditional unless the conditions for their activation have been triggered. We can make sense of the existence of these separate duties in Poison Assassins because there are possible concrete courses of action to which the secondary duty might apply. In the original version of Assassins, however, there are no courses of action to which secondary duties might apply. This shows us that the duties which pertain to Assassins do not naturally admit of a division into primary and secondary duties.

2.5.  Two Caveats I close the discussion of the Irrelevance Intuition with two caveats.

2.5.1.  Ignorance and Irrationality There are cases, involving ignorance and irrationality, which appear to challenge my argument. What if agents are acting, or attempting to act, from malicious motives, but, due to ignorance or irrationality, what they do or attempt to do is utterly harmless? These are hard cases. For the moment, my argument has only established the bare principle that differences in blameworthiness in acts that are not relevantly flawed, in respect of huge ignorance or irrationality, and that reflect differences in aim-­sensitive luck, do not constitute any offence against morality. That leaves us with much more to do.33 But this is already a significant development in the battle against the Anti-­Luck Account.

2.5.2.  Non-­Foreseeability Now for a second caveat.34 The Restricted Account is not forced to maintain that, when the Internal Claim has been satisfied, an agent has to be held accountable 32  Some of these issues are explored by Tadros (2012). 33  But see Yaffe (2010), ch. 9, for an instructive discussion. Sartorio (2012), pp. 81–3, discusses a somewhat related range of cases involving deviant causation. 34  Thanks to Jon Robson for making it particularly salient to me.

80  Resultant Luck and the Irrelevance Intuition for each and every harmful consequence of the act he or she has performed. We can imagine unusual cases in which, although performed from objectionable mental states, an agent’s act unleashes extreme and quite unforeseeable levels of harm. Consider, for example, a variant inspired by Judith Thomson’s Day’s End case, with a more morally objectionable preface and a deadlier ending: Mildly Malicious Day’s End: Jason always comes home at 11.00 p.m., and usually goes immediately to bed. This evening, however, he stays up in order to compose a rude and insulting email to Marie, his next door neighbour, about a slammed door in Marie’s house the night before, which had disturbed his sleep. Just as Jason presses the ‘Send’ button on his ageing computer, a circuit closes. By virtue of an extraordinary series of coincidences, unpredictable in advance by anybody, the closing of the circuit causes a release of electricity (a small lightning flash) in Marie’s house next door. Unluckily, Marie is in its path and is killed.35

We can agree that Jason ought not to have written the insulting email to Marie, or attempted to send it. Yet it is intuitively implausible to hold him blameworthy for her death. Though he may have intended to cause her some pique, he certainly did not intend to kill her. Sending the email was a completely unforeseeable route to her death by electrocution. The Restricted Account is not committed to holding Jason to be blameworthy for Marie’s death, even though her death causally ensues from his objectionable act (namely, sending the insulting email). Consider a contrasting case. What if, being in an unusually sensitive or vulnerable mood, Marie had reacted to Jason’s email by taking her own life? This reaction, while not intended by Jason, would actually be more foreseeable relative to the intention on which he had acted, which was to upset her, or disturb her emotional equilibrium. As a result, he might be deemed to be partly blameworthy for it. The degree of blameworthiness should be sensibly proportioned to the outer limits of ordinary foreseeability.36 Now it will plainly be a difficult matter to draw a principled line between harms which might intelligibly, though unexpectedly, arise out of culpable behaviour, for which an agent can be properly blamed,37 and harms which lie entirely beyond the boundary of ordinary human foreseeability, for which it would be improper to blame the agent. Imagine that an unjustified serious physical assault which would normally be non-­deadly, causing only a black eye or a broken jaw, actually causes the victim to 35  See Thomson (1990), p. 229, for the original Day’s End case. (I have also borrowed from her original wording here and there.) 36  See also Otsuka (2009), pp. 380–1, and Sartorio (2012), pp. 77–81. 37  Imagine that an unjustified physical assault which would normally be non-­deadly, causing only a black eye, actually causes the victim to suffer a heart attack and die. It would not be prima facie unjustified to hold this agent liable for manslaughter.

Two Caveats  81 suffer a heart attack and die. There would be no principled barrier to holding the assailant liable for manslaughter. But now imagine another case of improper physical contact: Julia gives Julian a mild shove, leading Julian to fall against Julie, who then stumbles into Judy, who suffers a heart attack and dies. This case seems harder to classify. Though Julia should not, by assumption, have shoved Julian, the outcome consisting in Judy’s death was not foreseeable from her point of view. There are various cases to explore here, and I do not make light of the difficulties in mapping out this sort of territory in a satisfactorily detailed way. The main point I wish to emphasize is that the Restricted Account is not precluded from acknowledging that there is such a line, or from attempting to establish roughly where this line lies. The Restricted Account’s non-­ hostility to the existence of such a line counts in its favour; without such an acknowledgement, the Restricted Account would be much less plausible.

3

Resultant Luck and the Fairness Intuition 3.1.  Tackling the Fairness Intuition In this chapter, I turn to the Fairness Intuition. The initial concerns about Assassins and Drivers reflect the involvement of both the Irrelevance Intuition and the Fairness Intuition. But the Anti-­Luck Account tends to assume that these concerns are stationed together: different degrees of blameworthiness between the agents in Assassins and in Drivers are held both to reflect the influence of irrelevant factors, and also to be unfair. The important question is whether these two complaints constitute an all-­or-­nothing, indivisible package, or whether they can come apart. Even if the charge of unfairness does not straightforwardly collapse into the charge of irrelevance, a relation of dependence may still obtain between them: if the Irrelevance Intuition is defeated, then there may be an absence of material for the Fairness Intuition to be exercised over. I have argued, in Chapter 2, that the Irrelevance Intuition simply cannot be upheld: the unchosen, luck-­reflecting factors which are at play in Assassins and Drivers are morally relevant to blame­ worthi­ness. These factors also help to determine the moral quality of these agents’ wills, which is what anti-­luckists tend to prize. This development now raises the question of whether the Anti-­Luck Account can hold the line by relying on the Fairness Intuition alone. I will argue that the Fairness Intuition cannot, in and by itself, uphold the Anti-­ Luck Account. The argument will unfold as follows. First, I will suggest that fairness, if it taken to prescribe equalized blameworthiness, faces a dilemma, where it either offends against the Relevance Claim, or else incurs the ‘Division Objection’. This dilemma is explained in Section  3.2. Next, in Section  3.3, I argue that the category of interpersonal luck on which the Fairness Intuition relies is troubled by the ‘Fate-­Sharing Problem’, an awareness of which is directed by reflection on Susan Hurley’s ‘Egalitarian Fallacy’. In Section  3.4, I argue, via what I label the ‘Moral Cost Argument’ that the differences in comparative luck between the agents in these cases are rendered acceptable by the liability which agents incur when they act in ways triggering the concerns of the Internal Claim. I make a quick comparison between this argument and luck egalitarianism in Section 3.5, before recapitulating some of this material in Section 3.6.

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0004

A Dilemma for the Fairness Intuition  83

3.2.  A Dilemma for the Fairness Intuition The boldest and most dismissive line that might be taken with the Fairness Intuition is to insist that, because luck-­reflecting differences in outcomes do make a difference to blameworthiness, blameworthiness assignments for agents’ acts are not properly regulated by an independent concern about fairness, any more than our standards of evaluation in other domains are necessarily regulated by concerns about fairness. Perhaps the proper distribution of blameworthiness is no more reflective of, or answerable to, the demands of fairness than the distribution of innate musical talent. Fairness simply has nothing to do with it. But there is a more nuanced line to take, and I would rather pursue that. To that end, I now outline a dilemma for the Fairness Intuition.

3.2.1.  First Horn: Dismissing the Relevance Claim On one horn of it we find the dismissal of the Relevance Claim. To illustrate it, we can return to Drivers. Fred’s and Thomas’s acts turn out differently. Fred kills a pedestrian, and Thomas does not. Since the Fairness Intuition insists that they are equally blameworthy because it would be unfair to hold them unequally blame­ worthy, does it now follow that Fred is not blameworthy for killing a pedestrian? Is he blameworthy only for the common element which he shares with Thomas— that is, driving while drunk? That view should strike us as unsatisfactory. Outcomes matter. We are affected principally by outcomes, and we routinely act either to bring about outcomes, or to prevent them from happening. Outcomes, rather than mental attitudes to the prospect of them, end lives, cause injury, and diminish our well-­being and the well-­being of others. As I see matters, a concern with fairness cannot bounce us out of this fundamental interest in outcomes. By insisting on an equalized degree of blameworthiness, we risk non-­accommodation of blameworthiness for the bad outcome which is attributable to Fred, but not to Thomas. This seems implausible.

3.2.2.  Second Horn: The Division Objection Now perhaps, on second thoughts, we need not deny that Fred is blameworthy for killing the pedestrian. Fred is blameworthy for killing the pedestrian and for driving drunk, whereas Thomas is blameworthy simply for driving drunk. But this proposal is now exposed, on the other horn of the dilemma, to the ‘Division Objection’: how are we to divide the same degree of blameworthiness among a different number of blameworthy items without causing distortions? On my view,

84  Resultant Luck and the Fairness Intuition this objection basically scuppers the appeal of Zimmerman’s distinction between the scope and the degree of responsibility, which we have already encountered.1 As applied to Drivers, Zimmerman’s distinction suggests that, while Fred’s blameworthiness may enjoy greater scope than Thomas’s, they are blameworthy to precisely the same degree. But how, in detail, is this division supposed to work? Both Fred and Thomas are guilty of drunken driving; that fact alone qualifies them as being blameworthy to some degree. The mental states which feed into their activity are morally objectionable. But if there is no denial that Fred is blameworthy for killing a pedestrian, and if the two agents are held to be equally blameworthy, then it appears to follow that the badness of the recklessness embodied in Thomas’s drunken driving must exceed the badness of the recklessness embodied in Fred’s drunken driving: Fred, after all, drives recklessly, and he also kills a pedestrian. If Fred’s killing of the pedestrian collects some positive degree of blameworthiness, then his drunken driving—a separate item in his blameworthiness ledger—must be less blameworthy than Thomas’s drunken driving. This position is perplexing. By appealing to the Determination Claim, I have argued that the badness of Fred’s act is enhanced by the badness of the outcome. The Fairness Intuition, however, appears to contend for precisely the opposite view. If blameworthiness assignments are equalized, the badness of Thomas’s act is indirectly enhanced by the fact that his act does not produce a bad outcome. The relative enhancement of his blameworthiness for driving while drunk, by comparison with Fred, will be required for the assignment to him of a degree of blameworthiness which is equal overall to that of Fred’s. That implication is absurd. It appears, then, that if the appeal to fairness delivers an insistence on equalized blameworthiness, it simply distorts what needs to be accounted for. It either omits the relevance of actual outcomes, thus offending against the Relevance Claim, or it provides an unconvincing division of blameworthiness among the items for which these agents can be blamed, thus inviting the Division Objection. The fundamental question is this: does a concern for fairness obviously recommend equalized blameworthiness? I do not, in fact, think that fairness speaks with a unified voice on this matter. I explain why in the next section.

3.3.  The Fairness Intuition and Fate-­Sharing Luck The following remarks focus, once more, on Drivers, though the points could be illustrated just as easily by Assassins. 1  See Zimmerman (2002), p. 560. See Chapter 4, Sections 4.5 and 4.6, and also Appendix II, for further discussion of Zimmerman’s argument.

The Fairness Intuition and Fate-­S haring Luck  85

3.3.1.  Equal Worlds and Lucky Worlds I am going to distinguish between two worlds, or two types of world. In the Lucky World, Fred is deemed more blameworthy than Thomas, in ways which take account of the fact that the outcome he produces is worse. The Equal World sought by the Anti-­Luck Account deems Fred and Thomas to be equally blame­ worthy. The Lucky World is, of course, the world that would be recommended by the Restricted Account, while the Equal World is the world that would be recommended by the Anti-­Luck Account. In being reassigned from the Lucky World to the Equal World, the inequality between Fred’s blameworthiness and Thomas’s blameworthiness is eliminated. But a certain scruple needs to be added immediately. It does not follow that, in the Equal World, Thomas is more blameworthy than he would be in the Lucky World. Neither does it follow that Fred is less blameworthy than he would be in the Lucky World. What follows is only that, in the Equal World, Thomas is no less blameworthy than Fred, and that Fred is no more blameworthy than Thomas. The Equal World collapses the relative difference in blameworthiness between them, but the elimination of this relative difference does not carry any immediate implications for the absolute level of blameworthiness that applies either to Fred or to Thomas. That is because the Equal World imposes only a constraint on the level of blameworthiness that can be properly extended to Thomas and Fred: neither of them is more or less blameworthy than the other. Consider the following proposal: we should extend a level of blameworthiness to both Fred and Thomas that focuses only on what lies within the control of these agents. That policy will not answer all the questions we need to ask, for there is still uncertainty about which level of blameworthiness should now be extended to each of them. True, Fred lacked control over the fact that a pedestrian appeared on the road before him. But Thomas also lacked control over the non-­ appearance of the pedestrian on the road before him. Whose case should we consult in order to fix blameworthiness assignments? In other words, to accept the advice that Fred and Thomas are blameworthy only for that which falls within their control does not tell us how bad their respective offences were, or even what their offences should encompass. Was Fred’s offence only as bad as Thomas’s? Perhaps Fred’s offence was less bad than we tend to suppose. Or was Thomas’s offence, in fact, every bit as bad as Fred’s? Perhaps his offence was worse than we tend to suppose. Again, whose case should we be consulting in order to fix the appropriate level of blameworthiness?2

2  For a searching discussion of these issues, see Wolf (2001). Wolf explores somewhat similar worries about the indeterminacy of the levels of blame in the anti-­luckist picture, though the precise lessons I am trying to recover differ from hers.

86  Resultant Luck and the Fairness Intuition To further consider these possibilities, I distinguish between three central candidates for blameworthiness assignments in the Equal World.3 First, there is maximal blameworthiness if the blameworthiness level for both Fred and Thomas in the Equal World is tied to the overall level of blame­worthi­ ness which is merited in the Lucky World by what Fred does in the Lucky World (that is, act recklessly and kill a pedestrian). Second, there is minimal blameworthiness if the blameworthiness level for both Fred and Thomas in the Equal World is tied to the overall level of blame­ worthi­ness which would be merited in the Lucky World by what Thomas does in the Lucky World (that is, act recklessly and kill nobody). Third, there is middling blameworthiness if the blameworthiness level for both Thomas and Fred in the Equal World falls somewhere between maximal and minimal blameworthiness, as these have already been defined. These possibilities, with some arbitrarily chosen cardinal numbers that respect the ordinal magnitudes of blameworthiness, are represented below, in Figure 3.1:  

Lucky World

Equal World (Maximal) 

Equal World (Middling)

Equal World (Minimal)

Fred’s blameworthiness  Thomas’s blameworthiness 

10 4

10 10

7 7

4 4

Figure 3.1  Blameworthiness Assignments in the Equal World and Lucky World

These values for the Equal World are chosen on the assumption that the equalized assignments will fall within the range of blameworthiness assignments in the Lucky World. Thus, the equalized assignments fall somewhere between 4 and 10. In the Minimal Equal World, Fred is less blameworthy than he would be in the Lucky World. In the Maximal Equal World, by contrast, Fred is every bit as blameworthy as he would be in the Lucky World. Here it is Thomas who is more blameworthy than he would be in the Lucky World. In the Middling Equal World, as it has been defined here, Thomas is more blameworthy than he would be in the Lucky World, and Fred is less blameworthy than he would be in the Lucky World. There will be a large series of other Equal Worlds as well, with the blame­ worthi­ness values falling somewhere between 4 and 10. In each of them, Fred will be less blameworthy than he is in the Lucky World, while Thomas will be more blameworthy. We could also populate the Equal World with other numerical values that fall outside the range of blameworthiness assignments established by this version of the Lucky World: perhaps the Super-­Maximal Equal World will assign 3 These options are not exhaustive, since super-­maximal and ultra-­minimal variants could be sketched. I briefly revisit such possibilities later on in this sub-­section.

The Fairness Intuition and Fate-­S haring Luck  87 blameworthiness to degree 50 for both Fred and Thomas, while the Ultra-­ Minimal Equal World will assign blameworthiness to degree 1 to both Fred and Thomas. Finally, of course, the Lucky World could also be calibrated differently. The particular calibrations here have been chosen simply for illustrative purposes. The main lesson to be recovered is that the replacement of the Lucky World with the Equal World does not offer much, if any, immediate guidance about the changing individual blameworthiness profiles of both Fred and Thomas. A further lesson is that we will usually be better off referring, in the plural version, to Equal Worlds, rather than the Equal World. Similarly, it will be sensible to refer to Lucky Worlds, rather than the Lucky World. One worry which is likely to arise at this point, accordingly, is that it is simply arbitrary which Equal World is favoured by the Anti-­Luck Account. But I do not want to make too much of this worry, and it will not be investigated any further. Defenders of the Anti-­Luck Account may concede that such potential arbitrariness renders their account incomplete, but this concession would not establish that their account cannot be completed. Further considerations may help to achieve full determinacy of blameworthiness. (Perhaps anti-­luckists might be guided, in assigning blameworthiness to tokens of a particular act-­type, by expected levels of harm, or the expected benefits of deterrence.4) I want instead to point to another problem—a deeper and less easily noticed one, concerning fairness.

3.3.2.  Fate-­Sharing in Equal Worlds In any of the Equal Worlds, the gap in blameworthiness between Fred and Thomas is eliminated. The interpersonal luck which would be manifested in one of the Lucky Worlds is therefore neutralized. The equalization of blameworthiness in this range of worlds should serve to assuage the following complaint from Fred: Fred’s Complaint: Whatever we should say about the relevance of outcomes, it is simply not fair that I should be deemed more blameworthy than Thomas. The difference between us, as everyone concedes, is due to factors beyond my control. He cannot take any credit for these differences, and they are not to my discredit. Again, it is difficult to see why there should be any differences in our respective levels of discredit or blameworthiness, given these unchosen differences between us. So it is reasonable for me to insist that I should not be deemed any more blameworthy than he is.

4  I return to this idea below, at the end of Section 3.3.2.

88  Resultant Luck and the Fairness Intuition But the equalized blameworthiness assignments which belong to an Equal World might invite, in turn, the following complaint from Thomas: Thomas’s Complaint: True, in any one of the Lucky Worlds, Fred will be more blameworthy than I am, and this is due, I concede, to unchosen differences between us. But the existence of these unchosen differences gives rise to more than one puzzle. In an Equal World, it is guaranteed that I am no less blame­ worthy than Fred is, despite the fact that there are these unchosen differences between us. Why should I be blamed as much as he is, given these unchosen differences? Why should my blameworthiness assignment be tied to his?5

It is easy enough to suspect an element of special pleading in Thomas’s complaint. Even so, I think he has managed to put his finger on something significant. Another form of luck is about to make its appearance. As we have seen, if the Fairness Intuition is to have any chance of enjoying an independent life from the Irrelevance Intuition, then we must treat Fred’s and Thomas’s claims as comparative claims, rather than non-­comparative claims.6 Comparative fairness demands that Fred should be blameworthy to no greater degree than Thomas, and that Thomas should be blameworthy to no lesser degree than Fred. What matters to comparative fairness is that there should be no differences in the degree to which these two agents are blameworthy. Now we have seen that facts about relative blameworthiness do not stand in any obvious relationship with facts about absolute levels of blameworthiness: if Thomas is as blameworthy as Fred in an Equal World, it does not necessarily follow that he will be more blameworthy than he would have been in a Lucky World. It all depends. When the inequality in their blameworthiness in the Lucky World is neutralized, then interpersonal luck is dealt with. But it cannot be denied that there is still a definite source of lucky impact on the level of blameworthiness assigned to both Fred and Thomas: each of them is blameworthy for what he did, but in ways that are tied to the blameworthiness of what the other did. And that is puzzling, because the relation between what Fred does and what Thomas does also lies beyond the control of either of them. As we know, Fred lacks full control over the outcome of his act. The outcome of his reckless act is due to various contingencies of the world over which he lacks full control. We assume that he enjoys control, in the relevant sense, over his initial decision, though not over the outcome of his act. But Fred also lacks control over the outcome of Thomas’s act. Indeed, he arguably has less control over the outcome of Thomas’s act, since he lacks control over any part of its history. Fred has no control over whether Thomas acted 5 As we shall see, an echo of this sort of reasoning will reappear when I come to explore Zimmerman’s anti-­luckism: see Chapter 4, Section 4.6. 6  This point was laboured in Chapter 1, Section 1.2.

The Fairness Intuition and Fate-­S haring Luck  89 culp­ably in the first place: this was Thomas’s decision to make, not his. His lack of control over how things turn out for Thomas therefore goes deeper than his lack of control over how things turn out for him. Exactly the same goes for Thomas. Like Fred, he lacks control over the outcome of his own act, due to various contingencies over which he lacks full control. But he also lacks control over the outcome of Fred’s act. Indeed, once more, Thomas arguably has less control over Fred’s act than his own, since he lacks any sort of control over its history. More specifically, he lacks control over whether Fred acted culpably in the first place. He has no control over the making of that initial decision, or over Fred’s acting on his decision. When luck is equalized, Fred will be blameworthy for what he does, in ways which are now sensitive to how blameworthy Thomas will be for what he does. Similarly, Thomas will be blameworthy for what he does, in ways which are now sensitive to how blame­ worthy Fred is going to be for what he does. So there is a sense in which these two agents are forced to share each other’s fate.7 Fred’s level of blameworthiness, whatever it turns out to be, is guaranteed to be the same as Thomas’s level of blame­ worthi­ness. But the fact that they are forced to share each other’s fate by being blamed to an equal degree is enough to confirm that this model of fairness is luck-­reflecting. To eliminate the interpersonal luck that is the particular concern of fairness, Fred is no more blameworthy than Thomas, and Thomas is no less blameworthy than Fred. But the equality in blameworthiness between them should not distract us from the fact that Fred’s level of blameworthiness is tied to Thomas’s level of blameworthiness, and that Thomas’s level of blameworthiness is tied to Fred’s level of blameworthiness. We can call this form of luck interpersonal fate-­sharing luck, or fate-­sharing luck for short. To remove the taint of unfairness, according to the Anti-­Luck Account, Thomas should be held to be as blameworthy as Fred. So, however much or little Fred is deemed blameworthy, Thomas will be just as blameworthy. Similarly, however much or little Thomas is deemed blameworthy, Fred will be no more blameworthy than that.

3.3.3.  Is Fate-­Sharing Luck Substantive? This argument invites the following response: there is no substantive form of fate-­ sharing luck that is enjoined by equalized blameworthiness assignments, because the coincidence between Thomas’s level of blameworthiness and Fred’s level of blameworthiness reflects only the deeper truth that, however the exercise in reckless drunken driving turns out, their levels of blameworthiness should be the 7  The echo here is a deliberately Rawlsian one: see Rawls (1999a), p. 88. Fate-­sharing can, I believe, be given a more benign interpretation in theorizing about social justice. See Chapter 7.

90  Resultant Luck and the Fairness Intuition same. As a result, there is nothing special about the actual outcome of Fred’s reckless act, and so Thomas’s fate is not being tied to that outcome, in particular. Similarly, there is nothing special about the actual outcome of Thomas’s reckless act, and so Fred’s fate is not being tied to that particular outcome. However the journey turned out, Fred would be just as blameworthy as Thomas. Again, the insistence that interpersonal luck be eliminated carries no automatic implications for how individually blameworthy Fred and Thomas are deemed to be. This response shows that the worry about fate-­sharing luck is chimerical. Fate-­sharing luck is not a substantive form of luck. It merely imposes a structural constraint on assignments of individual substantive blameworthiness. So it is not an embarrassment for the Anti-­Luck Account. This response proves too much. There are two points to make. First, if it is the Fairness Intuition which is meant to be calling the shots here, then the form of luck whose neutralization is being sought simply has to be interpersonal luck. Even if it is only Thomas who drinks and drives, rather than Thomas and Fred, the reason for setting the blameworthiness assignment where it is will be to ensure that, if Fred did drink and drive, then he would be no more blameworthy than Thomas, regardless of what transpired from his recklessness. To deny that point would be simply to obviate any appeal to fairness. It would also risk the admission that blameworthiness is determined only by quality of will, and not outcomes. That admission would be unwise, since it would offend against the Relevance Claim. Second, even if it is not the actual outcome of Fred’s reckless act in this particular case which strongly determines Thomas’s level of blameworthiness, the level of blameworthiness Thomas incurs is still likely to be sensitive to the averaged-­out data of harm and injury relating to drunken car journeys performed by different agents. To ensure that the blameworthiness assignment is non-­ arbitrary, it needs to bear an intelligible relationship to all the tokens of that same act-­type that have transpired, or are likely to transpire in the future. Why is drunken driving a significant offence? It is because drunken driving is apt to lead to avoidable injuries and fatalities. How, then, do we determine how blameworthy an individual act of drunken driving is to be? However that question is going to be answered, it seems unlikely that we will be able to avoid consulting the general track record of drunken car journeys and the fatalities which ensue from them. That track record will surely help to determine the scale of blameworthiness for an individual drunken car journey, regardless of whether that particular token ­journey is harmful. In this way, then, Thomas’s level of blameworthiness is obliquely fixed by what happens to everyone else. This must still be a form of luck. How could it fail to be?

The Fairness Intuition and Fate-­S haring Luck  91

3.3.4.  Two Models of Fairness With these points in mind, we can compare two models of fairness. This is the first of them: Actual Outcome Model: For any agent, X, X is assigned a fair degree of blame­ worthi­ness if and only if (a) X is not deemed blameworthy unless she acts culp­ ably, (b) X is in a position to avoid acting culpably, and (c) X’s level of blameworthiness is (roughly) proportioned to the badness of the actual outcome.8

And this is the second model: Equal Blameworthiness Model: For any agent, X, X is assigned a fair degree of blameworthiness if and only if (a) X is not deemed blameworthy unless she acts culpably, (b) X is in a position to avoid acting culpably, and (c*) X’s level of blameworthiness is tied to the same level of blameworthiness that would be assigned to anyone else who acted culpably in respect of the same offence in the same circumstances.

Notice that two of these claims, (a) and (b), are common property to both the Actual Outcome Model and the Equal Blameworthiness Model. Blameworthiness will not be fairly assigned unless the act was culpable, and unless the culpable act was avoidable. The two models differ only with respect to the third claim: the Actual Outcome Model embraces (c), whereas the Equal Blameworthiness Model accepts (c*). So the question now is whether (c) is preferable to (c*). I am not claiming that the Actual Outcome Model states a credible account of fairness merely because it establishes common ground rules for the treatment of Thomas and Fred. We need more than simply generality, or the existence of common ground rules, to arrive at a credible substantive theory of fairness. Fairness has to amount to something more than simply susceptibility to a common description.9 What needs to be added, more specifically, to the generality enshrined by (c) is a concern about the distribution of luck. Under the Actual Outcome Model, Thomas and Fred are assigned a level of blameworthiness which is permitted to vary along with the level of harm each of them produces. The Actual Outcome Model will of course introduce potential discrepancies between 8  The caveat here was rehearsed in Chapter 2, Section 2.6.2. 9 Raz (1986), pp. 218–21, makes a similar point about the distinction between generality and equality.

92  Resultant Luck and the Fairness Intuition them, due to interpersonal luck. However, it will also protect each of them against the fate-­sharing luck which ensues from the equalization of blameworthiness. Even as a theory of fairness, the Actual Outcome Model is not notably inferior to the Equal Blameworthiness Model. When we add the weaknesses of the Irrelevance Intuition to the picture, the case for its adoption seems to me to be compelling.

3.3.5.  The Egalitarian Fallacy There is more to say about the source of fate-­sharing luck, and this will also help to prepare us for the trajectory of argument explored in the chapters in Part Two of this book, when we will be concerned with distributive justice. The truth that equalized blameworthiness may reflect luck is made by Susan Hurley, in her unmasking of what she calls the ‘Egalitarian Fallacy’: The fact that people are not responsible for difference does not entail that they are responsible for non-­difference.10

Hurley’s immediate concern is with distributive justice, and thus with responsibility, or non-­responsibility, for benefits that are encompassed by the metric of distributive justice. We shall revisit it in Chapter 6. But the point she is making also enjoys an application to the assignment of blameworthiness: differences in assigned blameworthiness between Fred and Thomas are no more revealing of non-­responsibility, or luck, than non-­differences in blameworthiness assignments between them are revealing of non-­responsibility, or luck. Neither Fred nor Thomas is responsible for the difference between their blameworthiness assignments in Lucky Worlds. This difference reflects interpersonal luck. But the same goes for any patterned distribution between Fred’s and Thomas’s levels of blame­ worthi­ness. Equal Worlds erase the differences between their blameworthiness assignments, thus establishing parity between them, but the fact that their blame­ worthi­ness assignments are equal is also a matter of luck. This non-­difference is due to fate-­sharing luck.

3.4.  Higher-­Order Luck and the Mechanism of Liability In the story we have been telling so far about the Fairness Intuition, fairness under-­performs in its envisaged role as saviour for the Anti-­Luck Account. If 10  Hurley (2001), pp. 56–7; original emphases. See also the discussion in Hurley (2003), pp. 151ff.

Higher-­O rder Luck and the Mechanism of Liability  93 we settle for guidance by the Fairness Intuition alone, we risk either an inability to accommodate the relevance of outcomes, or we risk sponsoring an unconvincing division of labour among the items for which agents can be held responsible. Moreover, even in a world in which interpersonal luck is elim­in­ ated and blame­worthi­ness is equalized, there is no elimination of luck sim­plici­ ter. As soon as interpersonal luck is expunged, there is an eruption of fate-­sharing luck. Luck is ubiquitous, and the remedies sought by the Anti-­Luck Account cannot fully contain it. The appropriate conclusion to draw is that an appeal to fairness cannot underwrite any adjustment in the Restricted Account’s verdicts. Though the verdicts advanced by the Restricted Account support different assignments of blameworthiness to agents whose differences are due only to luck, the appeal to fairness is insufficient to close the gap between them. There is more to say, however, about why these agents have nothing to complain about if they are deemed unequally blameworthy. We can begin to see what that further story contains if we contrast Two Buttons with a more benign version of it. I am about to argue for a surprising asymmetry.

3.4.1.  Two Better Buttons Outcomes matter, and they matter to how we should appraise the agents who prod­uce them. That is what the Relevance Claim teaches us. So Fred is more blame­worthy than Thomas, because the outcome he causes is worse than the outcome Thomas causes. These blameworthiness cases can be called ‘Bad Cases’. Bad Cases contrast with ‘Good Cases’, in which good outcomes are caused by agents acting in well-­ intentioned ways, and which admit of different degrees of praiseworthiness. It would be natural to assume that Bad Cases and Good Cases are entirely symmetrical: just as differences in bad outcomes warrant differences in blame­worthi­ ness in Bad Cases, so differences in good outcomes warrant differences in praiseworthiness in Good Cases. In fact, I am reluctant to embrace this sym­metry, and am going to argue for a partial asymmetry between Bad Cases and Good Cases. In some Good Cases, differences in the goodness of the outcomes p ­ rod­uced by human action will indeed be partnered with differences in the p ­ raiseworthiness of the agents who produce these outcomes. But this will not be true of all Good Cases. In some of them, differences in outcomes will not be accompanied by differences in praiseworthiness. The difference is due, at bottom, to the interaction between the Relevance Claim and the Denial Claim in Bad Cases, and to the fact that Good Cases are free of this interaction: nothing plays the particular role of the Denial Claim in Good Cases.

94  Resultant Luck and the Fairness Intuition To help us think about these ideas, consider a case which is structurally parallel to Two Buttons: Two Better Buttons: There are two agents, Adina and Kristin, and there is a benign machine, with two buttons. Pressing Button A will ensure that ten innocent strangers are saved from death. Pressing Button B will ensure that an eleventh innocent stranger is saved from a severe injury: a broken arm. The buttons do not carry this precise information. Both Button A and Button B are adorned with the same sign, in large lettering: ‘OTHER PEOPLE, PERHAPS SEVERAL OF THEM, WILL BE SAVED FROM HARM IF YOU PRESS THIS BUTTON’. Both Adina and Kristin are disposed to press one of these buttons, because each of them has the beneficent aim of aiding others. Adina presses Button A, and Kristin presses Button B.

In Two Better Buttons, Adina produces a better outcome than Kristin.11 Like the agents in Two Buttons, the fact that she does so is just a matter of luck. She has no way of knowing that Button A produces a better outcome than Button B. The same goes for Kristin: though no less well-­intentioned than Adina, she had no way of knowing that pressing Button B would produce an outcome which is less good than the outcome caused by pressing Button  A.  Does Adina qualify for greater praise than Kristin? If they are both acting under the duty to prevent harm at no cost to themselves, does Adina’s act constitute superior performance of that duty than Kristin’s act? Can Kristin be criticized for not pressing the button which would have produced better consequences?12 Certainly, Kristin is not open to criticism for not pressing Button A. For related reasons, I am also hesitant to embrace the view that Adina is more praiseworthy than Kristin. In this case, the Relevance Claim interacts differently with the other claims that were yielded by reflection on Two Buttons. In Two Buttons, the Denial Claim denied Dylan a defence against his act of pressing Button 1. Dylan cannot appeal to Ought Implies Can as a means of deflecting his putatively greater blameworthiness, if he has already defied morality by pressing one of these buttons in the first place. For Dylan, there is no need to accommodate Ought Implies Can at the expense of the Relevance Claim. In Two Better Buttons, by contrast, the Denial Claim—or some other principle that would satisfy the role played by the Denial Claim—would have to work

11 This claim might inspire a complaint from anti-­aggregation Taurekian quarters: see Taurek (1977). But the claim will be safe enough for most people. See Lang and Lawlor (2016) for an attempt to tackle Taurek at source; our view is that Taurek’s own commitment to fairness demands his adoption of an interpersonal perspective which will, in turn, be consistent with the existence of morally credible and robust betterness judgements about states of affairs. 12  Imagine that Button A can be pressed twice, as long as it is pressed by separate agents. If it is pressed twice, then it will produce twice the amount of good that ensues from pressing it just once.

Higher-­O rder Luck and the Mechanism of Liability  95 differently. If Kristin was acting on a duty of beneficence, it does not make sense to say that she should settle for her lesser praiseworthiness, or even her openness to some criticism, by reflecting on Ought Implies Can. It is unhelpful to tell her that she ought to have satisfied her duty of beneficence by pressing Button A rather than Button B. By assumption, she had no way of knowing which of them would be better. Contrast, again, Dylan: Dylan did not know what the consequences of pressing either Button 1 or Button 2 in Two Buttons would be, but his duty not to press either of them was much more easily satisfied. He simply had to refrain from pressing either button. The same sort of hesitation should attach, I believe, to the Determination Claim. I am less convinced that there is a satisfactory case for externalizing Adina’s and Kristin’s internal states than for externalizing Dylan’s and Akira’s internal states. Adina is not obviously more praiseworthy than Kristin, and the internal states embodied in her act are not obviously better because of the greater good which ensued from pressing Button A, rather than Button B. If there is an asymmetry between the agents in Good Cases and the agents in Bad Cases, what could it consist in? What explains it? We need to locate a deeper story.

3.4.2.  The Costs of Culpability: The Moral Cost Argument The picture we have assembled so far can be stated in terms of the following argument, which I will call the Moral Cost Argument. The bulk of it is outlined here, but it will be eventually completed, after some unavoidable twists and turns, in Section 3.4.4. The argument will now be slowly outlined, with supporting commentary where required. We begin with these two claims, which simply recapitulate the central commitments of the Restricted Account: (A) Agents in Bad Cases are culpable if they act from internal states which are in some way motivationally or attitudinally flawed: they may embody malicious aims, or betray recklessness or negligence. (B) Agents’ final degree of blameworthiness in Bad Cases is established roughly in proportion to the actual outcomes of acts embodying culpable in­tern­al states.13

13  Strictly speaking, there is a small amount of redundancy here: the case will not count as a Bad Case unless the agents’ internal states are culpable, and agents’ possession of culpable internal states will then be sufficient to qualify their case as a Bad Case. These redundancies will, accordingly, be excised from the following discussion.

96  Resultant Luck and the Fairness Intuition Claim (A) makes an agent a candidate for blame in the first place, by providing a test for her culpability, and supplies a separate reason for making that agent blameworthy in virtue of her culpability. What (A) yields, in effect, is a ‘base rate’ of blameworthiness. But claim (B), which simply restates the External Claim, also permits the actual degree of blameworthiness to be enlarged beyond the base rate, in response to how acts turn out, and it permits different agents to differ in respect of their overall degree of blameworthiness, depending on the same facts. I should hope that it is clear by now that it is not inconsistent to combine (A) with (B). But the need for additional commentary on what their combination delivers is invited by the intuitive differences that we have noted between Good Cases and Bad Cases. To help to secure that explanation, my suggestion is that we supplement (A) and (B) with the following claim: (C) In acting from culpable inner states, agents lose their antecedent entitlement to be assigned a level of blameworthiness which would neutralize comparative luck.

It follows from (C) that (D) For any pair of agents, X and Y, who are in a Bad Case, and whose acts embody relevantly similar-­looking internal states,14 X and Y lose their ante­cedent entitlement to have their interpersonal luck neutralized, and thus lose their entitlement to be deemed equally blameworthy.

But why embrace (C) and, by implication, (D)? It seems to me that a suitable grounding for them is given by: (E) The moral cost that X and Y should pay for acting from culpable internal states—a cost which is due to the fact that acting from culpable inner states already constitutes an offence against morality—is that X and Y are denied, in the calibration of their blameworthiness, comparative justice, which would be achieved by the neutralization of their interpersonal luck.

We need to say more about (E). I have repeatedly stated in Chapters 1 and 2 that the Restricted Account does not collapse into the Strict Liability Account: consequences are not the only things that matter. Internal states can be morally corrupt, or flawed, even if they do not result in any actual harm. For this reason, Thomas and Phoebe can, and should, be deemed blameworthy, though their acts are not in 14  According to the Determination Claim, the moral quality of internal states will of course be different if the acts in which they are embodied turn out differently. Here I am emphasizing the descriptive similarity between those states before that information is added. These types of similarity and difference can be established without taking the actual consequences of these acts into account.

Higher-­O rder Luck and the Mechanism of Liability  97 fact harmful. It follows that all four of the agents in Drivers and Assassins are blameworthy. There is one further feature which Thomas and Phoebe share with their partnered agents in Drivers and Assassins: in acting in these ways, they are liable to be blamed for whatever actual bad consequences occur.15 Their liability to be blamed for the actual bad consequences of what they do then carries the further consequence that they cannot rely on comparative justice. I return to this idea after briefly considering a broadly friendly rival to it, proposed by Troy Jollimore.

3.4.3.  Deserving to Deserve? In an intriguing blog post, Troy Jollimore has suggested that two agents—let us call them Betty and Fergus—who have acted wrongly . . . are equal with respect to whether they deserve to deserve blame. They are equal, that is, with respect to second-­order desert: the matter of whether one deserves to be (and not simply whether one is) blameworthy.16

Jollimore’s proposal allows him to reconcile the intuition that Betty might be deserving of blame in a way that Fergus is not with the intuition that both Betty and Fergus are equally deserving from a moral point of view. The reconciliation is effected by the switch from the first-­order level to the second-­order level. Betty and Fergus can be regarded as being equally morally deserving at a second-­order level, rather than a first-­order level: at a second-­order level, they are equally morally deserving of different assignments of blameworthiness at the first-­order level, depending on how their acts turn out. Jollimore’s argument is suggestive, and contains insight. As it stands, however, it also contains at least two significant weaknesses.17 First, the apparatus of levels is important to Jollimore because it allows him to retain the intuition that Betty and Fergus are, in a significant sense, morally equal, despite being blameworthy to different degrees. But the moral equality between Betty and Fergus enshrined by this proposal risks being chimerical. Consider the following case: Restricted Franchise: The State permits Fergus, but not Betty, to cast a vote in democratic elections.

15  Subject to the caveats noted in Chapter 2, Section 2.5. 16  Jollimore (2004); original emphasis. 17  The second of these criticisms was raised in anonymous (or at least semi-­anonymous) replies to Jollimore’s blog post on PEA Soup.

98  Resultant Luck and the Fairness Intuition Now the State cannot parry the complaint that Betty and Fergus are plainly pol­it­ ic­al­ly unequal in Restricted Franchise by ascending to a second-­order level at which they are nonetheless declared equal. The State could not credibly pretend that Betty and Fergus are still politically equal by suggesting that they are equal at the second-­order level, where they can be regarded as equal at this second-­order level in the sense that, at the first-­order level, each of them may or may not be denied a vote. There is simply no content to this notion of second-­order equality: there is only inequality between Betty and Fergus at the first-­order level, with no substantive higher-­order equality between them to offset or justify that first-­order inequality. The equality enshrined at the second-­order level simply reports a conjunctive claim ranging over Betty’s and Fergus’s disjunctive prospects, according to which it is true of both Betty and Fergus that they may, or may not be, granted permission to vote in democratic elections. As a proposal for defending the arrangements in Restricted Franchise, this is just sophistry. Why should we consider Jollimore’s argument about resultant luck to be any less vulnerable to such a charge? Second, the ‘deserving to deserve’ locution is problematic. To see why, we can apply Jollimore’s suggestion to the agents in Drivers. If Fred and Thomas are equally deserving of deserving blame, then this equality in respect of their second-­order desert can be satisfied either by making them blameworthy to a differential degree at the first-­order level, or by not making them blameworthy at all. Why should we favour the first over the second way of looking at things? Jollimore’s proposal fails, then, but it does yield a useful checklist for taking the argument further. First, we need to construct a more illuminating pathway between what Fred and Thomas have in common, and the quotient of blame­ worthi­ness deserved by each of them in turn. Second, we must surrender the aim of squaring circles. There is no interesting substantive sense in which the pairs of agents in Bad Cases are equal, and we should abandon the ambition to demonstrate otherwise. Third, we should also de-­emphasize the role of desert, which introduces unnecessary complications. Though I think he has misidentified what it is, Jollimore is nonetheless correct to suggest that there is some common feature, connecting all these agents, which explains why we are entitled to reach the verdicts we do about their differential blameworthiness. As we are about to see, he is also correct to hold that the div­ ision of levels has something to teach us about why these agents lack a complaint about the influence or distribution of resultant luck.

3.4.4.  Liability and Comparative Justice: Completing the Moral Cost Argument As I see it, the further feature which these agents all share is that they are all liable to be blamed to different extents, depending on how their acts turn out. I shall

Higher-­O rder Luck and the Mechanism of Liability  99 call this consequential liability. In finding them blameworthy in proportion to their consequential liability, we do not wrong these agents.18 This is simply the price they pay for acting wrongly. Due to their consequential liability, they are not entitled to the neutralization of interpersonal luck.19 This is the key to understanding (E). The concept of liability may appear to be, and has often been charged as being, explanatorily meagre. After all, liability-­talk is often treated as freely exchangeable with forfeiture-­talk,20 and a central and traditional accusation against the notion of forfeiture has been that it does not explain why an agent, Noah, who has supposedly forfeited his right against T is not wronged by then being subjected to T.21 The claim that Noah is not wronged by being subjected to T is not properly explained by the claim that Noah has forfeited his claim against being subjected to T.  The two claims are just variants of each other, couched in slightly different terms. The forfeiture-­containing claim does not offer any explanation of the claim about Noah’s being wronged. What holds for forfeiture appears to hold for liability: similarly, then, the claim that Noah is liable to being subjected to T does not explain why Noah would not be wronged by being subjected to T.22 It seems to me that the charge of explanatory vacuity is unfair. Take a self-­ defensive situation in which Noah launches a potentially harmful attack against Nico, who is innocent. Most of us think that Nico is entitled to take proportionate self-­defensive action, if there is no other way in which he can repel the threat posed by Noah. (The permissions usually also extend to other-­defensive action, but I will not address the details of this extension here.23) Noah would normally be morally immune to such violence, but in this case he has lost this immunity: Noah is liable to be harmed, in virtue of the fact that he threatens Nico, who was 18  Cf. McMahan (2009), p. 10. 19  Herdova and Kearns (2015), pp. 364–5, suggest that the common moral valence of agents in Bad Cases—the fact that these agents are all blameworthy, however their acts turn out—imposes some intellectual pressure to adopt an anti-­luckist position, through getting us to reflect more deeply on what these agents have in common, rather than what makes them different from each other. I propose to exploit the common moral valence of these agents in a different way. As I see matters, the common valence of these agents is evidence for the existence of a mechanism of liability which can then deliver non-­equalized blameworthiness at the first-­order level. In other words, I interpret the common valence as an invitation to sustain these agential differences, not to collapse them in a further round of theorizing. 20  See, for example, the practice adopted in McMahan (2009). 21  For a fuller treatment of the history of complaints against forfeiture, see Lang (2014b). I argue there that the notion of forfeiture is, in fact, perfectly defensible against many of the complaints that have been advanced against it, but I acknowledge that the central charge of explanatory vacuity is a trickier one to dispel (2014b, pp. 51–5). In Lang (2017), I contend for a more complete adjustment in focus between the attacker’s liability and the victim’s non-­liability by arguing for a ‘Non-­Liability First Account’ of defence. 22  True, it is, specifically, forfeiture, rather than liability, which has borne the brunt of these attacks. That may be, in part, because the category of liability has tended, in recent debates, to be invoked by those who offer a broadly distributive theory of harm, to which many writers in the current debate are broadly sympathetic, whereas the category of forfeiture has been usually associated with rights-­based theories of self-­defence, such as that offered by Thomson (1991), and about which many writers are (wrongly, I think) sceptical. 23  The extension is complicated, however, and we should not rush to conclusions: see Parry (2017).

100  Resultant Luck and the Fairness Intuition not liable to be harmed. This larger picture does have explanatory force, if we relate Noah’s liability to the defensive measures which we think Nico should enjoy. We do not have to invoke Noah’s liability as the single explanatory key which unlocks the entire problem. Noah’s liability ensues from facts which relate to Nico’s non-­liability, and with the measures which Nico can take in order to protect himself against the violation of his right.24 Thus understood, Noah’s liability can be an element in a larger picture which can claim real explanatory force. Returning to the agents in our central Bad Cases—Drivers, Assassins, and Two Buttons—we can note that they have all acted wrongfully, and are therefore all liable. Liable to what? We are not dealing here with a species of defensive liability. They are liable, simply, to blame for the results of what they do. They are consequentially liable. How blameworthy they are will now depend, in part, on how their acts actually turned out. To make sense of the relationships between these various pairs of agents, it is helpful to invoke a Jollimore-­style division of levels. At a second-­order level, these agents have acted wrongfully, and are therefore liable to be blamed for whatever eventuates from their acts.25 At the first-­order level, they are blameworthy to different degrees, depending on how their acts turned out. The second-­order level provides us with a common normative explanation of why the differences at the first-­order level cannot sustain complaints from any of these agents about differential blameworthiness, but it does not pretend to establish any substantive form of equality between them which can offset the various inequalities at the first-­order level. That is not the function of the second-­order level. The function of the second-­order level is, rather, to reassure us that these agents have not been treated unjustly, though their actual treatment may indeed be unequal. They lack a complaint about differential blameworthiness, and about the suspension of the comparative fairness which they could otherwise rely upon, due to a truth which unites them: they have acted wrongly. Thus the final step in the Moral Cost Argument can be stated as follows: (F) The moral cost that X and Y pay for acting from culpable internal states—a cost which consists in their non-­immunity to interpersonal luck—is due to their liability to being blamed for the results of what they do: their consequential liability.

As we have seen, in Good Cases, such as Two Better Buttons, agents can rely on evaluation which is roughly proportioned to the quality of will which is em­bodied in their acts, over which they are assumed to have control. That is why Adina and Kristin in Two Better Buttons are not strongly qualified, as I see it, for differential 24  Again, see Lang (2014b), pp. 55–9; this picture has been augmented in Lang (2017). 25  Subject, once more, to the foreseeability condition: see Chapter 2, Section 2.5.2.

Higher-­O rder Luck and the Mechanism of Liability  101 amounts of praiseworthiness. This does not mean that agents in Good Cases should be indifferent to what they achieve in the world. Their aim should be to make the world go better, not just to try to make the world go better.26 It is not as though an agent should be indifferent to the difference between a successful attempt and an unsuccessful attempt. Even so, we may not think that differential praise is the appropriate response to agents in Good Cases whose acts turn out differently. Such agents can be protected from these kinds of contingencies. In Bad Cases, by contrast, when agents are acting from culpable inner states, they can no longer count on the neutralization of comparative luck. For these agents, this kind of comparative justice has been suspended. The agents in Two Buttons, Drivers, and Assassins are denied the luck-­insensitive appraisals, and therefore the comparative justice, which those who act non-­maliciously can normally count upon. Each of these agents loses the antecedent protection of an assignment of blameworthiness which expels the influence of any luck, and therefore loses his or her right to be blamed to the same degree as any other agent whose inner states are descriptively similar. That is the price they pay for satisfying the culpability condition in the first place. If I am in a Bad Case, there is a real sense in which I am exposed to the cruel winds of such contingency and chance. In acting wrongly, I make the world, and worldly outcomes, relevant to my moral evaluation, in ways which imply that I may easily get to be rated differently from someone whose internal states were qualitatively similar at the time of action, but whose acts then turn out differently. It is easy enough to understand why anti-­luckists find this picture a harsh one. Agents in Bad Cases lack full control over the circumstances of their evaluation. In a sense, this picture is harsh. Anti-­luckists are wrong, however, to think that there is any meaningful way of subduing its harshness. Agents in Bad Cases have invited this harshness,27 and have thereby risked a certain form of alienation from the goods and standards of evaluation which agents in Good Cases can safely count upon. They have made the world of contingency and chance relevant to their evaluation. This should give them an incentive—an additional incentive— not to act in these ways. What (C), (D), (E), and (F) give us, in conclusion, is a way of reconciling two sorts of moral significance: the moral significance arising from the worldly consequences of acts which reflect culpable inner states, and the moral significance of choosing and performing acts, however they turn out, which reflect those culp­ able inner states. There is a sense in which, in deciding to do something wrongful, I am inviting the world in. I am making my moral evaluation dependent on factors over which I fully lack control. Morality’s remedy is obvious: I should not 26  Cf. Gardner (2004) for a helpful discussion of the difference between reasons to try, reasons to succeed-­by-­trying, and reasons to succeed, simpliciter. 27  I do not say, of course, that they have consented to it. Their consent is not needed.

102  Resultant Luck and the Fairness Intuition behave wrongfully. But then, we knew that already. The Moral Cost Argument simply traces out the consequences of ignoring that remedy, and invites us to be reconciled to the existence of luck in the amounts of blameworthiness which are then required.

3.4.5.  Blame, Praise, and Comparative Fairness Egalitarianism The conclusions reached in this chapter can be roughly stated as follows. Agents in Bad Cases are liable to be blamed for the badness of the consequences they have caused to happen, thus nullifying any appeal to interpersonal fairness. Agents in Good Cases, by contrast, are not strongly worthy of being praised for differences in the goodness of the consequences they have caused to happen, where those differences are due to resultant luck. One way of articulating this claim is to say that moral evaluation—laudability or blameworthiness, as appropriate—should be distributed equally, unless agents have made choices which explain the differences in moral evaluation. That articulation of the underlying principle of evaluation may look a lot like a recognizably responsibility-­sensitive principle of distribution for blame­worthi­ ness and praiseworthiness. Consider Larry Temkin’s recent statement of his principle of comparative fairness, or of his comparative fairness egalitarianism: . . . there is nothing good or bad about equality, per se, rather, equality should be approved only to the extent that it promotes, or corresponds with, a comparatively fair outcome, but it should be opposed to the extent that it promotes, or corresponds with, a comparatively unfair outcome.28

The ostensible similarity between Temkin’s comparative fairness egalitarianism and the principle of distribution for blameworthiness I have reached may foster the expectation that I will go on to defend a version of luck egalitarianism in Part II, when I discuss distributive justice. That expectation will be confounded. This is partly due to the different ambitions for, and the constraints operating upon, a theory of distributive justice. A theory of distributive justice has, or ought to have, a different focus from a theory of blameworthiness. Luck egalitarians’ construal and deployment of the concept of luck is not fit to serve that purpose. These arguments will be advanced in Chapter 6. But there is more to say about the application of comparative fairness egali­tar­ ian­ism even to the allocation of blameworthiness. I have argued that there is a normative mechanism, outlined by the Moral Cost Argument, for why agents in

28  Temkin (2017), p. 55; original emphases.

Conclusion  103 Bad Cases are liable to be blameworthy in proportion to the results of what they do, even if these results are then subject to resultant luck. These pressures do not apply to agents in Good Cases. But the asymmetry between agents in Bad Cases and Good Cases does not reflect a background commitment to anything like comparative fairness egalitarianism. We should not be interested in upholding equal opportunities for avoiding blame, and equal opportunities for earning praise. Either these tendencies betray a misguided attitude to what morality expects of us, or they are guilty of ignoring what is there to be morally evaluated. These arguments will be advanced in Chapter 4.29

3.5. Conclusion The argument traced through Chapters 2 and 3 has been long and complicated, so it may be helpful at this point to offer a brief recapitulation of the conclusions I have reached. The Anti-­Luck Account is primarily powered by two intuitions, the Irrelevance Intuition and the Fairness Intuition, which become less alluring when the underlying issues are squarely confronted. The Irrelevance Intuition is resistible. The wrongness of malicious attempts or reckless acts is not naturally detached from the actual outcomes of such acts, and there is, in turn, no reason to detach blameworthiness assignments from the wrongness of the acts performed by these agents. They lack a defence against their exposure to luck, given the fact that they have already ignored morality’s demand not to have acted in these ways. And they are also exposed to the plausible claim that the acts which embody the internal states differ in badness, depending on how these acts turned out. In the absence of support from the Irrelevance Intuition, the Fairness Intuition can add little to the cumulative case for the Anti-­Luck Account. First, the insistence that blameworthiness should be assigned to an equal degree introduces distortions which are exposed by the attack on the Irrelevance Intuition. Second, even a focus on fairness alone does not unambiguously support a case for equal blameworthiness. Allowing blameworthiness assignments to vary among individuals does not offer an obviously worse accommodation of the value of fairness. Equal Worlds avoid pure interpersonal luck, but instantiate interpersonal fate-­ sharing luck. Lucky Worlds instantiate pure interpersonal luck, but avoid interpersonal fate-­sharing luck. So far, then, I want to contend that the Anti-­Luck Account cannot be upheld, that resultant luck exists, and that the Restricted Account delivers a plausible

29  See, in particular, Chapter 4, Section 4.2.

104  Resultant Luck and the Fairness Intuition explanation of how it can best be accommodated. But the job for the Restricted Account is not yet complete. We need to take a closer look at the other categories Nagel enumerates in his taxonomy of moral luck, particularly circumstantial luck. The extension of concern from resultant luck to circumstantial luck poses difficulties, but also creates certain opportunities. I turn to that exacting task in Chapter 4.

4

Restraining Situational Luck 4.1. Introduction As I mentioned in the Introduction, the moral luck debate has been largely focused on resultant luck. This chapter will examine the significance of circumstantial luck, and will also take a more detailed look at a broader category of luck in the relevant vicinity that Michael Zimmerman has described in detail: this is ‘situational luck’, which encompasses circumstantial luck and constitutive luck.1 The overall trajectory of argument will have a defensive character, in the sense that my overriding aim is to demonstrate that the territory gained in my campaign for the existence and significance of resultant luck is not lost by additional reflection on circumstantial luck. Indeed, many of the lessons I have already drawn for resultant luck apply in a relatively straightforward way to circumstantial luck as well. But that leaves us with a number of challenges and features of local interest. And it creates opportunities as well as challenges for the Restricted Account, as we shall see. I start, in Section 4.2, by indicating why friends of the Anti-­Luck Account may be particularly exercised by circumstantial luck. In Section  4.3, I provide a broader diagnosis of the temptations offered by circumstantial luck, and begin to discredit them. Friends of the Anti-­Luck Account may not be satisfied with these various lines of response, however. To get deeper into the underlying concerns, I will then provide, between Sections 4.4 to 4.6, an exposition and internal critique of Zimmerman’s account, which pushes the anti-­luckist resistance to circumstantial luck to its furthest limit.2 The critique will be internal in the sense that it will flush to the surface the contradictory impulses of Zimmerman’s account. Having shown what these problems are, I will then offer a diagnosis of what may have misled him. (Some of these arguments are dealt with at further length in Appendix II.)

4.2.  Understanding Circumstantial Luck The concerns so far have been focused on resultant moral luck. What additional challenges are raised by circumstantial luck? In the Introduction, I explained why 1  Much of our business with constitutive luck will, in effect, be transacted in Chapters 6 and 7. 2  Zimmerman (2002).

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0005

106  Restraining Situational Luck there is a genuine distinction between resultant luck and circumstantial luck.3 Circumstantial luck should be understood as the luck of being exposed to circumstances in which intentions are formed, acts are performed or at least attempted, dispositions are actualized, and dispositions are developed.4 The crucial contrast here, roughly put, lies between potentiality and actuality. Such luck extends to both Bad Cases and Good Cases, as I have called them. In connection to Bad Cases, agents with the potentiality for actualizing morally problematic internal states may be the beneficiaries of the good luck of not being placed in circumstances in which those internal states would have been actualized. In connection to Good Cases, agents with a similar potentiality for actualizing morally benign internal states may have been denied the opportunity of being placed in circumstances in which those internal states would have been actualized. David Lewis thinks that ‘the most intelligible cases of moral luck are those in which the lucky and the unlucky alike are disposed to become wicked if tempted, and only the unlucky are tempted’.5 What Lewis is describing here is squarely circumstantial luck. There can be no question that circumstantial luck is a genuine form of moral luck. How should the Restricted Account deal with it?

