Principles and Persons: The Legacy of Derek Parfit 9780192893994

Derek Parfit, who died in 2017, is widely believed to have been the most significant moral philosopher in well over a ce

361 35 3MB

English Pages [471] Year 2020

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Contents
Introduction
Jeff McMahan
1
Special Concern and Personal Identity
David O. Brink
2
Separating Persons
James Goodrich
3
Personal Identity and Impersonal Ethics
Tim Campbell
4
Temporal Neutrality and the Bias Toward
the Future
Samuel Scheffler
5
What Is the Opposite of Well-Being?
Shelly Kagan
6
Parfit on Love and Partiality
Roger Crisp
7 Individualist Utilitarianism and Converging Theories of Rights Elizabeth Ashford
8
Parfit’s Reorientation
From Revisionism to Conciliationalism
Ingmar Persson
9
Parfit’s Final Arguments in Normative
Ethics
Brad Hooker
10 Parfit on Act ConsequentialismKatarzyna de Lazari-Radek and Peter Singer
11
Nonlegislative Justification
Liam Murphy
12
Doing Right by Wrong
Stephen Darwall
13
Giving Reasons and Given Reasons
John Broome
14
Reply to Parfit’s ‘Innumerate Ethics’
John Taurek
15
Defence Against Parfit’s Torturers
Jeff McMahan
16
Overdetermination and Obligation
Victor Tadros
17 What Is Harming?Molly Gardner
18
Prioritarianism, Risk, and the Gap Between
Prudence and Morality
Nils Holtug
19
Relational Egalitarianism
Telic and Deontic
Kasper Lippert-Rasmussen
20
Duties That Become Supererogatory
or Forbidden?
F. M. Kamm
21
More Supererogatory
Thomas Hurka and Evangeline Tsagarakis
Recommend Papers

Principles and Persons: The Legacy of Derek Parfit
 9780192893994

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

EDITED BY

Jeff McMahan Tim Campbell James Goodrich Ketan Ramakrishnan

OXFORD

Principles and Persons The Legacy of Derek Parfit Edited by

JEFF M CMAHAN TIM CAMPBELL JAMES GOODRICH and KETAN RAMAKRISHNAN

OXFORD UN IVERSI TY PRESS

OXFORD UNIVERSITY PRESS

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It fu rthers 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 ©Oxford University Press 2021 The moral rights of the authors have been asserted First Edition published in 2021 Impression: I 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 reprograph ics 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: 202 1932109 ISBN 978-0-19-289399-4 DOI: I 0.1093/oso/9780192893994.001.000 I Printed and bound in the UK by T) 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.

Contents

vii

Contributors

Introduction Jeff McMahan

1

I. PERSONAL ID ENTITY, PRUDENCE , AND ETHI CS 1. Special Concern and Personal Identity David 0. Brink

15

2. Separating Persons James Goodrich

39

3. Personal Identity and Impersonal Ethics Tim Campbell

55

4. Temporal Neutrality and the Bias Toward the Future Samuel Scheffler

85

5. What Is the Opposite of Well-Being? Shelly Kagan

115

6. Parfit on Love and Partiality Roger Crisp

151

II. NORMATIVE ETHICAL THEORY 7. Individualist Utilitarianism and Converging Theories of Rights Elizabeth Ashford

169

8. Parfit's Reorientation: From Revisionism to Conciliationalism Ingmar Persson

195

9. Parfit's Final Arguments in Normative Ethics Brad Hooker

213

10. Parfit on Act Consequentialism Katarzyna de Lazari-Radek and Peter Singer

233

11. Nonlegislative Justification LiamMurphy

247

vi

CONTENTS

III. REASONS 12. Doing Right by Wrong

277

Stephen Darwall 13. Giving Reasons and Given Reasons

299

John Broome IV. MORAL MATHEMATICS 14. Reply to Parfit's 'Innumerate Ethics'

311

John Taurek 15. Defence Against Parfit's Torturers

323

Jeff McMahan 16. Overdetermination and Obligation

355

Victor Tadros 17. What Is Harming?

381

Molly Gardner V. EGALITARIAN I SM AND PRIORITARIANISM 18. Prioritarianism, Risk, and the Gap Between Prudence and Morality

399

Nils Holtug 19. Relational Egalitarianism: Telic and Deontic

417

Kasper Lippert-Rasmussen VI. SUPEREROGATION 20. Duties That Become Supererogatory or Forbidden? F. M. Kamm

441

21. More Supererogatory

463

Thomas Hurka and Evangeline Tsagarakis Index

479

Introduction Jeff McMahan

When Derek Parfit died early in January 2017, he had for years been widely regarded as the best living moral philosopher. Many regarded him as the best moral philosopher since Sidgwick (whom he described as ‘my great, drab hero’), and I know several distinguished philosophers who believe him to be the best moral philosopher since Kant. He was also greatly beloved by his many former students and colleagues from his years at Oxford, Harvard, New York University (NYU), Princeton, and Rutgers. When he had to be hospitalized in New Jersey in 2014 after coming close to dying, the philosophers who appeared at his bedside were so numerous that a nurse was moved to exclaim, ‘Jesus Christ had only twelve disciples, but look at you!’ He was also held in great affection by the many beneficiaries of his legendary generosity in commenting on unpublished philosophical manuscripts. Several philosophers have expressed amazement at having received comments from him that were lengthier than the manuscript itself, often only a day or two after sending it to him. In at least one instance, his comments on a book manuscript were nearly as long as the book itself—though I assume that on this occasion the comments took more than a day or two to prepare. Although Parfit was well known, admired, and indeed revered in the world of academic philosophy, he was little known outside that small world. He worked obsessively but in comparative obscurity, seldom giving public lectures and almost never writing for popular media. He never sought celebrity; nor did it ever find him. While some philosophers have been awarded dozens of honorary doctorates, Parfit never received a single one. Near the end of his life, he was awarded the prestigious Rolf Schock Prize in Logic and Philosophy; but apart from that, his nominations for the other major prizes for which moral and political philosophers are eligible (Kyoto, Templeton, Berggruen . . .) were repeatedly passed over by the selection panels. Perhaps most surprisingly, there was never a Festschrift, or liber amicorum, published in his honour during his lifetime—not even on the occasion of his mandatory retirement from Oxford at the age of 67 in 2010. Nor were there even plans for such a volume when he died unexpectedly seven years later.

Jeff McMahan, Introduction In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0001

2  Jeff M c Mahan There have, of course, been various edited volumes, monographs, and special issues of journals devoted to examinations of Parfit’s work.1 Although many of these did not appear until after his death, he had, happily, seen the material in most of them. These many published discussions help to explain why his students and friends had not prepared a Festschrift. Because there was so much active discussion of his ideas, and because he remained intensely engaged and productive and seemed to everyone, including his doctors, to be in good health, his students and friends reasonably assumed that there was no urgency about publishing a volume of essays in his honour. But we were wrong. In December 2017, the philosophy departments at New York University and Rutgers University jointly held a two-­day conference in Parfit’s memory at which eight papers were presented. In coordination with the organizers of that conference, and with the generous and able assistance of Joseph Carlsmith and Ketan Ramakrishnan, I organized a parallel conference in Oxford. This conference, which was held in May 2018, was larger in scale, spanning three days and featuring twenty-­three speakers who all presented full-­length papers. I planned the Oxford conference with the explicit aim of publishing revised versions of selected papers in an edited collection in Parfit’s honour—the long overdue and sadly posthumous Festschrift.2 The project has, however, become more ambitious since then. The papers from the Oxford conference alone were almost too many for a single volume. And six of the eight speakers at the Rutgers–NYU conference wanted to submit papers for inclusion as well. There were, moreover, a number of other moral philosophers who were unable to present a paper at either conference but wanted to pay tribute to Parfit in some public way and have therefore written essays specifically for the Festschrift. Still others wanted to contribute a piece of a more biographical nature. The material I have gathered has thus increased well beyond the limits of the single volume I had initially envisaged. This book is therefore only the first of three volumes that will appear as memorials to Derek Parfit: two collections of philosophical essays and one volume of intellectual biography and memoir. In inviting speakers to the Oxford conference, I explicitly urged them not to assume that they should restrict their presentations to commentaries on Parfit’s 1  For example, Symposium on Derek Parfit’s Reasons and Persons, Ethics 96, no. 4 (1986); Jonathan Dancy (ed.), Reading Parfit (Oxford: Wiley, 1997); Jussi Suikkanen (ed.), Essays on Derek Parfit’s On What Matters (Oxford: Wiley-­Blackwell, 2009); Peter Singer (ed.), Does Anything Really Matter?: Essays on Parfit on Objectivity (Oxford: Oxford University Press, 2017); Simon Kirchin (ed.), Reading Parfit on What Matters (London: Routledge, 2017); Matthias Hoesch, Sebastian Muders, and Markus Rüther (eds.), Worauf es ankommt: Derek Parfits praktische Philosophie in der Diskussion (Hamburg: Felix Miener Verlag, 2017); Matthias Hoesch, Sebastian Muders, and Markus Rüther (eds.), Derek Parfit: Personen, Normativität, Moral: Ausgewählte Aufsätze (Berlin: Suhrkamp Verlag, 2017); Husain Sarkar, Kant and Parfit: The Groundwork of Morals (London: Routledge, 2018). 2  I am deeply grateful to several institutions and one individual for generous financial support for the Oxford conference: Corpus Christi College, All Souls College, Oxford University’s John Fell Fund, and Ludêk Sekyra.

Introduction  3 work. While I did not discourage the writing of critical or exegetical work, which was also welcome, I did encourage invitees to address and attempt to advance our understanding of issues that Parfit believed really matter. The result is that, while many of the philosophical essays in this and the forthcoming second volume do in part analyse, criticize, or build on Parfit’s ideas and arguments, all are directly engaged with issues in moral philosophy on which Parfit worked, and all seek to make progress with those issues. They are continuations of his efforts to understand the issues and they carry forward his legacy. In recent years several edited collections have been published in English that contain essays that comment on the first two volumes of the major work of Parfit’s later years, On What Matters.3 As these two volumes address issues primarily in metaethics and ethical theory, the essays in the edited collections naturally address the same issues. By contrast, the essays in this book and in the sequel are not much concerned with metaethics. There are various reasons for this, one being that the metaethical issues that were of most concern to Parfit have already been well and thoroughly discussed in the various other edited collections. Another is a matter of geography. Speakers for the larger conference in Oxford were drawn largely from friends and students of Parfit’s on this side of the Atlantic, who tended to be concentrated in Oxford and Scandinavia and who, for whatever reason, have in general worked more in normative ethics and population ethics than in metaethics. While some of these philosophers chose to write about issues raised by Parfit’s discussions of Kantianism, contractualism, and consequentialism in volume i of On What Matters, most chose to write on issues that Parfit addressed in Reasons and Persons or in papers published before On What Matters. Almost half of the contributors of philosophical essays chose to write on topics in population ethics, the area of ethics concerned with issues involving causing people to exist. Although the origin of population ethics as a distinct area of moral philosophy can be traced to a few seminal articles in the late 1960s and early 1970s, the locus classicus of the range of difficult problems and paradoxes in the area is part IV of Reasons and Persons, which remains the best and most important single work in the field.4 The second of the three memorial volumes will be devoted entirely to essays in population ethics. The essays in this first volume are more diverse. Most address issues that Parfit discussed in parts I, II, and III of Reasons and Persons, such as personal identity, the basis of rational prudential concern about the future, self-­interest and time, the nature of well-­being, special relations and partiality, the aggregation of value, the nature of harm, and individual responsibility for causally overdetermined 3  Cited in n. 1. 4 Perhaps the earliest essay is Jan Narveson, ‘Utilitarianism and New Generations’, Mind 76 (1967): 62–72.

4  Jeff M c Mahan outcomes. Others, as I noted above, discuss the major ethical theories that Parfit argued, in volume i of On What Matters, substantially converge in their implications when they are understood in their most defensible forms. The remaining essays discuss principles and issues that Parfit discussed in essays or in various places in his writings, such as equality, prioritarianism, supererogation, and the nature of reasons. Rather than offer my own summaries of the essays, I have appended the authors’ own abstracts below. With one exception, the essays in this first volume were written either for one of the conferences or specifically for inclusion here. Also with one exception, all are published here for the first time. The one essay that has appeared elsewhere is that by de Lazari-­Radek and Singer. Although it was written for the Rutgers–NYU conference, it has recently been published in a journal.5 The essay that was not written as a tribute to Parfit is that by John Taurek. This was written several decades ago. In 1977, Taurek published a celebrated paper called ‘Should the Numbers Count?’.6 This paper was perhaps the principal catalyst for discussions in contemporary moral philosophy of issues of aggregation and in particular of the relevance to the evaluation of an act of the number of people harmed or benefited by it. The position Taurek’s paper defended is at one end of the spectrum of possible views and is thus quite controversial. In 1978 Parfit published an influential response.7 Taurek wrote and presented a reply but soon abandoned philosophy and, to the best of my knowledge, never sought to publish this reply. He died in 2003 and has subsequently become something of a cult figure in philosophy. Frances Kamm preserved a typescript of Taurek’s reply and kindly offered to allow me to include it in this book. I have subsequently found a second copy of Taurek’s typescript among Parfit’s papers. I am most grateful to Taurek’s daughter, Davida Taurek, for granting me her permission to publish it and also for securing the permission of her siblings. I am confident that the appearance, after so many years, of Taurek’s response to Parfit’s criticisms of his view will be a source of excitement and deep interest to many philosophers. Most of the contributors to this book were once Parfit’s students or friends. All have endeavoured to engage with issues that he believed matter and to produce work worthy of serving as a tribute to him, though in the knowledge that each of our pieces could have been better with the benefit of his comments, however hastily written.8

5 Katarzyna de Lazari-­ Radek and Peter Singer, ‘Parfit on Act-­ Consequentialism’, Utilitas 32 (2020): 416–26. 6  John Taurek, ‘Should the Numbers Count?’, Philosophy & Public Affairs 6 (1977): 293–316. 7  Derek Parfit, ‘Innumerate Ethics’, Philosophy & Public Affairs 7 (1978): 285–301. 8  I am extremely grateful to Shixiang (Kida) Lin for his excellent work in preparing the index for this book.

Introduction  5

Chapter Abstracts David O. Brink: ‘Special Concern and Personal Identity’ As discussed by John Locke, Joseph Butler, and Thomas Reid, prudence involves a special concern for the agent’s own personal good that she does not have for others. This should be a concern for the agent’s overall good that is temporally neutral and involves an equal concern for all parts of her life. In this way, prudence involves a combination of agent relativity and temporal neutrality. This asymmetrical treatment of matters of interpersonal and intertemporal distribution might seem arbitrary. Henry Sidgwick raised this worry, and Thomas Nagel and Derek Parfit have endorsed it as reflecting the instability of prudence and related doctrines such as egoism and the self-interest theory. However, Sidgwick thought that the worry was unanswerable only for skeptics about personal identity, such as David Hume. Sidgwick thought that one could defend prudence by appeal to realism about personal identity and a compensation principle. This is one way in which special concern and prudence presuppose personal identity. However, as Jennifer Whiting has argued, special concern displayed in positive affective regard for one’s future and personal planning and investment is arguably partly constitutive of personal identity, at least on a plausible psychological reductionist conception of personal identity. After explaining both conceptions of the relation between special concern and personal identity, the chapter concludes by exploring what might seem to be the paradoxical character of conjoining them, suggesting that there may be no explanatory priority between the concepts of special concern and personal identity. James Goodrich: ‘Separating Persons’ In Reasons and Persons, Derek Parfit argues for a reductionist view of persons and that our ethical thinking should become more impersonal. While doing so, he argues that we may need to give up some widely shared intuitions about the Separateness of Persons and all of those views which crucially hinge upon it. However, this chapter argues that Parfit was mistaken. His reductionist views of persons and his more general claim that our ethical thinking should become more impersonal are in fact compatible with several plausible interpretations of the Separateness of Persons. Parfit’s project in Reasons and Persons should thus be understood not as undermining the Separateness of Persons, but as transforming our understanding of it. The chapter closes by considering the degree to which Parfit had reason by his own lights to accept some version of the Separateness of Persons. Tim Campbell: ‘Personal Identity and Impersonal Ethics’ On the Reductionist View, the fact of a person’s existence and that of her identity over time just consist in the holding of certain more particular facts about

6  Jeff M c Mahan physical and mental events and the relations between these events. These more particular facts are impersonal—they do not presuppose or entail the existence of any person or mental subject. In Reasons and Persons, Derek Parfit claims that if the Reductionist View is true, then ‘it is . . . more plausible to focus, not on persons, but on experiences, and to claim that what matters morally is the nature of these experiences’. But why think that the Reductionist View has this implication? As critics such as Robert Adams, David Brink, Mark Johnston, Christine Korsgaard, and Susan Wolf have suggested, it is not clear why the Reductionist View should have any implications regarding the moral importance of persons. This chapter argues that in contrast to Non-reductionist views, Psychological Reductionism, a version of the Reductionist View that assumes a psychological criterion of personal identity, supports the kind of impersonal moral outlook that Parfit describes. Samuel Scheffler: ‘Temporal Neutrality and the Bias Toward the Future’ Many philosophers have held that rationality requires one to have an equal concern for all parts of one’s life. In the view of these philosophers, temporal neutrality is a requirement of rationality. Yet Derek Parfit has argued that most of us are not, in fact, temporally neutral. We exhibit a robust bias toward the future. Parfit maintains that this future-bias is bad for us, and that our lives would go better if we were temporally neutral. Like other neutralists, he also believes that the bias is irrational, however widespread and robust it may be. This article assesses these criticisms and offers a qualified defense of the bias toward the future. Shelly Kagan: ‘What Is the Opposite of Well-­Being?’ Typically, discussions of the nature of well-being focus only on the positive elem­ents—those things that directly constitute someone’s being well off or better off. But an adequate theory of well-being also needs to give an account of ill-being, the negative elements that directly constitute being badly off, or worse off. This chapter asks how to extend a particular nonstandard theory of well-being— according to which well-being consists in the enjoyment of objective goods—so as to cover ill-being as well. In effect, then, it tries to discover the opposite of well-being, according to this nonstandard theory. This chapter tries to answer the question: what is ill-being when well-being is enjoying the good? Graphs are used throughout to illustrate the alternative possibilities, and to help display the surprising complexity of the most plausible answers. Roger Crisp: ‘Parfit on Love and Partiality’ It is generally held that in his 1984 book Reasons and Persons Derek Parfit was advocating greater impartiality in ethics. In his later work, On What Matters, he seems more inclined to accept that we have partial reasons, for example, to give priority to those we love. This chapter raises some questions concerning Parfit’s

Introduction  7 arguments for partiality, including whether affection is too contingent to be ­valuable in itself, and whether partial concern for others, shared histories, or commitments can plausibly be said to ground non-instrumental reasons or value. The paper ends with a discussion of gratitude and an argument based on Parfit’s reductionist conception of personal identity. Elizabeth Ashford: ‘Individualist Utilitarianism and Converging Theories of Rights’ The paper develops two core themes of Derek Parfit’s philosophy. The first is his goal of unifying the two main rival impartial moral theories, Kantian deontology and consequentialism, therefore reinforcing their claim to pertain to objective moral truths. The second is his focus on the moral significance of the combined effects of many agents’ behaviour, and on the challenges this poses to ordinary moral thinking. This is a theme that runs throughout his work, that he returns to at the very end of volume iii of On What Matters. Kantianism and consequentialism have been thought to fundamentally diverge on the issue of rights and tradeoffs. The chapter first outlines the version of consequentialism taken to be most plausible, calling it ‘individualist utilitarianism’, which differs from so-called ‘classical utilitarianism’ in taking the moral importance of well-being to be grounded on the moral importance of the persons whose well-being it is. This paves the way for a pluralist Kantian and utilitarian account of human rights, grounded on the moral significance both of persons’ well-being and their dignity as rational au­tono­mous agents. The chapter then turns to the topic of the threat to access to the means of subsistence, both for the current poor and future generations, posed by global as well as domestic socio-economic structures and anthropogenic climate change. This harm is the combined effect of the ongoing patterns of behaviour of a vast number of agents. The chapter argues that individualist utilitarianism and Kantianism converge on the conclusion that the duty to avoid harms of this kind should be analysed as a shared duty of basic justice, non-fulfilment of which constitutes a structural human rights violation. Ingmar Persson: ‘Parfit’s Reorientation: From Revisionism to Conciliationism’ This paper aims to show that between Reasons and Persons and On What Matters the orientation of Derek Parfit’s philosophy underwent a significant change. The approach of Reasons and Persons is largely revisionist, which is exemplified by his reductionist account of personal identity. This account is omitted in On What Matters apparently because it does not fit in with the conciliationist project of this work. The aim of the first two volumes of that work is to show that, on the basis of a non-naturalist theory of normative reasons, three supposedly irreconcilable moral theories—rule-consequentialism, Kantian and Scanlonian contractualism—could converge to form a single Triple Theory. In the third volume, the conciliationist approach is carried further by Parfit’s attempt

8  Jeff M c Mahan to show both that his metaethical position is in essential agreement with rivals, such as Gibbard’s expressivism, and to reconcile parts of common-sense morality and consequentialism in order to bring them together in the Triple Theory. This chapter argues that the failure of these attempts as well as the fact that the most controversial revisionist claims in Reasons and Persons are left out throw doubt on the feasibility of Parfit’s conciliationalist undertaking. Brad Hooker: ‘Parfit’s Final Arguments in Normative Ethics’ This paper starts by juxtaposing the normative ethics in the final part of Parfit’s final book, On What Matters, volume iii (2017), with the normative ethics in his earlier books, Reasons and Persons (1984) and On What Matters, volume i (2011). The paper then addresses three questions. The first is, where does the reflectiveequilibrium methodology that Parfit endorsed in the first volume of On What Matters lead? The second is, is the Act-involving Act Consequentialism that Parfit considers in the final volume of On What Matters as plausible as Rossian deontology? The third is, how is the new argument that Parfit puts forward for Rule Consequentialism supposed to work? Katarzyna de Lazari-­Radek and Peter Singer: ‘Parfit on Act Consequentialism’ In the first two volumes of On What Matters, Derek Parfit argues that three major normative theories—Kantianism, Contractualism, and Consequentialism—are, in their most defensible forms, compatible, and can be reconciled as a ‘Triple Theory’. The form of Consequentialism that Parfit argues is compatible with Kantianism and Contractualism is Rule Consequentialism. This has led many to assume that Parfit does not believe that Act Consequentialism is a defensible form of Consequentialism. We draw on personal correspondence to show that this assumption is incorrect. We then consider how, in On What Matters, volume iii, which Parfit completed shortly before his death, he seeks to narrow the differences between Act Consequentialism and the Triple Theory. One of the ways in which he does this is to suggest that Impartial Rationality may be an external rival to Morality, in much the same way as egoism is an external rival to morality. It is argued that this move undermines morality, as shown by Parfit’s own example of the judgements that we may make in the case of terror bombing. We conclude that Parfit’s attempts to bridge the gap between Act Consequentialism and Triple Theory meet with only limited success. Liam Murphy: ‘Nonlegislative Justification’ If moral theorists who otherwise disagree, all approach moral theorizing as a search for a set of desirable moral principles for the general regulation of behavior, then there is a sense in which they are all, as Parfit says, climbing the same mountain. But it is the wrong mountain. Morality should not be understood as hypothetical legislation; it is a mistake to set about constructing morality as if we

Introduction  9 were making law. Real legislators evaluate possible legal rules by considering the effects they would have. They can do this because enforcement and acceptance of law ensure a high level of compliance. Moral legislators have no reason to assume any particular level of acceptance; the effects of counterfactual acceptance of a principle are not morally relevant. The argument targets rule consequentialism and Scanlon’s official version of contractualism. The paper begins in a positive mode by arguing that a nonlegislative version of Scanlon’s approach, that seeks justification for conduct of such-and-such a kind in such-and-such circumstances by comparing the reasons in favor and the reasons others have to object, is a very attractive way to think about what we owe to each other. Stephen Darwall: ‘Doing Right by Wrong’ A striking contrast between  Reasons and Persons and  On What Matters is the vastly different attitude Parfit takes towards Act Consequentialism. Parfit’s defense of Act Consequentialism against a battery of criticisms in  Reasons and Persons  was legendary. In  On What Matters, however,  Parfit  remarks that Sidgwick’s act-consequentialist principle of rational benevolence is best regarded, like egoism, as an ‘external rival to morality’. What lies behind this remarkable change in attitude, if not in view, is Parfit’s focus in On What Matters on deontic moral concepts, like wrongness, and their relation to accountability and reactive attitudes like moral blame. This essay explores the details of Parfit’s later views, arguing that he did not go far enough in pursuing this line of thought and that doing so is necessary to bring out the distinctive normativity of deontic moral concepts. Parfit’s claim that the ‘ordinary’ concept of wrongness is indefinable threatens to rob the concept of normativity in the ‘reason-involving sense’. If, however, we understand wrongness in terms of there being reason to blame, lacking excuse, we can account for its distinctive normative contours. John Broome: ‘Giving Reasons and Given Reasons’ Derek Parfit, as a leader of the ‘reasons first’ movement, says that the concept of a reason is fundamental and indefinable. But his concept of a reason differs from most philosophers’. Most philosophers take a reason to be a fact, whereas Parfit says that reasons are given by facts, not that they are facts. This paper distinguishes Parfit’s concept of a reason, which it calls a ‘given reason’, from the more common one, which it calls a ‘giving reason’. It argues that, whereas the concept of a giving reason is easily defined, the concept of a given reason is not. Parfit is therefore better placed than most philosophers to defend the claim that the concept of a reason is fundamental and indefinable. John Taurek: ‘Reply to Parfit’s ‘‘Innumerate Ethics” ’ This is the text of a presentation by Taurek that replies to Parfit’s ‘Innumerate Ethics’, which is itself a reply to Taurek’s ‘Should the Numbers Count?’.

10  Jeff M c Mahan Jeff McMahan: ‘Defence Against Parfit’s Torturers’ In the literature on ‘moral mathematics’ prompted by the section with that title in Reasons and Persons, one issue is whether, and if so to what extent, it is wrong to cause a negligible harm to each of a large number of people, and in particular whether doing so could ever be as seriously wrong as causing a substantial harm to one person. The topic in this chapter will be the closely related issue of proportionality in defence against those who would inflict only such tiny harms, though on a large number of victims. For example, might a person who would otherwise inflict a tiny harm on each of a large number of people be liable to be killed in defence of those people? The chapter will suggest that such a person seems liable to be killed in some cases but not in others, depending on what other people might be doing or on other facts about the context in which the harms would occur. It will review a range of examples involving the infliction of tiny harms that reveal some surprising facts about the conditions and limits of liability to defensive harm. Victor Tadros: ‘Overdetermination and Obligation’ This chapter is concerned with circumstances where a person’s act makes no difference to the occurrence of a negative outcome, but is a member of a group of acts that does make such a difference. In the light of an analysis of these circumstances, it argues against two familiar ideas. One is Derek Parfit’s view that the wrongness of an act directly depends on the consequences of the group of acts of which it is a member. The other is the view that intentions are irrelevant to permissibility. The chapter suggests that wrongness and permissibility, in these cases, is distinguished by the intentions of those who act. It also argues that intentions make a difference to a person’s liability to punitive, compensatory, and defensive harm. Finally, it briefly considers cases involving mixed motives. Molly Gardner: ‘What is Harming?’ A complete theory of harming must have both a substantive component and a formal component. The substantive component, which Victor Tadros calls the ‘currency’ of harm, tells us what I interfere with when I harm you. The formal component, which Tadros calls the ‘measure’ of harm, tells us how the harm to you is related to my action. This chapter surveys the literature on both the currency and the measure of harm. It argues that the currency of harm is well-being and that the measure of harming is best captured by a causal account on which harming is causing a harm. A harm for you is the presence of something in­trin­ sic­al­ly bad for you or the absence of something intrinsically good for you. Thus, although a counterfactual account of the measure of harm need not distinguish between a harm and a harmful event, the causal account reserves the term ‘harm’, not for a harmful event, but only for its effect. Finally, the chapter shows how a complete theory of harming can help us to answer questions about whether we

Introduction  11 can harm people with speech, whether we can harm the dead, and how it is ­possible to harm future generations. Nils Holtug: ‘Prioritarianism, Risk, and the Gap Between Prudence and Morality’ According to a prominent objection to prioritarianism, it inappropriately implies a gap between prudence and morality, even in single-person cases. Thus, according to prioritarianism, we should sometimes sacrifice an individual’s expected welfare in order to protect her from the risk of a worse outcome. The present chapter presents a critical discussion of this objection. It first provides a more precise account of axiological prioritarianism and what it implies for the relation between prudence and morality. Then it provides an account of four prioritarian theories that (unlike axiological prioritarianism) have implications for risky choices, namely ex ante prioritarianism, ex post prioritarianism, pluralist prioritarianism, and factualist prioritarianism. It then presents the objection that prioritarianism implies a gap between prudence and morality in single-person cases in greater detail, which includes explaining the extent to which this objection applies to the four different versions of prioritarianism mentioned above. Finally, the chapter defends the view that the prioritarian gap between prudence and morality is unproblematic, even in single-person cases. Kasper Lippert-­Rasmussen: ‘Relational Egalitarianism: Telic and Deontic’ Derek Parfit famously introduced a now commonly adopted distinction between telic and deontic distributive egalitarianism. This chapter argues that we can draw  a similar distinction between telic and deontic relational egalitarianism. Interestingly, telic relational egalitarianism might be less vulnerable to the levellingdown objection than telic distributive egalitarianism. However, while some relational egalitarian concerns are best captured by telic relational egalitarianism, other concerns are better captured by deontic relational egalitarianism and yet others relating to intergenerational justice are better captured by telic distributive egalitarianism. Accordingly, insofar as we are egalitarians, we should be pluralist egalitarians in a more thoroughgoing way than Parfit entertained. F. M. Kamm: ‘Duties That Become Supererogatory or Forbidden?’ This chapter first examines certain of Derek Parfit’s views in his On What Matters, volume iii on the relation between not harming, aiding, and making personal ­sacrifices to achieve each. It compares his views with those of the author on two different measures of the stringency of duties and the distinction between supererogation and obligation. The chapter goes on to consider implications of these views for cases in which an agent must choose whether to save many people by either (i) not saving or harming someone else or (ii) suffering some large ­personal loss himself. The chapter continues by arguing against one way in which Parfit thinks an agent-relative deontological conception of one’s duty

12  Jeff M c Mahan incorrectly bars our having common aims by requiring each person to minimize the harm he does. Thomas Hurka and Evangeline Tsagarakis: ‘More Supererogatory’ If acts can be supererogatory, presumably some can be more supererogatory than others, or further beyond the call of duty. This paper explains how this is possible within a general account of supererogation that sees it arising when a prima facie duty, for example to promote other people’s good, is outweighed by a prima facie permission to promote one’s own good. An act is then more supererogatory when the permission outweighs the duty by more, or when the gap between its strength and that of the duty’s is larger. The paper contrasts its permission-based account of supererogation with a more common one typified by Parfit in On What Matters, which rests it on a conflict between two ‘reasons’ that, despite their differing contents, are of the same deontic type and have the same favouring force. Alongside several other weaknesses, Parfit’s account doesn’t allow differing degrees of supererogation but must treat all supererogatory acts as on a par.

Bibliography Dancy, Jonathan (ed.), Reading Parfit (Oxford: Wiley, 1997). de Lazari-Radek, Katarzyna, and Singer, Peter, ‘Parfit on Act-Consequentialism’, Utilitas 32 (2020): 416–426. Hoesch, Matthias, Muders, Sebastian, and Rüther, Markus (eds.), Derek Parfit: Personen, Normativität, Moral: Ausgewählte Aufsätze (Berlin: Suhrkamp Verlag, 2017). Hoesch, Matthias,Muders, Sebastian, and Rüther, Markus (eds.), Worauf es ankommt: Derek Parfits praktische Philosophie in der Diskussion (Hamburg: Felix Miener Verlag, 2017). Kirchin, Simon (ed.), Reading Parfit on What Matters (London: Routledge, 2017). Narveson, Jan, ‘Utilitarianism and New Generations’, Mind 76 (1967): 62–72. Parfit, Derek, ‘Innumerate Ethics’, Philosophy & Public Affairs 7 (1978): 285–301. Sarkar, Husain, Kant and Parfit: The Groundwork of Morals (London: Routledge, 2018). Singer, Peter (ed.), Does Anything Really Matter?: Essays on Parfit on Objectivity (Oxford: Oxford University Press, 2017). Suikkanen, Jussi (ed.), Essays on Derek Parfit’s On What Matters (Oxford: ­Wiley-Blackwell, 2009). Symposium on Derek Parfit’s Reasons and Persons, Ethics 96/4 (1986). Taurek, John, ‘Should the Numbers Count?’, Philosophy & Public Affairs 6 (1977): 293–316.

1

Special Concern and Personal Identity David O. Brink

A traditional philosophical narrative links personal identity and the sort of special concern each person has for herself, including her own future. Prudential concern for one’s overall good is one expression of this kind of special concern. Despite their disagreements about the nature of personal identity, John Locke, Bishop Joseph Butler, and Thomas Reid embraced the forensic role of personal identity as essential to our understanding of both responsibility and special concern, especially prudence.1 For instance, in his discussion of personal identity in An Essay Concerning Human Understanding, Locke comments on the forensic role of personhood and personal identity. In this personal Identity is founded all the Right and Justice of Reward and Punishment; Happiness and Misery; being that, for which everyone is concerned for himself, not mattering what becomes of any Substance, not joined to, or affected with that consciousness.2 I dedicate this essay to the memory of Derek Parfit, who had a profound influence on me early in my philosophical career. I first encountered Derek in Michaelmas term in Oxford in 1982 when I was a visiting graduate student and attended lectures in which he was presenting material on personal identity and its normative significance, as he made final revisions to Reasons and Persons. I was struck by his intellectual intensity and creativity and his unwavering commitment to get things right. We reconnected after I took a position at MIT and Derek began visiting at Harvard. Over several years, we discussed issues about personal identity and its significance, the separateness of persons, and prioritarianism, and he provided generous feedback on my developing views. These discussions had a tremendous effect on me, shaping not just my views on these topics but also my philosophical sensibility. We continued to correspond for a couple of years after I moved to the University of California, San Diego (UCSD) in 1994. This essay engages themes about special concern and personal identity that we discussed many times. Ideas in this essay were also discussed in seminars at UCSD and in a Pea Soup symposium about temporal neutrality with Preston Greene, Meghan Sullivan, and Tom Dougherty. Earlier versions were presented at a Moral and Political Philosophy Seminar session at UCSD in 2016, colloquia at St Andrews and York Universities in 2017, and the 2017 Nanyang Technological University Conference on Neutrality: Agency, Time, and Reasons. In particular, I would like to thank Craig Agule, Dick Arneson, Saba Bazargan, Amy Berg, Matt Braich, Tim Campbell, Jonathan Cohen, Cory Davia, Tom Dougherty, Andrew Forcehimes, Preston Greene, Kathryn Joyce, J. P. Messina, Dana Nelkin, Christian Piller, Theron Pummer, Eric Watkins, Alan Thomas, and Danny Weltman for helpful comments on earlier versions. 1 J.  Locke, An Essay Concerning Human Understanding, ed. P.H.  Nidditch (Oxford: Clarendon Press, 1975), bk II, ch. xxvii, sects. 18, 26; J. Butler, ‘Dissertation on Personal Identity’, in The Whole Works of Joseph Butler, ed. S. Halifax (London: William Tegg & Co., 1847), 263–7; and T. Reid, Essays on the Intellectual Powers of Man, ed. B. Brody (Cambridge, MA: MIT Press, 1969). 2 Locke, Essay Concerning Human Understanding, II.xxvii.18. David O. Brink, Special Concern and Personal Identity In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0002

16  David O. Brink Locke included the sort of special concern characteristic of prudence as part of this forensic role. So, he thought that special concern for my own future was predicated on personal identity. Locke was what we would now call a psychological reductionist about personal identity and thought that his own conception of personal identity in terms of sameness of consciousness or memory explained this forensic role. In particular, he thought that I would and should extend special concern to any future person who would be memory-­connected with me now. By contrast, Butler was a non-­reductionist, who denied that any informative analysis of personal identity is possible. He too thought personal identity is tied to the forensic role of persons and special concern. But he rejected the Lockean conception of personal identity as committed to identity in only a ‘loose and popular’ sense, rather than a ‘strict and philosophical’ sense.3 Butler concludes that Lockean reductionism about personal identity cannot support the forensic role of persons. And from hence it must follow, that it is a fallacy upon ourselves, to charge our present selves with anything we did, or to imagine our present selves interested in anything which befell us yesterday, or that our present self will be interested in what will befall us to-­morrow; since our present self is not, in reality, the same with the self of yesterday, but another like self or person coming in its room, and mistaken for it; to which another self will succeed to-­morrow. . . . [F]or if the self or person of to-­day, and that of to-­morrow, are not the same, but only like persons, the person of to-­day is really no more interested in what will befall the person of to-­morrow, then in what will befall any other person.4

Here, Butler also assumes that special concern presupposes personal identity but denies that Lockean reductionism is an acceptable conception of numeric identity. Locke, Butler, and Reid agree that a prudential form of special concern depends on personal identity. Prudence involves a special concern for the agent’s own well-­being or personal good that she does not have for others. But this should be a concern for the agent’s overall good that is temporally neutral and involves an equal concern for all parts of her life. In this way, prudence involves a combination of personal bias and temporal neutrality. This asymmetrical treatment of matters of interpersonal and intertemporal distribution might seem arbitrary. Henry Sidgwick raised this worry, and Thomas Nagel and Derek Parfit have endorsed it as reflecting the instability of prudence and related doctrines such as egoism and

3  Butler, ‘Dissertation on Personal Identity’, 263–7. 4  Butler, ‘Dissertation on Personal Identity’, 267.

Special Concern and Personal Identity  17 the self-­interest theory.5 However, Sidgwick thought that the worry was unanswerable only for skeptics about personal identity, such as David Hume. Sidgwick thought that one could defend prudence by appeal to an ecumenical form of realism about personal identity. In this way, Sidgwick develops an idea common to Locke, Butler, and Reid that prudence presupposes personal identity. In particular, realism about personal identity and a compensation principle, which Sidgwick also discussed, jointly rationalize prudence’s combination of agent bias and temporal neutrality. While there is no automatic interpersonal compensation for sacrifice, diachronic intrapersonal compensation is automatic.6 However, another interpretation of the relationship between personal identity and special concern appears to stand this traditional conception of the relationship on its head. Jennifer Whiting has argued that the psychological reductionist about personal identity should recognize that special concern is an important ingredient in both interpersonal and intrapersonal unity.7 If we develop this argument, it seems that we should conclude that personal identity presupposes special concern. Though each interpretation of the relation between special concern and personal identity might seem attractive, it seems problematic to combine them, because this seems to affirm that special concern presupposes personal identity and that personal identity presupposes special concern. To avoid circularity, it seems we have to reject at least one of these conflicting narratives. However, I  want to try to reconcile them. Two philosophical concepts cannot both be explanatorily fundamental with respect to each other in the same respect. Nonetheless, we can treat them as equally fundamental, especially if we treat them as explanatorily fundamental with respect to each other in different respects. On this reading, with some qualifications, we might claim that personal identity is normatively prior to special concern but that special concern is metaphysically prior to personal identity. In what follows, I will first explain the demands of special concern and prudence (Section  1) and then explore a puzzling structural asymmetry within prudence that Sidgwick, Nagel, and Parfit have noticed (Section 2). Next, I will examine Sidgwick’s resolution of this puzzle, which appeals to realism about personal identity and compensation (Section  3). Then I will explore just how metaphysically ecumenical Sidgwick’s rationale for prudence is and whether it is compatible with the truth of a psychological reductionist conception of personal 5 H. Sidgwick, The Methods of Ethics (7th ed., London: Macmillan, 1907), esp. 418–19; T. Nagel, The Possibility of Altruism (Princeton: Princeton University Press, 1970), esp. 16, 19, 99–100; and D. Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), esp. 137–44. 6  In developing this rationale for prudence, I will be adapting arguments from D. Brink, ‘Rational Egoism and the Separateness of Persons’, in J. Dancy (ed.), Reading Parfit (Oxford: Blackwell, 1997) and D. Brink, ‘Prospects for Temporal Neutrality’, in C. Callender (ed.), The Oxford Handbook of the Philosophy of Time (Oxford: Clarendon Press, 2011). 7  J. Whiting, ‘Friends and Future Selves’, Philosophical Review 95 (1986): 547–80.

18  David O. Brink identity (Section  4). This allows us to clarify Sidgwick’s claim that prudence depends on personal identity; personal identity is prior to prudence, because personal identity and compensation justify prudence and temporal neutrality (Section 5). Next, I will explore and defend Whiting’s suggestion that the psychological reductionist should claim that special concern is in a way prior to personal identity, because it helps constitute intrapersonal unity (Section 6). While special concern helps constitute personal identity, it need not take a temporally neutral form to do so (Section 7). Having explained each of these two narratives about the relative priority of special concern and personal identity, I will address the apparent tension between them, arguing that the tension can be resolved by distinguishing metaphysical and normative priority (Section 8).

1.  Special Concern, Prudence, and Egoism Special concern, as it figures in the tradition of thinking about personal identity, involves a kind or degree of concern for oneself that one does not have for others. This is a concern for one’s overall good and essentially involves a concern for one’s own further future. One common expression of this sort of special concern is prudence and its temporally neutral concern with the agent’s own good. Both special concern and prudence are sometimes connected with rational egoism. Rational egoism is a theory about practical reason that claims that one has reason to do something just insofar as that would promote the agent’s overall interest or personal good. As such, egoism treats intertemporal and interpersonal distribution differently. A conception is neutral with respect to persons or time if it assigns no intrinsic significance to how goods and bads are distributed among persons or over time, except insofar as this affects the value of one’s life as a whole. By contrast, we might say, a conception is biased or relative if it does assign intrinsic significance to issues about whom a benefit or burden falls upon or when it does, independently of affecting the value of the life in which it occurs. Rational egoism is interpersonally biased or relative, but intertemporally neutral. In principle, a conception of practical reason might be interpersonally biased or relative in any number of ways. For instance, it might assign special normative significance to benefits and harms that accrue to those to whom the agent stands in special relationships.8 But rational egoism is biased in a special way insofar as it insists that goods and bads must accrue to the agent in some way to be of rational significance to her. Egoism itself does not specify the forms that this kind of benefit might take. For instance, I might benefit by benefiting another if the other will reciprocate my cooperation and assistance. In such cases, the egoist has 8  One such theory is Broad’s self-­referential altruism. See C. D. Broad, ‘Self and Others’, repr. in Broad’s Critical Essays in Moral Philosophy, ed. D. Cheney (London: George Allen & Unwin, 1971).

Special Concern and Personal Identity  19 instrumental reasons to be concerned about and benefit others. However, it’s possible that the egoist might have constitutive reason to be concerned about and benefit others if she stands in the right sort of relationship to them such that their good is a part of hers. We sometimes adopt this way of looking at one’s relationship to intimates—parents and their children, spouses, and close friends—when we view the good of another as part of one’s own good. We can say that the egoist recognizes non-­derivative reason for concern only for herself and derivate reason to be concerned for others. But this derivative concern for others might be more or less robust depending on whether that concern is purely instrumental or also constitutive. Interesting as these issues about the shape of egoist derivative concern for others are, I propose to ignore them as much as possible for present purposes.9 Whereas the egoist is interpersonally biased, she is temporally neutral. She assigns no intrinsic significance to the temporal location of goods and bads within her life, except insofar as doing so contributes to the value of her life.10 She should be concerned with her own overall good and be equally concerned about all parts of her life. This means that she should reject various kinds of temporal bias, including bias toward the near.11 Temporal neutrality requires sacrificing a nearer good for a later, greater good. Call this now‑for‑later sacrifice. This sort of sacrifice provides us with some of our most compelling paradigms of rationality. It seems a mark of rationality to undertake actions, projects, and commitments to which one would otherwise be indifferent or averse for the sake of some later, greater good. This kind of rational planning is ubiquitous. We may not notice its more mundane applications, such as when we stand in line to get tickets to a film, when we stop to refuel our cars, or when we go to the dentist for routine preventive dental care. We are more likely to recognize now-­for-­later sacrifice when the sacrifice is more significant. For instance, I engage in such sacrifice when I undergo a medical procedure that involves an extended and painful recovery in order to regain full range of motion and the ability to participate in a fuller range of physical activities than would otherwise be possible. The training 9  I discuss these issues at greater length in D. Brink, ‘Self-­Love and Altruism’, Social Philosophy & Policy 14 (1997): 122–57. 10  Two qualifications. (a) As Sidgwick notes (Methods of Ethics, 381), I may adopt an unequal concern for different periods in my life if I am a more efficient converter of resources into utility at certain points in my life. (b) If lives with certain kinds of narrative structure (e.g. an upward trajectory) are more valuable than others, then one might accept intrapersonal distributional principles and attach more normative significance to some periods in a person’s life than others. See D. Velleman, ‘Well-­ Being and Time’, Pacific Philosophical Quarterly 72 (1991): 48–77. However, both of these ways of attaching special significance to certain periods in one’s life are fully compatible with temporal neutrality, understood as insisting that the temporal location of goods and bads within one’s life has no normative significance except insofar as it contributes to the value of that life. 11  Bias toward the near in which proximate goods and bads are assigned normative significance out of proportion to their actual magnitude should be distinguished from bias toward the future in which goods and harms have normative significance prospectively that is greater than their significance retrospectively. My focus here will be on bias toward the near. I discuss bias toward the future briefly in Section 7.

20  David O. Brink required for success in many vocations and avocations often requires various non-­negligible physical, financial, and personal sacrifices. Provided the later benefits genuinely do outweigh the near-­term costs, the sacrifices seem rational and failure to persevere, if understandable, nevertheless seems to be a form of weakness. Indeed, the emergence of the ability to recognize the rationality of now-­for-­later sacrifice and to regulate one’s appetites, emotions, and actions in accordance with this recognition is arguably a significant part of the process of normative development that marks the progress from adolescence to responsibility and maturity. In this way, rational egoism combines agent bias and temporal neutrality. As I conceive of prudence, it is the egoist dimension of practical reason. Like egoism, prudence involves a special concern for one’s overall good that is agent-­biased and temporally neutral. Rational egoism says that this is all there is to practical reason. By contrast, as I conceive it, prudence does not preclude other kinds of demands of practical reason, for instance, demands for more impartial regard. Prudence might be just one form of practical reasonableness among others. My focus here is on special concern and prudence, especially its demand of temporal neutrality. For present purposes, I hope to remain agnostic about whether prudence should be part of a more pluralistic conception of practical reason or whether it is all there is to practical reason, focusing on the prospects of prudence, rather than rational egoism.

2.  Prudence and Asymmetrical Distribution Prudence combines interpersonal bias and intertemporal neutrality. As such, it provides an asymmetrical treatment of two distinct distributional dimensions. It says that it makes all the difference to whom benefits and burdens accrue and none whatsoever when they occur. On reflection, this may seem arbitrary. Why not treat these two distributional dimensions the same way? Sidgwick recognized this issue in The Methods of Ethics in the context of his discussion of the proof of utilitarianism. I do not see why the axiom of Prudence should not be questioned, when it conflicts with present inclination, on a ground similar to that on which Egoists refuse to admit the axiom of Rational Benevolence.  If the Utilitarian has to answer the question, ‘Why should I sacrifice my own happiness for the greater happiness of another?’ it must surely be admissible to ask the Egoist, ‘Why should I sacrifice a present pleasure for a greater one in the future? Why should I concern myself about my own future feelings any more than about the feelings of other persons?’12 12 Sidgwick, Methods of Ethics, 418.