4.2.1.  Two Challenges For circumstantial luck, elements of Good Cases and Bad Cases can be combined. Imagine that Angela and Angelo both have similar moral character profiles: each of them is generous, but lacks courage.6 Now imagine two separate cases. In one of them—call it Generosity—Angela is confronted with circumstances which call for generosity, but in which there is no particular need for her to display courage. Her generosity is tested, but not her courage. She acts generously, as we might expect, and her lack of courage is neither here nor there, because it has no bearing on how she should act in this case. She is, accordingly, a candidate for praise. In the other case—call it Courage—Angelo is confronted with circumstances in which his courage, but not his generosity, is tested. He predictably fails to act in a courageous way, while his generosity is not given a proper outlet of expression. He is, accordingly, a candidate for blame.7 Across these two cases, Angela emerges with 3  See the Introduction, Section 3. 4  Some of these items may appear to belong to the territory mapped out by Nagel’s category of constitutive luck. The boundaries between circumstantial luck and constitutive luck are in fact somewhat porous, as we shall see when we discuss Zimmerman’s category of ‘situational luck’ later on in this chapter. 5  Lewis (1989), p. 56. 6  I leave aside concerns about the unity of the virtues: this thesis can only be extended to very well-­ developed forms of virtues, in any case. We can understand Angela and Angelo to be approximately generous, and relatively uncourageous. These are the dispositional states which our ordinary ascriptions of these terms would pick out. 7  He may have an excuse. These further details do not need to be investigated. All I need for my argument here is the underlying point about circumstantial luck and its effect on moral valence.

understanding circumstantial luck  107 moral credit, but Angelo emerges with moral discredit. Had their circumstances been inverted, with Angelo being exposed to the circumstances of Generosity and Angela being exposed to the circumstances of Courage, these evaluations would most likely have been reversed: it is Angelo who would have emerged with moral credit and Angela who would have emerged with moral discredit. The explanation of why there is this discrepancy between them must be due, in part, to luck: circumstantial luck. Is this a problem? Now both Angela and Angelo may be exposed, further down the line, to other circumstances in which their characters will be tested in different ways. In these other cases, the distribution of moral credit and discredit between them may be inverted. But we should not assume in advance that these agents’ long-­haul circumstantial records are somehow guaranteed to balance out, so that neither of them will be in the position to claim any more moral credit than the other overall. We need a sharper form of engagement with the problem. There are, in fact, two problems raised by Generosity and Courage. The first of these problems concerns the shift in moral valence as we move between the cases. In Generosity, Angela is praiseworthy. Had she been in Angelo’s circumstances in Courage, we can assume that she would most likely have acted in a blameworthy way. Exactly the same holds for Angelo. In Courage, he is blameworthy. Had he been placed in Angela’s circumstances in Generosity, it is likely that his actions would have been praiseworthy. That difference, between praiseworthiness and blameworthiness, involves a change of moral valence in the evaluations we make of these agents. We go from positive to negative, or vice versa. That change in valence was not evident in the central cases that were used to explore resultant luck. In Drivers, Assassins, and Two Buttons, each of the agents in the pairs of agents in question qualified for some degree of blame­ worthi­ness. Resultant luck merely altered the degree of blameworthiness that each of them collects.8 Circumstantial luck, by contrast, changes the entire valence of the agent’s moral evaluation, not just the degree of it. This point is actually clearer in Generosity and Courage than it is in Nagel’s original case. Transplanted to Argentina, the German émigré is not exposed to the moral criticism that could be heaped upon him had he remained in Germany and become an officer at a concentration camp. But Nagel’s case does not immediately place the German émigré in circumstances in which he not only avoids blameworthiness, but elicits praiseworthiness as well.9 8  This point is anticipated by Herdova and Kearns (2015), pp. 364–6. 9  Herdova and Kearns (2015), pp. 363–4, outline other reservations about Nagel’s case: they make the reasonable point that it is not just his circumstances, but other factors to do with his character, which would explain why he would have behaved differently had he stayed in Germany. I do not think Nagel meant to imply that it is this man’s circumstances alone which would explain his behaviour. His point is that, however his acts are to be explained, there will be nothing here to explain but for his circumstantial luck, because the more psychological factors would not be mobilized but for the prior existence of this circumstantial luck. That point suffices to make circumstantial luck potentially problematic.

108  Restraining Situational Luck The second problem is that circumstantial luck raises concerns about the distribution of moral burdens. It is understandable to experience situations in which one’s weaknesses are tested as burdensome. Many of us would prefer to live in ways that minimize temptation and exposure to bad company, but our ability to do so will be a matter of luck. In these respects, some people are luckier than ­others. This looks like a potential justificatory headache. It is entirely understandable why the Anti-­Luck Account wishes to take circumstantial luck seriously.

4.3.  Circumstantial Luck and the Distribution of Praise and Blame Now that circumstantial luck has been described, and its threatening nature has been explicated, we can begin to assess how forceful the worries are. I proceed by dividing the intellectual labour between cases of laudability, and cases of blame­ worthi­ness—or, as before, between Good Cases and Bad Cases.

4.3.1.  Distributing Praise: Five Lessons To begin, then, with Good Cases,10 and a familiar old case: the Pond case.11 Imagine that Jessica is walking through the park, on her way to an appointment, and passes a pond which contains a drowning child. She does the right thing and rescues him, at the cost of only muddy clothes and the missed appointment. Jessica is rightly praiseworthy for saving this child.12 But consider Edward, whose tendency to Good Samaritanism fully matches Jessica’s. If Edward had been in the park, and had passed the pond in which a child had been drowning, he would have done exactly the same thing as Jessica does, because those Good Samaritan dispositions would then have found a proper outlet of expression. Yet Edward is not praiseworthy; only Jessica is praiseworthy. Only Jessica, and not Edward, has an active role to play in Pond. Is it problematic that Edward is not eligible for praise which, through merely bad circumstantial luck—or so he might think— goes only to Jessica? Would this be a reasonable complaint for Edward to make? I do not think so. There are five points to be made.

10  A closer look will suggest that the division between praiseworthiness cases and blameworthiness cases is not exclusive: see Section 4.3.2. But there is still something to learn by proceeding in this way. Though I will sometimes be referring to praise rather than praiseworthiness in the passages to follow, it is really praiseworthiness that I am tracking. 11  It originates in Singer (1972). 12  She is at least rated positively, even if it is thought that praise is inappropriate for the fulfilment of a moral obligation. (I cannot see strong grounds for that thought in any case. It is not unreasonable to praise fulfilments of obligation that are burdensome or require effort.)

circumstantial luck and the distribution  109 First, Pond presents us with an obvious opportunity for act-­focused appraisal or evaluation. Jessica can and should be evaluated for saving the child, or for omitting to save her, if that is what she had decided to do. Edward may have the same Good Samaritan dispositions as Jessica, but he was not in circumstances relevant to the activation of these dispositions. Jessica performed an act which is apt for evaluation, and Edward did not. Call this the Opportunity Claim: cases which present us with opportunities for different types of moral evaluation should duly collect those evaluations.13 Second, the aptness of act-­focused evaluation does not displace disposition-­ focused evaluation. We can note that Edward, as well as Jessica, has dispositions which we consider to be commendable. The circumstances of Pond makes act-­ focused evaluation appropriate, but the act-­focused evaluation, in and by itself, does not show a disposition-­focused evaluation to be inappropriate. It does not take a view one way or another about that issue. If you wished to conduct a disposition-­focused evaluation of Edward, you would not be impeded by the Opportunity Claim, or by the reported facts in Pond. One exercise will not displace the other. Call this the Non-­Displacement Claim.14 Third, the inequality in Jessica’s praiseworthiness and Edward’s praiseworthiness—she gets some, while he gets none—does not reflect badly upon him. Not praising Edward is not equivalent to blaming him for not having merited the praise in the first place. The assignment of praiseworthiness does not invite these kinds of comparisons; it is not a zero-­sum game, in which the bestowal of praise on Jessica somehow entails the subtraction of praise from Edward. The very mention of the fact that Edward was not placed in the relevant circumstances of Pond, whereas Jessica was placed in these circumstances, should be enough to reassure him that he is not being morally short-­ changed. Call this the Non-­Zero-­ Sum Claim.15 The fourth point consists in an explanation of why Edward’s non-­collection of praiseworthiness does not reflect badly on him. Had Edward been in circumstances that called for an act displaying the qualities of Good Samaritanism, he would be evaluated for his performance or non-­performance of that act. But the fact that Edward is not in these circumstances does not, in and by itself, require us to make a negative evaluation of him. Edward is not at fault for not being in Pondtype circumstances. Correlatively, we are not praising Jessica for two things: first, finding herself in Pond-type circumstances, and second, responding to those

13  One may think of the Opportunity Claim as a relation of the Relevance Claim, which was discussed in Chapter 2. When acts are performed and have worldly consequences, they are morally significant, and call for moral evaluation. 14  Various traps may be waiting for us here, to do with the fundamental attribution error. But I will ignore these complexities. 15  Some writers, such as Domsky (2004), do apparently hold that our ordinary moral thinking is committed to a rejection of the Non-­Zero-­Sum Claim. See, further, Appendix I.

110  Restraining Situational Luck circumstances by saving the drowning child. We find her praise­worthy only for saving the drowning child, given that she was in circumstances where the child was drowning and needed to be saved, and where she was in a position to save him. It will often be the case that whether one is in these circumstances will be a morally neutral fact, calling for neither blame nor praise. Call this the Neutrality Claim. Edward’s non-­exposure to Pond-type circumstances does not make him blameworthy, just as Jessica’s exposure to these circumstances does not make her praiseworthy. These are, other things equal, just neutral background facts about them. Fifth, it is unclear why Edward’s non-­exposure to the circumstances of Pond should be counted as bad circumstantial luck. It is not a misfortune not to be praiseworthy—so long, of course, as non-­praiseworthiness does not entail the generation of blameworthiness. What morality demands of us is conformity to its standards, not equal opportunities for earning praise by conforming to its standards. It is not a source of bad circumstantial luck that we should live in quiet times, or that opportunities for outstanding moral performance should be in short supply. As a result, moral agents should not deplore their non-­exposure to disasters or misfortunes which would allow them to exhibit their moral qualities. Generous individuals should not object to the absence of neediness which would then make claims upon their generosity, and justice-­loving agents should not pine for injustices to correct, or to protest against. Such longings amount to a corruption of moral focus. According to what I will call the Focus Claim, morality demands primarily that we conform to it, not that we should enjoy equal opportunities for collecting praise for conforming to it. If the Focus Claim is true, then any complaint about unchosen inequalities in the distribution of praise due to circumstantial luck should strike us as morally myopic.16 If Edward needs reassurance or balm for his wounded pride, these five points should give him that reassurance, and also—more pointedly—demonstrate that he had no real grounds for wounded pride in the first place.

16  Does the Focus Claim undermine the asymmetry between Good Cases and Bad Cases addressed in Chapter  3, Section  3.4? No, but it does something to clarify the grounds for this asymmetry. According to the Focus Claim, it is not the case that Good Cases are unlike Bad Cases on the grounds that the agents in Good Cases deserve, and ought to get, equal opportunities for equal praiseworthiness. So it is not the case that Adina and Kristin in Two Better Buttons (see Section 3.4.1) qualify for equal amounts of praiseworthiness on equal opportunity grounds. That claim would indeed be vulnerable to the Focus Claim. There are three points to make about Two Better Buttons in this connection. First, the argument against differential praiseworthiness between Adina and Kristin involves the appeal to Ought Implies Can, not to equal opportunities. Second, agents in Good Cases are unlike agents in Bad Cases in the sense that they have done nothing to make themselves liable to moral criticism. As I see it, there is no specific analogue of the liability-­generating mechanism at work in Good Cases for delivering inequalities in praiseworthiness such as that which exists for inequalities in blameworthiness in Bad Cases. (Bad Cases, but not Good Cases, must accommodate the Denial Claim.) Third, had one of these agents in Two Better Buttons not been in a position to press either Button A or Button B as a result of merely circumstantial luck, she would not have a serious complaint.

circumstantial luck and the distribution  111

4.3.2.  Distributing Blame Now it is plainly not Good Cases but Bad Cases which typically generate the concerns among anti-­luckists about circumstantial luck. Nonetheless, I think we can learn a fair bit from our unsympathetic responses to Edward’s dissatisfaction when we turn our attention to Bad Cases. Most of the points enumerated in discussion of Pond have a reasonably straightforward application to Bad Cases as well. I shall now restate these points in the context of a Bad Case, rather than a Good Case. Imagine, then, that Dylan acts badly, while Akira would have acted badly had she been in the relevant circumstances. Or, to state the concern more carefully: imagine that Dylan acts badly, and that Akira was not placed in circumstances in which she would either have acted badly, or had to fight the temptation to act badly. Akira and Dylan have similarly objectionable dispositions, and do not always resist the temptation to manifest them in the relevant circumstances.17 More specifically, imagine that both Akira and Dylan have a deep tendency to be unpleasantly boorish and rude in social company, when a few alcoholic drinks have been consumed. This time round, Akira has simply had the luck to avoid the circumstances in which Dylan has been placed. Imagine that a surprise party, to which Akira has not been invited, is thrown for Dylan. Call this the Surprise Party case.18 Lo and behold, drinks are consumed, and Dylan disgraces himself yet again. If circumstantial luck is not taken into consideration in Surprise Party, Dylan is blameworthy, and Akira is not. That is because Dylan has acted badly, while Akira has not. Can Dylan complain about the fact that, through his ex­pos­ ure to circumstances which were just a matter of luck, he has been blamed, whereas Akira avoided blameworthiness? I think the answer to this question must be ‘No’. Dylan lacks a strong case. The five points we recovered from Pond should serve as our guide as we test the strength of his complaint. The first point concerns the Opportunity Claim. Dylan is evaluated negatively because Surprise Party gives rise to the opportunity, and thus the demand, for such appraisal. There can be no real question that his behaviour calls for appraisal. Akira, by contrast, has not indulged in such behaviour, and nor has she had to resist the temptation to act badly. Her circumstances simply do not give rise to a relevant opportunity for appraisal. The second point, which swiftly follows on, returns us to the Non-­Displacement Claim: although Dylan is evaluated for his behaviour, and Akira is not, that very 17  This remark should not be taken to imply that these impulses are literally irresistible. Again, the concern is that, through bad circumstantial luck, agents can be tested differently, and that these different tests can be experienced as morally burdensome. 18  The moral stakes are of course lower in Surprise Party than in Pond, but that should not matter. The points to follow should be readily applicable to cases with higher stakes.

112  Restraining Situational Luck fact does not exclude or displace an evaluation of Akira’s disposition to boorishness. The fact that Dylan is blameworthy for his behaviour does not carry any consequences for the evaluation of Akira’s disposition. Conducting an act-­focused evaluation of Dylan does not, in itself, constitute any sort of opposition to the re-­opening of Akira’s disposition-­focused moral dossier, if that is what we are inclined to do.19 Next, there is the Non-­Zero-­Sum Claim. We find Dylan, but not Akira, blame­ worthy for his behaviour. But not blaming Akira does not amount to praising her. The fact that Dylan receives a negative moral evaluation does not place Akira any further ahead in the moral evaluation game. Akira is not praised for avoiding circumstances in which she might then be eligible for blameworthiness. Dylan cannot therefore complain that, while Akira sits out this moral test as a result of circumstantial luck, she is unfairly benefited. She is not relevantly benefited, precisely because the distribution of blameworthiness is not a zero-­sum game. It is a mistake to suppose that Akira unfairly gets further ahead just because Dylan has fallen behind. Next, and in partial explication of the Non-­Zero-­Sum Claim, we turn to the Neutrality Claim. Akira is not blameworthy, but neither is she praiseworthy for not being blameworthy. She has not been placed in the circumstances of Surprise Party, but she does not admit of praiseworthiness for not being placed in those circumstances. Akira is not blameworthy, simply because there was no moral test for her to pass or fail, but it is not as though she is praiseworthy for avoiding that test. Her avoidance of those circumstances reflects neither well nor badly on her. It is in itself a morally neutral fact about her, just as the fact that Edward was not placed in the circumstances of Pond is also morally neutral. Correlatively, Dylan is not being blamed for being in the circumstances of Surprise Party, and then for failing that moral test. He is blameworthy only for failing that moral test, given the fact that he was exposed to it. There are, however, some additional complications to mention here. Imagine that Evelina, knowing about her tendencies, shared with Akira and Dylan, to social boorishness in social company where drinks have been taken, has an intermittent policy of avoiding them, either by not turning up in the first place, or by keeping an eye on her consumption of drink. Assume, then, in the RSVP case, that Evelina receives in advance an invitation to a gathering in which the triggering conditions for her boorishness will all be in place. Knowing this, she 19  But should we? We might wish to proceed cautiously here. Even if Akira has this disposition, she is not precluded from resisting its influence on relevant occasions, and a respect for her moral autonomy may require that we withhold criticism before seeing how her dispositions are manifested in action. Smilansky (1994) provides this sort of response to Christopher New’s views about the moral (if not the epistemic) adequacy of pre-­punishment: see New (1992) for the background, and New (1995) for a response to Smilansky. Statman (1997) notes connections between the pre-­punishment debate and the moral luck debate. For an interesting take on these issues which takes its cue from Steven Spielberg’s film Minority Report, see Mulhall (2016), ch. 6. See also the remarks in Section 4.3.3.

circumstantial luck and the distribution  113 can decline or accept. Imagine that, on this particular occasion, she accepts the invitation, and that the same bad behaviour transpires. If she does this, then we might indeed find her blameworthy for getting into the circumstances which elicit the bad behaviour. We would find her blameworthy for the bad behaviour, and we might also find her blameworthy to some extent for deliberately getting herself into a situation in which her frailties would be tested, or which made it ante­cedent­ly likely that she would behave badly. In RSVP, there is a two-­tiered moral test: agents are open to blame for deliberately exposing themselves to circumstances in which their frailties will be tested, and they can also be found blameworthy for any subsequent manifestation of those frailties. Imagine that Erica has the same dispositions as Evelina. If, in a case like RSVP, Erica politely turns down the invitation to attend the dinner party, she can be praised for that. She will not be praiseworthy for not failing the moral test which was simply not there for her to satisfy or fail, but she can be praised for seeing to it that she did not have to face this test, given her reasonable fears about failing it. Again, this circumstantial test will have its luck-­based counterfactual component: this is the non-­appearance of the initial invitation to the dinner party. Perhaps, due to circumstantial luck, Isabel is not given the opportunity to avoid the test: she is simply not given the opportunity to accept or decline the invitation, with the result that the circumstances applying to Evelina do not apply to her. Again, the Non-­Displacement Claim can be invoked: given Isabel’s dis­pos­ ition not just to boorishness, but to casualness about the avoidance of temptation, nothing stops us from re-­opening Isabel’s moral dossier. The point remains that it is Evelina, not Isabel, who is in the direct line of moral fire in RSVP, due to the act-­focused evaluation of her which is there to be made. These more complicated forms of the circumstantial test also apply to Good Cases. If there is suffering to be alleviated, many of us keep our heads down and engage in ostrich-­like activity in order to shield ourselves from that uncomfortable truth, whereas others are more directly responsive to those un­allevi­ated needs. We should find the former blameworthy, and the latter praiseworthy.20 Finally, we turn to the Focus Claim. Here there may appear to be a relevant asymmetry between Pond and Surprise Party. Edward is denied a hearing for his complaint that he is not praised as much as Jessica in Pond, because morality does not fundamentally care about a fair distribution of act-­focused praiseworthiness. What primarily matters is that we conform to what morality tells us to do, not that we have equal opportunities for collecting praise for doing so. Edward’s 20  The demandingness of morality (and related issues such as the scope of supererogation) cannot be tackled here. But however demanding morality turns out to be, these phenomena will apply: the obligation-­meeting line will be accompanied, on the two distinct sides of it, by a blameworthy ostrich-­ like avoidance of responsiveness to unalleviated need and a praiseworthy responsiveness to un­allevi­ ated need.

114  Restraining Situational Luck complaint seems wrongheaded. But when we turn to Surprise Party, it is less likely that we will take the same view about Akira. If Akira is a flawed but basically conscientious moral agent, she may rightly be glad, or relieved, that she was not exposed to a moral test that she risked failing. Her reaction thus displays no misunderstanding of morality’s focus that is on a par with Edward’s misunderstanding of morality’s focus. She is right to regard the circumstances from which she has not been exposed to as an unwanted threat from which she has been spared, whereas he is wrong to regard the circumstances from which he has not been exposed as a wasted opportunity. Does this then make Dylan’s situation any different? After all, he has been exposed to a moral test to which Akira was not exposed. There is an unequal distribution of moral burdens between them, from which he may infer that he is entitled to complain. Dylan may even suspect that he is the victim of double jeopardy: he is unfortunate by comparison with Akira for being placed in the (mildly) morally hazardous circumstances of Surprise Party in the first place, and his misfortune is then compounded by being regarded as blameworthy for the bad behaviour that duly transpires.

4.3.3.  Holding the Line I think this reasoning is wrongheaded. We need to hold the line. There are two points to make. The first of these points is probably overdue. Each of these main cases—Surprise Party, and Pond—has a dual aspect. Each of them calls for a response that will elicit praise, but also risks a response that will invite blame. In Pond, Jessica will be a candidate for praise if she saves the drowning child, but will invite blame if she fails to save him (or at least fails to make a serious attempt to save him). Now it was stipulated, in Pond, that Jessica saved the child, but she may have not done. Just as Edward is not in the circumstances that can make him praiseworthy for saving the child, he is also denied exposure to the circumstances that could have made him blameworthy for failing to save him. The same point applies to Surprise Party. Dylan will be eligible for blame if he acts boorishly and rudely, but he may be open to praise for resisting these tendencies. Akira will not be blamed for failing a test that was not there for her to fail, but neither will she be praised for passing a test that was there for her to pass. It was stipulated, in Surprise Party, that Dylan failed the test, but he might have passed it. If he had passed it, he would then have been a candidate for praise, not blame. The second point, which broadly ensues from the first point, is that Dylan’s complaint about double jeopardy must rest upon a mistake. His mistake is to suppose that there was a relation of determination between being placed in the circumstances of Surprise Party and acting in a blameworthy way. But it was not inevitable that Dylan behaved badly, given the obtaining of ordinary conditions

The Explosion Argument  115 of control over our acts which I have been assuming all along.21 He may have avoided doing so. Had he managed to avoid acting badly, then there would be no content to the double jeopardy worry: again, he would then have actually been a candidate for praiseworthiness, rather than blameworthiness. We may reasonably hope for undemanding moral circumstances, but if more demanding circumstances make an appearance, we cannot complain that we might have been luckier. Or can we? Very determined anti-­luckists might wish to resist these arguments. The combination of the Opportunity Claim, the Non-­Displacement Claim, the Non-­Zero-­Sum Claim, the Neutrality Claim, and the Focus Claim might not satisfy them. They will continue to insist that there is unfairness, and a sort of moral myopia, in steering clear of pairwise comparisons between agents who are separated only by circumstantial luck. This gauntlet has been cast with particular force by Zimmerman. The time has come to investigate his account in detail.

4.4.  The Explosion Argument Zimmerman provides instructive and intriguing lessons to his fellow moral anti-­ luckists on how far their anti-­luckism can be taken.22 As he puts it: ‘seldom have the implications of the denial of the relevance of luck to moral responsibility been pursued to their logical conclusion’.23 This is Zimmerman’s ambition. His conclusions are startling. If I dispute them, it is not simply because they are so surprising. My strategy will be to demonstrate that, by his own lights, Zimmerman is not entitled to embrace these conclusions.24

4.4.1.  Situational Luck Zimmerman’s argument awards a starring role to not only resultant luck, but to circumstantial luck and constitutive luck as well. As mentioned above, he refers to the combination of these two further types of luck as situational luck, which combines the circumstantial luck in one’s ‘external situation’ with the constitutive luck in one’s ‘internal situation’.25 Though it rests upon relatively simple foundations, Zimmerman’s argument is complex and far-­reaching. In Section 4.4.2, I will provide a reconstruction of its 21  See the Introduction, Section 4. 22 See especially Zimmerman (2002). Some of Zimmerman’s other work touches on the same issues, but this is his most detailed and thorough treatment. 23  Zimmerman (2002), p. 559. 24  For an instructive critical discussion of Zimmerman’s argument, see Hanna (2014). My points are, I believe, broadly compatible with Hanna’s, but my engagement with Zimmerman’s argument will be different. 25  Zimmerman (2002), pp. 559, 565.

116  Restraining Situational Luck central thrust. My aim here will be to draw out the full implications of Zimmerman’s account. As a result, the reconstruction will broadly shadow the progression of Zimmerman’s argument, but present it slightly differently, and in such a way that its underlying stages and principles of transition are rendered fully explicit. (The cases and the names of agents are also mine.) I then, in Section 4.4.3, add three further clarifications regarding the character and scope of the argument. The argument is criticized in Section 4.6.26 I will refer to Zimmerman’s overall argument as the Explosion Argument, since its ultimate aim, as we shall see, is to explode the notion of responsibility, or demonstrate a conceptual instability that is somehow internal to it. As I reconstruct it, the Explosion Argument unfolds in distinct phases, which cumulatively insist on an expanded role for the central anti-­luckist intuition. I enumerate four distinct cases, where the first case is used as a platform to generate three further cases. Before I begin, two assumptions made by Zimmerman need to be explicitly stated. First, luck is inversely correlated with control: as luck increases, control decreases, and vice versa. Second, blameworthiness is positively correlated with responsibility: if there is something for which an agent can be deemed blame­ worthy, then that agent can be held responsible for it, and vice versa. These assumptions do some of the heavy-­lifting in the development of the argument, so they need to be kept in mind.

4.4.2.  Four Cases In Case One, we start with a familiar pair of attempt cases.27 Imagine, in World 1, that Bonnie and Clyde are agents who both attempt some wrongdoing: let φ-ing be killing, robbing a bank, kidnapping, or whatever.

Case One: 1. In World 1, Bonnie attempts to φ. 2. In World 1, Clyde attempts to φ. 3. Bonnie’s attempt to φ is successful, and causes outcome O. 4. Clyde’s attempt to φ is unsuccessful, and does not cause outcome O. Now 5. The difference in outcomes between Bonnie’s successful attempt to φ in World 1 and Clyde’s failed attempt to φ in World 1 is simply a matter of luck: resultant luck.

26  To avoid taxing the reader’s patience unduly, some of the further argumentative details and possibilities will be reserved for Appendix II. 27  Zimmerman (2002), p. 560.

The Explosion Argument  117 But 6. A difference between Clyde and Bonnie in World 1 which is due simply to luck cannot uphold any difference between Clyde’s blameworthiness and Bonnie’s blameworthiness in World 1. And so, by 5 and 6: 7. Clyde is no less blameworthy than Bonnie in World 1. This will be a familiar sort of argument from the debate about resultant luck. Case One corresponds exactly to the structure of Assassins. But Zimmerman now pushes its underlying point further, to encompass areas of concern beyond successful and failed attempts.28 Imagine, then, in World 2, that Clyde does not attempt to φ, simply because he has not been given the opportunity to do so, or because he was thwarted from taking advantage of an opportunity that had been initially presented to him. (We might suppose that Clyde, like Imogen and Phoebe in Assassins, is a professional assassin, and would have willingly taken the shot, but then loses his opportunity.) If Clyde was presented with an unobstructed opportunity to φ, however, he would seize this opportunity with both hands. The difference between Clyde’s situation in World 1 and Clyde’s situation in World 2 can be regarded as simply a matter of luck: circumstantial luck. With this point in mind, we can expand the basic argument from Case One to take account of this new concern with circumstantial luck. This gives us Case Two.29

Case Two: 8. In World 2, Bonnie attempts to φ. 9. In World 2, Clyde does not attempt to φ, due to the lack of an unobstructed opportunity to do so. Now 10. The difference between what Bonnie does in World 2 (i.e. to attempt to φ), and what Clyde does in World 2 (i.e. not attempt to φ due to the lack of an unobstructed opportunity to do so) is simply a matter of luck: circumstantial luck. But 11. A difference between Clyde and Bonnie in World 2 which is due simply to luck cannot uphold any difference between Clyde’s blameworthiness and Bonnie’s blameworthiness in World 2. 28  As we shall see, there is a sense in which the expansion of the argument to follow will eventually come full circle. 29 Zimmerman (2002), p. 563. It would also be possible, using Zimmerman’s argumentative resources, to present an analogous ‘single-­agent, mixed world’ case, comparing Clyde in World 1 with Clyde in World 2. (The same goes for Case Three, when we get to it.) But to reduce needless complexity, I will restrict the exposition, as Zimmerman himself does, to the interpersonal case.

118  Restraining Situational Luck And so, by 10 and 11: 12.   Clyde is no less blameworthy than Bonnie in World 2. One immediate further area of expansion of the argument will now be noted, which will furnish us with the materials for another case.30 Imagine that, in World 3, Clyde does not attempt to φ, and would not be disposed to attempt to φ even when he is given an unobstructed opportunity to do so. This is because, in World 3, Clyde possesses the morally upstanding character which is such that he recoils from any opportunities for wrongdoing that came his way. In other words, Clyde is like us. (At least, he is similar to what most of us, in most situations, probably take ourselves to be like.) One might be forgiven for thinking that, in World 3, Clyde must be completely safe from blameworthiness, at least as far as φ-ing is concerned. He neither makes an attempt to φ, nor is constituted in such a way that he attempts to φ if and when suitable opportunities come along. But Clyde’s immunity to blameworthiness, surprisingly, is just what Zimmerman’s account denies. For the various differences between Clyde’s situation in World 3 and his situation in World 2 boil down, once more, to matters of luck: constitutive luck.31 This gives us Case Three.

Case Three: 13. In World 3, Clyde does not attempt to φ, even though he is given the opportunity to do so, because he is not disposed to φ. 14. In World 3, Bonnie does not attempt to φ, but would have attempted to φ had she been given an unobstructed opportunity to do so. Now 15. The difference between what Clyde does in World 3 (i.e. not attempt to φ, and to be such that he would not have been disposed to φ even when given the opportunity to attempt to φ) and what Bonnie does in World 3 (i.e. not attempt to φ, but to be such that she would have attempted to φ had she been given the opportunity to do so), is simply a matter of luck: constitutive luck.32

30  Zimmerman (2002), p. 565. 31  Alarm bells will probably be going off at this point in anyone who has concerns about the robustness of character traits: see, for example, Doris (2002). The question of whether we can have knowledge of counterfactuals is addressed in Zimmerman (2002), pp. 571ff. I will not pursue these doubts any further here. 32  The wording here might seem awkward: after all, Clyde does not do anything in World 2. We should perhaps understand the relevant contrast as amounting to ‘what Clyde does (or does not do)’ in World 1 and World 2. Or we could use a phrase such as ‘what Clyde gets up to’ in World 1 and World 2 in order to display the relevant contrast. But these are essentially semantic quibbles or obs­ tacles, and we should not be falsely detained by them.

The Explosion Argument  119 But 16.  A difference between Clyde and Bonnie in World 3 which is due simply to luck cannot uphold any difference between Clyde’s blameworthiness and Bonnie’s blameworthiness in World 3. And so, by 15 and 16: 17.   Clyde is no less blameworthy than Bonnie in World 3. Finally, there is one further case that deploys this reasoning for applications of situational luck, which combines circumstantial luck with constitutive luck. Zimmerman does not state this case explicitly, but he is implicitly committed to it, and it can be created from materials to which he is already pledged his loyalty. This, then, is Case Four, which compares Clyde in World 3 and Bonnie in World 1.

Case Four: 18. In World 3, Clyde does not attempt to φ, even though he is given the opportunity to do so, because he is not disposed to φ. 19. In World 1, Bonnie attempts to φ, and her attempt, being successful, causes O. Now 20. The difference between what Clyde does in World 3 (i.e. not attempt to φ, and to be such that he would not have been disposed to φ even when given the opportunity to attempt to φ), and what Bonnie does in World 1 (i.e. to attempt to φ, and to cause O, given that she is disposed to φ and has the opportunity to attempt to φ), is simply a matter of luck: a combination of situational luck (combining circumstantial luck with constitutive luck), and resultant luck (concerning the success of Bonnie’s attempt to φ). But 21. A difference between Clyde in World 3 and Bonnie in World 1 which is due simply to luck cannot uphold any difference between Clyde’s blame­ worthi­ness in World 3 and Bonnie’s blameworthiness in World 1. And so, by 20 and 21: 22. Clyde is no less blameworthy in World 3 than Bonnie is blameworthy in World 1. Note the particular starkness of the contrast in this final case, which explicitly contrasts what would appear to be the worst case with the most innocuous case: Bonnie’s situation in World 1, where she makes a successful attempt to φ and causes O, is no less blameworthy than Clyde’s situation in World 3, where he both refrains from attempting to φ and lacks the disposition to attempt to φ! If Bonnie has murdered someone, she is no more and no less blameworthy for it than Clyde,

120  Restraining Situational Luck who does not murder anyone, who refrains from even trying to murder someone, and who would not be disposed to attempt to murder anyone in case he was presented with the opportunity to do so. I now need to comment on some further features of Zimmerman’s argument, in order to prepare us for the critical discussion.

4.5.  Three Dimensions of the Explosion Argument In order to engage critically with the Explosion Argument, we need to be in possession of a deeper understanding of its structure and purpose. In particular, we need to make explicit why I have labelled Zimmerman’s argument the Explosion Argument. I start, in Section 4.5.1, with Zimmerman’s distinction between freedom and control. I then remind the reader, in Section 4.5.2, of his important distinction between the degree and the scope of blameworthiness. This discussion will pave the way, in Section  4.5.3, for an explanation of how and why Zimmerman’s argument is supposed to explode the concept of responsibility.

4.5.1.  Freedom versus Control Notice a striking feature contained by Case Three and Case Four: in these cases, Clyde’s blameworthiness (in World 3) is not affected by how free he is, or whether he is free in any recognizable sense at all. Our ordinary construal of the conditions that need to be in place for an agent to be responsible or blameworthy will avert to the capacity for reasons-­responsive action and deliberation, the existence of alternate possibilities, and perhaps other conditions. Call these the freedom conditions. In the World 3 cases, however, we were not paying attention to whether these freedom conditions obtain. The conclusions to those arguments were ultimately fuelled by the point that Clyde’s degree of blameworthiness cannot fail to be affected if the difference between his situation and the situation of an agent who is blameworthy is a mere matter of luck. So luck turns out to be more fundamental than freedom for the ascription of blameworthiness. Why is there this cleavage between freedom and luck? It is because Zimmerman thinks that the deeper principle connecting freedom to responsibility is one which insists upon control, and thus one which excludes luck. In slogan form: ‘luck is irrelevant to responsibility’.33 In slightly elaborated form: ‘the degree to which we are morally responsible cannot be affected by what is not in our control’.34 I shall call this the Control Principle. The Control Principle is intended 33  Zimmerman (2002), p. 559.

34  Zimmerman (2002), p. 559.

Three Dimensions of the Explosion Argument  121 to function as the master-­principle in Zimmerman’s argument. It is utterly central to his account. We shall return to it for more intensive investigation in Section 4.6.

4.5.2.  Blameworthiness: Degree versus Scope We have already encountered, and also provided some assessment of, Zimmerman’s distinction between scope and degree of responsibility.35 My purpose here is to explain how this distinction applies to this particular range of cases. There would appear to be morally significant differences in what these agents can be held to be responsible for as we move across their situations in the different possible worlds. Consider the simplest case, Case One: in this case Bonnie succeeds in φ-ing, whereas Clyde fails to φ. So Bonnie, unlike Clyde, can also be held responsible for outcome O.  Zimmerman immediately concedes that a certain deontic moral judgement can be made about Bonnie, but not Clyde, in World 1: by φ-ing, Bonnie causes O, and an action which produces this outcome is wrong.36 If φ-ing is an act of murder, then Bonnie is blameworthy for murder. If Clyde did not murder anyone, there is no act of murder for which he can be deemed responsible. It will be a tediously familiar point by now that we view murder, not just attempted murder, as an offence against morality, in addition to being an offence against the criminal law. It now appears to follow that Bonnie is more morally blameworthy than Clyde. This conclusion cannot be avoided if we permit judgements concerning responsibility or blameworthiness to track deontic judgements. If we already have the deontic judgement in place, concerning the wrongness of murder, and we are prepared to partner that deontic judgement with a judgement concerning responsibility or blameworthiness, then Bonnie will be more blame­ worthy than Clyde. The same proposed puzzle can be recovered in even starker form from Case Four. Here Clyde does nothing, so there is nothing for which he is responsible: he is not responsible for any attempt to φ, or any outcome ensuing from that attempt, or for the disposition to attempt to φ had a suitable opportunity come along. But if he is nonetheless not just responsible, but every bit as responsible as Bonnie, then what can he be responsible for? In this case, it will seem that there must be unequal blameworthiness between Clyde and Bonnie, simply because there is something for which Bonnie can be blamed and absolutely nothing for which Clyde can be blamed. Zimmerman coins the term ‘hypological’ to refer to judgements concerning responsibility or blameworthiness.37 So our question can be put like this: what 35  See Chapter 1, Section 1.2.3, and Chapter 3, Section 3.2.2. 37  Zimmerman (2002), p. 554.

36  Zimmerman (2002), p. 561.

122  Restraining Situational Luck is the principle, for Zimmerman, which connects deontic judgements with hypological judgements?38 Zimmerman concedes that, on an initial inspection of the territory, there does indeed seem to be a shortage of things for which Clyde, in these different cases, can be responsible for, if he is going to be as responsible as Bonnie. To preserve the verdict that Clyde is still nonetheless as responsible as Bonnie, he invokes the distinction between the scope and degree of responsibility.39 This distinction licenses him to say that, while Bonnie’s responsibility in Case One enjoys greater scope than Clyde’s responsibility, on the grounds that Bonnie can be held responsible for outcome O in addition to her attempt to φ, Clyde is not responsible to any lesser degree than Bonnie. Similarly, in Case Four, Bonnie’s responsibility enjoys much greater scope than Clyde’s: Bonnie is responsible for the attempt to φ, the production of O, and (perhaps) various aspects of her bad character, while Clyde is responsible for neither the attempt to φ nor for the bad character that would, together with propitious circumstances, have explained the attempt to φ. Even with the distinction between scope and degree in place, we are still en­titled to ask what Clyde can be responsible for when we get to World 3. Zimmerman’s proposal is that he is responsible tout court, rather than responsible for anything in particular.40 There is nothing in particular for which Clyde is responsible; he is instead responsible tout court.41 I will note here four features of responsibility tout court, in order to fill out the picture of what responsibility tout court actually amounts to. First, Zimmerman holds that responsibility tout court does not offend against the generally accepted relational structure of responsibility: if Clyde is responsible tout court in World 3, then he is responsible in virtue of the fact that it is only a 38  Zimmerman (2006) provides a fuller treatment of the relationship between deontic judgements and hypological judgements. 39  Zimmerman (2002), p. 560. This distinction is also anticipated in Zimmerman (1987), but the earlier treatment is more cursory. Zimmerman holds, for a case like Drivers, that ‘[t]he [pedestrian’s] death indicates the need for a negative evaluation only indirectly. . . while the decision indicates it directly; but an indirect confirmation is still an indication’ (1987, pp. 383–4). Perhaps so—but the ‘indirect confirmation’ offered by the pedestrian’s death is still more or less redundant. It adds nothing to the agent’s responsibility for the reckless decision. Everything turns on the agent’s responsibility for that initial decision. It seems to me that, in his later treatment, Zimmerman tries to secure a much more robust or exacting accommodation of the significance of outcomes, and of agents’ re­spon­si­bil­ ities for them. 40  Zimmerman (2002), p. 564. 41  Peels (2015), p. 76, expresses immediate scepticism about the notion of responsibility tout court: his view is that, if there is nothing that Clyde is responsible for in World 3, then he is not responsible. Ultimately, I tend to share Peels’s suspicions about responsibility tout court, but I will let the argument run for a while. I have a different fundamental objection to level against Zimmerman. See also Khoury (2018), pp. 1361–3, for criticism of Zimmerman’s distinction that is broadly complementary to mine. (Khoury’s deep commitments are of course starkly different from mine: he is an internalist, whereas I am an externalist. But his opposition to the distinction between scope and degree confirms that not every internalist will wish to take advantage of it.)

Three Dimensions of the Explosion Argument  123 matter of luck that he avoids having the character which is such that, if the opportunity came along, he would attempt to φ.42 Second, the category of responsibility tout court allows Zimmerman to uphold the claim that freedom is required for responsibility. For the sense in which freedom is required for responsibility concerns responsibility-for, rather than responsibility tout court. Responsibility-for concerns responsibility for a particular act, or outcome, and this sort of responsibility does indeed require freedom. Responsibility tout court, by contrast, does not require freedom: it requires only that one be connected, via the Control Principle, to a case in which one enjoys responsibility-for. Third, although responsibility-for may require freedom, and thus require that the freedom conditions obtain, Zimmerman thinks that the more detailed question concerning what freedom consists in becomes less urgent when responsibility tout court is inserted into the picture. As Zimmerman variously puts it, the question of what freedom consists in is then ‘diluted’, ‘loses a great deal of its significance’, or ‘loses much of its force’.43 This is because responsibility tout court is more significant than responsibility-for. As Zimmerman puts it: . . . whether there is something for one which is responsible is immaterial; all that matters, fundamentally, is whether one is responsible. Degree of responsibility counts for everything, scope for nothing, when it comes to such moral evalu­ ation of agents.44

Fourth, responsibility-for and responsibility tout court are not advanced as rival accounts of responsibility. Zimmerman does not have a strategy of bifurcation in mind. Instead, there is just one type of responsibility: in Case Four, Clyde and Bonnie are as responsible as each other, and are therefore to be evaluated in exactly the same way.45

4.5.3.  Conceptual Instability: Expansion versus Contraction There are two ways of grasping the trajectory of thought we have traced. According to the first and perhaps more obvious of them, the Explosion Argument takes us from more obvious cases to less obvious cases for the assignment of blameworthiness, thus expanding the realm of responsibility and

42  Zimmerman (2002), p. 570. 43  Zimmerman (2002), p. 566. 44  Zimmerman (2002), p. 568; original emphases. The point is reaffirmed in Zimmerman (2006). 45  Zimmerman (2002), p. 568.

124  Restraining Situational Luck blame­worthi­ness.46 Thus the Explosion Argument leads us, in Case One, from successful wrongful attempts (the most obvious case of blameworthiness) to unsuccessful wrongful attempts (a slightly less obvious case); it then proceeds, in Case Two, to encompass non-­attempts due to circumstantial luck (still less obvious cases), making these cases candidates for blameworthiness as well; then, in Case Three, it argues for the blameworthiness of non-­attempts due to constitutive luck (even less obvious cases); finally, in Case Four, it contends that the blameworthiness of non-­attempts due to constitutive luck is every bit as high as the blameworthiness of successful attempts (a wholly non-­obvious case). If the successful wrongful attempt in Case One is deemed to be blameworthy, then so are these other cases. Because this trajectory of blameworthiness-­assignment demonstrates how blame­worthi­ness is to be expanded across the board, we can call it the Expansion Strategy. This is not the only way to proceed. It should be noted that none of these cases asserted that there is a specific offence which warrants blameworthiness. That is not even true of Case One, in which Bonnie makes a successful wrongful attempt. The concern was always purely comparative: one agent is no more and no less blameworthy than the agent with whom she is being compared, if the differences between them reflect only luck. This permissive structure can be treated in more than just one way. Accordingly, a rival way of understanding these developments will proceed in the opposite direction. Take Clyde in Case Four: if he is held to be non-­blameworthy, and there is only a series of lucky differences between his situ­ ation and the situation of agents in the other cases, then these agents, in these other cases, will also be deemed non-­blameworthy. If Clyde in Case Four is held to be non-­blameworthy, then we will also be committed to holding Bonnie in Case One to be non-­blameworthy. Because this second trajectory of blameworthiness-­ assignment demonstrates how blameworthiness is to be reduced across the board, we may call it the Contraction Strategy. Other things equal, the Expansion Strategy and Contraction Strategy seem equally viable, at least if the fundamental concerns are comparative. Neither of these strategies is ruled out by the Explosion Argument. So we have uncovered a source of instability within the very concept of responsibility, thus effectively exploding the concept of responsibility, or demonstrating that it cannot be relied upon to generate stable verdicts. And this is precisely the conclusion Zimmerman draws: . . . since an indefinite number of counterfactuals about what one would do, if one were differently situated, can be true at once, one can be morally responsible

46  I should clarify: it expands the scope of responsibility and blameworthiness. It does not, in and by itself, insist upon generally harsher assignments of blameworthiness in any particular case.

Exploding the Explosion Argument  125 tout court—both positively and negatively—to an indefinite number of degrees at once.47

Responsibility and blameworthiness are supposedly shown to be fundamentally unreliable, and the attendant risks of conceptual contamination and moral confusion demand that we sharply disinvest in blaming people, and in practices which are built on these rickety foundations, such as punishment.

4.6.  Exploding the Explosion Argument These developments will seem deeply surprising, but we need argument, not an incredulous stare, if we are going to see how and indeed whether the Explosion Argument misfires. This section identifies the tensions in Zimmerman’s argument, and attempts to locate what I think must be his ultimate commitments.

4.6.1.  The No-­Difference Claim and the Control Principle The lynchpin of the Explosion Argument—or the common lynchpin in his several sub-­arguments, as they have been depicted here—is located at, respectively, claim 6 in Case One, claim 11 in Case Two, claim 16 in Case Three, and claim 21 in Case Four. (These were the claims that followed the ‘But’ in each of the sub-­ arguments.) In each case, Clyde gets pulled into an orbit of blameworthiness due to a merely lucky difference between him and Bonnie. Here is a more general representation of this claim: No-­Difference Claim: The level of blameworthiness which is due to X will also be due to any agent, Y, if the difference between X’s situation and Y’s situation is due to factors which are simply a matter of luck.48

Should we agree with the No-­Difference Claim? And how is it related to the Control Principle? To recapitulate, the Control Principle can be defined as follows: Control Principle: The degree to which X is morally responsible cannot be affected by what is not under X’s control. 47  Zimmerman (2002), p. 570. I should say that Zimmerman does not draw out the full texture of the Expansion Strategy and the Contraction Strategy in his essay, but these aspects of his argument are clearly and strongly implied, as the quoted passage attests. 48  Note that Zimmerman’s argument will permit cases in which X is identical with Y.  Like him, though, I will focus on interpersonal cases, in which X and Y are distinct.

126  Restraining Situational Luck That statement of the Control Principle is taken verbatim from Zimmerman,49 but it would be helpful to have a restatement of it which matches the concepts and terms embedded in the No-­Difference Claim. We can trade on the conceptual connections we have already been relying upon in order to restate it in this more helpful way. One of these connections is between responsibility and blame­worthi­ ness: if there is something for which an agent can be deemed blameworthy, then that agent can be held responsible for it, and vice versa. The second of them is between control and luck: luck is inversely correlated with control, in the sense that as luck increases, control decreases, and vice versa.50 Hence it is possible without any loss of accuracy to restate the Control Principle as follows: Control Principle (alternative version): The degree to which X is blameworthy cannot be affected by factors which are due simply to luck.

This is the version of the Control Principle with which I will now be chiefly concerned. Clearly, Zimmerman is committed to the Control Principle; he is prepared to describe it, after all, as the most fundamental principle in his argument, the key to his whole account. And clearly, in turn, the No-­Difference Claim is meant to be compatible with the Control Principle. The No-­Difference Claim does the heavy lifting in each of the four sub-­arguments dealing with the pairwise comparisons between the two agents. Zimmerman needs both these claims. A closer look, however, places their compatibility in considerable jeopardy. Take the most extreme or surprising case, Case Four: Bonnie attempts to φ, causing O, while Clyde neither attempts to φ, though he is given the opportunity to do so, because he is not disposed to φ. These differences between Clyde’s situ­ ation and Bonnie’s situation are unchosen; they are a matter of luck. The sources of luck are hybrid—they comprise resultant luck, circumstantial luck, and constitutive luck—but they are all sources of luck, and thus enjoy, for Zimmerman, an equivalent status. So, according to the No-­Difference Claim, Clyde is just as blameworthy as Bonnie. But this reasoning appears to be flatly incompatible with the Control Principle. Why so? We already know the answer. It draws on exactly the same materials as those which feed into the Control Principle. It is simply because Clyde lacks control over the existence of these differences between himself and Bonnie. These differences between Clyde and Bonnie are—as we know from the No-­Difference Claim—a matter of luck.51 By application of the Control Principle, the differences lying beyond Clyde’s control should make no difference to how 49  Zimmerman (2002), p. 559. 50  See Section 4.4.1. 51  This line of argument will also have considerable purchase in Chapter 6, when I examine luck egalitarianism.

Exploding the Explosion Argument  127 blameworthy he is. It follows that his blameworthiness cannot be affected by these lucky differences between Bonnie and himself. If that is so, then Clyde’s unchosen differences between him and Bonnie cannot make him as responsible as Bonnie is. The point can be expressed in a slightly different way. If Clyde had Bonnie’s temperamental constitution, and was placed in exactly the same circumstances as Bonnie, then he would be blameworthy, or at least would be a candidate for blameworthiness in virtue of those facts about him. So he is blameworthy, according to the No-­Difference Claim, due to his lack of control over the differences between his and Bonnie’s respective situations. But the Control Principle asserts that Clyde’s degree of blameworthiness cannot be affected by factors which are due simply to luck. Since Clyde lacks control over the difference between his situ­ ation and Bonnie’s situation—since the difference between their situations is just a matter of luck—it must follow that his degree of blameworthiness cannot be affected by this unchosen difference. Since he lacks control over the difference between his actual situation and his counterfactual situation, it will follow that his actual blameworthiness should not be affected, still less determined, by his counterfactual blameworthiness. If we are being urged to take the Control Principle seriously, the unchosen difference between his actual situation and his counterfactual situation ought to protect him against this assignment of blame­worthi­ ness, rather than rendering him vulnerable to it. This conclusion, moreover, is not affected by the direction of travel chosen by the anti-­luckist; it is irrelevant whether anti-­luckists are opting for the Expansion Strategy or the Contraction Strategy, or even whether they are trying to combine both. The Control Principle does not discriminate between these possibilities. It protects Clyde against assignments of blameworthiness which reflect factors beyond his control. It does not matter whether full exposure to these factors would serve to augment his blameworthiness, or to diminish it. In Appendix II, I pursue some further options for establishing compatibility between the No-­Difference Claim and the Control Principle. (To anticipate, I do not think these options can save Zimmerman’s account.) In what remains of this chapter, I want to pursue Zimmerman’s deeper motivations for the Explosion Argument.

4.6.2.  Deeper into the No-­Difference Claim First, let us remind ourselves of the No-­Difference Claim: No-­Difference Claim: The level of blameworthiness which is due to X will also be due to any agent, Y, if the difference between X’s situation and Y’s situation is due to factors which are simply a matter of luck.

128  Restraining Situational Luck In the following part of the discussion, I shall refer to X’s situation as ‘SX’, and Y’s situation as ‘SY’. I shall also make two further assumptions. First, I am going to assume that SY generates a greater scope of blameworthiness than SX: we can, accordingly, regard Bonnie’s situations in all four of these cases as suitable instances for SY, and Clyde’s situations in those cases as suitable instances for SX. Second, I will switch freely between ‘control talk’ and ‘choice talk’. Zimmerman himself tells us that ‘an agent exercises control directly over his choices, and in­dir­ ect­ly over the consequences of his choices’,52 so this policy is not at any variance from his commitments. Now, by assumption, X lacks control over whether he is in SX or SY, and Y lacks control over whether she is in SX or SY. According to the No-­Difference Claim, X and Y are therefore blameworthy to the same degree, even if there appear at first to be substantial moral differences between SX and SY. Why is X held to be as blameworthy as Y? What deeper significance, from X’s perspective, attaches to the fact that the differences between SX and SY are un­chosen? We might start with the fact that X lacks control over whether he is in SX or SY. Other facts follow swiftly from that basic fact. X does not choose to avoid SY. He takes no steps to ensure that he is not in SY. He does not choose not to be in SY. Are these facts relevant? Not immediately so. The central reason why X does not choose not to be in SY is that he is not presented with the choice to be in SX rather than SY, and that should not count against X, for two reasons. First, it was beyond X’s control whether he had a choice to make between SX and SY. He should not be held responsible for his failure to make a choice that he was denied any opportunity to make. (Think of Case Two.) Second, given that X already has a character which is incompatible with the freely willed occupation of SY, it is nat­ ural to surmise that he would most likely choose to be in SX rather than SY if he were given the choice between being in SX and being in SY. His choices will nat­ur­ al­ly reflect the character he has. If it is supposed to be important that X did not choose to be in SX rather than SY, then we must locate a deeper set of considerations which that fact tracks, but which cannot be reduced to that fact. The most likely source of significance attaching to the fact that X did not choose to be in SX rather than SY is that the ability to choose one’s situation is required for moral responsibility. So the deeper message of the No-­Difference Claim seems to be that X did not choose to have the sort of character which is such that, if he were given the choice between SX and SY, he would choose SX. X is held to be just as blameworthy as Y because he lacks deep control over the formation of his character. When we turn to Y, exactly the same considerations apply. Why is Y held to be no more blameworthy than X? What deeper significance, from Y’s perspective,

52  Zimmerman (2002), p. 564.

Exploding the Explosion Argument  129 lies in the fact that the difference between SX and SY is unchosen? First, like X, Y lacks control over whether she is in SX or SY. She does not choose not to be in SX. So, according to the No-­Difference Claim, she can be no more blameworthy than X is blameworthy. Given that Y has the character which is SY-compatible rather than SX-compatible, we can surmise that she would naturally choose to be in SY rather than SX if this was her choice to make. So what is the deeper message of the No-­Difference Claim for Y? As it did for X, it seems once again to consist in the fact that Y did not exercise choice in acquiring the sort of character which is such that, if she were given the choice between SX and SY, she would choose SX rather than SY. Because she did not choose to have a better character than she in fact does, she cannot be more blameworthy than X. This reasoning commits Zimmerman to ending up in a rather different place than just a thoroughgoing interpersonal anti-­luckism. Or, to put it another way, Zimmerman’s anti-­luckism, when its source is fully traced, appears to take the form of a commitment to an extremely demanding responsibility condition, a condition which is arguably impossible to satisfy. The Control Principle is committed to protecting agents from the ascriptions of responsibility which reflect factors beyond their control, but the No-­Difference Claim rests upon commitments which insist upon the power of radical self-­determination of moral personality. These are the kind of commitments which are manifested in Galen Strawson’s ‘Basic Argument’ for the impossibility of moral responsibility.53 The Basic Argument purports to show that moral responsibility is impossible due to the impossibility of authoring one’s moral personality from scratch, or from the ground up. If the ascription is reliable, the No-­Difference Claim rests on premises which purport to show that moral responsibility is effectively impossible.

4.6.3.  Two Final Remarks Two final remarks on Zimmerman’s argument. First, I lack the room here to debate the merits of accounts, such as Strawson’s, which insist on a strong self-­determination condition. So, if Zimmerman is indeed committed to such an account, I am not in a position to criticize him due to any demonstration that such accounts are untenable.54 But his very commitment to such an account does risk breaching the working assumption in moral luck debates that we can establish ordinary blameworthiness conditions for individuals one at a time. As I attempted to establish in the Introduction,55 this working

53  See Strawson (1994), (1998). 54  But see Wolf (1990), Hurley (2003), Part I, Dworkin (2011), ch. 10, and Lang (2015b) for op­pos­ ition to this account of moral responsibility. 55  See Introduction, Section 4.

130  Restraining Situational Luck assumption is needed in order to explore the domain of moral luck without inviting too many distractions from the free will debate. Zimmerman gives the impression that it is moral luck he is mapping, and that his conclusions about the instability of responsibility are driven by the explorations of moral luck in particular. I do not think this ambition can be upheld. In my view, everything is driven, when all is said and done, by his very strong implicit assumptions about individual moral responsibility and freedom. Second, if my analysis of Zimmerman’s argument is sound, we are in fact not left with an unstable concept of responsibility that is capable of sustaining conflicting ascriptions of responsibility and blameworthiness. We have instead uncovered a stable, immovable conception of responsibility: a conception of responsibility which is so demanding as to be incapable of being satisfied on any occasion, or in any possible world. If moral responsibility is impossible, then no one is blameworthy. If no one is blameworthy, then, trivially, everyone will be equally blameworthy, and no one will be more blameworthy than anyone else. But Zimmerman has misled us if he thinks that this result is independently inspired by diligent reflection on the phenomenon of moral luck.

5

Gauguin’s Lucky Escape 5.1. Introduction As I remarked in the Introduction, the moral luck debate that has emerged since the Nagel and Williams ‘Moral Luck’ symposium in 1976 is in fact largely inspired by Nagel, not Williams. Part of the problem consists in the obliqueness of Williams’s argumentative strategy. Most of the essay, by his own admission, is concerned with a more general phenomenon, namely, the rationality or otherwise of retrospective justification for the decisions one makes. It is only towards the end of the essay, in fact, that Williams addresses the specific lessons for morality, as he sees them, and this part of his argument is strangely compressed. Both the nature of the connections between the more general phenomenon of retro­spect­ ive justification and morality, and the resulting nature of the damage which is supposedly inflicted upon morality, remain imperfectly understood. Or so I contend. The argument unfolds as follows.1 Section  5.2 will provide some basic back­ ground to Williams’s argument, concerning both his treatment of moral luck and his attack on the morality system. In Section  5.3, I will outline his central case involving Gauguin—Gauguin’s Escape, as I call it—and note some qualifications. Section 5.4 outlines and then quarrels with three preliminary interpretations of the justification which Williams thinks Gauguin may end up with. In Section 5.5, I introduce a further significant dimension of Williams’s argument, concerning regret and ‘agent-­regret’. The connection between justification and regret yields two further interpretations, though these also prove to be open to weighty objec­ tions. Section  5.6 canvasses two further interpretations, which are concerned with the transformation of Gauguin’s perspective as a result of his success. I believe that these final interpretations manage to capture Williams’s commit­ ments more convincingly than the other interpretations, but they are still vulner­ able to serious objections, including objections which Williams himself, in other contexts, showed signs of endorsing. Much of this material may seem only tenuously connected to the arguments I have been considering across Chapters 1 to 4. It is still worthwhile treating Williams in detail, however, given the very large shadow he has cast over the 1  For a partly overlapping treatment of this material that connects Williams’s work on moral luck with the broader attack on the ‘morality system’ spelled out in Williams (1985), see Lang (2019a).

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0006

132 Gauguin ’ s Lucky Escape moral luck debate. (A further consideration is that, without this detailed exam­in­ ation, the initial claim that Williams’s critical targets are rather different from Nagel’s cannot be properly substantiated.) In addition, however, Williams’s account does contain a significant challenge for the Restricted Account, concern­ ing the role of agent-­regret, which I duly tackle in Section 5.7.

5.2.  Moral Luck and the Morality System How does Williams’s concern with moral luck fit into his larger programme? And where does it fit into the moral luck debate? We might start with a comparison between him and Nagel.

5.2.1.  Williams versus Nagel Both Williams and Nagel appear to have a common critical target. They wish to question the claim that ‘moral evaluations, properly understood, do not apply to those aspects of a situation that are matters of luck’.2 But they proceed differently. In the sorts of case Nagel is principally associated with, it is the degree of agents’ blameworthiness that appears to differ, depending on luck. This concern is clearly manifested by Assassins and Drivers, which were intensively investigated in the earlier chapters. In Drivers, Fred is more blameworthy than Thomas, but on grounds which are ultimately due to luck. The same goes for Imogen and Phoebe in Assassins. Williams is not engaged with such debates—or at least not directly. As I read him, his leading aim in ‘Moral Luck’ is to show, inter alia, that it may be a matter of luck whether moral evaluation applies in the usual way to an agent. True, Nagel’s concerns and Williams’s concerns are partly coeval, and in two ways. First, both Williams and Nagel are concerned with actual outcomes. Assassins and Drivers are initially animated by the fact that, in each case, one outcome is mor­ ally worse than the other. This fact then creates the pressure to acknowledge that, for each pair of agents, one of the agents is more blameworthy than the other, even though the difference between them may be just a matter of luck. Similarly, as we shall see, Williams’s discussion largely revolves around a case where a cer­ tain sort of actual outcome is necessary to generate a justification. Second, if it can be a matter of luck whether a moral evaluation applies to an agent, as Williams contends, then that fact can still be described in ways that appear akin to Nagel’s 2  Williams (1993), p. 252; original emphasis.

Moral Luck and the Morality System  133 concerns. This is because, for Williams, the degree of this agent’s blameworthiness in these circumstances will indeed be affected: his blameworthiness will drop from whatever the usual assignment of blameworthiness would be to zero.3 It should not be doubted, then, that there is a way of assimilating Williams’s concerns to a structure which also accommodates Nagel’s concerns. But this structural similarity should not encourage us to conclude that Nagel and Williams are after the same thing. Another way of putting this point is that Williams’s aim is not to defend the Restricted Account, or something like it. His critical quarry, for the most part, is quite different. So far, we have not been fundamentally concerned with cases involving agents who stand any serious chance of standing outside moral evaluation altogether. Thomas is still blameworthy, but just not as blameworthy as Fred. Similarly, Phoebe is also blameworthy, only less blameworthy than Imogen.4 Nagel’s con­ cerns might therefore be described as being roughly internal to morality, whereas Williams’s concerns are concerned roughly with the limits of morality, on a cer­ tain construal of what morality is. Williams’s leading contention is that certain outcomes, arrived at in certain circumstances, may help to protect agents from the sort of unadulterated moral assessment to which they would otherwise be exposed. Having sketched a case in which an agent can somehow escape morali­ ty’s orbit—we shall turn to the details of this case shortly—Williams’s main busi­ ness is then to re-­evaluate the strength of the moral forces that were supposed to prevent this from happening.