Special Concern and Personal Identity  21 The egoist asks the utilitarian: Why should I sacrifice my own good for the good of another? The egoist doubts that concern for others is non-­derivatively rational. But one can ask the egoist on behalf of one’s present self: Why should I sacrifice a present good for myself for the sake of a future good for myself? Here, one might doubt that concern for one’s future is non-­derivatively rational. These doubts may seem parallel. We must decide where among lives and when within lives to locate goods and harms. Because both are matters of position or location, we may think that they should be treated the same. We might express this concern about prudence in terms of its hybrid structure, involving the asymmetrical treatment of interpersonal and intertemporal distribution. If we distinguish neutral and biased treatment and focus on biased treatment in favor of the agent or the present, we can contrast the hybrid structure of prudence with two purebred rivals (Table 1.1). These are structural conceptions of practical reason that make claims about the distribution of goods and bads across persons and times. They are agnostic about what the good consists in, as between hedonism, desire-­satisfaction, perfectionism, or objective list views. In principle, there are two purebreds—rational benevolence or neutralism and egoism of the present moment—and two hybrids—prudence or egoism and benevolence of the present moment. Benevolence of the present moment has had no champions (of whom I am aware) and holds little appeal, so I shall ignore it going forward, focusing on the contrast between the hybrid structure of prudence and the two purebreds.13 In chapter 7 of Reasons and Persons Parfit endorses Sidgwick’s worry about the hybrid structure of prudence or egoism, which he applies to what he calls the Self-­interest Theory (S). As a hybrid S can be attacked from both directions. And what S claims against one rival may be turned against it by the other. In rejecting Neutralism, a Self‑interest Theorist must claim that a reason may have force only for the agent. Table 1.1  Time   Persons

Neutrality Bias

Neutrality

Bias

Neutralism or Rational Benevolence Egoism or Prudence

Benevolence of the Present Moment Egoism of the Present Moment

13  It’s not clear that benevolence of the present moment has a plausible rationale. The rationale I will reconstruct for the hybrid character of prudence (Section 3) does not extend to this other hybrid conception.

22  David O. Brink But the grounds for this claim support a further claim. If a reason can have force only for the agent, it can have force for the agent only at the time of acting. The Self-­interest theorist must reject this claim. He must attack the notion of a time-­ relative reason. But arguments to show that reasons must be temporally neutral, thus refuting the Present-­aim Theory, may also show that reasons must be neutral between different people, thus refuting the Self-­interest Theory.14

If present sacrifice for future benefit is rational, why isn’t sacrifice of one person’s good for the sake of another’s? In this way, the appeal to parity may support rational benevolence. This is roughly the view Thomas Nagel adopts in The Possibility of Altruism. His primary aim is to argue against egoism’s agent bias and in favor of impartial­ity or altruism, and he relies on the parity of intertemporal and interpersonal distribution to do so. Just as the interests of an agent’s future self provide him with reasons for action now, so too, Nagel argues, the interests of ­others can provide him with reason for action. Failure to recognize temporal neutrality involves temporal dissociation—failure to see the present as just one time among others—and failure to recognize impartiality or altruism involves personal dissociation—failure to recognize oneself as just one person among others.15 Alternatively, we might treat time and person as parallel and argue from the agent bias that egoism concedes to temporal bias, in particular, present bias. If my sacrifice for another is not rationally required, it may seem that we cannot demand a sacrifice of my current interests for the sake of distant future ones. If so, we will think that it is only the present interests of the agent that provide her with non-­derivative reason for action. Though Parfit mentions Nagel’s fully neutral response to parity, it is the fully biased response that he develops and thinks Sidgwick anticipated.16

3.  Rationalizing the Asymmetry Whereas Parfit thinks that one cannot defend the hybrid character of prudence, Sidgwick thinks that this challenge to prudence is unanswerable only if we accept 14  Reasons and Persons, 140. 15 Nagel, The Possibility of Altruism, 16, 19, 99–100. (a) I have often thought that the real value of Nagel’s discussion lies in its adequacy as a description of developmental psychology. It seems to me that the process of turning children into mature and responsible adults is in significant part the process of overcoming temporal and personal solipsism. (b) Nagel’s remarks about the ‘combinatorial problem’ (The Possibility of Altruism, 134–42) show that he is skeptical of an impersonal interpretation of impartiality. Nonetheless his appeal to parity seems to require neutralism and not just impartiality. He appeals to parity to argue from egoism’s temporal neutrality to non-­derivative concern for others. But if intertemporal and interpersonal distribution must be isomorphic, and we accept a temporally neutral interpretation of intertemporal impartiality, then we seem forced to accept a person neutral interpretation of interpersonal impartiality. 16  Reasons and Persons, 137–44. Here, I ignore differences between Parfit’s fully relative Present-­ aim Theory (P) and egoism of the present moment, which actually obscure the structural issues I want to focus on.

Special Concern and Personal Identity  23 Humean skepticism about personal identity over time. After stating the challenge to the hybrid character of egoism (quoted above), Sidgwick continues this way: It undoubtedly seems to Common Sense paradoxical to ask for a reason why one should seek one’s own happiness on the whole; but I do not see how the demand can be repudiated as absurd by those who adopt the views of the extreme empirical school of psychologists . . . Grant that the Ego is merely a system of coherent phenomena, that the permanent identical ‘I’ is not a fact but a fiction, as Hume and his followers maintain; why, then, should one part of the series of feelings into which the Ego is resolved be concerned with another part of the same series, any more than with any other series?17

Sidgwick thinks that prudence is defensible provided we recognize the separateness of persons. It would be contrary to Common Sense to deny that the distinction between any one individual and any other is real and fundamental, and that consequently ‘I’ am concerned with the quality of my existence as an individual in a sense, fundamentally important, in which I am not concerned with the quality of the existence of other individuals: and this being so, I do not see how it can be proved that this distinction is not to be taken as fundamental in determining the ultimate end of rational action for an individual.18

What exactly is this appeal to the separateness of persons, and how does it rationalize the hybrid structure of prudence? Sidgwick’s rationale appeals to two claims: a compensation principle and realism about personal identity. His request for a justification of sacrifice19 suggests a concern with compensation. Prudence rejects the sort of extreme temporal bias characteristic of egoism of the present moment. Now-­for-­later sacrifice is rational, because the agent is compensated later for her earlier sacrifice. To see how this rationale works, it will help to consider a familiar analogy between intrapersonal and interpersonal balancing. Whereas prudence is temporally neutral, utilitarianism is person-­neutral. Prudence is temporally neutral and assigns no intrinsic significance to when a benefit or burden occurs in a person’s life. It says that we should balance benefits and harms, where necessary, among different stages in a person’s life and pursue the action or policy that promotes the agent’s overall good best. Utilitarianism is interpersonally neutral; it assigns no intrinsic significance to whom a benefit or burden befalls. Just as temporal neutrality requires intrapersonal balancing, so too person neutrality requires interpersonal balancing. It requires that benefits to some be balanced against 17 Sidgwick, Methods of Ethics, 418–9. 19 Sidgwick, Methods of Ethics, 418.

18 Sidgwick, Methods of Ethics, 498.

24  David O. Brink harms to others, if necessary, to produce the best interpersonal outcome overall. Utilitarianism’s person neutrality thus effects a kind of interpersonal balancing akin to the intrapersonal balancing that prudence’s temporal neutrality requires. But many think that this sort of interpersonal balancing is unacceptable because it ignores the separateness of persons. For instance, John Rawls famously makes this claim in A Theory of Justice. This view of social cooperation [utilitarianism’s] is the consequence of extending to society the principle of choice for one man [i.e. prudence], and then, to make this extension work, conflating all persons into one . . . Utilitarianism does not take seriously the distinction between persons.20

Nagel agrees, as do Robert Nozick and Bernard Williams.21 They all accept prudence’s intrapersonal balancing, at least for the sake of argument, but reject utilitarianism’s interpersonal balancing. We can see how to deny the parity of intrapersonal and interpersonal cases and provide a rationale for the temporal neutrality of prudence by highlighting the role of compensation in the separateness of persons objection. Nozick’s discussion is especially instructive here. Individually, we each sometimes choose to undergo some pain or sacrifice for a greater benefit or to avoid a greater harm. . . . Why not, similarly, hold that some persons have to bear some costs that benefit other persons more? But there is no social entity with a good that undergoes some sacrifice for its own good. . . . To use a person in this way does not sufficiently respect and take account of the fact that he is a separate person, that his is the only life he has. He does not get some overbalancing good from his sacrifice, and no one is entitled to force this upon him.22

Like the others, Nozick is invoking claims about compensation to explain the asymmetric treatment of intrapersonal and interpersonal balancing. Whereas balancing benefits and harms is acceptable within a life, balancing benefits and harms across lives appears unacceptable. In the intrapersonal case, benefactor and beneficiary are the same person, so compensation is automatic. In the interpersonal case, benefactor and beneficiary are different people; unless the beneficiary reciprocates in some way, the benefactor’s sacrifice will not be compensated. Whereas intrapersonal compensation is automatic, interpersonal compensation 20 J. Rawls, A Theory of Justice (Cambridge, MA: Harvard University Press, 1971), 27. 21 Nagel, The Possibility of Altruism, 134, 138–42; R. Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974), 31–4; and B.  Williams, ‘Persons, Character, and Morality’, repr. in Moral Luck (Cambridge: Cambridge University Press, 1981), 3. 22 Nozick, Anarchy, State, and Utopia, 32–3.

Special Concern and Personal Identity  25 is not. This fact about compensation appears to rationalize intrapersonal neutrality without rationalizing interpersonal neutrality. 1.  Realism about personal identity is true. 2.  Sacrifice is rational if and only if and insofar (iffi) it is compensated. 3. Hence, intrapersonal temporal neutrality is required by the sufficiency of compensation for rational sacrifice; because beneficiary and benefactor are the same in the intrapersonal case, sacrifice is automatically compensated. 4. Hence, the necessity of compensation for rational sacrifice blocks interpersonal neutrality; because benefactor and beneficiary are distinct in the interpersonal case, sacrifices are not automatically compensated. 5.  Hence, we have a rationale for the hybrid character of prudence.

The necessity of compensation for rational sacrifice rationalizes agent bias, whereas the sufficiency of compensation for rational sacrifice rationalizes temporal neutrality. The necessity of compensation for sacrifice is a controversial claim, because it would imply that it is irrational for one person to make a very small sacrifice for the sake of an enormous benefit to another, as in a case of easy rescue in which I can save the life of another at little or no risk to myself. But if we remain agnostic about whether prudence exhausts practical reason, we may be less invested in defending the necessity of compensation for the rationality of sacrifice and agent bias. By contrast, the sufficiency of compensation for rational sacrifice is a more plausible requirement and seems sufficient to justify intrapersonal temporal neutrality. Or does it? Couldn’t doubts about interpersonal balancing be extended to intrapersonal balancing? If the separateness of persons defeats interpersonal balancing, why doesn’t the separateness of different periods within a person’s life defeat intrapersonal balancing? After all, me-­now and me-­later are distinct parts of me.23 But then it is hard to see how me-­now is any more compensated for its sacrifices on behalf of me-­later than I am compensated by my sacrifices for you. Just as doubts about interpersonal balancing lead to a distributed concern with each person, perhaps doubts about intrapersonal balancing should support a distributed concern with each part of a person’s life. There are different interpretations of what this distributed concern requires in the interpersonal context, such as 23  I intend talk about temporal parts of a person or person’s life to be metaphysically ecumenical in two ways. First, it is convenient to talk about persons and their temporal parts whether persons are four-­dimensional entities that literally have temporal parts (as three-­dimensional entities have spatial parts) or whether they are three-­dimensional entities that have no temporal parts but do have lives, histories, or careers that have temporal parts or stages. Talk about a person’s temporal parts can refer to temporal parts of persons or to parts of lives or careers of persons. Second, my talk of temporal parts is neutral in the debate among those who treat persons as four-­dimensional entities having temporal parts about whether persons or their temporal parts are prior in order of explanation.

26  David O. Brink equality or priority to the worse-­off. Perhaps we need to explore comparable interpretations of distributed concern in the intrapersonal context.24 However, this challenge to temporal neutrality requires thinking that we can and should adopt a sub-­personal perspective when reckoning compensation. But there are problems with this idea. Once we go sub-­personal and appeal to full temporal relativity, there seems no reason to stop until we reach the sub-­personal limit—a momentary time slice of the person. But notions of compensation have no application to person slices, which do not persist long enough to act or receive the benefits of earlier actions. Moreover, many of the goods in life, especially the pursuit and achievement of worthwhile projects, seem to be realized only by tem­por­ al­ly extended beings. But if we stop short of momentary time slices and appeal to larger sub-­personal entities—call these person segments—other problems arise. One question is just where to stop. If we don’t fully relativize, why relativize partially? Moreover, if we do relativize partially, we introduce indeterminacy. This is because the careers of person segments overlap, with the result that any one point in time is part of the career of indefinitely many different segments. Assume that the time is t25. Normally, we would ask what the person has reason to do at this time. But there are continuum-­many possible overlapping segments present at t25, of which the diagram in Figure 1.1 depicts just six. We can determine whether compensation has occurred and determine the person’s reasons for action at t25 if our unit of agency is the person, but if we adopt the sub-­personal conception of the units of agency as person segments, we seem to lack a de­ter­ min­ate agent.

t1

t25

t50

Figure 1.1 

Of course, persons are just maximal segments. They also seem to be the most salient segments. Many of the things we value and that structure our pursuits are certain sorts of lives. We aim to be certain sorts of people. Insofar as these ideals structure our beliefs, desires, and intentions, the correct perspective from which to assess success would seem to be the perspective of a whole life. Even when persons have more parochial aims and ambitions, the successful pursuit of these aims and ambitions requires interaction and cooperation among segments, much as persons must often cooperate with others to achieve individual, as well as collective, aims. They do interact and cooperate, much as distinct individuals 24  For some useful discussion of intrapersonal distributional principles, see D. McKerlie, ‘Equality and Time’, Ethics 99 (1989): 475–91.

Special Concern and Personal Identity  27 interact and cooperate in groups, in order to plan and execute long‑term projects and goals. They must interact and cooperate if only because they have to share a body and its capacities in order to execute their individual and collective goals, much in the way that individuals must sometimes interact and cooperate if they are to use scarce resources to mutual advantage.25 Indeed, both the ease and necessity of interaction among person segments will be greater than that among persons, because the physical constraints and the reliability of fellow cooperators are greater in the intrapersonal case. But this means that person segments will overlap with each other; they will stand to each other and the person much as strands of a rope stand to each other and the rope.26 Though we can recognize the overlapping strands as entities, the most salient entity is the rope itself. So too, the most salient entity is the person, even if we can recognize the overlapping person segments that make up the person. But then if we appeal to realism about personal identity and compensation, we can rationalize the hybrid character of prudence.

4.  Prudence and Psychological Reductionism This is one way to vindicate Sidgwick’s claim that the hybrid character of prudence is problematic only if we deny the separateness of persons. As we have seen, Sidgwick says that this rationale for the prudential commitment to temporal neutrality would be undermined by Humean skepticism about personal identity. But he suggests that all prudence requires is the separateness and persistence of persons, not any particular analysis of what personal identity consists in. If so, then the rationale for prudence that appeals to personal identity appears metaphysically ecumenical. Is it? In chapter  14 of Reasons and Persons Parfit argues that psychological reductionism about personal identity further undermines the hybrid character of prudence. His focus is on temporal neutrality, but reductionism raises interesting questions about both temporal neutrality and agent bias. To assess these claims, we need a working understanding of psychological reductionism. Psychological reductionism requires two concepts. Psychological Connectedness.  Persons Px and Py are psychologically connected iffi the mental states and actions of Py are counterfactually dependent in the appropriate way on the mental states and actions of Px.

25  See C.  Korsgaard, ‘Personal Identity and the Unity of Agency’, Philosophy & Public Affairs 18 (1989): 101–32. 26  Wittgenstein employs the rope metaphor but does not apply it to personal identity or agency. See L. Wittgenstein, The Blue and Brown Books (Oxford: Blackwell, 1958), 87.

28  David O. Brink Connectedness takes many forms—experiential memory, retention of skills and traits, persistent beliefs and desires, actions that reflect the influence of prior deliberations and intentions, and reasoned modification of one’s beliefs, values, and goals. We can then define psychological continuity in terms of psychological connectedness. Psychological Continuity.  Persons Px and Pz are psychologically continuous iffi Px and Pz are elements in a series of persons such that temporally contiguous pairs of persons are psychologically connected.

Psychological continuity is the ancestral relation to psychological connectedness. Psychological reductionism should be formulated in terms of psychological continuity, because, as Reid pointed out, identity is a transitive relation, as continuity is and connectedness is not.27 Psychological Reductionism I:  P1 and P2 are identical iffi they are (sufficiently) psychologically continuous.

Parfit offers several arguments for thinking that psychological reductionism undermines temporal neutrality. His most interesting argument claims that psychological reductionism justifies an intertemporal discount rate of concern, because psychological connectedness normally diminishes over time.28 My concern for my future may correspond to the degree of connectedness between me now and myself in the future. Connectedness is one of the two relations that give me reasons to be specially concerned about my own future. It can be rational to care less, when one of the grounds for caring will hold to a lesser degree. Since connectedness is nearly always weaker over long periods, I can rationally care less about my further future.29

Parfit’s defense of an intertemporal discount rate seems to have the following structure. 1.  The psychological reductionist’s relation R involves continuity and connectedness. 2. Connectedness normally diminishes over time: P3 is typically less connected to P1 than P2 is. 27  This is the standard moral drawn by reductionists from Reid’s worry that Locke’s memory cri­ter­ ion is not a transitive relation. See Reid, Essays on the Intellectual Powers of Man, 357. 28  For a fuller discussion of Parfit’s arguments, see Brink, ‘Rational Egoism and the Separateness of Persons’. 29 Reid, Essays on the Intellectual Powers of Man, 313.

Special Concern and Personal Identity  29 3. Hence, one is normally justified in caring less about one’s further future than one’s nearer future. 4.  Hence, temporal neutrality is false.

Indeed, the natural conclusion of this argument would seem to be that temporal discounting is normally rationally required and not merely rationally permissible. But there are several problems with this argument. First, we should reject (1). Psychological reductionism should be formulated in terms of continuity, rather connectedness, because only continuity meets Reid’s transitivity worry. Whereas connectedness may depreciate over time, continuity does not. Second, (2) is questionable. It’s not clear that connectedness does depreciate over time. Parfit seems to think that psychological connectedness requires psychological similarity and that psychological similarity normally diminishes over time. But psychological connectedness requires counterfactual dependence of later actions and intentional states on earlier ones, not psychological similarity or fixity of character. Indeed, if similarity were essential to prudential concern, Parfit’s argument would imply that we have less prudential reason to undertake psychological change the greater the change is. On such a view, we would have no little or no reason to undertake dramatic self-­improvement, because the later person would be too different from the earlier one. But self-­improvement seems to be the paradigm of prudential concern. Third, (3) does not follow from (1) and (2). Even if relation R consisted in connectedness and connectedness diminished over time, this argument would require the assumption that the normatively relevant unit of agency is sub-­ personal. For we can see how a person segment might be more connected with an immediate successor than a distal successor. But we have argued that whole lives are the normatively relevant unit of agency, and earlier and later segments of a life are equally parts of that life. This confirms Sidgwick’s belief that the justification of temporal neutrality is metaphysically ecumenical. However, prudence combines temporal neutrality and agent bias, and psychological reductionism might seem to undermine agent bias. We saw that Locke, Butler, and Reid believe that prudential special concern presupposes personal identity in the sense that P1 can have prudential concern for P2 iffi P1 = P2. But the psychological reductionist analysis of fission cases, in which there is symmetrically branching continuity, appears to undermine this claim. Consider a case of fission in which Tom’s mental life is split symmetrically between two persons Dick and Harry. The psychological reductionist believes that psychological continuity constitutes identity. What happens in fission? There seem to be four principal possibilities. 1. Tom survives as Dick and as Harry. 2. Tom survives as Dick, rather than Harry.

30  David O. Brink 3. Tom survives as Harry, rather than Dick. 4. Tom survives as neither Dick nor Harry.

The psychological reductionist might be tempted by (1). If Tom survives in a non-­branching case, it might seem he should survive as Dick and as Harry in the fission case. But this would violate the transitivity of identity. If Tom is identical to Dick and Tom is identical to Harry, then, by the transitivity of identity, Dick must be identical to Harry. But that is plainly not the case. They are distinct persons with different spatio-­temporal locations. If I pinch Dick, Harry does not feel it, and vice versa. By Leibniz’s Law, they are not identical. But a psychological reductionist has no reason to accept (2) or (3), as Dick and Harry have precisely symmetrical claims to being Tom. That leaves (4), which is initially hard to believe, because if non-­branching continuity is a success in terms of Tom’s survival, how can a double success be a failure? However, (4) is the consensus description of fission for psychological reductionists. If we are psychological reductionists, we should deny Tom survives as Dick or as Harry, because of the transitivity of identity, and conclude that fission is a case of interpersonal psychological continuity. Whereas fission preserves psychological continuity, which can be one-­many, it cannot preserve identity, which must be one-­one. For psychological continuity to constitute personal identity, it must take a non-­ branching form. Psychological Reductionism II:  P2 is identical with P1 iffi P2 is (sufficiently) psychologically continuous with P1 and there is no other continuer of P1 that is as continuous with P1.

But fission also teaches us that special concern tracks continuity, rather than identity. Tom should have special concern for the lives of Dick and Harry as he would normally have for his own future self. But this suggests that special concern does not presuppose personal identity. P1 can have special concern for P2 even if P1 ≠P2. In this way, fission challenges the agent bias of prudence and egoism. Prudence can accommodate fission and other cases of interpersonal psychological continuity if psychological continuity extends a person’s interest. In non-­branching cases, psychological continuity extends Tom’s life and thereby extends his interests. In fission, continuity cannot extend Tom’s life, because of the logic of identity. But it can extend his interests. We can see Tom’s interests preserved in the lives of Dick and Harry. On this view, Tom has posthumous interests in the lives of Dick and Harry. I find this kind of extended self-­interest and prudence plausible.30 But I realize that it is controversial.

30  I discuss these issues at greater length in Brink, ‘Self-­Love and Altruism’.

Special Concern and Personal Identity  31 Alternatively, we might agree that prudence presupposes personal identity and that fission shows that special concern does not presuppose personal identity but introduce a quasi-­egocentric form of prudence that presupposes psychological continuity, rather than personal identity. Here, we could take a page from the reductionist reply to Butler’s worry about the circularity of Locke’s memory criterion.31 The standard reductionist reply, first clearly articulated by Sydney Shoemaker, is to appeal not to memory but quasi-­memory, which is otherwise like memory except that it presupposes continuity, rather than identity.32 In a similar way, quasi-­prudence might presuppose psychological continuity, rather than personal identity. Quasi-­prudence would be agent-­relative, because it makes special concern depend on the psychological relationship between benefactor and beneficiary, but it would not be strictly egoistic.

5.  How Special Concern Depends on Personal Identity So perhaps we should demur about whether prudential special concern presupposes personal identity, inasmuch as fission suggests that we could have special concern without personal identity. However, we shouldn’t overstate this result. We may be able to reconcile fission with the agent bias of prudence if we claim that psychological continuity extends a person’s interests, even when it doesn’t extend her life. And even if we don’t accept this extended sense of self-­ interest, we can reconcile fission with a quasi-­egocentric commitment to quasi-­ prudence. But we should allow that there are some cases in which special concern does not presuppose personal identity. But even if we cannot justify prudence’s agent bias by appeal to personal identity, we have seen that personal identity helps provide a rationale for a temporally neutral form of special concern. Specifically, an ecumenical form of realism about personal identity and the claim that compensation is sufficient for sacrifice are sufficient to justify the intrapersonal temporal neutrality characteristic of prudence. In this way, personal identity is prior to temporally neutral special concern, in general, and prudence, in particular.

6.  How Personal Identity Depends on Special Concern In this way, Sidgwick’s rationale for prudence fits within a larger tradition that claims that personal identity is prior to special concern and prudence. But the psychological reductionist might want to invert this claim, asserting that there is a sense in which special concern is prior to personal identity. 31  Butler, ‘Dissertation on Personal Identity’, 264. 32  Sydney Shoemaker, ‘Persons and Their Pasts’, American Philosophical Quarterly 7 (1970): 269–85.

32  David O. Brink To see this, let’s return to interpersonal psychological continuity. We described fission as a case of interpersonal psychological continuity in which there is just as much continuity as in the normal intrapersonal case. What makes fission an interpersonal case is that continuity takes a one-­many form, whereas the logic of identity is one-­one. At least for now, fission is an exotic, hypothetical case. But interpersonal psychological continuity is quite common and found in many forms of interpersonal association—between spouses and domestic partners, family members, friends, colleagues, neighbors, etc. In these cases, the mental states and intentional actions of each person depend on those of her associates in various ways. Associates share experiences and influence each other through discussion, argument, and example. They care about each other and share each other’s sorrows and joys. Psychological continuity makes for interpersonal unity. Associations differ in how tight their bonds are and how much interpersonal interaction and influence they involve. Indeed, we might take a page from the Stoic Hierocles and think of our relations to others in terms of a set of concentric circles, with ourselves in the innermost circle, those near and dear to us in nearby circles, and weaker relations in outer circles.33 This suggests that the psychological reductionist about personal identity can and should say that what distinguishes intrapersonal and interpersonal unity in normal circumstances (not involving fission or other forms of duplication) is the degree or strength of psychological continuity. In ‘Friends and Future Selves’ Whiting wants to make an analogy between intrapersonal and interpersonal psychological continuity, drawing lessons for personal identity from claims about friendship. She claims that special concern for one’s friend is an important ingredient in friendship. Part of what she has in mind is the sort of positive affective regard that is normal among friends. Friends are fond of each other and wish each other well. But presumably one might also understand special concern for one’s friend to include behavioral dispositions to help them when in need, invest in the relationship, and sacrifice one’s interests for the sake of the friend. One such manifestation of interpersonal special concern is trust. Friends rely on each other for information, advice, constructive criticism, support, and help. To be sure, there are forms of interaction that are part of friendship—spending time together, keeping in touch, and sharing experiences—that can be characterized without mentioning special concern. And we may not maintain this sort of care and concern for our friends at every moment in our friendships. There are moments within friendships when good will and mutual concern lapse. But it seems hard to imagine friendships persisting that did not generally exhibit this sort of special concern among friends. 33  See A. A. Long and D. N. Sedley (trans.), The Hellenistic Philosophers, i (Cambridge: Cambridge University Press, 1987), 349. For discussion of this Stoic claim, see D.  Brink, ‘Eudaimonism and Cosmopolitan Concern’, in D. Brink, S. Meyer, and C. Shields (eds.), Virtue, Happiness, and Knowledge: Themes from the Work of Gail Fine and Terence Irwin (Oxford: Clarendon Press, 2018).

Special Concern and Personal Identity  33 Just as special concern is partly constitutive of interpersonal psychological continuity, Whiting argues, special concern for one’s future self is partly constitutive of intrapersonal psychological continuity. Presumably, again, she has in mind both positive affective regard for one’s own future and behavioral dispositions to plan ahead, invest in one’s future, and make sacrifices for the sake of one’s future. Here too, one manifestation of such intrapersonal special concern involves trusting oneself. Intrapersonal trust is an important element of psychological continuity. Though we can reassess our earlier deliberations, beliefs, and intentions, it is important that we not do so continuously, or we would never make new inquiries or execute plans. It is essential that our later selves be disposed to trust the theoretical and practical deliberations of our earlier selves unless there is some good reason to think that we made some mistake earlier or significantly new evidence or options have become available in the meantime. Indeed, any temporally extended piece of reasoning or planning involves this sort of tacit reliance on one’s earlier self. And the formation of plans and intentions involves trusting that one’s later self will act on one’s current deliberations, unless something significant occurs to rebut that presumption. As in the interpersonal case, intrapersonal continuity would be anchored by psychological continuity that does not involve special concern—remembering past experiences, acting on earlier intentions, the persistence of skills, beliefs, and tastes, and the reasoned modification of beliefs, values, and plans. Nonetheless, in normal intrapersonal cases caring about and planning for one’s future are essential aspects of intrapersonal continuity. Of course, this sort of self-­directed good will and concern can lapse for periods of time in cases of depression or self-­loathing. But if these lapses persist and become the norm, we might take this to impair intrapersonal unity.34 Whiting claims that this stands Sidgwick’s claim about special concern depending on personal identity on its head, claiming that personal identity depends on special concern. Her thesis about the priority of special concern to personal identity requires two important qualifications. First, special concern does not exhaust continuity. Presumably, friendship also consists in various kinds of shared experiences and history that might be prior to and independent of special concern and that might, indeed, make special concern reasonable. Perhaps some friendships begin with special concern, as in love at first sight. But in many, if not most, cases, special concern is anchored in and emerges out of other forms of interpersonal continuity, such as shared experiences, cooperation, and familiarity. Nonetheless, we might agree that

34  One might even think that that there is a sense in which self-­loathing actually displays special concern inasmuch as it seems to reflect a form of disappointment in oneself that is out of proportion to the sort of disappointment one might have in a stranger. These issues merit more reflection and attention than I can give them here.

34  David O. Brink special concern plays a constitutive role in normal, healthy forms of friendship. Similarly, there are forms of intrapersonal psychological continuity that might be prior to and independent of self-­concern. There are memories of earlier experiences, and various skills, knowledge, preferences, and ideals are retained or modified over time. These forms of intrapersonal psychological continuity do not seem to depend on self-­concern and can, indeed, make self-­concern possible and reasonable. Nonetheless, self-­concern seems to play a constitutive role in normal, healthy intrapersonal unity. Second, in the intrapersonal case special concern takes the form of self-­ concern. But self-­concern seems to presuppose personal identity, inasmuch as one can only have self-­concern for one’s own future. But then it seems that special concern could not constitute personal identity without circularity. We can answer this circularity charge by again taking a page from the Lockean reply to Butler’s circularity worry, appealing to the concept of quasi-­self-­concern, which presupposes psychological continuity, rather than identity. Quasi-­self-­concern is special concern for another with whom one is psychologically continuous. It presupposes other forms of intrapersonal psychological continuity, but then further contributes to intrapersonal psychological continuity.

7.  Does Personal Identity Depend on Prudence? We’ve now explored the plausibility of thinking that special concern is an important ingredient in personal identity. Must this metaphysical role for special concern take a specifically prudential form? Does intrapersonal unity require agent bias and temporal neutrality? That seems unlikely. Special concern does not require agent bias. It seems essential to special concern that one have concern for one’s future that is different in kind or degree from one’s interest in strangers. But this needn’t preclude non-­derivative concern for others. Moreover, we’ve seen that special concern can extend to friends and others with whom one is continuous. Could there be persons who display no special concern for themselves whatsoever? Even if most of us display special concern for ourselves and our associates, isn’t it possible to be guided by a completely impartial interest in humanity in the manner of Mother Teresa or some selfless altruist, whose overriding goal is to do the most good in the world? Sometimes I wonder if a concern for maximizing total good makes sense without the assumption that each person has some special concern for herself. For if there is no personal good, independent of the good of others, it seems that there is nothing to ground beneficent impulses. Perhaps this only requires that not everyone be a selfless altruist. In any case, it seems that the selfless altruist does have some forms of special concern, even if this is only derivatively justified. For the selfless altruist must focus on her own life as an expression of altruistic

Special Concern and Personal Identity  35 commitments and as a means of effectuating the most good. She will be invested in her own contributions to maximizing the good in a way in which she is not invested in others, if only because she has more control over her own activities than those of others. This will require the forms of intrapersonal trust, planning, and sacrifice that we have discussed above. But if she is a true selfless altruist, this self-­concern will be regulated by her ultimate aim in which her own good is just one among many, with no privileged position. Most of us are not selfless altruists. We display non-­derivative forms of special concern for our own futures, which is part of what makes us diachronically unified persons. But must this special concern take a temporally neutral form? There are two main kinds of temporal bias—a bias toward the near in which proximate goods and bads are assigned normative significance out of proportion to their actual magnitude and a bias toward the future in which goods and harms have normative significance prospectively that is greater than their significance retrospectively. Temporal neutrality views both kinds of biases as mistakes. As we have seen, Sidgwick provides a rationale for thinking that bias toward the near is mistaken. Future bias leads to temporal instability in the prices we are willing to pay for goods and services, depending on whether we pay in advance or after-­the-­ fact, and it can turn agents with this bias into money pumps.35 However, even if temporal discounting is irrational, it is a common part of lives that seem sufficiently unified so as to count as persisting persons. So even if special concern for one’s own future is an ingredient in personal identity, it seems unlikely that temporally neutral concern is necessary for personal identity. Friends of temporal neutrality can and should defend it as a normative claim, noting that it can be understood as an idealization of the sort of intertemporal special concern that is partly constitutive of personal identity. Temporal neutrality is arguably a regulative ideal present in actual planning and investment, and so we might say that lives are more fully unified insofar as they display forms of planning and investment that are temporally neutral.

8.  Special Concern, Personal Identity, and Explanatory Priority Special concern and, in particular, prudence have seemed intimately connected with personal identity to many philosophers. How should we understand this relationship? On the one hand, as Locke, Butler, and Reid claim, it seems that we 35  Parfit discusses future bias in Reasons and Persons, sect. 64. For criticisms of future bias, see Brink, ‘Prospects for Temporal Neutrality’, sect. 7; T. Dougherty, ‘On Whether to Prefer Pain to Pass’, Ethics 121 (2011): 521–37 and ‘Future Bias and Practical Reason’, Philosophers’ Imprints 15 (2015): 1–17; P. Greene and M. Sullivan, ‘Against Time Bias’, Ethics 125 (2015): 947–70; and M. Sullivan, Time Biases (Oxford: Clarendon Press, 2018).

36  David O. Brink have special reason to be concerned about our own future selves. Sidgwick defends a version of this claim by arguing that an ecumenical version of realism about personal identity and the sufficiency of compensation for the rationality of sacrifice justify temporally neutral special concern characteristic of prudence. On the other hand, according to some plausible forms of psychological reductionism, special concern for one’s own future or at least for selves with whom one is psychologically continuous is partly constitutive of diachronic personal identity. On this view, special concern contributes to personal identity. But can we accept this symmetrical account of the relation between special concern and personal identity? It seems that it cannot be true both that special concern presupposes personal identity and that personal identity presupposes special concern. That would be circular. It might seem that at most one of these claims could be true. This might seem to be the lesson of the Euthyphro problem— that when two concepts are biconditionally related, we need to determine which concept is explanatorily prior. Suppose that X and Y are biconditionally related. X iff Y

In principle, the relationship between X and Y could be further specified in any of three ways. 1.  X is prior to and explains Y. 2.  Y is prior to and explains X. 3. X and Y are equally explanatorily fundamental in relation to each other; neither is prior to the other.

It’s not possible for both (1) and (2) to be true, at least if each is prior to the other in the same respect. That would apparently involve a viciously small circle. But neither might be prior to the other if each is equally fundamental in relation to each other. This is the No Priority thesis in (3) and sometimes seems like a plausible analysis of concepts that are biconditionally related.36 Indeed, No Priority can even endorse the explanatory priority of both X and Y, provided that each is explanatorily prior in a different respect. Special concern and personal identity might be said, on this analysis, to be prior to each other in different ways. On the one hand, special concern is metaphysically prior to personal identity, because special concern helps constitute personal identity. In the case of positive affective regard, special concern is one ingredient among others in intrapersonal 36  This may be the way to interpret the methodological assumptions that Wiggins makes when he endorses a non-­reductive biconditional relationship between being good and being such as to produce approval in the appropriate sort of appraiser, eschewing analysis and aspiring only to provide philosophical ‘commentary’ or ‘elucidation’. See D.  Wiggins, ‘A Sensible Subjectivism?’, in Needs, Values, and Truth (Oxford: Blackwell, 1987), esp. 188–9.

Special Concern and Personal Identity  37 psychological continuity. This sort of occurrent concern does not exhaust intrapersonal unity, because there are other attitude-­independent forms of psychological continuity, and it may not even be necessary for personal identity. But it is a normal consequent of these attitude-­independent forms of continuity and further contributes to intrapersonal unity. If we extend the notion of special concern to include dispositional concern of the sort manifested in relations of trust, co­oper­ation, planning, and investment, then special concern is arguably necessary for the kind of intrapersonal psychological continuity that makes for personal identity. At least, that kind of special concern is a major factor in intrapersonal psychological continuity, without which our lives would be substantially different and less unified. Special concern of this sort need not take a temporally neutral form in order to unify a life. On the other hand, personal identity is normatively prior to special concern and prudence, because the normative justification or rationalization of special concern and prudence rests on realism about personal identity and the compensation principle. In particular, realism about personal identity and the sufficiency of compensation for sacrifice rationalize temporally neutral concern for oneself. It is not just any form of special concern that is rationalized in this way; personal identity and compensation justify the temporal neutrality characteristic of prudence, because in the intrapersonal case there is automatic compensation for sacrifice. But if special concern and personal identity are each prior to the other in different ways, then there need be no incoherence or circularity in this kind of symmetrical dependence. Indeed, if each is prior to the other in different respects, then we can conclude that neither is prior to the other simpliciter. This is the No Priority thesis.

Bibliography Brink, D., ‘Self-Love and Altruism’, Social Philosophy & Policy 14 (1997): 122–57. Brink, D., ‘Rational Egoism and the Separateness of Persons’, in J. Dancy (ed.), Reading Parfit (Oxford: Blackwell, 1997). Brink, D., ‘Prospects for Temporal Neutrality’, in C.  Callender (ed.), The Oxford Handbook of the Philosophy of Time (Oxford: Clarendon Press, 2011). Brink, D., ‘Eudaimonism and Cosmopolitan Concern’, in D.  Brink, S.  Meyer, and C. Shields (eds.) Virtue, Happiness, and Knowledge: Themes from the Work of Gail Fine and Terence Irwin (Oxford: Clarendon Press, 2018). Broad, C. D., ‘Self and Others’, repr. in Broad’s Critical Essays in Moral Philosophy, ed. D. Cheney (London: George Allen & Unwin, 1971). Butler, J., ‘Dissertation on Personal Identity’, in The Whole Works of Joseph Butler, ed. S. Halifax (London: William Tegg & Co., 1847).

38  David O. Brink Dougherty, T., ‘On Whether to Prefer Pain to Pass’, Ethics 121 (2011): 521–37. Dougherty, T., ‘Future Bias and Practical Reason’, Philosophers’ Imprints 15 (2015): 1–17. Greene, P., and Sullivan, M., ‘Against Time Bias’, Ethics 125 (2015): 947–70. Korsgaard, C., ‘Personal Identity and the Unity of Agency’, Philosophy & Public Affairs 18 (1989): 101–32. Locke, J., An Essay Concerning Human Understanding, ed. P.H.  Nidditch (Oxford: Clarendon Press, 1975). Long, A., and Sedley, D. (trans.), The Hellenistic Philosophers, i (Cambridge: Cambridge University Press, 1987). McKerlie, D., ‘Equality and Time’ Ethics 99 (1989): 475–91. Nagel, T., The Possibility of Altruism (Princeton: Princeton University Press, 1970). Nozick, R., Anarchy, State, and Utopia (New York: Basic Books, 1974) Parfit, D., Reasons and Persons (Oxford: Clarendon Press, 1984). Rawls, J., A Theory of Justice (Cambridge, MA: Harvard University Press, 1971). Reid, T., Essays on the Intellectual Powers of Man, ed. B. Brody (Cambridge, MA: MIT Press, 1969). Shoemaker, S., ‘Persons and Their Pasts’, American Philosophical Quarterly 7 (1970): 269–85. Sidgwick, H., The Methods of Ethics (7th ed., London: Macmillan, 1907). Sullivan, M., Time Biases (Oxford: Clarendon Press, 2018). Velleman, D., ‘Well-Being and Time’, Pacific Philosophical Quarterly 72 (1991): 48–77. Whiting, J., ‘Friends and Future Selves’, Philosophical Review 95 (1986): 547–80. Williams, B., ‘Persons, Character, and Morality’, in Moral Luck (Cambridge: Cambridge University Press, 1981). Wiggins, D., ‘A Sensible Subjectivism?’, in Needs, Values, and Truth (Oxford: Blackwell, 1987). Wittgenstein, L., The Blue and Brown Books (Oxford: Blackwell, 1958).

2

Separating Persons James Goodrich

By the time Derek Parfit’s Reasons and Persons was published in 1984, a ­near-­consensus had emerged that traditional forms of consequentialism are false because they fail to respect the ‘separateness of persons’.1 Parfit held that the reductionist views of personal identity he defended in Reasons and Persons undermined this emerging consensus: The fact that we live different lives is the fact that we are not the same person. If the fact of personal identity is less deep, so is the fact of non-­identity. There are not two different facts here, one of which is less deep on the Reductionist View, while the other remains as deep. There is merely one fact, and this fact’s denial. The separateness of persons is the denial that we are all the same person. If the fact of personal identity is less deep, so is this fact’s denial.2

The aim of this essay is to argue that Parfit’s reductionist views of personal identity are in fact compatible with a plausible interpretation of the separateness of persons. My thesis does not, however, leave those of us who are inclined to accept Parfit’s reductionist views of personal identity in the same place as those who have traditionally wielded the separateness of persons as an objection to consequentialism. This is because the separateness of persons, on the Parfitian view of personal identity, can arise within a life. We thus have reason to reject moral views which treat the interpersonal and intrapersonal distribution of goods as different in kind. The legacy of Reasons and Persons should therefore not be understood as undermining the significance of the separateness of persons, but as transforming our understanding of its role in our moral and prudential thinking. Thanks are owed to many people for conversations regarding the content of this paper. Chief among them are Krister Bykvist, Simon Beard, D’i Black, Timothy Campbell, and Daniel Ramoller. Many thanks are also owed to Jeff McMahan who gave invaluable comments on an early draft. 1  See e.g. D. Gauthier, Practical Reasoning (Oxford: Clarendon Press, 1963), 123–7; T. Nagel, The Possibility of Altruism (Oxford: Oxford University Press, 1970), 134; R.  Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974), 32–3; J. Rawls, A Theory of Justice (Cambridge, MA: Harvard University Press, 1971), 26–7; and B.  Williams, ‘A Critique of Utilitarianism’, in J.  J.  C.  Smart and Bernard Williams (eds.), Utilitarianism: For and Against (Cambridge: Cambridge University Press, 1973), 108–18. 2 D. Parfit, Reasons and Persons (Oxford: Oxford University Press, 1984), 339. James Goodrich, Separating Persons In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0003

40  James Goodrich My argument proceeds by considering a series of cases. The first set of cases, which I dub ‘Diamond Cases’, show that the separateness of persons has intuitive support. The second set of cases, which I dub ‘Fission Cases’, illustrates Parfit’s reductionist views of personal identity. The third set of cases, which I dub ‘Diamond-­Fission Cases’, combine Diamond Cases and Fission Cases. These cases will illustrate that there is intuitive support for the separateness of persons in the very kinds of cases which motivates Parfit’s reductionist views of personal identity. I propose that the lesson that those of us attracted to Parfit’s reductionist views of  personal identity should draw from Diamond-­ Fission cases is that the ­fundamental concern underlying the separateness of persons can also arise within a single individual’s life. Thus, on the Parfitian view, we should accept that individuals are divided in a normative sense, not just a metaphysical sense.

1.  Diamond Cases This section introduces the Diamond Cases. These cases are meant to illustrate the intuitive appeal of the separateness of persons. Thus, to introduce Diamond Cases properly, we should get a firmer grip on what it means to respect the separateness of persons in the first place. John Rawls is often cited as the originator of the objection that Classical Utilitarianism fails to respect the separateness of persons. Here’s his most cited passage on the topic: [Classical Utilitarianism] is the consequence of extending to society the principle of choice for one man, and then, to make this extension work, conflating all persons into one through the imaginative acts of the impartial sympathetic spectator. Utilitarianism does not take seriously the distinction between persons.3

The objection, quite roughly, is that the Classical Utilitarian conflates what is morally appropriate for the distribution of goods or utility among different persons with what is prudentially appropriate for the distribution of goods or utility within a single life over time. The best examples to illustrate Rawls’s point were introduced in Peter Diamond’s critique4 of John Harsanyi’s defense of Average Utilitarianism.5 Diamond objects to Harsanyi by offering a pair of cases. Here’s a structurally analogous version of the first of Diamond’s cases: 3  See Rawls, A Theory of Justice, 27. 4 See P.  Diamond, ‘Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparison of Utility: Comment’, Journal of Political Economy 75 (1967): 765–6. 5 See J.  Harsanyi, ‘Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparisons of Utility’, Journal of Political Economy 63 (1955): 309–21. Interestingly, Rawls believed that Harsanyi’s

Separating Persons  41 Intrapersonal Coin-­Flip:  You can flip one of two coins, φ or ψ. If φ lands heads, Anne receives 10 units of welfare on Monday and 5 units of welfare on Tuesday. If φ lands tails, Anne receives 5 units of welfare on Monday and 10 units of welfare on Tuesday. No matter what side of ψ lands face up, Anne will receive 10 units of welfare on Monday and 5 units of welfare on Tuesday.

So long as Anne doesn’t have a preference for which day of the week she receives her welfare, it doesn’t matter whether one flips φ or ψ. This is because there’s nothing objectionable in thinking that one’s reasons to prefer different outcomes might be neutral with respect to times. Most who invoke the Separateness of Persons as an objection to Classical Utilitarian will agree with the intuitive verdict in this case. But consider a structural analogue of Diamond’s second case: Interpersonal Coin-­Flip:  You can flip one of two coins, φ or ψ. If φ lands heads, Beth receives 10 units of welfare and Carol receives 5 units of welfare. If φ lands tails, Beth receives 5 units of welfare and Carol receives 10 units of welfare. No matter what side of ψ lands face up, Beth will receive 10 units of welfare and Carol will receive 5 units of welfare.

Many believe that it does matter which coin you choose to flip in this case. You should choose to flip φ because it gives each potential receipt of the welfare benefit a ‘fair shake’ at receiving the larger benefit. From a completely person-­neutral standpoint—that is, a standpoint that does not recognize the separateness of persons—it should not matter which coin you flip. Each way the coin lands would confer the same sum total of benefits impersonally construed. Therefore, highly impersonal views like Classical Utilitarianism seem to run counter to powerful and important intuitions about the separateness of persons in cases like Interpersonal Coin-­Flip. On a slightly different way of thinking about it, the problem is not that Classical Utilitarianism is impersonal per se. Instead, the problem is that the Classical Utilitarian does not recognize that there are moral reasons to prefer that one distribution of equally valuable prospects to another. Because individuals are ‘separate’ they must each receive an equal chance, insofar as it’s possible, of receiving a given benefit. The lesson of Intrapersonal Coin-­Flip and Interpersonal Coin-­Flip, according to those who have traditionally invoked the separateness of persons as an objection to Classical Utilitarianism is that these two cases illustrate that there’s an important moral difference between how we should distribute goods within a life and how we should distribute goods between lives. There is, in other words, a view did not violate the separateness of persons for reasons most of us who now discuss the separateness of persons often ignore. See Rawls, A Theory of Justice, 183–94.

42  James Goodrich kind of asymmetry between intrapersonal and interpersonal cases. However, notice that we don’t need to posit an asymmetry between intrapersonal and interpersonal cases in order to accept our intuition about Interpersonal Coin-­ Flip. We could accept this intuition while also rejecting our intuition about some versions of Intrapersonal Coin-­Flip, thereby adopting a non-­standard, but symmetrical view. Thus there are three possible views: The Asymmetry View:  The distribution of goods in intrapersonal cases and interpersonal cases should be treated differently. For example, we have moral reasons to give two different individuals a fair shake in cases like Interpersonal Coin-­Flip. But we have no moral reason, ceteris paribus, to give two distinct time slices of the same person a fair shake. The Old Symmetry View:  We should think of the distribution of goods in interpersonal cases on the model of intrapersonal cases. On this view, we have no moral reasons to give two different individuals a fair shake. We also have no moral reasons, ceteris paribus, to give two distinct time slices of the same person a fair shake. The New Symmetry View:  We should think of the distribution of goods in intrapersonal cases on the model of interpersonal cases. On this view, we do have moral reasons to give two different individuals a fair shake in the rights. However, we also have moral reasons, ceteris paribus, to give two distinct time slices of the same person a fair shake.

Classical Utilitarians hold the Old Symmetry View. They believe that how welfare is distributed within a life as opposed to between lives is not make a moral difference. And they would reject the view that we should give fair shakes in cases of intrapersonal distribution. Those who have wielded the separateness of persons as an objection to Classical Utilitarianism hold the Asymmetry View. They believe it’s a matter of fairness that we don’t treat the distribution of goods between individuals as we treat the distribution of good within a life. The New Symmetry View holds, in effect, that the concern underlying the separateness of persons can arise within a life. If that’s right, the Classical Utilitarian is not wrong because they fail to accept an important difference in the distribution of goods in intrapersonal and interpersonal cases, but because they fail to recognize the moral concern underlying the separateness of persons in all cases.