5.2.2.  The ‘Morality System’ What does Williams take morality to be?5 Williams’s particular critical target in ‘Moral Luck’ is a certain version of the Kantian conception of morality. Williams dispenses with much detailed or exegetically exact description of that project: what most matters for his purposes are his assumptions that Kantian morality has universal scope, and issues categorical obligations. Moral requirements apply to agents regardless of whether they conform to those requirements. Kantian moral­ ity does not offer any sort of inducement, or happiness to agents; moral require­ ments apply categorically. Since Kant takes moral requirements to be categorical, and also believes in a strong form of the principle of Ought Implies Can, it follows that these requirements must be such that they can be perfectly followed by the 3  Perhaps, for Williams, blameworthiness is not entirely obliterated, but it is certainly moderated; he is not perfectly clear about this. 4  Except, of course, by the lights of the Strict Liability Account: see Chapter 1, Section 1.2.1. 5  See, in particular, Williams (1981a), pp. 20–2.

134 Gauguin ’ s Lucky Escape agents to whom they apply. But moral requirements cannot be perfectly followed if successful compliance with them depends upon the cooperation of factors which lie beyond agents’ control. So, as a corollary of these other theses, morality must be entirely resistant to luck. Williams argues that the appeal of the Kantian vision goes beyond the claim that it refuses to fudge or distort what we intuitively perceive to be the categorical nature of morality. For Williams, Kantian morality is not just inescapable, but seductive, since it offers us a way of understanding ourselves, and of arranging our relations with others, that is immune to luck. The appeal of that thought requires more than an acknowledgement of the inescapability of morality; it involves advocacy of morality’s fundamental importance. For its advocates, Kantian moral­ ity does not just refuse to sell morality short, but also refuses to sell us short. We can be vindicated in ways bypassing the disturbing vicissitudes of fortune that infect our non-­moral practical lives. That offers us consolation for the unfairness and randomness of the world.6 It gives us both a level playing field to occupy, and the sense that there is no worthier playing field to gather in. In short, moral value is not just a value, to be placed alongside other values; it is more than just a ‘doss-­ house of the spirit’,7 or, as Williams later puts it, ‘merely a consolation prize you get if you are not in worldly terms happy or talented or good-­humoured or loved’.8 It is the value. Morality ‘has to be what ultimately matters.’9

5.3. The Gauguin’s Escape Case Most of Williams’s discussion, and of the subsequent discussion of Williams’s article, is focused on the Gauguin’s Escape case. We need not make a fuss about the biographical accuracy of Williams’s treatment of this fragment of Paul Gauguin’s life.10 In Gauguin’s Escape, a budding painter, Gauguin, abandons his family in order to pursue his artistic dreams in Tahiti. At the time the decision is made, Gauguin does not know whether he will be a successful artist; the likelihood of success ‘has not unequivocally declared itself ’.11 If he is successful, however, then he will be in a significant sense immune to the moral complaints to which he would otherwise be exposed. Moreover, Gauguin’s non-­moral justification for what 6  Williams (1981a), p. 21. 7  Williams (1981a), p. 21. 8  Williams (1985), p. 195. 9  Williams (1985), p. 195. See Athanassoulis (2005) for an extended exploration of Aristotelian and Kantian approaches to moral luck, and the suggestion that may have more in common than is sometimes thought. Nussbaum (1986) also offers an important discussion of ancient approaches to luck and ethics, which for Williams tend to contrast more promisingly with the morality system. 10  In a later commentary on ‘Moral Luck’, Williams (1993), p. 255, is happy to refer to him as ‘pseudo-­Gauguin’. Incidentally, Gauguin—or ‘pseudo-­Gauguin’—enjoys a brief walk-­on role in an earlier discussion: see Williams (1972a), p. 71. 11  Williams (1981a), p. 23.

The Gauguin ’ s Escape Case  135 he does is then supposed to put pressure on the Kantian claim. This is Williams’s strategy in outline.

5.3.1.  Gauguin’s Escape: Five Further Features There are five additional significant details about Gauguin’s Escape, and it is important to draw attention to them at this early stage. They will be revisited at later points. First, Williams does not make the straightforward claim that Gauguin will be justified by his future artistic success, but, rather, and more subtly, that he has a basis for the thought that what he did was justified. If, by contrast, Gauguin fails, then he lacks any basis for that thought.12 Second, Williams tells us that Gauguin might experience two types of failure.13 There will be an extrinsic failure if Gauguin’s talent is not put to the test through accident or misfortune: if he suffers a shipwreck on his way to Tahiti, for example. By contrast, there will be an intrinsic failure if his talent is put to the test but then proves insufficient to sustain his original aim. The contrast between these two types of failure is accompanied by a distinction concerning the nature of Gauguin’s relationship to the justification which, in each case, will elude him. If the failure is extrinsic, then Gauguin will lack a justification, but the failure will not unjustify him. If the failure is intrinsic, then the failure will be deeper; he will be unjustified, not just someone who happens to lack a justification. Williams adds: ‘what would prove him wrong in his project would not just be that it failed, but that he failed’.14 It is not immediately apparent why Williams places such emphasis on these different sources of failure. The interpretations to come will attempt to make some progress on this score. Third, Gauguin is not depicted as simply an amoral agent who cares about nothing except his art, and who is prepared to go any lengths to pursue it. Williams explicitly denies that his version of Gauguin is amoral; he describes him as someone who is ‘concerned about these [moral] claims and what is involved in their being neglected’,15 and who ‘shares the same world of moral concerns’ as the rest of us.16 The risk Gauguin runs, Williams tells us, ‘is a risk within morality, a risk which amoral versions of these agents would not run at all’.17 This more refined characterization of Gauguin is important, for it offers us potential

12  Williams (1981a), p. 23. The final interpretation I consider will pay particular attention to this subtlety. 13  Williams (1981a), pp. 25–6, 36. 14  Williams (1981a), p. 25; emphases added. 15  Williams (1981a), p. 23. 16  Williams (1981a), p. 38. 17  Williams (1981a), p. 38; emphasis added.

136 Gauguin ’ s Lucky Escape ma­ter­ials for sidestepping the looming possibility that Gauguin’s partial exit from the moral world is bound to say much more about him than it does about morality.18 Gauguin’s Escape, upon initial inspection, may strike us as one where Gauguin’s attachment to morality turns out to be incomplete. He may not be wholly amoral, but it turns out that he is not sufficiently attached to morality to prevent him, in certain circumstances, from acquiring the false belief that he can be justified for reneging on morality’s demands of him. This defensive manoeuvre is a standing possibility for defenders of the morality system, and our interpretations of Williams’s argument will need to be aware of it. The challenge for Williams is complex: Gauguin needs to be sufficiently removed from morality’s orbit if his acts are going to teach us something about the limits of morality, but he needs to be sufficiently aligned with morality if his acts are going to teach us something about the limits of morality. Gauguin’s con­ cerns and attitudes must therefore do double duty; they must straddle the gap between morality and something which lies beyond morality. We will see whether any such characterization of Gauguin, however refined or subtle, can pull off this trick. Fourth, Gauguin’s Escape does not attack only the morality system, which is largely concerned with impartial morality, but also implicates wider notions of the ethical, which will ‘include . . . a concern for people directly affected by one’s action, especially those to whom one owes special care’.19 Notice that this descrip­ tion of the broader category of the ethical is actually tailor-­made for Gauguin’s Escape. This case is not concerned with the obligations of impersonal beneficence emphasized by the central representatives of the morality system. Gauguin’s deci­ sion, after all, is to abandon his family, not to ease off on his duties towards s­ trangers. We shall come to see what difference this makes to Williams’s conclusions. Fifth, Gauguin’s Escape is not the only case of moral luck to which Williams alludes. Another example, admittedly treated in less detail, is the Anna Karenina case.20 In Tolstoy’s novel of the same name, Anna’s elopement with Vronsky, and her consequent estrangement from her family, fails to lead to lasting happiness with him. Anna’s life ends, and ends in failure. But Williams’s discussion implies that, had the relationship with Vronsky worked out,21 Anna would have been in a prima facie position to collect, like Gauguin, a justification for what she had decided to do. In that alternative version of the story, Anna would have enjoyed moral luck as well. Though most of the following discussion will be concerned with Gauguin’s Escape, Anna’s case will also be revisited. We should not settle

18  This is a point heavily emphasized by Nagel (1976), pp. 137, 142; and Nagel (1979), p. 28, n. 3. Nagel’s concerns are addressed at Williams (1981a), pp. 36–9. 19  Williams (1993), p. 255. 20  Williams (1981a), pp. 26–7. 21  I leave aside here any worries about the truth of counterfactuals about fictional characters. This issue is discussed at length in Nehamas (1985).

Non-­Moral Value and Self-­Realization  137 for an interpretation of Williams’s argument that fails to accommodate her case as well.22

5.3.2.  Two Constraints Over the next few sections, I will outline seven different interpretations of what Gauguin’s justification can consist in. I take myself to be answerable to two types of constraint. First, I want to identify what, according to Williams, Gauguin’s jus­ tification amounts to, in the light of his various bits of commentary and the other bits of textual evidence available to us. We can call this the ‘internal constraint’, since it is constrained by Williams’s own conception of what he is up to. Second, I am interested in how plausible the interpretation is in its own right, or whether defenders of the morality system targeted by Williams have any reason to be ­troubled by it. This is the ‘external constraint’. I shall attend to these constraints roughly in that order: my procedure, first, is to worry about the internal constraint, and second, to see how the preferred in­ter­ pret­ation fares against the external constraint. It is difficult to advance an in­ter­ pret­ ation which perfectly satisfies the internal constraint, though some interpretations certainly fare better than others. As we shall see, even when we have managed to do that, the questions for Williams are far from over.

5.4.  Non-­Moral Value and Self-­Realization I shall kick off the critical journey by canvassing three succeeding interpretations. The second interpretation builds on, and arises out of, faults in the first in­ter­pret­ ation, and the third interpretation does likewise with the second interpretation.

5.4.1.  The Non-­Moral Value Interpretation In Gauguin’s Escape, Gauguin ends up producing great art. His production of this art then serves to justify him, in a retrospective way, for abandoning his family. This relatively simple line of thought yields a first interpretation. 22  More recent critical discussions of Williams, such as those offered by Manne (2018) and Callcut (2018), have been exercised by the gendered dimension of Williams’s central example. Is Gauguin’s Escape too carelessly or boorishly male? Does it provide a pretext or excuse for men’s selfish abandon­ ment of their responsibilities, or reflect the influence of a culture that tends to applaud this sort of behaviour while hemming in or subordinating women? Certain unwelcome or distracting overtones do seem to attach to Gauguin’s Escape, and it is a real possibility that this case is not the optimal case Williams could have selected to get us to think about what he wanted to establish. Though we must work with what Williams left us, we should also be prepared to locate the deeper story connecting Gauguin and Anna Karenina. Williams himself urges us to do this, after all, and it would be un­char­it­ able not to respond positively to that invitation.

138 Gauguin ’ s Lucky Escape Non-­Moral Value Interpretation: Gauguin’s decision to abandon his family in Gauguin’s Escape is justified by the non-­moral value of what he goes on to produce.

What does the Non-­Moral Value Interpretation have going for it? It boasts two significant advantages. First, it clearly accommodates the importance of what Williams refers to as the ‘determination by the actual’.23 There may be other species of non-­moral value which other agents, in other circumstances, could produce, but great art is the only particular species of non-­moral value that Gauguin stands any chance of producing. Moreover, to be justified, Gauguin must actually produce great art; his intention of producing great art is not enough. Nothing less than the production of such art will give Gauguin a way of acquiring a justification for what he has decided to do. If he fails in this endeavour, then he can expect no justification for the decision he made. The Non-­Moral Value Interpretation has a secondary advantage as well: it serves as an example of the gratitude we should feel that morality does not always get its way.24 On this view, Williams implicitly expects us to agree that perfect conformity to morality will have the effect of inhibiting the generation of signifi­ cant non-­moral goods, such as great art. Morality does not, of course, recoil against the existence of great works of art, and moral agents are typically glad that these works exist. They greatly enrich our lives, and provide us with a source of consolation and inspiration; they manifest human creativity at its most profound and refined. But anyone with some experience of the world should be in a pos­ ition to concur that, if great artists were always morally well-­behaved, they would probably have fallen far short of their artistic potential. The great works of art we now cherish would simply not have been created in the first place. Examples from other domains can also illustrate this point. If efficacious politicians were not pre­ pared to cut corners or act in shady or even ruthless ways, they would probably be less efficacious. We often have reasons to be grateful that they were prepared to sacrifice moral scruple for the achievements of effective policy.25 In short, there is quite a lot going for the Non-­Moral Value Interpretation. Even so, it faces severe interpretive problems. First, it fails to explain why Williams decides to place such heavy emphasis on the distinction between intrinsic and extrinsic failure. If Gauguin does not prod­ uce great art, then it should not matter to proponents of the Non-­Moral Value Interpretation how and why that failure occurred. If it is only the product that 23  Williams (1981a), p. 30. 24  Williams (1981a), pp. 23, 37. 25  Williams (1981a), p. 37, and Williams (1981d). Of course, it might be countered that political goods still constitute a species of moral value. In some of his later work, Williams develops a critique of political moralism which appears to suggest that political value is not straightforwardly continuous with moral value: see the essays collected in Williams (2005). I have my doubts about the force of this critique, but these issues cannot be addressed here.

Non-­Moral Value and Self-­Realization  139 matters, when all is said and done, then it should be a matter of indifference to the Non-­Moral Value Interpretation whether Gauguin’s talent was not properly put to the test, or whether the failure ensues from the mediocrity of the canvases he produces after several years of beavering away in his studio. It will not be in­trin­ sic­al­ly important, as Nagel might put it, whether Gauguin’s failure is owed to bad constitutive luck, or to bad circumstantial luck.26 Whatever the particular source of the failure, there will still be a gap between what he needed to accomplish in order to justify his abandonment of his family, and what in fact he does accom­ plish, and it is only the existence of this gap which determines whether he acquires the relevant justification. The second problem is that Williams explicitly denies that the value of Gauguin’s art is, in fact, doing the requisite justificatory work; he denies, that is, that the Gauguin case exemplifies the view ‘that moral values have been treated as  one value among others, not as unquestionably supreme’.27 This denial is explained by Williams’s emphatic determination to depict Gauguin as a moral agent who is running a moral risk.28 He goes out of his way to emphasize that Gauguin is not amoral; his case is not meant to represent ‘a welcome incursion of the amoral’.29 Now it is still far from clear how we are expected to understand Gauguin’s psychological profile, with its mixture of moral and non-­moral com­ mitments. However we propose to describe it, though, it looks as though the Non-­Moral Value Interpretation fails the internal constraint. It does not accom­ modate Gauguin’s Escape in the right way. A third problem is that the Non-­Moral Value Interpretation does not accom­ modate the Anna Karenina case.30 Had Anna achieved a justification, in virtue of her successful relationship with Vronsky, it would not have been through the pro­ duction of non-­ moral value. The happiness she was hoping to realize with Vronsky was not a gift to the world in the way that Gauguin’s art (or indeed Tolstoy’s novel) was a gift to the world. Anna’s luck, if she had experienced it, was only ever going to be the luck of achieving personal fulfilment.

5.4.2.  The Self-­Realization Interpretation The Non-­Moral Value Interpretation cannot be the full story, then, though it might continue to be part of it. Taking our cue from the Anna Karenina case in particular, we might propose a second interpretation. 26 Nagel (1979), p. 28. Williams (1981a), pp. 26–7, argues that intrinsic luck is not always c­ o-­extensive with constitutive luck. These categories of luck come apart in the Anna Karenina case, for example. But they do pretty much coincide in Gauguin’s Escape. 27  Williams (1981a), p. 37. 28  Williams (1981a), pp. 37–8. 29  Williams (1981a), p. 38. 30 Williams (1981a), p. 37, acknowledges these differences between Gauguin’s case and Anna Karenina’s case.

140 Gauguin ’ s Lucky Escape Self-­Realization Interpretation: Gauguin’s decision to abandon his family in Gauguin’s Escape is justified by his self-­realization, and perhaps, in turn, by the integrity upheld by such a life.

The Self-­Realization Interpretation gets a number of things importantly right. First, it obviously accommodates Anna’s case as well as Gauguin’s Escape. Second, it neatly dovetails with other prominent strands of argument we associate with Williams, concerning integrity and his protests against the various confinements imposed on us by the morality system.31 Third, the Self-­Realization Interpretation is nicely equipped to deal with the breadth of Williams’s target. If Gauguin is retro­spect­ively justified by his self-­realization, then the demands and expectations of the broadly ethical life, not just the obligations of the narrower life envisaged by the morality system, will also be endangered. We can add a fourth point, which looks at things from the other direction. The Self-­Realization Interpretation attends to a concern we might expect a decent account of human morality to take seriously. Kantian morality may satisfy us by offering us a form of value, and thus a form of self-­affirmation, which is insulated from the unfairness of the world, but there are of course other forms of self-­affirmation and self-­realization which depart from that model. Gauguin’s Escape exemplifies the appeal of these alterna­ tive models of self-­realization, and it is not unreasonable to expect broader accounts of morality, if not the Kantian morality system itself, to seek to accom­ modate these other forms of self-­realization.32 The Self-­Realization Interpretation is not without its problems, however. There are two main issues. First, the Self-­Realization Interpretation can enjoy no deep connection to the success condition which is so central to Gauguin’s Escape. Other implications immediately ensue from that failure: the Self-­ Realization Interpretation will struggle to find gainful employment for the role of luck and the importance of retrospective justification, and it will not explain the significance of the distinc­ tion between intrinsic and extrinsic failures of luck. Imagine a further fictional character loosely based on Gauguin, whom we may call Gauguin 2. Gauguin 2 wishes to spend his life painting, rather than producing valuable paintings as a result of this life. If Gauguin 2 is left free to paint, then he expects to achieve self-­ realization. Of course, he may not achieve self-­realization. Perhaps he will tire of it all. Even so, this risk seems small, and appears to fall far short of the success condition offered by Williams’s Gauguin.

31  For the classic discussion of integrity, see Williams (1973), esp. pp. 93–118. 32  Even then, we might be left with a complaint about morality, based on the thought that we are somehow being forced, humiliatingly, into pleading with morality to provide us with room for these forms of self-­fulfilment: see Wolf (2012) for this intriguing line of argument.

Non-­Moral Value and Self-­Realization  141 Second, the Self-­Realization Interpretation does not easily uphold Williams’s avowed interest in agents who are running moral risks. Gauguin 2 has the choice between moral duty and creative effort, however mediocre the results of the latter are likely to be. But agents who ignore their moral obligations in order to chase self-­realization are often going to be agents who have already, in effect, negotiated their exit from morality. This is particularly the case if there is not any substantial provision for luck. Gauguin 2 does not have to wait to see what happens: he is ready to leave the moral fold now. But that ‘wait and see’ element is important to Williams’s calculations, since it allows him greater leverage to insist that these agents, at the time their initial decisions are taken, are still firmly in the moral fold, and are committed to remaining within it unless their actions have certain consequences. Should Gauguin 2 be offered sanctuary by the Self-­Realization Interpretation? Here is a possible objection: if Gauguin 2 were indifferent to the value or quality of the paintings produced, then he would risk being disconnected in a deeper sense from the point of artistic activity, which at some level must involve the commitment to the creation of serious art, not just random and casually distrib­ uted daubs of paint on a canvas. There are two replies to this objection. First, Gauguin 2 might still be aiming at producing something better than, in his heart of hearts, he suspects he is capable of, but this is the life which he might nonetheless still enjoy most. It is not unintel­ ligible for him to value a life of painterly effort, even if only mediocre work is generated by his efforts. We can imagine that he likes turning up to the studio every day and seeing how far he can get, and that he is temperamentally well equipped to deal with the fact that he never makes all that much progress.33 Second, even if Gauguin 2 is disconnected from the deeper aim of artistic effort, this is no automatic barrier to self-­realization, unless Williams is prepared to say more. Being a bad painter, even if it is the activity of painting that one is pursuing, does not prevent an agent from achieving self-­realization. A much more outland­ ish case, but one which is familiar from the well-­being literature, is the Rawlsian grass-­counter, who elects to spend his days counting blades of grass despite his abundance of mathematical talent.34 Grass-­counting may offer this agent self-­ realization, though it is difficult to see how this activity is connected to any dis­ tinctive human excellence. Now even in the grass-­counter case, there is still the

33  Another possible trajectory for Gauguin is sketched in Section 5.6.2. 34  See Rawls (1971), pp. 432–3. The challenge I am considering here is not directly concerned with the adequacy of subjectivist accounts of well-­being. The problem is that the grass-­counter case seems to fall within the writ of the Self-­Realization Interpretation, but outside the scope of the cases Williams is interested in.

142 Gauguin ’ s Lucky Escape element of risk: perhaps he will stop enjoying counting blades of grass.35 But this element of risk seems small, and it appears in any case to create the wrong kind of hostage to fortune.

5.4.3.  The Valuable Self-­Realization Interpretation A third interpretation attempts to preserve some of the advantages of the Self-­ Realization Interpretation, while avoiding its weaknesses. Valuable Self-­Realization Interpretation: Gauguin’s decision to abandon his fam­ ily in Gauguin’s Escape is justified by the combination of the self-­realization it provides for him and the non-­moral value produced by the activity that earned him this justification.36

The Valuable Self-­Realization Interpretation represents a fusion of the Non-­Moral Value Interpretation and the Self-­Realization Interpretation. The Valuable Self-­Realization Interpretation winnows out some problems, but generates new problems in turn. It excludes the grass-­counter case, as well as Gauguin 2, from the justificatory sanctuary, but it also appears to exclude the Anna Karenina case. Perhaps we might attempt to finesse the value condition; valuable forms of self-­realization might encompass distinctly human forms of excellence or endeavour. That would allow the Valuable Self-­ Realization Interpretation to accommodate both Gauguin’s Escape and the Anna Karenina case. But we would need to finesse the condition with some care to ensure that the various elements of Williams’s argument—the roles for luck, retrospective justifi­ cation, and the differences between intrinsic and extrinsic failure—were being given sufficiently substantial roles. I do not doubt that some progress can be made on this score, or that the final interpretation of moral luck will be an interpretation that is largely coincident with the Valuable Self-­Realization Interpretation. Yet this interpretation is largely a place-­holder, rather than the interpretation which fully explains what Williams thinks he is arguing for. As it stands, it offers no real illumination of how Gauguin can be characterized in ways which incline us to the view that morality is limited, rather than that he is limited. And it does not have much to say about the connec­ tions of these issues to another important constituent in Williams’s argument, concerning regret. I turn to this issue next. Its introduction is already overdue.

35  But if he liked it in the first place, who can tell? We are already well outside the range of ordinary human possibilities. 36  Roger Crisp suggested this interpretation to me.

Justification and Regret  143

5.5.  Justification and Regret 5.5.1.  Gauguin’s Justification We need to take a closer look at the phenomenology, or how things are supposed to stand on the ground, when Gauguin has achieved success, and when, as a result, he acquires a justification for the decision he made. We already know that, on Williams’s view, Gauguin’s success has somehow catapulted him beyond the orbit of ordinary moral criticism. Gauguin is now no longer vulnerable to the unadulterated forms of moral criticism to which he would have otherwise been exposed. But we have done little to examine what this justification actually achieves for Gauguin, or what psychological difference it should make to him. This is what I want to investigate now. What, in acquiring justification, is Gauguin supposed to gain? Is he envisaged as no longer caring about abandoning his family? Or is he someone who is no longer afflicted by guilt, though not necessarily drained of concern? Alternatively, does his success mean that he now possesses the right to no longer care about abandoning them, or the right not to feel guilt for abandoning them, even if in fact he does continue to experience these feelings? Or are most of the implica­ tions in fact relevant to his family, rather than to him? Does it now follow that they no longer have the right to complain about his abandonment of them?37 Perhaps, even though his family were initially wronged, they fail to take into consideration the fact that Gauguin is justified in what he has done. But Williams is emphatic that Gauguin’s success does not exempt him from his family’s unhap­ piness: they are entitled to go on complaining. He makes this claim more than once.38 Williams does not, then, entertain any analogue of the ‘justification defeats liability’ claim, according to which an agent is not liable to moral sanc­ tions—in this case, moral blame—if he enjoys an overall justification for what he does.39 Joseph Raz proposes that Williams’s stance on this issue is best explained by his ethical anti-­realism, rather than his specific views on moral luck.40 But it seems to me that the metaethical issues and the normative issues cannot be so neatly sep­ar­ated. First, even if it is the metaethical theses which ultimately explain why Gauguin’s justification does not entail the removal of his family’s right to complain

37  Of course, in Gauguin’s case, the physical distance might make some difference: Gauguin does not have to face tense exchanges by Skype or Zoom, and communications are likely to be more spor­ ad­ic. But this contingent and relatively superficial ‘out of sight, out of mind’ feature of the Gauguin case cannot disperse these deeper questions. 38  Williams (1981a), pp. 23–4, 36–7. 39  See McMahan (2009), pp. 38–51, for a discussion and defence of this claim in the context of self-­defence and war. 40  Raz (2012), p. 159.

144 Gauguin ’ s Lucky Escape about his treatment of them, the normative or first-­order issue of where that leaves Gauguin does not become any less interesting, and the questions we can raise in this connection are not displaced. If his wife and children are complain­ ing, and do not lack the right to complain, how is Gauguin benefited? What can he say to them? What is the point of the justification he has gone to such lengths to secure if the fact that he has it gives them no reason to retire their complaints about him? Perhaps Gauguin, armed with his justification, is supposed simply to switch off at some point. But we might wonder how he would be able to do this, if they are still complaining, and do not lack a justification for doing so. Second, the meta-­normative source Williams himself invokes in this connection concerns his views on ethical consistency.41 Of course, the denial that morality issues a consist­ ent set of obligations does undermine one of the major platitudes about morality which moral realists have traded upon. But again, the interest for us, given these pressures on ethical consistency, lies in which bit of psychological space we can then expect to find Gauguin. Even if Gauguin is justified in what he does, his fam­ ily is justified in complaining about what he has done, and we need to know what difference that makes to him: what is the value to Gauguin of having a justifica­ tion, as opposed to lacking one?

5.5.2.  Two More Puzzles These questions are actually sharpened by Williams’s remarks in his later ‘Postscript’, in which he attempts to clarify the argument presented in ‘Moral Luck’: . . . what is the point of insisting that a certain reaction or attitude or judgment is or is not a moral one? What is it that this category is supposed to deliver? . . . [I]nvoking this category achieves absolutely nothing, unless one has some account of the singular importance of morality in this restricted sense. I still cannot see what comfort it is supposed to give to me, or what instruction it offers to other people, if I am shunned, hated, unloved, and despised, not least by myself, but am told that these reactions are at any rate not moral.42

In the present context, we should be puzzled by these remarks. In fact, they give rise to two problems. First, if it is the content of the complaints coming Gauguin’s way that are all-­ important, rather than their fittingness to count as examples of, specifically, moral criticism, then we face additional pressure to explain how and why Gauguin is benefited by his acquisition of a justification that reduces his vulnerability to 41  See Williams (1981a), p. 31, and Williams (1972b). 42  Williams (1993), p. 254; original emphases.

Justification and Regret  145 moral criticism. It appears no longer to matter which specific register, whether moral or non-­moral, his family’s complaints are best interpreted as falling in. However one describes their reactions, they will be aggrieved, and Gauguin is going to have to come to terms with that. What is his defence going to be, and what is the value of that defence to him? Second, if we can show after all that it is important that Gauguin’s justification has somehow defeated the full undiluted moral force of his family’s complaints, it does indeed seem to follow, contrary to Williams’s official stance, that there is something special about morality. There must be something special about moral­ ity if there is something special about having a defence against undiluted moral criticism—which is precisely what Gauguin is supposed to gain if he succeeds in his artistic endeavours.

5.5.3.  Agent-­Regret Williams’s discussion of regret, and agent-­regret in particular, might help us to pick a path out of this maze. While the general, constitutive thought which lies behind expressions of regret is something like ‘How much better if it had been otherwise’,43 Williams is especially interested in a more specific form of regret, which he calls ‘agent-­regret’. Agent-­regret has a first-­person character, concerning what the agent has done, perhaps not voluntarily, but is further distinguished by certain attitudes and emotional inflexions. Williams’s central example of someone who experiences agent-­regret is the lorry-­driver who accidentally, and through no fault of his own, runs over and kills a child.44 The lorry-­driver is not to blame, but he is naturally anguished by what he has done. He experiences, in acute form, feelings of regret: he wishes that he had not done what he did. He wishes he had not been at the wheel when the child crossed his path. He has a sense of involve­ ment with the fatality in ways from which he cannot easily detach himself, and which go beyond the more impersonal sort of sadness that a bystander might experience. An important lesson provided by the lorry-­driver case, as Williams sees it, is that it exposes the realm of strictly voluntary agency, so cherished by the Kantian morality system, as a relatively superficial one.45 Our various involvements with the world cannot be so primly contained. But the role which agent-­regret plays in our attempts to make sense of Gauguin’s Escape actually places Gauguin in a strik­ ing contrast with the lorry-­driver: unlike the lorry-­driver, Gauguin does not regret what he did, because his project turned out successfully. It is the absence of this sort of regret, then, which is meant both to justify him in what he did, and to 43  Williams (1981a), p. 27. 44  Williams (1981a), p. 28. 45  Williams (1981a), p. 29. See also Williams (1995a).

146 Gauguin ’ s Lucky Escape explain why this justification will presumably provide him with some sort of psychological protection against the awareness that he has imposed these costs on his abandoned family. There is a further feature of Williams’s discussion which fills out Gauguin’s moral-­cum-­psychological profile in a promising way. Gauguin’s profile must somehow combine his moral sensitivity with the intelligibility of his departure, in certain circumstances, from the grip of morality. The added feature concerns the conditional nature of Gauguin’s agent-­regret: we are told that Gauguin will experi­ ence agent-­regret, unless his project turns out successfully.46 As a result, there need be no denial that Gauguin is still involved in the moral world. The fact that he will be left with acute regrets if his project is unsuccessful, with nothing to place alongside those regrets, confirms that he is still a member of that world. Yet the fact that he will be unable to regret what he did if his artistic project turns out successfully is also evidence that, despite his engagement with moral concerns, he may yet be able to escape from the clutches of the morality system.

5.5.4.  The Absence of Regret Interpretation These reflections suggest a fourth interpretation. Absence of Regret Interpretation: Gauguin’s justification of his decision to aban­ don his family in Gauguin’s Escape is constituted by two facts about that deci­ sion: first, his decision being such that he knows that he will not regret making it if he is successful; and second, the decision being such that he knows that he will regret making it if he is unsuccessful.47

The Absence of Regret Interpretation represents real progress with respect to the composition of Gauguin’s moral personality; it is sensitive to the roles of luck and retrospective justification; and it also provides some much needed further psy­ chological details about the benefits of the justification which Gauguin hopes to acquire. All of this indicates that the Absence of Regret Interpretation represents an improvement over the Valuable Self-­Realization Interpretation. However, one immediate problem with the Absence of Regret Interpretation should persuade us to seek a more refined version of it. The problem is that Williams’s confident claim that Gauguin’s agent-­regret is conditional does not explain why Gauguin’s regrets do not spill beyond this

46  Williams (1981a), p. 30. The same goes for Anna Karenina. 47  If the failure is extrinsic, though, then it may be regret, rather than agent-­regret, which Gauguin comes to experience. The same qualification should attach to subsequent interpretations. Thanks to Jake Wojtowicz for this point.

Justification and Regret  147 conditional structure. The very neatness of this conditional structure is at odds with Williams’s usual insistence on psychological complexity and emotional loose ends. Williams’s own work has done a great deal to encourage us to be suspicious of such neatness. How does Gauguin know that he will not experience such regret? How can he be so confident that his regrets can be primly contained within this conditional structure? Gauguin seems, at the time when the original decision is made, to be in no better a position to predict that he will be uncon­ taminated by regret for what he has decided to do than to predict that his artistic career will be successful. In an essay which invests so extensively in the partial unknowability of the future, it seems strange that Gauguin should be in a position to make agent-­regret conditional in this way. Now perhaps Gauguin is banking on both things: on being both artistically successful, and on the cauterization of his regrets as a result of being successful. Still, he is not in a position to know that artistic accomplishment will lead to the cauterization of his regret. He does not know that success will cure him of regret, and, for that matter, he does not know that success is actually needed for the removal of regret. Upon closer examination, these things look like independent variables, not distinct elements in a single structure that contains them both. If the success and the cauterization of regret are somehow contained within a single structure, it is puzzling how Gauguin can be in any position to know about it. Correlatively, if Gauguin thinks he is in a position to vouch for this conditional structure, and to know that he will be immune to regret if he succeeds as a painter, then it is natural to suspect that he has already washed his hands of his family. He has already disinvested in them.48 He will then be ripe for Nagel’s worry that this Gauguin, however interesting in other ways, does not reveal morality to be limit­ed in the way that Williams supposes. The behaviour of this Gauguin does not challenge morality’s authority, but rather confirms that he was not prepared beyond a certain point of self-­sacrifice or personal frustration to continue to acknowledge morality’s authority. Subsequent events reveal that morality only ever enjoyed a tenuous and incomplete hold over him.

5.5.5.  The Modified Absence of Regret Interpretation These are serious worries. It seems to me that a revised version of the Absence of Regret Interpretation should concede that Gauguin is not really in a position to affirm the conditional structure of his agent-­regret in advance. So let us imagine that Gauguin, buoyed up by the success of his project, does not in fact experience regret over the decision he made, and that this lack of regret is in fact explained by

48  This was a major source of concern about the Self-­Realization Interpretation.

148 Gauguin ’ s Lucky Escape his success. Gauguin will not in fact be in a position to know this at the time he makes his initial decision, but he will be in a position to benefit from his success, and from the fact that his success causes his regret to evaporate, if this is indeed what happens to him. So Gauguin’s immunity to undiluted moral criticism will be owed to two things: first, the evaporation of regret; and second, the accomplish­ ments which provide him with his justification, and which cause his regret to evaporate. If we assume these facts, we will have a fifth interpretation. Modified Absence of Regret Interpretation: Gauguin’s justification of his decision to abandon his family in Gauguin’s Escape is constituted by the fact that he does not regret that decision, if he is successful, in a way which is due to this success. If his decision does not lead to artistic success, then he will be left with only regrets for the decision he made.

The Modified Absence of Regret Interpretation keeps the worry about conditional justification at bay. It does not pretend that Gauguin’s decision already embeds this conditional structure. The proof is in the pudding, not in the recipe which he concocts in advance. Nonetheless, the Modified Absence of Regret Interpretation is still open to other worries, which are also worries for the original Absence of Regret Interpretation. To further explore the Modified Absence of Regret Interpretation, it may be useful to pursue a more extensive comparison between Gauguin and the lorry-­ driver. The fact that the lorry-­driver does not lack justification for what he did does not spare him agent-­regret for what has happened. The driver is blameless, but that fact alone will not render him immune to feelings which will be phenom­ enologically akin to guilt or moral anguish.49 The driver’s sense of himself as an individual whose path through the world involves significant interaction with other individuals, both intimates and strangers, is surely bound to spill over into a feeling of responsibility for the outcomes he helps to cause. But if this is the case, we suddenly have the makings of a pairwise comparison between him and Gauguin which is not the one that Williams had in mind. Gauguin, Williams tells us, does not lack justification for what he does, if he is successful. If he has such a justification, then he is not blameworthy in the normal way. But he might still be vulnerable to a form of agent-­regret. After all, he was the one who abandoned his family, even if, as things have turned out, he was not lacking in justification for doing so. Similarly, the lorry-­driver kills the child, even

49  As we have already seen, in Chapter 1, Section 1.5.3, Enoch (2012) makes sense of these agent-­ regret intuitions by pressing hard on the distinction between an agent’s being responsible for X and that agent’s taking responsibility for X. I have already expressed some reservations about Enoch’s views. See, further, Section 5.7.

Justification and Regret  149 if, as things stood, he was not to blame, and did not lack a justification for being at the wheel when the fatality occurred. So why doesn’t Gauguin, like the lorry-­ driver, also experience agent-­regret? Williams invites us to look at the lorry-­driver through the lens of voluntary agency. But why not look at both the lorry-­driver and Gauguin through the lens of justification instead? If the lorry-­driver’s possession of a justification does not spare him agent-­regret, then what prevents us from reaching the same conclusion about Gauguin? It would seem that Gauguin’s having a justification should not make all the difference. This creates an interpretive puzzle. One asymmetry between Gauguin’s situation and the lorry-­driver’s situation is that, while it would be clearly inappropriate to pin any blame on the lorry-­driver, Gauguin can still be blamed by his family. But this particular asymmetry between Gauguin and the lorry-­driver does not imperil the deeper symmetry between them proposed by this line of thought. If anything, it makes it even more tempt­ ing to affirm that symmetry. If we expect the lorry-­driver to feel agent-­regret, despite his blamelessness, then we would if anything appear to have more reason, rather than less reason, to expect Gauguin to feel agent-­regret as well, especially in light of the fact that he can still be deemed blameworthy by his family. Presumably Williams’s answer would be that these agents stand in different relations to justification. The lorry-­driver was blameless, and was acting with jus­ tification at the time the fatality took place. But, given what the lorry-­driver has caused to happen, he wishes things had turned out differently. He wishes that he had not been at the wheel when the fatality occurs. He would, as it were, willingly trade in his justification for a state of affairs in which he did not have to appeal to it. By contrast, Gauguin does not seek to trade in his justification. His justifica­ tion, after all, is constituted by his absence of regret. He thinks he has been justi­ fied in the decision he initially made, which is just another way of saying that he does not regret making it.

5.5.6.  Means, Ends, and Regret This response does not remove all the questions we might ask. Gauguin’s decision imposes costs on his family, and they do not lack the right to complain about these costs. Gauguin might not regret his decision all things considered. But what about the means he employed to his end? Does he regret them? On the one hand, his end had to be secured by the means he employed. It was not going to be realized just by itself, without taking deliberate steps to achieve it. So perhaps, given this relationship between means and ends, Gauguin regrets nei­ ther the end for the sake of which he acted, nor the means which facilitated the achievement of that end. On the other hand, the means he took are separable from the end for the sake of which he acted, and may qualify for attitudes that

150 Gauguin ’ s Lucky Escape differ from the attitude Gauguin takes towards his end. While Gauguin does not regret his decision all things considered, or experience all-­in regret, that still leaves room for him to regret the means he took to achieve his end.50 Which resolution of this question does the Modified Absence of Regret Interpretation favour? There is a problem either way. Imagine that Gauguin does not regret his decision to abandon his family, since he has achieved what he set out to do, but he does regret the means he took to achieve that end. That puts Gauguin in an awkward relationship with morality—and indeed with himself. If he regrets the means he took, then he can still count as falling within morality, but now his case does not provide clear evidence of any escape from morality. Rather, this Gauguin risks psychological incoherence: he does not regret acting for the sake of the end, but he does regret the means he took to that end.51 We should not complain about the mere fact that Gauguin experiences conflict and ambivalence. However, some sources of conflict and ambivalence are going to say more about the agents who experience them than about how things are, generally speaking, in the practical realm. If Gauguin, both at the time he makes the decision and also in retrospect, thinks that there are strong moral reasons not to abandon his family, such that he regrets taking these means to his end, then it is difficult to square these thoughts with the further thought that he was justified, after all, in aban­ doning them. Imagine now that the means and ends are more tightly stationed together within a single justificatory structure: because Gauguin does not regret the end for the sake of which he acted, he also fails to regret the means he took to that end. But now it is difficult to avoid the thought that, at the time when the decision has supposedly been retrospectively justified, Gauguin has simply withdrawn his original concern for them. It is also difficult to avoid the suspicion that, even when the original decision was made, the sacrificable status of his family had already been budgeted for. On this showing, Gauguin is not as firmly stationed in the moral fold as Williams wishes us to suppose. There is another problem with the Modified Absence of Regret Interpretation, concerning the external constraint rather than the internal constraint. The failure to experience regret may be simply a poor guide to whether a decision has been justified. In other words, even if the Modified Absence of Regret Interpretation has managed to make some progress with respect to the internal constraint, how does it turn in a credible performance against the external constraint? How can these agents’ beliefs that they have acquired such a justification often fail to say more about them than it does about morality? The appeal in the first instance is to 50  Here I am indebted to Wallace (2012). I also borrow the phrase ‘all-­in regret’ from Wallace. 51  The status of means-­end reasoning is, of course, complicated, but it seems to me that something like Raz’s ‘facilitative principle’ may fit the bill: see Raz (2005). I cannot firmly pronounce on these issues here: see, for example, Broome (2013), Kiesewetter (2017), Wedgwood (2017), and Lord (2018) for suitably detailed treatments.

Regret and the Transformation of Perspective  151 a psychological fact: should these circumstances come to prevail, then Gauguin will no longer come to regret what he did. So he is off the hook. But why is he off the hook? He may think that he is off the hook, as indicated by the absence in him of profound psychological anguish. But the morality system does not depend on the actual acknowledgement of its authority. The fault may lie in agents, rather than in the demands that morality makes of them. While there is plenty of evidence that Williams is prepared to challenge the categorical nature of moral requirements,52 he can hardly fail to agree that, at least in some circumstances, not caring about morality is going to reveal more about the agents who do not care about it than it does about the limited authority of morality. What Williams needs to do is to convince us that Gauguin stands on the right side of this line. More is required to convince us that Gauguin’s failure to experience regret is not, in this instance, an example of a moral blind-­spot or deterioration, rather than a demonstration that one’s vulnerability to ordinary moral assessment can be genuinely a matter of luck.53

5.6.  Regret and the Transformation of Perspective 5.6.1.  The Transformation Interpretation One problem with the Modified Absence of Regret Interpretation is that it im­pli­ cit­ly assumes a static moral psychology: it is not an easy matter to make sense of the description of Gauguin as someone who is morally attuned to morality, whilst also arranging, though conditionally, his departure from it. A further element in Williams’s argument may be in a position to help. This further element concerns the transformation of Gauguin’s perspective as a result of the success of his project. Gauguin may lack access to a perspective that can justify, or even fully make sense of, his abandonment of his family at the time of his original decision. That perspective is acquired only if and when his artistic project bears fruit. So, though Gauguin can only deliberate from here, in his pre­ sent circumstances, he can still anticipate, if he is successful, an alteration in his

52  See, most famously, Williams (1981c); Williams (1995b) is also relevant. 53  See Harman (2009), and also Wallace (2012), for a powerful argument that the absence of regret can be an unreliable guide to justification. The case Wallace pursues at length is the ‘Non-­Identity Problem’ associated with Parfit (1984), ch. 16: specifically, the case of the fourteen-­year-­old girl who gives birth to a child raised in adverse circumstances, due to the mother’s extreme youth. Wallace observes that this case has little to do with luck, since even at the earlier point it is perfectly foreseeable, provided it turns out that her child has a life worth living and she enjoys a healthy relationship with him, that the fourteen-­year-­old girl will be unable, when she is older, to regret her original decision. But the fact that she cannot subsequently regret that decision certainly does not show that it was the right decision to make at the time, prior to the formation of this relationship. Wallace’s argument is interestingly extended in Wallace (2013).

152 Gauguin ’ s Lucky Escape ‘stand-­point of assessment, [which] will be from a life which then derives an important part of its significance for him from that very fact’.54 Yet the materials for the justification that he eventually ends up with will not be in place until this transformation in his perspective actually occurs. It is also a matter of luck whether he will end up with this transformation of perspective. At the time he makes his original decision, he is not in a position to know.55 This additional ingredient in Williams’s argument yields a sixth interpretation. Here is a first pass at it: Transformation Interpretation: Gauguin’s decision to abandon his family in Gauguin’s Escape is justified by his evaluative transformation, which secures for him a new perspective that releases him from all-­in regret about what he did.56

The Transformation Interpretation clearly provides for luck, since the trans­form­ ation in Gauguin’s perspective is a matter of luck. This interpretation also pro­ vides to some extent for the significance of the distinction between intrinsic and extrinsic failure. If Gauguin suffers bad extrinsic luck, the thought that he might have been right to attempt his project is still available to him, since this thought can be accompanied by the thought that there is some justifying perspective that he might have successfully occupied. If he does not occupy this perspective, then he will lack a justification, but he will not be straightforwardly unjustified. If he suffers bad intrinsic luck, by contrast, then he will know that he was simply wrong to think that there was any such perspective that he might have occupied, such that he was justified in what he did, consistently with him having these projects and being the sort of person he was in the first place. How about the relationship between means and ends? Again, the Transformation Interpretation can shed some light on these issues which goes beyond that offered 54  Williams (1981a), p. 35. 55  Salow (2017) suggests that we can properly distinguish between what he calls ‘restricted’ and ‘unrestricted’ interpretations of retrospective justification. Unrestricted retrospective justification allows retrospective justification to display a sensitivity to both the values an agent acquires later on and the factual information available to her only later on, whereas restricted retrospective justification permits sensitivity to the values she later acquires, but not to the factual information that she will acquire only later. By concerning ourselves only with restricted retrospective justification, we can eliminate any uncomfortable closeness between Gauguin’s Escape and cases of fluky and reckless gam­ bles that happen to turn out successfully (2017, p. 11). It seems to me, in response, that the division between new values and new factual information is murkier than Salow thinks it is, since whether new values achieve any uptake in an agent may depend on factual alterations whose obtaining the agent at an earlier time is not in any position to know about. This is actually obvious from the central case: Gauguin, for example, can only endorse his life as an artist if he actually succeeds in producing worthwhile art, but he does not know at the time of the original decision whether he will succeed in this venture. While Salow is aware of these issues (2017, p. 15, n. 13), I confess to finding his treatment of them unclear. 56  There have been significant recent additions to the literature on transformative experience. See Paul (2014), Chang (2015), Harman (2009), Kauppinen (2015), Barnes (2015), and Shupe (2016) for interesting examples of them. Here I will stay focused on Williams’s treatment of the issues.

Regret and the Transformation of Perspective  153 by the Modified Absence of Regret Interpretation. Here we need to distinguish between the roles and possibilities associated with each perspective. Gauguin’s original perspective is one which reflects a genuine concern for the moral costs of what he decides to do. Now that perspective can endure, since there is no reason to deny that Gauguin still enjoys some sort of imaginative and emotional access to it, even after the transformation of his perspective as a result of his success. So, from this original perspective, Gauguin regrets the means he employed to his end. From the new, transformed perspective, however, Gauguin is in a position to affirm both the justifiability of the end he pursued, and the means he took to attain it. Thus the Transformation Interpretation avoids psychological incoherence, because the ends are connected to the means in a coherent way under each perspective. According to the earlier perspective, the means are morally problematic, and the end is problematic as well. According to the later perspective, the end is vindicated, and the means taken to achieve it, in light of that fact, are also acceptable. The Transformation Interpretation also allows Gauguin to be depicted as a full member of the moral community, at least initially, given his original perspective. Moreover, it makes sense of what his justification achieves for him, given the arrival of his new perspective. At this later point, he endorses both the end, and the means he took. He may continue to be concerned for his family—the Transformation Interpretation does not make the justifiability of what Gauguin does depend on his ability to switch off or to acquire an indifference to them—but these concerns should not get him to think that, after all, he was wrong to take these means to his end. Gauguin is still likely to experience some psychological ambivalence, as he shuttles between the two perspectives which are available to him, but this is not a refutation of the Transformation Interpretation. It simply confirms that things have become emotionally and psychologically complex. The Transformation Interpretation also allows us to make sense of the subtlety of Williams’s formulation of what Gauguin’s success will achieve for him: it will give Gauguin a basis for justification, rather than, more simply, a justification. Having a basis for justification is consistent with his ability to grasp, and to regret, the costs which his decision has imposed on his family.

5.6.2.  The Improved Transformation Interpretation The Transformation Interpretation can be restated to make these features more explicit: Improved Transformation Interpretation: Gauguin’s decision to abandon his fam­ ily in Gauguin’s Escape is justified by his evaluative transformation, which (a) secures for him a new perspective vindicating his end, releasing him from all-­in regret about what he did, as well as directing him not to have all-­in regret for the

154 Gauguin ’ s Lucky Escape means he took to his end, and (b) supplants his original perspective, according to which the end for the sake of which he acts is not yet vindicated, and which renders the means he takes to that end morally problematic.

The Improved Transformation Interpretation is probably the most refined in­ter­ pret­ation we can recover from Williams’s argument, and this is indeed the point, for us, where the interpretive trail will end. Unfortunately for Williams, it remains vulnerable to certain problems. One problem with the Improved Transformation Interpretation arises out of another possible trajectory for Gauguin. Let us call this version of our central character Gauguin 3. Imagine that Gauguin 3 devotes his life to artistic efforts which come to very little. At first, he is disappointed with his returns. It looks as though he will be left with only regrets for what he has done. But then he comes to the view that the journey was still worthwhile, even though the destination was not what he thought it would be. As a result, his ultimate regrets evaporate. The perspective of Gauguin 3 has been transformed: at first he thinks the decision will be justified only by the results, but now he thinks that it is the quest for results, whether he attains them or not, which is all-­important.57 So Gauguin 3 qualifies for justificatory refuge under the Improved Transformation Interpretation. Is this the right result? It offers little scope for Williams’s emphasis on the ‘determination of the actual’, and it does little to accommodate the thought, con­ veyed by the original Gauguin case, that cases of moral luck are cases in which moral commitments can be defeated only by non-­trivial achievements. Gauguin 3 will have sidestepped these challenges by simply revising his thoughts about what might end up doing the justificatory work. It need not be conceded that a mere effort of will can produce this transformation in the perspective of Gauguin 3. That would indeed look cheap—an instant recipe for moral backsliding and ex post facto rationalization. But even if Gauguin 3 can boast the luck of having his perspective transformed, given the fact that this transformation of perspective occurred without his voluntary efforts, his case does not provide an obviously strong challenge to the morality system. The lesson readily generalizes. Not every transformation of perspective can expect to be taken equally seriously by the morality system. We can have de­teri­or­ ations of perspective that morality should dismiss, as well as alterations of per­ spective that might issue it with a sturdier challenge. In other work, addressing slightly different concerns, Williams obliquely appears to agree.58 In Ethics and the Limits of Philosophy, for example, when he is engaged in surveying the possi­ bilities for grounding ethics in well-­being, he strikes the following cautious notes about ‘self-­validating’ changes in one’s evaluative orientation: 57  The reasoning of Gauguin 3 has a MacIntyre-­style inflexion: see MacIntyre (1981). 58  See Williams (1981b); and Williams (1985), pp. 40–3.

Regret and the Transformation of Perspective  155 If an agent does not now acknowledge that a certain change would be in his interest and if, as a result of the change, he comes to acknowledge that it was in his interest, this will show that the change was really in his interest only on con­ dition that the alteration in his outlook is explained in terms of some general incapacity from which he suffered in his original state, and which has been removed or alleviated by the change.59

Williams’s principal concern in this passage is with induced changes in individu­ als’ outlooks, and with the adequacy of the defence that these changes often merely uncover these individuals’ ‘real interests’. Clearly this defence risks abuse. Williams therefore wishes to find a way of distinguishing between brainwashing or conditioning cases from cases which genuinely track real interests. Williams’s immediate concerns here are, at bottom, political,60 but his commentary is still relevant to the moral luck issue, because he still faces the question of how we might distinguish between changes in perspective that morality should be pre­ pared to live with, and changes in perspective that morality can rightly condemn. In this passage, it is the agent’s general incapacity which stands as the secure invariant feature, the Archimedean fulcrum that can help us to divide the relevant changes into brainwashing cases and cases involving real interests. What is going to play that role in the moral case? Williams does not give us enough to work with here. A second problem with the Improved Transformation Interpretation is that we have to settle for a hazy account of Gauguin’s grounds for undertaking the project in the first place. As the Improved Transformation Interpretation tells the story, at that earlier point Gauguin can count as a moral agent, with ordinary full-­blooded moral concerns. The end for the sake of which he acts is not yet vindicated; it may turn out to be vindicated, or it may not. It is unclear without additional commen­ tary why Gauguin’s concerns for the costs he imposes on his family do not stop him from making the decision, or why, for that matter, and from his earlier per­ spective, the non-­vindicated end stands any chance of being vindicated. (He will, naturally, feel unhappy and divided about this, if he abandons his plans rather than his family.) If the means he takes to his end are not yet justified, how can he take them while retaining the understanding of himself as a moral agent, answer­ able to moral concerns? A third problem is that, even though Gauguin may to some degree be able to shuttle back and forth between the earlier perspective and the later perspective, it is nonetheless the later perspective, not the earlier perspective, with which he comes to be ultimately identified. According to this later perspective, Gauguin 59  Williams (1985), pp. 42–3; original emphases. 60  One detects in this passage the likely influence of Berlin’s worries about abusive invocations of positive liberty and real interests: see, most famously, Berlin (1969).

156 Gauguin ’ s Lucky Escape takes himself to be justified in doing what he did: he is in a position to affirm the end for the sake of which he acted, and also the means he took to those ends. The fact that Gauguin embarked on this biographical journey from a perspective that was originally firmly stationed within morality does not change the fact that this journey brings him to a place where those concerns no longer register with him in the same way. Of course, this is the very point Williams wishes us to accept. But the challenge for him is to persuade us of the aptness of his preferred description of the case. On Williams’s view, Gauguin is originally stationed within morality, but, due to moral luck, finds himself beyond morality. According to the morality system’s preferred version of events, Gauguin is not someone who has escaped morality’s authority, but someone who has simply left morality. Gauguin had moral con­ cerns but no longer has them in the same way. The Improved Transformation Interpretation attempts to make sense of this transition in terms which are favourable to Williams’s version of events by distinguishing between two perspec­ tives that govern Gauguin’s concerns at different times. These perspectives may belong to the same agential biography, but they are still distinct, and the very fact of their distinctness, and the fact that one supplants the other, should make us pause before embracing Williams’s version of the story. Even in this most refined version of his account, I do not think that Williams manages to show us how there can be lucky escapes from morality.

5.7.  Agent-­Regret and the Restricted Account For the most part, my exploration of Williams’s argument has maintained a polite distance from the Nagel-­inspired debates, and from the concerns and commit­ ments of the Restricted Account. But we are about to re-­engage with these con­ cerns, for Williams’s arguments contain a hidden challenge for the Restricted Account. I want, in conclusion, to address this challenge. The challenge returns us to the lorry-­driver’s experience of agent-­regret when he runs over the child. Williams rails against any picture of agency that simply absolves the lorry-­driver from agent-­regret at what, through no fault of his own, he has caused to happen. The Restricted Account does not, or need not, protest against the existence of agent-­regret. But it does affirm that there is a big differ­ ence between Williams’s case, in which the lorry-­driver was not initially at fault, and a case such as Drivers, when the fatality-­causing drunken driver was initially at fault: there is a morally significant difference between the lorry-­d river’s situation and Fred’s situation in Drivers. According to the Restricted Account, it is right to hold Fred blameworthy for the pedestrian’s death, because he is conse­ quentially liable for it. As a result of his consequential liability, he can be held to

Agent-­Regret and the Restricted Account  157 be blameworthy for consequences over which he lacks full control. We might also expect Fred’s blameworthiness to be matched by feelings of guilt and anguish which register his greater moral blameworthiness. The lorry-­driver is in a differ­ ent situation. Whatever agent-­regret amounts to, it does not contain the moral­ ized elements of the feelings and emotions which it would be appropriate for Fred to experience. We may also recall, more generally, the differences between Fred and Thomas, from Drivers, and Noah, from Sober Driver.61 Fred and Thomas are blameworthy for their recklessness, whereas Noah is not reckless, and therefore not blame­ worthy. There is a sizeable gulf, according to the Restricted Account, between Noah, on the one hand, and Fred and Thomas, on the other hand, even if it is sensible to expect some degree of emotional disturbance in all of their lives. One way to parse these cases is simply to settle for a comfortable and consist­ ent division of territory. According to this proposal, Williams’s lorry-­driver and Noah from Sober Driver may experience agent-­regret, but they are not blame­ worthy, whereas Fred and Thomas from Drivers are both blameworthy, and are apt for the experience of a more strongly moralized form of results-­sensitive remorse. Furthermore, there are no real tensions between these different phe­ nomena in this moral-­psychological heat map: agent-­regret is one thing, and results-­sensitive moralized remorse is another thing, and neither of these items impugns or necessarily constrains the operation of the other. There may be a greater range of data in moral psychology to explain in total, but there is no necessary tension between the explananda. Agent-­regret and results-­sensitive remorse constitute non-­overlapping domains. We can call this the No-­Priority Model. It cannot be ruled out, however, that Williams means to make a deeper and more critically aggressive point: perhaps it is the existence of agent-­regret which is fundamental, while the existence of moral remorse is simply an extension of agent-­regret into a more obviously moralized or morally loaded context. That alternative way of looking at things—call it the Agent-­Regret Model—would indeed present a challenge to the Restricted Account, since it would impugn the Restricted Account for drawing distinctions in the way it does between Drivers and Sober Driver.62 The Agent-­Regret Model would place these two cases much closer together than I have done in my account. I am inclined to think that the Agent-­Regret Model is probably what Williams means to endorse. It can be 61  See Chapter 1, Section 1.2. 62  Crisp (2017), p. 14, appears to make friendly overtures to what I have called the Agent-­Regret Model, though it should be noted that this is in the course of arguing for an anti-­luckist debunking of the relevant intuitive data. Enoch (2012) might also, I think, favour the Agent-­Regret Model over the No-­Priority Model, so long as the former can be squared with his specific views about the role of penumbral agency. See Chapter  1, Section  1.5.3, for the earlier discussion of Enoch’s Taking Responsibility Strategy, which specifically engages with it in its guise as an anti-­luckist accommoda­ tion strategy.

158 Gauguin ’ s Lucky Escape detached from his main argument about Gauguin’s escape from morality, but it also challenges morality from the inside, by attempting to re-­order what we think matters to the generation of culpability and liability. On what grounds can friends of the Restricted Account favour the No-­Priority Model over the Agent-­Regret Model? The Agent-­Regret Model is banking on a certain view about how best we accommodate the manifestations of agency: on this view, we make any causal upshots from agency, no matter how compromised or fault-­free that agency may be, the bedrock of moral answerability or experi­ ence, and in so doing we may be able to make sense, further down the line, of the differential blameworthiness between Fred and Thomas in cases such as Drivers, or between either one of these agents and Noah in cases such as Sober Driving. Outcomes may be relevant, but their relevance is displayed, first and foremost, through brutely causal connections to agents, as manifested in the lorry-­driver’s case or in Noah’s case. Even if this is Williams’s ambition, I do not see how the Agent-­Regret Model represents any improvement over the No-­Priority Model. First, my preferred Restricted Account proposes to explain the differential blameworthiness between Fred and Thomas in Drivers by a form of consequential liability which is triggered precisely by Fred’s and Thomas’s culpability. If Fred and Thomas were not culp­ able, this triggering mechanism would not be in place. Second, the Agent-­Regret Model does not provide any easy way of accommodating the all-­important differ­ ences between Fred and Noah, on the one hand, and between Thomas and Noah, on the other hand. We shall take these comparisons in turn. Now both Fred and Noah kill someone, but that similarity between them must not push us into the arms of the Strict Liability Account. It is hard to see how the Agent-­Regret Model can avoid such a fate, unless Williams’s account is supposed merely to suggest to us what is going on in the experiential or psychological ledger of blameless killers such as the lorry-­driver, or Noah. The point empha­ sized by the Restricted Account is that blameworthiness for the pedestrian’s death which Fred causes is the proper price which he pays for his recklessness. It is a reflection of his consequential liability. Again, though Noah may find it difficult to extricate himself from the fact that he has caused a death, we should be able to register a difference between what he has done and what Fred has done, and we should expect corresponding differences between how he feels and how Fred feels. How is the Agent-­Regret Model going to construct a difference between Fred and Noah out of materials which emphasize, in the first instance, their simi­ larity? Moreover, there is a ready-­made explanation of why Noah’s anguish might seem partly indistinguishable from Fred’s, and the Restricted Account can take advantage of this. Perhaps Noah cannot be altogether sure, without further reflection and others’ reassurance, that he did in fact avoid recklessness, and we would expect

Agent-­Regret and the Restricted Account  159 morally conscientious agents not to be too ready to give themselves the benefit of the doubt when their ordinary activity has had such lethal consequences.63 Now let us consider Noah and Thomas. Noah kills someone, and Thomas does not. The Agent-­Regret Model is committed to making that distinction a signifi­ cant one, but it is going to be difficult to recover a position which is committed to Thomas’s blameworthiness and Noah’s non-­blameworthiness from those initial commitments. Again, if the Agent-­Regret Model is supposed to describe the emotional soil from which moral flora can be grown, we appear to have embarked from a point which is too closely stationed to the Strict Liability Account to give us what we are after. The Restricted Account, I believe, gives us a better way to go. This is because it purports to explain why, of the three agents—Fred, Thomas, and Noah—Noah is very distinct, morally speaking, from both Fred and Thomas, though Fred also morally differs from Thomas in a way that budgets for the grisly outcome of his reckless car journey. The Restricted Account is attuned to these differences, whereas the Agent-­Regret Model will struggle to register them and accord them their proper significance. So it is the Restricted Account which is better placed, as I see it, to capture how luck makes, or fails to make, a difference to agents’ blameworthiness.

63 See Sussman (2018) for an insightful discussion. This particular concern is also evident in Williams’s account.

6

Justice, Luck, and Pairwise Comparisons 6.1. Introduction We now turn to political philosophy, and more particularly to distributive justice, the domain which absorbs much of the intellectual energy regarding constitutive luck when the fundamental worries about free will have been noted and the relevant promissory notes have been issued. Instructive similarities between the  debate about blameworthiness in normative ethics and the debate about ­justice will be noted as we proceed. As we shall see, there are similar hazards encountered in these debates. These hazards ultimately rest, in my view, on an over-­extensive investment in pairwise comparisons between agents. Anti-­luckists in normative ethics are led astray by their determination to annul the luck-­based inequalities in praise or blame which their standing interest in pairwise comparisons encourages them to notice. I shall suggest that the same basic fault infects a certain class of responsibility-­sensitive theorists about distributive justice, known as luck egalitarians. The chapter—a rather long chapter—will unfold as follows. In Sections 6.1 to 6.3, I attend to some stage-­setting about distributive justice, and provide some preliminary analysis of luck egalitarianism’s distinctive commitments and features. I also raise a few preliminary worries about the grounding for luck egalitarianism. In Section 6.4 I outline the ‘Boring Problem’, which is the name of a problem for luck egalitarianism I owe to Susan Hurley.1 It seems to me that the Boring Problem—a problem so named both because Hurley thought it obvious, and because she thought it lacked much significance—is an interesting and deep problem that necessitates a fundamental alteration in the standard structure of luck egalitarianism. I call this new version ‘baseline-­relative luck egalitarianism’. The various twists and turns which are relevant to this material will occupy me between Sections 6.5 and 6.7. In Section 6.8, I will take a further critical look at baseline-­ relative luck egalitarianism. In my view, baseline-­ relative luck egalitarianism offers only a Pyrrhic victory for luck egalitarians. What might at first have held the promise of a tactical retreat is better regarded, in the cold light of day, as closer to a defeat. Responsibility-­sensitive theorists of justice should be encouraged to look elsewhere, as I suggest in Sections 6.9 and 6.10. The resulting 1  See Hurley (2003), ch. 6. As will be obvious, my thinking about these issues has been heavily influenced by Hurley. This chapter draws heavily on Lang (2015a), though my ultimate conclusions about luck egalitarianism here are more unambiguously critical than they were in that article. Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0007

164  Justice, Luck, and Pairwise Comparisons challenge will be taken up in Chapter 7, when we turn to Rawlsianism, which has a different way of dealing with luck. I should say at the start of this discussion that not every interesting or live issue concerning luck egalitarianism will be discussed in depth in this chapter. Some substantial accusations that have been levied against luck egalitarianism—for example, that it demeans or abandons the imprudent—will barely be touched upon.2 What I want to pursue above all else is the question of structure.3 My central contention will be that luck egalitarianism is structured in such a way that it cannot coherently achieve what it sets out to do, and the arguments to follow are all in the service of demonstrating that contention. We can learn from these failures, and re-­orientate the investigation accordingly. Moreover, a sensible re-­ orientation of the underlying concerns will be much less invested in the neutralization of luck, as we shall see.

6.2.  Justice, Equality, and Luck Inequalities between agents that reflect luck are unjust: that is the basic commitment of luck egalitarianism. We need more material to provide a fully accurate and comprehensive characterization of the position, however. In this section, I will provide a preliminary analysis of the structure, problems, and prospects for luck egalitarianism. Before I turn to luck egalitarianism itself, however, some further stage-­setting will be helpful.

6.2.1.  The Parameters of Justice: What, How, Who Following Susan Hurley, I distinguish among three parameters of distributive justice: these are the ‘what’ question, the ‘how’ question, and the ‘who’ question.4 The ‘what’ question is concerned with what gets distributed in theories of justice: which goods command the attention of justice, and which goods, if any, are irrelevant to it? Justice-­relevant goods comprise the currency of justice. The usual candidates for the currency of justice are resources, welfare, capabilities, and hybrid currencies such as G. A. Cohen’s ‘access for advantage’.5 But we should 2  See, for example, Anderson (1999), Wolff (1998), Voigt (2007), Segall (2007), Bou-­Habib and Olsaretti (2004), White (2004), and Lang (2009a). 3  Not every possible path of investigation under the structural dimension will be pursued, either: another possible path, which traces out the distributively destabilizing effects of various cyclical and counter-­cyclical variations in luck, is investigated in Lang (2006). 4  See Hurley (2003), pp. 9–10. 5  Following Sen (1980), this is sometimes referred to as the ‘equality of what?’ debate. I will abstain from an examination of the currency question here. It might be remarked, however, that much early theorizing in the luck egalitarian tradition, such as Dworkin (1981a) and (1981b), and Cohen (1989),

Justice, Equality, and Luck  165 be prepared to cast the net even more widely, and at least entertain other candidates for the currency of justice, such as political influence or moral authority.6 The ‘how’ question is concerned with the favoured pattern of justice, or recipe for generating just distributions: what is the shape of the distributions aimed at? Every theory of justice will have a way of answering the ‘how’ question, even ‘non-­patterned’ theories of justice such as Robert Nozick’s entitlement conception of justice.7 The entitlement conception of justice still returns an answer, admittedly defined in procedural but not end-­state terms, to the ‘how’ question. Egalitarian theories answer the ‘how’ question, of course, by recommending an equal amount of something-­or-­other. Luck egalitarianism, which we will be examining in this chapter, belongs to this company. The ‘who’ question is concerned with the constituency of justice: to whom are the goods recommended by the ‘what’ question, and governed by the ‘how’ question, distributed? These questions will be revisited in Chapter 8. Until then, most of our business will be transacted with the ‘how’ question.

6.2.2.  Egalitarian Theories and the Egalitarian Plateau Luck egalitarianism, as we shall see, is an egalitarian theory of justice. I am going to pursue the question of what egalitarianism is for a short while, before examining where luck fits into the picture. The class of egalitarian theories will be broader than is commonly thought unless we immediately add further characterization and structure. Will Kymlicka has suggested that every serious theory of political morality actually submits to an egalitarian characterization of it.8 Utilitarianism, for example, counts everyone’s utility equally: as far as utility goes, each counts for one, and no one for more than one. That particular description of utilitarianism will not be surprising, since it is one which was adumbrated by classical utilitarians such as Jeremy Bentham and John Stuart Mill.9 Moreover, in the light of empirical facts about humans’ diminishing marginal utility, utilitarianism will tend to favour, in typical conditions, relatively equal over relatively unequal unfolded as a series of attempts to reach greater precision in the currency debate. I am not convinced that this partial elision of ‘how’ concerns and ‘what’ concerns was sensible. 6 See Carter (2011), p. 540. The relationships among these categories can be complicated. For example, Sher (2014) nominates the ‘ability to live one’s life effectively’ as the equalisandum, which requires in turn, sufficientarian policies with respect to the other standard equalisanda, such as resources. 7  See Nozick (1974), ch. 7. 8  Kymlicka (2002), ch. 1. Dworkin (1973), p. 531, also adverts to the fact that many theories that can be made to adapt to egalitarian description. 9  See, for example, Mill (1861/1998), ch. V, para. 36, p. 105, where Mill suggests in a footnote that the principle of utility affirms that ‘equal amounts of happiness are equally desirable, whether felt by the same or by different persons’.

166  Justice, Luck, and Pairwise Comparisons distributions. Still, utilitarianism is not commonly regarded as an egalitarian theory. Its preference for equal over unequal distributions is just a contingent corollary of its fundamental concern with something quite different: namely, the maximization of utility. Moreover, and as a result, in at least some circumstances utilitarianism will endorse an unequal distribution, precisely in order to maximize utility. Other theories of political morality can also lend themselves to egalitarian description. Nozick-­ style libertarianism, for example, is usually taken to be hostile to egalitarianism, on the obvious grounds that it is opposed to any redistribution of property holdings acquired through free market exchanges.10 Libertarianism is fantastically relaxed, indeed insouciant, about inequality of condition. Yet libertarianism may also be described as a theory of equals: one which treats each person with equal respect by making her no more and no less a self-­owning agent than any other agent. Yet other theories, however unobvious their egalitarian credentials may seem, can be added on similar grounds to the theories which congregate on what Kymlicka calls the ‘egalitarian plateau’.11 At this rate, the egalitarian plateau is going to be very crowded. Indeed, it will seem over-­crowded, for it would be eccentric to classify libertarianism and utilitarianism as egalitarian theories. It seems likely, then, that there will be a presumptive difference between the standing of a political theory which may qualify for admittance to Kymlicka’s egalitarian plateau, and a political theory whose advocates and foes alike tend to understand it as egalitarian. The additional ingredient egalitarians are looking for is likely to be comparability: to the extent that equality matters, and with respect to the goods whose distribution it claims authority over, egalitarianism takes an interest in how one agent fares relative to another agent, and vice versa.12 Egalitarian theories insist upon an egalitarian distribution in that comparable sense. Comparability-­ catering forms of egalitarianism can be distinguished from theories of justice that exemplify an extended egalitarian structure in Kymlicka’s more relaxed sense. We can refer to the former as strongly egalitarian theories, and the latter as weakly egalitarian theories. This is not to dismiss Kymlicka’s suggestion as a red herring. It may indeed be a deeply important feature of all these theories of political morality that they can be represented in the first instance as theories of equality. An awareness of this theoretical feature may provide us with theoretical apparatus for comparing the merits of strongly egalitarian theories with only weakly egalitarian theories. It may help us to figure out what strongly egalitarian theorists are after, and how they can best secure it. For the moment, however, we need to pay attention to the aspects of egalitarianism that self-­described strong egalitarians take seriously, on 10  Nozick (1974), ch. 7.

11  Kymlicka (2002), p. 4.

12  Temkin (2017), p. 44.

Justice, Equality, and Luck  167 pain of failing to identify their key commitments. Clearly, comparability is the key to their distinctive distributive creed. If comparability is the key, then it is natural to expect a role for pairwise comparisons. As we shall see, this will turn out to be an unwise investment.

6.2.3.  Defining Luck Egalitarianism Luck egalitarianism is a strongly egalitarian theory—perhaps better, a family of distributive theories.13 The term ‘luck egalitarianism’ was bequeathed to the literature by Elizabeth Anderson in a notably hostile discussion of these theories.14 Nonetheless the label has stuck, among friends and foes alike, and I will be using it too. Luck egalitarianism was influenced by Rawls’s theory of ‘justice as fairness’, particularly in respect of the hostility Rawls displays to distributive influences which are morally arbitrary.15 The propriety of that Rawlsian influence will be questioned in Chapter 7. For the moment, we will simply study the behaviour of the would-­be heir to the Rawlsian inheritance, and leave the more detailed questions of intellectual inheritance until later. As a strong form of egalitarianism, we should take luck egalitarianism to be committed to comparability, which will generate in turn an immediate interest in pairwise comparisons. But luck egalitarianism is not rigidly egalitarian; it does not condemn each and every instance of inequality. Its aim is to provide a basis for either favouring or disfavouring inequalities within pairwise comparisons between any two individuals. To that end, it proposes a distributive ‘cut’ between permissible and impermissible inequalities. What is that cut? According to the received or standard understanding of luck egalitarianism, inequalities are unjust if16 they reflect relative involuntary disadvantage among individuals. The flavour of luck egalitarianism can be conveyed by two familiar, definitional remarks. The first of them comes from Larry Temkin: It is bad—unjust and unfair—for some to be worse off than others through no fault (or choice) of their own.17

13  See Arneson (1989), (2004), Cohen (1989), Roemer (1996), Rakowski (1991), Van Parijs (1995), Knight (2009), Segall (2013), (2016), Stemplowska (2009), Lippert-­Rasmussen (2015), and the essays in Knight and Stemplowska (2011) for informative discussions. This is a small representative sample of a much larger literature. 14  Anderson (1999). 15  See Rawls (1971), especially Part One. Another possible influence, particularly on figures such as Cohen and Roemer, might be Marx (1891/1978). Thanks to Nicola Mulkeen for this lead. 16  But not ‘if and only if ’, on many views of luck egalitarianism: the reflection of relative involuntary disadvantage is a sufficient but not a necessary condition of injustice. See Lippert-­Rasmussen (2015), pp. 3–6, for an instructive discussion. 17  Temkin (1993), p. 13, n. 21.

168  Justice, Luck, and Pairwise Comparisons The second of them comes from G. A. Cohen: [T]he purpose [of egalitarianism] is to eliminate involuntary disadvantage, by which I . . . mean disadvantage for which the sufferer cannot be held responsible, since it does not appropriately reflect choices that he has made or is making or would make.18

It is important to get a little clearer on the basis of the distinction between the voluntary and the involuntary, at least as that distinction feeds into luck egalitarianism. What does that contrast amount to? For Cohen, the crucial distinction, or principal distributive ‘cut’, lies between choice and luck.19 As Cohen writes, ‘Egalitarians . . . object to all and only those inequalities that do not appropriately reflect choice’.20 A corollary of that claim is that ‘only differential responsibility can justify inequality’.21 In truth, it does not matter all that much whether we identify choice or responsibility as the item which properly contrasts with luck. Cohen elsewhere remarks that ‘egalitarian redress is indicated to the extent that a disadvantage does not reflect genuine choice’.22 The italicized ­qualification of the reference to choice implies that we may need to make further distinctions among different types of choice. What matters to Cohen, and what will matter to luck egalitarians generally, is that individuals need to have control over, or responsibility for, their actions, if the results of those actions are to escape distributive correction.23 Another way of putting this point is that agents’ actions, or the outcomes of their actions, should be free of luck. Whether we place the emphasis on choice, control, or responsibility24 is less important than the point that these actions are meant to be, in the relevant sense, luck-­free. Correlative with luck egalitarianism’s condemnation of inequalities that are rooted in luck is its view that other types of inequality, inequalities which are not rooted in luck, may be upheld. For luck egalitarians, disapproval of certain types of inequality (those that reflect luck) goes along with approval of other types of inequality (those that do not reflect luck). That is, luck egalitarians appeal to a single principle, governing distributions of the relevant goods, which determines when inequalities are permissible and when they are impermissible. 18  Cohen (1989), p. 916. Although Cohen is often taken as a paradigmatic luck egalitarian, his support for it actually appeared to waver in his later work: see, for example, Cohen (2011). 19  See Cohen (1989), (2006). 20  Cohen (2006), p. 439. 21  Cohen (2006), p. 442. 22  Cohen (1989), p. 934; original emphasis. 23  All the results of those actions? That is implausible, as Olsaretti (2007) instructively argues. Compare the caveat pursued in Chapter 2, Section 2.5.2. 24  We would also need to know what further conditions responsibility, or control, or choice have to satisfy in order to count as luck-­free, and luck egalitarians can plausibly diverge with respect to these matters. Perhaps a choice does not count, in the relevant sense, as luck-­free unless the agent has an alternative, or a reasonable alternative, or can act in a way that is free of duress or other types of influence. I will not pursue these further details here, since the problems I am about to raise will apply regardless of how exactly they are settled. Thanks to Nicola Mulkeen for commentary on these matters.

Grounding Luck Egalitarianism  169 In the luck egalitarian literature, there is a further important distinction to note between brute luck and option luck. This distinction comes to us originally from Dworkin.25 On Dworkin’s characterization of it, brute luck is luck which results from choices that an individual could not avoid making. For him, brute luck is associated with a person’s circumstances, as opposed to the person herself. Talents and ambitions are associated with the person, whereas talents and mental and physical characteristics are associated with the person’s circumstances. Brute luck contrasts with option luck, or luck resulting from choices or activities that an individual freely undertakes. Consider gambling. You have no control over which horse comes in first, but you do have control—barring the possibility that you are a gambling addict—over the decision whether to have placed the bet in the first place. Since you have no control over which horse comes in first, placing a bet involves luck, but this is an example of option luck. Luck egalitarianism standardly considers brute bad luck, but not bad option luck, to warrant compensation. Luck egalitarianism is probably the most prominent post-­Rawlsian theory (or family of theories) of distributive justice in the Anglo-­American philosophical tradition, and its initial appeal is not difficult to grasp. For its defenders, luck egalitarianism preserves potentially valuable or attractive freedoms habitually associated with the political right with the indispensable protections and concern for a robust form of equality of opportunity offered by the political left. It does so, moreover, in a theoretically unified way, appealing to a single distributive ‘cut’ between permissible inequalities which reflect voluntary behaviour, and impermissible inequalities which reflect involuntary behaviour. Luck egalitarianism allows room for the appeal to individual responsibility and the pursuit of personal ambitions, which the political left has often been accused of ignoring.26 But it also protects individuals against involuntary losses and relative deprivation due to factors over which they have no control. This moral package is attractive, and its theoretical unity is also attractive. Luck egalitarianism is clearly an important theory, regardless of whether it proves to be a satisfactory one.

6.3.  Grounding Luck Egalitarianism Clearly, luck egalitarianism is a complex and conceptually hybrid theory. It combines an ‘anti-­luckist’ component, which assesses the justice of distributions

25  For this distinction, see Dworkin (2000), pp. 73ff. It derives originally from Dworkin (1981a). Dworkin’s coinage of this terminology, and its powerful influence on the literature, help to explain why he is often assumed to be a luck egalitarian, despite his protests about this ascription in Dworkin (2003). I will take no final view here of how Dworkin should be classified, though there are definitely notable differences between him and Cohen. 26  See, especially, Scheffler (1992). I enlarge on the distinction between the endowment-­insensitive part and the ambition-­sensitive part of the luck egalitarian package in Lang (2009a), pp. 318–22.

170  Justice, Luck, and Pairwise Comparisons according to the presence or absence of interpersonal luck, with an ‘egalitarian’ component, which is committed, at least presumptively, to equality in some or other form. How do these components go together?

6.3.1.  The Egalitarian Fallacy and the Egalitarian Default There are two central worries to contend with here, and these are interconnected.27 One of these worries is that, even if we manage to eliminate luck, we will not necessarily have arrived at equality. A second major worry is that the luck-­ neutralizing commitments, whatever they amount to, will in fact be tracking values other than equality. The first of these worries is stated forthrightly by Hurley: The fact that people are not responsible for difference does not entail that they are responsible for non-­difference.28

This is the ‘Egalitarian Fallacy’. The Egalitarian Fallacy, less elliptically, is the fallacy of moving from the claim that (1)   It is a matter of luck that X and Y are unequal to the claim that (2)   It would not be a matter of luck if X and Y were equal. Hurley argues that the move from (1) to (2) is just as fallacious as the move from the claim that (3)   It is a matter of luck that X and Y are equal to the claim that (4)   It would not be a matter of luck if X and Y were unequal.29 Luck egalitarians, of course, will have no time for the move from (3) to (4). Hurley’s point is that the move from (3) to (4) is no more or less defensible than the move from (1) to (2). Luck egalitarians might respond to the Egalitarian Fallacy by accusing Hurley of under-­ describing the commitments of their theory. The purpose of luck 27  The discussion here draws here upon earlier discussions in Lang (2006), pp. 43–6, and Lang (2015a), pp. 701–2. 28  Hurley (2001), pp. 56–7; original emphases. 29  See Hurley (2001), p. 57; original emphasis. (I have slightly changed Hurley’s notation.) See also Hurley (2003), pp. 151ff., for another presentation of the argument. The Egalitarian Fallacy was briefly encountered in Chapter 3, in connection with resultant luck: see Section 3.3.5.

Grounding Luck Egalitarianism  171 egalitarianism, they are entitled to remind us, is not to eliminate all luck, but differential luck. The most fundamental purpose of luck egalitarianism is not to expunge luck altogether, but to eliminate distributions which contain disadvantages that are a matter of brute bad luck, and replace them with distributions which do not contain disadvantages that are a matter of brute bad luck. If X and Y are equal, as in (3), luck of a certain sort might be involved—the luck, presumably, that consists in these agents living in the sort of morally enlightened polity that implements luck egalitarianism, rather than some rival principle of distribution—but this is not the sort of luck that sustains disadvantage, as in (1). That is why luck egalitarians can, with a clear conscience, support an equal over an unequal distribution. And that is why they are entitled to remain untroubled by the formal similarity between the move from (1) to (2), and the move from (3) to (4). If it is differential luck on which luck egalitarians focus, they will have a way of avoiding the embarrassment of having to endorse the move from (3) to (4), and of showing how the elimination of inequality is fundamental to their purposes. That reply is anticipated by Hurley, who maintains her line that luck egalitarianism’s luck-­neutralizing commitments do not provide any support for the specifically egalitarian part of luck of egalitarianism. To do so, the luck-­ neutralizing aim would have to demonstrate how luck egalitarianism favours distributions with a roughly egalitarian character; it would have to show why relatively equal distributions are preferred, at least as a default, over relatively unequal distributions.30 Hurley’s view is that the luck-­neutralizing aim cannot do this. It can provide neither a justification for this default type of egalitarian patterning, nor a specification for it. (A specification for the egalitarian default would consist simply in a descriptive statement of the extension of possible distributions meeting the relevant egalitarian constraints.)31 The luck-­neutralizing aim cannot contribute to the justification for egalitarian patterning due to its exposure to what she calls the luck-­neutralizing dilemma.32 On one horn of this dilemma, we appeal to interpersonal luck between agents in order to demonstrate what is objectionable about the inequality that obtains between them. Consider two agents: Jack and Erum. Imagine that Jack ends up with A, and that Erum ends up with A + B. (Assume also that A and B are positive quantities of goods whose distribution matters to justice.) Hurley’s main response to this horn of the dilemma is to point out that bad luck is being defined in terms of inequality. Jack suffers bad luck by comparison with Erum because Jack is

30  Hurley (2003), p. 147. Hurley’s treatment suggests that the egalitarian patterning constraint and the egalitarian default, considered separately, must both be provided by the luck-­neutralizing aim: see especially Hurley (2003), p. 148. 31  I will focus on what Hurley has to say about justification for the remaining part of this sub-­ section, and revisit her concerns about specification in Section 6.4.2. 32  See Hurley (2003), pp. 155–9.

172  Justice, Luck, and Pairwise Comparisons worse off than Erum due to factors which lie beyond his control.33 And that gives the appeal to luck no explanatory purchase on why the inequality is morally objectionable. On the other horn of the dilemma, we appeal instead to counterfactual luck; we say that it is a matter of counterfactual luck that Jack gets only A, whereas he might, in other circumstances, have ended up with A + B instead, which would have put him on a par with Erum. The problem with this second horn of the dilemma is that, even assuming we can obtain determinate verdicts about what particular individuals would end up with if their histories were purged of counterfactual luck, there is no particular reason to think that the resulting interpersonal distribution would be recognizably egalitarian.34 The worry so far is that the neutralization of luck cannot be expected to deliver equality, or explain why we should operate with an egalitarian default. In more recent work, Larry Temkin further questions the combination of luckism and egalitarianism: The egalitarian’s fundamental concern isn’t with luck per se, or even with whether or not someone is worse off than another through no fault or choice of her own, it is with comparative fairness, but people have been confused about this because as it happens in most paradigmatic cases where inequality involves comparative unfairness it also involves luck, or someone being worse off than another through no fault or choice of her own.35

This not just a helpful reminder that luck should always be understood to be differential luck. The emphasis on comparative fairness points to a different sort of non-­egalitarian grounding, as Shlomi Segall notes. It is often comparative desert, either prudential or moral, which is meant to call the distributive shots. Temkin seems happy enough with this understanding, but it is frankly ruinous for card-­carrying luck egalitarians, as Segall explains: Suppose saintly Bob and vicious Sue are equally situated despite being unequally deserving. And then suppose that Sue is hit by a falling tree limb. Desertarians will not condemn the new distribution even though it is the product of differential luck . . . We might as well . . . forget about differential luck and focus on desert.36

There is another problem for luck egalitarianism, which I believe is even more taxing. But before turning to it, I want to discuss two responses which luck egalitarians might be tempted to pursue against the complaints so far. These 33  Hurley (2003), pp. 156–7. 34  Hurley (2003), p. 160. 35  Temkin (2017), p. 45; original emphases. 36  Segall (2016), p. 62.

Grounding Luck Egalitarianism  173 responses attempt to engineer a closer connection between anti-­luckism and egalitarianism than any we have managed to uncover so far.

6.3.2.  From Formal Equality to Substantial Equality The first of these responses comes from Samuel Freeman. Now Freeman is not a luck egalitarian; like me, his commitments lean in a Rawlsian direction.37 Nonetheless, he argues that luck egalitarianism, whatever its shortcomings may be, is not burdened in the particular way Hurley suggests. More particularly, Freeman suggests that luck egalitarians are entitled to fight their egalitarian corner by appealing to formal equality, or the dictum which requires us to treat like cases alike.38 He imputes the following defence to the aspiring luck egalitarian: Why shouldn’t we be treated equally with regard to this natural distribution [of talents] and its effects? For, after all, like cases should be treated alike. Why shouldn’t the benefits and burdens that result from the exercise of our natural endowments be equally shared? This is not simply . . . a ‘default position’ . . . It is, rather, required by a fundamental principle of justice. It is clearly less arbitrary to treat equal cases equally, and distribute the products of luck equally, than any other distribution.39

The formal equality dictum is usually regarded as too weak to deliver robust egalitarian results, since it completely under-­ specifies which dimensions determine a case to be ‘like’ some cases but ‘unlike’ other cases. This under-­ specification gives rise to several hostages to fortune. The account of what makes case A ‘like’ case B but ‘unlike’ case C might, for example, embed racist, sexist, or other malicious discriminatory standards. Perhaps person A and person B, who are both Aryans, should be treated unlike person C, who is Jewish. Anti-­Semitism is of course a paradigmatic form of malicious discrimination, and yet we cannot rely on formal equality for its indictment, because formal equality imposes no qualitative checks on acceptable standards for likeness and unlikeness. Can luck egalitarianism avoid this ignominious fate? It all depends on what determines standards for likeness between cases. To steer things in an egalitarian direction, the Freeman-­inspired luck egalitarian may at first be tempted to say that distribution X is like distribution Y, but unlike distribution Z, if and only if X and Y are free of luck-­ based distributive influence, but Z is not. Now that

37  See Freeman (2007b). We shall be spending more time with Freeman in Chapter 7. 38  Freeman (2007b), pp. 120–1. 39  Freeman (2007b), p. 120.

174  Justice, Luck, and Pairwise Comparisons preliminary proposal will not successfully install egalitarian commitments, for reasons that are plainly exposed by the Egalitarian Fallacy: there is simply no guarantee that, when purged of luck-­based influence, X and Y will thereby be equal. A revised proposal can build in egalitarian tendencies by supplementing these standards for likeness and unlikeness. On this revised proposal, X and Y will count as like cases if and only if they are both free of luck-­based distributive influence and are equal in the absence of such luck-­based influence. That further enrichment of the likeness standard will indeed guarantee that distributions maintain respectable contact with the egalitarian centre. But no magical fusion of anti-­ luckism with egalitarianism has been achieved here; the egalitarian prescription is simply added to the overall distributive package as a further and distinctive element. Neither does this revised proposal offer any vindication of luck egalitarianism from the humble materials offered by formal equality. Formal equality is not in a position to lend its authority to the addition of this egalitarian default. In truth, it has no real authority to bestow.

6.3.3.  The Asymmetrical View Segall wishes to provide a genuine grounding for luck egalitarianism, or a revised but still recognizable version of it. He sets out to demonstrate that luck egalitarianism is underpinned by, or anchored in, a non-­instrumental value that is also distinctly egalitarian.40 Part of what it is for that underlying value, whatever it is, to be distinctly egalitarian is that it must be capable of meeting the egalitarian proviso that this doctrine ‘other things being equal, must never favour a less equal distribution of X over a more equal distribution of it’.41 Segall intimates that ‘the beginning of a defence’ of luck egalitarianism lies in the suggestion that ‘[w]hat is bad about (unchosen) inequalities is precisely the fact that they leave individuals undeservedly disadvantaged compared to others’.42 His case proceeds from the existence of egalitarian complaints: the complaint that a worse-­off person has for being worse off than another for arbitrary, luck-­based reasons.43 This is his revised version of luck egalitarianism:

40  Segall (2016), pp. 53–6. Segall (2015) also covers much of the same territory. Elements of this position are also intact in Segall (2013), ch. 1. 41  Segall (2016), p. 56. This egalitarian proviso is weaker than Hurley’s egalitarian patterning constraint, mentioned in Section 6.3.1. 42  Segall (2016), pp. 49–50. 43  Segall (2016), p. 51. For the moment, ‘arbitrary’ and ‘luck-­based’ are being treated more or less synonymously. More will be done to disentangle them in Chapter 7.

Grounding Luck Egalitarianism  175 Asymmetrical view of egalitarianism: It is bad for one to be worse off than another through no fault or choice of one’s own. It is not bad, with respect to equality, for one to be equal to another through no merit or effort of one’s own.44

The Asymmetrical View, as I shall henceforth refer to it, differs from standard versions of luck egalitarianism because it imposes justificatory burdens only on inequality, not on equality. Equality either requires no justification, or else automatically collects it; inequality, by contrast, is acutely exposed to the risk of non-­justification. The Asymmetrical View is meant to represent an improvement over standard versions of luck egalitarianism because it rests upon a recognizable, distinctly egalitarian value: the value that consists in the correction of arbitrary relative disadvantage. To correct such disadvantage, there must be an egalitarian complaint; if there is no inequality, then there can be no complaint, and so there is nothing that needs to be justified or explained. The notion of arbitrary relative disadvantage is conceptually related to the existence of inequality, and is erased by the replacement of an arbitrarily caused unequal distribution with an equal distribution. In an equal distribution, after all, no egalitarian complaints can arise. In respect of equality, there will be nothing to complain about. Segall is not interested in arbitrariness, in and by itself, or in inequality, in and by itself. As he remarks: It is the twin fact of being arbitrarily disadvantaged that is the source of badness according to this account. It is bad for one to be arbitrarily worse off compared to others, and consequently arbitrary inequalities are always bad.45

Segall also insists that he has not simply embellished standard luck egalitarianism with a gratuitous and undefended preference for equality. This is because, as he sees it, some inequalities are not objectionable: inequalities which result from differential success at gambling, for example, or inequalities which result from gift-­giving.46 His argument here, in effect, takes the form of a modus tollens: if the value of equality was simply presumed, then inequalities would be pro tanto objectionable; but inequalities are not pro tanto objectionable, and therefore the value of equality is not simply presumed.47 The fundamental problem with the Asymmetrical View, as I see it, is that there is still an unanswered question about why equality automatically earns a justification, or else is assumed not to need one. Segall’s defence against this 44  Segall (2016), p. 65. Segall wishes to provide a telic or axiological version of luck egalitarianism, rather than a deontic version of luck egalitarianism. These differences do not matter here. 45  Segall (2016), pp. 68–9. 46  Segall (2016), pp. 58–61. 47  Segall (2016), p. 68.

176  Justice, Luck, and Pairwise Comparisons challenge seems to consist in two points. First, the Asymmetrical View provides a substantive account of the bad-­making features of arbitrary inequalities. Second, not every inequality is bad in any case. But the Asymmetrical View’s account of the bad-­making features of arbitrary inequalities will inevitably be incomplete unless more is said about the value of equality. We know that arbitrarily caused equalities are not problematic. So it is not the fact that inequalities are arbitrary that makes them bad. It is the combination of arbitrariness with inequality that creates the toxic bad-­making mixture. Even though Segall is perfectly candid about the fact that inequality needs to combine with arbitrariness in order to trigger a justificatory deficit, he still needs to do more to explain why arbitrariness is problematic when it is wedded to inequality, but not problematic when it accompanies equality. To be sure, without an inequality, there can be no egalitarian complaint, and it is the existence of an egalitarian complaint which is, for Segall, the sine qua non of the Asymmetrical View. The Asymmetrical View takes an interest in egalitarian complaints, not other types of moral complaint. But the existence of an egalitarian complaint cannot be more fundamental than the existence of an arbitrarily caused inequality. The function of an egalitarian complaint is simply to register a complaint about an arbitrarily caused inequality. But then Segall owes us a fuller story about why arbitrariness becomes toxic when it infuses inequality, but not otherwise. Without that further explanation, the particular importance of egalitarian complaints, as one kind of moral complaint that calls for particular attention, has not been accounted for. I tentatively conclude, then, that there are unanswered questions in Segall’s account, which should reduce our level of confidence in it.48

6.3.4.  The Question of Defaults So far, it has been argued that no successful union of anti-­ luckism and egalitarianism is in the offing. Now some of those who are sympathetic to luck egalitarianism might simply accept that anti-­luckism cannot, in and by itself, deliver the egalitarian default. The point they may then wish to press is that, in any theory of justice, there must be some form of default, or some starting point. We must start from somewhere.49 On this attitude to luck egalitarianism, the egalitarian default has to be established separately, and then the luck-­neutralizing criterion can be used to supervise acceptable departures from equality. The fact

48  See Preda (2017) for other thoughts about Segall’s position. 49  This will be true of even a radically non-­patterned theory of justice, such as Nozick’s entitlement conception, though the point is rather smudged by his artfully sly choose-­your-­own distributive starting point in his parable about Wilt Chamberlain: see Nozick (1974), pp. 160–1.

The ‘ Boring Problem ’ for Luck Egalitarianism  177 that the luck-­neutralizing aim is not involved in establishing that egalitarian default in the first place need not be considered embarrassing. It was always asking too much of the luck-­ neutralizing aim to vindicate the egalitarian component of luck egalitarianism.50 Further questions will inevitably arise from this strategy, but it should not be dismissed out of hand. Why shouldn’t there be this division of labour? One worry about it returns us to Freeman’s attempt to use formal equality to deliver substantial egalitarianism, which we explored in Section  6.3.2. For any theory that professes to be strongly egalitarian, it is not unreasonable to ask for robust connections between that theory’s attitude to inequality and its attitude to equality. The reasons why we adopt an egalitarian starting point should be intelligibly connected to the reasons why we permit some inequalities but do not permit other inequalities.51 What inequality corrupts, equality puts right; what equality joins together, inequality puts asunder. Thoughts such as these, which insist on instructive connections between the propriety of equality and the impropriety of inequality, should not be far from egalitarians’ minds.52 Luck egalitarians still need to do more, then, to earn this starting point. I am not claiming that this task is beyond them. As we shall see, in fact, there is a sense in which they are forced to embrace the egalitarian baseline, when the consequences of a further problem are fully explored. The costs and benefits of this approach will be tabulated towards the end of the chapter.

6.4.  The ‘Boring Problem’ for Luck Egalitarianism 6.4.1.  Do I Dare to Eat a Peach? The Boring Problem Explained David Miller has recently attacked luck egalitarianism for its incoherence: the problem, for him, is that the responsibility-­focused axis of luck egalitarianism cannot be reconciled with its egalitarian axis.53 Much of Miller’s discussion is concerned with the disruptive influence of gift-­giving. But there are certain inequalities which he thinks should not cause alarm for luck egalitarians. He describes one of these as follows:

50  In a direct reply to Hurley, Cohen (2006) offers something like this response. 51  Cohen relies upon just this thought himself in his critique of Rawls, as we shall see in Chapter 7, Section 7.7. 52  In the useful terminology advanced by Sher (2014), p. 4, I here adopt a ‘monistic’ rather than a ‘pluralistic’ stance on what the luck egalitarian justification should be for the combination of its ‘egalitarian conjunct’ and its ‘inegalitarian conjunct’. Sher offers his own critique of luck egalitarian pluralistic justifications in chs. 2–3, and a critique of Dworkin’s monistic justification in ch. 4. 53  See Miller (2014).

178  Justice, Luck, and Pairwise Comparisons [I]magine that you and I each begin with a bowl of unripe peaches, and whereas you wait carefully until the fruit has reached its ripest condition, and then begin to eat, I am either too hasty or too dilatory and therefore get less enjoyment from my fruit. Here the inequality arises simply from the impact of the choices that each of us make on our own resources: what you decide to do with your peaches has no effect on my peaches or on what I choose to do with them.54

Miller thinks that this peach-­consuming inequality is not one of the problematic types of inequality for luck egalitarianism. Though he admits there is an element of brute luck in the inequality between the two peach-­eaters, he notes that the impatient peach-­eater was not obstructed from waiting for the peach to ripen, and so has only himself to blame.55 Now Miller’s verdict on the peach-­eating case is intuitively compelling. No serious theory of justice should be detained by inequalities like this one. It would not be a badge of merit that a theory of justice was forced to indict this inequality; doing so would seem heavy-­handed, even ridiculous. Still, it is less obvious that there is nothing for luck egalitarians to worry about, given the structure of their commitments. This takes us to the ‘Boring Problem’ for luck egalitarianism.56 Hurley describes it in the following passage: I may or may not be responsible for my income level. But this is a very different question from whether I am responsible for the relation between my income level and your income level. . . . If Ernest is responsible for X and Bertie is responsible for X + Y, is either [of them] responsible for the difference between their goods positions?57

To explain further: Ernest’s choices may explain why he gets X, but they cannot explain why Bertie gets X + Y. Ernest has no direct control over what Bertie does, or earns; Ernest bears responsibility for his choices, not Bertie’s choices. If Bertie is making any choices at all, then they cannot be Ernest’s choices. Similarly, Bertie’s choices may explain why he gets X + Y, but they cannot explain why Ernest gets X. That is because Bertie has no direct control over what Ernest does, or earns. Bertie bears responsibility for his choices, not Ernest’s choices. If Ernest is making any choices at all, then they clearly cannot be Bertie’s choices. This point will hold 54  Miller (2014), p. 8. 55 In an otherwise critically hostile discussion of Miller’s argument, Elford (2017) accepts this point. This is because Elford has another way of understanding luck egalitarianism’s central distributive commitment. See, further, n. 71. 56  Miller (2014), p. 8, n. 26, notes the possible application of the Boring Problem to the peach-­ eating case. His view is that the Boring Problem fails to apply here, because this case involves only option luck and not brute luck. The Boring Problem cuts through the distinction between brute luck and option luck, as I will show. 57  Hurley (2003), pp. 159–60; emphases added.

The ‘ Boring Problem ’ for Luck Egalitarianism  179 even for the ostensibly innocuous inequality between the peach-­eaters. Let us suppose that Bertie wisely waits for the peach to ripen, while Ernest foolishly jumps the gun and eats it before it has ripened. The difference between their situations was chosen by neither Bertie nor Ernest. The point easily generalizes beyond this simple two-­person case: no inequality between individuals, including supposedly permissible ambition-­ sensitive inequalities, can be fully traced to the choices made by any individual in any comparison among a set of individuals. But this must surely be problematic for luck egalitarianism, since the inequalities among individuals which luck egalitarianism is supposed to permit are those that reflect choices, and the inequalities it is supposed to condemn are those that do not reflect choices. The relevant riposte should be: whose choices? By assumption, the income gap between Bertie and Ernest is Y, which is the difference between Ernest’s X and Bertie’s X + Y. This gap undoubtedly emerges out of the joint combination of their choices, since Bertie’s choice explains why he gets X + Y, and Ernest’s choice explains why he gets X. But it is still the case that the difference between what each of them gets, and thus the inequality which arises between them, is chosen by neither of them. Ernest’s choice explains why he ends up with X, but not why Bertie gets X + Y. Bertie’s choice explains why he ends up with X + Y, but not why Ernest ends up with X. Relative income gaps between Ernest and Bertie are not, and cannot be, fully explained by the choices made by either Ernest or Bertie: their respective choices explain the two income sums from which that gap is calculated, but not the gap itself. This seems problematic. It certainly warrants further investigation. The problem also applies to Miller’s peach-­eating case, as innocuous and trivial as it may seem.

6.4.2.  Why is the Boring Problem a Boring Problem? The Boring Problem earns its name from Hurley’s contention that the problem just described is, variously, ‘simple’, ‘uninteresting’, and ‘obvious’.58 Now obvious problems are not, for that very reason, unimportant or uninteresting problems, but the Boring Problem is, in addition, only a relatively minor cog in the wheel of Hurley’s larger enterprise. We have already taken a look at some of this larger enterprise, in which the luck-­neutralizing dilemma looms large. The luck-­neutralizing dilemma suggests that the aim of neutralizing luck cannot contribute to the justificatory basis of egalitarianism. Hurley then suggests that the luck-­neutralizing aim cannot even contribute to the specification basis of egalitarianism. That is, the luck-­neutralizing aim cannot even reveal a presumptive preference for equal over unequal patterns 58  Hurley (2003), p. 161.

180  Justice, Luck, and Pairwise Comparisons of distribution. It is in this relatively minor role that we are introduced to the Boring Problem for luck egalitarianism.59 The Boring Problem helps to confirm, to Hurley, that the luck-­neutralizing aim cannot contribute even to the specification basis of egalitarianism, due to the fact that ‘[j]udgments of responsibility seem prima facie not to have the right form to specify a pattern of distribution across persons’.60 (A second problem for the specification basis of egalitarianism emerges when we turn to counterfactual luck: once again, these worries oscillate between indeterminacy and non-­conformity to any recognizably egalitarian pattern.) For Hurley, the Boring Problem enjoys, at best, only a limited and supporting role in the overall drama she has constructed for luck egalitarianism. The luck-­ neutralizing dilemma looms largest in this list of complaints about the theory. Besides this, however, Hurley also briefly sketches a solution to the Boring Problem: A luck neutralizer may reply that so long as each person is responsible for her actual position, he is not concerned with whether relations between them are nevertheless partly a matter of luck. Fair enough.61

According to this solution to the Boring Problem, luck egalitarians may legitimately lose interest in the gap of Y between Ernest and Bertie, as long as Ernest is responsible for getting X, and Bertie is responsible for getting X + Y. Hurley’s solution to the Boring Problem will be appraised in Section 6.5.1.

6.5.  Two Fixes for the Boring Problem? Before arriving at my own proposal for tackling the Boring Problem, I want to consider two quick fixes for it. These fixes point in incompatible directions: the first fix advises luck egalitarians to ignore the income gap between Ernest and Bertie, whereas the second fix instructs luck egalitarians to eliminate it. In my view, neither fix is satisfactory.

6.5.1.  The Dismissive Response We have already encountered the first fix: it is Hurley’s own.62 According to this line of thought, luck egalitarians are entitled to suggest that the gap between what Bertie earns and what Ernest earns does not matter in itself. What matters is only 59 In his interesting exploration of luck egalitarianism, Seligman (2007) encounters a problem rather similar to the Boring Problem. See also Huseby (2016) for similarly flavoured concerns. 60  Hurley (2003), p. 160; emphases added. 61  Hurley (2003), p.162. 62  Miller (2014), p. 8, n. 26, also indicates support for this response.

Two Fixes for the Boring Problem?  181 the fact that Bertie is responsible for what he earns, together with the fact that Ernest is responsible for what he earns. Neither of them has to be responsible for the gap between their respective income levels, as long as they are responsible for the individual earnings which lie on either side of that gap. To spell out this line of thought in more detail: Ernest gets X, and Bertie gets X + Y. But it is not being assumed that Ernest’s failure to end up with X + Y is to be explained by luck rather than choice. The explanation of why Ernest gets X, rather than X + Y, proceeds, by assumption, through an appeal to the choices which Ernest has made. So even if Ernest did not choose that Bertie ends up with X + Y, we can still appeal to Ernest’s choices to explain why he did not end up with X + Y. Why is that not enough to shut down the Boring Problem? We can refer to this fix as the Dismissive Response to the Boring Problem. There is some truth to the Dismissive Response. The interim solution that I will offer to luck egalitarians in Section 6.6 has some affinities to it, as we shall see. But it will not quite do as it stands. There are two problems with it. First—and this will be a familiar complaint by now—it is easy to recognize the presence of choice-­sensitivity in the Dismissive Response, but it is more difficult to discern the egalitarianism in it. The solution, as described, consists squarely in appeal to the principle that what individuals end up with should be due to their choices, not to luck. But this proposal does not tell us why inequalities are of special interest to luck egalitarians, and we have already established that strongly egalitarian theories should have something to say about this issue. Imagine that Ernest gets X and Bertie also gets X, and that Ernest’s receipt of X is due to choice while Bertie’s receipt of X is due to brute luck. Call this Outcome A. Outcome A represents both an equal outcome, and an outcome which offends against the choice-­sensitive principle. Now compare Outcome A with Outcome B: in Outcome B, Ernest gets X, while Bertie gets X + Y, but Ernest’s receipt of X is due to choice, while Bertie’s receipt of X + Y is due to luck. Outcome B is an unequal outcome, which makes it dissimilar to Outcome A. But, like Outcome A, Outcome B also offends against the choice-­sensitive principle. Are Outcome A and Outcome B equal offenders against luck egalitarianism, as the Dismissive Response construes it? If so, we lack provision for the egalitarianism which must be part of luck egalitarianism. If not, we lack an explanation of how the egalitarianism is provided for. We need more than this. Second, the basic idea behind luck egalitarianism is that individuals who are on the losing side of inequalities with other individuals are not forced to explain these inequalities by pointing to factors lying beyond their control. But Ernest is forced to explain the gap between his income and Bertie’s income by pointing to a factor over which he all too obviously lacked control: namely, the fact that he had no control over, or responsibility for, the fact that Bertie’s choices turned out in the way they did, or indeed the fact that Bertie made any choices in the first place. The puzzle therefore remains in place. What Hurley may be getting right is that,

182  Justice, Luck, and Pairwise Comparisons on a sensible approach to justice, neither Bertie nor Ernest has a serious complaint about the gap that exists between them, if neither of them was obstructed in his choice-­sensitive income-­making efforts.63 Still, luck egalitarians cannot easily make sense of their acquiescence to these distributive arrangements, since it remains the case that the inequality between Ernest and Bertie is explained by factors for which neither of them is responsible. As it stands, luck egalitarianism cannot explain why the income gap of Y between Ernest and Bertie does not sustain a complaint from Ernest. The income gap between them is therefore also accompanied by a justificatory deficit. More, then, needs to be said before this style of response can be upheld.

6.5.2.  The Concessive Response The second fix takes the opposite tack. According to this fix, luck egalitarians can say that Ernest in fact suffers brute bad luck by not making choices which allow him to collect X + Y, rather than X. Given the existence of this brute bad luck, there will be a compelling luck egalitarian case for redistributing the gap of Y between them, so that, in this particular instance, both of them plausibly end up with X + 0.5Y.64 Given its readiness to embrace the positive distributive implications of the Boring Problem, we can refer to this fix as the Concessive Response. The Concessive Response is also unconvincing, and the interim solution I go on to outline will be markedly less sympathetic to it than it is to the Dismissive Response. Our first question must be this: what is the exact source of Ernest’s complaint about the fact that Bertie gets X + Y, whereas he only gets X? Luck egalitarians can support different answers. First, they might try to exploit the luck-­affected contingency of the fact that Ernest only ends up with X, rather than X + Y.  (Luck egalitarians nowhere deny the presence of such contingency in successful choice-­sensitive behaviour.65) Ernest might have ended up with X + Y, like Bertie, had his choice turned out differently, or had he made a different choice which it was within his power to make. This line of argument seems immediately exposed to Hurley’s indeterminacy problem.66 It is surely true of any level of income made by Ernest that he might 63  The same holds, and even more obviously, for Miller’s peach-­eating case. 64  Perhaps not a compelling conclusive case, since luck egalitarians’ commitment to justice will be tempered by the need to uphold other values. See, further, Section 6.8.3. 65  Dworkin’s distinction between option luck and brute luck seems to confirm this view. Those who have urged luck egalitarians to be suspicious of the distinction between option luck and brute luck, such as Lippert-­Rasmussen (2001) and Vallentyne (2002), seem more likely to favour the overall tenor of the Concessive Response. Though he does not discuss the Boring Problem in its own right, the drift of the discussion in Otsuka (2006) suggests that Otsuka might also affirm the Concessive Response. 66  Hurley (2003), p. 160.

Two Fixes for the Boring Problem?  183 have earned another level of income, through making different choices, or because the choices he made turned out differently. There is nothing special about the fact that Ernest’s achieved level of income is X, rather than X + Y. So why should the fact that Ernest gets X rather than X + Y be privileged, given the additional fact that Ernest might have ended up with any other level of income? It seems to me that the indeterminacy problem does not stop the Concessive Response in its tracks. The fact that Ernest collects X rather than X + Y may indeed count as a privileged fact, in this context, because luck egalitarians are justified in taking a greater interest in correcting indefensible inequalities that are actual rather than merely possible. Only indefensible inequalities that are actual provide luck egalitarians with something to do. Merely possible indefensible inequalities stand in no need of correction, because they are not actually instantiated. This line of argument is more promising, but it ultimately fails to hold the line. We are now appealing to the fact that Bertie’s income of X + Y is contingently higher than Ernest’s income of X, as a result of which Bertie’s income will be reduced, in the corrected distribution, to X + 0.5Y. Due to Ernest’s failure to earn more than X, Bertie is not entitled to hang on to his income of X + Y.67 Correlatively, if Ernest had been more successful, then Bertie would not have lost 0.5Y of income. That fact surely compromises Bertie’s pursuit of his ambitions and choices in ways that were not foretold by the original luck egalitarian vision. The Concessive Response concedes too much to the Boring Problem. In effect, it simply amputates the choice-­sensitive or ambition-­sensitive dimension of luck egalitarianism, which is an indispensable part of luck egalitarianism’s overall initial appeal. It not only reveals a commitment to the elimination of fate-­sharing luck,68 but demonstrates that the elimination of fate-­sharing luck has substantive and potentially unwelcome consequences. The Concessive Response severely restricts individuals’ room for manoeuvre when they engage in choice-­sensitive activity. Luck egalitarians need to explore other options if they are to have any hope of honouring the original dual vision of what their theory promised: distributions that displayed a sensitivity to ambition together with an insensitivity to endowment.

6.5.3.  The Boring Problem and Blameworthiness Before seeing what options luck egalitarians have for evading the sting of the Boring Problem, it would be salutary to take a brief backwards look at its 67  The ensuing restrictions on the pursuit of an individual’s ambitions recall Cohen’s criticisms of a ‘joint ownership’ account of world-­ownership: see Cohen (1986). 68  See Chapter 3, Section 3.3.2.

184  Justice, Luck, and Pairwise Comparisons implications for the debate about blameworthiness which we examined in Part I of this book. What the Boring Problem teaches us in the present context is that, necessarily, each of the two agents in a pairwise comparison has no relevant control over certain aspects of her distributive relationship with the other agent. Regardless of whether they are equal, or unequal—that is the specific lesson to be recovered from the Egalitarian Fallacy—each of them, at best, can have control over only one side of that distributive relationship. We can apply that lesson to the pairs of agents in Assassins and Drivers. I will focus here on Assassins. Perhaps the fact that there is already a prominent luck-­reflecting feature of Assassins—the fact that, for both Imogen and Phoebe, there is a lack of control over whether their attempts are successful—encourages us to overlook a source of luck which is actually independent of that feature. Regardless of the powers of control which Imogen and Phoebe bring to bear when they act, the difference or non-­difference between the outcome produced by Imogen and the outcome produced by Phoebe lies beyond the control of either of them. This source of luck is interpersonal luck, and it cannot be reduced by any enlargement in their individual agential powers. Moreover, this form of luck persists whether the outcomes between Imogen and Phoebe are unequal or equal. Moral luck was never the sort of problem that can be assimilated to the problem of free will and responsibility.69 But it is suspicious that the form of luck that ultimately lies at the source of resultant luck is one that could never be overcome in ways that had anything at all to do with individual agency, and which persists regardless of the pattern of outcomes produced by Imogen and Phoebe. The Boring Problem helps us to see that the interpersonal focus in resultant luck cases is misleading. We would have been better off all along by focusing simply on what Imogen does, and on what Phoebe does. Everything else is a distraction. The comparison between them does not matter.

6.6.  Baseline-­Relative Luck Egalitarianism Wherever the truth lies, it would be surprising, given the concerns that have arisen so far, if any solution to the Boring Problem could aspire to be a quick fix. Even if the Boring Problem is a reasonably obvious problem, the solution to it is actually not all that obvious. This consideration should motivate a more detailed and patient exploration of the problem. How, then, should luck egalitarians defeat the Boring Problem? As I see it, the only stable way of identifying distributions which satisfy the luck egalitarian ‘cut’ between responsibility and luck is to tie permissible distributive inequalities to 69  See the Introduction, Section 4.

Baseline-­R elative Luck Egalitarianism  185 what we can call an egalitarian baseline. The resulting form of luck egalitarianism will be a form of the theory I call baseline-­relative luck egalitarianism.

6.6.1.  The Egalitarian Baseline Call the egalitarian baseline, whatever it is,70 ‘N’, and assume, as before, that Ernest earns X and Bertie earns X + Y.  According to baseline-­relative luck egalitarianism, it does not matter that Ernest has less than Bertie on grounds which neither of them has any direct control over. What matters instead are the facts about each of these individuals’ relationship to N.  If X exceeds N, then Ernest’s receipt of the net difference (X – N) should be due to his choices, rather than brute luck; and, if X + Y exceeds N, then Bertie’s receipt of the larger net difference (X + (Y – N)) should be due to his choices, rather than brute luck. Similarly, if X is less than N, then Ernest’s receipt of X should be due to his choices, rather than brute luck. And, if X + Y is less than N, then Bertie’s receipt of X + Y should be due to his choices, rather than brute luck. Regardless of which of these cases obtains, no further distributive supervision of the income difference between Ernest and Bertie will be available. If Ernest’s receipt of X is due to his choices, and Bertie’s receipt of X + Y is due to his choices, there will still be a gap of Y between their income levels, but that gap will not matter in itself. Gaps between their respective incomes will be assessed as fairness-­preserving or fairness-­destroying only in a mediated way: it is the gap between their incomes and N which is of primary interest. Baseline-­relative luck egalitarianism represents a notable improvement over the two responses to the Boring Problem we have already considered, and it may be useful to compare and contrast it with each of them.71

6.6.2.  Reconsidering the Dismissive Response and Concessive Response Unlike the Dismissive Response to the Boring Problem, baseline-­relative luck egalitarianism explicitly restructures luck egalitarianism, so that it is no longer 70  I return to this point later, particularly in Sections 6.8 and 6.9. 71  Baseline-­relative luck egalitarianism has some affinities with the ‘egalitarian mantra’ outlined by Elford (2013), according to which luck egalitarian concerns can be met by the existence of an opportunity not to fall below a certain level, as opposed to the root and branch elimination of brute luck. The egalitarian mantra, though not the elimination of brute luck, permits gambling and certain forms of gift-­giving, which makes it less restrictive than the austere luck-­eliminating variant of luck egalitarianism. Still, defenders of the egalitarian mantra will also object to unchosen gaps between the beneficiaries and the non-­beneficiaries of gift-­giving. So it seems to me that the egalitarian mantra offers a less attractive, because less flexible, accommodation of luck egalitarianism’s ambition-­sensitivity than baseline-­relative luck egalitarianism. For further remarks on gift-­giving, see Section 6.8.2.

186  Justice, Luck, and Pairwise Comparisons the unchosen nature of the income gap between Ernest and Bertie which generates a complaint about the inequality. It also endows the egalitarianism in luck egalitarianism with a more perspicuous presence. Perhaps this position is what some proponents of the Dismissive Response, such as Hurley, had in mind all along. But the Dismissive Response, as I characterized it, does not sufficiently labour the significance of the egalitarian baseline, or the necessity of renouncing fundamental interest in the existence of the relative gap between individuals in pairwise comparisons. Unlike the Concessive Response to the Boring Problem, baseline-­relative luck egalitarianism does not simply amputate the choice-­sensitive character of luck egalitarianism. Baseline-­relative luck egalitarianism permits the retention of a single distributive cut between permissible inequalities and impermissible inequalities, but adds an important qualification to the notion of relative disadvantage which underlies the formulation of standard luck egalitarianism. Baseline-­ relative luck egalitarianism is also in a position to make some headway against the other constituents of Hurley’s case against luck egalitarianism. Because, in this version of it, the egalitarian baseline is theoretically central to luck egalitarianism, Hurley’s complaints about the lack of justification for the egalitarian character of the theory begin to seem beside the point. The egalitarian baseline is not a gratuitous addition to the theory, but lies at the very centre of it, and thus explains why this form of egalitarianism cannot fail to provide for the egalitarian default. To put it another way, luck egalitarianism’s real business is to say when and why inequalities are permissible, and the responsibility-­sensitive part of the theory is surely indispensable to that project.72 We established in Section 6.3.4 that the provision of a default distribution is an ineliminable element in a theory of justice. Baseline-­relative luck egalitarianism wears that commitment on its sleeve. But what is wrong with that? For we must surely pay attention to what Ernest and Bertie are entitled to before the results of their choices, or their failures to choose, can even be considered. For luck egalitarians, and indeed any patterned theorists of a non-­Nozickian sort, an independent part of the concern with justice will be the presumptive share of distributive resources to which each individual is entitled before we pay attention to the dynamic effects of economic interaction over time. Individuals intelligibly, and rightly, care about this question independently of the inequalities which stand at the end of the various processes of economic activity and transfer, and this concern explains why many theories of justice find it desirable to provide for the egalitarian default. Apart from its ability to solve the Boring Problem, does baseline-­relative luck egalitarianism offer any other types of advantage? I believe so. For example, it permits luck egalitarians to sidestep an important challenge that has been 72  Cohen (2006) makes a similar reply on behalf of standard luck egalitarianism. It seems to me that Cohen’s reply could have been exposed as less than entirely satisfactory had Hurley decided to award a fuller role to the Boring Problem.

Tackling the Paradox of the Baseline  187 forcefully pressed by Saul Smilansky. I explore that issue in the next section. But the fallout for luck egalitarianism is not altogether benign, as I will show in later sections.

6.7.  Tackling the Paradox of the Baseline How do we fix the egalitarian baseline? This is an appropriate juncture at which to examine Smilansky’s ‘Paradox of the Baseline’, which poses a serious challenge to the standard formulation of luck egalitarianism.73 Before outlining Smilansky’s argument, we need to introduce some terminology. There are a number of essential working ingredients in his argument. First, there is his distinction between ‘Effectives’ and ‘Non-­Effectives’. Effectives are defined as those individuals who have economically remunerative options, while Non-­ Effectives are defined as those individuals who lack any such options (the severely handicapped, for example). Second, there is the ‘Highest Potential Income’, which is the highest income earned by any Effective. (Effectives’ choices will not be equally remunerative, since nothing stops these choices from turning out differently. This is a point I will revisit.) Third, there is the ‘egalitarian baseline’, which is the income which Non-­Effectives must be compensated for failing to make, given the fact that their failure to earn this income will not reflect any choices they have made.