2.  What Did Parfit Believe? The passage I quoted of Parfit’s in my introductory remarks suggests that he held the Old Symmetry View. However, if we consider different passages of Reasons and Persons, whether it would be fully accurate too attribute to Parfit the Old

Separating Persons  43 Symmetry View becomes less clear. For example, Parfit summarizes part of his conclusion as follows: Thus it is more plausible to claim that great burdens imposed upon a child cannot be compensated, or fully compensated, by somewhat greater benefits in this child’s adult life. When we thus extend distributive principles so that they cover, both whole lives, and weakly connected parts of the same life, this makes these principles more important. This is a move away from the Utilitarian View. I have just discussed a second argument for a change in the scope of our distributive principles. This claims that only the deep further fact makes possible compensation over different parts of a life. Since there is no such fact, we ought, as Nagel suggests, to change ‘the size of the units over which a distributive principle operates’. The units shrink to people’s states at particular times.6

In this passage, Parfit is at least entertaining something very like the New Symmetry View. The idea that our distributive principles should operate over units much small than lives—that is, over time slices of a single life—is very much in keeping with the New Symmetry View. However, Parfit immediately follows this statement with a bit of hedging: ‘This conclusion is defensible, but so is its denial.’7 This statement suggests that Parfit was at least agnostic about which view to accept. And interestingly, he follows this claim up by reaffirming the passage I quote in my introductory remarks about how the separateness of persons may seem to do less work in justifying distributive principles on the reduction view of persons.8 However, here, once again, Parfit takes an interesting turn. He says, I earlier discussed whether, if we change our view, we had as much reason to be specially concerned about our own futures. Some writers claim that only the deep further fact justifies this special concern, and that, since there is no such fact, we have no reason to be specially concerned about our own futures . . . I then advanced another argument against the Self-­interest Theory about Rationality. This appealed to the fact that part of what is important in personal identity, psychological connectedness, holds over time to reduced degrees. When some important fact holds to a reduced degree, it cannot be irrational to believe this fact to have less importance. It therefore cannot be irrational to be less concerned, now, about those parts of our future to which we are now less closely connected.9

Here, Parfit seems to be saying that our future selves are less like us and moral like other people from the standpoint of prudence. He even takes this one step further 6 Parfit, Reasons and Persons, 346. 8 Parfit, Reasons and Persons, 346.

7 Parfit, Reasons and Persons, 346. 9 Parfit, Reasons and Persons, 346–7.

44  James Goodrich by claiming that any criticism of how one treats his or her distant future self can only be criticized on moral grounds: On the Revised Self-­interest Theory, which is not refuted, it may not be irrational to do what one knows will be worse for oneself. Great imprudence may not be irrational. If such acts are not irrational, they need to be criticized. I claimed that we should regard them as morally wrong. If such acts are morally wrong, this strengthens the case for paternalism.10

If Parfit is right about this, then perhaps there remains room on his view to accept the New Symmetry View. Of course, claiming that prudence concedes ground to morality does not by itself entail the New Symmetry View. Perhaps this explains why Parfit seems to hedge in other passages. He is agnostic about whether to accept the New Symmetry View, at least in part, because his views don’t entail it. What we would need is a further argument. I will argue that those who accept Parfit’s claims about personal identity should accept that the truth lies between the Asymmetry View and the New Symmetry View. In other words, we should hold tightly to our intuitions about cases like Interpersonal Coin-­Flip, but some intrapersonal cases should be thought of more on the model of interpersonal cases than others. The difference between the interpersonal and the intrapersonal, on this view, is a matter of degree, not kind. For now, let’s set this view to the side. We’ll return to it in due course. The next step in my argument is to consider Parfit’s argument against the separateness of persons. I’ll argue that, insofar as Diamond cases are the litmus test for whether we should accept or reject the core concern people have with the separateness of persons, Parfit’s argument doesn’t go as far as he suggests. We can accept a view very much like the separateness of persons on his reductionist view of personal identity. Once this discussion is complete, we can return to the argument that, for those attracted to Parfit’s views about the relationship between prudence and personal identity, the truth lies somewhere between what I call the Asymmetry View and the New Symmetry View.

3.  Fission Cases The best way to understand Parfit’s argument against the separateness of persons is that it suffers from a presupposition error. If the category of ‘person’ is to cast some kind of morally relevant boundary, as the separateness of persons suggests, there had better be something morally significant about the category of ‘person’

10 Parfit, Reasons and Persons, 347.

Separating Persons  45 in and of itself. But, Parfit argues, the category of person is not in and of itself so morally relevant. His argument for this claim proceeds by showing that facts about personal identity over time don’t form the basis of a theory of prudence over time. Because prudence is usually thought to be what’s fundamentally at stake in cases of the intrapersonal distribution of goods, facts about personal identity over time are not relevant to intrapersonal distribution. Thus, Parfit reasons in some of the passages I’ve quoted, the supposed basis of the Asymmetry View is undermined if facts about personal identity over time do not ground facts about prudence over time. Parfit rejected the claim that relations of personal identity cast prudentially relevant boundaries in favor of the view that ‘relation-­R’ formed the basis of rational prudential concern about the future. Relation-­ R is a relation of overlapping chains of psychological connectedness and/or continuity formed by the right causes.11 It’s important to note that relation-­R could not, by itself, be a relation of personal identity. Why? Consider: Fission:  Ed enters the replicator machine. The replicator machine scans Ed’s body, produces two perfect replicas of him, simultaneously destroying his body. The two replicas of Ed are called ‘Ted’ and ‘Fred’ respectively. Both Ted and Fred wake up in the replicator machine with all of Ed’s beliefs, desires, intentions, and memories. To each of them, it feels like they were Ed, blinked, and are now on the other side of the room.

Relation-­R—so long as we follow Parfit in understanding ‘the right kind of cause’ as any cause whatsoever—holds between Ed and Ted as well as between Ed and Fred. That is, it holds one-­to-­many—from one individual at a time T1 to multiple individuals at another time T2. Identity relations can only hold one-­to-­one. This is, in part, because identity is a transitive relation: If α is identical to β and β is identical to γ, then α is identical to γ. Relation-­R is not transitive. It obtains between Ed and Ted as well as between Ed and Fred. But it does not obtain between Ted and Fred. Relation-­R is therefore not a relation of identity. Ed, according to Parfit, should think of Ted (and Fred for that matter), prudentially speaking, as if he were going to be Ted. Parfit was right, so far as it goes, that his views about personal identity and prudence are relevant to the separateness of persons. If the category of ‘person’ is not itself morally relevant, this does seem to count against the idea that the boundaries between people could matter; morally irrelevant categories don’t have morally relevant boundaries. But Parfit’s point doesn’t carry him as far from the Separateness of Persons as he seems to think. For, as we’ve just seen, he recognizes

11 Parfit, Reasons and Persons, 215.

46  James Goodrich another possible boundary-­ setting relation in place of personal identity: relation-­R. The lesson to be drawn is thus not that boundaries between individuals are morally irrelevant, but that the morally relevant boundaries between individuals are not grounded in relations of personal identity. The Parfitian view can make room for the ‘separateness of R-­related beings’. Though he does not say so explicitly, perhaps this is part of what Parfit meant when he says claims that what appears to be great imprudence is actually morally wrong. Reconsider Intrapersonal Coin-­Flip and Interpersonal Coin-­Flip. Assume that Anne is R-­related to herself today and on Monday, today and on Tuesday, and on Monday and Tuesday. Beth and Carol, however, are not R-­related across the outcomes in Interpersonal Coin-­ Flip. Our intuition that there is something different about the trade-­offs in the distribution of goods in Intrapersonal Coin-­ Flip and Interpersonal Coin-­Flip can thus be explained by the prudential and moral significance of the R-­relation. So while Parfit is correct to reject the separateness of persons, he can do so while accepting the intuition that supports the separateness of persons. One might think that the separateness of R-­related beings will support the Asymmetry View, thereby forcing Parfit to ally himself with those who have appealed to the separateness of persons as an objection to traditional forms of consequentialism which would accept the Old Symmetry View. In the next two sections, I will argue that this isn’t quite right. While Parfit should reject the Old Symmetry View, he should accept that the truth lies between the Asymmetry View and the New Symmetry View.

4.  Diamond-­Fission Cases I’ve introduced two kinds of case. The first kind of case—Diamond Cases—are really a pair of cases. The first of the Diamond Case pair involves an intrapersonal trade-­off, while the second of the Diamond Case pair involves an interpersonal trade-­off. The point of the Diamond Cases is to illustrate that most of us believe there is some important difference in the distribution of goods in intrapersonal and interpersonal cases. The second kind of case I’ve introduced is Fission cases. Parfit used Fission cases to support his views about personal identity and prudence. Such cases also nicely demonstrate that the R-­relation—the relation Parfit believes to form the basis of rational prudential concern over time—is non-­ transitive and therefore not an identity relation. If we combine Diamond Cases with Fission Cases, we can see why Parfit has reason to reject the Asymmetry view. Here’s one way to combine the two kinds of case: Diamond-­Fission:  Ed can walk into one of two replicator machines, φ or ψ. If Ed walks into φ, then either (i) Ted will have a hundred years of good life and

Separating Persons  47 Fred will have a hundred years of similarly good life or (ii) Ted will have fifty years of good life and Fred will have fifty years of similarly good life. There’s a 50 per cent chance of each outcome occurring. If Ed walks into ψ, then no matter what, Ted will have a hundred years of good life and Fred will have fifty years of similarly good life.

Intuitively, does it matter whether Ed picks φ or ψ? It seems to me that it does. Ed should walk into φ. Ψ, after all, gives Fred a fair shot at a hundred years of good life while ψ does not. However, from the perspective of Ed who is making the decision, both Ted and Fred are R-­related to him. They are individuals for whom he should care in the same prudential way as he cares for himself. And yet, this prudential concern seems entirely susceptible to the fair-­shake intuitions in Interpersonal Coin-­Flip. Here’s a natural suggestion about what’s going on in this case. While it’s true that Ed should feel for Ted and Fred as he would feel for a future individual who was identical to him, it’s not true that Ted and Fred should feel that way about each other. After all, Ted and Fred don’t stand in relation-­R. That is, they are each psychologically continuous and connected by the right kind of cause to Ed, but not to each other. Thus the separateness of R-­related beings (or persons for that matter) is not fundamentally about the relationship that the individual making a decision bears to those affected by that decision. The separateness of R-­related beings instead applies only to those affected by a choice. On the Parfitian view, two individuals can fail to stand in the relevant prudence-­inducing relation to each other while nevertheless both standing in the relevant prudence-­inducing relation to the same third party. And this is because relation-­R is non-­transitive. Thus, on the Parfitian view, an agent can have prudential reasons to care about two distinct individuals for their own sake and yet the separateness of R-­related beings arises anyway because those two distinct individuals don’t stand in the prudence-­grounding relationship to each other. The lesson here, for the Parfitian view, is simply that the distinction between intrapersonal and interpersonal cases is far messier than the more traditional views would have us believe. Our intuitions about Diamond-­ Fission support the first premise in an argument that shows the Asymmetry View is on shakier ground, given Parfit’s views about personal identity and prudence, than it would first seem: the separateness of R-­related beings can obtain so long as the individuals affected don’t stand in relation-­R to each other, whether or not the individual making the decision stands in relation-­R to those affected. If that’s right, even when an individual stands in a prudence-­grounding relation to other individuals, this fact does not undermine the existence of the special concern felt by those who believe in the separateness of persons. Thus, the thought that there’s some deep divide between what one has reasons to do in interpersonal and intrapersonal cases is on shakier grounds that it may first seem.

48  James Goodrich However, the Asymmetry View is still an option. For the contrast between interpersonal and intrapersonal cases can be drawn along the lines of whether those affected by the given choice stand in the R-­relation or not. If they do, you don’t need to give the individuals a fair shake. If they don’t, you do need to give each individual a fair shake. In the next section, I will argue that Asymmetry View is false. The truth instead lies between the Asymmetry View and the New Symmetry View.

5.  Intrapersonal Separateness without Fission Suppose Zack is twenty years old and is making a big life choice. He will either become an academic or a lawyer. Each path has its own distinct virtues. Becoming an academic will allow Zack to travel, think about difficult questions, and read great books. The life of a lawyer, on the other hand, will allow Zack to help others with his legal expertise and provide him with a level of financial stability that will easily allow him to have a large family. Zack realizes that he will be a very different person if he decides to become an academic from the one he will be if he decides to become a lawyer. His tastes, political preferences, and even beliefs about the world are very likely to differ. And neither life strikes Zack as better than the other. There’s thus at least a poetic sense in which Zack the Academic and Zack the Lawyer are just different people even though either of them would in the relevant sense be Zack the Twenty-­Year-­Old. On the Parfitian view, this is more than poetry. Zack the Twenty-­Year-­Old would be R-­related to Zack the Academic if that’s what he chose. Similarly, Zack the Twenty-­Year-­Old would be R-­related to Zack the Lawyer if that’s what he chose. But Zack the Academic is not and would not be R-­related to Zack the Lawyer. And this isn’t only because Zack the Academic and Zack the Lawyer fail to both exist. For illustrative purposes, suppose that Modal Realism—the thesis that all possible worlds are equally concretely real—is true. If that’s right, then Zack the Academic and Zack the Lawyer both exist, just at different worlds. Of course, these two individuals could not fulfil the causal condition on relation-­R since there are no trans-­world causal relations. But even if there could be, Zack the Academic and Zack the Lawyer are not psychologically continuous or connected across those possible worlds. Now these same claims could be mirrored by a more modest metaphysical picture which treated those worlds as abstract objects or fictions. In either case, relation-­R wouldn’t hold among the abstracta or within the fictions. Thus, there’s no meaningful sense in which Zack the Academic and Zack the Lawyer would stand in the right kind of prudential relation to each other. If what I’ve said in the previous paragraph is correct, then from the standpoint of the separateness of R-­related beings, Zack the Twenty-­Year-­Old’s choice is relevantly like Diamond Fission. That is, choices regarding how one could

Separating Persons  49 potentially distribute goods across radically different lives one might lead is not importantly different than choices regarding how to distribute goods across two R-­related beings that will actually exist. Think about it like this. Zack the Twenty-­ Year-­Old is making a choice which affects two potential individuals: Zack the Academic and Zack the Lawyer. Zack the Twenty-­Year-­Old thus has reason to give both Zack the Academic and Zack the Lawyer a fair shake, whatever that means in this case. Perhaps this means one is, ceteris paribus, required to flip a coin in this case. (Ruth Chang claims that it runs counter to the phenomenology of hard life choices to use a randomizing procedure to decide them.12 However, if the reasoning I’ve proffered thus far is compelling, we should perhaps jettison Chang’s intuition.) Moreover, if I’m right about Diamond-­Fission, then the fact that Zack would be R-­related to Zack the Academic and Zack the Lawyer is irrelevant from the standpoint of whether Zack the Twenty-­Year-­Old has a reason to be concerned about the separateness of R-­related beings. After all, Ed was R-­related to both Ted and Fred. But intuitively, this fact was insufficient to undermine reasons that Ed had to give both Ted and Fred a fair shake. So the fact that Zack the Twenty-­Year-­Old would be R-­related to both Zack the Academic and Zack the Lawyer gives Zack the Twenty-­Year-­Old no less reason to give Zack the Academic and Zack the Lawyer a fair shake. Therefore, the separateness of R-­related beings arises within lives as well. Here’s an important caveat. For flipping a coin to be appropriate it must be that Zack the Academic and Zack the Lawyer are equally or sufficiently nearly equally well-­off. If Zack the Academic would have a wonderful life and Zack the Lawyer’s life would be terrible, then flipping a coin wouldn’t be appropriate. Reconsider Interpersonal Coin-­Flip. Suppose that if φ lands heads, Beth will still receive 10 units of welfare and Carol will still receive 5 units of welfare. But if φ lands tails, Beth will receive −100 units of utility and Carol will receive 10 units of welfare. In such a case, no one would think that we should flip φ on the grounds that it gives Carol a fair shake. Even if it would give her a fair shake, the normative force of this point is clearly outweighed by the potential costs to Beth. Thus the position defended in this section should not be mistaken for the highly implausible view that we should flip a coin no matter what the relevant costs would be to various non-­R-­related individuals. I don’t claim that the view I have stated in this section is intuitive. I instead claim that a series of powerful theoretical claims that are backed up by our ­intuitions lead those of us attracted to Parfit’s views of personal identity to this conclusion. And if that’s right, we have reason to think that the Asymmetry View, as it’s normally understood, is false. The separateness of R-­related being can arise within a life without fission. This suggests that the standard interpersonal/­ intrapersonal distinction cannot bear the explanatory burden the Asymmetry

12  See R. Chang, ‘Hard Choices’, Journal of the American Philosophical Association 3/1 (2017): 15.

50  James Goodrich View has asked of  it. Before considering this implication in more detail, let’s address some objections to the line of argument presented in this section.

6.  Objections to Intrapersonal Separateness One might claim that I’ve been conflating questions of prudence across times with questions of prudence across possibilities. These, one might insist, are simply different kinds of questions. Perhaps that’s right. But we need to hear more about why they’re so different. Here’s what seems pretty well settled that challenges this objection. If Parfit doesn’t believe that individual persons have bare essences which travel with them across times to underlie what matters about prudence, then he shouldn’t believe that such bare essences travel with the very same individuals across possibilities. In other words, it’s difficult to see the motivation for being a reductionist about persons across times, but a not a reductionist about persons across possibilities. Furthermore, it’s natural to appeal to facts about modality and relation-­R when explaining why a given option is what an individual has most prudential reason to choose. I have prudential reason to eat more healthfully because there would later be an R-­related individual to me who would benefit from doing so. In fact, it’s difficult to say most of what we want to say about prudence and relation-­R without appealing to what R-­relations would or would not obtain. Thus, to reject my claim, one must explain why relation-­R should be the relevant prudence-­grounding relation across times, but not across possibilities. Here’s a different objection. Zack the Twenty-­ Year-­ Old’s choice is clearly different than Diamond-­ Fission. In Diamond-­ Fission, Ted and Fred would concretely coexist at the same world. This isn’t true of Zack the Academic and Zack the Lawyer. Thus, Fred can have a ‘fair-­shake’ complaint against Ed because Fred will actually exist in either outcome. But if Zack the Academic were to exist then Zack the Lawyer would not. Thus, Zack the Lawyer couldn’t have a complaint against either Zack the Twenty-­Year-­Old or Zack the Academic and vice versa. What this objection suggests is that there may be a kind of intrapersonal variant of the more familiar interpersonal Non-­Identity cases introduced by Parfit.13 In interpersonal Non-­Identity cases, we’re usually asked to consider whether to have one of two children who are non-­identical. The first would lead a much better life than the second. In such cases, most of us believe you should have the child that has the life with more good in it. However, we cannot explain why this is the case by appealing to what would be better or worse for either child since if the first 13  See Parfit, Reasons and Persons, ch. 16 for the original Non-­Identity Problem. I owe the idea that there may be an intrapersonal analogue of the Non-­Identity Problem to Jeff McMahan’s third Rutgers Lectures, entitled ‘Killing, Saving, and Causing to Exist’ (manuscript).

Separating Persons  51 child exists the second does not and vice versa. Zack the Twenty-­Year-­Old’s choice is seemingly similar because Zack the Academic and Zack the Lawyer cannot coexist as separate entities. Therefore it’s incumbent upon me to explain why Zack the Academic would have a claim to a fair shake without appealing to the view that Zack the Academic has a complaint against Zack the Twenty-­Year-­ Old or Zack the Lawyer. Some interpersonal Non-­ Identity cases elicit fair-­ shake intuitions.14 For example: Non-­Identity Coin-­Flip:  You can flip one of two coins, φ or ψ. If φ lands heads, Anders will be born and will live a hundred years of high quality life. If φ lands tails, Bjorn will be born and receive thirty years of high quality life. If ψ lands heads, Anders will born and will live sixty-­five years of high quality life. If ψ lands tails, Bjorn will be born and live sixty-­five years of high quality life.

The expected utility of both coins is the same. Thus, from a purely utility ­maximization standpoint, it doesn’t matter which coin you choose. Intuitively, however, it seems like you should flip ψ. Ψ gives Bjorn a fair shake at receiving a much better life. Non-­Identity Coin-­Flip thus seems to mirror Interpersonal Coin-­Flip. (Intuitions about this case are, of course, compatible with our simply being risk-­averse. However, we can change the probabilities, remove knowledge of the probabilities entirely, or try other cases entirely to remove this potential confound in the case. For ease of presentation, however I will proceed with this simpler case.) If our fair-­shake intuitions persist in Non-­Identity Coin-­Flip, this suggests to me that whatever we care about when we care about fair shakes doesn’t have to do just with outcomes. That is, the separateness of R-­related beings should be understood as a deontic, not telic constraint. This is suggested by the fact that the expected utility of both coins is the same. While I concede that we cannot appeal to a relation between Anders and Bjorn, since only one will exist, we can appeal instead, not to what is brought about, but to how it is brought about. We can object to φ as a kind of act that violates the separateness of R-­related individuals. Similarly, we can object to Zack the Twenty-­Year-­Old deciding in some way other than simply flipping a coin. Just as the fact that makes no difference to a victim whether they were killed or merely allowed to die, so too does it make no difference to the value of the outcome which coin is flipped. But just as with the distinction between killing and letting die, we shouldn’t understand the separateness of R-­related beings as a distinction concerned with outcomes. It’s a

14  I owe both this case and the thought that fair-­shake intuitions can arise in Non-­Identity Cases to Daniel Ramoller.

52  James Goodrich distinction in how outcomes are brought about. The separateness of R-­related beings is, in other words, still a non-­consequentialist constraint. Of course, more needs to be said in order to properly draw the boundaries between what does and does not count as a fair shake. That the separateness of R-­ related beings would be understood as a non-­ consequentialist constraint, however, would not surprise those who appealed to the separateness of persons as an objection to traditional forms of consequentialism. Perhaps it’s more surprising that, as I understand it, the separateness of R-­related beings is an agent-­centred as opposed to victim-­ centred constraint. In any case, that the separateness of R-­related beings is an agent-­centred constraint that doesn’t apply directly to what is good or bad is sufficient to evade the objection at hand. Now, those who are happy to consequentialize non-­consequentialist constraints are free to do so here. This would make little difference since such a non-­ traditional version of consequentialism would need to define outcomes differently or adopt agent-­ relative values. With these objections answered, let’s move on to tease out the implications of the view of the separateness of R-­related beings that I have been describing.

7.  Between Asymmetry and New Symmetry I said earlier in this essay that the truth lies between the Asymmetry View and the New Symmetry View. What I’ve been arguing in the last two sections is that the Asymmetry View is not the whole truth. In this section, I will explain why I  believe the Asymmetry View to be part of the truth and thus why the New Symmetry View is not the whole truth. Reconsider Intrapersonal Coin-­Flip. In Intrapersonal Coin-­Flip, we have a decision to make between two distributions of goods within a single life. What defines these distributions are the days of the week on which the various units of welfare fall. Anne right now is R-­related to herself on Monday, Anne on Monday is R-­related to Anne on Tuesday, and Anne on Tuesday is R-­related to Anne right now. Thus, no concerns about the separateness of R-­related beings arises on this case. This is the truth in the Asymmetry View. So long as the R-­relations are indeed in place between the relevant individuals, we can maintain that there’s a difference between what is required when we distribute goods in interpersonal and intrapersonal cases. This allows us to capture the distinction illustrated by the most plausible versions of the Diamond cases. And insofar as we can capture this distinction, we can hold onto a form of the Asymmetry View, but one with a more limited scope than is usually assumed. Next, the relations that underlie relation-­R—relations of overlapping chains of psychological continuity and connectedness—are matters of degree. I can stand

Separating Persons  53 in weaker or stronger chains of this form. The more abrupt and wide-­sweeping the changes are in my psychology, the less strongly the relations underlying the R-­relation obtain between me and my future self. Thus, whether the relations underlying the R-­relation obtains between any two individuals is not an all-­or-­ nothing matter. This leaves us with at least two possibilities. The first is that there is a threshold of degrees of overlapping chains of psychological continuity and connectedness that must be met in order for relation-­R to obtain. The second possibility is that relation-­R itself comes in degrees. Of these two possibilities, I prefer the second. We should avoid appealing to arbitrary thresholds where we can and I don’t see a principled way to determine the relevant threshold in this case. I also don’t think there a particularly worrying implications of accepting that the prudence-­grounding relation comes in degrees. If I’m right that we should think of relation-­R as coming in degrees, this illuminates another way in which the truth lies between the Asymmetry View and the New Symmetry View. The degree to which we should be concerned about the separateness of R-­related beings will depend upon the degree to which beings are or are not R-­related. The weaker the R-­relation, the stronger our concern should be for giving the various parties a fair shake. In this way, we might think of Intrapersonal Coin-­Flip and Interpersonal Coin-­Flip as forming the opposite extreme ends of a spectrum. Intrapersonal Coin-­Flip is the end of the spectrum in which no rational concern for the separateness of R-­related beings arises. Interpersonal Coin-­Flip is the end of the spectrum in which the rational concern for the separateness of R-­related beings is at its strongest. Many cases of prudence, if not most, lie somewhere between these two ends of the spectrum. Moving from the separateness of persons to the separateness of R-­related beings thus supports a position that lies between the Asymmetry View and the New Symmetry View. This is because, on the Parfitian view, the category of ‘persons’ doesn’t carve at the joints of prudence and morality. Relation-­R does. Parfit was thus correct to doubt the separateness of persons as it was originally conceived. And he was thus right to say, There is still a difference between my life and the lives of other people. But the difference is less. Other people are closer. I am less concerned about the rest of my own life, and more concerned about the lives of others.15

Those affected by my choices are often not so strongly R-­related. Possible future versions of myself and others are all owed a fair shake.

15 Parfit, Reasons and Persons, 347.

54  James Goodrich

Bibliography Chang, R., ‘Hard Choices’, Journal of the American Philosophical Association 3/1 (2017): 1–21. Diamond, P., ‘Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparison of Utility: Comment’, Journal of Political Economy 75 (1967): 765–6. Gauthier, D., Practical Reasoning (Oxford: Clarendon Press, 1963). Harsanyi, J., ‘Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparisons of Utility’, Journal of Political Economy 63 (1955): 309–21. McMahan, J., ‘Killing, Saving, and Causing to Exist’ (manuscript). Nagel, T., The Possibility of Altruism (Oxford: Oxford University Press, 1970). Nozick, R., Anarchy, State, and Utopia (New York: Basic Books, 1974). Parfit, D., Reasons and Persons (Oxford: Oxford University Press, 1984). Rawls, J., A Theory of Justice (Cambridge, MA: Harvard University Press, 1971). Williams, B., ‘A Critique of Utilitarianism’, in J.  J.  C.  Smart and B.  Williams (eds.), Utilitarianism: For and Against (Cambridge: Cambridge University Press, 1973), 77–150.

3

Personal Identity and Impersonal Ethics Tim Campbell

On what Parfit calls the Reductionist View although persons exist, the fact of a person’s existence, and the fact of her identity over time, just consist in the holding of certain more particular facts about physical or mental events and the relations between these events. Moreover, these more particular facts are impersonal—they do not have persons, or mental subjects, as constituents, and hence, do not presuppose or entail the existence of any person or mental subject.1 In contrast, the Non-­Reductionist View holds that persons exist and the Reductionist View is false—either the fact of a person’s existence doesn’t just consist in any more particular facts, or, any more particular facts in which this fact consists imply, or presuppose, the existence of some person, or some mental subject. In Reasons and Persons Parfit claims that the distinction between the Reductionist and Non-­Reductionist Views has moral significance. He claims that if the Reductionist View is true and the Non-­Reductionist View is false, then ‘it is . . . more plausible to focus, not on persons, but on experiences, and to claim that what matters morally is the nature of these experiences’.2 But it is not clear how a metaphysical thesis like the Reductionist View can intrude so violently into Derek Parfit’s ground-­breaking work on personal identity and population ethics inspired this chapter. I was introduced to Derek’s work as an undergraduate when Reasons and Persons was recommended to me. The book inspired me to continue in philosophy. But then I never imagined I would have the opportunity to meet Derek and interact with him in an academic setting. To have had this opportunity is a great privilege. I am eternally grateful to Derek for many things—for the philosophical discussions we have had since I first met him in 2008, for his invaluable comments on papers I wrote for his seminars, for serving on my dissertation committee, and for being a constant source of ­philosophical inspiration. While writing this chapter, I benefited greatly from discussions with Vuko Andric, Gustaf Arrhenius, Paul Bowman, Jonathan Dancy, Krister Bykvist, Tomi Francis, Karim Jebari, Anders Herlitz, Frances Kamm, Julia Mosquera, Wlodek Rabinowicz, and Larry Temkin. I am grateful to James Goodrich and Theron Pummer for written feedback on earlier drafts of this chapter, and especially to Jeff McMahan for providing many helpful comments and suggestions while I was developing my ideas and arguments. Finally, I would like to ac­know­ledge generous funding from The Swedish Research Council (421-­2014-­1037) and from Riksbankens Jubilemsfond (M17-­0372:1). 1 Parfit, Reasons and Persons, ‘How We Are Not What We Believe’, 225–6 (page references to Parfit 1984 will be to Reasons and Persons). Parfit clearly thinks that none of the particular facts by itself presupposes or entails the existence of any person. He seems to think that even the conjunction of all these particular facts doesn’t presuppose or entail the existence of any person. However, elsewhere he claims that if we know all of these particular facts then we can infer the existence of a person if we also possess the concept person (213). 2  Reasons and Persons, 446. Tim Campbell, Personal Identity and Impersonal Ethics In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0004

56  Tim Campbell morality.3 Parfit’s main reason for thinking it can is that the Reductionist View renders facts about personal identity ‘less deep’ metaphysically than does the Non-­Reductionist View.4 But as Parfit’s critics have pointed out, it is not obvious that the moral importance of persons diminishes with the metaphysical depth of facts about personal identity.5 I’m not sure whether the distinction between the Reductionist and ­Non-­Reductionist Views has the moral significance that Parfit claims it has. But, I will argue, there is a metaphysical distinction in the same ballpark that does. This is the distinction between Psychological Reductionism, a version of the Reductionist View that assumes a psychological criterion of personal identity, and what I call Subject-­Involving Views.6 According to the latter, psychological phenomena have mental subjects as constituents. Hence, on Subject-­Involving Views, the existence of psychological phenomena implies (or presupposes) the existence of some mental subject; facts about psychological phenomena are not, as Psychological Reductionism implies, impersonal. Thus, Subject-­ Involving Views are non-­reductionist insofar as they rule out Psychological Reductionism. Indeed, in Reasons and Persons Parfit classifies such views as non-­reductionist, and his classification is introduced at a point in the book after he argues for Psychological Reductionism, and at which he seems to use the term ‘the Reductionist View’ as a label for Psychological Reductionism.7 I will argue that Psychological Reductionism, in contrast with Subject-­ Involving Views, supports what I call The Impersonal Value of Experience:  If two outcomes differ in their experiential properties, then this difference by itself can make one of these outcomes worse (or better) than the other. 3  For more on this issue, see Adams (1985: esp. 457–8). 4  Reasons and Persons, 346. 5  The critics include Adams (1985: 455–8), Brink (1997: 117–18), Johnston (1992: 592), Korsgaard (1989: 108–9), and Wolf (1986: 705). Holtug (2010: 298–302) might also be among the critics. He rejects a Parfitian argument for the claim that the Reductionist View renders ethics more impersonal because it undermines principles of inter-­temporal compensation. Johnston (2016) has recently argued that even if Parfit’s claim about the ethical significance of the Reductionist View is false, the metaphysics of persons has important implications for ethics. Specifically, Johnston argues, recognizing four-­ dimensional mereological sums of person-­stages would undermine many aggregative ethical theories as well as certain claims about inter-­temporal compensation. 6  Here I follow Eric Olson (1997: ch. 2) in adopting a very broad construal of ‘the psychological criterion of personal identity’. I understand the psychological criterion as a broad disjunction of more specific criteria that includes non-­branching psychological continuity but also, for example, the continuous physical realization of a minimal capacity for consciousness (see e.g. Unger 1990: 116, 147–52, and McMahan 2002: 63). The psychological criterion of personal identity, as I understand it, might even include what Parfit calls the Physical Criterion, according to which what is necessary for personal identity ‘is not the continued existence of the whole body, but the continued existence of enough of the brain to be the brain of a living person’ (Reasons and Persons, 204). The continuous existence of ‘enough of the brain’ might involve the continuous existence of the capacity for consciousness. In that case, the Physical Criterion would be a version of the psychological criterion broadly construed. 7  For Parfit’s classification, see Reasons and Persons, 225.

Personal Identity and Impersonal Ethics  57 The locution ‘by itself ’ implies that the difference in experiential properties between the two outcomes can make one worse (or better) than the other ­independent of any other potentially evaluatively significant difference between these outcomes, such as a difference in individual well-­being. When I say that Psychological Reductionism in contrast with Subject-­Involving Views supports the Impersonal Value of Experience, I mean specifically that the Impersonal Value of Experience is more plausible (more likely to be true) if Psychological Reductionism is true than if any Subject-­Involving View is true. Given plausible assumptions about the relation between outcome value and morality, if a difference in experiential properties can, by itself, make one ­outcome better than another, then such a difference can, by itself, matter morally. For example, it seems plausible that when other things are equal, the fact that one outcome is better than another implies that one has more moral reason to bring about the former. Thus, other things being equal, the Impersonal Value of Experience supports the claim that an experiential difference between two outcomes can by itself provide a moral reason to choose the better of those outcomes. Here is my central argument for the claim that Psychological Reductionism, in contrast with Subject-­Involving Views, supports the Impersonal Value of Experience: P1.  Psychological Reductionism (in contrast with Subject-­Involving Views) supports what I call Shared Experience: Each of two different persons can be a subject of one and the same token experience. P2.  Shared Experience (in contrast with its negation) supports the Impersonal Value of Experience. P3.  If P1 and P2 are true, then Psychological Reductionism (in contrast with Subject-­Involving Views) supports the Impersonal Value of Experience. Therefore, C:  Psychological Reductionism (in contrast with Subject-­Involving Views) supports The Impersonal Value of Experience.

Section 1 further characterizes Psychological Reductionism and Subject-­Involving Views. Sections 2–4 defend, respectively, P1–P3 of my central argument. Section 5 considers and rejects an argument for the claim that even if Psychological Reductionism supports the Impersonal Value of Experience, the degree of ­support is at best very weak and so the conclusion of my central argument has little theoretical or practical significance. Section 6 discusses some implications of the argument’s conclusion for the evaluation of outcomes and for morality. Section 7 concludes.

58  Tim Campbell

1.  Psychological Reductionism and Subject-­Involving Views Suppose there exists a person, S. On the Reductionist View, the fact that S exists just consists in certain more particular facts, f1…fn. And f1…fn are impersonal; none of f1…fn includes any person, or subject, as a constituent. The Reductionist View provides a similar treatment of other facts regarding S, such as the fact that S exists at some particular time, and the fact that S persists from one time to another. This raises at least three questions. First, what is a person—what kind of object is the Reductionist View about? Second, what are the ‘more particular facts’ in which the fact of a person’s existence (and other related facts, such as the fact that the person persists from one time to another) just consists? Third, what is it for a fact to just consist in certain other facts? The third question is perhaps the hardest to answer. Parfit seems to think of the relation ‘just consists in’ as weaker than identity and stronger than both supervenience and ontological dependence. However, he does not provide a detailed account of this relation in Reasons and Persons or in later work.8 For the purposes of this paper, I shall remain neutral regarding how we should understand the relation ‘just consists in’ and shall focus mainly on the first and second questions stated above. What is a person? The language that Parfit uses throughout part III of Reasons and Persons suggests that he intends the Reductionist View to be about our nature and identity, about what we are. This suggests that for Parfit ‘persons’ refers to whatever ‘we’ refers to. And he seems to think that ‘we’ refers to humans that are persons in Locke’s sense—i.e. in the sense of a thinking, intelligent being with reason and reflection ‘that can consider itself as itself the same thing in different times and places’.9 That we are persons in Locke’s sense doesn’t tell us what our general identity conditions are. It doesn’t tell us whether, for example, we could survive Star Trek-­style teletransportation, or whether we once existed as two-­ week-­old embryos.10 To answer such questions, reductionists need an answer to my second question: what are the more particular facts in which the fact of a person’s existence, and other related facts, such as the fact that she persists from one time to another, consist?

8  See the three Reductionist claims stated in Reasons and Persons, 211 as well as Parfit’s discussion (on the same page) of these claims. An expanded discussion of the relation ‘just consists in’ can be found in ‘Experiences, Subjects and Conceptual Schemes’ (Parfit 1999). There, Parfit suggests that the relation ‘just consists in’ is analogous to the relation of material constitution that is sometimes thought to hold between a statue and the matter of which it is made. However, the discussion in ‘Experiences, Subjects and Conceptual Schemes’ adds little in terms of substance to the one in Reasons and Persons. 9  Locke (1975: 39). 10  Human embryos are not persons in Locke’s sense, but this fact does not imply that no Lockean person was ever a human embryo. Analogously, human embryos are not moral philosophers, but this doesn’t imply that no moral philosopher was ever a human embryo.

Personal Identity and Impersonal Ethics  59 Some reductionists, including Parfit, endorse Psychological Reductionism. They claim that for any persisting person (or mental subject), the facts of her existence and identity over time consist in facts about particular psychological phenomena, such as token experiences, memories, beliefs, desires, and certain relations between these phenomena.11 Given that the psychological phenomena in which a person’s identity consists include token experiences, according to Psychological Reductionism, facts about token experiences are impersonal—they do not ­presuppose or imply the existence of any person or mental subject. This implication of Psychological Reductionism is controversial. For instance, in a critical notice of Reasons and Persons, Sydney Shoemaker objects to Psychological Reductionism on the grounds that some facts about token experiences are not impersonal, that they imply the existence of a mental subject. For example, he claims, many experiences are experiencings, and hence, are adjectivally ­dependent on mental subjects in the way that seeings are adjectivally dependent on seers and deeds are adjectivally dependent on doers.12 Elsewhere, Shoemaker endorses the stronger view that any experience is an experiencing by a subject and is adjectivally dependent on that subject.13 Other philosophers, such as Galen Strawson, also endorse this stronger view. Indeed, Strawson claims that the view is ‘obvious’.14 Some controversial views of experience echo the view endorsed by Shoemaker and Strawson. Take, for example, adverbialist views of experience. On such views, sentences that seem (in light of their surface grammar) to assert or presuppose the existence of an experience qua object and to attribute certain properties to this object should be understood instead as expressing ontological commitment only to a subject who is experiencing in a certain way. For example, the sentence ‘there is a painful experience’ should be understood as expressing commitment only to a subject who experiences in ‘a painish way’, or who, as Chisholm would say, senses painly.15 Other views include those that treat token experiences as Kimian events—i.e. structured entities consisting of a particular object, a property that this object exemplifies, and a time of exemplification.16 On such views, the object in question is an individual (e.g. a person), and the property is that of being in a certain mental state—a state such that there is something it is like to be in that 11  For examples, see Reasons and Persons, 219–26. Psychological Reductionism isn’t the only version of the Reductionist View on offer. One could, for example, be a biological reductionist—i.e. someone who claims that the facts of our existence and identity over time consist only in biological facts that do not include any facts about psychology. Here I shall ignore such alternatives and focus only on Psychological Reductionism. 12  Shoemaker (1985: 446). 13  Shoemaker (1986: 10). 14  Strawson (2017: ch. 13, esp. pp. 253–59). E. J. Lowe raises a similar criticism of Psychological Reductionism. He argues that an experience cannot be identified and individuated from certain other experiences without reference to a unique person who is the subject of that experience, and that therefore the project of constructing persons (or selves) out of experiences and other psychological states is doomed by ‘fatal circularity’. See Lowe (1996: 8). 15  Chisholm (1963: 97–112). 16  Zimmerman (2010: 119–50), and see Kim (1976).

60  Tim Campbell state. Whenever there is an individual who exemplifies this property there is an experience—a particular event consisting of the individual, the property, and the time at which the former exemplifies the latter. The views described above are Subject-­Involving Views. On these views, a token experience intimately involves a particular subject; for example, it is an instance of that subject experiencing, or is made up of that subject and a property that she instantiates. On Subject-­Involving Views, facts about experiences include subjects as constituents. For example, if the sentence ‘there is a pain experience’ expresses a fact, then on the adverbialist view of experience, it expresses the fact that there exists a subject who is experiencing ‘painly’, and on the view of experiences as Kimian events it expresses the fact that there exists a subject who (at some time) instantiates the property of being in pain. Parfit rejects Subject-­Involving Views. His Psychological Reductionism appears to treat persons (and other subjects) as mere logical constructions out of experiences. Following Georg Christoph Lichtenberg, Parfit suggests that he can recognize the existence of experience (thought, perception, etc.) without being committed to the existence of any individual that is experiencing (thinking, perceiving, etc.). This is similar, Lichtenberg and Parfit thought, to how one can recognize the existence of thunder without being committed to the existence of any individual that is thundering.17 The distinction between Psychological Reductionism and Subject-­Involving Views has relevance for what determines the value of outcomes. Specifically, if Psychological Reductionism is true, and Subject-­Involving Views are false, then there is more support for what I call the Impersonal Value of Experience: if two outcomes differ in their experiential properties, then this difference by itself can make one of these outcomes worse (or better) than the other. In Section 2 I turn to my argument for this claim.

2.  P1: Psychological Reductionism (in contrast with Subject-­Involving Views) Supports Shared Experience In this section, I defend P1 of my argument. Psychological Reductionism, in contrast with Subject-­Involving Views, supports Shared Experience: each of two different persons can be a subject of one and the same token experience. Let us first consider what an instance of Shared Experience might look like. Suppose that two distinct persons are connected at the head and share certain parts of the brain. These shared brain-­parts (neurons, synapses, etc.) give rise to a 17  See Parfit, ‘Experiences, Subjects, and Conceptual Schemes’, 233 and 238 for discussion of the thundering example. This example was originally provided by Lichtenberg in his Schriften und Briefe and later modified and reintroduced by Parfit.

Personal Identity and Impersonal Ethics  61 particular experience, and the neural correlates of the experience involve only these shared parts. There is no neuroscientific or psychological basis for assigning the experience only to one of these persons or only to some other person, and there are no other ostensible relational properties in virtue of which the experience can be assigned only to one of these persons or to some other person. For example, each person stands in (roughly) the same relation to her surrounding environment, and each exhibits behavior consistent with her having an experience of the ­relevant type. It would seem that on pain of arbitrariness the experience should be assigned to each person. We should say that each is a subject of one and the same token experience.18 A potential real-­world example of the kind that I have described involves the genetically identical conjoined twins Tatiana and Krista Hogan. No one doubts that the names ‘Tatiana Hogan’ and ‘Krista Hogan’ refer separately to two persons—two young girls—each with her own body and her own unique psychological tendencies. Yet, because their cerebrums are partially fused within a single cranium, and connected neurologically, sometimes these twins appear to have the same experience—not just the same type of experience but the same token experience. For example, if one’s eyes are covered and the other is shown an object, e.g. a stuffed animal, the former can, reliably, name it, suggesting that one’s visual experience of the object is, at the same time, had by the other. If one eats or drinks something they both like, both express pleasure. This suggests that they may have shared gustatory experiences. And if one is stuck in the arm with a needle both wince at the same time, indicating what may be a shared pain experience.19 The case of the Hogans may not be a case of the kind that I have described. There are alternative interpretations of the phenomena. Perhaps when one twin appears to have the same token experience as the other, what really happens is that external stimuli are transmitted from one twin’s sensory organs into unshared parts of the other’s brain as well as unshared parts of her own brain. If this is true, then rather than a single neural event that involves only parts of each twin, there might instead be two different but very similar neural events, each involving parts of only one twin. But even if the case of the Hogans is not a case of the kind that I have described, such a case is metaphysically possible. We can imagine a case— call it the Twins Case—involving genetically identical conjoined twins T and K. These  twins are like the Hogan twins in that each has her own body and psychology, and each initiates some thoughts and actions independent of the other. But T and K are certainly also related in the way that I described: in at least 18  For another purported example of shared experience that is much more elaborate and stranger than mine, see Unger (1990: 177–84). 19  These examples are from Hershenov (2013), who uses the example of the Hogans in an argument against certain views of personal identity. For a first-­hand account of many examples of what may be a single experience had by each twin, see Dominus (2011).

62  Tim Campbell one instance, the neural correlates of a token experience involve only parts of both twins, and there are no ostensible neurological or psychological facts in virtue of which the experience is had only by T or only by K, or only by some other person.20 I have described what seems to me to be an instance of Shared Experience. However, if any Subject-­Involving View is true, then neither the Twins Case nor the actual examples in the lives of the Hogan sisters include any instances of Shared Experience. For example, in the Twins Case, when each of T and K has an experience, there are at least two experiencings, and hence, at least two experiences. Let’s suppose that the neural correlates of an experience that involve only parts of both T and K are correlates of a certain pain experience. Then those who accept an adverbialist view of experience will claim that what I am calling a pain experience should be understood as an instance of a person experiencing ‘painly’. And there are two instances of a person experiencing painly in this case— one instance of T experiencing painly and one instance of K experiencing painly. For the adverbialist, there are two pain experiences, not one, and each pain experience is associated with only one person; there is no genuine sharing of experience. Next, consider the view of experiences as Kimian events. On this view, a token pain experience consists of a particular person and the property of being in pain that she exemplifies at a certain time. In the Twins Case, T and K both exemplify, at the same time, the property of being in pain. So on the view of experiences as Kimian events, there are at least two pain experiences in this case—one consisting of T and the property of being in pain at a certain time, and one consisting of K and the property of being in pain at that time. T and K are subjects of ‘the same pain’ only in that each exemplifies, at the same time, the same property of being in (a certain kind of) pain. But this is not an instance of genuine experience sharing. In fact, all Subject-­Involving Views are incompatible with Shared Experience. If Shared Experience is true, then it is metaphysically possible that there is at most one experience and two persons each of whom is its subject. But for reasons of the kind that we just saw, on Subject-­Involving Views, this is not possible. On

20  One could come up with interpretations of the Twins Case and the actual case of the Hogan sisters according to which these cases involve more, or less, than two persons. For example, one could claim that these cases are plausibly described as having only one person who is controlling two bodies, and who is at least partially cut off from herself psychologically. Alternatively, one could claim that these cases are plausibly described as having three persons, one associated most closely with a certain body, another associated most closely with the other body, and a third person who is associated in roughly the same way with each body. This third person might be made up of both bodies and therefore be larger than either of the two other persons. Or she might be smaller than either of the two other persons—for example, she might be lodged somewhere inside the skull cavity. But I find these interpretations extremely implausible.

Personal Identity and Impersonal Ethics  63 such views, if there are at least two experiencers then there are at least two experiences.21 On the other hand, Psychological Reductionism is compatible with Shared Experience. Consider, for instance, a version of Psychological Reductionism that identifies a person with a certain ‘bundle’ of experiences.22 It is generally true that an event of a certain kind can be part of two distinct bundles of events of that kind. The event known as the Battle of Stalingrad is part of the bundle of events known as the Great Patriotic War, but it is also part of the bundle of events known as World War II.23 Or, building on one of Parfit’s examples, suppose that a river is just a bundle of particular movements of water connected in a certain way. A particular movement of water (an event) could be part of two intersecting rivers (two bundles of connected movements of water) if it occurs right at their intersection.24 Subject-­ Involving Views rule out Shared Experience, but Psychological Reductionism does not. If Psychological Reductionism is true, then Shared Experience has some chance of being true. But if any Subject-­Involving View is true, Shared Experience has no chance of being true. Thus, P1 of my central argument is true: Psychological Reductionism, in contrast with Subject-­Involving Views, supports Shared Experience.

21  One might question this claim. For example, one might wonder whether, even if a Subject-­ Involving View is true, it is nevertheless possible for multiple persons to jointly ‘participate’ in a single token experience, just as it is possible for multiple persons to jointly participate in a token event of some other kind, such as a barn-­raising. This comparison is not apt, however. When multiple persons participate in a barn-­raising, it is not the case that the entirety of the barn-­raising is attributable to each person. Rather, each person does something that constitutes part of the barn-­raising. The Twins Case is not a case in which each of multiple persons does something that is part of an experience. Rather, each person experiences. But what if each part of a token barn-­raising is causally overdetermined by the activities of multiple individuals, such that each raises the barn in its entirety? Wouldn’t there then be only one token ­barn-­raising attributable to each of these individuals? This analogy is illuminating only if a token barn-­raising intimately involves a person in the way that, according to Subject-­Involving Views, a token experience intimately involves a person. Let’s suppose, then, that a token barn-­raising is just an instance of a particular person raising a barn (the adverbialist view of barn-­raisings), or a complex entity consisting of a particular person, the property of raising a barn, and a time at which the person exemplifies this property (a Kimmian event). Then if it is true of each of multiple persons that she raises the barn in its entirety, then there are in fact multiple barn-­raisings in this example (but only one barn)—for there are multiple instances of a person raising a barn, and multiple complex entities consisting of a person, the property of raising a barn, and a time at which the person exemplifies that property. The barn-­raising example does not help us defend the compatibility of Shared Experience and Subject-­Involving Views. 22  The view that a person is a certain bundle of perceptions or other mental events is commonly referred to as ‘the bundle view’. It is often attributed to Hume, and is sometimes classified as a version of the Reductionist View. See e.g. Lowe (1996: ch. 2). 23  Thanks to Karim Jebari for this example. 24  Parfit, ‘Experiences, Subjects, and Conceptual Schemes’, 228.

64  Tim Campbell

3.  P2: Shared Experience Supports the Impersonal Value of Experience In this section, I argue for P2: Shared Experience supports The Impersonal Value of Experience: If two outcomes differ in their experiential properties, then this difference by itself can make one of these outcomes worse (or better) than the other. Let us revisit our twins T and K, and consider two different possible ­outcomes in which T and K are the only persons who exist. The first outcome is what I will call SHARED PAIN. In this outcome, T and K share certain parts of the brain that are involved in the production of an intensely unpleasant pain experience, such as the pain associated with a pounding migraine headache. Each person has exactly one unpleasant pain experience. And each person seems to be a subject of the same token unpleasant pain experience—what appears to be the only token unpleasant pain experience in that outcome. I write ‘seems to be’ and ‘appears to be’ because whether it is true that there is only one token unpleasant pain experience and that each person is the subject of this experience depends on whether Shared Experience is true (more on this below). The second outcome is what I will call TWO PAINS. In this outcome, there are exactly two token unpleasant pain experiences. T is the only subject of one of these experiences and K is the only subject of the other. Moreover, each of these pain experiences is equally as unpleasant as (what appears to be)  the only token pain experience in SHARED PAIN. So, while (it appears that) each person is equally well off in TWO PAINS as she is in SHARED PAIN, TWO PAINS (it appears) has more unpleasant pain experience than SHARED PAIN. Let us stipulate that the only potentially evaluatively significant difference between SHARED PAIN and TWO PAINS concerns their respective amounts of unpleasant pain experience. We cannot point to any other potentially evaluatively significant difference between these outcomes. For example, SHARED PAIN and TWO PAINS are equally good with respect to each dimension of T’s well-­being and each dimension of K’s well-­being; hence, they are equally good for each person with respect to mental state welfare, desire-­satisfaction, virtue, autonomy, personal achievement, etc. Hence, each person has the same level of well-­being in SHARED PAIN as in TWO PAINS. Moreover, there is perfect equality of well-­being in both outcomes. Thus, SHARED PAIN and TWO PAINS are equally good with respect to both total and average well-­being, as well as priority for the worse-­off. They are also equally good with respect to equality of opportunity, desert, respect for nature, and a multitude of other potentially evaluatively ­significant factors. We can imagine that in TWO PAINS (unlike in SHARED PAIN), there is no physical overlap between T and K. Figure 3.1 provides a visual representation.

Personal Identity and Impersonal Ethics  65 SHARED PAIN

TWO PAINS

Figure 3.1 

We may find it hard to believe that SHARED PAIN and TWO PAINS are equally good with respect to each dimension of each person’s well-­being. We might think that being physically conjoined with another person diminishes one’s autonomy and is therefore worse for one, other things being equal. However, we can also imagine that in TWO PAINS each person’s autonomy is diminished to the same degree as in SHARED PAIN, albeit for reasons unrelated to her being physically conjoined with someone else. Alternatively, we can imagine that in each outcome T and K are conjoined in whatever way reduces personal autonomy, but that in TWO PAINS each has an unpleasant experience that is realized in parts of her that are not parts of the other, such that she is clearly the only subject of that unpleasant experience. We can assume that SHARED PAIN and TWO PAINS are equally good (or bad) with respect to personal autonomy. I shall now present my argument for P2: Shared Experience supports the Impersonal Value of Experience. This argument concerns the relationship between Shared Experience and the following two claims: More Pain: There is more unpleasant pain experience in TWO PAINS than in SHARED PAIN. Worse: TWO PAINS is worse than SHARED PAIN.

Here is what I call The Shared Pain Argument: (1)  Shared Experience supports More Pain. (2)  More Pain supports Worse. (3)  If (1) and (2), then Shared Experience supports Worse. Therefore, (4)  Shared Experience supports Worse. (From (1)–(3)) (5) If (1) and (4), then Shared Experience supports the conjunction of More Pain and Worse.

66  Tim Campbell Therefore, (6) Shared Experience supports the conjunction of More Pain and Worse. (From (1), (4), and (5) ). (7) The conjunction of More Pain and Worse supports the Impersonal Value of Experience. (8) If (6) and (7), then Shared Experience supports the Impersonal Value of Experience. Therefore, (9) Shared Experience supports the Impersonal Value of Experience. (From (6)–(8)) As in this chapter’s central argument, the support relation mentioned in each premise of the Shared Pain Argument is contrastive. In the Shared Pain Argument, each instance of ‘p supports q’ means that q is more plausible (more likely to be true) given p than given p’s negation. In the remainder of this section, I defend the major premises of The Shared Pain Argument: (1)–(3), (5), (7), and (8). The argument for (1) is straightforward: given our stated background assumptions about SHARED PAIN and TWO PAINS, if Shared Experience is false then More Pain is false, but if Shared Experience is true, then More Pain is true. One of our background assumptions is that in SHARED PAIN there are two persons and each has an unpleasant pain experience. If Shared Experience is true, then each person has this unpleasant pain experience (and there are no other unpleasant experiences in the outcome). But if Shared Experience is false, then in SHARED PAIN there are two unpleasant pain experiences, each had by only one of the two persons. If that is true, then SHARED PAIN has just as many unpleasant pain experiences (two) as TWO PAINS. Since the pain experiences in each outcome are all equally unpleasant, More Pain is false; it is not the case that TWO PAINS has more unpleasant pain experience than SHARED PAIN. On the other hand, if Shared Experience is true, then in SHARED PAIN there is only one token unpleasant pain experience that is had by each of two different persons, and in TWO PAINS there are two token unpleasant pain experiences each had by only one person. Each of the two unpleasant experiences in TWO PAINS is equally as unpleasant as the one in SHARED PAIN, so if Shared Experience is true, TWO PAINS has more unpleasant pain experience than SHARED PAIN. Next, consider (2)—i.e. the claim that More Pain supports Worse. I suspect that Worse is a polarizing claim. Among those with whom I have discussed the Shared Pain Argument, some find Worse plausible, but others find this claim implausible. However, the case for (2) does not depend on the truth or plausibility of Worse. Those who find Worse implausible should accept the claim that Worse

Personal Identity and Impersonal Ethics  67 is less implausible if More Pain is true than if More Pain is false. After all, if More Pain is false, then SHARED PAIN and TWO PAINS contain the same amount of unpleasant pain experience, and hence, given our stipulation that the two outcomes are otherwise of equal value, there is no potentially evaluatively significant respect in which SHARED PAIN and TWO PAINS differ. If TWO PAINS has more unpleasant pain experience than SHARED PAIN, this provides some reason to believe that TWO PAINS is worse than SHARED PAIN. If More Pain is true, then Worse has some chance of being true; but if More Pain is false, then Worse has no chance of being true.25 This is sufficient to establish (2). However, while the degree of Worse’s plausibility is inessential for establishing (2), (2) may seem uninteresting to those who find Worse implausible. Suppose that you accept (2) based on the reasons stated in the previous paragraph. Still you might think that Worse is so implausible on its face that if More Pain increases the plausibility of Worse, it does so only infinitesimally. I think that this attitude is mistaken. I suspect that those who find Worse implausible are attracted to the view that how much goodness or badness an experience contributes to an outcome is determined by its relation to a person or personal perspective, such that when an unpleasant experience is had by two persons, this experience contributes twice as much badness to the outcome as an equally unpleasant experience had by only one person. I will argue in Section 5 that this view is problematic, and that Worse is more plausible than it may at first glance seem.26 Let us continue with the Shared Pain Argument. Having defended (1) and (2), I turn to (3): if Shared Experience supports More Pain, and More Pain supports Worse, then Shared Experience supports Worse. One should not be tempted by the thought that the support relation is generally transitive, such that for any claims x, y, and z, if x supports y, and y supports z, then x supports z. The support relation is non-­transitive.27 To borrow an example from Roche (2012), the claim that a certain card drawn at random from a standard playing card deck is a heart supports (probabilistically) the claim that the selected card is a red card. And the claim that it is a red card supports the claim that it is a diamond. However, the claim that the card is a heart does not support the claim that it is a diamond.28 25  Or, if no claim has zero chance of being true, then Worse has less chance of being true. 26  The claim that an experience contributes to outcome value in this way may lead to the problems of overcounting discussed in Johnston (2016). As Johnston points out, if persons have many overlapping person-­stages, then many of these person-­stages will have the person’s experiences; and if we care not only about the experiences themselves but the individuals that have them, then there may be no good reason for excluding these person-­stages in our calculation of the value of the outcome in which they exist. 27  For discussion of the transitivity of probabilistic support, see Roche (2012). Although Roche is concerned with probabilistic support in particular, it seems that the lessons of Roche’s discussion are quite general and carry over to claims about the structure of other forms of epistemic support (e.g. abductive support). 28  Roche (2012: 111–12).

68  Tim Campbell The reason that transitivity fails here seems obvious. Although the claim that the card is a heart supports the claim that it is a red card, which in turn supports the claim that it is a diamond, the claim that the card is a heart also rules out that it is a diamond, thereby making it less plausible that it is a diamond. Does Shared Experience, either by itself or in conjunction with More Pain make Worse less plausible? The answer seems to be ‘no’. Given our description of the two outcomes, as we saw, Shared Experience seems required for More Pain. And if More Pain were false, we would lose what seems to be our only source of support Worse. In this case (although not in general) transitivity holds. Thus, having established (1) and (2), we can infer (3). And (4) follows from (1)–(3). Next, consider (5): if Shared Experience supports More Pain and Shared Experience supports Worse then Shared Experience supports the conjunction of More Pain and Worse. Here, we cannot appeal to a general principle of the form: if x supports y, and y supports z, then x supports (y & z). This general principle is false. For example, suppose we have forensic evidence supporting the claim that Professor Plum is the killer, and suppose our evidence also supports the claim that Colonel Mustard is the killer. (Perhaps our evidence indicates that the killer is a man.) Our evidence might fail to support the conjunction of these two claims. This might be because our evidence supports the claim that the killer acted alone, which is incompatible with the conjunction (assuming the professor and the colonel are not identical). But again, our case is different from this one. Shared Experience does not undermine the conjunction of More Pain and Worse. Moreover, neither conjunct undermines the other. There are no impediments here to (5). We should accept it. And (6) follows from (1), (4), and (5). According to (7), the conjunction of More Pain and Worse supports the Impersonal Value of Experience. More Pain states that TWO PAINS has more unpleasant pain experience than SHARED PAIN. There is an experiential difference between the two outcomes. But we are assuming that SHARED PAIN and TWO PAINS are equally valuable in other respects. Hence, if, as Worse states, TWO PAINS is worse than SHARED PAIN, then it seems that what makes TWO PAINS worse than SHARED pain is the experiential difference between these outcomes. If this is true, then the Impersonal Value of Experience is true—an experiential difference between two outcomes can by itself make one of these outcomes worse (or better) than the other. Thus, (7) is true—the conjunction of More Pain and Worse supports The Impersonal Value of Experience.29 29  One could claim that if TWO PAINS is worse than SHARED PAIN then this is not only because, in the former outcome, there is one more unpleasant pain experience, but also because this experience is had by some person. One could claim that, therefore, although the experiential difference between the two outcomes accounts for their evaluate difference, this evaluative difference is not wholly impersonal. However, the evaluative difference between TWO PAINS and SHARED PAIN is, in an im­port­ ant sense, impersonal. In each outcome, the same persons exist and each has the same level of well-­being. If the experiential difference between TWO PAINS and SHARED PAIN makes the former outcome worse than the latter, this cannot be because it makes a difference to someone’s well-­being.

Personal Identity and Impersonal Ethics  69 Finally, according to (8), if Shared Experience supports the conjunction of More Pain and Worse, and this conjunction supports the Impersonal Value of Experience, then Shared Experience supports the Impersonal value of Experience. This claim is highly plausible even though, as we saw, support is not transitive in general. Transitivity would fail in this particular instance if Shared Experience or the conjunction of Shared Experience and More Pain, somehow undermined The Impersonal Value of Experience, or if Shared Experience somehow undercut the support relation between More Pain and The Impersonal Value of Experience. But this does not seem to be the case. Thus, (8) seems true. From (6), (7), and (8), we derive the conclusion, (9): Shared Experience supports the Impersonal Value of Experience.

4.  P3: If P1 and P2, then Psychological Reductionism Supports the Impersonal Value of Experience P3 is the least controversial premise in my central argument. It states that if Psychological Reductionism (in contrast with Subject-­Involving Views) supports Shared Experience, and Shared Experience (in contrast with its negation) supports the Impersonal Value of Experience, then Psychological Reductionism (in contrast with Subject-­Involving Views) supports the Impersonal Value of Experience. This inference would fail if Psychological Reductionism (in contrast with Subject-­Involving Views) somehow undermined the Impersonal Value of Experience, or if Psychological Reductionism (in contrast with Subject-­Involving Views) somehow severed, or weakened, the relation of support between Shared Experience and the Impersonal Value of Experience, or if the conjunction of Psychological Reductionism (in contrast with Subject-­ Involving Views) and Shared Experience rendered the Impersonal Value of Experience less plausible. But none of these claims seems true; and unless some such claim can be demonstrated, it seems we ought to accept the inference. We should think that in this instance, support carries over. This concludes my defense of P1–P3. Each premise seems highly plausible, and these premises jointly entail the conclusion of my central argument: Psychological Reductionism (in contrast with Person-­Involving Views) supports the Impersonal Value of Experience.

5.  A Doubt about the Significance of the Central Argument In Section 3, I briefly considered a view that challenges the theoretical significance of my central argument, as well as its conclusion. My conclusion depends on P2: Shared Experience supports the Impersonal Value of Experience. And my defense

70  Tim Campbell of P2 depends on what I called The Shared Pain Argument. A crucial premise of this argument is that More Pain, i.e. the claim that there is more unpleasant pain experience in TWO PAINS than in SHARED PAIN (an outcome in which each of two persons, T and K, is the subject of a single unpleasant pain experience), supports Worse, i.e. the claim that TWO PAINS is worse than SHARED PAIN. However, as I acknowledged, one might think that Worse is implausible. Perhaps the main source of resistance to Worse is based on the view that the value that any experience contributes to an outcome is determined by its relation to a person or personal perspective, such that an unpleasant experience that is had by each of two persons adds twice as much badness as an equally unpleasant experience had by only one person. Let us call this view of how experiences contribute to outcome value the person-­centered approach. The person-­centered approach suggests the following method for evaluating SHARED PAIN and TWO PAINS. First, identify the persons in those outcomes. Next, put yourself in the shoes of each person, one at a time, and ask whether, if you were her, you would have reason of an egoistic kind to prefer SHARED PAIN to TWO PAINS.30 When you do this, you should find that the answer is ‘no’. There is a one-­to-­one correspondence between SHARED PAIN and TWO PAINS such that from the perspective of each person in SHARED PAIN there is no reason of an egoistic kind to prefer SHARED PAIN to TWO PAINS and from the ­perspective of each person in TWO PAINS there is no reason of an egoistic kind to prefer TWO PAINS to SHARED PAIN. Someone who evaluates outcomes in this way would conclude that SHARED PAIN and TWO PAINS are equally bad outcomes, and that therefore Worse is false. Notice that those who are attracted to the person-­centered approach can still accept the Shared Pain Argument, as well my central argument for the conclusion that Psychological Reductionism supports the Impersonal Value of Experience. These arguments are concerned only with the relations of support between different claims. Thus, one can accept both Psychological Reductionism and my conclusion that Psychological Reductionism (in contrast with Subject-­Involving Views) supports The Impersonal Value of Experience, and yet reject The Impersonal Value of Experience. The claim that Psychological Reductionism (in contrast with Subject-­ Involving Views) supports the Impersonal Value of Experience does not entail that if Psychological Reductionism (rather than a Subject-­Involving View) is true, then the Impersonal Value of Experience is true. It entails only that the Impersonal Value of Experience is more plausible, or less implausible, if Psychological Reductionism is true than if any Subject-­Involving View is true. But someone who accepts this conclusion might claim that the Impersonal Value of Experience remains implausible even if Psychological 30  For discussion of this approach to evaluating outcomes, see Hare (1981: 87–129), and Holton and Langton (1999).

Personal Identity and Impersonal Ethics  71 Reductionism renders this claim somewhat less implausible. One might claim that even if Psychological Reductionism supports the Impersonal Value of Experience, the support that it provides is at best very weak. Psychological Reductionism may therefore give us a very weak reason to believe the Impersonal Value of Experience, but not a strong enough reason to take this claim seriously. I suspect that the attraction of the person-­centered approach is the main source of resistance to accepting the Impersonal Value of Experience. It is therefore worth considering potential problems for the approach. In this section, I will focus on problems related to vagueness. Some non-­reductionists deny that it can be vague how many persons there are. But if the Reductionist View is true, then, it seems, it can be. And, in at least some cases involving vagueness of the count of persons, the evaluations given by the person-­centered approach are implausible. To see this, consider an outcome that we will call BORDERLINE. This outcome is very similar to SHARED PAIN, in which persons T and K are subjects of the same token unpleasant pain experience. But in BORDERLINE the degree of overlap between the two conjoined bodies and brains is greater than in SHARED PAIN, and hence, it is vague whether the number of persons (and mental subjects) in BORDERLINE is one or two. Even if we knew everything about BORDERLINE that can be known based on the empirical sciences, e.g. biology and psychology, we would be unable to say whether the number of persons in that outcome is one or two, although we could say that it is definitely at least one and at most two.31 We are somewhat inclined to say that there is just one person in BORDERLINE—a person with two partially overlapping brains, and two faces, each associated most closely with only one of her two conjoined bodies—someone who can think and act in a somewhat unified way, e.g., by performing imperfectly synchronized smiles and multibrachiate gesticulation. But because there is also some degree of disunity between the bodies and brains in BORDERLINE, we are somewhat inclined to say that there are two separate persons in that outcome—each with her own mind and body. If there are two persons in BORDERLINE, they are unusually closely connected to each other, much like the actual conjoined twins, Tatiana and Krista Hogan. Although most of us seem to care deeply about persons, metaphysically speaking, vagueness of the count of persons is nothing special. There are more mundane examples of vagueness of count. To borrow an example from Quine, sometimes the saddle between two mountain peaks is deep and broad enough that it is clearly between two mountains; sometimes, it is shallow and narrow enough that it is clearly in the middle of a single mountain with two peaks. And sometimes, it is of medium depth and breadth so that the number of mountains is neither clearly one nor clearly two, and so it is vague how many mountains 31  For further discussion of vagueness of the count of persons in cases of conjoined twinning, see Blatti (2007) and Campbell and McMahan (2016).

72  Tim Campbell there are.32 BORDERLINE is of course much stranger than this situation involving mountains. But both are instances of the same general phenomenon, or so I shall assume.33 Now suppose that, vagueness aside, all my stipulations about the comparison of SHARED PAIN and TWO PAINS are true of the comparison of BORDERLINE and TWO PAINS. Thus, although it is vague how many persons exist in BORDERLINE, the following claims are true. There is exactly one unpleasant experience in BORDERLINE; there are exactly two unpleasant experiences in TWO PAINS; there are exactly two persons in TWO PAINS, and each is the subject of only one unpleasant experience; and each unpleasant experience in TWO PAINS is equally unpleasant as the one in BORDERLINE. To simplify the discussion, I will also assume that if there are two persons in BORDERLINE, then each is a subject of the single unpleasant experience in that outcome.34 It is hard to compare BORDERLINE and TWO PAINS by reference to the perspectives of persons. It seems one cannot compare these outcomes by imagining what it is like to be in the shoes of each person (one at a time) if one cannot determine how many pairs of shoes there are. More important, if the person-­centered approach is correct, and if, as we are assuming, nothing other than the number of persons and the nature of their experiences determines the ranking of BORDERLINE and TWO PAINS, then given that it is vague how many persons are in BORDERLINE, it should be vague whether TWO PAINS is worse than BORDERLINE. My intuition, which I think at least some readers will share, is that TWO PAINS is worse than BORDERLINE. Even if it is vague how many persons exist in BORDERLINE, the evaluative ranking of BORDERLINE and TWO PAINS is not vague. If there are cases of vague evaluative rankings, this is not one of them.35 32  Quine (1960: 114). 33  One could argue the number of persons is determined at least in part by people’s attitudes, and that this makes the case of persons importantly different from the case of mountains. Mark Johnston (2010) has argued that one’s values can partially determine facts about personal identity. Rawls (1974) gestures toward a similar view. The position defended by Johnston (and perhaps by Rawls) is an important one, but I will not consider it in this paper. 34  I am here ignoring the possibility that although two persons exist in BORDERLINE, only one of them has the unpleasant experience in that outcome but it is indeterminate which one. There is a further complication that I am ignoring. In my comparison of Shared Pain and Two Pains I assumed that the same persons exist in each outcome. But vagueness of the number of persons might also yield vagueness of trans-­world personal identity. Thus, it might be vague not only how many persons are in BORDERLINE, but also, regarding each person in TWO PAINS, whether she exists in BORDERLINE. 35  One argument for the claim that TWO PAINS is worse than BORDERLINE draws inspiration from John Broome (1997: 71). Broome defends what he calls The collapsing principle, special version: For any x and y, if it is false that y is Fer than x and not false that x is Fer than y, then it is true that x is Fer than y. If we assume that it is false that TWO PAINS is better than BORDERLINE and not false that BORDERLINE is better than TWO PAINS, then by the collapsing principle, special version, we can conclude that it is true that BORDERLINE is better than TWO PAINS. We can then conclude that it is true that TWO PAINS is worse than BORDERLINE. (If it is true that x is better than y, then it is true that y is worse than x.) While I have a strong intuition that the statement ‘TWO PAINS is worse than BORDERLINE’ is true, serious doubts have been raised about Broome’s principle. See e.g. Carlson (2013) and Gustafsson (2018).

Personal Identity and Impersonal Ethics  73 But whether one shares my intuition may depend on one’s view of vagueness. I am attracted to semanticism—the view that vagueness is semantic indecision.36 Although the world itself is determinate, no one has decided whether, in certain bizarre cases, the term ‘person’ refers to a thing with two partially fused human bodies or to each of two things that has only one human body.37 The term ‘person’, like the term ‘mountain’, exhibits vagueness of individuation. The statements ‘There is only one person in BORDERLINE’ and ‘There are two persons in BORDERLINE’ are both indeterminate, neither true nor false. If I were to accept the person-­centered approach I would conclude that due to semantic indecision there is no fact of the matter as to whether TWO PAINS is worse than BORDERLINE. But I find this conclusion extremely implausible.38 The main alternative to semanticism is epistemicism. On this account, vagueness is a kind of ineliminable uncertainty.39 If epistemicism is true, then the number of persons in BORDERLINE is either definitely 1 or definitely 2, but we cannot know which. There is a fact of the matter, but this fact is unknowable.40 Epistemicists who accept the person-­centered approach will claim that since we 36  I should clarify that I am here referring to non-­normative vagueness. Normative vagueness poses special challenges. See e.g. Schoenfeld (2016) and Sud (2019). For a characterization of semanticism, see Lewis (1986: 212–13) and Akiba (2015: 566). 37  I am assuming that any person in BORDERLINE has a human body. Having a human body may be understood either as being identical to a body or as standing in some other intimate relation to a body (e.g. being constituted by a body). 38  Parfit probably would have agreed. Although he never explicitly endorsed semanticism, I suspect he would have said that one could (in principle, if not in practice) know all the facts about BORDERLINE without knowing whether the number of persons in BORDERLINE is one or two and that this is because there is no fact of the matter. In support of this claim about what Parfit would have said, I refer to his discussion of his imaginary case My Division in Reasons and Persons, 254–9. In this hypothetical case, the two halves of Parfit’s brain are disconnected and each is transplanted into a different body. Each body appears to belong to a different person with Parfit’s psychological profile. Discussing the case, Parfit asks ‘What happens to me?’ (255). He considers four different answers (256): (1) I do not survive; (2) I survive as one of the two people; (3) I survive as the other; (4) I survive as both. After considering some arguments against each of (1)–(4), Parfit writes ‘Perhaps there are not here different possibilities, each of which might be what happens, . . . Perhaps, when we know that each resulting person would have one half of my brain, and would be psychologically continuous with me, we know everything. . . . I believe that there cannot be different possibilities, each of which might be the truth, unless we are separately existing entities, such as Cartesian Egos’ (258). Since Parfit rejects the view that we are separately existing entities, his final view seems to be that (1)–(4) are not different possibilities, each of which might be what happens. The facts about My Division do not include any of (1)–(4). Parfit also thought that the evaluation of My Division does not depend on any of (1)–(4). I believe Parfit would say something similar about the hypothetical cases that I have presented. For example, in response to the question ‘How many persons are there in BORDERLINE?’ He would say that the different answers, e.g. ‘one’ and ‘two’, do not refer to different possibilities, each of which might be the truth, and that the evaluation of BORDERLINE does not depend on our answer to this question. 39  For a defence of the account, see Williamson (1994). 40  Parfit seems committed to the denial of epistemicism. When discussing cases of indeterminacy of personal identity, Parfit often claims that if, with respect to such cases, we know all the facts about mental and bodily states and how these states are related to each other, then we ‘know everything’, and questions such as ‘Is this person identical to that one?’ and ‘Does this person continue to exist?’ are ‘empty’. See e.g. Reasons and Persons, 259–60. This seems incompatible with epistemicism. If epistemicism is true, then although the answers to such questions may be unknowable, the questions are not ‘empty’; they have answers. See Alter and Rachels (2004) for further discussion of this issue.

74  Tim Campbell cannot know how many persons exist in BORDERLINE, we cannot know exactly how BORDERLINE and TWO PAINS compare evaluatively. An epistemicist might argue as follows. If there is only one person in BORDERLINE, then TWO PAINS is worse than BORDERLINE since in TWO PAINS there are more persons with an unpleasant experience, and hence more persons suffering, but if there are two persons in BORDERLINE and each is the subject of the unpleasant experience in that outcome, then BORDERLINE and TWO PAINS are equally bad since they contain the same number of suffering persons and each person in BORDERLNE suffers just as much as each person in TWO PAINS. Since we cannot know which situation obtains, we cannot know whether TWO PAINS is worse than BORDERLINE or whether these outcomes are equally bad. I find this argument unconvincing for a variety of reasons. Here I shall focus on just one. If epistemicism is true and the person-­centered approach is correct then even a very small natural difference between two outcomes can ground a very large difference in their respective values.41 This is because according to epistemicism, even the smallest natural difference can make a difference to the number of persons. Suppose that in BORDERLINE the degree of separation between the two bodies is sufficiently great that there are two persons, each of whom is the subject of the single unpleasant experience in that outcome. Now compare BORDERLINE with an outcome that we will call BORDERLINE*. The only difference between BORDERLINE and BORDERLINE* is that in BORDERLINE* the degree of overlap between the two human bodies is slightly greater—they share just one more atom. This slightly greater degree of overlap brings us slightly closer to a situation in which there is only one person. A sufficiently long sequence of slight changes, each of which increases the degree of overlap between the bodies by just a single atom, would result in an outcome in which there is clearly just one person who is the subject of the unpleasant experience. If epistemicism is true, then at some point in this sequence, the number of persons jumps from two to one, although the exact point of the jump is unknowable. For the sake of argument, let’s assume that the jump happens between BORDERLINE and BORDERLINE*; there are two persons suffering pain in BORDERLINE, but only one person suffering pain in BORDERLINE*. Next, compare three outcomes—BORDERLINE, BORDERLINE*, and TWO PAINS. Suppose that the unpleasant experiences in these outcomes are extremely unpleasant, that they are among the worst-­feeling experiences one can have. Then, on the person-­centered approach, TWO PAINS should be much worse than BORDERLINE* since there are two persons suffering greatly in TWO PAINS, but only one person suffering greatly in BORDERLINE*. But the person-­centered approach also implies that BORDERLINE (in which the two human bodies 41  For a similar criticism of the conjunction of epistemicism and certain evaluative judgments, see Sider (2002: 64–7).

Personal Identity and Impersonal Ethics  75 overlap by just one fewer atom than in BORDERLINE*) is equally bad as TWO PAINS, since the number of persons suffering greatly in BORDERLINE is the same as in TWO PAINS. The placement of just a single atom makes a huge difference in badness. That is hard to believe.42 If there can be vagueness regarding the number of persons, then on the two most prevalent accounts of vagueness, the person-­ centered approach has implausible implications regarding the evaluation of outcomes. If semanticism is true, then due to semantic indecision, the approach can fail to give a determinate evaluation even when there is clearly one to be given. If epistemicism is true, then some evaluations given by the approach are implausibly sensitive to arbitrarily small natural differences. Of course, the defender of the person-­ centered approach can argue that the fault lies not with her approach to evaluation but with these accounts of vagueness. She might treat the abovementioned concerns as reasons for accepting the claim that vagueness about the number of persons in BORDERLINE is ontic—i.e. that the vagueness is in the world rather than in language and other forms of representation. If this were true, it would support the claim that vagueness of the evaluative ranking of BORDERLINE and TWO PAINS is also ontic.43 But the commitment to ontic vagueness would be a very strong one. Many have thought that no plausible account of ontic vagueness can be given. Although there have been attempts to provide such an account, these attempts have either been criticized for failing to explain what ontic vagueness is, or else have involved very strong ontological commitments, such as a commitment to non-­actual concrete possible worlds.44 Those who accept the view that there is ontic vagueness will face a further difficulty if they also accept the view that identity statements can be indeterminate due to ontic vagueness. Evans (1978) famously demonstrates the difficulty via the following argument. Suppose that the statement ‘A is identical to B’ is ­indeterminate on account of its vagueness. Then if its vagueness is ontic, there is a thing in the world, A, which has the property of being such that it is indeterminate whether A is identical to B. But the statement ‘B is identical to B’ is clearly true, and so not indeterminate. Thus, it is not the case that B has the property of being such that it is indeterminate whether B is identical to B. But if these claims are 42  Some non-­reductionists might justify this huge leap by claiming that the removal (or relocation) of a single atom could make a difference to whether, for example, there is one more Cartesian soul that is suffering. But in that case, the difference between BORDERLINE and BORDERLINE* wouldn’t be just the removal (or relocation) of one atom; it would be the removal (or relocation) of one atom plus the addition of a Cartesian soul. If a person is identical to a Cartesian soul (or to a compound of soul and body) it seems to me less plausible that there is vagueness of the number of persons. 43  Miriam Schoenfield (2016) argues that if moral realism is true then assuming there is moral vagueness, it is implausible to suppose that this vagueness is semantic or epistemic. She concludes that the vaguness (if it exists) is ontic. See Sud (2019) for a critique of Schoenfield’s arguments against the semanticist’s treatment of moral vagueness. 44  For an example of the former, see Barnes and Williams (2010), and the critical response by Akiba (2015). For an example of the latter, see Akiba (2000).

76  Tim Campbell true, then A has a property that B lacks, namely the property of being such that it is indeterminate whether it is identical to B.  Given Leibniz’s law, things with different properties are not identical. Thus, it seems, the statement ‘A is identical to B’ is not indeterminate but false. This contradicts our initial assumption that the statement is indeterminate.45 To avoid the contradiction we must reject at least one of the assumptions that leads to it. But all of the assumptions, except for the assumption of ontic vagueness, seem plausible.46 As stated above, I believe that the attraction of the person-­centered approach is the main source of resistance to accepting the Impersonal Value of Experience. If this is correct, then the worries about the person-­centered approach that I have raised seem to warrant increased confidence in the Impersonal Value of Experience, other things being equal. In any case, I have shown that how much confidence one should have in the Impersonal Value of Experience depends significantly on what other philosophical claims one accepts.

6.  Impersonal Ethics In this section, I discuss some applications of my central argument to value theory and to ethics. In part IV of Reasons and Persons, Parfit considers a multitude of fascinating and important puzzles in what is now known as axiology—the part of value theory that is concerned with ranking outcomes in terms of their overall goodness. One question that Parfit considers is whether axiology should take an impersonal form or a person-­affecting form.47 For both impersonal and person-­ affecting forms, Parfit considers various criteria and general principles of that form. I shall first discuss some different person-­affecting axiological claims and contrast them with some impersonal axiological claims. I shall then explain how the Impersonal Value of Experience undermines the former, and hence, provides contrastive support for the latter. Finally, I shall argue that this has implications for morality. One family of person-­affecting principles that Parfit considers is captured by The Narrow Person-­Affecting Restriction:  Outcome A is better than outcome B only if the following claim is true: If A obtains, then there exists some person S who is better off than S would have been if B had instead obtained.

45  Evans (1978). 46  For an illuminating discussion and criticism of Evans’s argument, see Akiba (2014). 47 See Reasons and Persons, 395–401. Although Parfit doesn’t use the term ‘axiology’, the principles that he considers in the discussion cited here are stated in terms of the goodness of outcomes, and hence, are axiological principles.

Personal Identity and Impersonal Ethics  77 For reasons that Parfit and many others have emphasized, this kind of restriction is implausible. The biggest problem with it concerns the creation of new persons. For example, consider the following two cases from Parfit: The Happy Child.  A couple are trying to decide whether to have another child. They can assume that, if they did, they would love this child, and his life would be well worth living. Given the size of the world’s population when this case occurs (perhaps in some future century), this couple can assume that having another child would not, on balance, be worse for other people. When they consider the effects on themselves, they have several reasons for having another child; but they also have several contrary reasons, such as the effect on their careers. Like many others, this couple cannot decide between these two sets of conflicting reasons. They believe that, if they had this child, this would be neither better nor worse for them.48 The Wretched Child.  Some woman knows that, if she has a child, he will be so multiply diseased that his life will be worse than nothing. He will never develop, will live for only a few years, and will suffer pain that cannot be wholly relieved.49

Here, by ‘worse than nothing’ Parfit means that the wretched child’s life is overall bad for him, and hence, worth not living. Now consider the following claims. The Happy Child Claim:  The outcome in which the happy child exists is better than the outcome in which he does not exist, other things being equal. The Wretched Child Claim:  The outcome in which the wretched child exists is worse than the outcome in which he does not exist, other things being equal.

Many, but not all, would accept the Happy Child Claim; and almost everyone would accept the Wretched Child Claim.50 But the Narrow Person-­Affecting Restriction is incompatible with these claims. If the outcome in which the happy child exists obtains, he is not better off than he would have been if the alternative outcome, in which he never exists, had obtained. And if the wretched child exists, then he is not worse off than he would have been if the alternative outcome, in which he never exists, had obtained. That one is better off, or worse off, than one would otherwise have been presupposes that one would otherwise have been 48  Reasons and Persons, 381. 49  Reasons and Persons, 391. 50  Those who reject the Happy Child Claim but accept the Wretched Child Claim include Benatar (2006) and possibly Frick (2014). See also Broome (2004: 140–9) on the view that creating certain people with positive welfare makes the world neither better nor worse. Some reject both the Happy Child Claim and the Wretched Child Claim on the grounds that outcomes containing different numbers of individuals are non-­comparable—i.e. there is no true positive statement about how these outcomes compare evaluatively. See e.g. Bader (2019).

78  Tim Campbell off—i.e. that one would otherwise have existed.51 For this reason, the Narrow Person-­Affecting Restriction leads to the unfortunate conclusion that the Happy Child Claim and the Wretched Child Claim are both false. How might a person-­affecting axiology accommodate the Happy Child Claim and the Wretched Child Claim? In response to this question some who are attracted to a kind of person-­affecting view adopt a ‘comparativist’ notion of ‘better for’ that allows for comparisons of existence and non-­existence with respect to personal good. This allows them to say that if the happy child exists, then the outcome in which he exists is better for him than the outcome in which he does not exist, and, similarly, that if the wretched child exists, then the outcome in which he exists is worse for him than the outcome in which he does not exist. According to these philosophers, the claim that one outcome is better (or worse) for a person than another does not imply that she is, or would have been, better off (or worse off) in the former than she is in the latter. One can state a person-­ affecting restriction in terms of this comparativist notion of ‘better for’ as follows: The comparativist person-­affecting restriction:  Outcome A is better than outcome B only if: EITHER (i) if A obtains, then there exists a person S such that A is better for S than B, OR (ii) if B obtains then there exists a person S* such that B is worse for S* than A.52

Since the comparativist person-­affecting restriction can accommodate both the Happy Child Claim and the Wretched Child Claim, it may seem more plausible than the Narrow Person-­Affecting Restriction.53 However, some philosophers, such as Krister Bykvist and Jeff McMahan, reject the comparativist notion of ‘better for’.54 Bykvist argues that the comparison ‘A is better for one than B’ requires that A and B each have some value for one, but that one’s non-­existence has no value (not even neutral value) for one. He argues that an outcome in which one does not exist is undefined with respect to what is valuable for one, and that therefore one’s existence and one’s non-­existence are incomparable with respect to what has value for one.55 But those who reject the comparativist notion of ‘better for’ can still accept a kind of person-­affecting restriction. An outcome can be overall good (bad) for a person if she has positive (negative) well-­being in that outcome. As Parfit points out, one outcome X can be better (worse) than another Y if it would be more 51  For more on the logic of ‘better off ’ and ‘better than’ see Reasons and Persons, 489 and Broome (1993: 77). 52  For discussion of comparativist person-­affecting restrictions, see Holtug (2010: ch. 6) as well as Arrhenius and Rabinowicz (2015). 53  Comparativists who wish to accept the Wretched Child Claim but deny the Happy Child Claim can modify the Comparativist person-­affecting restriction by omitting clause (i). 54  See Bykvist (2007) and McMahan (2013). 55  Bykvist (2017); Bykvist and Campbell (2019).

Personal Identity and Impersonal Ethics  79 (less) good for the X-­people (those who exist if X obtains) than Y would be for the Y-­people. In this case, X would be better than Y for people in what Parfit calls ‘the wide sense’. Those who claim that the wide sense of ‘better for people’ is the sense that matters for axiology could adopt The wide person-­affecting restriction: Outcome A is better than Outcome B only if: EITHER (i) If A obtains, then there exists a person S such that A is overall good for S, and either S is better off than S would have been if B had obtained or if B had obtained S would not have existed, OR (ii) If B obtains, then there exists a person S* such that B is overall bad for S*, and either S* is worse off than S* would have been if A had obtained or if A had obtained S* would not have existed.

This restriction is person-­affecting in that it implies that whether one outcome is better than another always depends on facts about personal goodness and personal badness.56 Since the claim that a person’s existence is overall good (or bad) for her does not imply that her non-­existence is worse (or better) for her, those who accept the wide person-­affecting restriction can avoid problematic comparisons of existence and non-­existence, with respect to personal betterness. Corresponding to the different person-­affecting restrictions, there are different person-­ affecting axiologies. For example, those who are attracted to the wide  person-­affecting restriction might accept a corresponding wide axiology that ranks outcomes according to how good or bad they are for persons. One example would be a wide total axiology that ranks one outcome as better than another if and only if, and because, the sum total of well-­being of everyone in the first outcome is greater than the sum total of well-­being of everyone in the second outcome. Alternatively, one could adopt a wide prioritarian axiology that differs from the wide total axiology in that it accords more value to the well-­being of a person the worse off she is.57 In contrast to person-­affecting axiologies, there are impersonal axiologies. According to such axiologies, the ranking of outcomes is determined not by how good they are for persons, but by something else, such as their respective balances of pleasure over pain. If the Impersonal Value of Experience is true, it is easier to see how one outcome can be better than another even if it is not better for persons in the narrow, comparativist, or wide sense. For example, I argued that TWO PAINS is worse than SHARED PAIN because it has more unpleasant pain experience, and I 56  For discussion of ‘wide’ person-­affecting principles see Parfit, Reasons and Persons, 396–401, as well as Temkin (2012: 437). 57  See e.g. Holtug (2010: esp. ch. 8).

80  Tim Campbell identified this as one instance of the Impersonal Value of Experience. But SHARED PAIN and TWO PAINS are equally good for persons. They are equally good for persons in the narrow sense: no matter which outcome obtains, each person is equally as well off in that outcome as she would have been if the other outcome had instead obtained. They are also equally good for persons in the comparativist sense: no matter which outcome obtains, for each person this outcome will be equally good as the other outcome. Finally, they are equally good for persons in the wide sense. The sum total of well-­being (or ill-­being) of everyone who exists in TWO PAINS is the same as the sum total of well-­being (or ill-­being) of everyone who exists in SHARED PAIN. The Impersonal Value of Experience appears to rule out the narrow, ­comparativist, and wide person-­affecting restrictions as well as the corresponding narrow, comparativist, and wide person-­affecting axiologies. Consider, for example, the wide total and wide prioritarian axiologies. These axiologies rank SHARED PAIN and TWO PAINS as equally good, since SHARED PAIN and TWO PAINS are equally good with respect to total well-­being, each outcome has the same number of persons, and there is perfect equality of well-­being in each. This seems to show that, contrary to what Parfit suggests in part IV of Reasons and Persons, the difference between a wide person-­affecting axiology and an otherwise similar impersonal axiology is not simply a matter of the rhetoric that is used to state each axiology.58 While an impersonal axiology can accommodate the claim that TWO PAINS is worse than SHARED PAIN, it is not clear that a person-­affecting axi­ology can. Certainly, the person-­affecting axiologies that we have considered in this section cannot. This difference between the impersonal and person-­affecting axiologies is substantive, not merely terminological. Finally, while the Impersonal Value of Experience clearly has relevance for axiology, it also seems to have relevance for morality, and hence, for ethics. Axiology feeds into morality. That one outcome is better than another can provide a moral reason to bring about the former rather than the latter. In some cases, this reason will be decisive in that it determines what one ought to do all things considered. Insofar as axiology is at least partly impersonal, it stands to reason that ethics, which concerns what we ought to do and what reasons we have, is also at least partly impersonal. For example, if TWO PAINS is worse than SHARED PAIN, and all else is equal, then it seems one would have moral reason to bring about SHARED PAIN rather than TWO PAINS. One would have moral reason to reduce the number of intensely unpleasant experiences. Correspondingly, I think, one would have moral reason to increase the total number of pleasant experiences, other things being equal. I mentioned above that some reject the Happy Child Claim. Some also deny that the fact that a person 58 See Reasons and Persons, 393–400. The claim that the difference between impersonal and wide person-­affecting axiologies is merely a matter of rhetoric is also made by Arrhenius (2003: 192).

Personal Identity and Impersonal Ethics  81 would be happy (i.e. have positive well-­being) provides a moral reason to create that person; but they maintain that the fact that some person would be miserable (i.e. would have negative well-­being) provides a moral reason to refrain from creating that person. These writers hold a position known as the Procreation Asymmetry.59 The justification that is usually offered for the Procreation Asymmetry is that if one were to create a miserable person, then this person would be a victim of one’s act, whereas if one were to refrain from creating a happy person, then, other things being equal, there would be no victim of one’s act.60 But this asymmetry in terms of the existence of a victim cannot explain why one would have a moral reason to bring about SHARED PAIN rather than TWO PAINS. Since each of these outcomes is equally good for everyone, there would be no victim of the choice to bring about TWO PAINS rather than SHARED PAIN. The moral reason to bring about SHARED PAIN rather than TWO PAINS would instead be that this would reduce the number of intensely unpleasant experiences. Someone could accept this claim but still deny that there is ever a moral reason to increase the number of pleasant experiences. But then one must come up with a different kind of justification for this asymmetry. The justification cannot be in terms of the existence or non-­existence of a victim.

7. Conclusion Parfit claimed that on the Reductionist View ethics is ‘more impersonal’ than on the Non-­Reductionist View. He thought that if the Reductionist View were true and the Non-­Reductionist View false, it would be more plausible to focus on experiences and to claim that what matters is the nature of experiences. If my central argument is sound, then Parfit is, to some extent, vindicated. There is an important distinction in the ballpark of the one between the Reductionist and Non-­ Reductionist Views, namely the distinction between Psychological Reductionism and what I have called Subject-­Involving Views, where the latter are non-­reductionist if one assumes (as Parfit did) a psychological criterion of personal identity. If Psychological Reductionism is true and Subject-­Involving Views are false, it is more plausible that a difference in experiential properties between outcomes can by itself make one of these outcomes worse or better than the other. This in turn should make it more plausible that such experiential differences have moral significance independent of what they imply about differences in personal well-­being. Thus, Psychological Reductionism seems to push toward a more impersonal axiology as well as a more impersonal moral outlook. 59 For an extended discussion of the Procreation Asymmetry, see McMahan (2013) as well as Frick (2018). 60  See e.g. Benatar (2006) and Narveson (1973).

82  Tim Campbell

Bibliography Adams, Robert (1985). ‘Should Ethics Be More Impersonal? A Critical Notice of Derek Parfit, Reasons and Persons.’ Philosophical Review 98/4: 439–84. Akiba, Ken (2000). ‘Vagueness as a Modality.’ Philosophical Quarterly 50: 359–70. Akiba, Ken (2014). ‘A Defense of Indeterminate Distinctness.’ Synthese 191: 3557–73. Akiba, Ken (2015). ‘How Barnes and Williams Have Failed to Present an Intelligible Ontic Theory of Vagueness.’ Analysis 75/4: 565–73. Alter, Torin, and Rachels, Stuart (2004). ‘Epistemicism and the Combined Spectrum.’ Ratio 17/3: 241–55. Arrhenius, Gustaf (2003). ‘The Person-Affecting Restriction, Comparativism, and the Oral Status of Potential People.’ Ethical Perspectives 10: 185–95. Arrhenius, Gustaf, and Rabinowicz, Wlodek (2015). ‘The Value of Existence.’ In The Oxford Handbook of Value Theory, edited by Iwao Hirose and Jonas Olson, 424–44. Oxford: Oxford University Press. Bader, Ralf (2019). ‘Person-Affecting Utilitarianism.’ Working Paper. Barnes, Elizabeth, and Williams, J.  R.  G. (2011). ‘A Theory of Metaphysical Indeterminacy.’ Oxford Studies in Metaphysics: Volume 6, edited by Karen Bennett and Dean. W. Zimmerman, 103–48. Oxford: Oxford University Press. Blatti, Stephen (2007). ‘Animalism, Dicephalus, and Borderline Cases.’ Philosophical Psychology 20/5: 595–608. Brink, David (1997). ‘Rational Egoism and the Separateness of Persons.’ In Reading Parfit, edited by Jonathan Dancy, 96–134. Oxford: Blackwell Publishers. Broome, John (1993). ‘Goodness is Reducible to Betterness: The Evil of Death is the Value of Life.’ In The Good and the Economical, edited by Peter Koslowski and Yuichi Shionoya, 70–84. Berlin: Springer-Verlag. Broome, John (1997). ‘Is Incommensurability Vagueness?’ In Incommensurability, Incomparability and Practical Reason, edited by Ruth Chang, 67–89. Cambridge, MA: Harvard University Press. Broome, John (2004). Weighing Lives. Oxford: Oxford University Press. Bykvist, Krister (2007). ‘The Benefits of Coming into Existence.’ Philosophical Studies 135/3: 335–62. Bykvist, Krister (2017). ‘Person-Affecting Morality and Non-Identity Cases.’ Working Paper. Bykvist, Krister, and Campbell, Tim (2019). ‘Persson’s Merely Possible Persons.’ Working Paper. Campbell, Tim, and McMahan, Jeff (2016). ‘Animalism and the Varieties of Conjoined Twinning.’ In Animalism: New Essays on Persons, Animals, & Identity, edited by Stephan Blatti and Paul F. Snowdon, 229–51. Oxford: Oxford University Press. Carlson, Erik (2013). ‘Vagueness, Incomparability, and the Collapsing Principle.’ Ethical Theory and Moral Practice 16/3: 449–63.