6.7.1.  The Paradox of the Baseline Armed with this terminology, Smilansky’s argument can now be reconstructed as follows: 1. The Highest Potential Income, earned by the most highly remunerated Effective, whom we can call Bill Gates, is H. Claim 1 is true by stipulation. 2. By the tenets of luck egalitarianism, the egalitarian baseline (i.e. N) must be set as high as the Highest Potential Income, i.e. N = H. Why? Because 3. If the egalitarian baseline were set any lower than the Highest Potential Income—call this lower amount H*—then Non-­Effectives would suffer 73  See Smilansky (2003), and (2007), ch. 7. Smilansky usually refers to choice egalitarianism, but I will treat luck egalitarianism and choice egalitarianism as interchangeable. The argument in Smilansky (2007), on which I will focus here, was also discussed in Lang (2009b), pp. 287–90, but the force of the Boring Problem in that earlier discussion had not fully dawned on me.

188  Justice, Luck, and Pairwise Comparisons relative involuntary disadvantage (due purely to their bad luck) in comparison to all incomes falling between H* and H. It is also the case that 4. The Effectives who earn less than Bill Gates do not need to be compensated for failing to earn H. Why is that? It is because 5. Effectives, as Effectives, make choices which are not guaranteed to be equally remunerative (since these outcomes reflect choices). Now plausibly, 6. Non-­Effectives can only be protected against the relative involuntary disadvantage of getting less than H if the Effectives’ income is redistributed to ensure that Non-­Effectives are given H. So 7. The Effectives must subsidize equality between Bill Gates and the Non-­ Effectives while foregoing any realistic expectation that they will come to have the same degree of reward. But plausibly, 8. The fact that, under luck egalitarianism, Effectives are morally required to engage in this subsidizing type of activity is ‘absurd and morally repugnant’.74 And so, plausibly, 9. Luck egalitarianism yields a false account of distributive justice.

6.7.2.  Degrees of Effectiveness The Paradox of the Baseline seems open, at first, to a challenge concerning degrees of effectiveness among the Effectives. Nearly every Effective is, in the relevant sense, non-­effective in comparison to, specifically, Bill Gates. After all, the point of selecting Gates as the baseline-­setter, rather than any other Effective, is because Gates’s choices generate a higher income than is generated by the choices of any other Effective. Smilansky might dismiss this objection as unconvincing, since, after all, every Effective can make choices, and since there was no antecedent guarantee that Effectives’ choices would prove to be equally remunerative.75 The fact that Bill Gates is the most highly performing Effective is a contingent one. But as we have 74  Smilansky (2007), p. 74. 75  For a more detailed challenge of this kind, and Smilansky’s reply to it, see Manor (2005) and Smilansky (2005).

Tackling the Paradox of the Baseline  189 seen, differences in the degree of effectiveness of the different Effectives lead directly to the Boring Problem. The ‘degrees of effectiveness’ challenge therefore constitutes a deep challenge to standard luck egalitarianism. And, to deal with the Boring Problem, I have suggested that luck egalitarians must embrace baseline-­ relative luck egalitarianism. As we shall see, the substitution of baseline-­relative luck egalitarianism for standard luck egalitarianism provides crucial resources for subduing the threat posed by the Paradox of the Baseline. To see why, we can consider two pairwise comparisons: between Tom and Daisy, where Tom is a Non-­Effective and Daisy is an Effective; and between Daisy and Jay, where both Daisy and Jay are Effectives. In the first pairwise comparison, between Tom and Daisy, imagine that a ‘middling’ egalitarian baseline has been selected, in which the baseline is 20. As a Non-­Effective, Tom is automatically awarded 20. Imagine that Daisy earns 25. According to Smilansky, luck egalitarians will object to the discrepancy between Tom’s income and Daisy’s income: the income gap between them of 5 (i.e. 25–20) does not reflect any choices that Tom makes, since in the relevant sense Tom, as a Non-­Effective, is not making any choices at all. Moreover, that conclusion is supposed to discredit all low-­to-­middling egalitarian baselines. Next, consider the pairwise comparison between Daisy and Jay, and imagine that Jay earns 30. The outcome of this particular pairwise comparison is supposed to be free of difficulty, since both Daisy and Jay are Effectives. But note—and this will be an entirely familiar point by now—that Daisy did not choose that Jay earn more than her. Jay’s ending up with 30, after all, was by assumption due to Jay’s choices, and Daisy did not choose that Jay make any choices at all, let alone choose that Jay’s choices would prove more remunerative than Daisy’s. In short, Jay’s attainment of 30 is explained squarely by Jay’s choices, not Daisy’s choices. Similarly, Daisy’s attainment of 25 is explained squarely by Daisy’s choices, not Jay’s choices. So neither Daisy nor Jay has any degree of direct control over the size of the income gap between them. Neither the existence nor the size of that gap is determined by the choices either of them makes. What follows from all this, if luck egalitarians are to see to it that the income gap between Daisy and Jay remains unchallenged, is that they must, in general, reject the idea that an income gap between any two agents will be permissible if it is the object of choice of either or both agents. What matters instead is that any earnings made by Daisy and Jay above or below the egalitarian baseline, N, should reflect their respective choices. But if that is true, then the first pairwise comparison, between Tom and Daisy, now stands in need of urgent review. True, Daisy earns more than Tom, and that inequality between them cannot be traced to any choices Tom has made. But we have just established that this particular truth does not offend against a condition which is generally operative in any pairwise comparison between any two subjects of justice in the luck egalitarian scheme, even when those two subjects are both Effectives.

190  Justice, Luck, and Pairwise Comparisons Of course, the permissible inequality between Daisy and Jay is constructed in some sense out of choice-­sensitive activity: Daisy chooses to act in ways which yield 25, and Jay chooses to act in ways which yield 30. That condition is not upheld in the comparison between Tom and Daisy, since Tom is incapable, in the relevant sense, of making any choices at all. Plainly, Non-­Effectives cannot be deleted from the scope of egalitarian concern; they must be awarded an income.76 But by now it is not so clear why the income awarded to Non-­Effectives must match that of any particular Effective, let alone the most effective or highly performing Effective.

6.7.3.  Non-­Effectives and Income Levels So which income is it appropriate to award Non-­Effectives, such as Tom? On a plausible view, the income it is appropriate to guarantee to Non-­Effectives is simply N, or the egalitarian baseline. This would entail that any permissible higher levels of income earned by Effectives such as Daisy and Jay reflect these Effectives’ choices. And it would entail that the income gap between Tom and any Effective is explained only by the choices made by that Effective, which would restore the significance of both the comparison between Tom and any Effective and also the comparison between any two Effectives. (The income gap between Daisy and Jay is constructed out of the choices made separately by Daisy and Jay.) If the income awarded to Tom was any less than N—call this lower amount N*— then the gap between Tom and any Effective earning more than N would be explained, not just by this Effective’s choices, but also by Tom’s brute bad luck. Call this randomly selected Effective, with a higher income than N, Jordan. Imagine that Jordan collects N+, which exceeds N (and also N*, of course). Tom’s brute bad luck would explain the part of the income gap represented by the difference between N* and N, while Jordan’s choices would explain the remaining part of the income gap, between N and N+. This would make the comparison between Tom and Jordan significantly unlike the comparison between Jordan and any other Effective. For these reasons, it is natural to conclude that baseline-­ relative luck egalitarians will only be satisfied if Non-­Effectives such as Tom are awarded the egalitarian baseline, N. These arrangements leave Effectives free to earn more or less than N, depending on how their choices turn out. Thus baseline-­relative luck egalitarianism is not exposed to the claim that Tom must be awarded the Highest Potential Income, and that relieves luck egalitarianism, in its baseline-­relative version, from the

76  This point does, however, raise further issues about who gets to count as a subject of justice. See Chapter 8 for further discussion.

Three Further Problems  191 challenge posed to standard luck egalitarianism by Smilansky’s Paradox of the Baseline.

6.8.  Three Further Problems Baseline-­relative luck egalitarianism survives some challenges. But there are other challenges waiting just around the corner. In this section, we will take a look at three of them.

6.8.1.  The Underdetermination Problem Baseline-­relative luck egalitarianism need not succumb to the Paradox of the Baseline, but its engagement with it leaves us with a problem, or at least a puzzling lacuna. Baseline-­relative luck egalitarianism seems exposed to what I will call the Underdetermination Problem: the selection of the egalitarian baseline is underdetermined by the content of luck egalitarianism. It is simply unclear, at this stage, what value of N is to be selected. It thus seems compatible with luck egalitarianism both that N is low and relatively undemanding, or that N is high and relatively demanding. Luck egalitarian schemes with low values for N will display a certain similarity to sufficiency-­based distributive schemes of the sort recommended by Harry Frankfurt,77 while luck egalitarian schemes with high values for N may start to generate Smilansky-­style concerns. However, luck egalitarian schemes with higher values of N may fare better with the worries, concerning domination, hierarchy, and status, which egalitarians typically, and rightly, emphasize. It is striking, but perhaps not immediately ruinous, for the theory of luck egalitarianism to be underdetermined in this way. There are two points which bear particular emphasis. On the one hand, it might seem that there is little for subjects of justice to rally round, or to take as normatively fundamental, if the specific value of the all-­ important egalitarian baseline cannot even be identified by the apparatus of baseline-­relative luck egalitarianism. But it is vital to emphasize that the problem faced by baseline-­relative luck egalitarianism is not the problem of indeterminacy, but rather that of underdetermination. Underdetermination is not the same as indeterminacy. There is no case for thinking that N cannot be assigned a particular

77  See Frankfurt (1987) for an early statement of sufficientarianism, and Shields (2016) for a more recent and detailed exploration of it. Baseline-­relative luck egalitarians will be keener than sufficientarians are, however, to keep tabs on the incidence of choice-­sensitive behaviour in income levels that diverge from the egalitarian baseline. See also Section 6.9.3.

192  Justice, Luck, and Pairwise Comparisons value. When we do give N a specific value, individuals can then care about whether the resulting distributive profiles reflect the appropriate choice-­sensitive relationship with N. On the other hand, it also needs to be emphasized that the Underdetermination Problem might not be wholly a problem for luck egalitarians, depending on what the really strong challenges to luck egalitarianism are taken to be. Some brief indication of what these challenges are, and how they might possibly be assuaged by baseline-­relative luck egalitarianism, is attempted in the next two sub-­sections.

6.8.2.  The Partiality Problem Some recent work on luck egalitarianism has suggested that the theory is committed to a level of supervision of inequalities between subjects of justice which is so dense and interventionist as to be deeply unattractive, even for those who have no complaints about egalitarian patterning as such. Such worries become particularly prominent when inequalities among individuals are explained by acts of non-­material sacrifice or assistance or friendship. As Hugh Lazenby remarks: Helping old ladies across the street, editing a novel for a friend or fixing their car might all be instances of brute luck giving that are objectionable from a luck egalitarian perspective.78

Why is this? Imagine, as a way of developing one of Lazenby’s examples, that Jordan edits Nick’s novel, but not Myrtle’s novel. Nick is now in the happy position of having had his novel edited, and thereby improved, whereas Myrtle’s novel remains unedited, and therefore less good than it might be. Let us also assume that the welfare gap that has now opened up between Nick and Myrtle falls within the scope of the currency of justice employed by luck egalitarianism.79 Does it follow that luck egalitarians must disapprove of, or at least seek to correct, Jordan’s uneven distribution of advantage between Nick and Myrtle?80 Because this worry affects, not just relatively ceremonial acts of gift-­giving, but also very basic and informal forms of partiality-­driven advantage, I will call it the Partiality Problem. 78  Lazenby (2010), p. 280. Lazenby here enlarges the scope of the discussion from the usual gift-­ giving context. Other discussions of gift-­giving and luck egalitarianism are offered by Miller (2014), Elford (2013), (2017), Williams (2013), and Segall (2015). 79  Though my stance on the currency issue is officially agnostic, it should be noted that many card-­ carrying luck egalitarians favour welfare-­involving metrics, on the grounds that welfare is what most obviously matters for its own sake to individuals. 80  Even if luck egalitarians do not propose to implement an ex ante prohibition of such behaviour, it will still seem counterintuitive to many people that, in the absence of distributive correction, Jordan’s actions introduce any injustice between Nick and Myrtle.

Three Further Problems  193 How is the Partiality Problem to be dealt with? Consider standard luck egalitarianism first. Here the welfare gap between Nick and Myrtle clearly lies beyond the control of either of them. If Lazenby is correct to suggest that ‘the fundamental issue from the luck egalitarian perspective is the unfair gain of the recipient relative to non-­recipients’,81 then Myrtle would appear to have a complaint. How does baseline-­relative luck egalitarianism address these concerns? There are two possible issues we need to distinguish between. The first of them is concerned with, specifically, the Boring Problem. Regardless of the choices made by Nick and Myrtle, the gap between them will be a matter of brute bad luck. Baseline-­relative luck egalitarianism can afford to dismiss this particular worry. But the substitution of baseline-­relative luck egalitarianism for standard luck egalitarianism does not retire a second worry, regarding Nick’s receipt of a benefit which seems not to be choice-­sensitive. Assume that there is nothing we can point to in Nick’s behaviour or choices which explains why he is advantaged over Myrtle. After all, this was Jordan’s decision to favour Nick, not Nick’s own decision. It follows that baseline-­relative luck egalitarianism, no less than standard luck egalitarianism, will be exposed to the Partiality Problem. Or so it may seem. Even so, a closer look at these issues suggests that baseline-­ relative luck egalitarianism may be able to relieve some of this critical pressure. Assume, then, that luck egalitarians are initially minded to assign Nick’s receipt of his benefit to the category of ‘choice’ rather than ‘luck’ in order to show that Jordan’s unevenly distributed largesse does not generate an injustice between Nick and Myrtle. How might they do this? It is natural to say that Nick’s receipt of this benefit cannot possibly reflect his choices, since it is Jordan, not Nick himself, who chooses to bequeath this benefit on Nick. This is of course true. But the benefit still arises in a perfectly intelligible way out of activities for which Nick is responsible: namely, his pursuits of friendship and artistic creation. That consideration will help baseline-­relative luck egalitarians, if they are so minded, to situate Nick’s receipt of Jordan’s non-­material gift in the category of choice, rather than luck, and therefore to declare themselves at peace with the resulting inequality between Nick and Myrtle. Since baseline-­relative luck egalitarians already have a principled basis for their acceptance of the fact that Nick’s and Myrtle’s choices can turn out differently, they have at least the beginnings of a response to the Partiality Problem. Of course, the success of this counter-­response depends on how outcomes are related to choices, which will depend, in turn, on how agents’ choices are individuated. It will not be theoretically satisfactory if the troublesome cases which trigger the Partiality Problem are automatically assigned to the category of ‘choice’,

81  Lazenby (2010), p. 280.

194  Justice, Luck, and Pairwise Comparisons while the non-­troublesome cases which intuitively call for egalitarian correction are automatically assigned to the category of ‘luck’, with no further rationale for the division between choice and luck. So this proposal remains, at best, a schema, which stands in need of further disciplined development. But it is better to have a rough-­and-­ready schema which offers some hope against the Partiality Problem than to be simply exposed to an undiluted form of it. In summary, baseline-­ relative luck egalitarianism has at least nascent resources for dealing with the Partiality Problem.

6.8.3.  The Pluralism Problem A further possible advantage of baseline-­relative luck egalitarianism emerges from the Underdetermination Problem. Against some criticisms that have been made of it—for example, that it is committed to making demeaning inquiries into the source of subjects’ disadvantages82—luck egalitarians have often appealed to a pluralist version of luck egalitarianism, where the value of justice has to be honoured along with, and thus tempered or kept in check by, the other values which are relevant to political morality, such as respect for privacy and economic efficiency. Faced with such objections, luck egalitarians often seek refuge in the pluralist character of their theory.83 Typically, they do not adopt the maxim of fiat justitia ruat caelum. Even for the just liberal state, there are more values that need to be upheld than simply distributive justice. There are at least two potential problems with this pluralist manoeuvre, as I will call it. First, it seems to encourage an explanatorily indistinct, intuition-­dependent approach to the weighing of these different values: when does justice cede its place to other values, and when, by contrast, does justice reign supreme?84 Does justice beat a retreat whenever the results of prioritizing it look embarrassing? That seems explanatorily unsatisfying. Call this the Intuition Problem. Second, it might already seem embarrassing that justice recommends morally counterintuitive courses of action which only other values can save the luck egalitarian from having to endorse in conclusive form. Embarrassment can surely await values in their pro tanto form, not just in their conclusive or all-­things-­considered form. One pertinent example, for egalitarians, is the levelling down objection: the objection that egalitarianism will be just as satisfied with egalitarian distributions that prescribe levelling down as with egalitarian distributions that avoid levelling 82  See, for example, Wolff (1998), Anderson (1999), and Lang (2009a) for some relevant discussion. 83  See, for example, Cohen (1989), p. 908. 84  In other words, pluralist luck egalitarianism appears to re-­embrace the intuitionism that Rawls (1971) was so keen—and surely for understandable reasons—to leave behind.

Three Further Problems  195 down.85 It seems embarrassing that egalitarians could endorse levelling down. Many egalitarians appear to agree. A common reply to the levelling down objection among strong egalitarians is that they will have other evaluative and normative commitments which ensure that they do not have to lend their all-­in support for levelling down. Should we leave it there? The problem with this response is that the levelling down objection does tease out a distinctive and troubling implication of egalitarianism. If the implications of egalitarianism are disturbing, as this style of response implicitly concedes, then we might expect egalitarians, as egalitarians, to be troubled by them. Again, it is suspicious or embarrassing that it is only other values—values that are usually given a lower billing in the list of political values endorsed by egalitarians—that can save egalitarianism from having to endorse these harsh and counterintuitive verdicts. Call this the Pro Tanto Problem. I will refer to these two worries together as the Pluralism Problem. Baseline-­relative luck egalitarianism can save luck egalitarians from the Pro Tanto Problem, precisely in virtue of the facts which give rise to the Underdetermination Problem. Because the value of N is not fixed in advance, the other values which need to be combined with justice in a satisfactory overall theory of political morality admit of theoretically easy combination, without these types of ad hoc adjustment. Moreover, justice, or the specific value of the egalitarian baseline that forms the heartland of justice in baseline-­relative luck egalitarianism, does not beat a retreat whenever the consequences of prioritizing justice seem problematic. Justice can endure as an equally weighty concern throughout our moralizing, though the other values which demand recognition in political decision-­making will have an effect on which value of N we operate with. This is a sizeable theoretical advantage. Of course, that does not entirely retire the Intuition Problem. We must still engage in something that looks exactly like intuitive judgement or weighing when the relevant values appear to be in collision. But the overall complexion of the Pluralism Problem is still improved, because it is not justice itself, but only the specific value of N instantiating its egalitarian centre, which admits of variation in situations when justice appears to collide with these other values.

85  See Parfit (1991), pp. 17ff., for the original statement of the concern in the contemporary literature. In the intervening period, the problem has generated a small industry. See Hirose (2014), ch. 3, for a reliable discussion. As a referee has suggested, the difference between avoiding harm and failing to help seems pertinent to our intuitive responses to the levelling down objection. I cannot explore these issues any further here, but it should be noted that at least some writers with luck egalitarian sympathies are happy to evade the worry about levelling down by switching from an egalitarian framework to a prioritarian framework: see, for example, Arneson (2000).

196  Justice, Luck, and Pairwise Comparisons

6.9.  Should We Settle for Baseline-­Relative Luck Egalitarianism? Baseline-­relative luck egalitarianism has some advantages. It solves the Boring Problem and meets the challenge provided by the Paradox of the Baseline. It also half-­solves the Pluralism Problem. The advantages do not end there. The Underdetermination Problem may not actually be all that problematic, and baseline-­relative luck egalitarianism is not obviously undone by the Partiality Problem. But baseline-­relative luck egalitarianism does face another problem, which ought to reduce our overall level of confidence in it, and to motivate a change in theoretical direction.

6.9.1.  Baseline-­Relative Luck Egalitarianism and Normative Focus One central question we must grapple with is this. Does the concern with individuals’ various relationships with the egalitarian baseline, as opposed to a direct concern with the actual inequalities that obtain in pairwise comparisons among individuals, fully make sense of the specifically egalitarian tendencies in luck egalitarianism? One might suspect that it is the existence of relative inequalities, whose nature is fully revealed in pairwise comparisons between individuals, which generate the lion’s share of complaints about inequality: these are the problems of low self-­esteem, of hierarchy, deference, risk of domination, and so on.86 These are also the particular social problems with inequality that prompt many to embrace egalitarianism in the first place. They help to create the emotional and moral energy for the pursuit of political egalitarianism. Now these problems are often associated with relational egalitarianism, rather than distributive egalitarianism, but luck egalitarians are not debarred from taking an interest in them, and will want to accommodate these particular concerns to their own purposes.87 From an egalitarian perspective, these problems may appear to arise among individuals because some are better off than others for reasons that are beyond the control of the worse-­off, rather than because some of these individuals bear the wrong sort of relationship to an egalitarian baseline that is largely hidden and off-­stage. The relevant problems will therefore be revealed in facts about pairwise 86  For good discussions of such matters, see Scanlon (2003), (2018), O’Neill (2010), and Wolff (2013). 87 The debate between relational and distributive egalitarians has been pursued by Anderson (1999), (2010), Schemmel (2012), Scheffler (2005), (2015), as well as the other papers in Fourie, Schuppert, and Wallimann-­Helmer (2015), Lippert-­Rasmussen (2015), ch. 7, and Lippert-­Rasmussen (2018). As Lippert-­Rasmussen has emphasized, it is actually far from obvious what keeps relational egalitarians apart from distributive egalitarians, if there is nothing to stop relational concerns from being assimilated to (perhaps non-­obvious) distributive metrics. But I cannot resolve these tricky issues here.

Should We Settle for Baseline-­R elative Luck Egalitarianism?  197 comparisons among individuals, rather than facts about different individuals’ relationships with a common egalitarian baseline. In short, baseline-­relative luck egalitarianism stands accused of having the wrong normative focus. Call this the Normative Focus Problem.

6.9.2.  Parrying the Normative Focus Problem The Normative Focus Problem is a serious one, and it is not obvious how it should be handled. In my view, responsibility-­sensitive egalitarians who might otherwise be attracted to baseline-­ relative luck egalitarianism can make at least some headway against it. There are three relevant points to make. First, and most boldly, baseline-­relative luck egalitarians might simply deny that egalitarian concern is properly engaged by the fact that some individuals are better off than others for reasons that are beyond the control of the worse-­off. The Boring Problem squarely implies, as we have seen, that this cannot be the genuine site of concern, even if that is far from obvious upon preliminary examination. So if, as a result, baseline-­relative luck egalitarianism requires egalitarians to abandon their traditional concerns, then this is perhaps what egalitarians are best advised to do. The second point is that baseline-­relative luck egalitarianism still boasts the sort of structural credentials which allow it to count as some form of egalitarianism. The common relevance of the egalitarian baseline—the fact that the egalitarian baseline possesses a common significance for everyone—ensures that, in a just distribution, different individuals’ distributive shares display the same sort of choice-­sensitive relationship to the same baseline. That complex structural feature may qualify baseline-­relative luck egalitarianism to satisfy the moral spirit of egalitarianism, even if this form of egalitarianism must largely disinvest in pairwise comparisons between individuals. I have campaigned on a number of fronts against the central relevance of pairwise comparisons in this book. If egalitarians are better off without them, then why aren’t they entitled simply to abandon them, without apology or regret? Won’t they be better off if they do so? Isn’t that precisely what I am arguing for? The third point is that, plausibly, not every unchosen inequality between individuals contributes in a similar way to the problems, concerning domination, hierarchy, and unequal status, that tend to be enumerated by egalitarians. Inequalities which engage egalitarian concern will tend to be persisting and explainable in ways which suggest that those on the losing side of the inequalities were denied a real opportunity to have avoided domination, hierarchy, and lower status. Much may depend on where the value of N is set. But, if so, then the relevant remedies may be within the control of baseline-­relative luck egalitarianism. Such

198  Justice, Luck, and Pairwise Comparisons egalitarians can simply insist on a value of N that is high enough to avoid severe problems for those who are on the losing end of permissible inequalities.

6.9.3.  Sharpening the Normative Focus These replies prevent any immediate collapse in the prospects for baseline-­relative luck egalitarianism. But there may still be some damage to contend with, or else missed opportunities which some alternative calibration of egalitarian concerns is able to articulate in a more convincing way. The Normative Focus Problem encourages us to reconsider the question of defaults. When this issue first arose,88 it was suggested that an egalitarian theory’s objections to certain types of inequality should be significantly related to its commitment to a default egalitarian distribution. One of these commitments should cast light on the other. Baseline-­relative luck egalitarianism is unlikely to deliver satisfaction on this score. The reasons for adopting the egalitarian default have little to do with the explanation of why and when inequalities are permissible and why and when they are impermissible. Baseline-­relative luck egalitarianism explains permissible inequalities in terms of choice-­sensitive departures from the egalitarian baseline. But that explanation does not cast much light on why it is important to proceed from an egalitarian baseline in the first place. It would appear, then, that there is a lack of balance in the normative-­explanatory picture assembled by baseline-­relative luck egalitarianism. The default commitment to the egalitarian baseline and the account of bad-­making properties of inequalities are not in critically productive contact with each other. Here is another way of getting at the same complaint. It will be a tediously familiar point by now that, if there is a morally permissible inequality between Ernest and Bertie, only one side of this inequality can only ever be under the control of either one of these agents. That is what the Boring Problem teaches us. The mere fact that this inequality is an interpersonal one will then guarantee that every permissible inequality combines elements of option luck and brute luck. Brute luck cannot be expelled from the account of permissible inequalities, if there are going to be any permissible inequalities at all. And that fact, in turn, weakens luck egalitarianism’s capacity to understand or rationalize what it is aiming to track. Though I have indeed campaigned long and hard against the relevance of pairwise comparisons, I have done so for different reasons. In the context of resultant luck, it will be clear by now that interpersonal luck is essentially a distraction. This form of luck appears to present us with additional relevant data

88  See Section 6.3.4.

Justice and Structures  199 for determining the blameworthiness of individual agents, but its relevance is simply chimerical. Most of what we need to know is fully intact in the blameworthiness profiles we construct for each of these agents. In distributive justice, by contrast, pairwise comparisons are more difficult to dismiss. That is because, in the domain of distributive justice, we are centrally interested in equal and unequal relations, and in the inequalities that hold between different individuals. The problem with baseline-­relative luck egalitarianism is that it cannot make full sense of interpersonal inequalities, because for each instance of inequality it can make sense of only one side of it. Inequalities are bounded on two sides, and to have a fully satisfactory account of them, we need an account that can encompass both sides. Here is a slightly different way of getting at the same point. One helpful and uncontroversial way of summarizing the luck egalitarian programme is provided by Brian Barry’s ‘principle of responsibility’: ‘the principle that unequal outcomes are just if they arise from factors for which individuals can properly be held responsible, and are otherwise unjust’.89 The problem with the luck egalitarian programme, we are now in a position to see, is that these ‘factors for which individuals can properly be held responsible’ are shared among different individuals. If responsibility is shared, then it cannot be owned by specific individuals. No individual can bring it about, by her own choice or efforts, whether she will be in distributively equal or unequal relations with others. When all is said and done, this is the basic problem—the hard problem—for luck egalitarianism. Baseline-­relative luck egalitarianism displays the lacuna as a badge of honour, rather than trying to keep it hidden, and that feature actually allows it to make progress along a number of different fronts. It still deserves to count as a lacuna, however, and it exposes a structural weakness in the luck egalitarian programme. Luck egalitarianism aims to understand certain phenomena while depriving itself of the full set of tools to do so.

6.10.  Justice and Structures Luck egalitarianism, in any recognizable version of it, begins with individual responsibility. Its account of permissible and impermissible inequalities must be constructed out of that basic notion. But that model seems destined to fail to make sense of the broader questions of why equality is desirable, if it is desirable, and why certain forms of inequality are undesirable, if they are undesirable. We need, I think, to take a substantially broader view of the relevant phenomena if

89  According to Scheffler (2005), p. 7, Barry alluded to this principle in a presentation he gave in Berkeley in 2003.

200  Justice, Luck, and Pairwise Comparisons we are to have any chance of attaining a lucid understanding of the proper distributive relations. We need above all to make sense of justice and injustice as properties of a social system. It is the social system containing the various distributive relations among individuals with which we should start. This is for different reasons. First, in order to make sense of what we do as agents and social actors, we need to make sense of the system to which we belong, and which casts us in social positions, with certain fields of influence and impact. Even when we think that we are able to pick out the distributive results of our choices, these choices would not have the results they do unless other people were involved, in either a demand-­focused or supply-­ focused role.90 We contribute, along with many others, to larger outcomes that we might, variously, rejoice in or deplore: wealth, housing, economic stability, and culture on the one hand, but also poverty, homelessness, environmental damage, and various forms of deprivation and discrimination on the other hand. As Iris Young persuasively argues, we cannot make sense of acute housing shortages for single parents, for example, without paying attention to macro-­ factors, not just micro-­factors such as individual acts of kindness and thoughtfulness.91 Housing shortages are not solved by kindly landlords (though they may be exacerbated by callous ones). They are ultimately directed by economic and social policies, and they need to be assessed at that more structural level. A more general lesson follows about justice, which Young states as follows: Justice and injustice concern primarily an evaluation of how the institutions of a society work together to produce outcomes that support or minimize the threat of domination, and support or minimize everyone’s opportunities to develop and exercise capacities for living a good life as they define it. Social justice concerns the actions of particular individuals on the policies of particular institutions only secondarily, as these contribute to constituting structures that enable and constrain persons.92

Another relevant point is about the basic explananda of justice. Why is equality valuable, if it is valuable? What is wrong with inequality, at least for some instances of inequality? Baseline-­relative luck egalitarianism has too little to say about the former question, and it has only a partial purchase on the latter question. Moreover, the only answers that luck egalitarianism is in a position to return to these two questions are not in critically productive contact with each other. The

90 Think of Smilansky’s example of Bill Gates as the most successful Effective in Section 6.7. Whatever talents and energies Bill Gates brings to the table, his choices will not be remunerative at all, let alone as remunerative as they are, without the contributions of other people. 91  See Young (2003). 92  Young (2003), p. 7. See also Young (2011).

Justice and Structures  201 switch to institutions and structures helps us to build explanatory bridges, and allows us to think about the larger social and political significance of inequality. John Rawls tells us that the subject of justice is the basic structure of society, which concerns ‘the way in which the major social institutions distribute fundamental rights and duties and determine the division of advantages from social cooperation’.93 The basic structure is comprised of the major socio-­ economic institutions which constitute the matrix from which life-­chances are distributed. Rawls’s famous views about justice arise from his assumption that we are all equal members of the society in which these structure exist. Since these structures are malleable, we can ask how they should be structured in such a way as to express our status as equal subjects of justice. The resulting account of inequalities takes its cue and direction from that more fundamental concern. Rawls has often been treated as an intellectual ancestor for luck egalitarianism, because his account, on a quick summary of it, seems to be committed to the neutralization of luck. His primary distributive target is the influence of morally arbitrary factors on distributions, and moral arbitrariness is often thought to be roughly equivalent to luck. Perhaps all the relentless study of Rawls’s work over the last few decades has actually blunted our receptivity to it, because I do not think that Rawls should be read in this way. I shall attempt to show in Chapter 7 that the truth on this score is more complicated, and that Rawls’s commitment to luck-­ neutralization is much more attenuated than this standard interpretive picture suggests. 93  Rawls (1971), p. 7.

7

What is Arbitrary about Moral Arbitrariness? 7.1. Introduction A striking influence on the development of luck egalitarian doctrine, according to many of its proponents, was John Rawls’s theory of ‘justice as fairness’. This was a point foreshadowed early in Chapter 6, and re-­affirmed at the end of it.1 Standard readings of Rawls propose deep similarities between his hostility to the distributive influence of morally arbitrary factors and luck egalitarians’ hostility to brute bad luck. Rawls is often characterized as an ‘embryonic’ luck egalitarian,2 and we are told often enough that luck egalitarians ‘see themselves as Rawls’s intellectual descendants’,3 or that luck egalitarianism ‘obviously draws inspiration’4 from justice as fairness. We have, by now, identified some difficulties with luck egalitarianism. Does the infection spread to Rawlsian views about justice? I want to deny this. It seems to me a mistake to interpret Rawls as an embryonic luck egalitarian, or as someone who was gesturing at ideas which attained a more refined and theoretically satisfactory form in the writings of Dworkin, Arneson, Cohen, and others. Now Rawls is definitely interested in luck; he thinks that luck ought not to enjoy distributive significance. This fact alone makes his views relevant to the concerns of this book. But his way of dealing with luck is different from many of his left-­ liberal responsibility-­sensitive descendants, and these differences can help to save him from the objections to luck egalitarianism that we have already uncovered. It can also allow him to press a new sort of case for political egalitarianism. In arguing for this revised understanding of Rawls, I will have other fish to fry as well. This should not be surprising, given the highly holistic, systematic nature of Rawls’s theorizing. Many things hang together in his system, and adjustments to our understanding of one of these items are likely to reverberate elsewhere. In particular, it is often suggested in textbook discussions of distributive justice that the contractarian apparatus Rawls employs in A Theory of Justice for the presentation 1  See Chapter 6, Section 6.2.3. 2  For this phrase, see Freeman (2007b), p. 119, and Scheffler (2003), p. 8. Like me, Freeman and Scheffler do not support this interpretation of Rawls. Their views will be consulted as we go along. 3  Seligman (2007), p. 292. 4  Sher (2014), p. 2.

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0008

Introduction  203 of his justice as fairness risks being theoretically dispensable, or redundant.5 On this view, the contract is little more than an eye-­catching though potentially distracting presentational device, modelling arguments whose normative source lies entirely elsewhere. A further claim, which often accompanies this view, is that the relevant normative source for Rawls’s contractarian argument derives entirely from Rawls’s opposition to distributive outcomes which are owed to morally arbitrary factors. It is the anti-­arbitrariness argument, not the involvement of a hypothetical contract, which does all the interesting justificatory work. The contract itself, strictly speaking, is redundant. Call this the Redundancy View.6 The Redundancy View need not, in and by itself, be considered an embarrassment for Rawls. Rawls himself describes the contractarian apparatus, on more than one occasion, as a ‘device of representation’.7 He does so, admittedly, under cumulative pressure from critics of A Theory of Justice. Even so, the remark does not obviously constitute a retraction of his theory, but rather a clarification of what is doing the heavy lifting in it. Despite Rawls’s readiness to de-­emphasize the significance of contractarian thinking, I nonetheless believe that he risks being too concessive on this score. By being more exacting about the status of the contract, and in aligning the contract to the anti-­arbitrariness argument in a different way, I think it may be possible to locate even more depth and weight in his system. The most important aim of this chapter is to argue for what I call the ‘Irrelevance Interpretation’ of moral arbitrariness, in preference to the standard ‘Neutralization Interpretation’ of the same notion.8 The Irrelevance Interpretation adopts a different view about the scope and the significance of Rawls’s hostility to moral arbitrariness than the familiar one advanced by the Neutralization Interpretation. I shall contend that Rawls was not aiming to expunge the influence of morally arbitrary factors on distributions, but rather denying their relevance to the selection of principles of justice. The two aims are crucially different. Rawls’s treatment of luck is different from that taken by luck egalitarianism, and he makes no extensive investment in the use of pairwise comparisons. When we accept the Irrelevance Interpretation, we are also in a better position to see why the Redundancy View should be dismissed. This is the second aim of the chapter, which is facilitated by the adoption of the Irrelevance Interpretation: 5 See, for example, Kymlicka (2002), pp. 60–75, for a well-­explained version of this story; cf. Gauthier (1977), p. 139, n. 7. 6  See Barry (1988) for a lively defence of this view. 7  Rawls (1993), p. 27, and Rawls (2001), p. 17. 8  The critical position adopted in Cohen (2000) and (2008) is perhaps the most conspicuous recent example of the Neutralization Interpretation. But Cohen did not expend much effort to secure it on purely exegetical grounds. It struck him as obvious, unlike the further implications for Rawls’s the­or­ iz­ing which he then went on to explore. Cohen’s basic approach is understandable. The relevant ideas he draws upon had been kicking around ever since the publication of A Theory of Justice.

204  What is Arbitrary about Moral Arbitrariness? it is to rebut the Redundancy View, and to reposition the role of the contract in his system. So a revised understanding of Rawls’s opposition to moral arbitrariness will, in due course, be partnered with an adjusted understanding of why the hypothetical contract matters to Rawls’s justice as fairness. The Irrelevance Interpretation establishes a comfortable conceptual distance between justice as fairness and the luck egalitarian theories which were examined in Chapter  6. Because the role that luck plays in his system is very different, Rawls’s justice as fairness is not beset with the problems faced by luck egalitarianism.9 Partly as a result of this different orientation towards luck, Rawls’s theory strikes me as largely defensible. I cannot defend it in full in these pages; this task would probably require more than one book, let alone one chapter. There will be enough room, even so, to advert to some significant advantages of a broadly Rawlsian approach to distributive justice. Two notes before going any further. First, I will tend to focus on remarks in A Theory of Justice,10 and will largely abstain from any very detailed commentary on continuities and discontinuities between Rawls’s early work and the ‘political turn’ in his later work, in which he develops his distinctive political liberalism.11 As it happens, I do not think the interpretation offered here imposes any intolerable strain on the project of establishing respectable continuities between the earlier and the later work, but no special attempt will be made to uphold that claim. Second, I am not seeking copper-­bottomed exegetical evidence for my interpretation. The reflections to follow should be regarded for the most part as a sympathetic reconstruction of Rawls’s theory, rather than as a primarily exegesis-­ driven project. However, my interpretation does claim some exegetical basis: Sections 7.4.1 and 7.4.2 will offer important and familiar textual evidence that, when fully digested, should lead us away from the Neutralization Interpretation. The chapter is organized as follows. In Section 7.2, I will do some further brief stage-­ setting, and state the ‘Redundancy Challenge’ about the status of the contract, as well as some other central worries.12 The orthodox Neutralization Interpretation of moral arbitrariness will be outlined in Section 7.3. In Section 7.4, I explore a striking piece of textual evidence in Rawls’s A Theory of Justice which is routinely noted, but which proves perplexing when viewed through the lens of the Neutralization Interpretation. In Section 7.5, and building on that 9  Tomlin (2012) thoughtfully explores various ways in which luck egalitarianism and Rawls’s just­ ice as fairness may in fact be mutually consistent, differing only in focus and scope. In my view, there may still be some degree of theoretical rivalry between them, since differences of emphasis and focus can often testify to a higher-­order disagreement about aims and purpose, rather than to a division of labour which everyone can happily agree upon. 10  I will also focus on the original edition of A Theory of Justice, rather than the 1999 Revised Edition (Rawls 1999a). Since my interventions in this debate tackle a critical picture of Rawls with a long history, we might as well focus on the text that directly inspired that history. 11  See, in particular, Rawls (1993). 12  The Redundancy View, already mentioned, is one way of meeting the Redundancy Challenge.

Redundancy in Justice as Fairness  205 discussion, I outline the rival Irrelevance Interpretation of moral arbitrariness. In Section 7.6, I consider how Rawls can withstand the Redundancy Challenge. In Section 7.7, I will appeal to the preceding argument to defeat a searching challenge to it due to G. A. Cohen. Some of the strands of this story will carry over to Chapter 8, when other problems of arbitrariness will be tackled.

7.2.  Redundancy in Justice as Fairness So that we are reminded of Rawls’s destination, I begin with the conclusion of his argument. His ‘general conception of justice’ is stated as follows: All social primary goods—liberty and opportunity, income and wealth, the bases of self-­respect—are to be distributed equally unless an unequal distribution of any or all of these goods is to the advantage of the least favoured.13

Rawls also goes on to develop a ‘special conception of justice’, consisting of two principles, ranked in lexical order: First Principle Each person is to have an equal right to the most extensive total system of equal basic liberties compatible with a similar system of liberty for all. Second Principle Social and economic inequalities are to be arranged so that they are both: (a)   To the greatest benefit of the least advantaged, consistent with the just savings principle, and (b)   Attached to offices and positions open to all under conditions of fair equality of opportunity.14 The various details concerning the interpretation and defence of the special conception of justice are not of supreme interest to us here. In fact, the main thing we need to keep in mind about justice as fairness is sufficiently plain in the unqualified general conception of justice: this is its strong presumption in favour of equality in respect of the social primary goods of wealth, income, and the bases of self-­respect. As Nozick put it, equal distribution of the social primary goods is the rest position of the Rawlsian system, deviation from which may be caused only by

13  Rawls (1971), p. 303.

14  Rawls (1971), pp. 302–3.

206  What is Arbitrary about Moral Arbitrariness? moral forces.15 But why does Rawls think social justice demands equal, or at least roughly equal, shares of liberty, income, wealth, and so on?

7.2.1.  The Official Argument Rawls’s official argument is the contractarian argument.16 Under the guise of a hypothetical contract, justice as fairness consists in the principles of justice that would be chosen by the parties assembled in the original position, choosing in a mutually self-­interested way under a thick veil of ignorance. The veil of ignorance permits the deliberators to have general knowledge about their society, the principles of political economy, and so on, but conceals from them some extensive items of information about themselves which Rawls deems to be ‘morally arbitrary’, or ‘arbitrary from a moral point of view’. These encompass people’s ‘natural’ endowments, their ‘social’ endowments, and also their ‘comprehensive conception of the good’. It is the fact that conceptions of the good, as well as endowments, are encompassed by the veil of ignorance which makes the veil ‘thick’, rather than ‘thin’. A person’s natural endowments are constituted by her collection of natural or genetic features or qualities, while her social endowments are constituted by her collection of social features or qualities. Both sets of endowments are simply given to us; we inherit them, and cannot take credit for them. It might be remarked immediately that this is not a clean distinction, since features we might be tempted to place on the ‘natural’ side of the endowment collection will still tend to reflect the influence of social sources of nurture or maintenance. Rawls can agree. He treats them in a similar way, which means that he need not worry about the exact relationship between them.17 The effect of imposing this veil of ignorance on us is to achieve a uniformity of characterization among the contracting parties. In essence, each party is nothing less and nothing more than a moral personality: an individual with a sense of justice, and a comprehensive conception of the good, with a set of natural and social endowments on which she can draw in order to pursue that conception of the good. Given this uniformity, there is no bargaining as such in the original position. Each party aims to maximize her index of social primary goods in ways which cannot fail to produce the same results for everyone else.18

15  Nozick (1974), p. 223. 16  The main lineaments are laid out in Rawls (1971), ch. III. See also Rawls (2001), Parts I and II. Helpful commentary is offered by Freeman (2007a), esp. chs. 2–4. 17  Nagel (1997) places greater emphasis on the correction of social deficits than natural deficits. But he would then appear to face the problem of distinguishing between natural and social contribution. This is a challenge not to be taken on lightly. 18  In his influential critique, Sandel (1998) makes the fact that the decision-­making in the original position makes no real demands on the plurality of contractors a source of worry about Rawls’s

Redundancy in Justice as Fairness  207 In these circumstances, Rawls thinks that the parties in the original position will maximin their shares: maximize the share of the worst-­off member. The presence of maximin reasoning is specifically reflected in the difference principle in the second principle of justice.19 Thus the question of what principles of justice there should be is resolved in terms of decision-­making under conditions of uncertainty in the original position. By reducing the problem of justice to an exercise of decision-­making under conditions of uncertainty, Rawls thinks it is possible to make more determinate progress.

7.2.2.  The Unofficial Argument Mention of morally arbitrary factors brings us to the unofficial or intuitive argument for justice as fairness. This argument tells us why it is appropriate to remove all this information from the contracting parties. The official argument has already instructed us that Rawls denies the parties access to facts about their own social and natural endowments, as well as their more comprehensive conceptions of the good. Due to the imposition of this thick veil of ignorance, the parties in the original position are symmetrically situated, and are thereby motivated to choose roughly egalitarian principles to govern their common economic and public life, where the ‘social primary goods’ of wealth, income, social liberties, and the bases of social self-­respect are roughly equally distributed. I will focus on the argument which quarrels in turn with the distributive influence of social endowments, and then natural endowments.20 First, Rawls argues for the superiority of ‘fair equality of opportunity’ over ‘careers open to talents’. These are rival versions of equal opportunity. Equality as careers open to talents forbids legal or formal discrimination, but allows life chances to vary along with unequal social endowments. This distributive hostage to fortune is quarrelled with in the following passage: The existing distribution of income and wealth . . . is the cumulative effect of prior distributions of natural assets—that is, natural talents and abilities—as these have been developed or left unrealized, and their use favoured over time by social circumstances and such chance contingencies as accident and good fortune. Intuitively, the most obvious injustice of the system of natural liberty is that it permits distributive shares to be improperly influenced by these factors so arbitrary from a moral point of view.21 theorizing. As I see matters, the same feature happily confirms that Rawls is not falling into the traps that were waiting all along for luck egalitarianism. 19  This protects Rawls’s egalitarianism from the worry that it might support levelling down. See Chapter 6, Section 6.8.3, for mention of the levelling down objection. 20  See, in particular, Rawls (1971), §12. 21  Rawls (1971), p. 72; emphasis added.

208  What is Arbitrary about Moral Arbitrariness? Once fair equality of opportunity is favoured, there are still important decisions left to make. In particular, fair equality of opportunity is consistent with both ‘liberal equality’ and ‘democratic equality’. Liberal equality aims to neutralize or correct the effects of unequal social endowments, but is untroubled by the impact of unequal natural endowments: [D]istributive shares are decided by the outcome of the natural lottery; and this outcome is arbitrary from a moral perspective. There is no more reason to permit the distribution of income and wealth to be settled by the distribution of natural assets than by historical and social fortune.22

So it would seem that the same moral reasoning which opposed the uncorrected influence of arbitrary social inequalities can now be re-­applied to arbitrary natural inequalities. Rawls appears to agree: [O]nce we are troubled by the influence of either social contingencies or natural chance on the determination of distributive shares, we are bound, on reflection, to be bothered by the other. From a moral standpoint the two seem equally arbitrary.23

We are left with democratic equality, which attempts to neutralize the influence of unequal social and natural endowments. Democratic equality combines equality of fair opportunity with the difference principle, which only permits in­equal­ ities that raise the expectations of the worst-­off, whoever they happen to be.

7.2.3.  Two Challenges Rawls attaches priority to the contractarian argument. He says, of the unofficial argument, that . . . none of the preceding remarks are an argument for this conception [of justice], since in a contract theory all arguments, strictly speaking, are to be made in terms of what it would be rational to choose in the original position.24

It has struck many thinkers, however, that justificatory priority must lie in the other direction. I will cite two familiar reasons.25 22  Rawls (1971), p. 74; emphasis added. 23  Rawls (1971), pp. 74–5; emphasis added. 24  Rawls (1971), p. 75. 25  Other questions remain. For example, why does Rawls think that the parties in the original position would actually embrace maximin reasoning? Why not take a moderate risk that maximized their expected benefits? See, for example, Harsanyi (1975), and Hirose (2014), pp. 26–8, for a helpful

Moral Arbitrariness: The Neutralization Interpretation  209 First, the moral significance of hypothetical arguments is notoriously elusive. Dworkin claimed that hypothetical contracts are not pale forms of contract, but rather no forms of contract at all.26 They enjoy none of the binding power of ordinary contracts. Unlike some uses to which hypothetical contracts have been put, such as the question of how the state can acquire legitimacy, there is no obvious sense in which this hypothetical contract can be relied upon as disclosing what the parties in the original position ‘really’ want, given the absence of opportunity to ascertain what these people actually want. What each of them really wants from a theory of justice is likely to be quite different from what other people want. Call this the Hypothetical Challenge. Second, the key feature to the contractarian argument is decision-­making under the veil of ignorance. To enjoy any authority over the principles of justice, these conditions must be justified. To justify them, we need to attend to what the unofficial argument says. But when we have done that, the contractarian apparatus seems merely ornamental. A related way of getting at this point that is that Rawls seems all too ready to concede that the outcome of the contractarian apparatus is delivered by the way the apparatus is set up.27 The outputs of the contractarian procedure are only as good as its inputs, all of which are supplied by the intuitive, anti-­arbitrariness argument. Such description-­rigging seems problematic if Rawls holds that the outcome of the contractarian procedure has normative significance over and above the moral credibility of the premises mobilized in its construction. As Kymlicka succinctly says: All the major issues of justice . . . have to be decided beforehand, in order to decide which description of the original position to accept. But then the contract is redundant.28

This is the Redundancy Challenge, which we have already anticipated.

7.3.  Moral Arbitrariness: The Neutralization Interpretation To tackle these standard criticisms of the contract, it is sometimes urged that we should allow Rawls’s hostility to morally arbitrary influence to shoulder the entire

summary of some of the concerns. I will ignore this important question, which has generated an enormous subsequent literature, in order to maintain focus on the role of moral arbitrariness and the nature of the contract, which is immediately downstream from it. 26  Dworkin (1973), p. 501. 27  See Rawls (1971), p. 121. 28  Kymlicka (2002), p. 68. See Hampton (1980), Buchanan (1990), and Freeman (1990) for other ideas about the interpretation of justice as fairness.

210  What is Arbitrary about Moral Arbitrariness? normative-­explanatory strain.29 This commits us, in turn, to figuring out what he means by ‘moral arbitrariness’. The Neutralization Interpretation of moral arbitrariness insists that Rawls is committed to the full neutralization or annulment of morally arbitrary inequalities among individuals. But what is so bad about moral arbitrariness, such that it cannot be permitted to have any unequal distributive effects? The possibilities to follow provide sharpened readings of the relevant considerations, in such a way as to support the Neutralization Interpretation.

7.3.1.  The Desert-­Sceptical Variant It has sometimes been suggested, especially in older literature,30 that Rawls neutralized the operation of morally arbitrary factors in the design of his original position because of his scepticism about individual desert, or because of his doubts about our full moral ownership of our various traits. On this Desert-­ Sceptical Variant, if an individual’s possession of certain traits is morally arbitrary—if she cannot take any credit for their acquisition or emergence—then Rawls denies that this individual can deserve anything in virtue of the deployment of these traits. This is how, in effect, Nozick reads Rawls.31 His interpretation of Rawls can be summarized as (N): (N) I do not deserve the benefits that flow from my benefit-­generating activities, and therefore have no prior right to keep those benefits, since what enables me to engage in these activities is that I possess endowments which are morally arbitrary, and for which I can therefore claim no credit.

If (N) is taken as the correct interpretation of Rawls, then it seems exposed to Nozick’s rejoinder that desert need not ‘go all the way down’.32 Desert has to start somewhere; the foundations underlying desert do not themselves have to be deserved. They simply have to be possessed by the person, or related to her in the right way.

29  Admittedly, this line of interpretation does not tackle the complaint that Rawls does not take sufficiently seriously our ownership of, or identification with, our particular traits. See Nozick (1974), pp. 213–31, Sandel (1998), chs. 1–2, and Sandel (2005). Some of the components of that line of argument will be encountered in this section. 30  See, in particular, the references to Nozick and Sandel listed in n. 29. See also Sher (1987), ch. 2, and Pogge (1989), chs. 1–2 for valuable discussions. 31  Nozick (1974), pp. 183–231. 32  Nozick (1974), p. 225.

Moral Arbitrariness: The Neutralization Interpretation  211 We can add two further points. First, we do not want to end up with the self-­ defeating conclusion that no one ever deserves anything, since that would jeopardize the prospects of the legitimacy of any distribution, egalitarian or otherwise. Second, Rawls himself denies that moral desert plays any role, positively or negatively, in his argument. He writes: The principles of justice that regulate the basic structure and specify the duties and obligations of individuals do not mention moral desert, and there is no tendency for distributive shares to correspond to it.33

This tells us both that Rawls is not a sceptic about moral desert, and that moral desert is not supposed to be doing any work for him. These challenges seem sufficiently powerful to look for critical possibilities other than the Desert-­ Sceptical Variant.

7.3.2.  The Inequality Variant So what other materials can we use? Some Rawlsian passages emphasize, not the fact that we do not deserve our natural assets, but the fact that we do not deserve whatever exact place is assigned to us on the spectrum of the interpersonal distribution of those assets: It is one of the fixed points of our moral judgments that no one deserves his place in the distribution of natural assets any more than he deserves his initial starting place in society.34

According to Samuel Freeman, what Rawls is specifically targeting here is not the claim that we are entitled to, or enjoy full ownership over, our natural assets, but the inequalities which exist between our natural assets.35 How does that help? It helps because it demonstrates that Rawls is not operating with the assumption, imputed to him by Nozick, that our natural assets are being treated by justice as fairness as a ‘common’ or ‘collective’ asset, morally owned by society and not by the individuals from which they are drawn. Even if we concede that, say, Fran is the proper owner of her natural assets, NF, and that Dan is the proper owner of his natural assets, ND, we must surely accept that Fran does not deserve to be better off than Dan in case it turns out that NF exceeds ND. And, by symmetry, we must surely accept that Dan does not deserve to be 33  Rawls (1971), p. 311. 34  Rawls (1971), p. 311; emphasis added. 35  Freeman (2007b), esp. pp. 115 ff.

212  What is Arbitrary about Moral Arbitrariness? better off than Fran in case ND exceeds NF. Neither Dan nor Fran could reasonably dissent. Rawls clearly thinks the point is obvious. As he asks in a later discussion: ‘Do people really think that they (morally) deserved to be born more gifted than others?’36 Clearly any inequality between Fran and Dan in respect of their natural assets has nothing to do with the levels of responsibility and desert, or facts about ownership, which it would be appropriate to assign to either of them. The gap between ND and NF is a normative no-­man’s-­land. No single individual has any title to this territory based on self-­ownership, desert, or any other consideration that could intelligibly function as the basis for that individual’s claim. But if that is so, then this is territory which is apt to be colonized by moral forces of the sort Rawls is offering. We should pause to note an interesting difference between Rawls’s theory and luck egalitarianism in this respect: the gaps between individuals in Rawls’s theory seem ripe for a moral intervention, whereas these gaps actually pose a conceptual problem—the Boring Problem—for luck egalitarianism. This point may serve to motivate certain features concerning the design of the original position, where one effect of the veil of ignorance is to make the inequalities between levels of natural endowment unknown, thereby constraining the parties’ selection of principles of justice. We can call this the Inequality Variant. The Inequality Variant appears much more promising, but it can be coherently questioned unless more is added to it. To see why, it may be helpful to contrast Nozick’s Rawls with Freeman’s Rawls. As against Nozick’s (N), Freeman’s Rawls is embodied, in effect, in (F): (F) I do not deserve to keep the additional benefits (relative to others) that flow from my benefit-­generating activities, since what enables me to engage in those activities is that I possess superior endowments which are morally arbitrary, and for which I can therefore claim no credit.37

Now (F) seems to mark a decisive improvement over (N), since (F) is not obviously undermined by Nozick’s point that desert does not have go all the way down. In fact, and as we have seen, (F) shies away from making any general conceptual claim about desert, being concerned instead with inequalities between individuals’ endowments. It objects to a system of distributive justice under which some people should have more than others simply because they have superior endowments for which they can take no credit.

36  Rawls (2001), p. 74. 37  I have italicized the words in (F) that help to distinguish it from (N).

Moral Arbitrariness: The Neutralization Interpretation  213 This ostensibly impressive showing for (F) is, I believe, largely illusory. As it stands, (F) does not manage to escape the orbit of the Desert-­Sceptical Variant. We have already seen that the foundations underlying desert need not themselves be deserved. Defenders of (F) will say, of course, that they can accommodate Nozick’s basic point, since their concern is with inequalities in respect of the distributive implications of those natural assets. But this position is harder to uphold than they realize. Defenders of (F) face the objection that, if they accept that the foundations underlying desert need not themselves be deserved, then they may actually have left it too late to complain about inequalities in respect of those foundations. If the foundations underlying desert are sound, considered for each and every agent in himself or herself, then it is none too clear why these foundations should suddenly start to look shaky just when, and because, other agents are brought into the picture. Applying this suspicion to our two-­agent example, we appear to have already agreed that Fran’s possession of NF does not disbar her from getting what she can get with NF, and that Dan’s possession of ND does not disbar him from getting what he can get with ND. Both Fran’s and Dan’s desert-­claims seem to be perfectly in order when considered separately. It is suspicious to start complaining just at the point when these desert-­claims are placed alongside one another. This query need not be accompanied by the claim that Fran has done anything to deserve to be more talented and therefore more successful than Dan, in case it turns out that Fran is more talented than Dan; we can say simply that Fran deserves whatever flows from the talents she possesses, and that no reason has yet been given for complaining about the size of these talents, and therefore the size of the benefits they generate, after it has been revealed that they exceed Dan’s. (Exactly the same points would apply, of course, should it turn out that Dan is more talented than Fran.) The Inequality Variant, without further supplementation, also runs into trouble. It needs to be buttressed in order to maintain a respectable distance from the Desert-­Sceptical Variant.

7.3.3.  The Anti-­Luck Variant Defenders of (F) need to change tack. One particular reformulation or re-­casting of (F) may enrol the notion of luck. This interpretation is of special interest, since it places the Rawlsian enterprise within touching distance of luck egalitarianism. On this proposal, which we can call the Anti-­Luck Variant, those who are sympathetic to (F) may protest that what they object to is not the fact that Fran has a certain level of talent for which she can claim no credit, but the fact, rather, that people who are less talented than Fran, such as Dan, have done nothing to deserve their bad luck. Similarly, Fran has done nothing to deserve her good luck.

214  What is Arbitrary about Moral Arbitrariness? What (F) really expresses, according to this variant, is the aspiration to neutralize the differential effects of luck. Once we enlarge the basis for equality in this way, we move beyond the restrictions of the Inequality Variant. A new variable—luck, good or bad—is introduced, in order to complete the disinvestment from individual desert. This, in effect, is the resulting view: (L) I do not deserve to keep the additional benefits (relative to others) that flow from my benefit-­generating activities, since what enables me to engage in those activities is that I possess superior endowments which are morally arbitrary, and for which I can therefore claim no credit, and which therefore are such as to generate merely good luck.

What (L) has going for it is that it is prepared to say more about why the unequal desert-­foundations are morally problematic, but in a way that goes beyond dependence on the Desert-­Sceptical Variant. But there are still problems to deal with here. The problem with (L) is that good luck is simply being defined in terms of getting more, as a result of superior natural endowments, and that bad luck is simply being defined in terms of getting less, as a result of inferior natural endowments. But if this is so, (L) barely advances on the claim that inequality is objectionable. It does not succeed in telling us why inequality is objectionable, because the categories of good luck and bad luck that have been used to illuminate (L) are cashed out, at the end of the day, simply in terms of equality and inequality. In particular, the opposition between good luck and bad luck is cashed out in terms of inequality, while the notion of luck is cashed out in terms of the undeserved foundations of desert-­claims. As interpreted above, then, (L) embodies a disappointingly uninformative connection between endowment, luck, and equality.38 The Anti-­Luck Variant does not get Rawls where he needs to be. We must aim to do better than this.

7.4.  Rawlsian Justice and ‘Simply Natural Facts’ The Neutralization Interpretation seems unpromising. We have by now earned the remit to look for a different interpretation. My pursuit of an alternative interpretation fastens, in the first instance, on two pieces of textual evidence. Though both of them are well-­known, I will suggest that it is not a straightforward

38  Since Freeman campaigns against the enrolment of Rawls into the camp of luck egalitarians, presumably he would resist (L) as well.