Personal Identity and Impersonal Ethics  83 Chisholm, Roderick (1963). ‘The Theory of Appearing.’ In Philosophical Analysis: A Collection of Essays, edited by Max Black, 97–112. Ithaca, NY: Cornell University Press. Dominus, Susan (2011). ‘Could Conjoined Twins Share a Mind?’ New York Times Magazine, 25 May. Evans, Gareth (1978). ‘Can There Be Vague Objects?’ Analysis 38: 208. Frick, Johann David (2014). ‘ “Making People Happy, Not Making Happy People”: A Defense of the Asymmetry Intuition in Population Ethics.’ Doctoral dissertation, Harvard University. Frick, Johann (2018). ‘Conditional Reasons and the Procreation Asymmetry.’ Working Paper. Gustafsson, Johan (2018). ‘Does the Collapsing Principle Rule Out Borderline Cases?’ Utilitas 30/4: 483–92. Hare, R. M. (1981). Moral Thinking. Oxford: Clarendon Press. Hershenov, David (2013). ‘Who Doesn’t Have a Problem of Too Many Thinkers?’ American Philosophical Quarterly 50/2: 203–8. Holton, Richard, and Langton, Rae (1999). ‘Empathy and Animal Ethics.’ In Singer and His Critics, edited by Dale Jamieson, 209–32. Oxford: Blackwell Publishers. Holtug, Nils (2010). Persons, Interests, and Justice. New York: Oxford University Press. Johnston, Mark (1992). ‘Reasons and Reductionism.’ Philosophical Review 101/3: 589–618. Johnston, Mark (2010). Surviving Death. Princeton: Princeton University Press. Johnston, Mark (2016). ‘The Personite Problem: Should Practical Reason be Tabled?’ Noûs 51: 617–44. Kim, Jaegwon (1976). ‘Events as Property Exemplifications.’ Action Theory, Synthese Library Book Series 97: 159–77. Korsgaard, Christine (1989). ‘Personal Identity and the Unity of Agency: A Kantian Response to Parfit.’ Philosophy & Public Affairs 18/2: 101–32. Lewis, David (1986). On the Plurality of Worlds. New York: Oxford University Press. Lichtenberg, Georg Christoph (2005). Sudelbucher I–II: Material Books and Diaries. Register zu den Sudelbüchern: Dreibändige Gesamtausgabe im Schmuckschuber Taschenbuch. Locke, John (1975). ‘Of Identity and Diversity.’ In Personal Identity, edited by John Perry, 33–52. Berkeley and Los Angeles: University of California Press. Lowe, E. J. (1996). Subjects of Experience. Cambridge: Cambridge University Press. McMahan, Jeff (2002). The Ethics of Killing: Problems at the Margins of Life. New York: Oxford University Press. McMahan, Jeff (2013). ‘Causing People to Exist and Saving People’s Lives.’ Journal of Ethics 17: 5–35. Narveson, Jan (1973). ‘Moral Problems of Population.’ The Monist 76: 62–72.

84  Tim Campbell Olson, Eric (1997). The Human Animal: Personal Identity Without Psychology. New York: Oxford University Press. Parfit, Derek (1984). Reasons and Persons. Oxford: Oxford University Press. References are to page no. unless otherwise stated. Parfit, Derek (1999). ‘Experiences, Subjects, and Conceptual Schemes.’ Philosophical Topics 26/1: 217–70. Quine, W. V. O. (1960). Word & Object. Cambridge, MA: MIT Press. Rawls, John (1974). ‘The Independence of Moral Theory.’ Proceedings and Addresses of the American Philosophical Association 48: 5–22. Roche, William (2012). ‘A Weaker Condition for Transitivity in Probabilistic Support.’ European Journal for Philosophy of Science 2: 111–18. Schoenfield, Miriam (2016). ‘Moral Vagueness is Ontic Vagueness.’ Ethics 126/2: 257–82. Shoemaker, Sydney (1985). ‘Critical Notice of Reasons and Persons.’ Mind 94: 443–53. Shoemaker, Sydney (1986). ‘Introspection and the Self.’ In The First-Person Perspective and Other Essays, 1–24. Cambridge: Cambridge University Press. Sider, Theodore (2002). ‘Hell and Vagueness.’ Faith and Philosophy 19: 58–68. Strawson, Galen (2017). The Subject of Experience. New York: Oxford University Press. Strawson, P. F. (1959). Individuals. London: Methuen. Sud, Rohan. (2019). ‘Moral Vagueness as Semantic Vagueness.’ Ethics 129/4: 684–705. Temkin, Larry (2012). Rethinking the Good, Moral Ideals and the Nature of Practical Reasoning. New York: Oxford University Press. Unger, Peter (1990). Identity, Consciousness, and Value. New York: Oxford University Press. Williamson, Timothy (1994). Vagueness. London: Routledge. Wolf, Susan (1986). ‘Self-Interest and Interest in Selves.’ Ethics 96/4: 704–20. Zimmerman, Dean (2010). ‘From Property Dualism to Substance Dualism.’ Proceedings of the Aristotelian Society, supplementary vol. 84: 119–50.

4

Temporal Neutrality and the Bias Toward the Future Samuel Scheffler

1.  Introduction: Rawls and Sidgwick on Temporal Neutrality The conviction that rationality requires an equal concern for all parts of one’s life marks a rare point of agreement among leading Kantian and utilitarian philosophers. John Rawls disagrees with Henry Sidgwick about many things, but the rationality of temporal neutrality is not one of them. In A Theory of Justice, Rawls makes his agreement with Sidgwick on this point explicit. He writes: In the case of an individual the avoidance of pure time preference is a feature of the rational. As Sidgwick maintains, rationality implies an impartial concern for all parts of our life. The mere difference of location in time, of something’s being earlier or later, is not in itself a rational ground for having more or less regard for it. Of course, a present or near future advantage may be counted more heavily on account of its greater certainty or probability, and we should take into consideration how our situation and capacity for particular enjoyments will change. But none of these things justifies our preferring a lesser present to a greater future good simply because of its nearer temporal position.1

Sidgwick, for his part, thinks that the principle ‘of impartial concern for all parts of our conscious life,’2 as he calls it, represents a commonsense consensus. He This paper was originally written for the Conference in Memory of Derek Parfit that was held at Rutgers in December 2017. I am grateful to the Rutgers audience as well as to audiences at MIT and the University of Pittsburgh for helpful discussion. I am also indebted to Jeff McMahan, Jake Nebel, and Joseph Raz for written comments on earlier drafts. 1 J.  Rawls, A Theory of Justice (Cambridge, MA: Harvard University Press, 1971), 293–4. Rawls does not distinguish here between tensed and un-­tensed temporal preferences. That is, he does not distinguish between a preference for having certain things occur either in the past or in the future and a preference for having certain things occur either earlier or later. The distinction is significant, however. The bias toward the future is not the same as a bias toward the later, and some un-­tensed preferences may be compatible with temporal neutrality, if such neutrality is understood as an impartial concern for all parts of one’s life. See Section 4. 2 H. Sidgwick, The Methods of Ethics (7th edn, London: Macmillan & Co., 1907; repub. Indianapolis: Hackett Publishing Company, 1981), 381. Samuel Scheffler, Temporal Neutrality and the Bias Toward the Future In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0005

86  Samuel Scheffler says that an ‘equal and impartial concern for all parts of one’s conscious life is perhaps the most prominent element in the common notion of the rational—as opposed to the merely impulsive—pursuit of pleasure.’3 On this last point, at least, Sidgwick was almost certainly mistaken. The principle of impartial concern may sound commonsensical, but most of us have robust preferences that are inconsistent with it. That is one lesson of Derek Parfit’s discussion of temporal bias in Reasons and Persons.4

2.  Parfit on the Bias Toward the Future In Parfit’s famous example, you are asked to imagine that you awake in your hospital bed. You are told by a nurse that either you underwent ten hours of extremely painful surgery without anesthesia yesterday, after which you were given a drug to cause you to forget the experience, or else you will undergo one hour of extremely painful surgery without anesthesia today, after which you will be given a drug to cause you to forget the experience. Which would you prefer? Most of us would prefer to have undergone the longer operation yesterday than to be faced with the shorter operation today. Using this and other examples, Parfit argues that we display a bias toward the future in the following respects. We would prefer to have experienced pain of a given intensity and duration in the past than to experience it in the future. We would even prefer to have experienced a longer period of pain in the past than to experience a shorter period of pain in the future. And we would prefer that our lives contain more total hours of pain, if that meant less of it were still to come. With respect to pleasurable sensations, we have the reverse preferences. We would prefer to experience pleasure of a given intensity and duration in the future than to have experienced it in the past. We would even prefer to experience a shorter period of pleasure in the future than to have experienced a longer period in the past. And we would prefer our lives to contain fewer total hours of pleasure, if that meant more of it were still to come. Taken together, and setting aside various refinements and qualifications, these claims ascribe to us a general preference that our pains be in the past and our pleasures in the future. With respect to pleasure and pain, we are, so to speak, more solicitous of our futures than of our pasts. In that sense, we are biased toward the future. As Parfit makes clear, our attitudes toward time form a complex network, of which the bias toward the future is just one element, and it is not immediately 3 Sidgwick, The Methods of Ethics, 124n. 4 D. Parfit, Reasons and Persons (Oxford: Oxford University Press, 1984), esp. ch. 8. As Tom Hurka has pointed out to me, Parfit’s position was anticipated by J. M. E. McTaggart, who identified the phenomenon of future-­bias very clearly in The Nature of Existence, vol. ii (Cambridge: Cambridge University Press, 1927), ch. 59.

Temporal Neutrality  87 apparent how the various elements fit together. What is evident is that care must be taken in delineating the scope and limits of the bias and its relations and interactions with various of our other temporal attitudes. Three things that Parfit says deserve special attention in this connection. First, he says that the bias toward the future is not the only form of temporal bias we exhibit. Many of us also exhibit a bias toward the near. We would prefer a smaller pleasure sooner to a larger pleasure later, and a larger pain later to a smaller pain sooner. To the extent that we exhibit both of these temporal biases, there are questions about the relations and possible interactions between them. And in assessing the rationality of the bias toward the future, we must be sure that our assessment is consistent with our assessment of the bias toward the near. Second, Parfit indicates that the bias toward the future is limited in scope. It applies to certain of our experiences, and especially to our experiences of pleasure and pain, but it does not apply to all the good and bad things in our lives.5 For example, he says, it is not a feature of our attitudes toward ‘events that give us either pride or shame; events that either gild or stain our picture of our lives’.6 Although Parfit may here be underestimating the extent to which experiences of pleasure or pain can themselves be the objects pride or shame—think, for example, of people who feel shame about certain experiences of sexual pleasure or pride about having endured certain episodes of pain and discomfort—it is not difficult to see the sort of thing he has in mind. Would you prefer already to have written five excellent novels or to write one such novel in the future? Would you prefer to have betrayed your friends five times in the past or to do so once in the future? If our bias toward the future applied to these cases, then we would prefer to write one excellent novel in the future rather than to have written five in the past. And we would prefer to have betrayed our friends five times in the past rather than to do so once in the future. But it is not clear that most people would have these preferences. Third, Parfit says that the scope of the bias is limited in another way as well. It does not, he asserts, manifest itself as clearly in third-­person cases as it does in first-­person cases. Imagine you had been informed that your elderly mother, who was living in a distant land and about whom you seldom received news, was suffering from a fatal illness and, moreover, that she was going to have to undergo several months of severe pain before succumbing to the illness. Imagine that you then received a second communication informing you that the initial report had been mistaken about the timing of your mother’s ordeal. In fact, she had already undergone several months of severe pain and had then died. Would you be greatly relieved to learn that her suffering, instead of being in the future as you had 5  Indeed, as I have argued in Why Worry about Future Generations? (Oxford: Oxford University Press, 2018), 124–9, there are many good and bad experiences to which it also does not apply. 6  Reasons and Persons, 160.

88  Samuel Scheffler initially supposed, was instead in the past? Parfit doesn’t believe that you would. More generally, he writes: My examples reveal a surprising asymmetry in our concern about our own and other people’s pasts. I would not be distressed at all if I was reminded that I myself once had to endure several months of suffering. But I would be greatly distressed if I learnt that, before she died, my mother had to endure such an ordeal.7

This seems to show that the bias toward the future is not as evident in our attitudes toward other people’s experiences as it is in our attitudes toward our own experiences. Although it is a great relief to think that our own pains lie in the past rather than the future, it is not a comparably great relief to think this about those we love. Despite these complications, the bias toward the future appears to be quite robust. Virtually everyone seems to exhibit the bias. And it does not appear to be subject to direct volitional control. That is, it does not appear to be something one could simply decide to give up. Caspar Hare says that ‘you may as well try to lose your skin’.8 But while this casts doubt on the commonsense credentials of Sidgwick’s principle ‘of impartial concern for all parts of our conscious life’, it doesn’t show that the principle is mistaken. It doesn’t show that temporal neutrality is not a requirement of rationality. It may be that the bias toward the future, widespread and robust though it is, is nevertheless irrational. Perhaps we are all subject to the bias, but perhaps we are all, to that extent, irrational. What does Parfit think? Does he mean to be claiming not only that the bias is deep and widespread but also that it is rationally required or at least rationally defensible? Or does he, in the end, side with Sidgwick and Rawls, and suppose that the bias is irrational and that temporal neutrality is, after all, a requirement of rationality?9 In fact, Parfit’s position is surprisingly difficult to pin down. This is largely because of the dialectical context in which his discussion of the bias toward the future is embedded. He raises the issue in the course of an extended argument against what he calls S, or the Self-­interest Theory. And rather than 7  Reasons and Persons, 182. 8  C. Hare, ‘Time—The Emotional Asymmetry’, in H. Dyke and A. Bardon (eds.), A Companion to the Philosophy of Time (Oxford: John Wiley & Sons, Inc., 2013): 507–20, at 513. Compare McTaggart: ‘[While] allowing that past good or evil does tend to produce present happiness or unhappiness, it remains the case that future good or evil tends to produce them to a much greater extent. I do not know that any reason can be assigned for this greater present importance of the future. It may have a reason which is not yet discovered. Or it may be an ultimate fact. But it cannot be denied to be a fact’ (The Nature of Existence, ii. 350). 9  In fairness to Rawls, it should be noted that, when he says that temporal neutrality is a requirement of rationality, it appears that his main concern is to insist on the irrationality of the bias toward the near. He does not explicitly mention the bias toward the future, and it is not clear whether he is taking account of the fact that the requirement of temporal neutrality condemns both biases.

Temporal Neutrality  89 assessing the bias directly, or in his own voice, often what he does instead is to draw conclusions about what the S-­theorist must say about the bias. For example, he writes: If [the S-­Theorist] condemns the bias toward the near because it cannot have rational significance when some pain is felt, he must condemn the bias toward the future. He must claim that it is irrational to be relieved when some pain is in the past. Most of us would find this hard to believe. If the S-­Theorist insists that we should be temporally neutral, most of us will disagree.’10

In this passage, Parfit seems to be treating it as an implausible implication of S  that, on certain assumptions, it must condemn the bias toward the future as irrational. So it is natural to conclude that he himself thinks the bias is not irrational. And that is how some readers have interpreted him.11 But he never actually says this. Nor does he say explicitly in Reasons and Persons that the bias toward the future is irrational. But he does say, somewhat surprisingly in light of passages like the one just cited, that S’s requirement of temporal neutrality is ‘plausible’, and that ‘this part of S may be either true or, if we think that such claims cannot be true, part of the best or best justified theory’.12 And he devotes a section to arguing that we should not be biased toward the future, because the bias is bad for us. In developing this argument, he asks us to consider a person he calls Timeless, who lacks the future-­bias. For Timeless, looking backward to past enjoyments is just as pleasant as looking forward to future ones, and looking backward to past pains is just as distressing as looking forward to future ones.13 If we were like Timeless, Parfit says, things would be in some ways worse for us. For example, we would not be relieved when bad things were in the past. On the other hand, we would not be sad when good things were over. In addition, we could afford to be selective about which events we chose to look backward to. We ‘could allow ourselves to forget most of the bad things that have happened, while preserving by rehearsing all our memories of the good things’.14 So the net effect of these changes would be positive. Furthermore, we would gain enormously in our atti-

10  Reasons and Persons, 170. 11  Some others have interpreted him as being neutral about the rationality of future-­bias. See e.g. P. Greene and M. Sullivan, ‘Against Time Bias’, Ethics 125 (2015): 947–70, at 967n. 12  Reasons and Persons, 194. 13  Parfit seems to understand looking backward to past enjoyments as not requiring memory of those enjoyments. Although, as he emphasizes, the memory of a pleasurable experience can itself be pleasurable, and although Timeless experiences this sort of pleasure, he also enjoys knowing that he had pleasant experiences even if he has no memory of them. Or at any rate, that is how I interpret the relevance of Parfit’s discussion in the final complete paragraph on p. 172 of Reasons and Persons for his subsequent characterization of Timeless on p. 174. 14  Reasons and Persons, 175.

90  Samuel Scheffler tudes toward our deaths. Although, as death approached, we would have less and less to look forward to, we would have more and more to look backward to. On our deathbeds, we would have our whole lives behind us, and that would be almost as good as having our whole lives ahead of us.15

3.  Is the Bias Toward the Future Bad for Us? However, this accounting of the advantages and disadvantages of Timeless’s situation—or of what I will call Timelessness—is not comprehensive. There are several points to consider. First, Parfit says that Timeless is never relieved when bad things are in the past. When he ‘is reminded that he once had a month of agony, he is as much distressed as when he learns that he will later have a month’.16 Now to me this sounds like a prescription for universal PTSD (post-­traumatic stress disorder). Parfit’s belief that it would merely be a disadvantage that would be outweighed by the various advantages of Timelessness rests on the thought that we could choose to preserve our memories of the good things that have happened and to ‘allow ourselves to forget’ the bad things. But this assumes without argument a strongly voluntaristic conception of memory. Moreover, even if we make generous assumptions about people’s ability to choose which past events to dwell on when engaged in private contemplation, and even if we ignore questions about the status of unconscious memories and the possible psychic costs of suppressing them, there is also the role of other people to consider. For example, other people’s sufferings often trigger memories of one’s own, and Timelessness would ensure that such memories were just as distressing as an awareness of one’s future sufferings. Furthermore, the fact that reminding someone of his past suffering could produce extreme distress suggests new possibilities for malicious behavior. In order to torment someone one disliked, one would need only to issue such reminders. (‘Remember the time you got food poisoning after eating bad seafood, or the day your partner left you, or the time you made a fool of yourself at the party, or the day you were denied tenure?’) For Timeless and those like him, time heals no wounds, and the only alternative cure

15  Why ‘almost’? Parfit’s reasoning may seem to show that it would be just as good. But there is a complication: ‘This reasoning ignores those emotions which are essentially future-­directed. It would not apply to those people for whom the joy in looking forward comes from making plans, or savouring alternatives. But the reasoning seems to be correct when applied to more passive types, those who take life’s pleasures as they come.’ Accordingly, Parfit’s more guarded conclusion is that, to the extent that we are like these more passive types, ‘we would be happier if we lacked the bias towards the future. We would be much less depressed by ageing and the approach of death. If we were like Timeless, being at the end of our lives would be more like being at the beginning. At any point in our lives we could enjoy looking backward or forward to our whole lives’ (Reasons and Persons, 176–7). 16  Reasons and Persons, 174.

Temporal Neutrality  91 would be a form of selective amnesia so powerful that even the interventions of other people could not revive the forgotten memories. Second, it is worth wondering what the effects of Timelessness would be on human motivation. Once one had lived long enough to have accumulated some good experiences, would one still have the same motivation to seek out additional good experiences, with all the attendant effort, uncertainty, and risk of failure that involves, when one could just look back with pleasure to the old ones instead? Of course, Timelessness would not mean that looking back was bound to be just as pleasant as the new experiences themselves might be, only that it would be as pleasant as looking forward to those experiences. But, of course, the new experiences themselves would have no advantage as such over the old experiences, and so the question is how, if looking forward also had no advantage over looking backward, this would affect the structure of our motivations and our incentives to seek out new experiences. To put it another way, what are the cognitive, motivational, and developmental roles played by attitudes like anticipation in our psychic economy, how would Timelessness affect those roles, and with what consequences? (Parfit himself seems undecided, in his discussion on pp. 172–4 of Reasons and Persons, between a thin conception of anticipation as consisting simply in the thought that something will happen and some thicker, though not fully specified, conception.) I don’t suppose that these questions can be answered from the philosophical armchair. But they do need to be answered if we are fully to assess the advantages and disadvantages of Timelessness for creatures like us. Third, there is also the question of how human relationships would be affected by the elimination of the bias toward the future. I have already suggested that Timelessness would open up new possibilities for malicious behavior. Equally, of course, it would also open new possibilities for kindness: one could reliably cheer people up by reminding them of good times they’d experienced in the past. But malice and kindness aside, there is a more general question about the effects of temporal neutrality on personal relationships. Consider the simple fact that we often look forward to seeing our friends and family members in the future. Timelessness would mean that it would be just as satisfying to look backward to having seen them in the past. How would this affect people’s interactions and their desires to spend time together? How would it affect the structure of human attachments and our responses to loss? Again, I don’t pretend to know the answers to these questions, but I am reluctant to draw conclusions about the net advantages or disadvantages of Timelessness without knowing them. Fourth, consider the appetites. Parfit’s discussion may create the impression that Timeless looks backward to, say, past gustatory pleasures with just as much enjoyment as he looks forward to future gustatory pleasures. But this is misleading. When I am hungry, it is not just my future-­bias that leads me to look forward to eating my next meal more avidly than I look backward to eating my last one. There is also the fact that my next meal can satisfy my present appetite,

92  Samuel Scheffler while my last meal cannot. I assume that Timeless is no different from me in this respect. When he is hungry, he too looks forward to eating his next meal with greater enthusiasm than he looks backward to eating his last one. But if that is right, it suggests that Timeless’s neutrality with respect to past and future appetitive pleasures is narrower in scope, and its advantages are therefore more limited, than we might initially have assumed. Fifth, Parfit says that, as death approached, the fact that Timeless would have less and less to look forward to would be offset by the fact that he would have more and more to look backward to. But of course, this works only if Timeless has accumulated a sufficient store of good experiences to look back to. Those people whose lives have been filled with suffering would have little to look forward to and little to look backward to, so for them the advertised advantages of Timelessness would be elusive. The point is not that these people would be better off if they retained a bias toward the future. The point is rather that the alleged advantages of Timelessness, with respect to our attitudes toward aging and death, would not be available to everyone. They would accrue only to those who had been fortunate enough to lead reasonably good lives. Sixth, when assessing the effects of Timelessness on our attitudes toward death, it is worth bearing in mind Parfit’s caution that the bias toward the future applies to certain of our experiences, and especially to our experiences of pleasure and pain, but that it does not apply to ‘events that give us either pride or shame; events that either gild or stain our picture of our lives’. With respect to these events, we exhibit no bias toward the future even now. This means that, as death approaches, we are already in as good a position as Timeless is to take solace from the things we have done that give us pride. We are also, of course, in just as good a position, if that’s the right expression, to be tormented by the things we have done that give us shame. But many people do derive solace, in the face of their impending deaths, from their justifiable pride in things they have accomplished during their lifetimes. Yet I think it is fair to say that, even for such people, this does not always extinguish the fear of death. Far from it. And so it is important to emphasize that Timelessness would confer no additional advantage with respect to attitudes like pride or shame, or with respect to the things that gild or stain our picture of our lives. The extra advantage it would confer on us, in looking backward, is merely that we might derive greater pleasure from contemplating our past experiences of pleasure. But if even justified pride in one’s past achievements is not enough to banish the fear of death, it is perhaps not obvious that the pleasure of thinking about past pleasures would be more effective. Finally, Parfit may be underestimating the advantages of our existing attitudes toward death. Consider that people who are aware of the imminence of their deaths can no longer sensibly occupy themselves with trying to shape their lives. To the extent that they are inclined to engage in self-­reflection, their gaze must be directed primarily backward rather than forward. Given their biased tendency to discount past pleasures and pains, Parfit thinks they are unlikely to find this

Temporal Neutrality  93 enforced backward gaze rewarding. Accordingly, he imagines that it would be a great improvement for them if, like Timeless, they could relish their past experiences just as avidly as they had once relished the prospect of future experiences. But this neglects the fact that the pressure to attend to the past presents a temporally biased person who is nearing death with a distinctive challenge and, perhaps, a distinctive opportunity. The challenge is to come to terms with one’s life as it has actually turned out. As a practical matter, one is no longer in a position to change one’s life in fundamental ways. Yet as an epistemic matter, one is in a better position than ever before to survey the entire trajectory of one’s life; over time, ignorance of one’s future has been almost entirely supplanted by knowledge of one’s past. Finding oneself in these circumstances— with one’s practical position diminished but one’s epistemic position enhanced— and lacking Timeless’s inclination to dwell lovingly on the hedonic high points of yesteryear, the challenge is simply to reckon with one’s life, in all its awe-­inspiring contingency, finitude, and immutability. If one is fortunate, this challenge may bring with it a certain opportunity: the opportunity to achieve a kind of self-­ acceptance, or to make peace with what one has become. If, by contrast, we were like Timeless, we might just go to our graves looking back with delight on experiences of great meals, great music, or great sex. And I am sure some will agree with Parfit that that would be a vast improvement over the current arrangement. But others may wonder whether the opportunity to take the measure of one’s life, to view it as a (nearly) complete object and to try to make sense of it as such, does not have a value of its own. They may wonder whether the advantages of Timelessness, as death approached, would be quite as clear-­cut as Parfit suggests. Of course, the considerations I have been rehearsing do not show that Parfit is wrong in thinking that the bias toward the future is bad for us and that we would be better off if we were like Timeless. But they do suggest that, in order to arrive at defensible conclusions about the overall advantages and disadvantages of the bias, a more comprehensive accounting would be required.

4.  Is the Bias Toward the Future Irrational? Suppose, however, that Parfit is correct, and that the bias toward the future is indeed bad for us. How exactly does this bear on questions about the rationality of the bias? As I’ve noted, Parfit concludes his discussion by saying that the requirement of temporal neutrality is plausible and may indeed be true. And he said subsequently that he did mean to be endorsing the view that rationality requires temporal neutrality and that the bias toward the future is irrational.17 But 17  Personal communication, September 26th, 2015. See also D. Parfit, On What Matters (Oxford: Oxford University Press, 2011), i. 495 (n. to p. 57).

94  Samuel Scheffler if that is his view, then what are the arguments in its favor that he finds persuasive? One obvious possibility is that he thinks the bias is irrational because it is bad for us. The difficulty with this suggestion is that Parfit explicitly rejects it. After concluding that the bias is bad for us, he adds that this ‘does not beg the question about the rationality of this bias. On any plausible moral view, it would be better if we were all happier. This is the sense in which, if we could, we ought not to be biased toward the future.’18 In other words, we would be happier if we lacked the bias, and so it would be better from a moral point of view if we did, but this leaves open the question of whether the bias is rational. To emphasize the point, he says a few pages later that ‘the rationality of an attitude does not depend on whether it is bad for us’.19 But if it is not the (presumed) fact that the bias is bad for us that leads Parfit to think it is irrational, then what are his reasons for that conclusion? The text of Reasons and Persons leaves this obscure. As I’ve said, many of his arguments are ad hominem arguments directed against the S-­theorist. They are about what the S-­theorist must say about the bias rather than about the bias itself. And by emphasizing how difficult most of would find it to believe that the bias is irrational, so that it would be awkward for the S-­theorist to be committed to that view, those arguments create the impression—mistaken, as it happens—that Parfit himself thinks the bias is not irrational, rather than that it is. Perhaps, like Sidgwick and Rawls, Parfit is simply struck by the force of the thought that whether an experience is in the future or in the past is not itself a reasonable ground for having ‘more or less regard for it.’ But nothing has done more to call the force of that thought into question than Parfit’s own arguments about the strength and robustness of our bias toward the future. Can we really suppose that rationality requires us, in Parfit’s hospital case, to hope that we will undergo the painful surgery today? Given the strong preference that many of us will have to the contrary, a preference we are likely to affirm upon reflection, more may be required to persuade us of this than the bare assertion of the rational irrelevance of temporal location. One argument against the rationality of future-­bias appeals to the limitations on the scope of the bias that Parfit mentions. If the bias does not apply to all goods, and if it does not apply in the same way to our third-­person judgments about other people’s pleasures and pains, then this may be said to show that it is rationally anomalous. David Brink, who makes an argument of this kind, observes that our biased preferences are also diachronically unstable. If I view the options in Parfit’s hospital case from a temporal perspective where both options lie in the future—say, before I enter the hospital—then I will prefer the shorter operation to the longer one, even if I know that shorter operation would take 18 Parfit, Reasons and Persons, 177.

19 Parfit, Reasons and Persons, 185.

Temporal Neutrality  95 place at a later time than the longer one. So too if I am asked about my preference when both options lie in the past—as, for example, when I am leaving the hospital. The bias emerges only when, as in Parfit’s original version of the example, the options ‘straddle’ the present: when one of them lies in the past and one of them lies in the future. As Brink says, ‘it’s not about preferring earlier pain to later pain; instead, it’s about preferring past pain to future pain’.20 This, he says, makes the bias ‘more narrow or isolated’.21 Summing up, he concludes that the bias ‘does not generalize well and remains limited in scope and unstable’.22 One thing this form of diachronic instability suggests is that, when we are contemplating alternative actions we might perform or experiences we might undergo or things that might happen to us, we can think of the alternatives in two different ways and can ask two different questions about them. Abstracting from our temporal position, we can ask: which alternative would make my life as a whole go better? But viewing the alternatives from a temporally situated perspective, in which it is salient that some of our life lies in the past and some of it in the future, we can also ask: which one would make the rest of my life go better? We can, for convenience, label these two different ways of thinking the whole-­life perspective and the future-­facing perspective, although we should take care not to reify these ‘perspectives’ or to think of them as constituting elements or modules of the self. They are simply two different ways of thinking that are available to us. Most of the time, there is no conflict between them. We will prefer the same alternative no matter which question we ask. The same alternative that would make my future better would also make my life as a whole better. But sometimes the answers may diverge. The bias toward the future manifests itself in cases where such divergence arises and I give priority to the future-­ facing perspective. When this happens, then we will see the form of diachronic instability

20  D.  Brink, ‘Prospects for Temporal Neutrality’, in C.  Callender (ed.), The Oxford Handbook of Philosophy of Time (Oxford: Oxford University Press, 2011), 352–81, at 377. This point is significant because some people think that the temporal order of one’s experiences can make a difference to the quality of one’s life. For example, some people think that a life filled with misery in the early years but joy in the later years is better than a life filled with joy in the early years but misery in the later years, even if the amounts of misery and joy in the two lives are the same. They think, in other words, that a life with an upward trajectory is better than a life with a life with a downward trajectory. Similarly, some people think that it is better if satisfactions are distributed evenly throughout one’s life than if they are clustered in one temporal period. These are sometimes described as views according to which the ‘narrative structure’ of a life contributes to its value. (See e.g. D. Velleman, ‘Well-­Being and Time’, Pacific Philosophical Quarterly 72 (1991): 48–77, repr. in J. M. Fischer (ed.), The Metaphysics of Death (Stanford, CA: Stanford University Press, 1993), 329–57.) The ‘narrative preferences’ just mentioned are not equivalent to, and are not always compatible with, the bias toward the future. And although they may seem to violate temporal neutrality, they do not. Although they assign significance to the temporal location of good and bad experiences, they do so only insofar as the temporal location of those experiences affects the overall value of one’s life from a neutral perspective. For that reason, they are compatible with an impartial concern for all parts of one’s life. See Brink, ‘Prospects for Temporal Neutrality’, 357–8. 21  Brink, ‘Prospects for Temporal Neutrality’, 377. 22  Brink, ‘Prospects for Temporal Neutrality’, 380.

96  Samuel Scheffler that Brink identifies. If I am asked which of Parfit’s two operations I would prefer before I enter the hospital, my preference will be for the shorter operation. From both the whole-­life and future-­facing perspectives, that is the better ­alternative. But if, as in Parfit’s original example, I am asked the same question when I know that the shorter operation is in the future and the longer operation is in the past, then my preference will be for the longer operation. From the future-­facing but not the whole-­life perspective, that is the preferable alternative. Here, in other words, the two perspectives diverge, and I give priority to the future-­facing perspective. I display a bias toward the future. Even if I know all this in advance, the same pattern of preference is likely to persist. That is, if I am asked for my preference before I enter the hospital, I will continue to prefer the shorter operation even if I know that, should I end up in the situation Parfit describes, I will then prefer to have had the longer operation. And if I am asked for my preference in that situation, then I will continue to prefer to have had the longer operation, even if I know that, had I been asked for my preference before I entered the hospital, I would have preferred the shorter operation. Mere self-­awareness does not produce any change in my pattern of preference. The idea that we can see our alternatives from two different perspectives, and that giving priority to the future-­facing perspective in cases where the two perspectives conflict produces diachronic instability, has been developed and extended by other writers, including Tom Dougherty23 and Caspar Hare.24 Dougherty says that one perspective we take toward ourselves and our experiences is temporal and perspectival, while the other involves seeing ourselves as temporally extended agents. Hare speaks instead of a contrast between our attitudes toward predicaments and our attitudes toward lives. These distinctions are similar to my distinction between the future-­ facing and whole-­ life perspectives, and for present purposes we can treat them as different ways of characterizing the same distinction. Both Dougherty and Hare describe complex cases in which the way we respond to divergence between the two perspectives seems to result in cycling or intransitivity of a kind that most people regard as irrational. The problem arises because, when the perspectives diverge, it appears that we don’t always give priority to the future-­facing perspective. Sometimes we see things from the whole-­life perspective instead. In consequence, we display the bias toward the future in response to some but not all cases of divergence, and it is this that leads to intransitivity. Both Dougherty and Hare believe that the solution, in the cases they describe, is to give one of the perspectives consistent priority

23  T. Dougherty, ‘Future-­Bias and Practical Reason’, Philosophers’ Imprint 15/30 (2015): 1–17. 24 C.  Hare, ‘A Puzzle About Other-­Directed Time-­Bias’, Australasian Journal of Philosophy 86 (2008): 269–77. See also Hare, ‘Time—the Emotional Asymmetry’.

Temporal Neutrality  97 over the other. Dougherty believes that, both in his cases and more generally, we should always give priority to the whole-­life perspective. He agrees with those who think that the bias toward the future is irrational. Hare, by contrast, believes that we should always give priority to the future-­facing perspective, at least in cases like those he describes.

5.  The Tolerant Stance I will return to the question of intransitivity. But first let me put some of my own cards on the table. I take it for granted that the temporal dimension of human life is, for anyone trying to lead such a life, one of its most mysterious and perplexing aspects. Philosophy aside, we experience ourselves as moving from the past into the future, and we struggle to understand what it means, and what it means to us, to have a past and a future: to be self-­conscious creatures extended in time but living always in the present moment. We struggle to understand what our pasts mean to us, what kind of reality they have for us, and what resources, beyond the fragile tool of memory, sustain our ongoing connections to our past. At the same time, we are moving always into a future that presents itself to us as open, in ways that seem both fundamental and elusive, and we are preoccupied with the implications of that openness. We worry and we wonder about what the future will bring, and about how we can and cannot influence it. And we have a whole range of attitudes that we deploy in orienting ourselves to that future: from hope and fear to anticipation and dread to determination and resignation. Moreover, as we look backward to the past and forward to the future, our attitudes and actions are guided by values—including but not limited to conceptions of what would we be good for us—that are themselves shaped by our self-­understanding as temporally extended creatures and by our experience of temporality. We would not have the values we have if we did not understand the temporal dimension of our lives in the ways that we do. At the same time, our values serve in turn to shape our attitudes toward time. We would not have the temporal attitudes that we have if we did not have the values that we do. In my view, and here I am doing no more than articulating an unargued methodological predisposition, the first task of philosophical reflection in this area is to try to understand this complex terrain, and we should not be too quick to assume that every manifestation of temporal bias in our desires or valuing attitudes is irrational. Indeed, since the very term ‘bias’ imports a certain bias, the tendency of philosophers to use the term in this context is to that extent unfortunate (although, in deference to the prevailing usage, I will continue to do so myself). As with studies of rational judgment and decision-­making in other areas, the aim should be to navigate between the complacent assumption that our ordinary thinking must be in good order and the revisionist application of oversimplified

98  Samuel Scheffler models that lack any authority over our actual practices and tendencies of thought.25 Against this background, it may seem tempting to say something like the following about our bias toward the future. It is a basic feature of our practical thought that we can assess our alternatives from both the whole-­life and future-­ facing perspectives. So if it is true that, by virtue of our responses to cases in which the two perspectives diverge, we exhibit certain forms of diachronic instability or are apt to form intransitive preferences in certain contexts, perhaps it is a mistake to try to eliminate this complexity or to insist that one perspective must rationally supersede the other in all cases of conflict. Perhaps we should be less preoccupied with subjecting our responses to rational criticism. Perhaps we should simply accept the fact that, in this respect as in many others, we are complex creatures, and focus our efforts on trying to understand the roles of the two perspectives in our lives. I will call this the tolerant stance toward future-­bias. The tolerant stance urges us to accept both our future-­bias and its limits: both the fact that we exhibit the bias in some contexts and the fact that we do not exhibit it in others. A word of clarification is in order here. As I have said, the bias toward the future manifests itself in cases where the whole-­life and future-­facing perspectives diverge and we give priority to the future-­facing perspective. When we do not exhibit future-­bias, that may be either (a) because the two perspectives do not diverge or (b) because they do diverge and we give priority to the whole-­life perspective. In particular cases, it may be open to debate which of these conditions applies. For example, consider the case of goods that ‘gild or stain our picture of our lives’. Suppose I would prefer to have written five excellent novels in the past rather than to write one such novel in the future, and so I do not manifest a bias toward the future in this instance. It might seem that the reason I do not manifest such a bias is that I am giving the whole-­life perspective priority over the future-­ facing perspective. But another possibility is that, in this case, the two perspectives do not actually diverge, because, say, having written an excellent novel is an achievement that benefits a person throughout the person’s life and not merely at the time when it is accomplished. According to this view, I will actually have a better future, as well as a better life, if I have written five novels in the past. The question of which diagnosis is correct, in this case and others like it, raises deep issues which lie beyond the scope of his paper. The tolerant stance is neutral on this question: it urges acceptance of the limits of our future-­bias however those limits are explained.26

25  This paragraph is drawn, with some modifications, from my discussion in Why Worry about Future Generations?, 121. 26  I am grateful to Jake Nebel for illuminating discussion of these points.

Temporal Neutrality  99

6.  The First Obstacle to Endorsing the Tolerant Stance: Near-­Bias Even for someone of my methodological predisposition, however, there are at least two obstacles to endorsing the tolerant stance. To appreciate the first obstacle, consider the other temporal bias that Parfit discusses, the bias toward the near. Manifestations of the bias toward the near also lead us to exhibit various forms of instability in our preferences. Yet it seems clear that manifestations of this bias can indeed expose us to rational criticism. But if that is right, then more needs to be said to explain why the bias toward the future is different, and why it should be tolerated or accepted rather than criticized. One difference lies in our own attitudes toward the two biases.27 As I have noted, most people are strongly inclined, even upon reflection, to reaffirm their future-­biased reactions in cases like Parfit’s hospital example. And few of them feel that they are making a mistake in doing so. Instead, their preference to have had the longer operation yesterday rather than the shorter operation today continues, with one qualification that I will mention later, to make sense to them. Even upon reflection, in other words, most people believe that the preference to have had the longer operation yesterday makes sense in the situation Parfit describes. This is, of course, compatible with recognizing that that preference is diachronically unstable. To reflectively reaffirm the preference is not to deny that one would have different preferences if both options were either in the past or in the future. Those preferences too seem appropriate upon reflection. With the bias toward the near, things are different. Manifestations of this bias are often subject to criticism by the very people who manifest it. Even when our bias toward the near leads us to postpone dentist appointments, fail to save for retirement, and avoid doing chores today that will only be more difficult to do tomorrow, often we feel uneasy about what we are doing and regard it as revealing a weakness or flaw or vice of some kind. Moreover, many people try to resist this bias, and many people are successful. They do undergo routine dental care and save for retirement, and they rarely put off until tomorrow what they can do today. What might explain this difference in our attitudes toward the two biases?28 Consider the case of pain, and let me begin with some undigested phenomenology. Phenomenologically speaking, both biases reflect a concern to keep pain at bay. We think of our present consciousness as moving forward into the future and, as we move forward, we are concerned to avoid having any pain impinge on that consciousness. We might call this our defensive impulse. Since we experience our 27  This point is also made by C. Heathwood, in ‘Fitting Attitudes and Welfare’, in R. Shafer-­Landau (ed.), Oxford Studies in Metaethics, iii (Oxford: Oxford University Press, 2008), 47–73, at 62. 28  It is noteworthy that, while many people have proposed psychological or evolutionary explanations for each of the two biases considered separately (see, for discussion, Greene and Sullivan, ‘Against Time Bias’, 966–70), the availability of such explanations by itself does not explain the difference in our own attitudes toward the two biases.

100  Samuel Scheffler conscious selves as moving forward in time, the defensive impulse is satisfied by the recognition that an episode of pain is in the past. Past pains will never again impinge directly on our consciousness (which is not to deny that the memory of pain can itself be disturbing). They are over and done with. Future pains, however, pose a threat, since we will eventually cross paths with them; they will indeed impinge on our consciousness. So, from a defensive standpoint, it makes perfect sense to prefer past pains to future pains. Future pains pose a threat in a way that past pains do not. That is why future-­bias makes sense to us. The bias toward the near, meanwhile, also has defensive roots. By postponing our pains, we keep as much temporal distance as possible between us and them, and this has an obvious defensive point. Hence the bias toward the near. Yet, at the same time, this bias is open to criticism from the defensive standpoint itself. By postponing our pains, we do increase the temporal distance between us and them, but it remains the case that we will eventually encounter them, and, by hypothesis, postponement ensures that when we encounter them, they will be worse than they would otherwise have been. So there is a clear sense in which the bias toward the near, as a defensive strategy, is self-­defeating. That is why, even when we are in the grip of the defensive impulse, we may be uneasy about the bias toward the near and may withhold our reflective endorsement of it. We may think of it as constituting a flaw or defect of character. Just to be clear, the phenomenology I have described presupposes that we experience ourselves as moving forward in time, and that we therefore have asymmetrical attitudes toward past and future. It makes no attempt to vindicate that view. The thought is simply that, given that we experience ourselves that way, we can see why the bias toward the near strikes us as problematic in a way that the bias toward the future does not. But we need not appeal to phenomenology to account for our different ­attitudes toward the two biases. We may simply observe that the bias toward the near, unlike the bias toward the future, is not a response to divergence between the whole-­life and future-­facing perspectives. In cases where we ­manifest the bias toward the near, we display a preference for an alternative that is condemned by both perspectives. That is, the alternative we prefer is one that will both make our future go worse and make our life as a whole go worse. No wonder, then, that we are likely to withhold our reflective endorsement of the bias. The considerations I have been discussing suggest that the bias toward the future is a deeper or more securely-­entrenched feature of our attitudes than is the bias toward the near. And they help us to understand how and why that is so. In so doing, they make it easier for someone with my methodological predisposition to endorse the tolerant stance toward future-­bias. They address the first of the two obstacles I mentioned to such endorsement.

Temporal Neutrality  101

7.  The Tolerant Stance Revisited Before describing the second obstacle, let me say a bit more about what the tolerant stance amounts to. In suggesting that we should accept our bias toward the future, I am not suggesting that the bias is rationally required or mandatory.29 There are several reasons why I would not want to say that. First, and most straightforwardly, I simply cannot think of any plausible norm of rationality that might require future-­bias. Second, although I don’t think I have ever met anyone who lacked such a bias, I am in no position, from my armchair, to exclude the possibility that there are such people. Perhaps complete temporal neutrality is the sort of thing one can achieve through a sufficiently disciplined and rigorous regimen of training and meditation. Or perhaps there are rare individuals who are just naturally like Parfit’s Timeless. I doubt it, but I can’t rule it out. And if there are people who genuinely lack a bias toward the future, either as a result of undergoing some training regimen or because that is simply their nature, then I am not inclined to say that they are being irrational or to subject them to rational criticism: not, at least, without knowing a lot more about them. My firmest conviction about such people is that, if indeed they exist, they must be very different from the rest of us in many ways, and I would be immensely curious to know what their lives are actually like. There is a third reason I would not want to say that the bias toward the future is rationally required or mandatory, and it has to do with the qualification I alluded to earlier when asserting that most of us reflectively affirm our future-­biased preferences. What I had in mind was that, despite our reflective endorsement, these preferences may nevertheless be sources of puzzlement or bemusement. One needn’t be a professional philosopher to be struck at times by the thought that there is something strange about our attitudes toward our past and future experiential states and about the broader attitudes toward time that they reflect. It can seem mysterious, at least if one is in a certain sort of mood, that we invest future experiential states with such great emotional and attitudinal significance— that we make them the objects of attitudes like fear, dread, longing, and anticipatory excitement—only to dial back the emotional temperature nearly to zero, and to turn most of our attention away from them, once we have actually undergone them. We may wonder: How can they have merited such intense reactions in advance if we find them so forgettable once we have actually experienced them? 29  The view that the bias is rationally mandatory is strongly suggested by Heathwood, who writes that he is ‘not convinced that reason permits Parfit to prefer to be the patient whose operation is later today’ (‘Fitting Attitudes and Welfare’, 61). He also writes, ‘Parfit is being completely reasonable in preferring that his pain be in the past. In fact, even his no longer caring at all that it occurred is perfectly fitting—not at all inappropriate. Why should he care about it now? No reason—it’s over and done with’ (56–7).

102  Samuel Scheffler Why doesn’t our prospective awareness that this abrupt change will happen do anything to diminish the intensity of our anticipatory attitudes? Why doesn’t my awareness of the fact that I will be indifferent to tomorrow’s pleasure or pain once the following day dawns make me more nearly indifferent today? What does it say about us that we are more focused on future experiences that we have not undergone than on experiences that already belong to our personal histories? What does this say about our relation to our own pasts, about the significance for us of those pasts, and about the sense in which they are ours? I rehearse these questions to make what is only a limited point. Most of us have future-­ biased preferences, and we are not inclined to abandon them upon reflection, but they can also strike us as mysterious, and as exemplifying broader attitudes toward time and toward our lives as temporally extended creatures, that are equally mysterious. And this is another reason why, although I am inclined to accept our future-­bias, I would be reluctant to say that it is rationally mandatory or required, and the tolerant stance does not in fact say that. But if the tolerant stance does not view the bias as rationally required, then neither does it view it as a form of human irrationality to which we must simply resign ourselves. The acceptance that is part of the tolerant stance is not a world-­weary recognition of the inevitable flaws and foibles of human beings. That attitude, or something like it, may be an appropriate response to some familiar forms of human irrationality, such as akrasia, or the bias toward the near, or some of the deliberative anomalies documented by psychologists like Kahneman and Tversky. Many of us do our best on a case-­by-­case basis to avoid falling prey to those forms of irrationality, but we don’t suppose that human beings will ever cease altogether to exhibit them. So while we resist them on the micro-­level, we despair of eliminating them on the macro-­level, and in that sense we accept them. But that is not the kind of acceptance that is urged by the tolerant stance toward future-­bias, because the tolerant stance does not condemn that bias as irrational in the first place. If accepting the bias toward the future amounts neither to thinking that it is rationally required nor to thinking that it is an ineliminable form of human irrationality, then what’s left? I see two possibilities. The most obvious suggestion is that accepting future-­bias means judging it to be rationally permissible though not required. The alternative is to say, not that future-­bias is rationally permissible though not required, but rather that it is one of those things—like, perhaps, our need for love or companionship, or our vulnerability to boredom—that help to define the type of rational creature we are rather than something that is itself up for rational assessment. I find this suggestion tempting, but to make good on it, one would need to explain what the difference between being rationally permissible and being ineligible for rational assessment really amounts to. Since the issues here are subtle and space is limited, I won’t pursue this question any further. Instead, I will assume that, in urging that we accept the bias toward the future as part of our nature as complex creatures, the tolerant stance is

Temporal Neutrality  103 recommending that we think of the bias in one of the two ways just mentioned: either as rationally permissible but not required or as not subject to rational assessment at all.

8.  The Second Obstacle to Endorsing the Tolerant Stance: Intransitivity Let me turn now to the second of the two obstacles I mentioned to endorsement of the tolerant stance. As I said earlier, both Dougherty and Hare describe cases in which, it seems, we respond to divergence between the two perspectives by giving priority to the future-­facing perspective at some moments but to the whole-­life perspective at others, and this results in cycling or intransitivity of a kind that is usually seen as irrational. If that is correct, then the tolerant stance, which neither condemns our future-­bias nor insists that we must display it in every case of divergence, may seem simply mistaken. If the tolerant stance is to be defensible, something must be said to take the sting out of the cycling arguments. It is important to remember that what the cases described by Hare and Dougherty appear to show is not that future-­bias by itself leads to intransitive preferences, but rather that the disposition to manifest future-­bias in response to some but not all instances of divergence between the two perspectives can lead to intransitivity. Dougherty concludes that, in order to avoid such intransitivity, we should give priority to the whole-­life perspective in all cases of divergence and should reject future-­bias as irrational. Hare, on the other hand, argues that we should display more future-­bias, at least in cases like the ones he describes, than most of us already do. The tolerant stance suggests still another possibility, namely, that we need not revise our responses at all. If it is true, as Hare and Dougherty suggest, that we display future-­bias in response to some but not all instances of divergence between the two perspectives, then we can accept that pattern of response even if it leads to the formation of intransitive preferences in cases like those they describe. Whether this is plausible depends on the force of the considerations in favor of the tolerant stance and the force of our reasons to avoid the intransitivities in question. The case in favor of the tolerant stance is simple. The whole-­life and future-­facing perspectives are both undeniably available to us, and our tendency to manifest future-­bias in response to at least some cases of divergence between them is a deep and perhaps ineliminable feature of our agency. Furthermore, most people are strongly inclined to affirm their future-­biased preferences upon reflection. Yet, at the same time, our future-­bias is limited in a number of ways. Those limits too are deeply entrenched features of our attitudes, and ones that we are strongly inclined to affirm upon reflection. In both cases, the tolerant stance asserts, the appropriate response to our attitudes is one of acceptance.