Rawlsian Justice and ‘Simply Natural Facts’  215 matter to grasp the commitments they contain. Again, familiarity may have blunted our receptivity to them.

7.4.1.  The Principle of Redress One obvious place to launch a Rawlsian reply to the Neutralization Interpretation, discussed by both Freeman and Samuel Scheffler,39 is Rawls’s insistence that justice as fairness is not equivalent to what he calls the ‘principle of redress’. According to this principle of redress: . . . undeserved inequalities call for redress; and since inequalities of birth and natural endowment are undeserved, these inequalities are somehow to be compensated for . . . The idea is to redress the bias of contingencies in the direction of equality.40

Rawls remarks that the principle of redress is ‘only . . . a prima facie principle, one that is to be weighed in the balance with others’.41 This does suggest that Rawls was not aiming at the Neutralization Interpretation. It helps to tell us what Rawls was not. But it does comparatively little to tell us what Rawls was, and why he refused to award the principle of redress an even fuller role in his theory. If morally arbitrary distributive influence is inimical to justice, then why not aim at its full neutralization? If it is on the moral radar at all, such that Rawls has reasons for seeking to reduce its influence, why leave any of it lying around? Why not seek to eliminate all of it? By the lights of the Neutralization Interpretation, Rawls’s hostility to morally arbitrary influences seems curiously half-­hearted. We need more of a narrative to cast light on these matters. There are glimpses of my preferred interpretation in Freeman and Scheffler. This is what Freeman has to say: There is no hint . . . of an explicit luck egalitarian commitment by Rawls to equalize the effects of arbitrary inequalities. Rather the implication is simply that differences in natural talents by themselves are not a reason for any pattern of distribution.42

And this is Scheffler’s view:

39  Freeman (2007b), pp. 118–19, and Scheffler (2003), p. 25. The point is also noted by James (2005), p. 288. 40  Rawls (1971), pp. 100–1. 41  Rawls (1971), p. 101. 42  Freeman (2007b), p. 116.

216  What is Arbitrary about Moral Arbitrariness? Rawls’s emphasis on the moral arbitrariness of people’s natural attributes and social starting points is meant to undercut our tendency to treat those factors as morally authoritative, especially when doing so would compromise something morally fundamental.43

The main claim in these passages seems to be that morally arbitrary assets have no authority to uphold an unequal distribution. That is indeed a key thought behind the Rawlsian system. But why should the arbitrariness of our assets prescribe an equal distribution? And why, having demonstrated a connection between anti-­arbitrariness and equality, would the anti-­arbitrariness argument then permit any deviations from equality?44 Why wouldn’t the tendency to equality be anything less than whole-­hearted, rather than being qualified in the way Rawls insists upon? These are the central questions to investigate.

7.4.2.  The Non-­Applicability Claim To make further progress, I want to focus on another crucial though less discussed piece of textual evidence for the Neutralization Interpretation. In many or even most expositions of Rawls’s justice as fairness, the following passage is frequently cited but rarely revisited, even by those who can barely wait to apply their critical scalpels to the Rawlsian system: The natural distribution is neither just nor unjust; nor is it unjust that persons are born into society at some particular position. These are simply natural facts. What is just and unjust is the way that institutions deal with these facts.45

We can call the claim announced in this passage the ‘Non-­Applicability of Justice Claim’—the ‘Non-­Applicability Claim’ for short—since it concerns the lack of applicability of the category of justice to these facts about natural and social endowments. I want to suggest that, when placed against the background of standard interpretations of justice as fairness, the Non-­Applicability Claim is surprising. To provide a satisfactory accommodation of it, we need to look at Rawls’s system in a new way. A terminological distinction before going any further: I will distinguish between the ‘endowments facts’ and the ‘lottery facts’. The endowments facts are simply the collection of facts about individuals’ natural and social endowments, while the lottery facts are facts about the distribution of those individuals’ natural 43  Scheffler (2003), p. 26; emphasis added. 44  See Section 7.7 for a discussion of Cohen’s treatment of these ideas. 45  Rawls (1971), p. 87.

Rawlsian Justice and ‘Simply Natural Facts’  217 and social endowments. (We can further subdivide these two types of facts, if we like, into natural endowments facts, natural lottery facts, social endowments facts, and social lottery facts.) The Non-­Applicability Claim is puzzling because it seems to be in conflict with certain passages which are often cited in connection with the Neutralization Interpretation: The existing distribution of income and wealth . . . is the cumulative effect of prior distributions of natural assets—that is, natural talents and abilities—as these have been developed or left unrealized, and their use favoured over time by social circumstances and such chance contingencies as accident and good fortune. Intuitively, the most obvious injustice of the system of natural justice is that it permits distributive shares to be improperly influenced by these factors so arbitrary from a moral point of view.46 [U]ndeserved inequalities call for redress; and since inequalities of birth and natural endowment are undeserved, these inequalities are somehow to be compensated for . . . The idea is to redress the bias of contingencies in the direction of equality.47

We have already encountered these passages.48 Take, for example, the second of them, which is concerned with the principle of redress. Even if Rawls does not pursue a full expression of this principle in his theory of justice, he is sympathetic to a qualified form of it, and his principles of justice are supposed to embody that sympathy. As a result, it would be hard to deny that there must be some sort of deficiency or shortcoming with the lottery facts. As they stand, they are not fit to radiate distributive influence. This is why—it is natural to surmise—the distributive implications of the lottery facts need to be corrected through the device of the veil of ignorance. But if there are reasons to fix or correct the distributive implications of the lottery facts, then it is natural to assume that those reasons must reflect the existence of a deeper complaint about the lottery facts themselves. The natural thought here is that, in the absence of these remedies, the lottery facts would be a source of distributive injustice. It is easy to imagine some critical resistance at this juncture. To take the more obvious case, natural endowments facts are clearly not authored into existence by any particular agent or process, and for that reason it may be held that natural lottery facts also resist any characterization in terms of justice. Natural endowments are not brought about by unjust acts, and they do not reflect unjust agency; people are simply born with whatever endowments they happen to have. Neither is the final distribution of natural endowments—the lottery facts—the 46  Rawls (1971), p. 72; emphasis added. 48  See Section 7.2.2.

47  Rawls (1971), pp. 100–1.

218  What is Arbitrary about Moral Arbitrariness? product of unjust agency or unjust processes. There is no active collaboration between procreators in order to produce the inequalities manifested by the lottery facts. Aren’t these the unremarkable truths which must lie behind Rawls’s endorsement of the Non-­Applicability Claim? In response, there is a difference between denying the injustice of the acts or processes which bring about the distribution of natural endowments, and denying the injustice of the distribution itself. The process might be morally blameless, but that does not prevent us from saying that the distribution to which that process gives rise is problematic: that distribution might at least be bad, even if wrongdoing was not involved in its commission. And the interpersonal ­ ­distribution of natural endowments seems problematic, from the point of view of  justice, because it risks allowing some people to be privileged over others in  morally arbitrary ways. But if the distributive consequences are morally ­unsatisfactory when they reflect too closely the influence of the lottery facts, then there would seem to be a line of complaint about the distributive implications of the lottery facts which terminates in a complaint about the lottery facts themselves. If the uncorrected distribution is problematic, as it would be if we were to operate with the system of careers open to talents, then we must surely accord the same verdict to the lottery facts which lie behind, or subvene, the problematic distribution. Or so it may seem. Rawls’s claim about the social lottery facts seems even more puzzling. Even if no one has deliberately authored into existence the social lottery facts—as Nozick often reminds us, these will usually arise without much intentional oversight— they are still inequalities within social space, and they are the products of a social system which, as we know, might have been structured differently. They are not the products of natural, procreative activity that unfolds, whenever it occurs, largely outside the domain of social justice.49 If the social lottery facts are problematic, then they are unjust. Yet Rawls appears to deny even that. The writ of the Non-­Applicability Claim extends to both the social lottery facts and the natural lottery facts. Both sets of facts are deemed to be entirely blemish-­free from the perspective of justice. By now, it is difficult to avoid confrontation with the challenge posed by the Non-­Applicability Claim. It seems that Rawls is determined to fix something which, according to his own view, does not need to be fixed. He appears to both affirm and deny that there is something to correct. Now we can agree that we cannot fix the natural lottery at source without risking severe intrusions into human procreative activity or bodily integrity, and we can also agree that we cannot fix the social lottery without institutional oversight. But we need not deny,

49  This is not to deny that questions of justice can arise with respect to procreative activity.

Rawlsian Justice and ‘Simply Natural Facts’  219 in either case, that there is a problem at source, which for various reasons we are in a position to tackle only retrospectively, institutionally, or obliquely. Moreover, if there is a problem at source, then it would appear that this problem must be counted as a problem of injustice. The injustice of distributions that are downstream from the lottery facts must surely be reflective, in part, of something that is deficient about those lottery facts. That deficiency, in turn, cannot be unrelated to matters of justice. We can at least say that the lottery facts are such as to be incapable of sustaining just distributions of the social primary goods. If that is so, then the lottery facts surely fall short of justice. But a plainer way of saying that the lottery facts fall short of justice is to say that they are not just, and there is only a short conceptual distance between saying that they are not just and saying that they are unjust.50 Rawls plainly wishes to deny any such claim. But how can he make good on this denial? And why does he want to?

7.4.3.  Deontic versus Telic Equality One way in which we can make sense of Rawls’s commitments is to call upon Derek Parfit’s distinction between Deontic and Telic Equality.51 According to Parfit’s terminology, Telic Egalitarians hold that ‘when we should aim for equality, that is because we shall thereby make the outcome better’;52 Deontic Egalitarians, by contrast, hold that ‘though we should sometimes aim for equality, that is not because we shall thereby make the outcome better, but is always for some other reason’.53 The other reasons tend to be concerned with wrongdoing: if no wrongdoing has been involved in the commission of an inequality, then it cannot be one to which Deontic Egalitarians will object. Parfit summarizes the distinction in the following way: ‘On the Telic View, inequality is bad; on the Deontic View, it is unjust’.54 Does Parfit’s distinction suggest that, when Rawls commits himself to the Non-­ Applicability Claim, he has Deontic Equality in mind? Perhaps the Non-­ Applicability Claim is explained simply by the fact that no wrongdoing was involved in the generation of the lottery facts. But Parfit also claims to find evidence in the following passage that Rawls was in fact a Telic Egalitarian, at least in part:

50  It may be protested that, for example, Beethoven’s string quartets are not just, but are not unjust, either. They lie entirely outside the category of justice. The same conclusion might apply to endowments facts as well. But the Neutralization Interpretation is not encouraging to that thought. 51  See Parfit (1991). 52  Parfit (1991), p. 3. 53  Parfit (1991), p. 8. 54  Parfit (1991), p. 9.

220  What is Arbitrary about Moral Arbitrariness? Aristocratic and caste societies are unjust because . . . the basic structure of these societies incorporates the arbitrariness found in nature. But there is no necessity for men to resign themselves to these contingencies.55

This passage is encountered just below the passage I have used to ascribe the Non-­Applicability Claim to him. Parfit thinks that this passage, with its references to incorporation and resignation, suggests that Rawls assumes that natural inequality is bad (if not wrong).56 He draws the conclusion that ‘[a]n objection to natural inequality is . . . one of the foundations of [Rawls’s] theory, and one of its driving forces’.57 Parfit’s interpretation requires some contortion. As he sees it, Rawls affirms both that lottery facts are neither just nor unjust, and also that aristocratic and caste societies are unjust because those lottery facts are bad. But if Rawls thought that the badness of lottery facts is such as to generate injustice in societies that preserve that badness, then why would he go to the trouble to deny that the lottery facts admit of talk of justice and injustice? I think there is a way of making sense of what Rawls says that does not involve this sort of contorted interpretation, and that resolves these passages in a more straightforward way. We should now begin to see what this proposal consists in.58

7.4.4.  Arbitrariness and Valence Rawls’s commitment to the Non-­Applicability Claim is not, I believe, the product of carelessness or obtuseness in his thinking. The best interpretation of it will ascribe to him a subtly different approach, and subtly distinct commitments. What the Non-­ Applicability Claim reveals is that the critical target of his broadside against moral arbitrariness is not what many Rawlsian critics have taken it to be. The data used to generate the challenge in Section 7.4.1 must be re-­described in order to tackle the underlying concerns. Rawls’s deep point, as I see it, is not that the arbitrariness of lottery facts gives us reasons to correct the effects of that distribution at a later point, on the grounds that lottery facts, if unchecked, will give rise to injustice. His point is better understood in the following way: the ‘simply natural’ facts about our endowments and the distribution of our endowments do not contain within them any hidden distributive recipe. They do not point in the direction of any resulting distribution. These facts have, in short, no valence in the domain of justice. Thus there is no distribution which can either reflect, or else break ranks with, the endowments facts or the lottery facts. There is simply no sense in which endowments facts or 55  Rawls (1971), p. 102. 56  See Parfit (1991), p. 10. 57  Parfit (1991), p. 10. 58  I return to the talk of ‘resignation’ and ‘incorporation’ in Section 7.4.4.

Rawlsian Justice and ‘Simply Natural Facts’  221 lottery facts could lead, in an unchecked way, to any particular distribution of social primary goods. In this sense, then, endowments facts and lottery facts can fall short of justice without being unjust. Justice has to be created by something else. We shall presently get to what else is needed to create justice. For the moment, we need to flesh out the idea that endowments facts and lottery facts have no valence in the domain of justice. It is uncontroversial that, in order to get to a subsequent distribution in which the subjects of justice end up with goods, we need rules to take us from endowment inputs to distributive outputs. We can call these ground rules, which regulate the terms of economic interaction between individuals.59 It is the ground rules which settle who gets what; the ground rules settle distributive shares among the individuals involved, as well as the rules or principles which explain which distributive changes are acceptable and which distributive changes require correction. But these rules are not grounded in, or justified by, or reflective of, any facts about endowments. They cannot be. Facts about endowments lack any such directedness. It might seem otherwise. Take, again, Fran and Dan. Imagine, as before, that Fran’s natural endowments facts can be summarized as ‘NF’, and Dan’s natural endowments as ‘ND’. And imagine, as before, that we are also inclined to say that NF is superior to ND. Imagine now two distributions: D1 and D2. In D1, Fran ends up with G+, while Dan ends up with G. G+ exceeds G. In D2, by contrast, the distributive shares are inverted: Fran only gets G, whereas Dan gets G+. We might think that D1, in which Fran ends up with more than Dan, provides a better ‘fit’ with the distribution of their natural endowments than D2, in which Dan ends up with more than Fran. Since we are assuming that NF is superior to ND, we will not be surprised by D1, in which Fran gets G+ and Dan gets only G. It is D2, where Fran only gets G and Dan gets G+, which carries the surprise. But this reasoning is flawed. At bottom, the calibration of Fran’s and Dan’s endowments is not a neutral exercise. To impute superior endowments to Fran over Dan is already to have taken sides on distributive fit. There are three basic points which need to be emphasized here. The first of them is this. Relative to D1, NF is superior to ND, if Fran’s possession of NF explains why she gets G+ and Dan’s possession of ND explains why he gets only G.  But relative to D2, ND is superior to NF, if within that distributive framework Dan’s possession of ND explains why he gets G+ and Fran’s possession of NF explains why she gets only G.  Judgements about the inferiority and superiority of endowments facts are internal to the set of rules that take us from endowment inputs to distributive outputs. To pretend that D1 is the ‘natural’ distribution, while D2 is an ‘unnatural’ or surprising distribution, is to labour 59  For an illuminating discussion, see Pogge (1989), esp. ch. 1. The phrase ‘ground rules’ is also taken from Pogge.

222  What is Arbitrary about Moral Arbitrariness? under the illusion that one level can be settled by another. It cannot be. The relationship between them has to be fixed by something else. This takes us to the second point. The fact that Fran ends up with G+ under D1, and only G under D2, cannot be wholly explained by the fact that her natural endowments level is NF. Where Fran ends up is dependent on the ground rule which takes us from NF to G+ in D1, and on the separate ground rule which takes us from NF to G in D2. (The ground rules will be complicated, of course, by the presence of social endowments and other facts about the social environment. But the point generalizes even when these are added.) The ground rule in D1 plots a connection from NF to G+, and the ground rule in D2 plots a connection from NF to G. The fact that NF leads to G+ in one distributive scheme and to G is another distributive scheme simply confirms that Fran’s distributive entitlements are not the product purely of her natural endowments. They are not settled by her natural endowments. They are the product of ground rules which, in effect, decide to install these connections between natural endowments and distributive entitlements. That decision is not fixed by facts about natural endowments themselves. And the generalized version of this claim also holds: when we combine the natural endowments facts with the social endowments facts about Fran, we still do not have a set of facts which settle the case for Fran’s collection of any particular level of distributive entitlement. The ground rules belong to a different category, and outrun any information which is supplied by facts we find within the category of endowments. Before leaving this point, we need to return to the passage used by Parfit to generate support for the Telic Egalitarian reading of Rawls. As a reminder: Aristocratic and caste societies are unjust because . . . the basic structure of these societies incorporates the arbitrariness found in nature. But there is no necessity for men to resign themselves to these contingencies.60

Resignation and incorporation can take different forms. If we note the distribution of natural assets and then propose ground rules which preserve that distributive structure as we plot the conversion to distributive entitlements, then in one important sense we have indeed resigned ourselves to lottery facts and incorporated their arbitrariness into our account of distributive entitlements. But these claims can be easily squared with the Non-­Applicability Claim. The point that the lottery facts do not, and cannot, settle the content of ground rules does not establish that the ground rules cannot be influenced by the lottery facts. The ground rules, after all, admit of any influence whatsoever. The lottery facts can be one such influence. We can make them exercise this influence. But it should not be maintained that, if

60  Rawls (1971), p. 102.

Rawlsian Justice and ‘Simply Natural Facts’  223 ground rules are modelled on the descriptive lottery facts, those ground rules can be represented as a response to the badness of those lottery facts. Now for the third point. We need not deny that Fran has a higher IQ than Dan, if that fact is meant to belong to the collection of information conveyed by the fact that she has NF, while he has ND, and by the claim that NF is superior to ND. But what makes IQ level a natural asset, in part, depends on what one can do with it in a social system—on whether it is a source of benefit or cost.61 And that will partly be dependent on what ground rules are operating. When all is said and done, the description we offer of Fran’s and Dan’s respective endowments, and our readiness to describe these endowments as sources of advantage or disadvantage, reflects only the fact that one set of endowments is correlated with a higher degree of benefit in the justice distribution. If we end up with D2 rather than D1, where Fran gets G and Dan gets G+, then we do not have a puzzling failure of ‘fit’ between their endowments and the assignments of goods in the distribution arising from them. Rather, we have simply confirmed that their endowment levels were not the sources of benefit which we might have taken them to be. The endowments facts do not have any independent pattern-­forming life of their own. They are just a shadow of the ground rules we have adopted. But the endowments themselves do not tell us how the ground rules should be formed. It is up to us. Thus we should be aware that endowments facts carry a double aspect. On the one hand, they are ‘simply natural facts’, just as Rawls refers to them. This description of the facts does not point us in any distributive direction. On the other hand, they can be described as endowments, which have already collected some form of distributive identity: an endowment is the kind of thing such that having more of it rather than less of it will get you further ahead in the distributive game. When Rawls says that our endowments are morally arbitrary, or arbitrary from a moral point of view, he should be interpreted as referring to the facts about endowments under their pre-­distributive description. And then his substantive point about the ‘arbitrariness’ of our endowments is that they do not provide guidance. They do no genuine arbitration about justice. They are normatively inert. The problem of ‘simply natural’ facts, for Rawls, is a problem of radical underdetermination. Their influence does not need to be neutralized, or corrected, or annulled; the problem with them, rather, is that they can have no influence. They lie within a category which enjoys no normative authority. 61  It also makes sense to think of IQ levels as a test, not of raw brainpower—whatever that might amount to—but of effective performance, or at least the capacity for effective performance, in a selection of cognitive tasks and exercises that happen to be valued or emphasized by our social system. That selection of tasks and exercises will, in turn, be culturally or socially specific. All of this can be admitted without succumbing to scepticism about the existence of robust facts about individuals’ natural constitution.

224  What is Arbitrary about Moral Arbitrariness?

7.5.  Moral Arbitrariness: The Irrelevance Interpretation There is a broader and less specialized sense of arbitrariness of which we need to be aware. This sense of arbitrariness arises from the platitudinous fact that any moral theory will be opposed to actions and practices which do not conform to it. These deficient actions and practices will reflect the influence of considerations which, relative to the standards of the theory, ought not to hold sway. The moral theory will be understandably hostile to the influence of these considerations. These considerations might then, without linguistic impropriety, be described as ‘arbitrary’, or as ‘arbitrary from a moral point of view’, since by the lights of the theory itself they enjoy an unmerited moral influence. We can take utilitarianism as an example. Utilitarians often condemn ordinary moral agents for adopting patterns of beneficence which are less than fully impartial. These patterns of beneficence can be interpreted as resting on the implicit judgement that spatial distance matters: the more distant the needy are from the donor, the less morally pressing their condition is. Utilitarians tend to condemn that judgement, because as they see it spatial distance has nothing to do with the strength of moral claims from moral patients. The weal and woe of moral patients far away from us is no less morally relevant than the weal and woe of moral patients nearer to hand. Distance is therefore irrelevant; it is arbitrary.62 One lesson to draw here is concerned with the scope of claims about arbitrariness: each theory is entitled to make its own determinations of arbitrariness. The other lesson to recover is concerned with the source of claims about arbitrariness: claims about arbitrariness will in this sense be theory-­ dependent. These ideas will now be applied to Rawls.

7.5.1.  Arbitrariness and the Veil of Ignorance For Rawls, the descriptive facts about our natural constitution and social background do not determine the principles of justice. So how does Rawls propose to derive principles of justice? As we know, he installs a contractarian apparatus, and then makes justice the product of the hypothetical contract. The significance of the hypothetical contract will be addressed in Section 7.6. For the moment, I want to return to the conditions under which the contracting parties choose the principles of justice. They do so, of course, under the veil of ignorance. A feature of the veil of ignorance which is under-­emphasized by proponents of the Neutralization Interpretation is that the veil encompasses, not just individuals’ knowledge of their endowments, but also knowledge of their comprehensive 62  See, again, Singer (1972) for a well-­known argument along these lines.

Moral Arbitrariness: The Irrelevance Interpretation  225 conceptions of the good. This feature suffices to make the veil thick, not thin. But why should the veil be this thick? It would be fatuous to observe that an individual’s comprehensive conception of the good, like her endowments, is inherited rather than deserved. That cannot be a significant characterization of an individual’s attachment to her conception of the good. It is none too clear what it would even mean to say that I do not deserve my beliefs that philosophy is worth doing, or that music nourishes the soul, or that you should hold on to your friends. If we can identify Rawls’s reasons for excluding individuals’ information about their respective conceptions of the good, then these reasons ought, in turn, to shed considerable light on his reasons for also excluding information about individuals’ social and natural endowments. This is because the imposition of the veil of ignorance over all these items of information is likely to have a common explanation. Our task is to figure out what that explanation is. Rawls assumes that the contracting parties in the original position have self-­ interested motivations: they are concerned to maximize their share of social primary goods. This is not to impute to ordinary people psychological egoism, but simply to ensure that everyone’s claim to justice is properly registered and expressed in social space. No one is overlooked or sacrificed for the sake of the many. Certain implications follow from this assumption, together with the unanimity assumption which requires that everyone in the original position agrees on the principles of justice. If everyone is assumed to be self-­interested, then we need to leave aside information which will obstruct the aim of achieving unanimity among the parties in the original position over the principles of justice. Otherwise we risk gridlock. But it is this aim of ensuring the removal of obstructions to unanimity-­ realizing deliberation which explains why the veil of ignorance encompasses, not just our endowments, but our conceptions of the good as well. These facts are just a distraction: if they were not removed from the awareness of the parties in the original position, then each of these parties would select different principles, depending on which principles happened to favour her. Principles that will favour me will not necessarily favour you, and vice versa. So the relevant sources of obstruction need to be removed. Even though we may agree that certain information needs to be jettisoned, we have not so far been given much of a guide about what we can afford to leave out, and what we cannot afford to leave out. We turn to that issue next.

7.5.2.  Interpreting Equal Standing There is a choice of which ground rules we adopt. Different ground rules will favour different individuals. As the Non-­Applicability Claim demonstrates, these

226  What is Arbitrary about Moral Arbitrariness? choices cannot be settled by any squarely descriptive information about what we are like. Is there a way forward which can keep everyone happy? Rawls proposes that the decision-­making in the original position must reflect our standing as free and equal persons. Why does Rawls start with freedom and equality? We can loosely reconstruct the central line of moral reasoning in a number of stages. At Stage 1, a certain group of individuals, who are in a position to interact with each other in social, economic, and political ways, come to realize that they need principles of justice to regulate their common existence. Their purpose is to regulate the common life they share with other people. They see their lives with other people as shared because they seek common benefits, and wish to protect themselves from common dangers. A form of equality is manifested even at this stage. Presumably there must be some explanation of why each party in the original position is qualified to be there,63 and it cannot be assumed in advance that any party to the original position is more qualified to be there than any other. In virtue of what are they all equally qualified to be there? The highest common denominator description of each of them is as a contributor to a scheme of social cooperation. That is an all-­or-­nothing property, and that is what principles of justice are for, in the first instance: to supervise the distribution of benefits and costs arising from a scheme of social cooperation. Each of the parties in the original position is no more and no less a contributor to the scheme of social cooperation than anyone else. At Stage 2, these principles of justice are agreed to be more than just a modus vivendi. These individuals are after justice, not just principles that everyone can live by. They wish the principles they agree upon to be reasonably permanent, and to be stable over time. To see whether a collection of principles, P, is one which these individuals could or would agree to live by, we would have to know what P was competing with. What is the relevant alternative to P? If the alternative to P was a Hobbesian (or even a Lockean) state of nature, it might be easy enough to justify P. For each individual, P would be much better than the state of nature, whether the problems with the state of nature were due to unrestrained violence, or gross inefficiency, or insufficient production of public goods, or a mixture of all of these. But that would leave us with a large range of possibilities for P. If the state of nature is the worst that a social world can be, there are lots of improvements to it that we can think of. Which of them deserves to be regarded as the most suitable realization of P? Any particular one of these principles will favour some individuals over others. Which of them should be chosen?

63  I return to this issue in Chapter 8, Section 8.4.

Moral Arbitrariness: The Irrelevance Interpretation  227 Plausibly, if there is a value of P which will treat everyone as an equal, then this value of P would be chosen by those parties, if they were in pursuit of a value of P that treated them as equals. But the parties do not yet know what equal concern amounts to.64 That is the task they have given themselves: to find an interpretation of equal concern. At Stage 3, because these individuals are not certain what the appropriate principles of justice are, they agree to set aside all extant conceptions of what equal concern consists in. They further agree that principles of justice are not settled, at this stage, by any existing differences of bargaining power. The capitulation to such bargaining power would reflect, in turn, either capitulation to a morally unstructured state of nature, or else capitulation to differential holdings accumulated under an existing scheme of justice whose choiceworthiness cannot be vouched for at this stage. At Stage 4, and in order to make further progress, these individuals decide not to consult certain pieces of information which will obstruct the attempt to arrive at principles they can all agree upon. For this reason, they lay aside their respective conceptions of the good, and also information about their social and natural endowments. To explain Stage 4 further: imagine that Tom has artistic talent, and Daisy has engineering talent. Tom organizes his life around the pursuit of artistic activity, while Daisy organizes her life around the pursuit of engineering activity. Now it would be in Tom’s interest to have arrangements whereby artistic talent would be more highly rewarded than engineering talent, and it would be in Daisy’s interest to have arrangements whereby engineering talent would be more highly rewarded than artistic talent. But they both swiftly come to realize that they cannot get anywhere in their attempt to arrive at agreement over principles of justice unless each of them retires the ambition to reach this agreement by privileging the particular talents they possess. Such information must be discarded. After all, it is not as though either of them has any antecedent right to shape the eventual settlement by privileging their particular talents, given the fact that they both have equal status. This is because each of them came to possess those endowments in the same way, by simply coming to inherit them.65 There is nothing that either of them can say to favour the antecedent privileging of their particular talents and endowments. Both Tom and Daisy are connected to these endowments in the same way, and both of them count equally as contributors to the scheme of social cooperation.66

64  See, again, Kymlicka (2002), ch. 1, and also Chapter 6, Section 6.2.2. 65  This is not to suggest, specifically, a model of genetic inheritance. The remark encompasses roles for both ‘nature’ and ‘nurture’. One of the advantages of Rawlsian political philosophy is that it is not forced to make difficult calls on how the division between nature and nurture is settled. 66  Nagel (1997) offers a good discussion of this point.

228  What is Arbitrary about Moral Arbitrariness? But why not allow the resources of bargaining theory some play here? I am not denying that relative bargaining advantages between Tom and Daisy, based presumably on facts about their subjective valuation of artistic and engineering activity, could produce determinate outcomes. It is because, once again, of their commitment to a background principle of equal concern or consideration, and their hostility to bargaining power with a morally uncertain provenance.

7.5.3.  Arbitrariness as Inconclusiveness In the light of these considerations, we should take ‘morally arbitrary’ to mean simply ‘morally inconclusive’. On this view, morally arbitrary characteristics, such as natural and social endowments, do no fundamental moral arbitration. The items of information encompassed by the veil of ignorance—information about these endowments, as well as conceptions of the good—are retired for the purpose of deliberation because they will simply obstruct the common aim of the deliberators to produce a consensus on the principles of justice. On the Irrelevance Interpretation, justice as fairness is not directly luck-­ neutralizing, and it does not insist on the full neutralization of morally arbitrary inequalities. Justice as fairness does not seek to neutralize or expunge the effects of a prior distribution of morally arbitrary endowments; rather, it merely denies their relevance to the choice of principles of justice. If it turns out that the principles of justice, when reached in the conditions just outlined, permit some degree of correlation between unequal endowments and benefits, we have been given no necessary cause for concern. The principles were not reached by privileging any of those endowments, and that was the point of the deliberative procedure. It was never the aim of the procedure to eradicate entirely inequalities in respect of them. As an alternative way of glossing these conclusions, notice that facts about individuals’ endowments and conceptions of the good do not yield any particular set of ground rules. As we already know, there is no entailment or even plausible case for any particular set of ground rules if we look at facts about different individuals’ endowments. Something else must be done in order to shed light on what principles of justice ought to obtain.

7.6.  Revisiting the Redundancy View So far, I have been describing the reasoning which, on Rawls’s view, motivates the design of the original position. But why, when all is said and done, bother actually going into the original position? It might be alleged that all the Irrelevance Interpretation manages to do, in effect, is to supply an alternative interpretation of the intuitive argument which underwrites the contractarian argument.

Revisiting the Redundancy View  229 Furthermore, why isn’t that enough? What value is added by the fact that these principles would be agreed to in the original position? In short, we still lack an explicit reply to the Redundancy View. I want to devote this section to a speculative line of thought concerning a potential lacuna in the argument which might serve, in turn, to allow some independent authority to migrate to the contractual procedure itself. To motivate the worry, it may be helpful to restate or recapitulate some of the argument as it has unfolded so far. I start with claims (1) to (4): (1) Morally arbitrary endowments can combine to produce an equal distribution only via a set of ground rules, which set the terms of economic interaction between agents. (2) Ground rules are not settled by facts about endowments, and are not any obvious function of facts about endowments. Since every individual has inherited her endowments in the same way, and since every individual is to be treated as an equal, we cannot appeal to the existence of those endowments in order to justify the decision to implement one set of ground rules rather than any other. By (1) and (2): (3) Social and natural endowments are revealed to be morally arbitrary; social and natural endowments do no fundamental moral arbitration in thinking about justice. And, by (2) and (3): (4) Individuals have no positive claim arising from their particular social and natural endowments that any particular set of ground rules be implemented. We reach (4), roughly speaking, at the conclusion of the intuitive argument, according to the Irrelevance Interpretation. But whether claims (1) to (4) have collectively established a conclusive case for roughly egalitarian ground rules is a moot point. Earlier in the discussion, I suggested that parties agree to resist unequal sources of bargaining power. The decision to allow bargaining power, construed as being proportionate to individuals’ differential holdings, to influence deliberations, would capitulate too much to distributive arrangements whose probity had not been established. But just as the parties cannot vouch for those extant arrangements, neither can they definitively rule them out. Of course, one might hold that it is simply a brute normative fact that unequal bargaining power is irrelevant to the search for the fundamental structures of social justice. But even if we were to concur with that claim, it might still be the case that we have reached a stand-­ off between the parties, rather than any decisive reason for adopting egalitarian ground rules (or indeed any ground rules at all). Thus (1) to (4) appear to license (5):

230  What is Arbitrary about Moral Arbitrariness? (5) No particular assignment of endowments can fix the appropriate set of ground rules, or show why the appropriate set of ground rules should be fixed in a certain way. This is where I think the idea of a contract—even a hypothetical contract—can earn its stripes. To avoid a stalemate, or stand-­off, it seems relevant that the parties would in fact agree to the principles of justice which they do agree to in the circumstances of the original position. In other words, the idea of a hypothetical contract might help to show how we can add, to (5), claims (6) and (7) in order to make further progress: (6) There should be no inequalities in bargaining advantage between the parties who are trying to establish a common set of ground rules. (7) Freed of inequalities in bargaining advantage, the parties would choose roughly egalitarian ground rules. Since, in the original position, the parties would agree to roughly egalitarian principles of justice, we have a way of building a bridge from (5) to (6) to (7). And that, in turn, permits the contractarian argument to be a non-­dispensable part of Rawls’s overall argument for justice as fairness. The actual fact that individuals would agree to egalitarian principles of justice in the original position does not simply summarize arguments which can be stated in non-­contractarian terms, but supplements those arguments in such a way that we can now explain how these egalitarian principles are reached, and why they get to carry normative authority. This reasoning answers both the Hypothetical Challenge and the Redundancy Challenge. In the light of these considerations, we can move, finally, from (7) to (8), which completes the argument: (8) The fact that the parties in the original position would choose roughly egalitarian principles just is the fact which explains why there are principles of justice which have that roughly egalitarian character.

7.7.  Cohen’s Challenge In this final section, I want to examine a potentially threatening argument to the Irrelevance Interpretation, and by extension the Redundancy Challenge, advanced by G. A. Cohen.67 As I hope to show, the considerations Cohen relies 67  See Cohen (2008), pp. 166–8. Cohen has mainly Samuel Scheffler’s views in mind here. (See the passage in Section 7.4.1.)

Cohen’s Challenge  231 upon ultimately fortify the credentials of the Irrelevance Interpretation and, in turn, the contract’s non-­redundancy.68

7.7.1.  From D1 to D2 and Back Cohen compares two views of how the Rawlsian hostility to moral arbitrariness contributes to his argument for justice as fairness. Imagine we start with an equal distribution, D1, which mutates into an unequal distribution, D2, as a result of the unregulated collection of free bilateral trades among different pairs of agents. According to the first view, which conforms roughly to the Irrelevance Interpretation, Rawls merely denies the relevance, or legitimating force, of morally arbitrary factors on distributions. This first view is within its rights to deny that D2 preserves justice, if we can assume that the transition from D1 to D2 is explained by morally arbitrary factors. So far, so Rawlsian. But, Cohen suggests, this view cannot advance any justification for starting off with D1 in the first place. It cannot offer any positive reasons in favour of D1. It can only offer reasons for refusing to accede to D2, if D1 is where we started from. For the denial that morally arbitrary factors play any legitimating role in the unequal distribution of D2 will not, by itself, sustain the claim that we had any reason for starting with the equal distribution of D1. The second of Cohen’s views conforms roughly to the Neutralization Interpretation. What is objectionable about D2, according to this second view, is its accommodation of morally arbitrary inequalities among the agents in question. That objection to the distributive influence of morally arbitrary factors also explains why it was right to choose the initial egalitarian distribution, D1, from which morally arbitrary inequalities had been purged. The objections to D2 double up as reasons for embarking from D1. The second view generates both reasons in favour of D1 and reasons against moving from D1 to D2. That constitutes a notable expansion of its normative credentials, compared to the first view. As Cohen puts it: My critic says that Rawls bases the starting point of equality on the proposition that (1) talent differences do not justify inequality, rather than on the proposition that (2) the arbitrariness of talent differences in some way justifies an initial equality—and (1) is consistent with the move to D2. I reply that (1) is indeed 68  See Lang (2014a), and, especially, Lang (2016) for detailed discussion of the ‘Freedom Objection’, according to which Rawls’s provision for the incentive-­seeking that is likely to install morally arbitrary inequalities is explained by his desire to maintain certain freedoms, including the freedom of occupational choice. According to the Freedom Objection, Rawls’s second principle of justice is not applied with the zealotry Cohen thinks it demands because it is held in check by Rawls’s first principle of justice. I think Cohen considerably underestimates the force of the Freedom Objection.

232  What is Arbitrary about Moral Arbitrariness? consistent with the move to an unequalizing D2, but that, by the same token, (1) does not motivate beginning with equality, that is, with D1. So either the case for the starting point of the . . . argument [which permits D2] is inadequate or the case for its starting point imperils its inference.69

Now if we were simply to appeal to (1), or the first view, then Cohen would be within his rights to point to a gap between the irrelevance of morally arbitrary factors and the justification for adopting any particular initial distribution (including D1). But we do not have to return to the luck-­neutralizing view in order to furnish a rationale for starting with D1.70 That justificatory gap can be plugged if we pay more attention to the contractarian apparatus which the moral arbitrariness argument is used to prepare us for.

7.7.2.  The Reappearance of the Contract In the absence of a positive justification for any particular distribution, we allow decisive significance to attach to the fact that parties under the veil of ignorance would choose Rawls’s principles of justice, and therefore D1, at least initially. And the parties in the original position would choose these principles of justice, rather than being simply stuck without any idea of what to settle for, because what they are left with, in the original position, is roughly equal bargaining power, as well as an awareness of the risk of indeterminacy, but with a determination to embrace concrete principles of justice and the remit to make their decisions of principles of justice stick even after the veil of ignorance has been lifted. We start with D1 because that is the distribution contracted among the contracting parties. D1 is what they would embark from because, in the absence of other considerations, it is what they would choose. D2 will be a justice-­ preserving extension, in turn, only if it is consistent with the principles that selected D1 in the first place. Thus, in particular, D2 will be acceptable only if it does not worsen the situation of the worst-­off. Hence there is no justificatory asymmetry or dissonance between D1 and D2. It is not the case that the contractarian argument rebuts D2, but fails to uphold D1. D1 is, in fact, the central application of the contractarian exercise. The principles thus chosen, which embrace D1, will also take a view as to the distributive adequacy of D2. If D2 diverges from the normative principles that

69  Cohen (2008), pp. 167–8; original emphases. 70  That may prove unwise in any case, given Hurley’s concerns: see Chapter 6, Section 6.3.1. But we can suspend this worry for the moment to see how the second view can respond to Cohen’s explanatory challenge.

Conclusion  233 explain the initial selection of D1, then D2 will not be deemed adequate. If D2 does not diverge from these principles, then D2 will be acceptable. Once again, the hypothetical contractarian procedure proves its explanatory worth. Without the contract, Rawls’s account is incomplete.

7.8. Conclusion That completes the exploration and defence of Rawls’s justice as fairness. But there are other issues which require attention, concerning the extension of his project to international justice and interspecies justice. These issues are variously concerned with the ‘who’ question for justice.71 The relationship between Rawls’s account of equal standing and the recent project which pursues ‘basic equality’ also needs attention, since this has a bearing on who qualifies for Rawlsian justice in the first place. These issues, which are all in one way or other triggered by the existence of arbitrary boundaries to moral standing or eligibility for moral treatment, will be dealt with in the next chapter.

71  See Chapter 6, Section 6.1.1.

8

Justice and Arbitrary Boundaries 8.1.  Extending Rawlsian Justice I argued, in Chapter  7, that Rawls’s justice as fairness should be regarded as a fundamentally contractarian exercise, just as Rawls’s original presentation of the argument suggests. The notion of moral arbitrariness should be interpreted, not simply as the harbinger to a commitment to eradicate the influence of luck, or in response to the absence of strong personal desert, but as a theoretical filter which is internal to that contractarian exercise. And the contract itself has an ineliminable importance: the principles of justice carry the authority they do because these are the principles selected by the parties in the original position. As a result, arbitrariness, or luck, enjoys only a highly qualified and specialized contribution to Rawlsian principles of justice. Most of us will see ourselves as free and equal citizens. Most of us are properly interested in the stability of a scheme of justice over time. And most of us think that there must be accommodation of the fact that our society is not mono-­ cultural or morally homogenous in any broad sense. These are Rawls’s starting points, and many if not most of us will share them. Though there needs to be substantial further exploration and defence of the theory he erects over those starting points, in order to be fully confident in the resilience of his system, it seems to me that the reputation that Rawls has acquired in the tradition of left-­ liberal theorizing about justice for obtuseness or clumsiness—however respectfully or carefully those charges are advanced—is largely undeserved. There are other worries in the air, however. These are broadly concerned with the ‘who’ question for justice (and for moral consideration more generally).1 One of Rawls’s last major contributions to political theory was The Law of Peoples.2 In this book, Rawls sketches a vision of international morality which has disappointed many of his supporters and well-­wishers. In particular, Rawls refuses to embrace a thoroughly cosmopolitan vision of global justice: he does not endorse the idea of a global original position which would obliterate the distinction between ‘insiders’ and ‘outsiders’ to a single worldwide scheme of justice. In Sections 8.2 and 8.3, I want to step away from the details of Rawls’s own position in The Law of Peoples,3 and to ask, more generally, what a Rawlsian 1  See Chapter 6, Section 6.2.1. 2  See Rawls (1999b). 3  For searching discussions of it, see the papers in Martin and Reidy (2007).

Strokes of Luck: A Study in Moral and Political Philosophy. Gerald Lang, Oxford University Press (2021). © Gerald Lang. DOI: 10.1093/oso/9780198868507.003.0009

Arbitrary Boundaries and the Incoherence Objection  235 theory should make of arbitrary moral boundaries. It is Rawls’s readiness to accommodate arbitrary national boundaries which, I believe, has disappointed or puzzled many commentators who are otherwise broadly sympathetic to his approach to justice within single societies. There are other worries about the ‘who’ question. What does it take for any human to qualify as a subject of justice? Why, and on what grounds, does Rawls think that qualification among human subjects is all-­or-­nothing? And what prevents non-­human animals from admission to the community of justice? Are humans, but not animals, singled out for special treatment simply because they belong to a certain species? These questions will be further investigated between Sections 8.4 and 8.6. The topics dealt with in this final chapter, then, are relatively heterogeneous. In one sense, and to a certain degree, they are spillover topics from Chapter 7: the Rawlsian system is invested, either explicitly or implicitly, in dealing with them in a certain way. But it should be obvious enough that the interest of these topics does not simply consist in seeing whether the Rawlsian system can be successfully filled out in these ways. These problems are interesting and vitally important in their own right, since they are centrally concerned with the moral significance, or otherwise, of our membership of various groups: in particular, national groups and species groups. These present us with among the most fundamental challenges in moral and political philosophy. These problems are also directly related to the more general concern with luck that has animated the whole book. It is plainly a matter of luck that we belong to certain groups and not others. This fact explains the understandable worry that ordinary common-­sense morality is committed to a collection of unacceptable patterns of discrimination between those individuals who, through luck, are situated inside these groups and those individuals who, again through luck, are placed outside them. The charges of discrimination arise, in the first instance, from the fact that membership of these groups is little more than a lucky or unlucky accident. This chapter will investigate whether these lucky facts about membership and non-­membership poison the ordinary moral standards which we commonly think arise from them. For the most part, I am going to maintain that they do not. The resilience of these ordinary moral standards can co-­exist easily enough with the fact that these membership facts are a product of luck.

8.2.  Arbitrary Boundaries and the Incoherence Objection 8.2.1.  The Incoherence Objection Rawls is often taken to task for having a theory of justice which displays hostility to morally arbitrary influences, but which also permits just shares to be affected

236  Justice and Arbitrary Boundaries by arbitrary facts about how particular schemes or communities of justice are bounded. Michael Blake states the worry like this: Allowing [state] boundaries to determine distributive shares seems to place an almost feudal notion of birthright privilege back into the heart of liberal theory.4

Other left-­liberal non-­cosmopolitans also display an awareness of this worry. Nagel, for example, explicitly admits that facts about national boundaries and populations are fixed ‘for all sorts of accidental and historical reasons’,5 from which it follows that ‘such [national] co-­membership is itself arbitrary, so an arbitrary distinction is responsible for the scope of the presumption against arbitrariness’.6 And yet that point, and the candour with which it is admitted, does not stop Nagel from adopting such a form of non-­cosmopolitanism: he thinks that arbitrariness of national boundaries does not constitute an insuperable barrier to the propriety of different schemes of justice, with different access to wealth and resources, each of them governed by the same anti-­arbitrariness principles of justice. This style of non-­cosmopolitanism does not deny that we have robust duties to outsiders. Nagel is committed to a ‘minimal humanitarian morality’, consisting of ‘basic duties of humanity’ that aim to ensure that outsiders are provided with the materials for a minimally decent life.7 Blake helpfully summarizes the relevant arrangements under the slogan ‘principles of sufficiency abroad and principles of distributive equality at home’.8 Even so, Nagel’s candour raises the question of why the avoidance of arbitrariness is so important when it comes to saying what justice requires, but relatively unimportant when it comes to determining who is going to count as a subject of justice for any particular scheme of justice governed by those same principles. The danger here is that the ‘how’ question for justice risks getting out of alignment with the ‘who’ question for justice. If we make the content of a theory of justice hostile to morally arbitrary factors, but allow the various schemes of just­ ice governed by that theory to be determined by arbitrary factors, then these Rawlsian non-­cosmopolitans risk exposure to the charge of incoherence.9 Call this the Incoherence Objection.

4  Blake (2001), p. 257. Blake thinks that the worry can be tackled: see Section 8.2.2. 5  Nagel (2005), p. 121. 6  Nagel (2005), p. 128. 7  Nagel (2005), pp. 131, 118. 8  Blake (2001), p. 258. 9 For earlier influential examples of dissatisfied cosmopolitan Rawlsians, see Beitz (1975) and Pogge (1989), Part Three. See also the various essays exploring these themes in Scheffler (2001).

Arbitrary Boundaries and the Incoherence Objection  237

8.2.2.  Justice for Insiders One form of response to the Incoherence Objection appeals to morally relevant properties of the relationships obtaining among insiders to a scheme of justice that do not obtain between insiders and outsiders. Some of these writers—Blake and Nagel, for example—place special emphasis on the coerciveness or non-­voluntariness of the relations which are enforced among insiders. For Blake, a central fact about the lives of insiders is that they are all confronted with the coercive power of the state, which intimately regulates their lives. This coerciveness is intrinsically problematic, because it poses a threat to our autonomy. To justify this coercive power, a state has to be such that it cannot be reasonably rejected by any of the individuals it governs; and, as a necessary condition for meeting that justificatory test, it must be the case that some are not worse off than others for morally arbitrary reasons.10 Nagel, by contrast, thinks that distributive arrangements among insiders must be such that we all can regard ourselves as co-­authoring laws and policies enacted by the liberal state. To satisfy that co-­authorship condition, there must in turn be morally non-­arbitrary ­reasons for the inequalities that obtain among insiders.11 Other writers, such as Sangiovanni, focus instead on the relations of reci­procity or collective provision which unite the citizens who are bound by those coercive structures. Sangiovanni holds that the coerciveness is incapable of doing any fundamental normative work, and the non-­voluntariness of the reciprocity-­based relations obtaining among insiders cannot fruitfully distinguish relationships which hold among insiders and the relationships which obtain between insiders and outsiders.12 What matters to him is the content of the relationships which insiders bear to one another: relationships of reciprocity, where we all contribute, where we can, and using the powers available to us, to the creation of goods to which we can then expect access on fair terms. For these non-­cosmopolitans, then, egalitarian pressures arise either as a result of the particular timbre of social relationships which obtain, or ought to obtain, among insiders, or else the coercive or non-­voluntary structure of these relationships, or perhaps both. These equality-­friendly considerations are absent when we consider the relationship between insiders and outsiders. Proponents of the Incoherence Objection may reply as follows.13 Even if there are morally relevant differences between the character of ‘insider-­insider’ relationships and the character of ‘insider-­outsider’ relationships, there are still further decisions to make about who gets to count as an insider and who is classified

10  Blake (2001), esp. pp. 266–85. 11  Nagel (2005), esp. pp. 128–30. 12  See Sangiovanni (2007), esp. pp. 17–19, and Sangiovanni (2012). See also Julius (2006), Cohen and Sabel (2006), and Valentini (2011) for more in this vein. 13  See, in particular, Arneson (2005), for a powerful articulation of this response.

238  Justice and Arbitrary Boundaries as an outsider. Perhaps justice demands that outsiders are reappointed as insiders, in order to allow them to be embedded in insider-­insider relationships, rather than insider-­outsider relationships. This reply need not quarrel with the significance of insider-­insider relationships to the case for egalitarianism. It can afford to admit, at least for the purposes of argument, that these relationships are relevant to the triggering of egalitarian standards, and that the elimination of morally arbitrary inequalities or involuntary relative disadvantages are particularly significant when those relationships are in place. But this admission does not tackle all the worries which congregate under the banner of moral arbitrariness. There can still be a proper refusal to allow justice schemes which are pledged to the elimination of morally arbitrary differences to have morally arbitrary boundaries. If the relationships obtaining among the insiders to a scheme of justice are important to justice, then it should also matter to justice where the boundaries of that scheme are drawn. It will be important that outsiders are not denied the opportunity to partake in those relationships, whatever they are, and whatever features they have, which trigger the relevant egalitarian distributive standards. To permit justice schemes to have such arbitrarily drawn boundaries is thus to maintain exposure to the Incoherence Objection. It might be protested that this picture assumes a non-­relational version of egalitarianism, rather than a relational form of egalitarianism.14 If the underlying theory of egalitarianism is relational, rather than non-­ relational, then some non-­cosmopolitans may say that there will be no pressure to admit a greater number of individuals into the insider-­insider relationships which trigger the more demanding justice standard. But this reply would be not quite correct. Even if the ultimate source of egalitarian demands is relational, rather than non-­ relational, there is surely a further question—and an open question—arising as to who gets to be embedded in those insider-­insider relationships. How is that question to be answered? Hurley’s ‘who’ question has not gone away. It is the existence and significance of that further question, rather than the dogmatic insistence that distributive justice needs to conform to some antecedent non-­relational pattern, that creates the difficulty for these brands of non-­cosmopolitanism.

8.3.  Towards A Rawlsian Non-­Cosmopolitanism The Non-­Applicability Claim, I believe, helps to defuse the Incoherence Objection. It does not license anything like a full argument for non-­cosmopolitanism, and 14  Both Sangiovanni and Nagel would, I think, be tempted by this response. See, for example, Sangiovanni (2007), pp. 36–7.

Towards A Rawlsian Non-­C osmopolitanism  239 that additional work will not be undertaken here. I adopt an official policy of agnosticism about the final outcome of the debate between cosmopolitans and non-­cosmopolitans. My argument does, however, defuse an important challenge to the viability of the non-­cosmopolitan project. To see how the Non-­Applicability Claim can tackle the Incoherence Objection, we need to investigate what the Incoherence Objection is really based upon. The Incoherence Objection begins with the observation that, according to non-­ cosmopolitans, insider-­insider relations are relevantly different from insider-­ outsider relations. In particular, insider-­ insider relations are thought to be governed by a more scrupulous attention to morally arbitrary differences among the insiders. These differences are then annulled, or neutralized, in justice-­ realizing insider-­insider relations. But the morally arbitrary differences between insiders and outsiders are treated less scrupulously. Those differences are not annulled, or neutralized. And what is surprising about all this is that, when all is said and done, the outsiders differ from the insiders in exactly the same ways that insiders differ from each other: from the point of view of individual insiders and individual outsiders, the differences are all morally arbitrary. Imagine that Catherine and Jim are the insiders to a justice scheme, and that Jules is the outsider. There are morally arbitrary differences between Catherine and Jim, which are neutralized in the implementation of correct distributive relations between them. The existence of these morally arbitrary differences is the source of the case for correcting them: in doing so, we will go from an unjust distribution to a just distribution. There are also morally arbitrary differences between Catherine and Jules, and between Jim and Jules, which are not neutralized, since Jules is an outsider. But how can this discrimination possibly be justified, given that Jules differs from Catherine and Jim in the very same ways—in only morally arbitrary ways—in which Catherine and Jim differ from each other? If morally arbitrary differences between Catherine and Jim demand correction, then how can the morally arbitrary differences between Catherine and Jim and Jules fail to demand correction? Complaints about the morally arbitrary nature of endowments facts are put to work for the relations between Catherine and Jim, but then ignored for the relations between Catherine and Jim and Jules. There can, on the cosmopolitan view, be no justification for this selective employment of the anti-­arbitrariness argument. If we have decisive reasons for being morally hostile to arbitrariness, then we must correct the resulting distributive blemishes wherever we find them. It should be reasonably clear that this line of reasoning assumes something like the Neutralization Interpretation of the anti-­arbitrariness considerations. On this view, the deep point of justice is to correct for the arbitrary advantages and disadvantage faced by subjects of justice. The impetus for justice is to level the playing field for those who are arbitrarily on the losing side of such inequalities, and

240  Justice and Arbitrary Boundaries pretty much the only qualification for being a subject of justice is that one is a human subject, or a moral personality, and is in a position to be arbitrarily advantaged or disadvantaged. In this respect, outsiders are no less qualified than in­siders. So, if Jim is arbitrarily disadvantaged relative to Catherine, and Jules is also arbitrarily disadvantaged relative to Catherine, then the correction of the arbitrary gap between Jim and Catherine, combined with the refusal to correct the arbitrary gap between Jules and Catherine, can only compound the amount of distributive arbitrariness in the system. It imposes, in effect, an arbitrary source of obstruction on the flow of the moral fuel that keeps the whole system running. This reasoning misidentifies, it seems to me, what is objectionable about being on the losing side of an arbitrary inequality. As the Non-­Applicability Claim demonstrates, Rawls does not believe that there is this sort of contamination at source. Our world may contain these morally arbitrary inequalities, but the purpose of justice—at least the immediate purpose of justice—is not to fix them. Morally arbitrary inequalities are not the nails which demand attention from the hammer of justice. This is because lottery facts are not the facts which, for Rawls, trigger the demand to install relations of justice in the first place. Rawls’s argument is different: for those agents who qualify on other grounds for inclusion in a scheme of justice, their relations in that scheme of justice should be governed by prin­ ciples of justice which would be chosen in the original position. At that point, morally arbitrary information is set aside, and the resulting principles will not privilege some endowments over others. But prior qualifications for membership in that scheme of justice may still have to be satisfied, and it is consistent with the role of anti-­arbitrariness in the content of principle of justice that these membership conditions may, in part, embed morally arbitrary factors. A theory of justice will not, of course, mandate these morally arbitrary factors, but neither need it exclude them from the very outset. It will all depend on what else is involved in appointing someone as a subject of justice in a particular scheme of justice. The reminder that it is the Irrelevance Interpretation rather than the Neutralization Interpretation which calls the shots in Rawlsian justice does not, as I have already said, settle the case for non-­cosmopolitanism. It merely tackles the Incoherence Objection, which will otherwise threaten to stop non-­ cosmopolitanism in its tracks. Rawls certainly thinks those who qualify for justice ought to be in a scheme of justice, governed by Rawlsian principles. But there is no immediate case for assuming that we all need to belong to the same scheme of justice.15 The non-­cosmopolitan position is not defenestrated by the Incoherence Objection, and the arguments between cosmopolitans and non-­cosmopolitans can unfold further.

15  A point made by Nagel (2005), p. 132.

Rawlsian Justice and Basic Equality  241

8.4.  Rawlsian Justice and Basic Equality Rawls’s brief remarks on the conditions required for inclusion in the original position have also raised eyebrows for other reasons. This section and the next one will, accordingly, be concerned with basic equality and interspecies justice. These issues are complicated and delicate, and the treatment of them offered here will have to be relatively brief.16 But it does give Rawlsians and others who are sympathetic to aspects of the Rawlsian programme a platform to build on.

8.4.1.  Qualifying for the Original Position What qualifies us to be subjects of justice? Rawls has comparatively little to say about this matter in A Theory of Justice, except for a brief section towards the end of it. He defines the ‘basis of equality’ as ‘the features of human beings in virtue of which they are to be treated in accordance with the principle of justice’.17 But what are those features? Rawls thinks that anyone who qualifies as a subject of justice must realize, or be capable of realizing to some degree or other, moral personality: she must be capable of forming a comprehensive conception of the good, and she must be capable of responding to the demands of justice. That specification of moral personality will not seem remarkable. These are the properties we would probably have expected to be associated with subjects of justice, governed by Rawls’s theory. But one problem with Rawls’s procedure lies in the fact that there are going to be scalar variations in moral personality. Take some property, ‘P’, which strikes us as a compelling candidate for the conferral of moral personality, but which can be realized to different degrees. If P is a broadly natural property, it will almost in­ev­ it­ably come in degrees.18 Assume also that Catherine realizes P to a higher degree than Jules. By assumption, the possession of P by Catherine and Jules makes both of them subjects of justice. But since Catherine instantiates P-­ness to a higher degree than Jules, we must now confront the question of whether Catherine enjoys a higher moral worth or standing than Jules. Why wouldn’t differences in P-­ness count, if P-­ness is the all-­important qualification for being a subject of just­ice in the first place? Rawls sees this problem. His way round it is sketched as follows:

16  For some earlier treatments, see Lang (2012) and Lang (2019b). I have drawn upon these earlier materials here and there in this chapter. 17  Rawls (1971), p. 504. 18  See Williams (1972c), Carter (2011), Arneson (1999), Cupit (2000), and Waldron (2008), (2017), for discussions of this problem.

242  Justice and Arbitrary Boundaries [I]t is not the case that founding equality on natural capacities is incompatible with an egalitarian view. All we have to do is to select a range property . . . and to give equal justice to those meetings its conditions. For example, the property of being in the interior of the unit circle is a range property of points in the plane. All points inside this circle have this property although their coordinates vary within a certain range. And they equally have this property, since no point in­ter­ ior to a circle is more or less interior to it than any other interior point.19

Range properties eliminate scalar variation among properties by describing them in a certain way. Both Catherine and Jules satisfy the second-­order range property of having P to some or other degree; both of them have the property of instantiating a value for P which falls within the possible range of P-­values. Range properties are therefore second-­order, all-­or-­nothing properties which testify to the positive presence of first-­order, scalar properties, paying no further attention to differences among those first-­ order properties. Both Catherine and Jules instantiate the range property of instantiating P to some degree, and thus they possess this second-­order property to an equal degree. So, although Catherine and Jules may instantiate to different degrees these first-­order elements of moral personality, the very fact that they instantiate them to any degree counts as equal satisfaction of the second-­order range property.

8.4.2.  Problems with the Range Property Solution Rawls’s range property solution has tended to produce dissatisfaction among his commentators.20 There are two sizeable problems with it. First, it may be suspected that there is something gimmicky about the focus upon the second-­order range property, rather than on the instantiation of the first-­order properties which also constitutes satisfaction of the second-­order property. An individual satisfies the second-­order property because she satisfies the first-­order property. If it is the realization of a certain first-­order property which qualifies an individual as having moral personality in the first place, and if realization of that property varies by degree, then why isn’t it basically cheating to level the playing field by pretending that each individual who instantiates the first-­order property to any degree has instantiated it to a relevantly equal degree? If we want to argue honestly for the egalitarian intuition, we need a more prin­ cipled reason for not deferring to the lower-­order scalar property, rather than 19  Rawls (1971), p. 508; emphasis added. A range property solution is also offered by Waldron (2017). See Lang (2019b) for a detailed discussion of Waldron’s position. 20 See, for example, Carter (2011), pp. 549–50; Lippert-­Rasmussen (2015), pp. 42–3; Nathan (2011), pp. 215–17; Cupit (2000), p. 110; and Arneson (1999), pp. 108–9. Responding to certain features of Singer’s work, Arneson calls the scalar problem the ‘Singer Problem’.