104  Samuel Scheffler What about our reasons to avoid intransitive preferences? As I have said, such preferences are widely taken to be irrational, although there has been some debate about this among philosophers in recent decades.30 I don’t want to intervene in this general debate, except to the extent of trying to ascertain how far considerations of transitivity cast doubt on the rationality of the bias toward the future. One place to begin is by asking why exactly intransitive preferences are thought to be irrational? If this were simply a stipulation about how the term ‘irrational’ was being used, then it would lack any substantive force. But it is more than that. The claim that intransitive preferences are irrational relies on the thought that if someone prefers A to B, B to C, and C to A, then it strikes us immediately that something has gone substantively awry. If it is not exactly that the person is reasoning badly, then perhaps it is that they don’t really grasp the concept of a preference. The important point is that it is the substantive thought that something has gone awry that underwrites the charge of irrationality. This thought is often illustrated or reinforced by arguments to the effect that a person with intransitive preferences will be vulnerable to ‘money-­pumping’. That is, such a person will be disposed to accept a sequence of trades that will leave her worse-­off than she was when she started. She will, say, pay one dollar to move from C to B, one dollar to move from B to A, and one dollar to move from A to C. So she will end up back where she started, only three dollars poorer. And then the cycle can simply start all over again. Clearly, something has gone wrong. This means that if, by manifesting future-­bias in response to some but not all cases of divergence between the whole-­life and future-­facing perspectives, we are liable to form intransitive preferences in certain situations, then this is a serious obstacle to endorsing the tolerant stance. Before trying to decide whether the obstacle can be overcome, it may help to look at some of the specific sorts of intransitivities that have been said to arise. Philosophers have provided examples of several different kinds, but I will consider just two: one of Hare’s and one of Dougherty’s. Hare’s example relies on an important observation about the alleged asymmetry, with respect to future-­bias, between first-­person and third-­person cases. As Hare observes, Parfit’s sense that he would experience no future-­bias in the case of his

30  For challenges to transitivity, see e.g. W. Quinn, ‘The Puzzle of the Self-­Torturer’, Philosophical Studies 59 (1990): 79–90; L. Temkin, ‘A Continuum Argument for Intransitivity’, Philosophy & Public Affairs 25 (1996): 175–210; S.  Rachels, ‘Counterexamples to the Transitivity of Better Than’, Australasian Journal of Philosophy 76 (1998): 71–83; A.  Friedman, ‘Intransitive Ethics’, Journal of Moral Philosophy 6 (2009): 277–97; T. Willenken, ‘Deontic Cycling and the Structure of Commonsense Morality’, Ethics 122 (2012): 545–61; and L. Temkin, Rethinking the Good (Oxford: Oxford University Press, 2012). For defenses of transitivity in the face of such challenges, see e.g. A. Norcross, ‘Comparing Harms: Headaches and Human Lives’, Philosophy & Public Affairs 26 (1997): 135–67; K. Binmore and A. Voorhoeve, ‘Defending Transitivity Against Zeno’s Paradox’, Philosophy & Public Affairs 31 (2003): 272–9; and M.  Huemer, ‘Transitivity, Comparative Value, and the Methods of Ethics’, Ethics 123 (2013): 318–45.

Temporal Neutrality  105 distant mother’s illness relies crucially on the fact that she is distant and that he cannot see or communicate with her. In cases where those features are lacking, Hare argues, we do experience a bias toward the future in behalf of other people. Imagine, he says, that you receive a letter on July 28th informing you that your daughter, who is spending the summer at a monastery in Japan, was to have either a more painful tooth extraction on July 27th or a less painful extraction on July 29th. Hare thinks that, in this situation, most people would prefer for their daughter to have the less painful extraction on the 29th. Here, as in Parfit’s case, we do not display future-­bias in behalf of someone else. But now Hare asks us to imagine that you receive the letter on July 26th and fly immediately to Japan, arriving at your daughter’s bedside on July 28th. You find her sleeping restlessly, but are uncertain whether she had the more painful operation on the 27th or is scheduled for the less painful operation on the 29th. Her restlessness might be a sign either of post-­operative discomfort or of pre-­operative anxiety. In this case, Hare believes, most people would prefer for their daughter to have had the more painful extraction on the 27th. If he is right about these cases, then what this shows is that we are prone to experience future-­bias in behalf of other people in some contexts but not others. In some contexts, it seems, we give the future-­facing perspective of someone else priority over that person’s whole-­life perspective. In other contexts, we do not. This is an important observation in its own right, because it complicates our understanding of, and so helps further to illuminate, the place of these two perspectives in our lives. But Hare also argues that our tendency to toggle back and forth between future-­biased and temporally neutral responses in third-­ person cases, depending on whether the person we are concerned about is either near or distant, may result in the formation of intransitive preferences. He illustrates the point with a variant of the previous cases. Suppose again that you receive the letter on July 26th informing you of your daughter’s situation, but that this time you know only that she is either in a monastery in the north of Japan or in a monastery in the south of Japan. In response, you and your spouse fly off to the two different monasteries. Your spouse, who is better than you at providing post-­operative comfort, goes to the southern monastery. You, who are better than your spouse at allaying pre-­operative anxiety, head for the northern monastery. You arrive there on July 28th and find a sleeping figure who may or may not be your daughter. What, in this situation, do you hope for? Your daughter may be nearby or she may be far away. And she may have had the operation already or she may be about to have it. So there are four possibilities: A.  B.  C.  D. 

She is nearby and has already had the more painful operation. She is far away and has already had the more painful operation. She is nearby and will have the less painful operation. She is far away and will have the less painful operation.

106  Samuel Scheffler In this case, Hare says, you will prefer B to A, because if she has had the more painful operation she will be better off with your spouse than with you. And you will prefer C to D, because if she is going to have the less painful operation she will be better off with you than with your spouse. You will prefer A to C, because when she is nearby your future-­bias leads you to prefer less pain in her future, and you will prefer D to B, because when she is distant, your neutralist stance leads you to prefer that she experience less pain overall. But this means that you prefer B to A, A to C, C to D, and D to B. Your preferences are intransitive. Notably, the intransitivity depends on the fact that, when comparing A and C, the whole-­life and future-­facing perspectives diverge. So too for B and D. Yet in preferring A to C, you give the future-­facing perspective priority over the whole-­life perspective, while in preferring D to B, you do the reverse. Some people may feel that their own reactions in the cases Hare describes would be either consistently future-­biased or consistently neutralist, so that, for them, no intransitivity would arise. Other people may have reservations about some of the details of what Hare says.31 Let’s suppose, however, that many of us would react in just the ways Hare thinks we would and, accordingly, that we would be led to form intransitive preferences in his final case. Before attempting to assess the significance of this conclusion, let me first describe one of Dougherty’s examples. Dougherty argues that although we display future-­bias with respect to hedonic experiences of pleasure and pain, we display no such bias in cases where we are contemplating exchanges between hedonic experiences and other goods. Suppose you are willing to exchange twenty dollars now for the gustatory pleasure that a restaurant meal will bring you tomorrow. So you think this future pleasure is worth twenty dollars. If you were relevantly future-­biased, however, you might well think, after having eaten the meal, that you had overpaid, because the meal and its associated gustatory pleasure would then be in the past, and a past meal, with its associated pleasure, would not be worth as much money to you as a future meal, with its associated pleasure. So, if you were relevantly future-­biased, you would regret having paid twenty dollars for the meal. After all, if the twenty dollars were still available to you, you could use it to purchase future gustatory pleasure, and you prefer future gustatory pleasure to past gustatory pleasure. Yet from a whole-­life perspective, there is no basis for such a preference and, accordingly, no basis for regretting your purchase of the past meal. From that perspective, it makes no difference whether your gustatory pleasure lies in 31  Will you really prefer C to D? Admittedly, your daughter will be better off with you than with your spouse during the pre-­operative period, but she will be better off with your spouse during the post-­operative period, and if she is going to have her operation tomorrow, she will need to go through both periods. Perhaps we could stipulate that, if C obtained, your spouse would be able to travel from the southern monastery to the northern monastery in time to provide your daughter with post-­ operative comfort. So your daughter would have the less painful operation and she would receive optimal pre-­operative and post-­operative care. If so, however, then is it still clear that you would prefer A to C?

Temporal Neutrality  107 the future or the past. And in fact, Dougherty says, you probably won’t regret your purchase. With respect to exchanges between hedonic and non-­hedonic goods, he claims, we generally see things from a whole-­life perspective and remain temporally neutral. Or, as he puts it, we maintain a ‘fixed exchange rate’ between hedonic and non-­hedonic goods, and we think it rational to do so. But if we have fixed exchange rates in such cases, then we cannot, on pain of violating the global requirement to have transitive preferences, be future-­biased in cases involving only hedonic goods. For if, given my fixed exchange rate, I am indifferent between a dollar and a minute of future gustatory pleasure, and equally indifferent between a dollar and a minute of past gustatory pleasure, then transitivity requires that I also be indifferent between a minute of past pleasure and a minute of future pleasure. If, like most of us, I prefer future pleasure to past pleasure, then my preferences are not transitive. Again, we may or may not agree with what Dougherty says about this example. For the sake of argument, though, let’s suppose that both he and Hare have provided examples in which people are likely to alternate between future-­ biased and neutralist responses and, in consequence, to form intransitive preferences. In Hare’s example, the intransitivity results from the fact that, in our attitudes toward our loved ones’ sufferings, we seem to toggle back and forth between neutralist and future-­biased preferences depending on whether our loved ones are nearby or far away. In Dougherty’s example, the intransitivity results from the fact that, in our attitudes toward exchanges of goods, we seem to toggle back and forth between neutralist and future-­ biased responses depending on whether the exchanges are between hedonic goods only or between hedonic and other kinds of goods. How troubled should we be by these intransitivities? How great an obstacle do they present to endorsement of the tolerant view of future-­bias? On one interpretation, the point of the examples is simply to illustrate how intransitive preferences can arise in cases involving future-­bias. It is not to establish the irrationality of those preferences, which is instead taken for granted. So interpreted, however, these examples won’t help us very much if what we are trying to do is precisely to decide whether the kind of intransitivity to which we may be liable if we are sometimes but not always future-­biased is a manifestation of irrationality. We have already agreed that, if someone has intransitive preferences, then it seems to most people as if something has gone seriously awry. Yet the examples depend for their force on the thought that, although future-­bias seems appropriate to most people in some cases of divergence between the whole-­ life and future-­facing perspectives, temporal neutrality seems appropriate in others. This means that, if we imagine trying either to eliminate future-­bias altogether or to extend its scope more widely, we are likely to meet with internal resistance: not only with a feeling that it would be difficult to enforce such a decision, but also with a sense that something would have gone awry if we did. It may look, then,

108  Samuel Scheffler like we are at an impasse, and we cannot overcome the impasse by simply assuming that intransitivity is always irrational. Sometimes when people give examples in which a certain pattern of concern is said to lead to the formation of intransitive preferences, a different interpretation of these examples may be intended. The point may be to establish that we are liable to engage in self-­defeating behavior if we exhibit the target pattern of concern and, accordingly, that our intransitive preferences are problematic not only theoretically but also practically. We may, for example, be vulnerable to the kind of exploitation that is involved in money-­pumping. On this interpretation, the point of such examples is not merely to illustrate the irrationality of intransitive preferences; it is to provide an independent argument for it. But in the examples of Hare’s and Dougherty’s that I have described, there is no possibility of money-­pumping, because the options among which one is forming one’s preferences include different past histories, and there is no way that a money-­pumper could credibly offer to bring about the past history of one’s choice.32 Even if we were vulnerable to money-­pumping in those cases, moreover, the claim that our intransitive preferences were irrational for that reason would take us back to the sort of argument that Parfit rejected, the argument that says future-­bias is irrational because it is bad for us. And even if we were prepared to admit such arguments as legitimate in principle, the badness of the specific intransitivities arising from future-­ bias (in combination with certain other attitudes) would have to be judged comparatively. We would have to decide whether the practical problems exemplified by the threat of the money pump in those cases were more severe than whatever practical disruptions would be produced by the elimination of future-­bias. This would require not only an assessment of the severity of the money-­pumping threat but also a comprehensive accounting of the advantages and disadvantages of Timelessness. And, as I have argued, we do not yet have such an accounting. Furthermore, even if, in addition to making the suppositions already enumerated, we also supposed that the money-­pumping threat was the greater practical problem, that would still not show that future-­bias was irrational: only that, in the relevant cases, we should display either more or less future-­bias than most of us are now disposed to do. Dougherty thinks that we tend to overlook the intransitivities to which he calls attention because our thinking is ‘compartmentalized’. When we think about exchanges between hedonic goods and goods of other kinds, we conceive of ourselves as temporally extended creatures and take a neutralist stance. But when we think about exchanges solely among hedonic goods, we view things 32  In ‘On Whether to Prefer Pain to Pass’ (Ethics 121(2011): 521–37), Dougherty gives a different example which is meant to show that people who are both future-­biased and risk-­averse can be turned into ‘pain pumps’.

Temporal Neutrality  109 perspectivally and exhibit a bias toward the future. Our tendency to treat the two sorts of exchange differently derives from these ‘twin aspects of our identities’,33 and our compartmentalized thinking leads us to overlook the ‘tension’ between them. By calling attention to the intransitivities that result when exchanges of the two different kinds are considered as part of a single set of options, Dougherty aims to ‘expose this compartmentalization’,34 and to help us overcome it. He wants us to recognize that ‘fundamentally we are temporally extended agents’,35 and, accordingly, that we should always give priority to what I have called the whole-­ life perspective. It is this perspective that reflects ‘who we really are’,36 and by giving it priority over the future-­facing perspective whenever conflicts arise we can overcome compartmentalization and avoid intransitivity. But whether compartmentalization is a bug or a feature may be in the eye of the philosophical beholder. Some will be tempted to turn Dougherty’s diagnosis on its head. Since the whole-­life and future-­facing perspectives are both available to us, and since, in that sense, both of them are part of ‘who we really are’, we cannot expect to eliminate either of them and compartmentalization may be a reasonably effective technique for managing the tension between them. If, by failing to apply the norm of transitivity across ‘compartments’ in the same way that we do within compartments, we accommodate both perspectives while minimizing agential dissonance, and if, in this way, we manage to stay out of practical trouble too, then it is not clear that this is irrational.

9.  Future-­Bias and Regret-­Avoidance Our tendency to compartmentalize is also relevant to an argument against future-­ bias that has been developed by Preston Greene and Meghan Sullivan.37 Unlike Hare and Dougherty, Greene and Sullivan do not focus on cases of cycling or intransitivity. In a somewhat similar spirit, however, they maintain that future-­ bias in combination with certain other attitudes can lead to self-­defeating behavior. On that basis, they argue that the bias is irrational. Simplifying a bit, their main argument is roughly as follows. It is rationally permissible to avoid acting in ways one is certain one will later regret, where regret is understood not as an affective state but rather as a preference that one had acted otherwise. But future-­ biased agents who wish to avoid certain regret will have to postpone their pleasurable experiences for as long as possible, if they are to avoid regret about those experiences being in the past. If Jack chooses to have an enjoyable meal on 33  ‘Future-­Bias and Practical Reason’, 15. 35  ‘Future-­Bias and Practical Reason’, 15. 37  In ‘Against Time Bias’.

34  ‘Future-­Bias and Practical Reason’, 13. 36  ‘Future-­Bias and Practical Reason’, 15.

110  Samuel Scheffler Monday rather than waiting until Friday, when those are the options available to him, then as soon as the meal is over, he will strongly prefer that it were in the future rather than the past, which is to say he will regret having scheduled it for Monday. And if future-­bias is rational, then Jack’s regret is rational. Greene and Sullivan call this the scheduling problem. Even worse, future-­biased agents who wish to avoid certain regret will have to forgo greater goods in the immediate future in order to secure lesser goods in the further future. Billy might forgo, say, two cookies now in order to be assured of having one cookie, or even a morsel of a cookie, at some point in the future. After all, if he chose to have two cookies now, then as soon as he had eaten them he would regret his choice, since his eating experience would then be in the past, and he could instead have had a cookie-­eating experience still ahead of him. Greene and Sullivan call this the meager returns problem. According to Greene and Sullivan, it is irrational to act in ways that lead to the scheduling and meager returns problems. Yet if one chooses to avoid certain regret and one is future-­biased, one will act in these ways. So either it is not permissible to avoid certain regret or future-­bias is irrational. But the former, they say, is implausible; surely there is no rational requirement that one not avoid acting in ways that one will certainly regret. It follows that future-­bias is irrational. One thing that is striking is that, although we are future-­biased, most of us do not act in ways that lead to the scheduling and meager returns problems. We do not postpone restaurant meals indefinitely and we don’t exchange present cookies for future cookie-­morsels. How is this to be explained? One suggestion is that, in general, we simply don’t care about avoiding certain regret. But this does not seem true. Although we sometimes do things we know we will regret, usually we try to avoid doing that. Another suggestion is that our bias toward the near offsets our bias toward the future, so the combination of the two biases prevents us from acting irrationally in the ways Greene and Sullivan describe.38 For example, our impulse to accept a cookie morsel later in preference to two whole cookies sooner, and thereby to avoid future regret, is offset by our near-­biased impulse to prefer cookie pleasure sooner to cookie pleasure later. But if that were the whole story, then we would expect people routinely to experience regret once their restaurant meals were over and their cookies consumed, since at that point their bias toward the near would no longer be engaged and their bias toward the future would lead them to prefer that they still had the relevant pleasures ahead of them. Yet it is not clear that most people are subject to all-­things-­considered preferences of this kind. If not, then perhaps the reason that future-­biased people who wish to avoid regret nevertheless do not fall prey to the scheduling and meager returns problems is that they do not actually expect to experience regret in the relevant cases. 38  Greene and Sullivan themselves appear to endorse this suggestion, at least as it applies to the scheduling problem. See ‘Against Time Bias’, 960.

Temporal Neutrality  111 Why wouldn’t they experience such regret? Presumably because their future-­ bias does not come into play, or does not come into play with full force, in these cases. But why not? It may be relevant that, in the cases that Greene and Sullivan present, the putatively regrettable outcome comes about, if it does, as a result of one’s own choice, rather than resulting from factors exogenous to one’s choices. This contrasts with Parfit’s hospital case, for whether you had the longer operation yesterday or will have the shorter operation today is not a function of any choice you did or will make. But one chooses to have two cookies now rather than a cookie morsel later. At the moment when one makes that choice, moreover, the two-­cookie alternative is (possible regrets aside) preferable from both the future-­ facing and whole-­life perspectives. In other words, both one’s future and one’s life as a whole will go better if one has two cookies now rather than a cookie morsel later. And perhaps our future-­bias does not come into play, at least not with full force, in cases with this structure. Perhaps, more precisely, we do not manifest future-­bias in cases where (a) the putatively regrettable outcome is one that we ourselves chose, and (b) we chose that outcome, at least in part, because it was (as good as or) better than the available alternatives from both the future-­facing and whole-­life perspectives. This may also help to explain why, in Dougherty’s case, we are unlikely to feel regret after eating and paying twenty dollars for a meal we had judged in advance to have been worth that much. If these suggestions are correct, then the reason we don’t incur the scheduling and meager returns problems, despite being future-­biased, is that we don’t expect our future bias to lead to regret in the relevant range of cases, and so there is no regret to be avoided. If that is the explanation, it reveals another limit on the scope of our future-­bias. But perhaps the fact that we don’t act in ways that lead to the scheduling and meager returns problems should be explained in a different way. Perhaps, even if we do expect to experience regret in cases of the kind described by Greene and Sullivan, we see no need to avoid regret in cases of that type. Does this mean that we deny the permissibility of avoiding regret in such cases? That is not what I am suggesting. Greene and Sullivan acknowledge that, rather than flatly denying the permissibility of avoiding certain regret, one might instead hold that ‘one can [permissibly] aim to avoid regret except in cases in which doing so leads to the scheduling or meager returns problems’.39 They allow that this modified norm would avoid the conclusion that future-­bias is irrational, but they also say that it would be unacceptably ad hoc. My suggestion, however, is not that we affirm such a norm, but rather that we simply don’t in practice seek to avoid certain regret in these cases. And while one might say that if this is so, then we are guilty of drawing ad hoc distinctions, one might instead say that it reveals yet another way in which our thinking is compartmentalized. Of course, a critic might insist that

39  ‘Against Time Bias’, 965.

112  Samuel Scheffler such compartmentalization is itself irrational. But perhaps the correct conclusion is that, for creatures with our repertoire of attitudes, it is a reasonably effective way of avoiding irrationality and staying out of practical trouble. In short, I have suggested two possible explanations for why we don’t generally act in ways that lead to the scheduling and meager returns problems. One possibility is that, owing to an implicit limit on the scope of our future-­bias, there is no future regret to be avoided in the relevant cases. The other possibility is that, owing to our tendency to compartmentalize, we simply don’t treat future regret as something to be avoided in those cases. The two explanations might be combined; perhaps, up to a point, we don’t expect to experience regret in the relevant cases and, beyond that point, we don’t treat the regret as something to be avoided. Consider again the cookie example. Perhaps, on the one hand, we don’t expect to experience much regret about having chosen the two cookies after we have eaten them, even though we will then have no cookie morsel to look forward to; and perhaps, on the other hand, even if we do anticipate some regret, we don’t see that regret as something to be avoided. Whatever the explanation, the fact remains that the combination of future-­bias and a desire to avoid regret does not in practice lead us to act irrationally. So it is not clear why reflection on that combination of attitudes should lead us to condemn our future-­bias as irrational.

10.  Conclusion It is time to draw this discussion to a close. I have not considered all the arguments that have been put forward against the rationality of future-­bias.40 For that reason and others, what I have said has not been conclusive. But I hope I have managed to explain why I am not yet ready to abandon the tolerant stance or to join Sidgwick, Rawls, Parfit, and the many others who have advocated temporal neutrality and who have, either explicitly or implicitly, condemned the bias toward the future as irrational. For now, at least, I am inclined to align myself instead with McTaggart, who writes: the anticipation of a good in the future produces greater happiness in the present than the memory of an equal good in the past. There may be no more reason for this than there is for a man’s preference of burgundy to claret, or claret to burgundy. But the absence of reason is not here contrariety to reason. It would, no doubt, be possible for a man to hold that his nature would be more admirable

40  For a sustained defense of temporal neutrality, see M.  Sullivan, Time Biases (Oxford: Oxford University Press, 2018). Many but not all of Sullivan’s arguments against future-­bias in that book recapitulate arguments I have addressed in this essay, including arguments of Dougherty’s that I have discussed as well as arguments advanced by Greene and Sullivan in ‘Against Time Bias’.

Temporal Neutrality  113 if good and evil affected him to the same extent when they were past as when they were future. But I do not know any reason why anyone should hold this.41

Bibliography Binmore, K., and Voorhoeve, A., ‘Defending Transitivity Against Zeno’s Paradox’, Philosophy & Public Affairs 31 (2003): 272–9. Brink, D., ‘Prospects for Temporal Neutrality’, in C.  Callender (ed.), The Oxford Handbook of Philosophy of Time (Oxford University Press, 2011), 352–81. Dougherty, T., ‘Future-Bias and Practical Reason’, Philosophers’ Imprint 15/30 (2015): 1–17. Dougherty, T., ‘On Whether to Prefer Pain to Pass’, Ethics 121 (2011): 521–37. Friedman, A., ‘Intransitive Ethics’, Journal of Moral Philosophy 6 (2009): 277–97. Greene, P., and Sullivan, M., ‘Against Time Bias’, Ethics 125 (2015): 947–70. Hare, C., ‘A Puzzle About Other-Directed Time-Bias’, Australasian Journal of Philosophy 86 (2008): 269–77. Hare, C., ‘Time—the Emotional Asymmetry’, in H.  Dyke and A.  Bardon (eds.), A Companion to the Philosophy of Time (Oxford: John Wiley & Sons, Inc., 2013), 507–20. Heathwood, C., ‘Fitting Attitudes and Welfare’, in R.  Shafer-Landau (ed.), Oxford Studies in Metaethics, iii (Oxford University Press, 2008), 47–73. Huemer, M., ‘Transitivity, Comparative Value, and the Methods of Ethics’, Ethics 123 (2013): 318–45. McTaggart, J.  M.  E., The Nature of Existence, volume ii (Cambridge: Cambridge University Press, 1927). Norcross, A., ‘Comparing Harms: Headaches and Human Lives’, Philosophy & Public Affairs 26 (1997): 135–67. Parfit, D., Reasons and Persons (Oxford: Oxford University Press, 1984). Parfit, D., On What Matters, volumes i and ii (Oxford: Oxford University Press, 2011). Quinn, W., ‘The Puzzle of the Self-Torturer’, Philosophical Studies 59 (1990): 79–90. Rachels, S., ‘Counterexamples to the Transitivity of Better Than’, Australasian Journal of Philosophy 76 (1998): 71–83. Rawls, J., A Theory of Justice (Cambridge, MA: Harvard University Press, 1971). Scheffler, S., Why Worry about Future Generations? (Oxford: Oxford University Press, 2018).

41  The Nature of Existence, ii. 350.

114  Samuel Scheffler Sidgwick, H., The Methods of Ethics (7th edn., London: Macmillan & Co., 1907; repub. Indianapolis: Hackett Publishing Company, 1981). Sullivan, M., Time Biases (Oxford: Oxford University Press, 2018). Temkin, L., ‘A Continuum Argument for Intransitivity’, Philosophy & Public Affairs 25 (1996): 175–210. Temkin, L., Rethinking the Good (Oxford: Oxford University Press, 2012). Velleman, D., ‘Well-Being and Time’, Pacific Philosophical Quarterly 72 (1991): 48–77; repr. in J.  M.  Fischer (ed.), The Metaphysics of Death (Stanford, CA: Stanford University Press, 1993), 329–57. Willenken, T., ‘Deontic Cycling and the Structure of Commonsense Morality’, Ethics 122 (2012): 545–61.

5

What Is the Opposite of Well-­Being? Shelly Kagan

1.  Introduction I call this paper What is the Opposite of Well-­Being? but I think of that as more of a nickname. Its real title—the one I secretly call it by—is cumbersome and impractical: What is Ill-­Being When Well-­Being is Enjoying the Good? That longer title is significantly more precise concerning my topic; but in addition to being rather unwieldy it is also relatively opaque. Many people reading the longer title would have little or no idea what I am talking about. So let me start by trying to explain it. That should give you a fairly good sense of the issue we’ll be exploring. Starting at the beginning, then, I use the term ‘ill-­being’ to refer to the negative aspect of individual welfare, those elements which directly constitute a person’s being badly or worse off.1 This is a topic that doesn’t get much attention in the philosophical literature. When we introduce the standard theories of well-­being, whether in our writings or in talking to our students, we typically focus on the positive aspects of welfare—those elements which constitute a person’s being well off. We talk about desires being satisfied, or the presence of pleasure, or the possession of various objective goods (like knowledge, or love, or achievement). Details aside, these theories agree that to the extent that these things obtain one is well off, or at least, one is better off than one would be without them. Presumably, however, a complete theory of well-­being would also include various negative elements. For there are things that can directly lower one’s well-­ being (as opposed to merely being the absence of the positive elements), leaving one badly off, or at least worse off than one might otherwise be without them. For example, hedonists think that pleasure increases well-­being, but pain reduces it. Pain thus belongs in the hedonist’s theory of ill-­being. But despite the familiarity of this idea, and the plausibility of thinking that each theory of well-­ being will need some account of ill-­being, the topic is rarely investigated with any care. We note the fact that hedonism views pain as bad for you, but then, having

1  I have argued elsewhere that we need to distinguish between questions about how well off a person is and how well their life is going. (See ‘Me and My Life’, Proceedings of the Aristotelian Society 94 (1994): 309–24.) In this paper I focus exclusively on the former. Shelly Kagan, What Is the Opposite of Well-­Being? In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0006

116  Shelly Kagan pointed this out, we normally limit further discussion to the things that are good for you, suggesting, perhaps, that we do this for ‘simplicity’. The assumption, I suppose, is that it is fairly straightforward to see how to extend our discussion of the positive aspects of individual welfare—well-being, strictly so called—so as to cover the negative aspects, ill-­being, as well. My own belief is that this assumption, rarely stated explicitly, is actually mistaken, and that it is often surprisingly difficult to say how best to extend a given theory of well-­being to cover the negative case. Or so I have argued elsewhere.2 But in any event, what I want to do in this paper is to attempt to extend a particular theory of well-­being in just this way, to see how it might best cover not just well-being, but ill-being as well. However, the particular theory of well-­being that I have in mind is not one of the standard, familiar theories. It is a hybrid theory, drawing on elements from some of the more familiar views, and while the basic idea behind this theory is not particularly original to me, the theory that I have in mind is not yet especially well known. Indeed, although others have discussed views like this, it has no standard name. I call it well-­being as enjoying the good.3 In a moment, I’ll give you a quick introduction to this view. But even before I do that, you should now be in a position to understand what I will be up to in the rest of the paper. I want to take a particular nonstandard theory of well-­being—well-­being as enjoying the good—and ask how to extend it to cover the negative case, ill-­being. Hence the question posed by the ‘secret’ title of the paper: what is ill-­being when well-­being is enjoying the good? Let me be clear, right at the outset, that my goal here is not to defend this nonstandard theory of well-­being, nor to explore its positive aspects in any detail. While I find it plausible and attractive, my purpose here is the rather more limited one of asking how best to extend the view. As we’ll see, the answer isn’t obvious. But if I am right that this theory is indeed a plausible account of well-­being (worthy of consideration, if nothing more, along with its more familiar rivals) then the task I’ve set for the paper should be worth undertaking.

2.  Well-­Being as Enjoying the Good, Introduced The nonstandard theory of well-­being that I want to explore emerges in a natural way out of some familiar concerns about simpler theories. According to hedonism, well-­being consists in the presence of pleasure and the absence of pain. One standard worry about such theories is that one’s pleasures may be built on an 2  See ‘An Introduction to Ill-­Being’, Oxford Studies in Normative Ethics 4 (2014): 261–88. 3  See ‘Well-­Being as Enjoying the Good’ (Philosophical Perspectives 23 (2009): 253–72) for a fuller introduction to the view as well as references to other statements of the same general idea.

What Is the Opposite of Well-Being?  117 illusion: you may take pleasure in the possession of goods that in fact you don’t actually possess. Many of us believe, for example, that life on an experience machine—which directly creates the relevant mental states without actually giving you the external goods you take yourself to have—is far from an ideal picture of individual well-­being.4 Another worry is that one may take pleasure in things that one does in fact possess, but which are unworthy or less worthy of being valued in this way. For example, the pleasure that the sadist takes in the pain of others strikes many people as having little or no value. In response to such concerns, people sometimes turn to objective or objective list theories, according to which well-­being consists in the actual possession of various objective goods. This neatly solves the problems just noted for hedonism. If you don’t actually have the goods you take yourself to have, then you aren’t really made better off. And if the ‘goods’ you possess are actually trivial—or worse still, ­ob­jec­­t­ive­ly bad—then you are made only minimally better off (or not made better off at all). Friends of this approach can of course argue about what goods are the most valuable ones—though friendship, love, knowledge, and achievement figure prom­in­ent­ly on most such lists—but they share a commitment to the idea that if you really do have the relevant objective goods in your life, then you are well off. But this raises the worry that on such an account you might be extremely well off even though you are utterly miserable. After all, the mere fact that your life contains the relevant goods does nothing at all to guarantee that you take any pleasure in the possession of those goods. If someone is loved, accomplished, and knowledgeable, but despises love, knowledge, and accomplishment—and is made miserable by the presence of these things in her life—that hardly strikes many people as an image of a person very well off. To be sure, one can soften the force of this objection by including pleasure on the list of objective goods, so that one is ideally well off only if one is pleased as well as having the various other goods. But this still seems to leave open the possibility that if one did well enough in terms of the other goods, and yet lacked pleasure nonetheless—or indeed was fairly unhappy—one could still be very well off indeed. And that strikes many people as unacceptable as well. It is at this point that the possibility of a hybrid view suggests itself. Perhaps well-­being requires both the actual possession of various objective goods and the taking of pleasure in the possession of those goods. Perhaps neither ingredient taken alone suffices for well-­being, but taken together they do. As I’ve already remarked, I call views of this sort well-­being as enjoying the good. According to this idea, even if you have an objective good in your life, the mere presence of that good doesn’t enhance your well-­being unless you also take pleasure in the good. No pleasure, no bump in well-­being. Here is a way of graphically displaying the basic idea. (See Figure 5.1.) 4  See Robert Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974), 42–5.

118  Shelly Kagan Well-being

Pain

G

Pleasure

Ill-being

Figure 5.1  Well-­being as Enjoying the Good

Let the x axis represent pleasure and pain. Points to the right of the origin represent the person’s taking pleasure in (being pleased by) some object; the further to the right, the greater the pleasure. Points to the left of the origin represent the person’s taking pain in (being pained by) the object; the further to the left, the greater the pain. Let the y axis represent the boost or hit to well-­being, with points above the origin representing increases to well-­being, and points below the origin representing increases to ill-being. (The further from the origin, the greater the boost or hit.) This graph displays the line for an objective good of a given size. If you have that good (in that size) in your life, then provided that you take pleasure in the possession of that good, your welfare is thereby increased. The greater the pleasure, the greater the boost to your well-­being. That’s why the line slopes up and to the right. But if you take no pleasure in the possession of the good, then there is no increase in your well-­being. That’s why the line passes through the origin. In Figure 5.1 the line is drawn straight, representing not only the idea that greater and greater pleasure in the given good results in a greater and greater increase in well-­ being, but more particularly, that well-­ being increases linearly with the increase in pleasure. Arguably, however, one might want to reject the latter thought. Perhaps as the amount of pleasure increases, the corresponding increase in well-­ being grows smaller, so that it takes larger and larger increases in pleasure to gain similar increases in well-­being. I find this a plausible idea. If so, then the line will need to curve to the right somewhat, rather than being drawn straight. More radically, perhaps at a certain point further increases in pleasure will bring no further increase in well-­being at all. The person will have ‘extracted’ all the well-­being that is to be had from that particular good. Beyond that point,

What Is the Opposite of Well-Being?  119 even though the person may be getting more pleasure from the good in question, this won’t actually involve any greater boost in the person’s level of well-­being. If we accept this idea then the line will actually plateau—go flat—at the point in question. Alternatively, perhaps the line only asymptotically approaches the upper bound, so that increases in pleasure always do make one somewhat better off, but by a vanishingly small amount. Even if so, there will still be a limit to how much well-­being can be found in the possession of any given good. I find one or another version of this more radical proposal plausible too, though it won’t play a significant role in what follows. In any event, there is no reason to assume that the line will be identical for all goods (or all sizes of goods). On the contrary, it seems plausible to think that greater goods will have lines ‘higher up’ the y axis, so that taking pleasure in a greater good will constitute a greater increase in well-­being than the same amount of pleasure taken in a lesser good. Similarly, if one does accept the idea that there is an upper bound to the amount of well-­being that can be gotten out of the possession of any given good, it seems plausible to suggest that the amount of this upper bound will be greater for larger goods. Accordingly, an accurate portrayal of well-­being as enjoying the good will involve more than a single line. Instead it might look a bit more like Figure 5.2 (where G is a greater good than G, which is in turn a greater good than g). Well-being

G G g

Pain

Pleasure

Ill-being

Figure 5.2  The Everywhere Differentiated View

Although this is one way the view could be developed, there are alternatives. Notice that in Figure 5.2 each line is distinct from the others (except at the origin). This means that for any given (nonzero) level of pleasure, the boost in well-­ being is greater for greater goods. Call this the everywhere differentiated view. But

120  Shelly Kagan it could be, instead, that the lines for different goods share initial segments, so that at low enough levels of pleasure you will get the same boost to well-­being as long as the good you are taking pleasure in is large enough. It is only as the level of pleasure increases that various lesser goods begin to ‘peel off ’ and are ‘left behind’, so that they will no longer yield boosts to well-­being comparable to what greater goods provide. This alternative view is shown in Figure 5.3. As this picture is vaguely reminiscent of a feather (with its central spine and diverging branches), we can call this the feather view. Well-being G G g

Pain

Pleasure

Ill-being

Figure 5.3  The Feather View

Unsurprisingly, there are still other alternatives. For example, some people might prefer to relax the assumption that all lines pass through the origin. Instead of insisting that possessing a good without taking pleasure in it provides no boost whatsoever to one’s well-­being, some may prefer to claim that there can be some slight boost after all, though not nearly as great as the boost one gets when one actually takes pleasure in the good in question. Although this issue is a fascinating and important one, I leave it aside in what follows. For simplicity—if nothing more—let us limit ourselves to the versions of the everywhere differentiated and feather views that are shown in Figures 5.2 and 5.3. In any event, it should be noted how both of these views incorporate a feature we can call object sensitivity. It matters what good you are taking pleasure in, for the size of the good can make a difference to the size of your boost in well-­being. Admittedly, on the feather view this will not always make a difference; at any given level of pleasure, there will still be various distinct goods that are

What Is the Opposite of Well-Being?  121 nonetheless great enough so that their lines have yet to diverge. But other goods will be too small in size, so that—given that level of pleasure—you will get less well-­being if it is one of these lesser goods in which you are taking pleasure. So even the feather view does display object sensitivity. And for the everywhere differentiated view, the object sensitivity is even more pronounced: goods of different sizes have completely distinct lines (except at the origin). I should emphasize the point that what these lines represent—or more precisely, what a particular location on one of these lines represents—is the particular boost to well-­being constituted by the fact that you take the given amount of pleasure in the relevant good. This doesn’t yet tell us how we should aggregate these various constituents of well-­being so as to determine someone’s overall level of well-­being at a given time, or for a life as a whole. That too is an important topic, but not our concern here. Finally, bear in mind in what follows that well-­being as enjoying the good is indeed a theory of well-­being. The claims being put forward are about the very constituents of welfare, not the mere means to it. Assuming you think that welfare has intrinsic value, then these are claims about intrinsic value, not mere instrumental value. This will have to do by way of introduction to the particular nonstandard theory of well-­being that I want to examine. Our goal, recall, is to see how best to extend this theory to cover not just the positive aspect of well-­being, but ill-­being as well. For it will be noted that the view as we have described it so far does indeed only concern well-being, strictly so called. We have said that well-­being—the positive aspect of well-­being—consists in the taking of pleasure in the possession of various objective goods. So far, however, we’ve said nothing about the negative case. If pleasure in the good is what it takes to have well-­being, what does it take to have ill-­being? That is our question. And we are ready, at long last, to turn to it.

3.  An Unacceptable Answer I want to approach the question by considering an answer that I think clearly won’t do. According to the theory we are examining, well-­being consists in taking pleasure in the good. There are, clearly enough, two main ingredients in this account: pleasure, on the one hand, and objective goods, on the other. It takes both to have well-­being. And that means, as well, that there are two main ways things can go wrong. We might have pain instead of pleasure, and we might have objective bads instead of goods. In effect, then, there are two variables that we can vary. We can have pleasure or pain, and we can have goods or bads. That makes for four broad types of cases. In the case we have been examining, of course, we take pleasure in the good. Here

122  Shelly Kagan everything comes together as it should, and we have well-­being. But that leaves three other cases, which we’ve yet to examine. It could be that there are indeed objective goods in our life, but instead of taking pleasure in them, they bring us pain. Or it could be that instead of goods there are objective bads. In one such case we take pleasure in the bad. And in the remaining case we are, instead, pained by the bad. So that makes three ways in which things can fail to come together as they should, three ways in which we might fall short—one way or another—of the ideal case. That much seems clear. But what, exactly, should we say about these three nonideal cases? Suppose we start by considering a simplified version of the positive part of our theory. For the time being, let’s examine the everywhere differentiated view, and let us suppose that the relevant lines should be drawn straight, rather than curved. Two representative lines are shown in Figure 5.4 (where, again, the capital G marks the line for the greater good). Well-being

G

g

Pain

Pleasure

Ill-being

Figure 5.4  Two Representative Lines

One natural proposal is to extend these lines beyond the origin and down into the lower left quadrant. (See Figure 5.5.) That would answer the question of what to say about cases in which one is pained by the various goods in one’s life, rather than taking pleasure in them. It represents the attractive and plausible thought that if one is pained by goods rather than being pleased by them then this is bad for you—it constitutes ill-­being rather than well-­being. (That’s why the lines for the two goods cross below the x axis when we move from pleasure to pain: points below the x axis represent ill-­being rather than well-­being.)

What Is the Opposite of Well-Being?  123 Well-being

G

g

Pain

Pleasure

g

G

Ill-being

Figure 5.5  Transvaluation with Pain

In effect, then, on the current proposal, if we hold the presence of an objective good constant, but toggle from pleasure to pain, this reverses the value that would otherwise obtain. When one appropriately takes pleasure in a good, that constitutes a bit of well-­being; but if, instead, one inappropriately takes pain in the good the value is reversed, and we have an instance of ill-­being rather than well-­being. That’s the suggestion, and as I say, this does seem to be a plausible position to take with regard to the first of our three nonideal cases. Turning now to another one of the nonideal cases, what should we say about the situation in which one takes pleasure, but takes pleasure in the bad, rather than in the good? Here too we can come up with a natural proposal by thinking about how to extend Figure 5.4; this time, however, we’ll extend the figure not by continuing the lines, but rather by drawing some additional ones. As I have previously noted, it is plausible to think that the line for a greater good will be higher than the line for a smaller good; that’s why the line for G is steeper than the line for g. Expressing the same point the other way around, as we move from greater to lesser goods we should expect lines with gentler and gentler slopes, with extremely small goods having lines very close to the x axis. Perhaps then when we finally reach the case of taking pleasure in something which is actually objectively neutral (rather than good), the line will actually coincide with (the right half of) the x axis. And then—continuing this thought—as we move to cases where one takes pleasure in bads rather than goods, the lines will continue their clockwise rotation, moving into the lower right quadrant. At first, presumably, with extremely mild bads, the resulting lines would move only minimally beyond and below the x axis; but as we consider greater bads the lines

124  Shelly Kagan would grow progressively steeper. The resulting picture is shown in Figure  5.6. (Hereafter, N represents objects which are objectively neutral, neither good nor bad; and B and b represent objective bads, with B a larger bad than b.) Well-being

G

g

Pain

Pleasure

N

b

Ill-being

B

Figure 5.6  Transvaluation with Bads

Here too we have a kind of reversal from the positive case. Well-­being, we are supposing, consists in taking pleasure in the good. Accordingly, if, inappropriately, it is instead the bad that one takes pleasure in, rather than the good, the situ­ation is reversed: we have ill-­being rather than well-­being. This too seems a plausible position to take. That leaves one more case. We’ve now considered being pained by the good, and taking pleasure in the bad. What then should we say about being pained by the bad? Here too, a natural proposal suggests itself. As we have seen, if we hold the presence of objective good constant, but toggle from pleasure to pain, we get a kind of reversal of value. And similarly, if we hold the presence of pleasure constant, but toggle from goods to bads, we get a reversal of value as well. In our remaining case, however, we move from pleasure in the good to pain in the bad: we are toggling both variables simultaneously. So it seems natural to suggest that we will get two reversals of value—from well-­being to ill-­being and back again, as it were—with the two reversals canceling each other out. Which is to say, our final case will be an instance of well-­being, rather than ill-­being. The resulting picture is shown in Figure 5.7 (which incorporates our earlier results as well).

What Is the Opposite of Well-Being?  125 Well-being B

G

b g

N

N Pain

Pleasure

g b

G Ill-being

B

Figure 5.7  Double Transvaluation

Figure 5.7 undeniably has its appeal. Just as Figure 5.5 extended the lines (initially drawn in Figure 5.4) in the obvious way from the upper right quadrant to the lower left quadrant, Figure 5.7 extends the lines since added to the lower right quadrant (initially, in Figure 5.6) in the obvious way into the upper left quadrant. The resulting diagram certainly has a pleasing symmetry. But it can’t possibly be correct! Think about what it means to have lines like the ones drawn in the upper left quadrant. They represent the claim that if someone has some objective bad in their life, and they are pained by this fact, then in and of itself that directly constitutes an instance of well-­being. And that simply isn’t credible. For example, imagine someone who has just lost a child—certainly a plausible candidate for an objective bad—and imagine that, understandably, they are utterly miserable: they are tremendously pained by the death of their child. But according to the view shown in Figure 5.7 our reaction should be: ‘Oh my, what a lucky man! How incredibly well off they are!’ And that, of course, is an absurd (and offensive) thing to suggest. So the view we’ve arrived at in Figure 5.7 can’t possibly be right. Something has gone wrong somewhere. But what?

4.  Two Hedonistic Constraints One obvious thought at this point is that the reason the position just sketched with regard to pain in the bad is utterly unacceptable is that it says that in the right circumstances pain can be good for you, and great enough pain can be very good for you. That seems very hard to believe.

126  Shelly Kagan Believe me, I’ve tried. I’ve considered the possibility that my intuitions about the case of the grieving parent might be mistaken. After all, giving up on intuitions like this, and embracing the idea that pain in the bad directly contributes to one’s well-­being, would allow for a simple and elegant theory. But I just can’t do it. I cannot take seriously the suggestion that being in that sort of misery can be good for you. It simply doesn’t seem credible. (It may be worth recalling in this connection that the issue here concerns what constitutes well-­being per se—not what might or might not turn out to be merely instrumentally valuable as a means to achieving greater well-­being in the long run.) On the contrary, it seems perfectly clear that pain cannot be intrinsically good for you. It may be appropriate, in certain circumstances, to feel pain. But it cannot be that your being in pain constitutes being well off—no matter what the circumstances. So here we have a plausible constraint on the nature of well-­being: being in pain (no matter what the object of the pain) cannot constitute well-­being. We can call this the hedonistic constraint with regard to pain. I believe that we should accept it. And if we do, that will rule out the lines in the upper left quadrant of Figure 5.7. (Even if being in pain cannot constitute well-­being, might it nonetheless sometimes be neutral—neither good nor bad for you? That seems implausible as well; accordingly, I will understand the hedonistic constraint as also ruling out this more modest possibility.) Now there is, in fact, a second possible hedonistic constraint that many people will be attracted to as well. This one concerns pleasure. For at least some people will also be troubled by the suggestion (displayed in both Figures 5.6 and 5.7) that pleasure can ever be bad for you. On the contrary, it might be suggested, just as being in pain cannot constitute well-­being, similarly, if one is taking pleasure in something then this cannot constitute ill-­being (no matter what the object of one’s pleasure). We can call this second constraint the hedonistic constraint with regard to pleasure. If we accept it then that will rule out the lines in the lower right quadrant (in Figures 5.6 and 5.7) as well. As it happens, I am not inclined to accept this second hedonistic constraint. I find it considerably less intuitively compelling than the first. It seems to me, on the contrary, that in certain circumstances taking pleasure is indeed bad for you. It just doesn’t seem correct to me to suggest that if the grieving parent were instead pleased by the death of their child then this would be good for them—that taking pleasure in the loss of a loved one could constitute an instance of well-­being. Nor does it seem right to suggest that such pleasure might make you neither better off nor worse off (a different possibility left open by the second hedonistic constraint— though we can hereafter interpret that constraint so as to rule this out as well). No, it seems clear to me that if your child has died then even if you are pleased by this fact—indeed, especially if you are pleased by this fact—this constitutes an instance of ill-being. So the second hedonistic constraint seems to me to be mistaken. There is, to be sure, a popular saying that takes the other side of this issue: ‘When life gives you lemons,’ the saying goes, ‘make lemonade.’ The second

What Is the Opposite of Well-Being?  127 hedonistic constraint is arguably a reflection of this sort of view: even if your life contains objective bads (it seems to be saying) a positive attitude can turn the situ­ation around, leaving you well off. Seen from this perspective, I guess, I am an anti-­lemonade theorist. (Recall, once again, that we are here concerned solely with claims about intrinsic value, not instrumental value.) It does not seem to me to be correct to insist that taking pleasure in something cannot constitute ill-­being. So the hedonistic constraint with regard to pleasure seems to me mistaken. This is not to say that we should reject the first constraint, however. Despite my belief that in certain circumstances pleasure can indeed be bad for you, I do still find myself confident that pain cannot be intrinsically good for you. So although I reject the second hedonistic constraint, I still want to embrace the first. Nonetheless, I readily recognize that some people will want to insist upon both constraints: pain cannot be good for you, and pleasure cannot be bad. So let’s explore this possibility a bit further. Let’s ask ourselves what our theory of well-­ being might look like if we took both hedonistic constraints to heart.