Rawlsian Justice and Basic Equality  243 settling for appeal to the higher-­order range property which conveniently flattens the lower-­order differences. Or so it may seem. Second, the range property solution generates further worries about the arbitrariness of the lowest point of satisfaction for the achievement of moral personality. Imagine a creature narrowly missing out on the Rawlsian range property test: by a small margin, this creature lacks the type of mental complexity and the particular capacities which would qualify it as a moral personality, and therefore a subject of justice. If we are prepared to tolerate quite large degrees of difference among those who qualify as subjects of justice by meeting the threshold, it will be more difficult to insist that there is nonetheless a sharp cut-­off point on the spectrum of mental characteristics, such that every individual who is located on the spectrum above this cut-­off point qualifies, and qualifies equally, for justice, but every individual who falls below this cut-­off point is wholly disqualified from just­ice. The radical discontinuity of moral status between those who are equally qualified and those who are wholly disqualified seems to be at variance from the underlying scalar facts about the psychological composition of moral personality. The sharp all-­or-­nothing cut-­off point seems jarring, and ad hoc.

8.4.3.  Basic Equality These objections to Rawls’s range property solution appear to reflect an attachment to a form of egalitarianism which has become known as ‘basic equality’.21 Basic equality denotes a deeper sort of equality which qualifies those who satisfy it for inclusion under ‘surface-­level’ principles of egalitarian distributive justice.22 Theorists who are committed to the basic equality project will tend to think that, in the absence of basic equality, our entitlement to egalitarian justice risks being unfounded. There are at least two important lines of thought in play here. The first line of thought runs a quality check on the ‘who’ question. The appointment of an individual as a subject of justice serves both to distribute bene­ fits and burdens; it amounts to the conferral of a significant sort of social or pol­it­ ical status on her. It is therefore important to know which individuals qualify as subjects of justice. It would be unfair to exclude some individuals from the constituency of justice if they possessed properties that made them relevantly similar to other individuals who are not excluded from the constituency of just­ice. It might also be unwise to include individuals who lack these properties, as Ian Carter explains: 21  See, in particular, Waldron (2008), (2017), and Carter (2011). Other discussions of these concerns are offered by Cupit (2000), Lippert-­Rasmussen (2015), ch. 3, and Nathan (2011). The worries which lie behind the development of basic equality, however, were well grasped by Williams (1972c), in an essay that was originally published in 1962. 22  The term ‘surface-­level’ is taken from Waldron (2017), p. 10.

244  Justice and Arbitrary Boundaries [S]uch an equality of entitlements must be viewed as appropriate in the light of certain characteristics of the bearers of those entitlements. If unequals ought nevertheless to be accorded equal entitlements, why not accord equal entitlements indiscriminately to humans and cats and oysters?23

Oysters surely fail to qualify as subjects of justice. So we must avoid this sort of over-­generation of subjects of justice. The justice constituency must not be flooded by creatures who plainly do not belong there. The second line of thought establishes a connection between the ‘who’ question and the ‘how’ question. If we are to return an egalitarian answer to the ‘how’ question, then our answer is naturally directed to the properties of the individuals over whom egalitarian principles are supposed to apply. The morally significant distribution of benefits which egalitarian theories of distribution aim to secure must be partnered by the thought that it is appropriate to treat subjects of justice in an egalitarian way, and thus inappropriate to fail to treat them in this way. But why would it be appropriate to treat these individuals as equals unless there is a sense in which they just are equals? It will be appropriate to treat them in this way if they have properties which are such that, in virtue of being equal, treating them unequally would constitute a slight to them. Correlatively, if an individual who is allegedly included in the scope of the ‘who’ question fails to have those properties, then it is difficult to see what complaint might be presented on her behalf for not being treated with egalitarian concern. These thoughts help to make sense of Carter’s determination to achieve greater clarity about the basis of equality: Why ought people to be treated as equals? Is there something about people that makes them equals, such that it is appropriate to accord them equal concern and respect?24

Carter takes these two questions to license a third question: Is there a property that they possess to an equal degree such that they can be reasonably described as equals?25

Carter’s background reasoning appears to come to something like this: if we wish to support the judgement that we should treat individuals as equals, then we should be prepared to explain our grounds for this judgement. An approximate explanation is that there is something about these individuals which explains why they should be treated as equals, or why they should command equal amounts of 23  Carter (2011), p. 541. 25  Carter (2011), p. 539.

24  Carter (2011), pp. 538–9.

Rawlsian Justice and Basic Equality  245 the favoured currency of justice.26 But if there is something—some particular property—possessed by these individuals that can explain why they have a claim to be treated as equals, then it must plausibly be contained by each of them to the same degree. Think of the property, whatever it is, as a kind of normative magnet. This property generates an entitlement to a certain share of the total stock of goods as prescribed by the currency of justice. Call the property ‘P,’ and call the total stock of justice-­relevant goods ‘G.’ Any individual who possesses the property of P-­ness has a justice-­based claim to some appropriate share of G. We might then expect this property, if it possesses the sort of magnetic power to qualify its possessor as a justice claimant in the first place, to be such that it is possessed to an equal degree by each claimant. If P is possessed to an unequal degree among claimants of justice, then there must, in turn, be a presumptive case for insisting that just distributions take account of the unequal distribution of P-­ness among those subjects of justice. Those who have more of the underlying entitlement-­generating property, or who instantiate P to a greater degree, should then get more, and those who have less of the underlying entitlement-­generating property, or who instantiate P to a lesser degree, should get less. At the very least, the question is left open whether those who instantiate P to a lesser degree genuinely fall within the scope of egalitarian concern. To put egalitarian concerns on more robust foundations, it would be preferable to find a value for P which was such that every justice claimant possesses it equally. That lesson is forced upon us if we admit that P-­ness is the property which qualifies an individual to be a justice claimant in the first place. If P-­ness makes you a justice claimant, then it generates your entitlement to a certain proportion of G. So, for the ‘how’ question to be awarded a reliably egalitarian character, it needs to be the case that egalitarian subjects of justice all instantiate P to the same degree. The metaphor of normative magnets is well placed to capture this feature of Carter’s reasoning. Magnets differ in their magnetic powers. So, if possession of P-­ness establishes individuals as justice claimants, then differences in P-­ness among these individuals should incline us to embrace, as an answer to the ‘how’ question, a final pattern of distribution which is sensitive to those differences in P-­ness. It follows, then, that if Catherine and Jules are to be treated as subjects of egalitarian justice, they must possess P to the same degree.

26  Here we should be prepared to interpret the currency question broadly. In addition to the usual candidates for the currency of justice such as resources, welfare, capabilities, and hybrid currencies such as Cohen’s ‘access for advantage’ (Cohen 1989), we should be prepared to cast the net even more widely than this, and at least entertain other candidates for the currency of justice, such as political influence or moral authority: see Carter (2011), pp. 539–­40.

246  Justice and Arbitrary Boundaries

8.4.4.  The Resilience of Range Properties Imagine, then, that Catherine’s realization of moral personality scores more highly on the relevant indices than Jules’s. Both Catherine and Jules fall within the Rawlsian range, and thereby fall equally within that range, since the question of whether they fall anywhere within the range can be described as whether they have satisfied this binary higher-­order property. But the underlying properties which gain them admittance to the range are different, and Catherine qualifies as a more fully realized moral personality than Jules. So there is, supposedly, a problem: intellectual honesty requires that we take this difference into consideration, or at least that we cannot discount the possibility that this difference is morally relevant. Differently described, the challenge must then be that being worse off in respect of the properties comprising moral personality risks prescribing an in­fer­ ior grade of concern or moral standing. One conspicuous problem with this reasoning is that these implications are roundly condemned by the substantive content of Rawls’s justice as fairness, as Rawls himself points out further down the passage in which he initially outlines the idea of the range property: [T]he description of parties in the original position identifies such a [range] property, and the principles of justice assure us that any variations in ability within the range are to be regarded as any other natural asset.27

If Catherine and Jules qualify as subjects of justice, then the differences among their natural abilities are immediately neutralized by justice as fairness: these differences will be deemed morally arbitrary, and they should be in no position to exercise particular influence on the selection of the principles of justice within the original position. Rawls’s point encourages us, I believe, to draw a more general lesson. He did not draw this lesson himself, but it strikes me as a sensible extrapolation from what he does say. One of the aims of Rawls’s justice as fairness is to neutralize the distributive effects of morally arbitrary differences in talent or endowment. For Rawls, these differences among endowments do not challenge the very possibility of his theory of justice, but simply constitute the raw descriptive data to which his theory is to be applied. The mere fact that these individuals are unequal in the first place, prior to the application of the theory of justice as fairness, does not invalidate the very possibility of that theory. How could it? If everyone was already equal, there would be nothing for an egalitarian theory of justice to do. And if the mere existence of inequalities among talents or endowments was

27  Rawls (1971), p. 508.

Rawlsian Justice and Basic Equality  247 supposed to endanger these individuals’ qualification as subjects of egalitarian just­ice, we would be making demands on the scope question for justice which were at odds with what must happen in the retail operation of justice. This is why Rawls does not have to worry about differences in degrees of moral personality among human beings. He can rely on the unremarkable and deceptively casual claim that justice applies to anyone who falls within the conceptual range of its operation.28 The deeper lesson to be recovered from Rawls’s relative insouciance on this score is that it is wrong-­headed to apply egalitarian justice to subjects only if these subjects are already equals in some descriptive aspect. This underlying ‘equality-­ out, equality-­in’ assumption seems confused.29 The point of egalitarian theories of justice is to arrive at egalitarian distributions. They do not have to proceed from an egalitarian distribution of something or other which qualifies the individuals in question as subjects of justice. If an egalitarian pattern of concern or respect is what we are aiming at, it does not follow that we have to embark from some form of descriptive equality, involving instantiation to the same degree of the same descriptive properties. The point of Rawls’s theory, along with other theories from the same left-­liberal tradition, is to treat with equal concern individuals who, in the absence of the application of the theories, would be treated, in effect, as unequal. The natural constituency for such theories of justice, then, is a set of individuals who exhibit different degrees of fortune and misfortune; for the theory to have its central and typical applications, the individuals over whom it ranges will differ in levels of fortune. The aim of these theories is precisely to nullify these relative degrees of misfortune. But the worries and ambitions expressed by theorists of basic equality risk an insistence on appointing, as subjects of egalitarian justice, a set of individuals who, at some deep level, do not differ in levels of fortune. Again, this is puzzling. Since the equality-­out, equality-­in assumption is not forced upon us, and since it generates this counterintuitive result, it should be rejected, however en­ticing it may seem at first. Rawls does the essential groundwork, I believe, in demonstrating that the basic equality project is wrongheaded. Contrary to what the basic equality project suggests, we need not insist that every subject of justice, egalitarian or otherwise, needs to satisfy a descriptive egalitarian profile before the theory of justice is even applied. Human subjects of justice can come in all shapes and sizes. This is just as well; justice is surely meant to assist those whose circumstances, both internal

28  There are still serious charges of exclusion for him to face, however: see Nussbaum (2007). But Rawls is not exposed to the scalar problem in particular, which is our concern here. 29  This phrase echoes the ‘obligation-­out, obligation-­in’ assumption which Williams ascribes to Kantian morality: see Williams (1985), pp. 174–81.

248  Justice and Arbitrary Boundaries and external, place them at a pre-­justice disadvantage relative to those whose personal circumstances are more fortunate.30

8.5.  Justice and Species Barriers Even if we feel that we have made decent progress on the problem of basic equality, we are still going to be left with urgent questions about who gets to qualify as a subject of justice. Is justice a value whose remit omits no human, but includes no animal? Why should that be? Won’t the ‘who’ question now turn upon arbitrary discrimination between species? The worry here is that the generous inclusion of humans who are down on their luck will be counterbalanced, less impressively, by the exclusion of creatures whose only substantive shortcoming is to belong to a different species. In this section, we briefly turn to the question of species-­based discrimination, or ‘speciesism’.

8.5.1.  Are Humans Morally Equal? Whatever our moral commitments, it is difficult not to evince a commitment to human moral equality. Morally speaking, every human counts, and counts equally. That is a secure datum of moral common-­sense thought. But its deep entrenchment may reflect, in part, the fact that it is not too obvious what we are signing up for when we invoke it. Humans come in all sorts of shapes and sizes: they differ in every conceivable way, as Williams memorably puts it, ‘from weight lifting to the calculus’.31 They differ in respect of morally significant properties—in capacities and skills, the wherewithal for the realization of well-­being, and in terms of moral performance. Furthermore, some human beings are so-­called ‘marginal’ cases: being very young, or very old, or cognitively disabled, or comatose, they may lack the occurrent capacities of normal Lockean personhood. Yet we tend to think in ordinary moral thought that these humans enjoy a common moral fellowship with the rest of us. We are all supposed to be morally equal, though our circumstances may be wildly different. In fact, these marginal persons might seem to constitute a particularly robust sub-­set of humans for qualification to equal moral standing: since

30  Again, it is not the direct aim of Rawls’s theory, as I understand it, to take individuals who differ in fortune and then to equalize their fortune on the grounds that they had no control over these different starting points. That is the interpretation embodied in the Neutralization Interpretation, rather than the Irrelevance Interpretation. My main point here is that it is no barrier to the implementation of justice of fairness under the Irrelevance Interpretation that the individuals to whom it is applied in any given scheme of justice differ in these antecedent ways. 31  Williams (1972c), p. 234.

Justice and Species Barriers  249 they are already vulnerable, or disadvantaged relative to the rest of us, it is particularly important that they not suffer a moral demotion. These are uncontroversial deliverances of moral common sense, which we can expect moral theory to explain and justify. The task here is more complicated than moral common sense acknowledges. What is human moral equality based on? Can it be based on the mere fact that we are all human? The claim that we are all human is, for all practical purposes, an all-­or-­nothing claim, and it seems to flatten the field immediately among all humans. That may seem to be an advantage. But that claim is also open to a number of immediate and taxing challenges. For one thing, to say that a human is a human seems to be just a trivial claim; it is a trivial platitude incapable of any genuine moral illumination. The property of being human does not explain why humans are such as to qualify for the status associated with basic equality. Such a specification will risk being morally unsatisfactory for a further reason: it involves a commitment to an invidious form of discrimination. It is a familiar claim in applied ethics that species divisions are not, in and by themselves, morally significant. To suppose that a mere difference in species gives rise to a substantive difference in moral standing is to be committed, the argument often runs, to a claim that is not unlike the noxious commitments underlying racism and sexism.32 It is worth paying further attention to the analogies between speciesism, ra­cism, and sexism proposed in these writings. What makes racism and sexism paradigmatically invidious forms of discrimination, and why is speciesism analogous to them? This is what Singer has to say: From the mere fact that a person is black, or a woman, we cannot infer anything else about that person. This . . . is what is wrong with racism and sexism.33 To give less consideration to a specified amount of pain because that pain was experienced by a black would be to make an arbitrary distinction. Why pick on race? Why not [pick] on whether one person was born in a leap year? Or whether there is more than one vowel in his surname? All these characteristics are equally irrelevant to the undesirability of pain from the universal point of view.34

These passages suggest that race and sex are morally arbitrary because they are irrelevant, and they are irrelevant because they can tell us nothing about what 32 These ideas are conveyed with considerable force in Singer (1974), and in the other essays reprinted in Parts 1 and 2 in Singer (2001), as well as Singer (1993), esp. ch. 3. Such arguments have proven to be very influential in contemporary ethics. The critical argument in Williams (2006) is also discussed and defended at length by Lang (2012), from which the present discussion draws. See also Kagan (2016) for a highly interesting discussion of the force of the speciesism charge. 33  Singer (1974), p. 107. 34  Singer (1978), p. 198.

250  Justice and Arbitrary Boundaries morally significant properties that individual possesses. An individual’s skin ­colour and sex, in and by themselves, will provide no instruction about that individual’s capacity to experience pain, or her other morally significant properties and traits. Race and sex are opaque properties; they are morally unrevealing, or uninformative. If we use them as principles of discrimination, we will be misled into overlooking these individuals’ genuinely relevant properties, which will remain concealed. Race and sex are thus a false guide to what matters. The same verdict must hold for the appeal to species membership.35

8.5.2.  Moral Individualism and Misfortune Once the spectre of speciesism is raised, the basic equality project may once again seem more enticing. This project attempts, for understandable reasons, to bypass any appeal to human chauvinism or speciesism in order to arrive at more hygienic and transparent foundations for human moral equality.36 Now I have already expressed some scepticism about this project. But how can speciesism be avoided if we are not going to pursue basic equality, or something like it? Don’t we need a more principled reply to the worries about speciesism, and won’t basic equality fit the bill? Perhaps we should reconsider the basic equality project, on the grounds that it permits us to escape the orbit of speciesism, if for no other reason. My view is that, while we do need to say more—a moral appeal to the moral importance of humanity is hardly self-­ratifying—we do not need to reinvest in the basic equality project. Instead we might question the reputation which spe­ cies­ism has acquired for being a moral bête noire. To see why, return to Singer’s diagnosis of the faults of speciesism. His diagnosis is dependent on a distinction between morally relevant and morally irrelevant properties. What are the morally relevant properties? For Singer, these properties will be non-­relational, welfare-­realizing properties of the individual. Singer’s understanding of what these properties must be reflects his attachment to the doctrine of ‘moral individualism’.37 No creature gets to be morally significant simply because of the association or kinship it enjoys with others; at the most fundamental level, relational properties do not matter. To settle questions of moral 35  See Lang (2012), pp. 297–301, for a lengthier discussion. 36  See Waldron (2017), pp. 87, 229–30, Arneson (1999), p. 103, Carter (2011), p. 541, Cupit (2000), p. 109, Lippert-­Rasmussen (2015), pp. 39–40, and Nathan (2011), p. 223. The treatment of why and to what extent humanity is morally significant offered by Waldron (2017), Lecture 6, is particularly detailed and interesting. It is discussed in Lang (2019b), pp. 257–9. 37  In addition to the works of Singer already cited, see, for example, Rachels (1990), Norcross (2004), and McMahan (2005). The term ‘moral individualism’ derives, I believe, from Rachels. (The other, welfarist input into Singer’s views about morally relevant properties reflects his commitment to utilitarianism. But it is the individualism which will attract non-­utilitarians and non-­welfarists to Singer’s cause.)

Justice and Species Barriers  251 standing, we compile an inventory of an individual’s internal properties.38 Moral individualism encourages the thought that, at bottom, moral standing is a meritocratic affair. You do not get admitted to the moral constituency simply by as­so­ci­ ation, or by virtue of group membership, or because other individuals have special ties to you. Moral individualism strikes me as seriously incomplete.39 It encourages us to appraise individuals for what they have, not for what they lack. But it can often be a morally important fact about individuals that there is something they lack, or something that cannot be included in an inventory of their internal properties. Imagine that an adult human has a vocabulary of only 300 words. His vocabulary is matched by a particularly intelligent gorilla. This vocabulary size offers some evidence for cognitive parity between the human and the gorilla. But, of course, we will also want to draw the further conclusion that the adult counts as unfortunate by human standards, while the gorilla count as fortunate by gorilla standards.40 Our treatment of the adult human may then reflect the fact that, unlike the gorilla, he is unfortunate, or vulnerable, or down on his luck, or disadvantaged by the standards of his community. And different standards will determine whether he can be reliably described in any of these ways. To be able to make these further determinations, we will need to know about the community to which he belongs, and also—most often—the species to which he belongs. Standards of fortune and misfortune are relative to standards of flourishing, and those standards will be largely community-­relative. This is what is wrong with the anti-­speciesism argument. Species membership is not morally irrelevant if it has something to teach us about which standards of fortune and misfortune apply to any given individual. These standards of fortune and misfortune do not write themselves. They cannot be settled by the properties actually possessed by any given individual. These standards of fortune and misfortune are also, in part, relational characteristics. They draw upon the standards for flourishing, fortune, and misfortune that can be ascribed to individuals in a

38  Advancing a watertight distinction between the ‘intrinsic’ and the ‘extrinsic’ or relational can prove to be a difficult task, if not a fool’s errand: to see what might be involved, see Langton and Lewis (1998); further additions to that literature are provided by Marshall and Parsons (2001) and Langton and Lewis (2001). I will not appeal to metaphysical scruples like these. Perhaps individualists can get by on more rough-­and-­ready criteria for ‘intrinsic’. I aim to discredit individualism in another way. 39  I say ‘incomplete’, because intrinsic properties can still of course function as a barrier to certain forms of treatment. If a creature can suffer pain, then individualism correctly implies that this creature is morally significant in virtue of its ability to suffer pain. The extrinsic or relational properties which I think are morally relevant augment, but do not supplant, the moral significance of a creature’s intrinsic properties. 40  Cf. Anderson (2004), pp. 281–2, and Lang (2012), p. 307. Again, the asymmetry between the human and the gorilla does not show that the gorilla is morally insignificant. I would reject any such implication. The point is only the difference in species are correlated with morally significant differences.

252  Justice and Arbitrary Boundaries moral community of a certain type. We need to be familiar with these standards in order to know what we owe to others.41 Species membership properties are not, unlike the categories of race and sex within the human community, morally opaque. Species membership is morally important information about an individual, because it allows us to locate which standards of fortune and misfortune properly apply to that individual. Speciesism is thus not structurally akin to racism and sexism. In summary, species membership is not, in the relevant sense, arbitrary. Of course, we did nothing to bring about that we belong to the human species. We do not deserve to be human. But it will not be surprising by now that this claim establishes hardly anything. We are human, and we belong to the human moral community. Although that fact is little more than a fait accompli, it is still one which immediately mobilizes a complex set of moral and political demands and expectations.

8.6.  Moral Equality and Justice It will be fairly obvious that we have by now left behind the relatively narrow focus of distributive justice. The focus in the last section has been on the broader issue of human moral equality. This is a broader issue than justice, because it is getting at a broader sort of moral equality which will be manifested in a broader range of contexts than simply distributive justice. But we should briefly revisit the topic of distributive justice, to reflect our concern with the original Rawlsian starting point. Rawls does not think that every human, regardless of her condition or circumstances, is eligible for inclusion in the original position. This is because not every human has a moral personality in the Rawlsian sense: many humans will not be capable of forming a comprehensive conception of the good, or of acting from a sense of justice. Rawls thinks their interests can and should be taken into account by the deliberators in the original position.42 It may be complained, in response, that this is still a troubling form of exclusion. The mere fact that they do not themselves appear in the original position may suggest that their interests are of secondary importance, or that they need, patronisingly, to be managed, or that they are simply something of an afterthought.43

41  Some accommodation of this point is offered by McMahan (1996), though his stance on these issues appears to have somewhat hardened in his later work: compare McMahan (2005). See Berkey (2017) for other ideas. 42  See, for example, the remarks at Rawls (1971), pp. 128–9, where parties to the original position are conceived of as ‘heads of families’, with all the further responsibilities implied by that title. 43  See Nussbaum (2007), esp. ch. 2.

Conclusion  253 This line of argument leaves Rawls free to maintain the following disjunctive line. On the one hand, if he has overlooked aspects of moral personality within the human community, his system can be easily adjusted to include these human individuals. It is natural to think that his system will be more accommodating of human vulnerability than it is often thought. Rawls, after all, is perfectly aware of our common vulnerability, and his system involves, at the most fundamental level, an articulation of what moral equality involves.44 On the other hand, even if there are certain humans whose cognitive formation makes it difficult to regard them as subjects of justice, he can still say that there are pressing moral duties on us, regardless of whether these duties belong strictly to the duties of justice. Justice is not the only value to which we need to pay attention. These moral duties will also reflect the humanity of the moral patients; they will have a different content, and may enjoy a higher degree of urgency, than our duties to other, non-­ human moral patients. It seems to me that this reply largely protects Rawls from such criticisms.

8.7. Conclusion My main business in Chapters 7 and 8 has been to re-­situate Rawls in the trad­ ition of luck-­sensitive theorizing about justice. As we have seen, Rawls is misleadingly classified as a luck egalitarian avant la lettre. His approach to justice does not reflect luck egalitarian concerns. Luck egalitarians’ concern with luck is vis­ ible, in the first instance, in pairwise inequalities among individuals that reflect involuntary disadvantage. This is not the starting point for the Rawlsian approach. The Rawlsian system starts from a commitment to political equality, and an awareness that political equality can only be achieved by starting with a broad structural view of how our socio-­economic institutions interact, and which distributive patterns they form. Rawls then advances an interpretation of the distributive consequences of equality by employing a contractarian apparatus. The luck-­neutralizing parts of the theory are internal to that contractarian apparatus. I have argued that this is a defensible approach. But it has hardly anything to do with a luck egalitarian approach. We have other things to learn from Rawls. Though the account he argues for in The Law of Peoples has not received an extended defence here, it is at least not exposed to the Incoherence Objection. His relatively relaxed conditions for qualifying as a subject of justice are also defensible. We do not need to embrace the basic equality project, and we need not worry that the humanity of those we wish

44  Cf. Thomas (2011), p. 140.

254  Justice and Arbitrary Boundaries to qualify as moral equals is playing too prominent a role. No objectionable chauvinism is involved in a commitment to the moral significance of humanity. Even if, in the contemporary Western world, the political influence of Rawlsian liberal thinking appears to be on the wane, it seems to me that the Rawlsian system displays an impressive solidity against philosophical objections to his deeper commitments.45 There is no reason to think that, if and when the winds of pol­it­ ical fortune change, the central components of the Rawlsian system will not be in a good position to reap the benefits. 45  For a recent comprehensive and spirited attack on Rawlsian philosophy, see Forrester (2019). My doubts about Forrester’s treatment are aired in Lang (2020).

APPENDIX I

An Empirical Challenge to Resultant Luck 1. Introduction In Chapter  1, Section  1.5, I tackled a wave of ‘accommodation strategies’, which seek to maintain anti-­luckist commitments at a deep level while explaining away the anti-­anti-­ luckist surface of some of our discourse. Some writers have enlarged the resources at the disposal of the accommodation strategies by drawing on the literature on empirical psychological biases to challenge accounts which permit resultant luck to enjoy moral relevance.1 This sort of work has already drawn significant critical fire,2 but I want to say something more about one particular prominent argument, by Darren Domsky. I will suggest that Domsky’s appeal to empirical biases cannot discredit anti-­anti-­luckism in general, or the Restricted Account in particular.

2.  Domsky’s Debunking Explanation Domsky’s analysis of the problem of resultant luck proceeds via two conflicting intuitions which, so he suggests, govern our attitudes to luck:3 Intuition 1: Moral blameworthiness is independent of luck. Equally negligent agents are equally blameworthy, even if by luck one brings about bad outcomes and one does not. Intuition 2: Moral blameworthiness is not independent of luck. Negligent agents who by luck bring about bad outcomes are more blameworthy than equally negligent agents who by luck do not.4 Intuition 1 denies a role for luck in the allocation of blameworthiness, while Intuition 2 affirms such a role. Intuition 1 is endorsed by the Anti-­Luck Account, and Intuition 2 is endorsed by the Restricted Account. The two intuitions, as they stand, are plainly inconsistent, and rationality demands that one of them be abandoned. Domsky’s strategy is to provide a debunking explanation for Intuition 2, thereby discrediting it. This will leave Intuition 1 without any relevant competition. Domsky’s overall aim, then, is to convince us

1 See, in particular, Domsky (2004), esp. pp. 453–64, and Royzman and Kumar (2004). For in­struct­ive criticism of Royzman and Kumar (2004), see Enoch and Guttel (2010). 2  See Statman (2005) and Enoch and Guttel (2010). 3  Domsky (2004), p. 445. 4  Negligence seems to be a relatively narrow category for Domsky to invoke here, as is noted by Statman (2005), p. 425: what about malicious attempts, of the sort manifested by Assassins? But it seems to me that we can safely expand Domsky’s definitions to encompass other morally problematic mental states, such as malicious attempts.

256  Appendix I that we should revise our ordinary moral thinking so that it conforms more thoroughly to Intuition 1. The heart of Domsky’s debunking strategy lies in his suggestion that Intuition 2 is the product of two separate biases: a selfish bias and an optimistic bias. These biases exert a persistent but subconscious influence on us. Even if we think that we are immune to them, there is the distinct danger, if left to our own devices, that they will dominate our thinking. If we are to eliminate their influence, we must cut the Gordian knot and renounce them in a more comprehensive and self-­conscious way. As Domsky sees it, no measures falling short of the wholesale rejection of Intuition 2 will suffice for this exercise in moral decontamination. The selfish bias, in this context, is the tendency to adopt moral beliefs which promote our self-­interest, while the optimistic bias is the tendency to assume that we are less vulnerable to bad luck than the average person. When these biases combine, Intuition 2 is favoured, because other people will be blamed more than we are. The optimistic bias is explained by the fact that we implicitly or subconsciously impute to ourselves powers of agency that render us invulnerable to the sort of luck plaguing those—that is, just about everyone else—whose powers of agency are less extensive. The optimistic bias is an instance of our general tendency to overrate ourselves; to believe of ourselves, relative to other people, that we are better or more talented than them, or simply blessed by better fortune than they are. Since, if I am in the grip of this bias, it will seem to me that my luck is better than that of others, then support for Intuition 2 will entail my belief that I will be less blameworthy than others across my practical career. It should be noted—and this will turn out to matter—that Domsky’s characterization of the content of the optimistic bias varies a little. In one formulation of it, the optimistic bias is described as ‘an underlying, tacit belief that we have a special mastery over luck’.5 On this view, luck is depicted as a determinate sort of force that some of us manage better than others. Call this the Pure Luck Interpretation. The Pure Luck Interpretation may appear to involve a needless form of reification. Napoleon is reputed to have said—though this story may be apocryphal—that he wanted lucky generals, rather than good generals. That hope cannot have been intelligently accompanied by the expectation that some of his generals had the disposition to master luck, as opposed to having greater powers of control that made them less vulnerable to it. However, the bias is meant to be unconscious, so perhaps its content really is described by the thought that, when luck comes my way, it is more likely to be good luck than bad luck. The content of this bias can hardly expect to survive reflective scrutiny, but that does not make the bias unintelligible. Each person may indeed unconsciously think of herself as lucky; she may unconsciously assume that while she is no less prone to luck than anyone else, she is somehow much more likely to be the beneficiary of good luck than the victim of bad luck whenever luck comes along. Domsky also characterizes the optimistic bias as the belief that we have the sort of character or qualities or skills which explain why we are less vulnerable to bad luck. On this view—call it the Control Interpretation—my relatively low vulnerability to bad luck is explained by the presence in me of more extensive powers of control, or a superior collection of skills, talents, and character traits, relative to others.6 According to the Control Interpretation, if Jules enjoys more extensive powers of agency than Jim, then Jules

5  Domsky (2004), p. 458. 6  Domsky tells us that ‘it is not just that we are especially confident in our abilities. We are also confident that we are somehow immune to bad luck’ (2004, p. 458). That looks like a version of the Pure Luck Interpretation. But then we are told, in further explication of this claim, that ‘[w]e subconsciously believe that we somehow have either the skill to prevent bad luck or the foresight to avoid it.

Appendix I  257 has, while Jim lacks, the ability to control certain outcomes and events. We may, if we like, describe these differences between Jules and Jim in terms of Jim’s greater vulnerability to luck, relative to Jules. But luck should be understood here as simply the complement to various limitations in these agents’ powers of control. The more control there is, the less luck there is. Correlatively, the less control there is, the more luck there is.7 Jim is more vulnerable to luck than Jules is simply because he has less control than Jules does, or because his powers are less developed than Jules’ powers. The Control Interpretation does not, in and by itself, insist on any particular distribution of good luck and bad luck between Jules and Jim as soon as the powers of agency have expired and it is luck that is heading in their direction. The optimistic bias could, of course, encompass both the Pure Luck Interpretation and the Control Interpretation: I might subconsciously think that I have more extensive powers than others, and that these reduce my vulnerability to luck tout court, and also subconsciously think that, whenever I do experience luck, it is much more likely to be good luck than bad luck. Still, we need to keep these interpretations apart, both out of a concern for conceptual hygiene, and because there are different ways in which they can be exploited in Domsky’s argument. The distinction between them will be revisited in the next section.

3.  Combining the Optimistic Bias With the Selfish Bias Daniel Statman has challenged Domsky’s argument by questioning the implications of the selfish bias. Statman’s suggestion is that a genuine regard for self-­interest would actually make it more rational to embrace Intuition 1 and to abandon Intuition 2.8 The optimistic bias has an egocentric character: for each of us in the grip of this bias, the bias has the effect that I judge myself to be less vulnerable to bad luck than other people. If that is so, then Statman thinks that a more sensible and secure self-­interested policy to adopt would make me blameworthy only for those outcomes which fall under my control. Since I can hardly pretend to be omnipotent—sometimes the results of what I intentionally do are bound to reflect various contingencies beyond my control—a better bet for me overall is simply to uphold Intuition 1, and to renounce Intuition 2 (and to hope that others do the same). If I am unwise enough to support Intuition 2, then at least on some occasions I must risk paying the price of greater blameworthiness. Why take that risk? While Statman is right to query the effectiveness of Domsky’s enrolment of the selfish bias, I think this particular argument can be resisted. There are two relevant points. First, it is none too clear how the optimistic bias is supposed to combine with the selfish bias to benefit me if Intuition 1 but not Intuition 2 is favoured. However extensive I take my powers of agency to be, support for Intuition 1 but not Intuition 2 would protect me from a blameworthiness assignment that reflected resultant luck. If I happened to regard my powers of agency as being unusually truncated, rather than optimistically thinking of them as being more extensive than average, then Statman’s reasons for adopting a policy of exclusive support for Intuition 1 would still apply to me in exactly the same way. Intuition 1 will give me the safe option, regardless of how extensive my powers of agency may be, or how prone to good luck I think I am. But the explanation of why we embrace Intuition 2 was supposed to draw upon the combination of the selfish bias with the optimistic bias. Since the interaction between the selfish bias and the optimistic bias is not provided for in 7  See the Introduction, Section 4.

8  Statman (2005), pp. 427–8.

258  Appendix I Statman’s argument, we are entitled to think that a piece of the interpretive picture is probably missing. A second problem arises if we adopt the Control Interpretation of the optimistic bias and then favour Intuition 1 rather than Intuition 2. If I credit myself with powers of agency which other agents do not possess, then I seem potentially committed to blaming myself more than others. Outcomes which may appear to be due to bad luck will in fact be due to my failure to have properly exercised my powers of agency. It may not be obvious to others that I possess such powers, but my possession of the optimistic bias implies that I should still be prepared to ascribe them to myself. So, if I were honest about the biases I harbour, I would admit that I am potentially blameworthy for these outcomes. Now compare other agents: I do not think that they enjoy these hidden powers of agency (even if that is what they think about themselves). The bad luck experienced by others will, by my lights, be merely bad luck; as far as they are concerned, what appears to be bad luck really is bad luck. So again, if I am honest, I will affirm that, by my lights, these agents are not blameworthy for these outcomes. The outcomes attributed to them do not genuinely manifest these others’ agency, though of course they will reflect facts about the limitations of their agency. In short, if I uphold Intuition 1, then I risk being exposed to blame while holding others to be non-­blameworthy. So it is not clear that I will come out ahead of others if I retain Intuition 1 and abandon Intuition 2. This reasoning implies that, for Domsky’s purposes, the Pure Luck Interpretation is the better interpretation of the optimistic bias. If I favour Intuition 2, then I will be comparatively benefited over others, because I assume that their greater exposure to bad luck entails that they will be on average more blameworthy than me.9 At bottom, then, Domsky must be understood as proposing that my only reason for upholding Intuition 2 is to increase my comparative advantage over others by ensuring that other people get into more trouble than I do. I will be better off, relative to them, if they are worse off relative to me, even if my absolute position is unchanged. Even if my absolute position is actually worse than it could otherwise be, I might still be benefited in comparative terms if I am only slightly worse off and they are substantially worse off. A comparative interpretation of benefit attributes to each of us, in effect, a radical form of Rousseau-­style amour-­propre.10 It imputes to us a model of self-­interest mixing ordinary self-­interest with a contemptuous dismissal of, or hostility towards, other people. On this view, it is not enough that I prosper; others must suffer. Indeed, it may not actually be necessary that I prosper, as long as others suffer more than I do. The only thing that matters to me is the gap between them and me. I am not losing, even if I am not prospering, just as long as they are losing more heavily than me. It strikes me as frankly outlandish for Domsky’s whole strategy for unmasking our anti-­ anti-­luckism to rest, when all is said and done, on this sort of extreme, comparative form of self-­ interest. The fact that Domsky is prepared to invest these biases with such significance suggests that there is something else missing in his argument.

As a result, we feel that it is to our credit that we are luckier than our peers’ (2004, p. 458). That looks more like evidence for the Control Interpretation. 9  Statman (2005), pp. 428–9, anticipates that Domsky might embrace a comparative model of benefit, but complains that his case for it is under-­argued. I agree with Statman that Domsky does not say enough on its behalf. But this is nonetheless the model we must adopt if Domsky is to be able to exploit the combination of the optimistic bias with the selfish bias. 10  See Hobbes (1651/1991), esp. ch. 13, and the appalled (and yet somewhat gleeful) application of such ideas in the political genealogy outlined by Rousseau (1754/1997).

Appendix I  259

4.  What is Missing? An implicit implication of Domsky’s argument is that we could have no reasons apart from Rousseau-­style amour-­propre for supporting Intuition 2. On Domsky’s view, it would appear that there is nothing to say in defence of Intuition 2 except that which can be recovered from the subconscious biases which he describes and then indicts. Because there is so little to be said on behalf of Intuition 2, our support for it must reflect the disproportionate influence of these grubby little recesses of our psychological lives. It seems not to have occurred to Domsky that we can assemble a more productive case for the relevance of resultant luck than that. Explanation may abhor a vacuum, but why pretend that we have nothing to go on except these psychological biases? These psychological biases may appear to be more explanatorily enticing than they deserve to be regarded because Domsky is blithely indifferent to all the other reasons we can have for upholding Intuition 2.11 It is natural to think that agents should be answerable for what they have done. The outcomes of what they do cannot be held to be irrelevant to the acts they have performed. And, perhaps, the outcomes of what they do cannot be held to be irrelevant to the moral significance of the mental states which compose their actions. Domsky’s argument pretends that any such attempt to argue for the Restricted Account reflects only the craven attempt to hold on to a few perks in this sub-­ domain of amour-­propre. But that interpretation would be persuasive only if there was nothing else to be said in favour of the Restricted Account. Nothing he says undermines in advance a more positive and patient argument for the Restricted Account. This is what I hope to have assembled in this book.

11  Domsky in fact considers no such arguments: those arguments he does disagree with are only those anti-­luckist strategies which are intended to appear to accommodate Intuition 2, while really siding with Intuition 1.

APPENDIX II

Zimmerman on Freedom, Control, and Luck 1. Introduction I now continue, at a more fine-­grained level of detail, the engagement with Zimmerman’s radical anti-­luckist account which I tackled in Chapter  4 in Sections  4.4 to  4.6. My discussion here assumes a basic familiarity with the contents of Chapter  4, but I will provide some basic recapitulation before moving forwards. The underlying problem with Zimmerman’s argument, I suggested there, is that there is a stark and unresolvable tension between two of his load-­bearing principles. These are the ‘Control Principle’ and the ‘No-­Difference Claim’, respectively.1 The Control Principle can be defined as follows: Control Principle: The degree to which X is blameworthy cannot be affected by factors which are due simply to luck. And the No-­Difference Claim can be defined as follows: No-­Difference Claim: The level of blameworthiness which is due to X will also be due to any agent, Y, if the difference between X’s situation and Y’s situation is due to factors which are simply a matter of luck. The inconsistency can be easily demonstrated. To fix a clear target, let us revisit Case Four, which most fully represents all the lucky differences that could be separating our two agents, Bonnie and Clyde.2 In Case Four, Bonnie is disposed, wrongfully, to φ, is given the opportunity to φ, and causes outcome O as a result of her attempt to φ. Clyde, by contrast, does not attempt to φ, despite having the opportunity to do so, because he is not disposed to φ. If these differences between Bonnie and Clyde are due merely to luck, then Clyde is no less responsible than Bonnie is. He may not be responsible for anything in particular, but he is responsible tout court. And, as far as Zimmerman is concerned, responsibility tout court is what ultimately matters. The No-­Difference Claim plainly implies that Bonnie and Clyde are just as responsible as each other, since the differences between their situations are due merely to luck. But it must also be true that the differences between them, being due to luck, lie beyond their control. Clyde lacks control over the fact that his situation is not Bonnie’s, and Bonnie lacks control over the fact that her situation is not Clyde’s. So Clyde cannot be as

1  These were first defined in Chapter 4, Sections 4.5.1 and 4.6.1. 2  This was first outlined in Chapter 4, Section 4.4.2.

262  Appendix II responsible as Bonnie, by the lights of the Control Principle. The Control Principle and the No-­Difference Claim therefore make inconsistent verdicts. They push in opposite directions. This is a substantial problem for Zimmerman. What, if anything, can be done to make it go away? We must think of ways of restoring compatibility between the No-­Difference Claim and the Control Principle. I can think of two options for Zimmerman, as well as a further coda to the second of them. I will argue that these solutions are unsuccessful.

2.  Restoring Consistency: First Attempt Perhaps the most obvious way of re-­establishing compatibility is to press hard on the conceptual distinction between the categories of responsibility-for and responsibility tout court. According to this suggestion, the Control Principle will be partnered with responsibility-for, while the No-­Difference Claim will be partnered with responsibility tout court. Zimmerman makes it abundantly clear that the existence of the freedom conditions matters only to responsibility-for, and not responsibility tout court. There is a world of difference between having responsibility for something, and having responsibility in virtue of something, as Case Four illustrates. In Case Four, Clyde is not held responsible for either the attempt to φ or the disposition to φ. He cannot be held responsible for these items, since he does not attempt to φ, and he lacks the disposition to φ. He can only be responsible in virtue of the fact that it is only a matter of luck that he does not have the disposition to φ, or that he is not placed in circumstances where, having the disposition to φ, he would then attempt to φ. As I see it, the mere existence of the distinction between responsibility-for and responsibility tout court will not secure the compatibility of the No-­Difference Claim and the Control Principle. We can immediately concede that it is open to Zimmerman to stipulate a new conception of responsibility which is entirely unconcerned with freedom. This new form of responsibility might then function as a purely aretaic or evaluative conception of responsibility, which aims simply to provide a description of what, morally speaking, agents are like: it will impute to them their moral track records, and on this basis enumerate the catalogue of virtues and vices which can be properly ascribed to them, or (more accurately) the combined catalogue of their actual virtues and vices, together with the virtues and vices they would manifest in other circumstances. However, that is not what Zimmerman is after. There are three problems with this approach. First—an ad hominem point, but one which bears upon the proper interpretation of his position—Zimmerman clearly distances himself from this purely evaluative or descriptive conception of responsibility which he identifies in the work of others, such as Robert Adams.3 Second, responsibility tout court is meant to be compatible with the Control Principle, and the Control Principle, in turn, is meant to teach us the deeper lessons about the relevance of freedom to blameworthiness. We already know that Zimmerman does not take himself to be offering a bifurcated account of responsibility: both Bonnie and Clyde

3 Zimmerman (2002), pp. 556–7; see Adams (1985) for an example of this approach. See the Introduction, Section 4, for more on such matters.

Appendix II  263 are responsible, and they are responsible to the same degree.4 They differ only in terms of how the relational nature of the appropriate responsibility claim is to be filled out for each of them. Bonnie is responsible for her attempt to φ, while Clyde is responsible in virtue of the fact that it is only a matter of luck that he is not in the same situation as Bonnie. Third, and in connection to the second point, Zimmerman writes:

Why is freedom commonly regarded as important to moral responsibility? The usual answer is simply that we cannot be morally responsible for what is not in our control. I believe that this answer is accurate but incomplete. There is a more general point to be made, and that is that the degree to which we are morally responsible cannot be affected by what is not under our control.5 The claim stated towards the end of this passage is just the Control Principle. According to this passage, the Control Principle is supposed to subsume, and thus to be compatible with, another claim appearing earlier in this passage which is deemed ‘accurate but incomplete’. We might refer to this ‘accurate but incomplete’ claim as the Freedom Principle:

Freedom Principle: X is responsible only for that which lies within X’s control. Is the Freedom Principle in a position to earn its description as ‘accurate but incomplete’, as this passage proclaims? It might be maintained that the Freedom Principle is ‘accurate but incomplete’ simply because, while the Freedom Principle is essential to one bona fide conception of responsibility (responsibility-for), it is irrelevant to a further bona fide conception of responsibility (responsibility tout court) which, for Zimmerman, is actually more significant than responsibility-for. We should reject this suggestion. This way of demonstrating how the Freedom Principle provides us with an ‘accurate but incomplete’ picture of responsibility does not help us to see how responsibility-for and responsibility tout court are cut out of the same basic cloth of the Control Principle. The Control Principle is reflected in the Freedom Principle, but it is not accommodated by responsibility tout court. The Control Principle goes unaccommodated by responsibility tout court for the familiar reason that Clyde, in Case Four, does not have control over the difference between his situation and Bonnie’s situation. If Clyde is blameworthy for that unchosen difference between him and Bonnie, then he is blameworthy for factors which lie beyond his control. That conclusion is plainly inconsistent with the Control Principle. This proposed remedy has failed. We are stuck with the conclusion that responsibility tout court will work only as a purely evaluative conception of responsibility; an evaluative conception of responsibility, we should add, that is arguably even less interested in control than the purely aretaic Adams-­style model. This is because, unlike the purely aretaic model, responsibility tout court is also sensitive to information about what Clyde would be like in other circumstances which actually lie beyond his control. Though Zimmerman has not lost his right to stipulate the existence of any conception of responsibility he sees fit, 4  See, again, Chapter 4, Section 4.5. 5  Zimmerman (2002), p. 559.

264  Appendix II the onus on him must be to show that the conception of responsibility tout court is significantly derived from the Control Principle. It is difficult to see how he is going to be able to do that.

3.  Restoring Consistency: Second Attempt A second route for restoring consistency between the Control Principle and the No-­ Difference Claim is to sacrifice the scope of the Explosion Argument. One understanding of why cases such as Case Three or Case Four present Zimmerman with such a headache is  that the counterfactuals are doing all the work. There is nothing for which Clyde is responsible in these cases, so the only responsibility that can be imputed to him is responsibility tout court. We have nothing to go on regarding Clyde as he is in the actual world, so  we can only appeal to Clyde as he would be in other possible worlds. The ledger for responsibility-for is empty, so the content for Clyde’s responsibility must be wholly generated by the ledger we generate for responsibility tout court. Those arrangements seem unsatisfactory, for the reasons that have already been rehearsed. One way to avoid this dead end for Zimmerman may be to ensure that the No-­Difference Claim only applies to agents who have non-­ vacuously satisfied the category of responsibility-for. We can use Case One and Case Two to illustrate this sort of approach. In Case One, Clyde makes an attempt to φ. He is, by assumption, responsible for that attempt. Clyde’s attempt to φ is unsuccessful, and does not cause O. But the fact that his attempt to φ is unsuccessful is due to factors beyond his control. So why, on the anti-­ luckist’s view, isn’t he blameworthy to the same degree as Bonnie? This seems to be a more promising case for the application of the No-­Difference Claim, because it offers a fighting chance that it can be squared with the Control Principle. As applied to Case One, the Control Principle says that Clyde’s blameworthiness should not be affected by factors which are a matter of luck. This principle appears to presuppose that we have a way of calibrating Clyde’s blameworthiness which applies to him independently of those luck-­ influenced factors. Now we are assuming that he is responsible for the attempt to φ. According to the No-­Difference Claim, the blameworthiness which should attach to him for attempting to φ should not be modified by factors which are beyond his control. So Clyde is just as blameworthy as Bonnie in Case One. How about Case Two? Here Bonnie attempts to φ, while Clyde has the disposition to φ but produces no attempt to φ due to the absence of any clear opportunity. What is Clyde responsible for in Case Two? Zimmerman wants Clyde to be blameworthy on the grounds that it is just a matter of luck that he does not attempt to φ. What Clyde may have a chance of satisfying in the responsibility-for ledger in Case Two is the disposition to φ if he is given the opportunity to attempt to φ.6 If he can be deemed blameworthy for having that disposition, then the Control Principle will say that his blameworthiness should not be affected by the lucky fact that he is not in circumstances where he would make an attempt to φ. It seems to me, however, that even this restricted form of the Explosion Argument will not pay dividends for Zimmerman. More particularly, Zimmerman faces a dilemma. On the first horn of it, he can dispense with responsibility tout court, but he will then encounter the problem of how to reconcile the degree of responsibility with the scope of responsibility. 6  Perhaps this will strike you as implausible, on the reasonable grounds that neither Bonnie nor Clyde has any deep control over the central features of their moral personalities. If so, Case Two must then be assimilated to the range of cases encompassing Case Three and Case Four, and the scope of the Explosion Argument will be even more tightly restrained.

Appendix II  265 This will expose him to the Division Objection, which we have already encountered in the discussion of the Fairness Intuition.7 On the second horn of the dilemma, Zimmerman’s problem of reconciling the degree of responsibility with the scope of responsibility is eased, but he will not be able to dispense with responsibility tout court, and will thus be unable to avoid the problem of achieving a satisfactory reconciliation between the Control Principle and the No-­Difference Claim. Let us imagine that Zimmerman is prepared to sacrifice the scope of the Explosion Argument, and to agree that his argument applies only to Case One and Case Two. He might then take himself to be able to dispense with responsibility tout court. He will be left with an anti-­luckist conception of responsibility-for, according to which the degree of responsibility-for should not be affected by matters which are only a matter of luck. But now he must face difficult questions concerning the scope and degree of responsibility. Take Case One. If Clyde is not blameworthy to any lesser degree than Bonnie in Case One, then that same degree of blameworthiness must be divided among the morally assessable items for which he is responsible. Bonnie is blameworthy for the attempt to φ, and also outcome O. Clyde is blameworthy only for his attempt to φ, since he does not cause O. Both Clyde and Bonnie are, by assumption, blameworthy to the same degree. But if they are blameworthy to the same degree, and if Bonnie’s blameworthiness for outcome O collects a non-­trivial degree of blameworthiness, it will follow that Clyde must be blameworthy to a greater degree than Bonnie for the morally appraisable item that they have in common: that is, the attempt to φ. So Clyde must be more blameworthy for his attempt φ than Bonnie is for her attempt to φ. That cannot be right. First, the Control Principle implies that Clyde should be no less blameworthy than Bonnie for his attempt to φ, given that it was beyond his control whether his attempt to φ would cause O. But it beggars belief that Clyde should be more blameworthy for his attempt to φ than Bonnie is blameworthy for her attempt to φ. How can a failed attempt be more blameworthy than a successful attempt, other things being equal?8 A second concern is that Clyde cannot be responsible to a greater degree than Bonnie for his attempt to φ if it was not under his control that his attempt to φ caused, or failed to cause, outcome O. The Control Principle, once again, implies that Clyde and Bonnie are not blameworthy to an unequal degree if the difference in what they respectively cause is due to luck, or factors beyond their control. But the fact that Bonnie’s attempt to φ produced O must, by Zimmerman’s own lights, be considered a matter of luck. So it cannot be right that Bonnie should be blameworthy to a greater degree than Clyde for their respective attempts to φ. Zimmerman may protest that the allocation of a greater or lesser degree of blameworthiness for any particular item does not fundamentally matter, given the fact that the total level of blameworthiness is invariant. However, that response seems unsatisfactory. The question he faces is what makes the total degree of blameworthiness invariant. At this stage, it must be responsibility tout court that takes up the slack. If Clyde is just as responsible as Bonnie in Case One, then he must be responsible in virtue of the fact that it is just a matter of luck that his attempt to φ did not produce O.  But we have already examined the problems with responsibility tout court, in which the relational nature of the responsibility claim is cashed out in ‘in virtue of—’ terms, rather than ‘for—’ terms.

7  See Chapter 3, Section 3.2.2. 8 Compare, again, Chapter  3, Section 3.2, when similar reasoning was applied to the Fairness Intuition.

266  Appendix II Responsibility tout court will not permit a genuine reconciliation between the No-­Difference Claim and the Control Principle.

4.  The Anti-­Disaggregation Strategy One interpretive possibility remains, as I see it. Perhaps Zimmerman might still be in a position to insist that Clyde is responsible to the same degree as Bonnie without being forced to say that Clyde’s attempt to φ is more blameworthy than Bonnie’s attempt to φ. Perhaps Zimmerman can say instead that Bonnie’s degree of responsibility is assigned to her for the fact that her attempt to φ produces O, while Clyde’s degree of responsibility is assigned to him solely for his attempt to φ. That is, we do not have to produce an analysis of what Bonnie is responsible for which is then disaggregated into a collection of discrete items: first, her attempt to φ, and second, the fact that her attempt to φ causes O.  By adopting this anti-­disaggregation strategy, we can avoid the problem of assigning to Clyde a higher degree of blameworthiness for his unsuccessful attempt to φ, as compared to Bonnie’s successful attempt to φ. We can then rely on the fact that, by the lights of the No-­Difference Claim, Clyde and Bonnie are equally blameworthy. Any further analysis of the components of their blameworthiness must be guided by that principal and non-­reducible fact. I am unpersuaded of the merits of this anti-­disaggregation strategy. First, we know that attempting to φ is blameworthy, and we also know that attempts to φ which cause O are blameworthy.9 In Case One, Clyde is blamed solely for his attempt to φ. Since Bonnie has also attempted to φ, what can justify our inattention to that fact? Why does Bonnie’s attempt to φ suddenly become completely uninteresting, when Clyde’s attempt to φ gives us our only reason for being morally interested in Clyde? Second, and in any case, the anti-­ disaggregation strategy does not successfully disperse the questions that motivated its initial formulation. Our original question was this: how can the fact that Clyde’s attempt to φ fails while Bonnie’s attempt to φ succeeds make Clyde more blameworthy for his attempt to φ than Bonnie is blameworthy for her attempt to φ? With the anti-­disaggregation strategy in place, we can still ask a modified form of that original question: how can the fact that Clyde produces a failed attempt to φ while Bonnie produces a successful attempt to φ render Clyde blameworthy for his attempt and Bonnie non-­blameworthy for her attempt? If we were puzzled by the original question, then we will be no less puzzled by the modified form of it. The same sort of story can be given for Case Two. Here Clyde does not attempt to φ, but merely has a disposition to φ which is not actualized. It is just a matter of luck that an opportunity for attempting to φ fails to present itself to him. Once again, we face the problem of how to align degree with scope, or how to divide the degree of responsibility among the different number of items for which these agents are respectively responsible. Bonnie is blameworthy for her attempt to φ, whereas Clyde is not. If we assume that Bonnie’s attempt to φ collects a non-­trivial degree of responsibility, then it will seem that Clyde must, in compensation of that fact, be blamed more for having the disposition which is such that he would attempt to φ if he had the opportunity to do so. But Bonnie has that disposition as well. So why should Clyde be more blameworthy for having the same disposition as Bonnie, on the grounds that this disposition is never actualized, whereas 9  Recall the Opportunity Claim from Chapter 4, Section 4.3.1, according to which different items in our practical lives call for moral evaluation. Attempts and outcomes both call for moral evaluation. The anti-­disaggregation strategy is hardly in a strong position to dislodge this claim.

Appendix II  267 Bonnie’s disposition is actualized? Again, that seems absurd. Moreover, there is no reason to think that the anti-­disaggregation strategy enjoys brighter prospects of achieving a better fit for Case Two than it did for Case One. These strategies for reconciling the No-­Difference Claim and the Control Principle have failed. We cannot escape the conclusion that the No-­ Difference Claim exhibits commitments that lie at some distance from the Control Principle. In Chapter  4, Section  4.5.3, I try to say what the deeper commitments of the No-­Difference Claim amount to. The evidence at our disposal suggests that Zimmerman is operating, at least implicitly, under a very austere set of assumptions about what is required for an agent to be morally responsible.

References Adams, R. M. ‘Involuntary Sins’, The Philosophical Review 94 (1985): 3–31. Anderson, Elizabeth. ‘What Is the Point of Equality?’, Ethics 109 (1999): 287–337. Anderson, Elizabeth. ‘Animal Rights and the Values of Nonhuman Life’, in Animal Rights: Current Debates and New Directions, ed. C. R. Sunstein and M. C. Nussbaum. Oxford: Oxford University Press, 2004. Anderson, Elizabeth. ‘The Fundamental Disagreement between Luck Egalitarians and Relational Egalitarians’, Canadian Journal of Philosophy 40 (2010): 1–23. Andre, Judith. ‘Nagel, Williams, and Moral Luck’, Analysis 43 (1983): 202–7. Arneson, Richard. ‘Equality and Equal Opportunity for Welfare’, Philosophical Studies 56 (1989): 77–93. Arneson, Richard. ‘What, if Anything, Renders All Human Beings Equal?’, in Singer and His Critics, ed. Dale Jamieson. Oxford: Blackwell, 1999. Arneson, Richard. ‘Luck Egalitarianism and Prioritarianism’, Ethics 110 (2000): 339–49. Arneson, Richard. ‘Luck Egalitarianism Interpreted and Defended’, Philosophical Topics 32 (2004): 1–20. Arneson, Richard. ‘Do Patriotic Ties Limit Global Justice Duties?’, The Journal of Ethics 9 (2005): 127–50. Arpaly, Nomy. ‘Moral Worth’, Journal of Philosophy 99 (2002): 223–45. Athanassoulis, Nafsika. ‘A Response to Harman: Virtue Ethics and Character Traits’, Proceedings of the Aristotelian Society 101 (2000): 215–21. Athanassoulis, Nafsika. Morality, Moral Luck and Responsibility: Fortune’s Web. Basingstoke: Pallgrave Macmillan, 2005. Barnes, Elizabeth. ‘What You Can Expect When You Don’t Want to be Expecting’, Philosophy and Phenomenological Research 91 (2015): 775–86. Barry, Brian. ‘Equal Opportunity and Moral Arbitrariness’, in Equal Opportunity, ed. N. E. Bowie. Boulder, CO: Westview Press, 1988. Beitz, Charles. ‘Justice and International Relations’, Philosophy & Public Affairs 4 (1975): 360–89. Bell, Malcester. ‘The Standing to Blame: A Critique’, in Blame: Its Nature and Norms, ed. D. Justin Coates and Neal A. Tognazzini. Oxford: Oxford University Press, 2013. Berkey, Brian. ‘Prospects for an Inclusive Theory of Justice: The Case of Non-Human Animals’, Journal of Applied Philosophy 34 (2017): 679–95. Berlin, Isaiah. ‘Two Concepts of Liberty’, in his Four Essays on Liberty. Oxford: Oxford University Press, 1969. Blake, Michael. ‘Distributive Justice, State Coercion, and Autonomy’, Philosophy and Public Affairs 30 (2001): 257–96. Bou-Habib, Paul and Olsaretti, Serena. ‘Liberal Egalitarianism and Workfare’, Journal of Applied Philosophy 21 (2004): 257–70. Boyd, Richard. ‘How to be a Moral Realist’, in Essays on Moral Realism, ed. Geoffrey SayreMcCord. Ithaca, NY: Cornell University Press, 1988. Brogaard, Berit. ‘Epistemological Contextualism and the Problem of Moral Luck’, Pacific Philosophical Quarterly 84 (2003): 351–70.