5.  Working Within the Two Constraints As we have already noted, the first hedonistic constraint rules out lines in the upper left quadrant, while the second rules out lines in the lower right. So if we accept both hedonistic constraints we are limited to lines in the two remaining quadrants: the upper right for cases involving pleasure, and the lower left for cases involving pain. The simplest view of this sort is classical hedonism. It is shown in Figure 5.8. Well-being

Pain

Pleasure

Ill-being

Figure 5.8  Classical Hedonism

128  Shelly Kagan According to this view, all pains are bad for you, all pleasures good for you. So both constraints are observed. You may of course have objective goods or bads in your life, but these have no genuine significance in this theory of well-­being. It matters not at all whether the pleasure that one has is taken in objective goods or in objective bads; as long as it is pleasure one takes, the result—well-­being—is the same. And indeed, all that matters is the amount of pleasure you take. (Trivially, though, this still entails the claim that possession of objective goods produces no well-­being in the absence of taking pleasure in those goods.) And similarly, of course—mutatis mutandis—for pain: it doesn’t matter whether one is pained by something bad or something good; so long as one is pained, the result—ill-­ being—is the same. All that matters is the amount of pain. Note, then, that although this simple hedonistic position incorporates both hedonistic constraints, it does this in a way that abandons the feature I earlier called object sensitivity. For as I have just pointed out, on this approach all that matters is the amount of pleasure or pain one has; it makes no difference at all what the object of one’s pleasure or pain might be. So it is important to realize that we could adopt instead a more complicated view, one which still accommodates both hedonistic constraints while nonetheless retaining object sensitivity. We simply need to multiply the number of lines drawn in the two permissible quadrants, with distinct lines for various goods and bads. The result is shown in Figure 5.9. Well-being

G N B

Pain

Pleasure

B N G Ill-being

Figure 5.9  The Standard View

In this diagram I have only drawn three representative lines, one for an objective good, one for an objective bad, and one for something objectively neutral. It would be a trivial matter to include more lines. The crucial point to grasp is that

What Is the Opposite of Well-Being?  129 by differentiating the lines in this way, we reintroduce object sensitivity. It makes a difference, for example, not just how much pleasure one has, but what it is that one takes pleasure in. For any given amount of pleasure there will be a boost to one’s well-­being (that’s the implication of accepting the second hedonistic ­constraint), but the size of that boost will depend on the goodness (or badness) of the object of one’s pleasure. The greater the size of the objective good, the greater the boost to one’s well-­being (holding the amount of pleasure c­onstant). Contrariwise, the greater the bad in which one takes pleasure, the smaller the boost. To be sure, given that we are working with the second hedonistic constraint, it follows that no matter how bad the object of one’s pleasure, it will remain the case, nonetheless, that as long as one does indeed take pleasure (rather than pain) in the bad, then one’s well-­being does still go up somewhat. No pleasure can be bad for you, no matter how bad the object of one’s pleasure. But for all that, greater and greater bads will yield smaller and smaller boosts to your well-­being. So it is better for you if you have something less bad rather than something more bad in which to take a given amount of pleasure, and better still if you have something neutral; and better still if there is something good in your life, rather than something bad or merely neutral. That, at any rate, is what we could have in the upper right quadrant. But what about the lower left one? It will be observed that in drawing Figure 5.9 I have once again extended the various lines in the upper right quadrant in the obvious way, pushing them through the origin and down into the lower left quadrant. And the result, obviously enough, accommodates the first hedonistic constraint, the constraint with regard to pain. Since all the lines in the left half of the graph are in the lower left half, pain is never good for you (nor merely neutral): all pains—no matter what their object—constitute instances of ill-­being. So the position shown in Figure 5.9 accommodates both hedonistic constraints, while still incorporating object sensitivity: pain is always bad for you and pleasure is always good for you, but it makes a difference to your well-­being what the object of your pleasure or pain might be. There is, nonetheless, a feature of Figure 5.9 that might be questioned. Consider the ordering of the lines in the lower left quadrant. The line for the representative bad is ‘higher’ (closer to the x axis) than the line for a neutral object, which in turn is higher than the line for something good. This means that for a given fixed level of pain, it would be less bad for you to be pained by something bad than it would be for you to be pained by something neutral, which in turn would be less bad for you than it would be to be pained by something good. Is this the right position to take here? I find myself uncertain; I have no clear, direct intuitions about the relevant cases. But I am somewhat moved by the thought that while it is appropriate to take pleasure in the good, it is inappropriate to be pained by it. Rather, it is the bad that one ought to be pained by. The ordering of the lines in the lower left quadrant of Figure 5.9 reflects this idea. After all, to be

130  Shelly Kagan pained by the bad is virtuous; but to be pained by the good is vicious. So insofar as one finds it plausible to think that virtue may be an ingredient in well-­being, one might find the ordering of the lines in the lower left quadrant a philosophically attractive one.5 To be sure, given the hedonistic constraint about pain, all pains must be instances of ill-­being; but we can still insist that, other things being equal, a vicious instance of pain will constitute a larger bit of ill-­being than a virtuous one. So the ordering of the lines in the lower left quadrant of Figure 5.9 is certainly a plausible one, and I imagine that many people will be attracted to it. Let’s call this the standard ordering. It has an undeniable appeal, but as I say, I think it can be questioned. Consider the alternative ordering, shown in Figure 5.10, which I call the mirror ordering. Well-being

G N B

Pain

Pleasure

G N B

Ill-being

Figure 5.10  The Mirror View

As the name suggests, under the mirror view the order of the lines in the lower left quadrant is the reverse of the one given by the standard view: for a given amount of pain, the drop in one’s well-­being constituted by this pain will be smaller if the object of one’s pain is good rather than bad. (To capture this idea, the lines bend as they pass through the origin; the sole exception is the line for a neutral object.) It is important to recognize that the view shown in Figure 5.10 still accommodates both of the hedonistic constraints, as well as displaying object sensitivity. So 5 See Thomas Hurka, Virtue, Vice, and Value (Oxford: Oxford University Press, 2001) for the account of virtue at work here. See also my ‘Well-­Being as Enjoying the Good’, 261–3, for the connection between virtue and well-­being.

What Is the Opposite of Well-Being?  131 the choice between the standard ordering and the mirror ordering is not yet settled by those earlier commitments; we must make it on independent grounds. And there is, I think, at least some philosophical motivation that might be offered on behalf of this new view as well. The relevant thought is simple enough: pain is never good for you; but it is still better—other things being equal—to have something objectively good in your life rather than something objectively bad. So even if one is in pain, the amount of ill-­being will be smaller if one possesses an objective good rather than an objective bad. That, at any rate, is the idea behind the ­mirror view. As I say, both of these views—the standard view and the mirror view—seem to me to have a certain appeal, and I don’t yet find myself able to choose confidently between them. But let me note an analogy, if nothing more, that may be helpful in thinking about this question. As we have just seen, the standard ordering can be thought of as giving a certain kind of weight to the role of virtue in well-­being, while the mirror ordering, in contrast, gives that weight instead to the presence of objective goods and bads. Both of these ideas are appealing, but unfortunately, as we now recognize, they favor incompatible orderings. However, instead of simply choosing between them—embracing one while rejecting the other—perhaps we can make progress here by asking which of these two competing ideas should be given greater weight. An analogous question arises within virtue theory. Suppose we accept the view that virtue is a matter of loving the good and hating the bad. And suppose we grant as well that such virtuous attitudes are themselves intrinsically valuable. We might then wonder whether the value of virtue is greater than, or less than, the (absolute) value of the underlying object of the given virtuous attitude. For example, if I have a positive attitude towards a first-­order good, is the value of that attitude greater than or less than the value of the first-­order good itself? Similarly, if I have a negative attitude toward a first-­order bad, is the positive value of the virtuous attitude greater or smaller in magnitude than the negative value of the first-­order bad? There have been those who think that the value of virtue is greater (perhaps even infinitely greater) than the value (or disvalue) of its object. But I find the opposite conclusion more plausible. If one feels compassion at the suffering of others, for example, that compassion may be intrinsically good; but it is not valuable enough to outweigh the disvalue of the suffering itself! Ordinarily, it seems, the value of a virtuous attitude is actually less than the value (or disvalue) of its object.6 Similarly, then, perhaps we should conclude that something analogous holds when it comes to well-­being. Perhaps the significance of actually having objective

6  See the discussion of this issue in ch. 5 of Hurka, Virtue, Vice, and Value.

132  Shelly Kagan goods in one’s life (and avoiding objective bads) is greater than the significance of having the appropriate attitudes toward those goods (and bads). If so, then the mirror ordering is to be preferred to the standard one. The analogy is suggestive, nothing more. I certainly don’t think that the judgment one makes with regard to virtue and its objects logically commits one to making the corresponding judgment with regard to well-­being. Still, it is sug­gest­ ive, and it provides at least some reason to conclude that the mirror view—as shown in Figure 5.10—is to be preferred. But for all that, I don’t think these considerations are decisive (though the analogy merits further investigation). So for the time being, at least, I think we need to keep both options on the table. One might wonder, in passing, why a similar choice wasn’t forced upon us with regard to the lines in the upper right quadrant. Why don’t we face a corresponding choice between the ‘standard’ ordering of those lines (as shown in Figures  5.9 and 5.10) and a mirror ordering, one where the line for pleasure in the bad would be higher than (rather than lower than) the line for pleasure in the neutral or pleasure in the good? The short answer, I suppose, is this: while such a view is logically coherent, it simply has no intuitive appeal. Assuming that one is sympathetic to object sensitivity in the first place, it seems far more intuitive to insist that one’s pleasure in an object constitutes a greater instance of well-­being when the object is an objectively good one rather than when it is an objectively bad one. What’s more, the fact of the matter is that the two philosophical motivations we’ve distinguished with regard to the ordering of the lines in the lower left quadrant actually converge, rather than diverging, when it comes to the upper right quadrant. On the one hand, if virtue consists in loving the good, then it is appropriate to take pleasure in the good rather than in the bad, so this consideration favors the standard ordering in the upper right quadrant. And on the other hand, insofar as it is simply better, other things being equal, to have something objectively good in your life rather than something objectively bad, then this consideration too favors the standard ordering in the upper right. In short, as far as the upper right quadrant is concerned, the two motivations coincide in their support for the standard ordering: so there is no need to choose between them. It is only when we move to the lower left quadrant that the two motivations support distinct orderings, with the one favoring the standard ordering and the other favoring the mirror ordering. So it is only with regard to the lower left quadrant that we face a genuine choice. Accordingly, this issue—standard versus mirror ordering in the lower left quadrant—is one that we will eventually need to resolve. But for the time being, as I say, we can leave both options open. Of course, that’s not to say that we should in fact accept either Figure 5.9 or Figure 5.10 as drawn. On the contrary, for reasons already noted I think it would

What Is the Opposite of Well-Being?  133 be more plausible to draw the relevant lines curved rather than straight. So I think that even if one prefers the standard view, instead of accepting Figure 5.9 it might be preferable to accept something more like Figure 5.11. Well-being G N B Pain

Pleasure

B N

G

Ill-being

Figure 5.11  The Standard View with Curved Lines

And if, instead, one prefers the mirror view, then instead of accepting Figure 5.10 one should accept in its place something closer to Figure 5.12. Well-being G N B Pain

Pleasure

G N

B

Ill-being

Figure 5.12  The Mirror View with Curved Lines

134  Shelly Kagan Once again, however, we should probably hesitate before moving too quickly from assumptions about lines in the upper right quadrant to conclusions about lines in the lower left. I do find it plausible to think that the lines in the lower left quadrant should be curved, and because of this I’ve drawn them that way in both Figures 5.11 and 5.12. Notice, however, that while the precise ordering of the lines differs in the two figures (with the standard view shown in the first, and the mirror view in the second), nonetheless in both cases the curves are similar, in that the lines bend down as we move away from the origin. Is that the right way to go? Or might it be instead that lines in the lower left quadrant should curve up as we move from the origin to the left? This alternative is shown in Figure  5.13. (To avoid needless duplication I’ve avoided labeling the lines; as far as I can see, questions about curvature are independent of questions about the appropriate ordering.) Well-being

Pain

Pleasure

Ill-being

Figure 5.13  Implausible Curved Lines

However, while it must be conceded that the view shown in Figure  5.13 is logic­al­ly coherent, I believe that reflection reveals it to be implausible (regardless of the preferred ordering of the lines). For the implication of having the lines curve up (rather than down) is the following: for any given good (or bad), as we move to greater and greater amounts of pain, although it remains true that one’s level of ill-­being is always increasing, nonetheless it is increasing by an ever decreasing amount. Indeed, depending on the precise degree of curvature, it could turn out that at high enough levels of pain, huge further increases in pain make for vanishingly small additional increases in ill-­being. None of that, it seems to me, is remotely plausible. On the contrary, I imagine that anyone attracted to the

What Is the Opposite of Well-Being?  135 hedonistic constraint with regard to pain will be unwilling to allow that greater and greater increases in pain might generate smaller and smaller increases in ill-­ being. In any event, I personally find the suggestion unacceptable. Accordingly, I will assume that if the lines in the lower left are to be curved, they should bend down, not up. But this means, of course, that I am treating pleasure and pain asymmetrically. I am insisting that greater and greater pain makes for an increasingly larger difference in ill-­being (or at least, at a minimum, it does not make for an ever smaller difference), but I am at the same time rejecting the analogous position with regard to pleasure: greater and greater pleasure does not, I believe, make for an increasingly larger improvement in well-­being. Who knows? Perhaps someone truly wed to both hedonistic constraints would find it more plausible to treat pleasure and pain fully symmetrically. Perhaps my own inclination to embrace diminishing returns on increasing pleasure while rejecting the corresponding position with regard to pain is nothing more than a reflection of the fact—already confessed—that I do not find the second hedonistic constraint—the hedonistic constraint with regard to pleasure—nearly as compelling as I find the hedonistic constraint with regard to pain. But be that all as it may, the fact remains that I find the curvature in Figure 5.13 implausible, while it does seem to me plausible to suggest that increasing pleasure generates diminishing increases in well-­being. So I conclude—albeit tentatively— that if one were to insist upon working within the confines set by the two hedonistic constraints, the most plausible resulting views would be similar, at least in broad outline, to these displayed in Figures 5.11 and 5.12. (Arguably, of course, the lines in the upper right quadrant should depart even further from those in the lower left. Perhaps, for example, they should be drawn as asymptotically approaching—or even reaching—upper bounds.7 In contrast, I  think, nothing like this is remotely plausible for the lines in the lower left quadrant.)

6.  Without the Second Constraint We’ve been exploring the nature of well-­being and ill-­being given the assumption that one accepts both the first and the second hedonistic constraints. But as I have already remarked, I find the second constraint considerably less compelling than the first; indeed, I am strongly inclined to reject it. Although it does seem clear to me that pain cannot be good for you—whatever its object—the corresponding 7  It seems especially important to insist upon this for those lines in the upper right quadrant that involve objective bads. For it strikes me as extremely implausible to claim that no matter how great the objective bads in your life, as long as you were sufficiently pleased by this there would be no limit to how well off this might make you. That’s taking the ‘lemonade’ approach to absurd extremes.

136  Shelly Kagan claim about pleasure seems mistaken. In certain circumstances taking pleasure in something can indeed be bad for you. Or so it seems to me. Recall the example of the parent who has lost a child and yet takes pleasure in this fact. It seems clear to me that this does not constitute an instance of well-­ being. Given the presence of this tremendous objective bad in your life, being pleased isn’t good for you, it’s bad for you. So the hedonistic constraint with regard to pleasure should be rejected. And I find myself wanting to say something similar (though perhaps more controversially) with regard to a sadist who takes pleasure in the suffering of others. One who accepts the second hedonistic constraint must say that this pleasure constitutes an instance of well-­being—despite its object. But that seems to me the wrong thing to say. On the contrary, I find myself wanting to insist that the sadist’s pleasure is bad for him, an instance of ill-­being rather than well-­ being. I recognize, of course, that by hypothesis the sadist is enjoying himself; but I want to claim, nonetheless, that taking pleasure in the suffering of others is not, in fact, good for him at all, it’s bad. (To forestall misunderstanding, let me explicitly note that I am not merely claiming that such pleasure is immoral, or that it is bad from the moral point of view, or that it makes the outcome worse. I am indeed making a claim about well-­being: I am not merely angry at the sadist; I pity him, in this regard.) Suppose, then, that we agree to jettison the second hedonistic constraint. Let’s allow for the possibility that taking pleasure in an objective bad will be an instance of ill-­being rather than well-­being. To do this is to allow lines that occupy the lower right quadrant. That much, at least, is clear. What is less clear is what the lines should look like.

Well-being

Pain

Pleasure

B Ill-being

Figure 5.14  Pleasure in an Objective Bad as an Instance of Ill-­being

What Is the Opposite of Well-Being?  137 In Figure 5.14, there is a single point, located in the lower right quadrant. We can suppose that it marks the amount of ill-­being constituted in some case by the fact that you take a given amount of pleasure in the possession of some objective bad. For example, this might be the ill-­being involved in taking pleasure in the loss of your child. The second hedonistic constraint would have disallowed having a point like this (a point in the lower right quadrant), but we are no longer working within that constraint: taking pleasure in the death of your child, we are supposing, is indeed bad for you. What we now need to ask, however, is this: would it then be better for you to be pained by your child’s death? Merely having abandoned the second constraint does not, in and of itself, commit us to an answer to this new question. In principle one could agree that it is bad for you to be pleased by the death of your child, while still insisting that to be pained by this fact would be even worse for you. Anyone attracted to this potential answer will therefore want to say that as we continue the line into the lower left quadrant—where one is pained by the death rather than pleased—these new points should be drawn at an even lower level along the y axis. But this seems to me the wrong position to take. It is inappropriate to be pleased by the death of your child. It would indeed be better—better for you—to be pained by such a loss, rather than pleased. So as we continue the line into the lower left quadrant the new points should instead move up the y axis, rather than down. That’s the view shown in Figure 5.15. Well-being

Pain

Pleasure

B Ill-being

Figure 5.15  Pain in an Objective Bad as Better than Pleasure

Undoubtedly, some friends of the first hedonistic constraint will find this idea unacceptable. For I am asserting that it can be better for you to be pained by a given object than to be pleased by it. Strictly, of course, this does not violate the hedonistic constraint with regard to pain. For as we have previously stated it, that constraint

138  Shelly Kagan only disallows the claim that pain can sometimes be good for you. And since all of the points in the line segment shown in Figure 5.15 are drawn below the x axis, they still affirm that the pain is indeed bad for you. So the first constraint is still satisfied. But for all that, some friends of the first constraint may want to impose a bolder restriction: not only can pain never be good for you—they might insist—it can never be better for you either (better than pleasure, that is, or better than a smaller amount of pain). Clearly, were we to accept this bolder version of the constraint we would have to disallow Figure 5.15, reverting instead to the view I rejected a moment ago. But I am myself unprepared to endorse this bolder version of the first constraint. While it does seem clear to me that being in pain can never be part of an instance of well-­being (strictly so called), it also seems clear to me that being in pain can nonetheless sometimes be better for you. It is better to be pained by the death of a loved one than to be pleased by the loss, even though in neither case is this (noncomparatively) good for you. So I propose that we retain only the more modest version of the first constraint, and allow the sort of line segment shown in Figure 5.15. Of course, the line in Figure 5.15 still needs to be completed, extended to the left and to the right so as to cover the full range of pleasures and pains. One such extension—the extension to the right—is straightforward, or at least straightforward once one accepts the idea that pleasure can be worse for you than pain. If it is inappropriate to be pleased at the death of your child, then it is plausible to suggest that as your level of pleasure grows greater and greater, the inappropriateness grows greater and greater as well, and so does the corresponding damage to your well-­being. So we can reasonably expect the line drawn in Figure 5.15 to continue going down as we move further and further to the right. But what about the second extension? Should we similarly expect the line to continue moving up indefinitely, as we go further and further to the left? That’s the position shown in Figure 5.16. Well-being

Pain

Pleasure

B Ill-being

Figure 5.16  If More Pain in an Objective Bad is Always Better

What Is the Opposite of Well-Being?  139 There is, admittedly, something initially tempting about this proposal. It is inappropriate to feel pleasure at the death of your child. It is better to be pained. Perhaps, then, the greater the pain, the better off you are? But Figure 5.16 makes it clear that if interpreted straightforwardly this idea is unacceptable. For if the line continues up and to the left without bending, then inevitably it will eventually cross the x axis and move into the upper left quadrant. And that means that if you are sufficiently pained by the objective bad in your life, then this will be an instance of well-­ being after all. In fact, if the pain is tremendous enough, this can leave you very well off indeed. Here, then, we really do have a violation of the first hedonistic constraint (even in its appropriately more modest incarnation). And that’s just not acceptable. It simply is not credible to suggest that if one is in sufficient pain, that can, in and of itself, be good for you. So the view shown in Figure 5.16 must be rejected. The line must be prevented from crossing over into the upper left quadrant as we move to the left. One way to do this, presumably, would be to allow the line to asymptotically approach the x axis without ever reaching it. That’s the view shown in Figure 5.17. (Alternatively, perhaps it could actually reach the x axis, and then go flat, still never crossing over into the upper left quadrant.)

Well-being

Pain

Pleasure

B Ill-being

Figure 5.17  Asymptotically Approaching the X Axis

This avoids the violation of the first hedonistic constraint, but I don’t think it does it in a particularly plausible way. For on reflection I think we should agree that ever increasing amounts of pain don’t actually make for ever decreasing amounts of ill-­being. After all, even if it is inappropriate to feel pleasure in the presence of some objective bad, and even if it would be better to feel some pain,

140  Shelly Kagan that doesn’t mean that it is appropriate to feel pain without limit, so that the greater the amount of pain, the better. On the contrary, for any given objective bad, I think it more plausible to suggest that there will be some amount of pain that is the appropriate amount of pain to take in that bad. Greater pain than that and you are pained more than you should be. So at levels of pain greater than the appropriate amount, the line should bend back down, with the amount of ill-­ being steadily increasing thereafter. In short, a more plausible way to draw our line is the one shown in Figure 5.18. For any given objective bad, there is an appropriate amount of pain that one should take in that bad. (In Figure 5.18 this level is indicated by an arrow.) Too little pain, and one’s ill-­being is greater than it needs to be. And if one feels pleasure in the bad (instead of pain), that is even worse. But that doesn’t mean that more and more pain leaves you better and better off. On the contrary, there is an ideal amount of pain that one should be feeling—a level fixed by the particular bad in question—and if one feels more pain than that, if one feels an inappropriately large amount of pain for the given objective bad, then one is indeed worse off than one would be, had one felt, instead, the appropriate amount of pain. Well-being

Pain

Pleasure

B Ill-being

Figure 5.18  The Appropriate Level of Pain

Notice that all of the points of the line in Figure 5.18 are below the x axis. This represents the thought that regardless of what your reaction to the objective bad might be—pleasure or pain—this cannot be an instance of well-being. So the first hedonistic constraint is satisfied. Pain is never good for you. It’s just that sometimes it is better for you than the alternatives. More precisely, given the presence

What Is the Opposite of Well-Being?  141 of the objective bad in your life, it is best to feel the appropriate amount of pain. Doing this will still constitute an instance of ill-­being, but a lesser instance of ill-­ being than you would have if you were pained too much or too little (or if you were pleased). Generalizing, we might conjecture that all objective bads should be represented by lines of this same basic sort, with each objective bad having its own appropriate level of pain. (Arguably, though, the lines should be drawn curved, rather than straight.) But this is not at all to suggest that the appropriate level of pain is always the same, regardless of the size of the bad. On the contrary, it seems natural to suggest that the greater the size of the objective bad, the greater the amount of pain that it is appropriate to feel. What’s more, it also seems plausible to suggest that the amount of ill-­being involved will depend on the size of the bad as well, so that when one undergoes the appropriate amount of pain for a given bad, the greater the bad, the greater the amount of ill-­being. Feeling the appropriate level of pain will minimize the amount of ill-­being that one must have (given the presence of the bad), but the larger the bad, the larger the inevitable damage to your well-­being. Well-being

Pain

Pleasure b B

Ill-being

B

Figure 5.19  The Right Half of the Line for Three Objective Bads

These ideas are illustrated in Figure 5.19, where I have drawn the right half of three lines, for three different bads (with B a greater bad than B, which is a greater bad than b). Since I’ve only drawn the right half of any given line, each line ter­ min­ates (on the left) at a point which represents the optimal or appropriate amount of pain for the given bad. We’ve just agreed that the greater the bad, the greater the amount of appropriate suffering, and the greater the amount of

142  Shelly Kagan ill-­being that exists when one feels that appropriate amount of pain. Accordingly, the points representing the appropriate amount of suffering move down and to the left as we move to larger and larger bads. Figure 5.19 also shows what happens to ill-­being when one feels less than the appropriate amount of pain, or when one feels pleasure instead. In keeping with our earlier suggestion that it is worse for you to feel too little pain, and worse still to feel pleasure, the lines each move downward as we go to the right. But Figure 5.19 incorporates a further plausible thought as well: the rate at which ill-­ being increases (as one inappropriately moves to the right) grows ever greater as well, as we deal with larger and larger bads. So ill-­being always increases if one is pained too little by some objective bad (or if one is pleased by it), but it increases at a faster rate with larger bads. That, at any rate, is what is happening in the right half of the lines that represent bads of various sorts. But what is happening in the left half of these same lines? To answer that question we will need to return to our original representations of well-­ being as enjoying the good, as presented in Figures 5.2 and 5.3.

7.  Well-­Being as Enjoying the Good, Extended We are now in a position to answer the question I posed at the start of this essay: if one accepts the approach to well-­being that I’ve called well-­being as enjoying the good, how is this basic account—originally introduced only in terms of positive well-­being—to be extended so as to cover ill-­being as well? In answering our question, I retain the various assumptions I have endorsed along the way. I accept the first hedonistic constraint, the constraint with regard to pain, but I reject the second. And I embrace that first constraint only in its more moderate version, which rules out the possibility that pain might be good for you, but allows nonetheless that pain might be less bad for you than indifference or pleasure. I assume, moreover, that for any given objective bad there is an appropriate amount of pain—larger for larger bads—so that if one feels more than the appropriate amount of pain then one’s ill-­being increases. But I also assume that if one feels less than the appropriate amount of pain (or if one feels pleasure instead) then this increases one’s level of ill-­being as well, and that the rate at which this happens grows larger with larger and larger bads. Finally, I assume that the larger the objective bad, the greater the inescapable damage to one’s well-­being. Given all of this, what exactly should someone who accepts well-­being as enjoying the good say about the nature of ill-­being? How is the positive account,

What Is the Opposite of Well-Being?  143 sketched in Section 2, to be extended so as to cover not just taking pleasure in the good, but the various nonideal cases as well? When I first introduced the idea behind well-­being as enjoying the good I suggested that there were at least two main variants worthy of our attention—the everywhere differentiated view (shown in Figure 5.2) and the feather view (shown in Figure 5.3). Let’s begin with the feather view. According to that view, the lines for objective goods all share an initial segment (the spine of the feather) though the line for any given good eventually peels off or breaks away at some point—with the relevant point occurring higher up and further to the right for greater goods. In light of what we have now learned about the shape and position of lines for objective bads, it is easy enough to see how the feather view is to be continued. The lines for bads can also share the common spine (on the left), but these lines will break away from the spine at progressively lower and lower points for greater and greater bads. The basic idea is shown in Figure 5.20.

G

Well-being G

g

Pain

Pleasure

N

b B Ill-being B

Figure 5.20  The Extended Feather View

A few points here deserve at least some brief comment. First of all, the right halves of the lines for bads slope down and to the right, in keeping with our thought that if one has less pain than is appropriate for the given bad (or if one inappropriately feels pleasure rather than pain) then this increases one’s level of

144  Shelly Kagan ill-­being. And the slopes grow steeper for larger bads. Second—also in keeping with our earlier conclusions—the larger the bad, the greater the amount of appropriate pain, and the greater the level of ill-­being that this involves, even when one does feel the appropriate level of pain. Third, although the lines for bads slope down as we move to the right, away from the spine, the lines for goods slope up, rather than down. This reflects the thought that while it is bad for you to take pleasure in the bad, and the more pleasure the worse, it is good for you to take pleasure in the good. And up to a certain point, at least, more and more pleasure yields greater and greater well-­ being. Of course, saying all of this is still compatible with thinking that for any given good there is a maximum amount of well-­being that can be gained from the good in question, so that even the line for an objective good may have an upper bound. Anyone attracted to such upper bounds (as I am) will want to tweak the lines in the upper right quadrant of Figure 5.20, so that they approach (or reach) those upper bounds, rather than continuing up without limit as we move to the right. Finally, I have drawn the line for neutral objects to coincide with the right half of the x axis. If we accept this idea then we are saying that positive well-­being can only be had if one takes pleasure in an objective good. In contrast, all pleasures in the bad, and even ‘harmless’ pleasures (pleasures taken in objects neither good nor bad), are incapable of generating any genuine well-­being. One might, however, feel that this is an overly ‘moralistic’ position to take. Even if we agree that it is bad for you to take pleasure in objective bads—especially if those bads are large—we might still think it going too far to insist that if one takes pleasure in something objectively neutral (like your home team winning a sports tournament) this cannot in and of itself enhance your well-­being. Surely, it might be objected, taking pleasure in something neutral can be good for you. Suppose we accept this thought. Then we would resist the proposal to have the line for neutral objects coincide with the right half of the x axis. Rather, we would place this line somewhat above the x axis—though perhaps only slightly above it—so that taking pleasure in a neutral object would indeed constitute (a small amount of) well-­being. Taking this step would have the following interesting implication. If we assume—as also seems plausible—that very small bads would break away from the common spine at points just below the breakaway point for neutral objects, then in moving the line for neutral objects above the x axis we would also be doing the same for the lines of sufficiently small bads. So not only would it be the case that taking pleasure in neutral objects would be good for you, it would also be the case that taking pleasure in certain bads could be good for you as well, provided that they were sufficiently small. (See Figure 5.21.)

What Is the Opposite of Well-Being?  145 G

Well-being

G g N b Pain

Pleasure

b

B

Ill-being

B

B

Figure 5.21  The Extended Feather View with Well-­being from Small Bads

Is this a plausible result? I think that people can reasonably disagree about this question. On the one hand, this might accommodate the thought that enjoying practical jokes (which often involve some embarrassment or discomfort on the part of the victim) can be a source of at least some small amount of well-­being. Similarly, it might allow us to say, for example, that enjoying the account of a well-­executed art heist can be a (limited) source of well-­being.8 On the other hand, some might prefer to insist that when we gain well-­being in such cases the objects of our pleasure are actually objective goods, rather than bads—since we take pleasure in the intelligence and cleverness of the joke, for example, or the ingeniousness and careful execution of the plan, in the case of the heist. And for the same reason it might be suggested in turn that even when we seem to gain well-­being from the enjoyment of a neutral object, what we really gain well-­being from is the enjoyment of the objectively good aspects of the otherwise neutral object (such as the skill of our favored team). Arguably, then, the line for genuinely neutral objects should be left at the x axis after all. But I won’t try to settle the issue here. Turn now to the task of extending the everywhere differentiated view. Here the lines have no common shared segment (though for objective goods, at least, the lines do intersect at the origin), so we can expect them to remain distinct as they continue into the lower left quadrant. But that means that we now face a question 8  I owe this example to Gwen Bradford.

146  Shelly Kagan about the ordering of these lines, identical to the question previously posed in Section 5. Should we endorse the standard ordering (with lines for greater goods further below the x axis in the lower left quadrant)? Or should we instead accept the mirror ordering (with lines for greater goods closer to the x axis)? Suppose we accept the mirror ordering. Then the extension of the everywhere differentiated view will end up looking something like the graph shown in Figure 5.22. Well-being G g

N Pain

Pleasure G g N

b

b B

Ill-being

B

Figure 5.22  The Extended Differentiated View with Mirror Ordering

The view shown in Figure 5.22 is surprisingly complicated, with the most striking feature being that the lines for objective bads look rather different from the lines for objective goods. Indeed, while the lines for goods all cross at the origin (a feature already in place in Figure 5.2) the lines for bads ‘float free’ of those other lines, never intersecting them (or one another) at all. But it is easy to see how these new lines—the lines for bads—are a natural extension of those already given by the initial everywhere differentiated account, once that account is combined with our further conclusions about the nature of ill-­being. Note, for example, how for each objective bad there is an appropriate amount of pain that one should feel, so that if one feels less pain than this (or if one feels pleasure instead) then this results in a greater amount of ill-­being. And note as well how the appropriate amount of pain is greater for larger bads, and how the smallest increase in ill-­ being it is possible to achieve for any given bad grows greater with larger bads. Furthermore, it is easy to see how the mirror ordering is maintained in the lower left quadrant, with the lines for bads being located further below the x axis than

What Is the Opposite of Well-Being?  147 the lines for goods. So despite its unexpected complexity, the view shown in Figure 5.22 seems to me a reasonable suggestion concerning the extension of the everywhere differentiated view. As always, of course, there remain details to work out. For example, it might be that we would want the various lines for goods in the upper right quadrant to approach or reach upper bounds. And here, too, some might want to move the right half of the line for neutral objects up a bit above the x axis, so that taking pleasure in something neutral—or even something only minimally bad—can actually constitute a slight increase in well-­being. These further issues remain important; but I won’t try to resolve them here. Finally, what if one accepts the everywhere differentiated view, but wants to combine this with the standard ordering of lines in the lower left quadrant? The resulting picture grows more complicated still, and is shown in Figure 5.23. Well-being G g

N

Pain

B

Pleasure

b

b N

g

G Ill-being

B

Figure 5.23  The Extended Differentiated View with Standard Ordering

At first glance the position shown here may seem to be nothing more than a bewildering array of intersecting and crisscrossing lines. Indeed, when I first drew it I found myself repeatedly thinking that it must contain some hidden contra­dic­tion, or must somehow violate some obvious claim about the nature of well-­being. But a more careful examination of the view quiets these concerns. Here too, after all, it remains the case that for each objective bad there is an appropriate amount of pain that one should feel, so that if one feels too little pain (or if one inappropriately feels pleasure instead) this creates a greater amount of ill-­ being. And here too, the greater the size of the bad, the greater the amount of pain that it is appropriate to feel, and the larger the amount of ill-­being produced when

148  Shelly Kagan one does feel the appropriate amount of pain. It is simply that on the current view the lines in the lower left quadrant reflect the standard ordering, where it is normally better to be pained by something bad rather than by something good. Admittedly, this ordering is only rigorously maintained by the left halves of the lines for objective bads. The right halves move down and to the right, and this produces the crisscrossing pattern that can seem so disconcerting. But for all that, it is easy to see that the sequence of lines—as we move from great bads to smaller ones, to neutral objects, to small goods, and then on to greater goods—changes smoothly and continuously as we move from one line to the next. As we move from bads to goods, the left halves of the lines grow ever steeper, and the right halves first grow closer to the x axis from below and then climb steadily above it. Here too, of course, there is room for quibbling about the details. Perhaps the right half of the line for neutral objects should be placed above the x axis, rather than coinciding with it. And perhaps lines in the upper right quadrant should be drawn so as to involve appropriate upper bounds. But as I have done throughout this essay, I leave such issues unresolved. I do, however, want to briefly consider a possible variant of Figure  5.23. As I have just pointed out, although the various lines do crisscross one another, the sequence of lines can be seen as evolving, from one line to the next, in a relatively constant and regular pattern. But we could make the pattern more regular still, if we relaxed the assumption that the lines for goods must all pass through the origin (after all, the lines for bads do not), and even more so if we also introduced a kink or bend into the lines for objective goods, similar to the bends in the lines for objective bads. The resulting picture is shown in Figure 5.24.

Well-being

G g N Pain

Pleasure

B

b

b N

g

G Ill-being

B

Figure 5.24  If Lines for Goods Don’t Go Through the Origin

What Is the Opposite of Well-Being?  149 Here the transformation of any given line—as we move from bads to goods—is very regular indeed. As we move from large bads to smaller ones, from bads to goods, or from small goods to larger ones, the lines continue to rotate counterclockwise. What’s more, as we go from bads to goods, each line shifts further to the right as well, as can easily be seen by tracing the movement of the bends in the lines. (In contrast, in Figure 5.23, although the lines do rotate counterclockwise as we go from bads to goods, they stop shifting further to the right once we reach objective goods.) So let us ask: should a friend of the everywhere differentiated view who is also drawn to the standard ordering find this a particularly congenial expression of their position? Unfortunately, despite its undeniable appeal, I think this view is problematic. Precisely because the lines for objective goods continue to shift to the right (rather than all passing through the origin), the left halves of these lines cross the x axis and pass into the lower right quadrant. This means that if you have an objective good in your life, but fail to enjoy it sufficiently, this can be bad for you, indeed very bad indeed. And that seems to me implausible. Admittedly, if someone fails to take sufficient pleasure in a good that may constitute a shortfall in virtue.9 But I find it implausible to suggest that someone who has a good and is enjoying it can nonetheless, by virtue of that very fact, be badly off. Accordingly, I think this particular variant should be rejected. We should retain the position shown in Figure 5.23 (and for that matter, in Figure 5.22 as well) according to which if one has an objective good in one’s life, then provided that one takes some pleasure in that good there will be at least some minimal increase in one’s well-­being.

8.  Conclusion Where then does that leave us? What is the opposite of well-­being? It seems to me that we are left with five basic positions, each of which deserves to be taken seriously. Many people, I am confident, will be drawn to both of our hedonistic constraints, so they will be limited to lines that pass through the upper right and lower left quadrants; the other two quadrants will be off limits. But even those who accept both of these constraints should accept object sensitivity as well, and this introduces the need to choose between the standard and the mirror orderings. Thus the basic choices here, I believe, are two, shown in Figures 5.9 and 5.10 (or, in their more plausible variants, Figures 5.11 and 5.12). Of course, as I have repeatedly noted, I personally believe that we would do well to reject the second constraint. Sometimes, it seems to me, taking pleasure in an object can be an instance of ill-­being rather than well-­being. Accordingly, we

9  Following Hurka, Virtue, Vice, and Value, 83.

150  Shelly Kagan can allow lines to enter the lower right quadrant. But for all that, of course, I do think we are right to embrace the first constraint—the constraint with regard to pain. So the upper left quadrant should remain empty. At the same time, however, I believe that the first constraint should be accepted only in its more moderate form. While pain can never be good for us, it can, in the right circumstances, be less bad for us than indifference or pleasure. So we should expect some lines to go down, as we move from the lower left quadrant (being pained by some object) to the lower right (taking pleasure in it). Given this more tolerant set of views concerning the nature of ill-­being, more options open up for us. If we accept the feather view, we are led to the view shown in Figure 5.20, and if instead we accept the everywhere differentiated view, we are led, I think, to the still more surprising views shown in Figures 5.22 and 5.23. These last three, and in particular the last two, deserve special comment. I find them gloriously asymmetrical, unexpectedly so. If all one had at the outset of theorizing was the basic account of well-­being as enjoying the good, one would have little reason to anticipate that extending that account so as to cover ill-­being might lead to overall views as complex and distinctive as the ones shown in Figures 5.22 and 5.23 (or even Figure 5.20). Discovering that these unexpectedly complicated positions are in fact the most plausible ways of extending that initial view reinforces the thought it is always important to ask, for any given account of well-­being, just how, exactly, it should be extended so as to cover ill-­being as well. All of which is just to say: when thinking about the nature of well-­being, we must never forget to ask about its opposite.

Bibliography Hurka, Thomas, Virtue, Vice, and Value (Oxford: Oxford University Press, 2001). Kagan, Shelly, ‘An Introduction to Ill-Being’, Oxford Studies in Normative Ethics 4 (2014): 261–88. Kagan, Shelly, ‘Me and My Life’, Proceedings of the Aristotelian Society 94 (1994): 309–24. Kagan, Shelly, ‘Well-Being as Enjoying the Good’, Philosophical Perspectives 23 (2009): 253–72. Nozick, Robert, Anarchy, State, and Utopia (New York: Basic Books, 1974).

6

Parfit on Love and Partiality Roger Crisp

1.  A Move Away from Impartiality? Like his ‘master’ Henry Sidgwick, Derek Parfit is often mistakenly described as a utilitarian. Sidgwick, ultimately, had no consistent first-­order ethical position, being unable to resolve the so-­called ‘dualism of practical reason’. But it is clear which positions he saw as contenders: rational egoism, and universal benevolence (i.e. utilitarianism). Parfit’s position is less clear. In Reasons and Persons, appendix F, on the Social Discount Rate, Parfit outlines what he calls ‘the Argument from Special Relations’, which includes the claim that each of us ought to give priority to the interests of people to whom we stand in certain special relations—our children, parents, pupils, patients, those we represent, or our fellow citizens (RP, 485).1 Parfit rejects this argument for the Social Discount Rate, but he does not explicitly reject the claim about special relations. Indeed he claims that the people to whom we have greatest obligations are our own family (RP, 486). Now it may be that he sees these partial obligations as derived from some impartial obligation, such as that to maximize the good. But he gives no hint here that this is his view. Further, at RP, 29, Parfit tells us that he is ‘not yet convinced that C [act consequentialism] is the best theory’. In one of his most famous passages, Parfit says of Reductionism about personal identity: Is the truth depressing? Some may find it so. But I find it liberating, and consoling. When I believed that my existence was a . . . further fact, I seemed im­prisoned in myself. My life seemed like a glass tunnel, through which I was moving faster every year, and at the end of which there was darkness. When I changed my view, the walls of my glass tunnel disappeared. I now live in the open air. There is This paper was presented at the conference on Derek Parfit convened in Oxford in May 2018. I am very grateful to Jeff McMahan, Joe Carlsmith, and Ketan Ramakrishnan for the invitation to speak, and to the audience for extremely helpful comments and discussion. I am deeply indebted to David Brink, Tim Campbell, Brad Hooker, and Theron Pummer for penetrating comments on and discussion of previous drafts. Let me close by expressing my deepest gratitude to Derek Parfit, for kindly guidance, encouragement, discussion, friendship, and inspiration over many years. 1  Page references to RP are to Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984). Roger Crisp, Parfit on Love and Partiality In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0007

152  Roger Crisp still a difference between my life and the lives of other people. But the difference is less. Other people are closer. I am less concerned about the rest of my own life, and more concerned about the lives of others.  (RP, 281)

Parfit does not here distinguish between partial and impartial concern for others. But it is reasonable to think that if the weakening of psychological links between selves in my own life decreases partial concern for my future self, then the weakening of psychological links between myself now and other future selves will also weaken partial concern for those future selves. Further, it is plausible to think that a general move in the direction of impartiality is supported by Parfit’s claims about the direct self-­defeatingness of M (common-­sense morality) arising out of its agent-­ relativity and about the problems caused by non-­identity for person-­affecting views. Parfit’s first-­order view comes more clearly into focus in On What Matters.2 In particular, Parfit explicitly commits himself to our having other-­regarding ­reasons of partiality: We can have strong reasons to care about the well-­being of certain other people, such as our close relatives and other people whom we love. Like self-­interested reasons, these altruistic reasons are person-­relative or partial in the sense that these are reasons to be specially concerned about the well-­being of people who are in certain ways related to us. (OWM i. 40)

If Parfit did not believe in Reasons and Persons that reasons grounded in special relations were not derived from ultimately impartial reasons, statements such as this constitute a change in direction away from greater impartiality in ethics. Whether this is the case or not is a matter primarily of Parfit scholarship. What I wish primarily to suggest is that partiality itself may be one of the elements of common-­sense morality which Parfit would have been wise to follow Sidgwick in rejecting.

2.  Special Relations and Love Later in On What Matters, volume i, Parfit clarifies the kind of partiality towards others he has in mind. Chapter 3 is on subjectivism about reasons, the view that reasons depend, in some sense, on our present desires. Parfit is of course highly sceptical about subjectivism, and hence wishes to make it clear that his claims do not apply to certain complex states which involve desires: When we love someone, for example, we are motivated to act in certain ways. We care greatly about this person’s well-­being, and we want to do what would be 2  References to OWM are, by volume and page number, to Parfit, On What Matters, vols. i and ii (Oxford: Oxford University Press, 2011) and vol. iii (Oxford: Oxford University Press, 2017).

Parfit on Love and Partiality  153 best for him or her. Though our loving someone partly consists in our having such desires, we have strong reasons, I believe, to care about, and try to promote, the well-­being of those we love. Such reasons are provided, not by the desires involved in loving someone, but by the ways in which love is in itself good, and by various other facts about our relations to those we love, such as facts about shared histories, or commitments, or reasons for gratitude, or by the facts that are involved in romantic or erotic love, or love for our parents, children, or other close relatives. To illustrate this distinction, we can suppose that I meet several strangers, all of whom need my help. If I had a strong desire to help one of these strangers, perhaps because I like her face, that would at most give me only a weak reason to help this stranger rather than any of the others. Love, in its various forms, is very different from such a desire.  (OWM i. 73)

Consider the following case: Burning House. You are in front of a burning house, able to save only one of the two people trapped inside. One, L, is someone you love; the other, S, is a stranger. All else is equal.

According to standard act utilitarianism, on a straightforward understanding of the ceteris paribus clause in this example, and on the assumption that saving either person would increase utility overall to the same degree, you have no reason to give either of these individuals priority over the other. It is true that if you save S, you will be sad and bereft; but we should assume that if you save L, a person who loves S will be equally bereft. Parfit’s view is that you have a stronger, non-­self-­interested reason to care about and save L.  Let me call this a partial reason. I take it that this reason is ultimate and not derived from some ultimately impartial reason. According to Parfit, this reason is not based on any desires or motivations involved in loving L. Just before the passage quoted above, Parfit says that in his arguments against subjectivism he is using ‘desire’ in a wide sense, to cover any state of ‘being motivated’ (Parfit perhaps means here ‘actually being motivated’), or any state ‘of wanting something to happen and being to some extent disposed to make it happen, if we can’. Love, he is suggesting at the start of the passage, includes such desires, broadly construed, and he instances being motivated to help that person, caring about them (e.g. wanting that they not experience pain), and wanting to do what is best for them. His position, then, is that we have ­reasons to have these desires, and should not be misled into thinking that these desires provide their own rational grounds. For no desire provides a rational ground for anything, whether it be an action, a feeling, or itself. If these desires themselves do not provide partial reasons, what does? Parfit has said that loving someone partly consists in our having desires to help that person, and so on, and so one might expect that he will claim that our reasons to have

154  Roger Crisp these desires are constituted by the other, non-­desiderative constituents of love. One such constituent might be, for example, our actually helping the other person, as opposed to the desire to do so, but it is not clear how doing any of the actions characteristic of loving someone could be said to be self-­grounding. My reason for helping someone cannot be provided by the fact that I am helping someone, just as my reason to desire to help someone cannot be provided by the fact that I am desiring to help someone.

3.  Affection and Liking Another common constituent of love Parfit may have in mind is affection. This feeling or emotion might itself be said to be partly constituted by certain mo­tiv­ ation­al states. But presumably there are elements of it that could be distinguished from such states, and these could be claimed to provide partial reasons. Consider liking. At the very end of the passage quoted above, Parfit seeks to illustrate the distinction between loving and mere desires with an example in which I meet several strangers who need help. I like the face of one of them, and this leads me to have a strong desire to help this stranger rather than any other. This desire, Parfit suggests, could give me at most only a weak reason to help this stranger. On the most straightforward reading, Parfit is here allowing for the possibility that a desire—albeit a strong one—could give me a reason—albeit a weak one. But the context of the passage makes it hard to accept this reading. First, Parfit has already made it clear that he believes that reasons are value-­based, not desire-­based. Second, the central point of the passage quoted above is to show that there are reasons to love even if there are no desire-­based reasons. If Parfit believed there could be a weak desire-­based reason in the case in which I like someone’s face, he might claim that there is a strong desire-­based reason in a case in which I feel the kind of deep affection for someone else which is characteristic of love. Where then does the possible reason to help the stranger come from? One plausible source is my liking her, or rather her face. Could the kind of liking someone which is characteristic of love ground a reason for partiality, perhaps even a strong reason? Consider the following case:3

3  For a similar case, see that of ‘Fred Simmons’, in N.  Kolodny, ‘Love as Valuing a Relationship’, Philosophical Review 112 (2003): 144–5. Some of the arguments that follow are adapted from my 2018 Lindley Lecture, ‘Against Partiality’, https://kuscholarworks.ku.edu/bitstream/handle/1808/26747/ Crisp_2018_AgainstPartiality.pdf?sequence=3&isAllowed=y. I am grateful to the Dept. of Philosophy at the University of Kansas for the invitation to deliver this lecture, for very helpful comments and discussion, and for permitting me to rely here on some of the arguments from the lecture.