270 References Broome, John. Rationality Through Reasoning. Oxford: Wiley, 2013. Browne, Brynmor. ‘A Solution to the Problem of Moral Luck’, The Philosophical Quarterly 42 (1992): 345–56. Buchanan, Allen. ‘Justice as Reciprocity versus Subject-Centred Justice’, Philosophy and Public Affairs 19 (1990): 227–52. Burge, Tyler. ‘Individualism and the Mental’, in Midwest Studies in Philosophy 4 (1979): 73–122. Burge, Tyler. ‘Individualism and Psychology’, The Philosophical Review 95 (1986): 3–45. Callcut, Daniel. ‘Living the Life Authentic: Bernard Williams on Paul Gauguin’, Aeon, 11 June 2018. Carter, Ian. ‘Respect and the Basis of Equality’, Ethics 121 (2011): 538–71. Chang, Ruth. ‘Transformative Choices’, Res Philosophica 92 (2015): 237–82. Coates, D. Justin and Tognazzini, Neal A. ‘The Contours of Blame’, in Blame: Its Nature and Norms, ed. D.  Justin Coates and Neal  A.  Tognazzini. New York: Oxford University Press, 2013. Coffman, E. J. ‘Strokes of Luck’, Metaphilosophy 45 (2014): 477–508. Coffman, E. J. Luck: Its Nature and Significance for Human Knowledge. London: Palgrave Macmillan, 2015. Cohen, G. A. ‘Self-Ownership, World Ownership, and Equality: Part II’, Social Philosophy and Policy 3 (1986): 77–96. Cohen, G. A. ‘On the Currency of Egalitarian Justice’, Ethics 99 (1989): 906–44. Cohen, G.  A. If You’re an Egalitarian, How Come You’re So Rich? London: Harvard University Press, 2000. Cohen, G.  A. ‘Luck and Equality: A Reply to Hurley’, Philosophy and Phenomenological Research 72 (2006): 439–46. Cohen, G. A. Rescuing Justice and Equality. London: Harvard University Press, 2008. Cohen, G. A. ‘Fairness and Legitimacy in Justice, and: Does Option Luck Ever Preserve Justice?’, in his On the Currency of Egalitarian Justice and Other Essays in Political Philosophy, ed. Michael Otsuka. Oxford: Princeton University Press, 2011. Cohen, G. A. ‘Casting the First Stone: Who Can, and Who Can’t, Condemn the Terrorists?’, in his Finding Oneself in the Other, ed. Michael Otsuka. Oxford: Princeton University Press, 2012a. Cohen, G. A. ‘Ways of Silencing Critics’, in his Finding Oneself in the Other, ed. Michael Otsuka. Oxford: Princeton University Press, 2012b. Cohen Joshua and Sabel, Charles. ‘Extra Rempublicam Nulla Justitia?’, Philosophy and Public Affairs 24 (2006): 147–74. Crisp, Roger. ‘Moral Luck and Equality of Moral Opportunity’, Proceedings of the Aristotelian Society, Supplementary Volume 91 (2017): 1–20. Cupit, Geoffrey. ‘The Basis of Equality’, Philosophy 75 (2000): 105–25. Davidson, Donald. ‘Agency’, in his Essays on Actions and Events. Oxford: Oxford University Press, 1980. Dewey, John. Outlines of a Critical Theory of Ethics. New York: Hillary House, 1957 (ori­gin­ al­ly published 1891). Domsky, Darren. ‘There is No Door: Finally Solving the Problem of Moral Luck’, Journal of Philosophy 101 (2004): 445–64. Domsky, Darren. ‘Tossing the Rotten Thing Out: Eliminating Bad Reasons Not to Solve the Problem of Moral Luck’, Philosophy 80 (2005): 531–41. Doris, John. Lack of Character: Personality and Moral Behavior. Cambridge: Cambridge University Press, 2002.

References  271 Dworkin, Ronald. ‘The Original Position’, University of Chicago Law Review 40 (1973): 500–33. Dworkin, Ronald. ‘What is Equality? Part 1: Equality of Welfare’, Philosophy & Public Affairs 10 (1981a): 185–246. Dworkin, Ronald. ‘What is Equality? Part 2: Equality of Resources’, Philosophy & Public Affairs 10 (1981b): 283–345. Dworkin, Ronald. Sovereign Virtue: The Theory and Practice of Equality. London: Harvard University Press, 2000. Dworkin, Ronald. ‘Equality, Luck, and Hierarchy’, Philosophy and Public Affairs 31 (2003): 190–8. Dworkin, Ronald. Justice for Hedgehogs. London: Harvard University Press, 2011. Elford, Gideon. ‘Equality of Opportunity and Other-Affecting Choice: Why Luck Egalitarianism Does Not Require Brute Luck Equality’, Ethical Theory and Moral Practice 16 (2013): 139–49. Elford, Gideon. ‘The Coherence of Luck Egalitarianism’, Ethical Theory and Moral Practice 20 (2017): 617–26. Enoch, David. ‘Luck Between Morality, Law, and Justice’, Theoretical Inquiries in Law 9 (2008): 23–59. Enoch, David. ‘Being Responsible, Taking Responsibility, and Penumbral Agency’, in Luck, Value, and Commitment: Themes from the Ethics of Bernard Williams, ed. Ulrike Heuer and Gerald Lang. Oxford: Oxford University Press, 2012. Enoch, David and Guttel, Ehud. ‘Cognitive Biases and Moral Luck’, Journal of Moral Philosophy 7 (2010): 372–86. Enoch, David and Marmor, Andrei. ‘The Case Against Moral Luck’, Law and Philosophy 26 (2007): 405–36. Feinberg, Joel. ‘Problematic Responsibility in Law and Morals’, The Philosophical Review 71 (1962): 340–51. Feinberg, Joel. ‘Noncomparative Justice’, The Philosophical Review 83 (1974): 297–338. Fourie, Carina, Schuppert, Fabian, and Wallimann-Helmer, Ivo. Social Equality: On What It Means to be Equal. Oxford: Oxford University Press, 2015. Forrester, Katrina. In the Shadow of Justice: Postward Liberalism and the Remaking of Political Philosophy. Princeton: Princeton University Press, 2019. Frankfurt, Harry. ‘Equality as a Moral Ideal’, Ethics 98 (1987): 21–43. Freeman, Samuel. ‘Reason and Agreement in Social Contract Views’, Philosophy & Public Affairs 19 (1990): 122–57. Freeman, Samuel. Rawls. London: Routledge, 2007a. Freeman, Samuel. ‘Rawls and Luck Egalitarianism’, in his Justice and the Social Contract: Essays on Rawlsian Political Philosophy. Oxford: Oxford University Press, 2007b. Gardner, John. ‘The Wrongdoing that Gets Results’, Philosophical Perspectives 18 (2004): 53–88. Gardner, John. ‘What is Tort Law For? Part 2. The Place of Distributive Justice’, in Philosophical Foundations of Tort Law, ed. John Oberdiek. Oxford: Oxford University Press, 2014. Gauthier, David. ‘The Social Contract as Ideology’, Philosophy & Public Affairs 6 (1977): 130–64. Greco, John. ‘A Second Paradox Concerning Responsibility and Luck’, Metaphilosophy 26 (1995): 81–96. Hales, Steven D. ‘A Problem for Moral Luck’, Philosophical Studies 172 (2015): 2385–403. Hales, Steven D. ‘Why Every Theory of Luck is Wrong’, Noûs 50 (2016): 490–508.

272 References Hampton, Jean. ‘Contracts and Choices: Does Rawls Have a Social Contract Theory?’, Journal of Philosophy 57 (1980): 315–38. Hankins, Keith. ‘Adam Smith’s Intriguing Solution to the Problem of Moral Luck’, Ethics 126 (2016): 711–46. Hanna, Nathan. ‘Moral Luck Defended’, Noûs 48 (2014): 683–98. Harman, Elizabeth. ‘“I’ll Be Glad I Did It” Reasoning and the Significance of Future Desires’, Philosophical Perspectives 23 (2009): 177–99. Harman, Gilbert. ‘Moral Philosophy Meets Social Psychology: Virtue Ethics and the Fundamental Attribution Error’, Proceedings of the Aristotelian Society 99 (1999): 315–31. Harman, Gilbert. ‘The Nonexistence of Character Traits’, Proceedings of the Aristotelian Society 100 (2000): 223–6. Harman, Gilbert. ‘Skepticism about Character Traits’, Journal of Ethics 13 (2009): 235–42. Harsanyi, John. ‘Can the Maximin Principle Serve as a Basis for Morality? A Critique of John Rawls’s Theory’, The American Political Science Review 69 (1975): 594–606. Hart, H. L. A. Punishment and Responsibility. Oxford: Oxford University Press, 1968. Hartman, Robert. In Defense of Moral Luck: Why Luck Often Affects Praiseworthiness and Blameworthiness. Oxon: Routledge, 2017. Herdova, Marcela and Kearns, Stephen. ‘Get Lucky: Situationalism and Circumstantial Moral Luck’, Philosophical Explorations 18 (2015): 362–77. Hieronymi, Pamela. Freedom, Resentment, and the Metaphysics of Morals. Princeton: Princeton University Press, 2020. Hirose, Iwao. Egalitarianism. Oxon: Routledge, 2014. Hobbes, Thomas. Leviathan, ed. R.  Tuck. Cambridge: Cambridge University Press, 1991 (originally published 1651). Hurley, Susan L. ‘Luck and Equality’, Proceedings of the Aristotelian Society, Supplementary Volume 75 (2001): 51–72. Hurley, Susan L. Justice, Luck, and Knowledge. London: Harvard University Press, 2003. Husak, Douglas N. ‘Is Drunk Driving a Serious Offense?’, Philosophy and Public Affairs 23 (1994): 52–73. Huseby, Robert. ‘Can Luck Egalitarianism Justify the Fact that Some are Worse Off than Others?’, Journal of Applied Philosophy 33 (2016): 259–69. James, Aaron. ‘Constructing Justice for Existing Practice: Rawls and the Status Quo’, Philosophy and Public Affairs 33 (2005): 281–316. Jollimore, Troy. ‘Second Order Desert and Moral Luck’, PEA Soup, 16 October 2004, online philosophy blog at: http://peasoup.typepad.com/peasoup/2004/10/second_order_ de.html. Julius, A. J. ‘Nagel’s Atlas’, Philosophy and Public Affairs 34 (2006): 176–92. Kagan, Shelly. ‘What’s Wrong with Speciesism?’, Journal of Applied Philosophy 33 (2016): 1–21. Kauppinen, Antti. ‘What’s So Great about Experience?’, Res Philosophica 92 (2015): 371–88. Khoury, A.  C. ‘The Objects of Moral Responsibility’, Philosophical Studies 175 (2018): 1357–81. Kiesewetter, Benjamin. The Normativity of Rationality. Oxford: Oxford University Press, 2017. Knight, Carl. Luck Egalitarianism: Equality, Responsibility, and Justice. Edinburgh: Edinburgh University Press, 2009. Knight, Carl and Stemplowska, Zofia. Responsibility and Distributive Justice. Oxford: Oxford University Press, 2011. Kumar, Victor. ‘Empirical Vindication of Moral Luck’, Noûs 53 (2019): 987–1007. Kupperman, Joel. ‘The Indispensability of Character’, Philosophy 76 (2001): 239–50.

References  273 Kutz, Christopher. ‘Against Political Luck’, in Reading Bernard Williams, ed. Daniel Callcut. London: Routledge, 2009. Kymlicka, Will. Contemporary Political Philosophy: An Introduction, 2nd edition, Oxford: Oxford University Press, 2002. Lackey, Jennifer. ‘What Luck is Not’, Australasian Journal of Philosophy 86 (2008): 255–67. Lang, Gerald. ‘A Dilemma for Objective Act-Utilitarianism’, Politics, Philosophy, and Economics 3 (2004): 221–39. Lang, Gerald. ‘Luck Egalitarianism and the See-Saw Objection’, American Philosophical Quarterly 43 (2006): 43–56. Lang, Gerald. ‘Luck Egalitarianism, Permissible Inequalities, and Moral Hazard’, Journal of Moral Philosophy 2 (2009a): 317–38. Lang, Gerald. ‘Thinking Big, Thinking Small: Smilanksy’s Paradoxes’, Iyyun: The Jerusalem Philosophical Quarterly 58 (2009b): 277–91. Lang, Gerald. ‘Discrimination, Partial Concern, and Arbitrariness’, in Luck, Value, and Commitment: Themes from the Ethics of Bernard Williams, ed. Ulrike Heuer and Gerald Lang. Oxford: Oxford University Press, 2012. Lang, Gerald. ‘Jobs, Institutions, and Beneficial Retirement’, Ratio 27 (2014a): 205–21. Lang, Gerald. ‘Why Not Forfeiture?’, in How We Fight, ed. H. Frowe and G. Lang. Oxford: Oxford University Press, 2014b. Lang, Gerald. ‘How Interesting is the “Boring Problem” for Luck Egalitarianism?’, Philosophy and Phenomenological Research 91 (2015a): 698–722. Lang, Gerald. ‘Owning Up: Dworkin on Responsibility’, Jurisprudence 6 (2015b): 298–308. Lang, Gerald. Review of Shaping the Normative Landscape, by David Owens, Analysis 75 (2015c): 175–7. Lang, Gerald. ‘Rawlsian Incentives and the Freedom Objection’, Journal of Social Philosophy 47 (2016): 231–49. Lang, Gerald. ‘What Follows from Defensive Non-Liability?’, Proceedings of the Aristotelian Society 117 (2017): 231–52. Lang, Gerald. ‘Gauguin’s Lucky Escape: Moral Luck and the Morality System’, in Ethics Beyond the Limits: Bernard Williams’ Ethics and the Limits of Philosophy, ed. SophieGrace Chappell and Marcel van Ackaren. Oxon: Routledge, 2019a. Lang, Gerald. Review of One Another’s Equals, by Jeremy Waldron, Mind 128 (2019b): 249–60. Lang, Gerald. Review of In the Shadow of Justice: Postwar Liberalism and the Remaking of Political Philosophy, by Katrina Forrester, Ethics 131 (2020): 143–8. Lang, Gerald and Lawlor, Rob. ‘Numbers Scepticism, Equal Chances, and Pluralism: Taurek Revisited’, Politics, Philosophy and Economics 15 (2016): 298–315. Langton, Rae and Lewis, David. ‘Defining “Intrinsic”’, Philosophy and Phenomenological Research 58 (1998): 333–45. Langton, Rae and Lewis, David. ‘Marshall and Parsons on “Intrinsic”’, Philosophy and Phenomenological Research 63 (2001): 353–5. Latus, Andrew. ‘Moral and Epistemic Luck’, Journal of Philosophical Research 25 (2000): 149–72. Lazenby, Hugh. ‘One Kiss Too Many? Giving, Luck Egalitarianism, and Other-Affecting Choice’, Journal of Political Philosophy 18 (2010): 271–86. Levy, Neil. ‘The Good, the Bad, and the Blameworthy’, Journal of Ethics and Social Philosophy 1 (2005): 1–16. Levy, Neil. ‘The Responsibility of the Psychopath Revisited’, Philosophy, Psychiatry, and Psychology 14 (2007a): 129–38. Levy, Neil. ‘Doxastic Responsibility’, Synthese 155 (2007b): 127–55.

274 References Levy, Neil and McKenna, Michael. ‘Recent Work on Moral Responsibility’, Philosophy Compass 4 (2009): 96–133. Levy, Neil. Hard Luck: How Luck Undermines Free Will and Moral Responsibility. Oxford: Oxford University Press, 2011. Levy, Neil. ‘Dissolving the Puzzle of Resultant Moral Luck’, Review of Philosophy and Psychology 7 (2016): 127–39. Lewis, David. ‘The Punishment That Leaves Something to Chance’, Philosophy & Public Affairs 18 (1989): 53–67. Lippert-Rasmussen, Kasper. ‘Egalitarianism, Option Luck, and Responsibility’, Ethics 111 (2001): 548–79. Lippert-Rasmussen, Kasper. Luck Egalitarianism. London: Bloomsbury, 2015. Lippert-Rasmussen, Kasper. Relational Egalitarianism: Living as Equals. Cambridge: Cambridge University Press, 2018. Lord, Errol. The Importance of Being Rational. Oxford: Oxford University Press, 2018. Manne, Kate. ‘Forgiving Masters: On Bernard Williams’ Moral Luck’, 18 March 2018, at http://as.cornell.edu/news/manne-williams-moral-luck. Manor, Tal. ‘Inequality: Mind the Gap! A Reply to Smilansky’s Paradox of the Baseline’, Analysis 65 (2005): 265–8. Marshall, Dan and Parsons, Josh. ‘Langton and Lewis on “Intrinsic”’, Philosophy and Phenomenological Research 63 (2001): 347–51. Martin, Rex and Reidy, David. Rawls’s Law of Peoples: A Realistic Utopia? Oxford: Blackwell, 2007. Marx, Karl. The Critique of the Gotha Programme, in The Marx-Engels Reader, 2nd edition, ed. Richard C. Tucker. New York: W. W. Norton & Co, 1978 (originally published 1891). Mason, Eleanor. Ways to be Blameworthy: Rightness, Wrongness, and Responsibility. Oxford: Oxford University Press, 2019. MacIntyre, Alasdair. After Virtue: A Study in Moral Theory. London: Duckworth & Co. Ltd, 1981. McGinn, Colin. ‘Charity, Interpretation, and Belief ’, Journal of Philosophy 74 (1977): 521–35. McMahan, Jeff. ‘Cognitive Disability, Misfortune, and Justice’, Philosophy and Public Affairs 25 (1996): 3–34. McMahan, Jeff. ‘“Our Fellow Creatures”’, The Journal of Ethics 9 (2005): 353–80. McMahan, Jeff. Killing in War. Oxford: Clarendon Press, 2009. Mele, Afred R. Free Will and Luck. Oxford: Oxford University Press, 2006. Mill, John Stuart. Utilitarianism, ed. Roger Crisp. Oxford: Oxford University Press, 1998 (originally published 1861). Miller, David. ‘The Incoherence of Luck Egalitarianism’, Centre for the Study of Social Justice Working Papers, SJ022, Department of Politics and International Relations, University of Oxford, 2014. Moore, Michael  S. ‘The Independent Moral Significance of Wrongdoing’, Journal of Contemporary Legal Issues 5 (1994): 237–81. Mulhall, Stephen. On Film, 3rd edition. Oxon: Routledge, 2016. Nagel, Thomas. ‘Moral Luck’, Proceedings of the Aristotelian Society, Supplementary Volume 50 (1976): 137–51. Nagel, Thomas. ‘Moral Luck’, in his Mortal Questions. Cambridge: Cambridge University Press, 1979. Nagel, Thomas. ‘Justice and Nature’, Oxford Journal of Legal Studies 17 (1997): 303–21. Nagel, Thomas. ‘The Problem of Global Justice’, Philosophy and Public Affairs 33 (2005): 113–47.

References  275 Nathan, Christopher. ‘Need There be a Defence of Equality?’, Res Publica 17 (2011): 211–25. Nehamas, Alexander. Nietzsche: Life as Literature. London: Harvard University Press, 1985. Nelkin, Dana Kay. ‘Moral Luck’, Stanford Encyclopedia of Philosophy, 2008, at: http://plato. stanford.edu/entries/moral-luck/. Last accessed 2 February 2016. New, Christopher. ‘Time and Punishment’, Analysis 52 (1992): 35–40. New, Christopher. ‘Punishing Times: Reply to Smilansky’, Analysis 55 (1995): 60–2. Norcross, Alastair. ‘Puppies, Pigs, and People: Eating Meat and Marginal Cases’, Philosophical Perspectives 18 (2004): 229–45. Nozick, Robert. Anarchy, State and Utopia. Oxford: Basil Blackwell, 1974. Nussbaum, Martha Craven. The Fragility of Goodness: Luck and Ethics in Greek Tragedy and Philosophy. Cambridge: Cambridge University Press, 1986. Nussbaum, Martha Craven. Frontiers of Justice: Disability, Nationality, Species Membership. London: Harvard University Press, 2007. Olsaretti, Serena. ‘Responsibility and the Consequences of Choice’, Proceedings of the Aristotelian Society, 109, Part 2 (2007): 165–88. O’Neill, Martin. ‘The Facts of Inequality’, Journal of Moral Philosophy 7 (2010): 397–409. Otsuka, Michael. ‘Prerogatives to Depart from Equality’, Royal Institute of Philosophy Supplement 58 (2006): 95–112. Otsuka, Michael. ‘Moral Luck: Optional, not Brute’, Philosophical Perspectives 23 (2009): 373–88. Owens, David. Shaping the Normative Landscape. Oxford: Oxford University Press, 2012. Parfit, Derek. Reasons and Persons. Oxford: Clarendon Press, 1984. Parfit, Derek. ‘Equality or Priority?’, Lindley Lecture, Department of Philosophy, University of Kansas, 1991. Parry, Jonathan. ‘Defensive Harm, Consent, and Intervention’, Philosophy & Public Affairs 45 (2017): 356–96. Paul, L. A. Transformative Experience. Oxford: Oxford University Press, 2014. Peels, Rik. ‘A Modal Solution to the Problem of Moral Luck’, American Philosophical Quarterly 52 (2015): 73–82. Pogge, Thomas. Realizing Rawls. London: Cornell University Press, 1989. Preda, Adina. Review of Why Inequality Matters: Luck Egalitarianism, Its Meaning and Value, by Shlomi Segall, Ethics 128 (2017): 276–81. Pritchard, Duncan. Epistemic Luck. Oxford: Oxford University Press, 2005. Pritchard, Duncan. ‘The Modal Account of Luck’, Metaphilosophy 45 (2014): 594–618. Putnam, Hilary. ‘The Meaning of “Meaning”’, in his Philosophical Papers, Vol. II: Mind, Language, and Reality. Cambridge: Cambridge University Press, 1975. Rachels, James. Created from Animals: The Moral Implications of Darwinism. Oxford: Oxford University Press, 1990. Railton, Peter. ‘Moral Realism’, The Philosophical Review 95 (1986): 163–207. Rakowski, Eric. Equal Justice. Oxford: Oxford University Press, 1991. Rawls, John. ‘Two Concepts of Rules’, The Philosophical Review 64 (1955): 3–32. Rawls, John. A Theory of Justice. Oxford: Oxford University Press, 1971. Rawls, John. Political Liberalism. New York: Columbia University Press, 1993. Rawls, John. A Theory of Justice, Revised Edition. London: Harvard University Press, 1999a. Rawls, John. The Law of Peoples. London: Harvard University Press, 1999b. Rawls, John. Justice as Fairness: A Restatement, ed. Erin Kelly. London: Harvard University Press, 2001. Raz, Joseph. The Morality of Freedom. Oxford: Clarendon Press, 1986. Raz, Joseph. ‘The Myth of Instrumental Rationality’, Journal of Ethics and Social Philosophy 1 (2005): 1–28.

276 References Raz, Joseph. ‘Responsibility and the Negligence Standard’, Oxford Journal of Legal Studies 30 (2010): 1–18. Raz, Joseph. ‘Agency and Luck’, in Luck, Value, and Commitment: Themes from the Ethics of Bernard Williams, ed. Ulrike Heuer and Gerald Lang. Oxford: Oxford University Press, 2012. Rescher, Nicholas. ‘Moral Luck’, in Moral Luck, ed. Daniel Statman. Albany, NY: State University of New York Press, 1993. Richards, Norvin. ‘Luck and Desert’, Mind 95 (1986): 198–209. Roemer, John. Theories of Distributive Justice. London: Harvard University Press, 1996. Rousseau, Jean Jacques. ‘Discourse on the Origin of Inequality’, in The Discourses and Other Early Political Writings, ed. Victor Gourevitch. Cambridge: Cambridge University Press, 1997 (originally published 1754). Royzman, Edward and Kumar, Rahul. ‘Is Consequential Luck Morally Inconsequential? Empirical Psychology and the Reassessment of Moral Luck’, Ratio 17 (2004): 329–44. Salow, Bernhard. ‘Partiality and Retrospective Justification’, Philosophy and Public Affairs 45 (2017): 8–26. Sandel, Michael. Liberalism and the Limits of Justice, 2nd edition. Cambridge: Cambridge University Press, 1998. Sandel, Michael. ‘The Procedural Republic and the Unencumbered Self ’, in his Public Philosophy: Essays on Morality in Politics. London: Harvard University Press, 2005. Sangiovanni, Andrea. ‘Global Justice, Reciprocity, and the State’, Philosophy and Public Affairs 35 (2007): 3–39. Sangiovanni, Andrea. ‘The Irrelevance of Coercion, Imposition, and Framing to Distributive Justice’, Philosophy and Public Affairs 40 (2012): 79–110. Sartorio, Carolina. ‘Resultant Luck’, Noûs 84 (2012): 63–86. Scanlon, T.  M. ‘The Significance of Choice’, Tanner Lectures on Human Values 8 (1988): 149–216, ed. Sterling M. McMurrin. Salt Lake City: University of Utah Press. Scanlon, T. M. What We Owe to Each Other. London: Harvard University Press, 1998. Scanlon, T. M. ‘The Diversity of Objections to Inequality’, in his The Difficulty of Tolerance: Essays in Political Philosophy. Cambridge: Cambridge University Press, 2003. Scanlon, T.  M. Moral Dimensions: Permissibility, Meaning, Blame. London: Harvard University Press, 2008. Scanlon, T. M. Why Does Inequality Matter? Oxford: Oxford University Press, 2018. Scheffler, Samuel. ‘Responsibility, Reactive Attitudes, and Liberalism in Philosophy and Politics’, Philosophy and Public Affairs 21 (1992): 299–32; reprinted in Scheffler, Boundaries and Allegiances. Scheffler, Samuel. Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought. Oxford: Oxford University Press, 2001. Scheffler, Samuel. ‘What is Egalitarianism?’, Philosophy and Public Affairs 31 (2003): 5–39. Scheffler, Samuel. ‘Choice, Circumstance, and the Value of Equality’, Politics, Philosophy, and Economics 4 (2005): 5–28. Scheffler, Samuel. (2015) ‘The Practice of Equality’, in Social Equality: On What It Means to be Equal, ed. Carina Fourie, Fabian Schuppert, and Ivo Wallimann-Helmer. Oxford: Oxford University Press, 2015. Schemmel, Christian. ‘Distributive and Relational Equality’, Politics, Philosophy, and Economics 11 (2012): 123–48. Segall, Shlomi. ‘In Solidarity with the Imprudent: A Defense of Luck Egalitarianism’, Social Theory and Practice 33 (2007): 177–98. Segall, Shlomi. Equality and Opportunity. Oxford: Oxford University Press, 2013. Segall, Shlomi. ‘What’s So Egalitarian About Luck Egalitarianism?’, Ratio 28 (2015): 349–68.

References  277 Segall, Shlomi. Why Inequality Matters: Luck Egalitarianism, Its Meaning and Value. Cambridge: Cambridge University Press, 2016. Seligman, Matthew. ‘Luck, Leverage, and Equality: A Bargaining Problem for Luck Egalitarians’, Philosophy & Public Affairs 35 (2007): 266–92. Sen, Amartya. ‘Equality of What?’, Tanner Lectures on Human Values 1 (1980): 197–220, ed. Sterling M. McCurrin. Salt Lake City: University of Utah Press. Sher, George. Desert. Princeton: Princeton University Press, 1987. Sher, George. Equality for Inegalitarians. Cambridge: Cambridge University Press, 2014. Shields, Liam. Just Enough: Sufficiency as a Demand of Justice. Edinburgh: Edinburgh University Press, 2016. Shoemaker, David. ‘Attributability, Answerability, and Accountability: Towards a Wider Theory of Moral Responsibility’, Ethics 121 (2011): 603–32. Shupe, Eli. ‘Transformative Experience and the Limits of Revelation’, Philosophical Studies 173 (2016): 3119–32. Singer, Peter. ‘Famine, Affluence, and Morality’, Philosophy and Public Affairs 1 (1972): 229–43. Singer, Peter. ‘All Animals are Equal’, Philosophic Exchange 5 (1974): 103–16; reprinted in Singer, Unsanctifying Human Life. Singer, Peter. ‘Is Racial Discrimination Arbitrary?’, Philosophia 8 (1978): 185–203; reprinted in Singer, Unsanctifying Human Life. Singer, Peter. Practical Ethics, 2nd edition. Cambridge: Cambridge University Press, 1993. Singer, Peter. Unsanctifying Human Life, ed. Helga Kuhse. Oxford: Blackwell, 2001. Smilansky, Saul. ‘The Time to Punish’, Analysis 54 (1994): 52. Smilansky, Saul. ‘Choice-Egalitarianism and the Paradox of the Baseline’, Analysis 63 (2003): 146–51. Smilansky, Saul. ‘Choice-Egalitarianism and the Paradox of the Baseline: a Reply to Manor’, Analysis 65 (2005): 333–7. Smilansky, Saul. Ten Moral Paradoxes. Oxford: Blackwell, 2007. Smith, Adam. The Theory of Moral Sentiments, ed. D. D. Raphael and A. L. Macfie. Oxford: Oxford University Press, 1976 (originally published 1759). Smith, Angela M. ‘Responsibility for Attitudes: Activity and Passivity in Mental Life’, Ethics 115 (2005): 236–71. Smith, Angela M. ‘On Being Responsible and Holding Responsible’, The Journal of Ethics 11 (2007): 465–84. Smith, Angela  M. ‘Control, Responsibility, and Moral Assessment’, Philosophical Studies 138 (2008): 367–92. Smith, Angela  M. ‘Attributability, Answerability, and Accountability: In Defense of a Unified Account’, Ethics 122 (2012): 575–89. Smith, Angela M. ‘Responsibility as Answerability’, Inquiry 58 (2015): 99–126. Statman, Daniel. ‘The Time to Punish and the Problem of Moral Luck’, Journal of Applied Philosophy 14 (1997): 129–35. Statman, Daniel. ‘Doors, Keys, and Moral Luck: A Reply to Domsky’, Journal of Philosophy 102 (2005): 422–36. Steinbock, Bonnie. ‘Drunk Driving’, Philosophy and Public Affairs 14 (1985): 278–95. Stemplowska, Zofia. ‘Making Justice Sensitive to Responsibility’, Political Studies 57 (2009): 237–59. Strawson, Galen. ‘The Impossibility of Moral Responsibility’, Philosophical Studies 75 (1994): 5–24. Strawson, Galen. ‘Free Will’, in Routledge Encyclopedia of Philosophy, vol. 3, ed. Edward Craig. London: Routledge, 1998.

278 References Strawson, P.  F. ‘Freedom and Resentment,’ Proceedings of the British Academy 48 (1962): 1–25. Sussman, David. ‘Is Agent-Regret Rational?’, Ethics 128 (2018): 788–808. Tadros, Victor. ‘Duty and Liability’, Utilitas 24 (2012): 259–77. Taurek, John. ‘Should the Numbers Count?,’ Philosophy and Public Affairs 6 (1977): 293–316. Temkin, Larry. Inequality. Oxford: Oxford University Press, 1993. Temkin, Larry. ‘Equality as Comparative Fairness’, Journal of Applied Philosophy 34 (2017): 43–60. Thomas, Alan. ‘Virtue Ethics and an Ethics of Care: Complementary or in Conflict?’, Eidon 14 (2011): 132–51. Thomson, Judith Jarvis. The Realm of Rights. London: Harvard University Press, 1990. Thomson, Judith Jarvis. ‘Self-Defense’, Philosophy and Public Affairs 20 (1991): 283–310. Thomson, Judith Jarvis. ‘Morality and Bad Luck’, in Moral Luck, ed. Daniel Statman. Albany, NY: State University of New York Press, 1993. Todd, Patrick. ‘Strawson, Moral Responsibility, and the “Order of Explanation”: An Intervention’, Ethics 127 (2016): 208–40. Todd, Patrick. ‘A Unified Account of the Moral Standing to Blame’, Noûs 53 (2019): 347–74. Tomlin, Patrick. ‘Can I be a Luck Egalitarian and a Rawlsian?’, Ethical Perspectives 19 (2012): 371–97. Valentini, Laura. ‘Coercion and (Global) Justice’, American Political Science Review 105 (2011): 205–20. Vallentyne, Peter. ‘Brute Luck, Option Luck, and the Equality of Initial Opportunities’, Ethics 112 (2002): 529–57. Van Inwagen, Peter. ‘The Incompatibility of Free Will and Determinism’, Philosophical Studies 27 (1975): 185–99. Van Parijs, Philippe. Real Freedom for All. Oxford: Clarendon Press, 1995. Voigt, Kristin. ‘The Harshness Objection: Is Luck Egalitarianism Too Harsh on the Victims of Option Luck?’, Ethical Theory and Moral Practice 10 (2007): 389–407. Waldron, Jeremy. ‘Basic Equality’, New York University School of Law, Public Law and Legal Theory Research Paper No. 08–61, 2008. Waldron, Jeremy. One Another’s Equals. Harvard: Harvard University Press, 2017. Walker, Margaret Urban. ‘Moral Luck and the Virtues of Impure Agency’, in Moral Luck, ed. Daniel Statman. Albany, NY: State University of New York Press, 1993. Wallace, R. Jay. ‘Justification, Regret, and Moral Complaint: Looking Forward and Looking Backward on (and in) Human Life’, in Luck, Value and Commitment: Themes from the Ethics of Bernard Williams, ed. Ulrike Heuer and Gerald Lang. Oxford: Oxford University Press, 2012. Wallace, R. Jay. The View from Here: On Affirmation, Attachment, and the Limits of Regret. Oxford: Oxford University Press, 2013. Watson, Gary. ‘Two Faces of Responsibility’, Philosophical Topics 24 (1996): 227–48. Webber, Jonathan. ‘Virtue, Character, and Situation’, Journal of Moral Philosophy 3 (2006): 193–213. Wedgwood, Ralph. The Value of Rationality. Oxford: Oxford University Press, 2017. White, Stuart. ‘What’s Wrong with Workfare?’, Journal of Applied Philosophy 27 (2004): 271–84. Williams, Andrew. ‘How Gifts and Gambles Preserve Justice’, Economics and Philosophy 29 (2013): 65–85. Williams, Bernard. Morality: An Introduction to Ethics. Cambridge: Cambridge University Press, 1972a.

References  279 Williams, Bernard. ‘Ethical Consistency’, in his Problems of the Self: Philosophical Papers 1956–1972. Cambridge: Cambridge University Press, 1972b. Williams, Bernard. ‘The Idea of Equality’, in his Problems of the Self: Philosophical Papers 1956–1972. Cambridge: Cambridge University Press, 1972c. Williams, Bernard. ‘A Critique of Utilitarianism’, in J. J. C. Smart and Bernard Williams, Utilitarianism: For and Against. Cambridge: Cambridge University Press, 1973. Williams, Bernard. ‘Moral Luck’, Proceedings of the Aristotelian Society, Supplementary Volume 50 (1976): 115–35. Williams, Bernard. ‘Moral Luck’, in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press, 1981a. Williams, Bernard. ‘Persons, Character, and Morality’, in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press, 1981b. Williams, Bernard. ‘Internal and External Reasons’, in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press, 1981c. Williams, Bernard. ‘Politics and Moral Character’, in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press, 1981d. Williams, Bernard. Ethics and the Limits of Philosophy. London: Fontana Press/ Collins, 1985. Williams, Bernard. ‘Postscript’, in Moral Luck, ed. Daniel Statman. Albany, NY: State University of New York Press, 1993. Williams, Bernard. ‘Voluntary Acts and Responsible Agents’, in his Making Sense of Humanity and Other Philosophical Papers 1982–1993. Cambridge: Cambridge University Press, 1995a. Williams, Bernard. ‘Internal Reasons and the Obscurity of Blame’, in his Making Sense of Humanity and Other Philosophical Papers 1982–1993. Cambridge: Cambridge University Press, 1995b. Williams, Bernard. In The Beginning Was The Deed, ed. Geoffrey Hawthorn. London: Princeton University Press, 2005. Williams, Bernard. ‘The Human Prejudice’, in his Philosophy as a Humanistic Discipline, ed. A. W. Moore. London: Princeton University Press, 2006. Wolf, Susan. Freedom Within Reason. Oxford: Oxford University Press, 1990. Wolf, Susan. ‘The Moral of Moral Luck’, Philosophic Exchange 31 (2001), at: http://digitalcommons.brockport.edu/phil_ex/vol31/iss1/1. Wolf, Susan. ‘ “One Thought Too Many”: Love, Morality, and the Ordering of Commitment’, in Luck Value and Commitment: Themes from the Ethics of Bernard Williams, ed. Ulrike Heuer and Gerald Lang. Oxford: Oxford University Press, 2012. Wolff, Jonathan. ‘Fairness, Respect, and the Egalitarian Ethos’, Philosophy and Public Affairs 27 (1998): 97–122. Wolff, Jonathan. ‘Scanlon on Social and Material Inequality’, Journal of Moral Philosophy 10 (2013): 406–25. Yaffe, Gideon. Attempts. Oxford: Oxford University Press, 2010. Young, Iris Marion. ‘Political Responsibility and Structural Injustice’, Lindley Lecture, Department of Philosophy, University of Kansas, 2003. Young, Iris Marion. Responsibility for Justice. New York: Oxford University Press, 2011. Zimmerman, Michael J. ‘Luck and Moral Responsibility’, Ethics 97 (1987): 374–86. Zimmerman, Michael J. ‘Taking Luck Seriously’, Journal of Philosophy 99 (2002): 553–76. Zimmerman, Michael  J. ‘Moral Luck: A Partial Map’, Canadian Journal of Philosophy 36 (2006): 585–608.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages. Absence of Regret Interpretation  146–8 Accommodation strategies  44–51, 157n.62, 255 Accountability  13–15, 17–18 Actual Outcome Model  91–2 Adams, R. M.  12n.34, 262–4 Agent-regret  1–2, 131–2, 145–9, 146n.47, 156–9 Agent-Regret Model  157–9 Aim-sensitive luck  40–2, 72–3, 79 Ambition-sensitivity  169n.26, 179, 183, 185n.71 Amorality  135–6, 139 Amour-propre 258–9 Anderson, E.  164n.2, 167, 194n.82, 196n.87, 251n.40 Andre, J.  36n.12 Anna Karenina  136–7, 139, 139n.26, 142, 146n.46 Answerability  16–18, 158 Anti-anti-luckism  8n.24, 18, 21, 58–61, 255, 258 Anti-disaggregation strategy  266–7 Anti-Luck Account  21, 32–3, 36, 72, 79, 82, 85, 87, 89–90, 92–3, 103–5, 108, 255–6 Anti-Luck Variant  213–14 Appraisability  4–5, 11–12, 17–19 Arneson, R.  167, 195n.85, 237n.13, 241n.18, 242n.20, 250n.36 Assassins (case)  29–30, 35, 37–46, 52–3, 64–76, 78–9, 82, 96–7, 100–1, 107, 117, 132–3, 184, 255n.4 Assessability 11–12 Asymmetrical View  174–6 Athanassoulis, N.  12n.33, 134n.9 Attributability  11–15, 17–18 Bad Cases  93, 95–6, 98, 99n.19, 100–3, 106–8, 110n.16, 111 Bargaining  206, 227–30, 232 Barry, B.  199, 203n.6 Baseline-relative luck egalitarianism  163–4, 185–201 Basic Argument  129 Basic equality  24, 233, 241–50, 253–4 Basic structure  201, 211, 220, 222 Beitz, C.  236n.9

Bell, M.  18n.52 Bentham, J.  165–6 Berlin, I.  155n.60 Blake, M.  235–7, 236n.4 Blame versus blameworthiness  18–19 Boring Problem  163–4, 177–89, 187n.73, 193, 196–8, 212 Bou-Habib, P.  164n.2 Boyd, R.  31n.7 Brainwashing 155 Brute luck  23–4, 169, 178, 178n.56, 181, 182n.65, 185, 185n.71, 192, 198 Burge, T.  64n.23 Callcut, D.  137n.22 Careers open to talents  207, 218 Carter, I.  165n.6, 241n.18, 242n.20, 243n.21, 244–5 Case One (case)  116, 121–5, 264–7 Case Two (case)  117, 123–5, 128, 264–7 Case Three (case)  118, 120, 123–5, 264, 264n.6 Case Four (case)  119–27, 261–4 Causal luck  7–9 Character  11–12, 14, 16, 24, 29, 35–7, 49, 57–63, 106–8, 118, 122–3, 127–9, 256–7 Choice  13, 60–1, 102, 128–9, 141, 167–9, 172, 175, 178–9, 181–3, 185–91, 193–4, 199–200, 225–6, 228 Choice egalitarianism  187n.73 See also Luck egalitarianism Choiceworthiness 227 Choice-sensitivity  181–3, 186, 190–3, 197–8 Circumstantial luck  6–7, 9–11, 21, 38n.14, 51n.39, 103–15, 117, 119, 123–4, 126–7, 138–9 Coates, D. J.  18n.52 Co-authorship (of laws)  237 Coffman, E. J.  3n.11 Cohen, G. A.  18n.52, 164–5, 167nn.13,15, 168, 169n.25, 177nn.50,51, 183n.67, 186n.72, 194n.83, 202, 203n.8, 204–5, 216n.44, 230–3, 245n.26 Coincidental control  55n.2

282 Index Comparability 166–7 Comparative fairness egalitarianism  102–3 Comparative justice  39, 96–7, 101 Comparative luck  43–4, 65–6, 71, 75, 82, 96, 101 Conceptions of the good  24, 206–7, 224–5, 227–8, 252 Concessive Response  182–3, 186 Consequential liability  98–100, 156–9 Consequential luck: see Resultant luck Consequentialism 47n.25 See also Utilitarianism Constitutive luck  7–11, 21–2, 105, 106n.4, 115, 118–19, 123–4, 126–7, 138–9, 163 Contractarian apparatus  23, 202–3, 209, 224, 232, 253 Contraction Strategy  124, 125n.47, 127 Control Interpretation  256–8 Control Principle  120–1, 123, 125–7, 129, 261–4 Control Requirement  36–40 Cosmopolitanism  23–4, 234, 236n.9, 238–40 Courage (case)  106–7 Crisp, R.  1n.1, 20–1, 142n.36, 157n.62 Culpability  33–5, 45, 47, 53, 63, 69–70, 72, 74–6, 80, 91, 95–7, 100–2, 157–8 Cupit, G.  241n.18, 242n.20, 243n.21, 250n.36 Currency of justice  164–5, 192, 244–5 Davidson, D.  63–5 Debunking explanations  72n.29, 157n.62, 255–7 ‘Deep’ responsibility  12–13 Defensive liability  100 Degree versus scope of blameworthiness or responsibility  35, 50n.37, 53n.1, 83–4, 121–3, 127–9, 261 Deliberate control  55n.2 Democratic equality  208 Denial Claim  54–6, 60, 65–8, 71, 74–5, 93–5, 110n.16 Deontic equality  219–20 Desert  97–8, 210–11 Desert-Sceptical Variant  210–11, 214 Determination Claim  61–6, 68–71, 74–6, 84, 95, 96n.14 Determinism 7 Deterrence Strategy  47–8 Dewey, J.  11–12 Difference principle  207–8 Diminishing marginal utility  165–6 Dismissive Response  180–2, 185–7 Dispositions  4–8, 10, 12, 24, 36–7, 62, 105–6, 106n.6, 108–9, 111–13, 119–21, 256, 262, 264, 266–7 See also Enacted dispositions

Division Objection  64n.22, 82–4, 264–5 Domsky, D.  7nn.20,21, 109n.15, 255–9 Drivers (case)  29–30, 35, 38n.14, 39–46, 49–50, 49n.30, 52–3, 59, 65–6, 71–8, 82, 84, 96–8, 100–1, 107, 122n.39, 132–3, 156–9, 184 Dworkin, R.  129n.54, 164n.5, 165n.8, 169, 202, 209 Egalitarian baseline  177, 184–92, 196–8 Egalitarian default  170–4, 176–7, 186, 198 Egalitarian Fallacy  82, 92, 170–4, 184 Egalitarian plateau  165–7 Elford, G.  178n.55, 185n.71, 192n.78 Enacted dispositions  62 Endowments  7, 173, 206–8, 210, 212, 214, 216–18, 219n.50, 220–5, 227–30, 239–40, 246–7 See also Natural endowments, Social endowments Endowments facts  216–18, 219n.50, 220–3, 239 Enoch, D.  3n.13, 8, 19–25, 59, 63, 148n.49, 157n.62, 255nn.1–2 Entitlement conception of justice  165, 176n.49 Envatted counterparts  64 Epistemic luck  2–5 Epistemic Strategy  45–7 Equal Blameworthiness Model  91–2 Equal Worlds  85–9, 92, 103 Equitable maxim  20–2 Ethical anti-realism  143–4 Ethical consistency  143–4 Evolution 46–7 Expansion Strategy  123–4, 125n.47, 127 Explosion Argument  115–25, 127, 264–6 External Claim  33–5, 56–7, 65, 69–70, 76, 96 Externalism  16, 56–7, 63–5, 75–6, 122n.41 Extrinsic failure (of luck)  135, 138–40, 142, 152 Fair equality of opportunity  205, 207–8 Fairness Intuition  21, 30, 37–8, 51, 53n.1, 71, 75, 82–93, 103, 264–5, 265n.8 Fate-sharing luck  89–93, 103, 183 Feinberg, J.  1n.2, 10 First principle (of Rawls’s justice as fairness)  205, 231n.68 First Single Two Buttons (case)  65–6 Focus Claim  110, 113–14 Forfeiture 99 Formal equality  173–4, 177 Free will  8–11, 36n.12, 129–30, 163, 184 Freedom conditions  120, 123, 262 Freedom Objection  231n.68

Index  283 Freedom Principle  263 Freeman, S.  173–4, 202n.2, 206n.16, 211, 214n.38, 215 Gardner, J.  75n.30, 101n.26 Gauguin, P.  134–5 Gauguin 2  140–2 Gauguin 3  154 Gauguin’s Escape (case)  131, 134–40, 142, 145–6, 148 Gauthier, D.  203n.5 Generosity (case)  106–7 Gift-giving  175, 177, 185n.71, 192–3, 192n.78 Good Cases  93, 95–6, 100–3, 106–8, 110n.16, 111, 113 Greco, J.  59n.10, 59n.12 Ground rules  91–2, 221–3, 225–6, 228–30 Hales, S. D.  2–3, 3n.11 Hankins, K.  1n.1 Hanna, N.  38n.14, 115n.24 Harsanyi, J.  208n.25 Hart, H. L. A.  47n.25 Hartman, R.  1n.2, 4n.16, 8, 59, 59n.12 Herdova, M.  107n.8, 107n.9 Hobbes, T.  258n.10 Hurley, S.  82, 92, 129n.54, 163–4, 170–3, 177–87, 232n.70, 238 Huseby, R.  180n.59 Hypological judgements  121–2 Hypothetical Challenge  209, 230 Hypothetical contracts  202–4, 206, 209, 224, 230, 233 Ignorance  16, 79 See also Veil of ignorance Improved Transformation Interpretation  153–6 Incoherence Objection  235–40, 253–4 Indeterminacy  70–1, 85n.2, 180, 182–3, 191–2, 232 Inequality Variant (of Neutralization Interpretation) 211–14 Insiders  234, 237–40 Insider-insider relationships  237–9 Insider-outsider relationships  237–9 Integrity  140, 218–19 Internal Claim  33–5, 45–6, 56–7, 65, 69–70, 72–3, 76, 79–80, 82 International justice  234–8, 253–4 Interpersonal fate-sharing luck: see Fate-sharing luck Interpersonal luck  42–4, 66, 82, 87–93, 96, 98–100, 103, 169–72, 184, 198–9 Intrinsic failure (of luck)  135, 138–40, 142, 152

Intuition Problem  194–5 Irrationality 79 Irrelevance Interpretation  203–5, 224–31, 240, 248n.30 Irrelevance Intuition  21, 30, 36–9, 51–2, 61, 71, 79, 82, 88, 92, 103 Jollimore, T.  97–8, 100 Judge Actual  6, 9–10, 19–20 Judge Counterfactual  6, 9–10, 19–20 Judge Possible  10 Justice as fairness  202–5, 241–8, 252–3 Official argument  206–7 Unofficial argument  207–8 ‘Justification defeats liability’ claim  143 Kant, I.  133–4 Kearns, S.  107nn.8,9 Khoury, A. C.  35n.10, 64, 122n.41 Kumar, V.  47n.25 Kutz, C.  31, 34 Kymlicka, W.  165–7, 203n.5, 209, 227n.64 Lack of Control Account (of luck)  2–5, 8 Langton, R.  251n.38 Lazenby, H.  192–3 Levelling down objection  194–5, 207n.19 Levy, N.  9n.25, 13, 15–16, 46–7 Lewis, D.  59n.12, 106, 251n.38, 48nn.27–28 Liability  82, 92–103, 110n.16, 143, 156–9 See also Consequential liability, Defensive liability Liberal equality  208 Lippert-Rasmussen, K.  18n.52, 167nn.13,16, 182n.65, 196n.87, 242n.20, 243n.21, 250n.36 Lottery facts  216–23, 240 Luck egalitarianism  19–20, 22–3, 82, 102, 126n.51, 163, 202–4, 206n.18, 212–13 See also Baseline-relative luck egalitarianism Luck-neutralizing dilemma  171, 179–80 Lucky Worlds  85–9 Maliciousness  4, 13, 17–18, 33–5, 42, 45, 48, 52–3, 56, 65–70, 72–4, 76, 78–80, 95, 103, 173, 255n.4 Manipulation 12–13 Manne, K.  137n.22 Manor, T.  188n.75 ‘Marginal’ persons  24–5, 248–9 Marmor, A.  8n.24, 37n.13, 49n.31, 59, 63 Marx, K.  167n.15 Mason, E.  11nn.28,29 Maximin  207, 208n.25

284 Index McGinn, C.  60n.13 MacIntyre, A.  154n.57 McMahan, J.  99nn.18,20, 143n.39, 250n.37, 252n.41 Means-end reasoning  150n.51 Means and ends  149–51 Mele, A.  9n.25 Metric of justice  92 See also Currency of justice Mildly Malicious Day’s End (case)  80 Mill, J. S.  165–6 Miller, D.  177–9, 180n.62, 182n.63, 192n.78 Minority Report (film)  112n.19 Misfortune  110, 113–14, 135, 247, 250–2 Modal Account (of luck)  3–4, 72n.29 Modified Absence of Regret Interpretation 147–53 Modus vivendi 226 Moral arbitrariness  23–4, 167, 201, 202–40, 246–7, 249–50 Moral backsliding  154 Moral Cost Argument  82, 95–103 Moral individualism  250–2 Moral personality  24, 129, 146, 206, 239–43, 246–7, 252–3 Moral track records  12, 17, 24, 32, 62, 90, 262 Morality system  131, 133–7, 140, 145–6, 150–1, 154, 156 Mulhall, S.  112n.19 Mulkeen, N.  167n.15, 168n.24 Nagel, T.  1–3, 5–9, 19, 22, 29, 29n.1, 103–4, 106n.4, 107, 131–3, 138–9, 147, 156, 206n.17, 227n.66, 236–7, 240n.15 Napoleon, B.  256 Nathan, C.  242n.20, 243n.21, 250n.36 Natural assets  207–8, 211–13, 217, 222–3, 246 See also Natural endowments Natural endowments  7, 173, 206–8, 214, 217–18, 221–2, 224–5, 227, 229 See also Endowments, Endowments facts, Natural assets Negligence  4, 17–18, 33, 48, 56, 58, 73, 95, 255n.4 Nelkin, D. K.  36n.11, 37n.13 Neutrality Claim  109–10, 112, 115 Neutralization Interpretation  203–5, 209–17, 219n.50, 224–5, 231, 239–40, 248n.30 New, C.  112n.19 No-Difference Claim  125–9, 261–2, 264–7 No-Priority Model  156–9 Non-Applicability Claim  216–20, 222–3, 225–6, 238–40 Non-comparative justice  39 Non-cosmopolitanism  23–4, 236–41

Non-Displacement Claim  109, 111–13, 115 Non-foreseeability 79–81 Non-identity problem  151n.53 Non-Moral Value Interpretation  137–9, 142 Non-Zero-Sum Claim  109, 112, 115 Normative Focus Problem  196–8 Normative magnets  245 Nozick, R.  165–6, 176n.49, 186, 205–6, 210–13, 210n.29 Nussbaum, M. C.  134n.9, 247n.28, 252n.43 Olsaretti, S.  164n.2, 168n.23 Opportunity Claim  109, 111, 115 Optimistic bias  256–9 Option luck  169, 178n.56, 182n.65, 198 Original position  206–10, 212, 225–6, 228–30, 232, 234, 240–2, 246, 252 Otsuka, M.  59–61, 80n.36, 182n.65 Ought Implies Can  36, 55, 74, 94–5, 110n.16, 133–4 Outcome luck: see Resultant luck Outsiders  23–4, 234, 236–40 Pairwise comparisons  19–23, 115, 126, 148, 163, 166–7, 184–6, 189, 196–9, 203 Paradox of the Baseline  187–91, 196 Parfit, D.  151n.53, 195n.85, 219–20, 222 Parity Requirement  36–40 Parry, J.  99n.23 Partiality Problem  192–4, 196 Paul, L. A.  152n.56 Peels, R.  2–3, 4n.16, 35n.10, 122n.41 Pluralist manoeuvre  194 Pluralism Problem  194–6 Pogge, T.  221n.59, 236n.9 Poison Assassins (case)  78–9 Pond (case)  108–15 Praiseworthiness  11, 19n.53, 93–5, 100–2, 107–11, 114–15 See also Good Cases Preda, A.  176n.48 Primary duties  75–8 Principle of redress  215–17 Pritchard, D.  2n.6, 3–4, 72n.29 Pro Tanto Problem  194–5 Pure comparative luck: see Comparative luck Pure interpersonal luck: see Interpersonal luck Pure Luck Interpretation  256–8, 256n.6 Putnam, H.  64n.23 Quality of will  36–7, 46–7, 90, 100–1 Rachels, J.  250n.37 Racism  24–5, 249, 252 Railton, P.  31n.7

Index  285 Range properties  241–3, 246–8 Rawls, J.  7, 23–4, 47n.25, 89n.7, 141–2, 163–4, 167, 194n.84, 201–26, 228–36, 238–48, 252–4 Raz, J.  30n.4, 91n.9, 143–4, 150n.51 Real interests  155 Reciprocity 237 Recklessness  6, 17–18, 31–3, 35, 42–3, 45–6, 50, 56, 72–8, 84, 86, 88–90, 95, 103, 122n.39, 152n.55, 157–9 Redundancy Challenge  204–5, 209 Redundancy View  202–4, 204n.12, 228–30 Regret  49–50, 131, 142–56, 197 See also Agent-regret Relational egalitarianism  196, 238 Relational properties  250–1 Relevance Claim  53–6, 58–60, 65–7, 71, 72n.29, 73–5, 77, 82–4, 90, 93–4, 109n.13 Responsibility as accountability: see Accountability Responsibility as answerability: see Accountability Responsibility as attributability: see Attributability Responsibility-for  123, 262–6 Responsibility tout court  122–3, 261–6 Restricted Franchise (case)  97–8 Restricted Luck-Sensitive Account  21–2, 33–6, 45–6, 49n.30, 50–2, 56–7, 59, 67, 69–70, 72, 75–6, 79–81, 85, 93, 95–7, 103–6, 131–3, 156–9, 255–6, 259 Resultant luck  5–7, 9–11, 17–18, 20–1, 29, 52, 82, 105–7, 115, 117, 119, 126–7, 170n.29, 184, 198–9, 255, 257–9 Retrospective justification  131, 140, 142, 146, 152n.55 Rousseau, J. J.  258n.10 RSVP (case)  112–13 Salow, B.  152n.55 Sandel, M.  205n.31, 206n.18, 210n.29 Sangiovanni, A.  237, 238n.14 Sartorio, C.  30n.4 Scalar variation  56n.4, 58, 241–3 Scanlon, T. M.  12, 14n.41, 17n.50, 36n.12, 196n.86 Scheffler, S.  169n.26, 196n.87, 199n.89, 202n.2, 215–16, 230n.67 Scheme of social cooperation  226–7 Second Single Two Buttons (case)  66 Secondary duties  75–9 Second principle (of Rawls’s justice as fairness)  205, 207, 231n.68 See also Difference principle Segall, S.  164n.2, 167n.13, 172–6, 192n.78

Self-disclosure  11–12, 15 Self-ownership  166, 212 Selfish bias  256–9 Self-Realization Interpretation  139–42, 147n.48 Self-validating changes  154–5 Seligman, M.  180n.59 Sen, A.  164n.5 Separate Duties Challenge  75–9 Sexism  24–5, 249, 252 Sher, G.  165n.6, 177n.52, 202n.4, 210n.30 Shields, L.  191n.77 Singer, P.  108n.11, 224n.62, 242n.20, 249–51 Single Assassin (case)  43 Single Driver (case)  43 Situational luck  7n.20, 21, 105, 106n.4, 115–16, 119 Smilansky, S.  112n.19, 186–91, 200n.90 Smith, A.  1, 20 Smith, A. M.  16–18, 18n.52, 12nn.34–35 Sober Driver (case)  31, 33, 49n.30, 156–9 Social endowments  7, 206–8, 216–17, 222, 228 See also Endowments, Endowments facts Social primary goods  205–7, 219–21, 225 Social system  199–201, 218, 223 Speciesism  24–5, 250–2 State of nature  226–7 Statman, D.  112n.19, 255nn.2,4, 257–9 Stemplowska, Z.  167n.13 Strawson, G.  129–30 Strawson, P. F.  14n.39, 36n.12 Strict Liability Account  30–4, 69, 96–7, 133n.4, 158–9 Strongly egalitarian theories  166–7, 177, 181 Sufficiency  165n.6, 191, 236 Surprise Party (case)  111–15 Sussman, D.  159n.63 Tadros, V.  79n.32 Taking Responsibility Strategy  48–51, 157n.62 Taurek, J.  94n.11 Telic equality  175n.44, 219–20, 222 Temkin, L.  39n.15, 102, 166n.12, 167, 172 Thomson, J. J.  6, 9–10, 10n.27, 19–20, 44n.21, 57–62, 79–80, 99n.22 Tognazzini, N. A.  18n.52 Todd, P.  14n.39, 18n.52 Tolstoy, L.  136–7, 139 Tomlin, P.  204n.9 Transformation Interpretation  151–3 Treating people as equals  37–40, 97–8, 165–7, 173–4, 201, 225–8, 241–8 See also Fairness Intuition

286 Index Two Better Buttons (case)  93–5 Two Buttons (case)  52–66, 53n.1, 68, 70–2, 74–5, 93–5, 100–1, 107 Two-Judgement Strategy  59–61 Unanimity 225 Underdetermination Problem  191–2, 194–6 Utilitarianism  165–6, 224, 250n.37 See also Consequentialism Valence  99n.19, 107, 220–4 Vallentyne, P.  182n.65 Valuable Self-Realization Interpretation 142–3 Van Inwagen, P.  7n.22 Veil of ignorance  206–7, 209, 212, 217, 224–5, 228, 232 Voigt, K.  164n.2 Volitionism 15

Voluntariness and involuntariness  145–6, 149, 167–9, 188, 196, 237, 253 Waldron, J.  241n.18, 242n.19, 250n.36, 243nn.21–22 Walker, M. U.  48–9 Wallace, R. J.  150n.50, 151n.53 Watson, G.  11–15 Weakly egalitarian theories  166–7 Williams, A.  192n.78 Williams, B.  1–4, 22, 31n.5, 49n.30, 131–59, 241n.18, 243n.21, 247n.29, 248, 249n.32 Wolf, S.  11–14, 48–9, 85n.2, 129n.54, 140n.32 Wolff, J.  164n.2, 194n.82, 196n.86 Young, I. M.  200 Zimmerman, M.  7n.20, 10n.27, 12n.31, 29n.3, 35, 37n.13, 53n.1, 55n.2, 83–4, 88n.5, 105, 106n.4, 115–30, 261