Parfit on Love and Partiality  155 Titania. Titania wakes. The first person she sees is Bottom. The second person she sees is Top. Because of the magic juice Oberon has sprinkled in her eyes, she has developed a deep affection for Bottom. Both Bottom and Top are in a burning house, and she can save only one of them. All else is equal.

Titania’s affection for Bottom is a highly contingent psychological fact about her.4 Had she seen Top first, she would have felt the same affection for him instead. Note that her new affection for Bottom may well have some practical significance for her. Imagine, for example, that she has to choose to marry either Bottom or Top. She has a straightforward self-­interested reason to marry Bottom, since that relationship is likely to be happier and more successful than a relationship with Top. The question is whether her affection for Bottom would provide an ultimate justification for her giving moral priority to Bottom over top, and it is hard to see, on reflection, why it should. Consider an analogous case, in which a judge finds herself liking one of two defendants much more than another.5 After the trial, if she has to spend a few minutes with one or other defendant, it might well be reasonable for her to choose to spend that time with the one she likes more, if we assume that neither defendant cares whether she chooses them or not. But if both defendants are found equally guilty, her greater liking for one is irrelevant to her judicial decision. Nor does it matter if the affection in question is reciprocated (imagine that Oberon had caused Bottom to feel for Titania what she feels for him). The mere fact that two people feel affection for each other is not a good reason for either of them to give the object of their affection moral priority. It might be objected that ordinary relations of affection are not like this. They are not usually brought about through manipulation, nor is their origin as independent of a person’s qualities. But ordinary relations of affection are not up to us. We just find ourselves liking or loving some people, and not others; so the role of Oberon in the story is not essential.6 And we may like or love someone just because we find ourselves liking or loving some apparently non-­evaluative feature of them: their face, for example, or the colour of their hair.

4  Contingency itself is not objectionable, of course. My promise to you may be a contingent matter. But that—if there is an ultimate duty to keep promises—makes my duty to you to fulfil that promise grounded and non-­arbitrary in a way that Titania’s alleged partial reason is not. 5  The judge could be said to have special role-­based or professional duties of impartiality. But those role-­based duties are most plausibly seen as themselves grounded on a prior duty of impartiality. The rationale for that prior duty may depend on the highly plausible view that individuals should be treated equally unless there is some reason to do otherwise. And the question then is whether Titania has such a reason. See n. 10. 6  Some might claim that what happens to Titania is especially objectionable since her will is dom­ in­ated by that of another person. A version of the story could then be told in which the juice happens to drip as sap from an overhanging branch.

156  Roger Crisp But do we not, at least in many cases, feel affection for someone because of certain good or ‘loveable’ qualities they have?7 I might love my wife, for example, because she is beautiful, kind, and witty. But in that case why should I be partial to my wife, and not to those who are beautiful, kind, and witty? (Imagine that in the Titania case, both Bottom and Top were equally well worthy of Titania’s love.) And if it is claimed that I should prioritize these groups, we can anyway ask why should they be given any moral priority, except in so far as their kindness and other beneficial qualities are the result of free and voluntary actions they have taken, in which case we are talking not about duties based on personal or social relations, but about duties based on moral desert, a notion which Parfit rejects (OWM i. 270). What lies behind Parfit’s rejection of subjectivism is in large part the idea that desires themselves are merely contingent psychological states, with no value or normative weight in themselves. Of course, they often aim at the good, or their satisfaction is pleasurable, but the evaluative and normative work here is done by the good and pleasure. It seems to me that the same goes for liking. The mere fact that one feels any form of affection for anything, whether it be another person or some other object, does not imbue that object with any special value8 or provide any ultimate reason for giving that object priority.

4.  Love as Non-­Instrumentally Good If we now return to an earlier part of the passage, we find that Parfit provides a list of potential sources of partial reasons, particularly those based on love. The first suggestion is that such reasons are provided by the ways in which love is non-­ instrumentally good. One way love might be good is through its contribution to the well-­being of the lover. In his famous appendix I to Reasons and Persons, Parfit is non-­committal on the correct theory of well-­being, though he ends the discussion by mentioning a theory according to which love is a component of well-­being if it is desired and found pleasurable by the subject (RP, 502). On this view, one’s reason to be partial could be ultimately grounded in self-­interest, in a way that it would not be were, say, hedonism to be the correct view of well-­being. According to hedonism, any self-­interested reason I have to be partial is derivative from my reason to promote my own net pleasure. One might combine with C the view that one element of 7  For a good defence of this view as applied to romantic love, see S. Keller,‘How Do I Love Thee? Let Me Count the Properties’, American Philosophical Quarterly 37 (2000): 163–73; For a helpful crit­ ic­al account of the view in general, see Kolodny, ‘Love as Valuing a Relationship’, 138–42. The role attachment to what is impersonally valuable may play in creating meaning is elucidated in J.  Raz, Value, Respect, and Attachment (Cambridge: Cambridge University Press, 2001), 19. 8  This is the case even if that object is itself impersonally valuable; cf. Raz’s view, cited in n. 7.

Parfit on Love and Partiality  157 well-­being consists, ultimately, in partiality, so that our reasons to be partial do not come into conflict with the reason to promote the good overall. But this is not Parfit’s view in On What Matters, where he accepts that there can be self-­interested reasons to favour one’s own good over the overall good (OWM i. 131). I myself can understand why Titania’s life might be improved by her new love for Bottom, if for example she enjoys thinking about him and spending time with him more than she would have enjoyed whatever she would have been doing without the application of the magic juice. But I find it harder to see why the mere fact of her loving Bottom, or being loved by him, makes her own life better for her, independently of other potential goods such as pleasure or understanding. And I would expect an anti-­subjectivist such as Parfit to agree: his inclusion of love as a potential element of well-­being is in apparent tension with his generally objectivist stance on value. Perhaps love or partial concern is, though not a component of well-­being, an impersonal good or value, the instantiation of which in a world makes an independent contribution to the value of that world. One way this could be the case would be if (a) there are ultimate reasons of partiality and (b) a world is improved by its inhabitants acting on the reasons they have. (a) here is an assertion which the impartialist can of course deny. Perhaps, however, there is something special about partiality of concern itself which can be seen if we compare worlds of greater and lesser actual partiality. Consider, then, the following three worlds: Non-­Concern. All human beings have evolved to be entirely self-­interested, showing no concern for others. They all lead lives worth living. Impartial Concern. All human beings have evolved to be impartially concerned for others. They all lead lives worth living. Partial Concern. All human beings have evolved to be partially concerned for others. They all lead lives worth living.

Assume that all else is equal in these worlds. For example, we should not imagine that Non-­Concern involves more vicious behaviour than the other worlds; perhaps its inhabitants have realized the Hobbesian advantages of morality. If you believe that partial personal relationships themselves contribute to well-­being, assume that Non-­Concern and Impartial Concern include certain other goods to compensate for their lack of such relationships. Even if that is the case, however, some will believe that, though the welfare levels in Non-­Concern and Partial Concern are the same, there is something impersonally bad about Non-­Concern, just as there might be if a world contained well-­being gained from certain sadistic or cruel activities. But even if we accept this, it is hard to see why Impartial Concern should be worse than Partial Concern

158  Roger Crisp in this respect. In Impartial Concern, each individual cares about others, and, we can assume, has relationships with them, but this care is not distributed unequally. In Partial Concern, this care is distributed unequally, and it is hard to see why this inequality should be a good-­making feature. If anything, it might seem impersonally worse than equality of concern. It might be claimed that the relationships in Impartial Concern must be shallow; each person would have many trivial acquaintances with others, but no deep friendships. But this need not be the case. In Impartial Concern, people might well find themselves liking certain people more than others, spending more time with them and getting to know them better. But that liking will not lead them to give priority to the interests of those they like.

5.  Shared Histories Parfit goes on to cite as sources of partial reasons several other facts about our relations to those we love. He first mentions shared histories, again repeating the phrase ‘facts about’. I shall take it that he means that the fact that two people who love one another have a shared history itself provides a reason for them to give each other priority over others. This is one fact about our relations to those we love, or at least many of them, but there will of course be other facts about shared histories. Since Parfit does not give any clues as to what other facts he might have had in mind, it is not possible to enquire further into them at this point. (This point applies also to the other two items he goes on to list in this clause: commitments and reasons for gratitude.) Consider first the following case: Blue and Green. Blue and I work together for many years, collaborating on many projects. We engage daily in many activities and conversations, some of them characteristic of friendship, and sometimes outside of work, in group outings. Though there is no spark in our relationship, we do not dislike one another in the slightest. We just feel no special concern for one another; nor do we have any commitments to one another, or reasons to be grateful. I have never met Green. I now find myself standing in front of a burning house, able to save either Blue or Green. All else is equal.

I see no reason for me to give priority to Blue over Green. But this case does not involve affection. So now consider: Titania 2. Titania 2 engages daily in joint activities and conversations characteristic of friendship, with two individuals, Bottom 2 and Top 2, though she feels

Parfit on Love and Partiality  159 no affection for either of them. Oberon 2 sprinkles some magic juice into Titania 2’s eyes while she is asleep, which will ensure (i) that she comes to develop deep affection for the first person she sees with whom she is daily engaging in joint activities and conversations characteristic of friendship, and for whom she currently feels no special affection, and (ii) that Titania’s feelings will come to be reciprocated by this person. When Titania 2 wakes, the first person she sees is Bottom 2. After some time, she has to decide between saving Bottom 2 and Top 2 from a burning house. All else is equal.

It does not seem unreasonable for Titania 2, before Oberon’s intervention, to have felt no affection for either Bottom 2 or Top 2, nor for them to have felt none for her. The cases of Titania, and Blue and Green, suggested respectively that neither mere affection nor a shared history can ground partiality. That of Titania 2 suggests that the same is true of the affection that often develops when people share histories. Note that this is not to say that such affection is unreasonable or objectionable in itself. What is unreasonable is believing that such affection grounds reasons to discriminate against those for whom one feels no such affection.

6.  Commitments In some cases, we make commitments to those we love, implicitly or explicitly. As I mentioned above, in Reasons and Persons Parfit said that he had not yet decided whether act consequentialism was the best moral theory. By On What Matters, it seems he had come to the view that it was not, and that there are deontological duties such as that of fidelity.9 If what justifies partiality to a person I love is my having made some commitment to them, the reason here will not be ultimately partial, since we may make promises to strangers, and the principle of fidelity will not justify partiality to someone I love if I have made no commitment to them. Further, it may be that for commitments to provide fidelity-­based reasons of partiality it must be the case that there are non-­commitment-­based reasons for such partiality. Assume that morality requires us to be impartial. It may then be immoral, other things being equal, to promise to give some other person priority. Imagine, for example, a burning house case in which I have seen both of the inhabitants enter, and have promised one, because I like her face, that I will save her rather than the other inhabitant.

9  See F. Kamm [Chapter 20 in this volume]. The ‘triple theory’ Parfit defends in OWM includes not the act consequentialist principle, but Kantian rule consequentialism (see e.g. OWM ii. 244).

160  Roger Crisp

7.  Gratitude Next on Parfit’s list of reasons for lovers to give priority to the object of their love is gratitude. Consider the following case: My benefactor. I am outside a burning house, where I can save one of my major benefactors,10 or a stranger. All else is equal.

Many will believe that I have stronger ultimate reason to save my benefactor. If I have benefited from someone’s else’s actions, would it not be objectionably ungrateful not to give them some degree of priority in my practical reasoning? Gratitude seems to require some sort of reciprocity. But, as Sidgwick notes, there is a question about whether a beneficiary’s return should be based on the effort or intention of the benefactor, or the service they provide.11 Sidgwick considers the service-­based view, and rightly points out that we would not think that a rich person saved from drowning by a poor one ought to give as a reward whatever she would have been prepared to pay in advance for her life. It might be claimed by a defender of the service view that what must be proportionate to the service is not the expression of gratitude through reward, but the degree of gratitude itself. But the fact that any reasonable reward may well be far less valuable than what the rich person would have paid suggests that more is in play in our common understanding of gratitude than merely the value of the service (such as, for example, the cost of providing the service—perhaps all the poor person had to do was push a button releasing a lifebelt). On reflection, it seems that gratitude should be seen entirely as a matter of intention on the part of the benefactor. Imagine, for example, that both P and Q sought to benefit me, by sending me a cheque, but Q’s cheque was lost in the post.12 Though in fact I shall probably be more grateful to P than to Q, and be more likely to reciprocate and to offer more in return, this seems quite unreasonable. Why should Q lose out just because of an inefficient postal service?13 10  Assume that all benefactors in the cases discussed below are major. 11 H. Sidgwick, Methods of Ethics (7th edn., London: Macmillan, 1907), 259–61. 12  The case is adapted from Abelard, Ethics, 48.13–30, in D. Luscombe (ed.), Peter Abelard’s Ethics (Oxford: Clarendon Press, 1971). 13  My argument against gratitude in this section is analogous in certain ways to what Scheffler calls ‘the distributive objection’ to the claim that we have certain special responsibilities to others (see e.g. S.  Scheffler, Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought (Oxford: Oxford University Press, 2001), 83–6). But my suggestion is not that priority based on gratitude is unfair, since this is consistent with there being a pro tanto reason to give such priority, but that the notion of gratitude is internally undermining to the point of practical irrelevance. I am relying on a broader version of what Adam Smith calls the ‘equitable maxim’, according to which blameworthiness and praiseworthiness depend on a person’s intentions, not on irrelevant factors over which they have no control (see The Theory of the Moral Sentiments, ed. D. Raphael and A. Macfie (Indianapolis: Liberty Fund, 1982), II.iii.introd.1–5). The maxim I have in mind applies to rewards other than praise, and is a matter not of equity or justice, but of reasonableness. It seems plain unreasonable to give

Parfit on Love and Partiality  161 Should we, then, understand gratitude in terms of the ‘quality of will’ of a bene­ fac­tor, to use P. F. Strawson’s phrase?14 To deal with the lost cheque case, we might think that my duty of gratitude is not to those who actually benefit me, but to those who attempt to benefit me. Consider now the following case: Potential benefactors. I am outside a burning house, and can choose to save only one of the following: my benefactor, P; Q, who tried to benefit me and failed; or R, who, given the opportunity, would have tried as hard to benefit me as P and Q. All else is equal.

It would seem unreasonable to treat R differently from P or Q, since the quality of her will towards me is as good as that of P and Q. Now consider a further case (Potential benefactors 2), the same as Potential benefactors, except that I can also choose to save another stranger, S, who would not have benefited me, given the opportunity, but only because they had been given certain drugs without their knowledge. Without those drugs, they would, given the opportunity, have benefited me. Again, it would seem unreasonable to treat S differently from R, since her failure to be well disposed to me is not her fault. What might seem to matter, then, is not the mere quality of a person’s will, but the quality of a person’s unmanipulated will. Consider one further case (Lack of concern), in which, in addition to P, Q, and R, there is another stranger, T, who would not have benefited me and who has not been manipulated by others. Here we have the intuition that it would not be unreasonable to give priority to P, Q, R, or S over T, perhaps choosing which of the four to benefit by lot. The quality of T’s will is less good than that of the others, but this is almost certainly to some practically relevant degree because of factors beyond T’s control. T is the way she is in part because of her genetic pre­dis­pos­ ition to concern for others, her early experiences, her moral education, various events that have taken place, and so on. It may be that her own choices and decisions have affected her current character, but the fact remains that had things gone differently the quality of her will could have been the same as the others. For that reason, it seems unreasonable to act in a way that amounts to indirect discrimination against her, or indeed anyone else, in distributing benefits on the basis of gratitude.15

someone priority over another merely because, by accident, she has succeeded in benefiting one. Here we see the same arbitrariness and contingency as in the Titania cases above. (The problem for partiality of differential luck is well stated by Rachels, ‘Morality, Parents, and Children’, repr. in his Can Ethics Provide Answers? (Lanham, MD: Rowman and Littlefield, 1997), 214–16.) 14 ‘Freedom and Resentment’, repr. in G.  Watson (ed.), Free Will (2nd edn., Oxford: Oxford University Press, 2003), 83. 15  Note again Parfit’s rejection of desert as a ground for distribution at OWM i. 258–63.

162  Roger Crisp

8.  Family Relations The final item on the list are facts about love for our parents, children, or other close relatives.16 Again, it is not clear which facts Parfit has in mind here. But if he believes that merely being genetically related to some other individual provides a reason to give that person priority over others, the following case suggests that he may be mistaken: Lost sperm. A man provides sperm for a randomized controlled trial, making it clear that he does not consent to this sperm being used for any other purpose. The test-­tube containing the sperm is blown by wind from an open window into a package of test-­tubes from sperm donors, which are then sent to a fertility clinic. The sperm is used to fertilize an egg, and a baby is later born. Several years later, the man finds himself outside a burning house, where he can save either his child or some other child. All else is equal.

It is difficult to believe that this man’s genetic relation to his child provides him with an ultimate reason to give priority to that child, whether or not he is aware of the events leading to his genetic parenthood. Note that my argument is not meant to rest on our judgement of this case in particular. As we have already seen, there are different examples which might suggest the opposite conclusion.17 The point of this case, and indeed most of the other cases I have described, is heuristic: here, to enable us to see that, on reflection, mere biological relationships are morally irrelevant. By isolating them, we can see them for what they are.18

9.  Psychological Connectedness I have been arguing against each of the items in the list of reasons for partiality Parfit provides at On What Matters, i. 73. Later in On What Matters, at i. 136,

16  Before this final item, Parfit mentions facts involved in romantic or erotic love. Parfit does not expand on which facts he has in mind, but I hope what has already been said might throw doubt on some of the more plausible candidates, such as affection or shared histories. 17 See e.g. J.  McMahan, The Ethics of Killing: Problems at the Margins of Life: Problems at the Margins of Life (Oxford: Oxford University Press, 2002), 376. 18  It might be claimed that some of the various factors I isolate may be morally relevant when brought together in a single case. They may be, but some argument would be required to show how two morally irrelevant properties could combine to become morally relevant. Note also that the un­usual nature of the cases helps avoid our assessment of them being distorted by common-­sense intuitions concerning partiality of the very kind we are seeking to hold up to the light (e.g. our view that a long-­lost child who knocks on our door asking to meet us should not be treated like a mere stranger). Moral philosophy requires the rational assessment of normative propositions, and such assessment is made more difficult by the engagement of one’s sentiments and emotions. This is why the abstract examples used by Parfit are often more helpful than those taken from history or literature.

Parfit on Love and Partiality  163 Parfit suggests that the psychological relations we have to certain others, such as our close relatives and those we love, to some degree mirror those that ground concern about our own future well-­being. So we can be said to have reasons to be partial to those with whom we have what he calls close ties. If we are to understand the earlier list as speaking of various psychological relations in which an individual stands to those they love, my arguments above against the items on the list will still be applicable. But perhaps we have here a separate argument, which extends self-­concern to other-­concern.19 Which psychological relations ground concern about one’s own future? One important one, of course, is memory. That one’s future self will remember one’s current experiences gives one some reason to be concerned about one’s future self. Other relevant relations might include intentions, beliefs, and desires (RP, 205). And there are various ways in which my psychological states may stand in an analogous relation to psychological states of others: we may remember some common event, we may have shared intentions, or my beliefs might have shaped theirs. But now consider the following case:20 Torture. A sadist has you in their basement. They offer you a choice: Either A: You will be severely tortured for one minute and then killed. Or B: You will be much more severely tortured for one year and then killed; but before the torture begins, all your particular memories, intentions, beliefs, desires, etc. will be painlessly erased.21

I find it hard to believe that B is the rational option for me to choose, from the self-­interested point of view, and that choosing A would be an act of great self-­ sacrifice to protect the subject of the torture in option B. This suggests that psychological relations of the kind Parfit mentions may not be necessary for future-­related concern. This example, however, concerns only hedonic states. Perhaps psychological relations are necessary for future-­related concern as far as non-­hedonic components of well-­being are concerned. Mere survival does not matter; but survival may matter conditionally, in so far as it enables me to acquire certain goods I

19 For a sophisticated development of this position, see D.  Brink, ‘Rational Egoism, Self, and Others’, in O. Flanagan and A. Rorty (eds.), Identity, Character, and Morality (Cambridge, MA: MIT Press, 1990), 339–78; ‘Self-­Love and Altruism’, Social Philosophy and Policy 14 (1997): 38–41; ‘Rational Egoism and the Separateness of Persons’, in J. Dancy (ed.), Reading Parfit (Oxford: Blackwell, 1997), 126–8; ‘Impartiality and Associative Duties’, Utilitas 13 (2001): 166–9. 20  See Williams, ‘The Self and the Future’, repr. in his Problems of the Self (Cambridge: Cambridge University Press, 1973), 51–3; P. Van Inwagen, Material Beings (Ithaca, NY: Cornell University Press, 1990), 183–4. 21  How long this removal takes seems to me irrelevant, as long as there is no psychological connection remaining at the end of the process. Once the torture starts, you will of course develop a strong desire for it to stop, but this will be a new desire with the same content as the old one.

164  Roger Crisp could not otherwise acquire. Consider, for example, accomplishment and the acquisition of understanding, which are often cited as constituents of well-­being in so-­called ‘objective list’ accounts (see e.g. RP, 499). To accomplish something, or to acquire and possess understanding of something, requires a high degree of psychological continuity and connectedness.22 But these goods are usually conceived of as components of an individual’s well-­being (see RP, appendix I). Could such values ground reasons for other-­regarding concern? We do speak of accomplishment and the acquisition of understanding by groups as well as individuals. Perhaps members of certain groups have reasons to be partial to their fellow members because such partiality will serve to advance the well-­being of the group. As far as I know, Parfit did not discuss the question whether groups might have a well-­being independently of the well-­being of the individuals constituting that group. But in discussing moral mathematics, Parfit claims that we should sometimes ask not, ‘What should I do?’, but, ‘What should we together do?’ (e.g. RP, 77). Above I suggested that partial concern on its own cannot plausibly be claimed to make a world better, and so ground reasons to act. The claim that accomplishment and understanding by groups can make a world better is more plausible, and might allow a place for partiality to be ultimately or non-­instrumentally valuable as a constituent of such accomplishment and understanding (though of course groups might well accomplish or understand without partiality). Such group goods might well ground reasons for groups to act.23 They would also give rise to some deep problems about the implications of group goods and group reasons for the practical reasoning of individuals about their own actions.24

Bibliography Abelard, P., Ethics, in D.  Luscombe (ed.), Peter Abelard’s Ethics (Oxford: Clarendon Press, 1971). Brink, D., ‘Rational Egoism, Self, and Others’, in O.  Flanagan and A.  Rorty (eds.), Identity, Character, and Morality (Cambridge, MA: MIT Press, 1990), 339–78. Brink, D., ‘Self-Love and Altruism’, Social Philosophy and Policy 14 (1997): 127–57. Brink, D., ‘Rational Egoism and the Separateness of Persons’, in J. Dancy (ed.), Reading Parfit (Oxford: Blackwell, 1997), 96–134. Brink, D., ‘Impartiality and Associative Duties’, Utilitas 13 (2001): 152–72. 22  For one to have a reason to care at some particular time about such accomplishment or understanding in the future, it also seems plausible that there should be psychological connection between the self at that time and the future accomplishing and understanding selves. 23  See e.g. F. Jackson, ‘Group Morality’, in P. Pettit, R. Sylvan, and J. Norman (eds.), Metaphysics and Morality: Essays in Honour of J. J. C. Smart (Oxford: Basil Blackwell, 1987), 91–110. 24 See RP, 70–3.

Parfit on Love and Partiality  165 Crisp, R., ‘Against Partiality’, Lindley Lecture 2018, https://kuscholarworks.ku.edu/ bitstream/handle/1808/26747/Crisp_2018_AgainstPartiality.pdf ?sequence= 3&isAllowed=y. Jackson, F. ‘Group Morality’, in P. Pettit, R. Sylvan, and J. Norman (eds.), Metaphysics and Morality: Essays in Honour of J.  J. C.  Smart (Oxford: Basil Blackwell, 1987), 91–110. Keller, S.,‘How Do I Love Thee? Let Me Count the Properties’, American Philosophical Quarterly 37 (2000): 163–73. Kolodny, N., ‘Love as Valuing a Relationship’, Philosophical Review 112 (2003): 135–89. McMahan, J., The Ethics of Killing: Problems at the Margins of Life (Oxford: Oxford University Press, 2002). Parfit, D., Reasons and Persons (Oxford: Clarendon Press, 1984). Parfit, D., On What Matters, vols. i–ii (Oxford: Clarendon Press, 2011). Parfit, D., On What Matters, vol. iii (Oxford: Clarendon Press, 2017). Rachels, J., ‘Morality, Parents, and Children’, repr. in his Can Ethics Provide Answers? (Lanham, MD: Rowman and Littlefield, 1997): 213–33. Raz, J., Value, Respect, and Attachment (Cambridge: Cambridge University Press, 2001). Scheffler, S., Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought (Oxford: Oxford University Press, 2001). Sidgwick, H., The Methods of Ethics (7th edn., London: Macmillan, 1907). Smith, A., The Theory of the Moral Sentiments, ed. D.Raphael and A.  Macfie (Indianapolis: Liberty Fund, 1982). Strawson, P.  F., ‘Freedom and Resentment’, repr. in G.  Watson (ed.), Free Will (2nd edn., Oxford: Oxford University Press, 2003): 72–93. Van Inwagen, P., Material Beings (Ithaca, NY: Cornell University Press, 1990). Williams, B., ‘The Self and the Future’, repr. in his Problems of the Self (Cambridge: Cambridge University Press, 1973), 46–63.

7

Individualist Utilitarianism and Converging Theories of Rights Elizabeth Ashford

This paper develops two core themes of Derek Parfit’s philosophy. The first is his goal of unifying the two main rival impartial moral theories, Kantianism and consequentialism, thereby reinforcing their claim to pertain to objective moral truths. The second is his focus on the moral significance of the combined effects of many agents’ behaviour, and on the challenges this poses to ordinary moral thinking. This is a theme that runs throughout his work, and that he returns to at the very end of volume iii of On What Matters. A principal respect in which Kantianism and consequentialism have been thought to be fundamentally incompatible is the issue of rights and trade-­offs. A long-­standing objection to so-­called ‘classical utilitarianism’ is that it overlooks the separateness of persons.1 In this paper, I outline the version of utilitarianism I take to be most plausible, which I call ‘individualist utilitarianism’, and contrast it with the view that has come to be known as ‘classical utilitarianism’. I aim to show that individualist utilitarianism is based on a more plausible understanding of the core tenets of utilitarianism, and of the nature of impartial sympathetic concern at the heart of the theory, than so-­called ‘classical utilitarianism’. I argue, on the one hand, that the ‘separateness of persons’ objection is indeed a clinching objection to classical utilitarianism. On the other hand, I argue that individualist utilitarianism offers a powerful account of rights, as grounded on the moral status of the individual right-­holder. I then argue that if we focus on individualist utilitarianism rather than classical utilitarianism, there is much more convergence between utilitarian and Kantian accounts of the nature of duties of justice imposed by rights than there is commonly taken to be. Core tenets of both theories entail the acknowledgement of certain general positive duties of basic justice. As I contend, many prominent Kantian arguments for the claim that there is a rigid dichotomy between general positive and negative duties of justice, such that only the latter can constitute duties of justice, give insufficient attention to the moral significance of what we 1  John Rawls, A Theory of Justice (rev. edn., Cambridge, MA: Belknap Press, 1999), 163. Elizabeth Ashford, Individualist Utilitarianism and Converging Theories of Rights In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0008

170  Elizabeth Ashford together do. Moreover, both theories should acknowledge the intrapersonal trade-­offs, between different interests of the individual right-­holder, that arise when the status quo is fundamentally unjust; and both theories highlight the moral urgency of recognizing and removing the background injustice that leads to such trade-­offs. What I shall be arguing for, then, is a pluralist approach, according to which each of the theories reaches the same conclusion from a very different theoretical foundation. This is a continuation of Parfit’s project of seeking convergence between seemingly rival impartial moral theories. It is also a continuation of the project explored most fully in Reasons and Persons but to which he returns in On What Matters, of applying impartial moral theorizing to challenge core aspects of common-­sense moral thinking, in order to suggest a more adequate moral response to some of the most prevalent and widespread threats to persons’ basic interests, in particular those that are the combined effect of many agents’ behaviour. Section 1 discusses individualist utilitarianism and the ways in which it contrasts with classical utilitarianism. Section 2 discusses the individualist utilitarian analysis of rights, and the debate between this account and that of Kantian deontology over the issue of trade-­offs. Section 3 defends the claim that there is considerable convergence between these two theories over the nature of perfect duties of justice correlative to basic human rights.

1.  Classical Utilitarianism, Individualist Utilitarianism, and the Separateness of Persons 1.1  Classical utilitarianism According to classical utilitarianism, the right action or policy is the one that maximizes the conglomerate sum total of welfare. No intrinsic weight is assigned to how the welfare is distributed among individual lives. The classical utilitarian conception of impartiality is that equal quantities of welfare have equal moral weight, however they are distributed across lives. This conception of impartiality has two implications, which tend not to be adequately distinguished by proponents of classical utilitarianism. The first implication is that a benefit or burden of a certain magnitude is of equal moral weight in whichever person’s life it is instantiated. This is an appealing expression of the claim that everyone’s well-­being matters equally. It simply implies that a certain benefit or burden matters just as much if it is instantiated in the life of a pauper or a prince. However, the second implication is much more problematic: that a given quantity of well-­being or suffering is of equal moral weight however it is divided up among lives: whether it is all instantiated in one person’s life, or whether it is thinly distributed among many lives. Thus, whereas the first implication is that

Individualist Utilitarianism  171 100 units of suffering are just as bad whether they are experienced by a prince or a pauper, the second implication is that 100 units of suffering are equally bad if they are all imposed on one person, or if they are thinly spread out among many persons’ lives; a hugely significant burden to one person is taken to be equivalent to the aggregate of much smaller burdens to several individuals. To bring out the problematic nature of this second implication, consider a variation of one of Parfit’s hell examples, in which each person involved experiences nothing but agony, and is benefited by dying as soon as possible. Consider two very different distributions of periods of pain of uniform intensity among different persons’ lives. We can set aside the complication of the impact the pain might have on other prudential values in the person’s life by stipulating that the persons’ lives contain no positive prudential values at all. We can also assume, for simplicity, that each ten-­second period instantiates the same total quantity of pain. In the first scenario, millions of people each suffer agony for ten seconds each (and then die), and in the second, one person suffers agony for thousands of years. Let us suppose that the first scenario—in which the pain is thinly spread out among different lives—instantiates a greater net sum total of suffering. Classical utilitarianism, which is concerned solely with the aggregate total quantity of well-­being and suffering, will judge the first scenario to be worse, and require choosing the second scenario. However, there is a widely held intuition, which I share, that the second scenario is far worse, because of the horrific fate of the one person. Moreover, the claim that the second scenario is in fact less bad seems to me to be a distortion of the nature of the impartial sympathetic concern at the heart of the theory. On the most natural and appealing conception of sympathetic concern, it is directed at the persons whose well-­being or suffering it is.2 The underlying concern is for persons (and sentient beings in general, but my focus here is persons). This concern is directed at those persons’ well-­being or suffering because well-­being is, by definition, what is beneficial to persons, and suffering is, by definition, what is bad for persons. In short, we value well-­being or suffering for the sake of those who will receive it. But it follows from that, I suggest, that well-­being or suffering matters because it is instantiated in the lives of individual persons, as distinct and enduring centres of consciousness. And that implies, in turn, that impartial sympathetic concern is directed at how persons fare. Generally, concern is directed at how persons fare over the course of their lives—that is, at persons’ lifetime well-­being or life prospects. (In some cases, such as those involving periods of severe pain within lives, the concern may be directed at how persons fare over the

2  For a forceful defence of this, see T. M. Scanlon, What We Owe To Each Other (Cambridge, MA: Belknap Press, 1998), 181; Stephen Darwall, Impartial Reason (Ithaca, NY: Cornell University Press, 1983), 161–2; Kai Draper, ‘Personal and Impersonal Dimensions of Benevolence’, Noûs 36/2 (2002): 201–27.

172  Elizabeth Ashford shorter time period of the duration of the pain; but the concern is still directed at the persons’ experience of pain over that entire period, rather than at quantities of pain experience considered in abstraction from the way in which they are instantiated within individual persons’ lives. For simplicity I focus here on how persons fare over the course of their lives.) The fundamental problem with classical utilitarianism, I contend, is that it does not assign ultimate value to how individual persons fare. Knowing the conglomerate net sum total of well-­being tells us nothing about the overall lifetime well-­being (or shorter-­term fates) of any of the persons involved. What this brings out, I suggest, is the implausibility of what John Broome calls ‘the assumption of separability’.3 The ultimate units of value, according to classical utilitarianism, are the individual pieces of well-­being (or of suffering). The way in which those pieces of well-­being are packaged into lives is a patterned good, which cannot be seen by looking separately at those individual pieces of well-­ being. From this perspective, the overall fates of which those temporal sequences are a part disappear from view and are assigned no intrinsic moral weight. In the hell example I have been discussing, classical utilitarianism takes the ultimate units of value to be the individual ten-­second experiences of pain, and takes the least bad outcome to be the one that instantiates the smallest total number of them. It gives no intrinsic moral weight to whether the quantities of pain are linked together within the same enduring centre of consciousness, so as to all be constitutive of one person’s overall fate of enormously prolonged agony, or whether they are thinly dispersed among a vast number of lives. I conclude that there is an important respect in which classical utilitarianism does indeed ignore the separateness of persons: it views the moral significance of well-­ being or suffering in abstraction from its instantiation in individual persons’ lives.

1.2  Individualist utilitarianism Individualist utilitarianism, by contrast, takes the moral importance of well-­being to be grounded in the moral importance of the persons whose well-­being it is. The ultimate units of value are not individual pieces of well-­being, but individual persons’ lifetime well-­being (or shorter-­term fates). In other words, whereas classical utilitarianism takes the ultimate units of value to be individual quantities of welfare, however they are distributed among lives, individualist utilitarianism takes the ultimate units of value to be how persons fare over the course of their lives. Each person’s lifetime well-­being is a distinct, intrinsic final value. To know persons’ lifetime well-­being, you have to know how the overall quantity of well-­being is packaged into lives. The significance of the distribution of 3  John Broome, Weighing Goods (Oxford: Wiley-­Blackwell, 1995).

Individualist Utilitarianism  173 well-­being among lives is therefore built into the general good. Whereas classical utilitarianism assigns equal moral weight to equal quantities of well-­being regardless of their distribution, individualist utilitarianism assigns equal moral weight to each person’s lifetime well-­being. Classical utilitarianism amalgamates the welfare of different individuals into a conglomerate lump sum, and takes the general good to be an unstructured net total quantity of well-­being. By contrast, individualist utilitarianism takes the general good to be a plurality of distinct intrinsic final values: how each person fares over the course of their lives. The general good is the inclusive combination of each person’s lifetime well-­being. Another central difference is that classical utilitarianism treats benefits or burdens to different persons of equal magnitude as fungible parts of the general good. Therefore a burden to one person can be cancelled out by a benefit of equal magnitude to someone else. By contrast, since individualist utilitarianism treats each persons’ lifetime well-­being as a distinct final value, a burden to one person is not cancelled out by an equal quantity of benefit instantiated in other lives. The burden remains regrettable for the person’s sake. The moral point of view is that of equal individualized concern for the lifetime well-­being of all. The impartial sympathetic spectator is concerned with every person’s well-­being, for the sake of the persons whose well-­being it is. The spectator’s sympathetic concern is therefore directed at how all the persons fare, rather than at the conglomerate sum total of well-­being considered in abstraction from its instantiation in individual persons’ lives. The impartial moral point of view might be described as ‘omnipersonal’. It is a representation of the equal moral importance of how each person’s life goes. In contrast, the classical utilitarianism moral point of view is, I have argued, in an important respect impersonal: well-­being is viewed, for its moral importance, in abstraction from its instantiation of the lives of individual persons, and accordingly no ultimate moral weight is assigned to the life prospects of any of the individuals affected. (It is worth emphasizing that individualist utilitarianism is distinct from person-­affecting utilitarianism, as characterized by Jan Narveson, according to which only the well-­ being of actual persons matters morally.4 Individualist utilitarianism is perfectly compatible with taking the welfare of possible people to matter morally. More importantly, the person-­affecting utilitarianism defended by Narveson has the same problematic implications as impersonal total utilitarianism in cases involving different distributions of well-­being to the same individuals. Narveson’s view is therefore still problematically impersonal in the sense that it fails to give ultimate moral weight to the life prospects of any of the individuals involved.)

4  Jan Narveson, ‘Utilitarianism and New Generations’, Mind 76/301 (1967): 62–72.

174  Elizabeth Ashford One implication of individualist utilitarianism’s omnipersonal conception of the impartial moral point of view is that sharp priority is given to the avoidance of deprivation and suffering, over the promotion of the well-­being of those whose lives are already reasonably good.5 A full defence of this is beyond the scope of this paper, but the general thought is that if persons’ lives are sufficiently good, those persons have reason to be contented with their lives, and the impartial sympathetic spectator is correspondingly contented. A further benefit increases the impartial spectator’s sympathetic pleasure and has positive value, but no special urgency is assigned to its promotion. By contrast, if persons are suffering deprivation, they have reason to be discontented and to assign a special urgency to removing the deprivation, and, correspondingly, so does the impartial sympathetic spectator.6 The blighting of one person’s life is not cancelled out by a benefit of equal magnitude to others, and the impartial spectator’s regret is not compensated. The impartial sympathetic spectator assigns an even greater priority to the prevention or alleviation of severe pain; it is the nature of such pain that it is experienced as unbearable and thus as intensely demanding to be removed, and the spectator assigns a corresponding urgency to its removal. At the same time, however, since the general good is the plurality of everyone’s lifetime well-­being (represented by the moral point of view of impartial concern for the well-­being of all rather than of each considered seriatim), it is intrinsically worse for a greater number of people to suffer a certain burden, from the moral point of view. While there is no conglomeration of different persons’ well-­being into a net total quantity of well-­ being, the number of persons who would experience a certain burden is assigned intrinsic moral weight. The overarching goal is to promote the general good as much as possible, understood as the inclusive combination of everyone’s lifetime well-­being, and this requires weighing up a plurality of competing principles, with no precise formula by which to do so. One principle is that each person’s well-­being over the course of their lives matters equally. Another is that each person’s life prospects should be promoted as much as possible. A third is that intrinsic moral weight is assigned to the importance of the interests at stake, and that special urgency is assigned to what Mill calls ‘the essentials of human well-­being’, lack of which constitutes deprivation or suffering. A fourth principle is that intrinsic moral weight is assigned to the number who would suffer a certain benefit or burden. 5  Roger Crisp, ‘Equality, Priority, and Compassion’, Ethics 113/4 (2003): 745–63; Harry Frankfurt, ‘Equality and Respect’, Social Research 64/1 (1997): 3–15. 6  Along similar lines, Roger Crisp offers a powerful analysis of the special moral importance of benefiting those who are badly off, in terms of the ideal spectator’s virtue of compassion: ‘a truly virtuous spectator would be more concerned for those who are badly off. What virtue lies behind such concern? The answer is compassion: the spectator puts himself or herself into the shoes of all those affected and is concerned more to the extent that the individual in question is badly off ’ (Crisp, ‘Equality, Priority, and Compassion’, 756–7).

Individualist Utilitarianism  175 This might seem so far from the standard modern conception of utilitarianism, which Rawls calls ‘classical utilitarianism’, that it is unclear how it could be understood as a version of utilitarianism at all. However, as I now suggest, the pluralist and imprecise welfare function I have described, and the individualist conception of welfarism and impartiality that underlies it, is reflected in core passages in Bentham and Mill.

1.3  Individualist utilitarianism and the classical utilitarians One of Bentham’s key formulations of the welfare function is ‘the equal greatest happiness of all, without exception’, which contains competing variables: maximizing happiness, ensuring that no one is left out from the distribution of happiness, and promoting persons’ happiness equally. In another important passage, Mill refers to each person’s right to ‘all the means of happiness’, which suggests a conception of impartiality that assigns equal weight to each person’s overall life prospects rather than to equal quantities of well-­being. These formulations were tidied up by Sidgwick, who understood the formula of ‘the greatest happiness’ as referring solely to the greatest conglomerate net sum total of happiness, and took this to be definitive of the general good. Sidgwick took the goal of moral theory to be that of providing a scientifically precise principle from which to evaluate outcomes and thereby deduce the right action in any situation. Sidgwick’s goal of providing scientific precision is expressed in other passages of Mill and Bentham, such as Mill’s description of the utilitarian principle of impartiality as ‘a purely arithmetic notion’. There is a tension between this arithmetic conception of impartiality, and the affirmation of each person’s equal right to ‘all the means of happiness’. This suggests that Mill and Bentham did not recognize or adequately think through the difficulty of reconciling a precise, arithmetic notion of impartiality and maximization with the individualist conception of welfarism and impartiality reflected in other core passages. Sidgwick’s formulation of the right action as the one that maximizes the net sum total of happiness subsequently came to be taken to be definitive of utilitarianism. Moreover, this is the position that has come to be known as ‘classical utilitarianism’. Nevertheless, I suggest that the individualist conception of welfarism and impartiality expressed in core passages in Mill and Bentham, and the competing principles it generates, indicates a more promising direction in which utilitarianism could have developed, and may in fact have a better claim to the title of ‘classical utilitarianism’. In his commentary on Parfit, Allen Wood offers an illuminating contrast between the Millian method of moral philosophy (together with Kant’s), and the Sidgwickian method. Sidgwickian theory, he argues, ‘sets itself the goal of providing a precise principle or set of principles which, along with a set of facts,

176  Elizabeth Ashford enable one to deduce the “right” conclusion about what to do under any conceivable situation’. By contrast, the Millian and Kantian methods appeal to a fundamental moral principle, which is ‘an articulation of a basic value (that of rational nature for Kant, that of happiness for Mill)’, that is ‘drawn from very general and fundamental considerations’. They then derive from that principle various moral rules and duties not deductively, but through ‘an interpretation of the normative principles applying that basic value under the conditions of human life’. Setting aside Mill’s ‘arithmetic’ account of impartiality, I suggest that his conception of impartiality as recognizing each person’s equal claim to all the means of happiness articulates the basic value of the equal moral importance of each person’s life.

2.  Rights and Trade-­Offs 2.1  Individualist utilitarianism and rights Classical utilitarianism, I have argued, does problematically ignore the separateness of persons, and therefore cannot ground an adequate account of rights. As the pain example illustrates, any amount of suffering could be outweighed by a relatively trivial burden to a sufficiently large number of others.7 By contrast, an individualist utilitarian account of rights takes them to be based on the importance of the well-­being of the individual right-­holder, which such rights protect. As we have seen, the general good is the inclusive combination of each person’s lifetime well-­being, and each person’s well-­being over the course of their lives is a distinct, final value. A burden to one person is not made up for by a benefit to someone else, and special weight is given to the avoidance of deprivation and suffering. Since rights protect, in Mill’s words, ‘the essentials of human well-­ being’, their fulfilment is the most important constituent of the general good. These are the kinds of interests the non-­enjoyment of which would be liable to blight persons’ lives. Rights are highly resistant to trade-­offs, given the importance of the interests they protect. ‘The essentials of human well-­ being’ that rights protect are discontinuously more valuable than other types of interests. Rights therefore trump those other interests.8

7  This is not to claim that ten seconds of agony is at all trivial in an absolute sense. My point is that it is not of comparable moral significance to thousands of years of suffering. 8  David Brink offers an illuminating analysis of Mill’s pluralist conception of happiness, according to which there is a hierarchy of values, such that the interests that rights protect trump lesser goods (‘Mill’s Deliberative Utilitarianism’, Philosophy & Public Affairs 21 (1992): 67–103). James Griffin offers a powerful analysis and defence of discontinuity in value, and its implications for a teleological conception of rights (Griffin, Well-­Being: Its Meaning, Measurement and Moral Importance (Oxford: Clarendon Press, 1986); On Human Rights (Oxford: Oxford University Press, 2008)).

Individualist Utilitarianism  177 Rights are not, however, immune to trade-­offs, when persons’ basic interests unavoidably conflict. But if we focus on individualist utilitarianism rather than classical utilitarianism, then the reason for such trade-­offs is not that individualist utilitarianism ignores the separateness of persons. Rather, it is because of two features of individualist utilitarianism. The first is its rejection of the claim that the distinction between acts and omissions is of intrinsic moral significance. The agent has to take just as much responsibility for harms she allows as for harms she herself inflicts. This means that in circumstances in which the agent is unable to protect each person’s basic interests, she cannot avoid being responsible for a trade-­off. The second feature of individualist utilitarianism that underpins its account of trade-­offs is that it assigns intrinsic moral weight to the number of people who are affected in various outcomes. Since the general good is the inclusive combination of everyone’s life prospects, the situation is worse the greater the number who suffer a certain burden. Therefore, in a situation in which it is not possible to protect everyone from suffering a severe burden, concern for the interests of each person has to be weighed up against concern for the interests of all the rest, from the impartial moral point of view. Although a severe burden to one person cannot be cancelled out by benefits to others, it might be tragically outweighed by an even greater burden to someone else, or by a comparably severe burden to a large number of others. Let us briefly consider some of the central implications of these two features. First, there are fundamental general positive rights to the provision of goods and services, such as health care and sanitation, given the importance of the interests such rights protect.9 Second, the identification of positive rights to the provision of scarce resources has to take into account not only the importance of the interest the right would protect, but the cost this would impose on others. For example, the identification of rights to various kinds of health care has to take into account the overall level of medical technology and resources, and the number of people with claims on that resource. (A core implication of the impartiality of utilitarianism is that the level of resources relevant to identifying a human right to health care is global, rather than domestic.) From the individualist utilitarian moral point of view, the interests of each person have to be weighed up against the interests of everyone who would be affected by the allocation of a scarce resource. To assert a right is not simply to play the trump card against any rival moral claims. Rather, it is a way of asserting the moral significance of one’s own life while simultaneously recognizing the equal moral significance of others’ lives. 9  It should be noted that, as Henry Shue has shown, it is highly problematic to draw a sharp distinction between positive and negative rights, and I have defended his argument elsewhere (Ashford, ‘The Alleged Dichotomy Between Positive and Negative Rights and Duties’, in Charles Beitz and Robert Goodin (eds.), Global Basic Rights (Oxford: Oxford University Press, 2009)).

178  Elizabeth Ashford James Griffin gives the example of someone suffering a very rare-­life threatening disease. Griffin argues that if saving the persons’ life required mounting a hugely expensive crash programme to try to find a cure for the disease, it would not be plausible to assume the person had a right to such a cure. On an individualist utilitarian analysis, although the person’s death would be deeply regrettable, he would not been wronged by not getting the cure, given the cost this would have imposed on others. At the opposite end of the spectrum are deaths from lack of safe drinking water or adequate nutrition. Such deaths are utterly gratuitous. Paul Farmer points out that in communities in which such deaths are endemic, they are referred to as ‘stupid deaths’, given that they could be prevented at such small cost, relative to the overall level of global resources.10 The failure to prevent such deaths is fundamentally incompatible with minimally adequate recognition of the moral importance of the lives that are blighted or destroyed as a result. It appropriately arouses moral indignation and outrage. There is thus a sharp distinction between tragic and unavoidable trade-­offs, and trade-­offs that are unacceptable and incompatible with adequately acknowledging the moral significance of each person’s life. Rights are protections against unacceptable trade-­offs. The role of human rights is to ensure that, barring tragic cases such as rare life-­ threatening medical conditions that would be hugely expensive to treat, the essentials of well-­being are protected for each person, to a reasonable degree of security. Fundamental human rights are protections of persons’ basic interests against what Henry Shue calls ‘standard threats’, where standard threats are threats that are reasonably common, and are socially eradicable at reasonable cost. As Shue argues, both the right to security and the right to subsistence are basic human rights.11 Rights themselves are not appropriately subject to trade-­offs, barring exceptional circumstances, if the status quo is just. Nevertheless, a third implication of individualist utilitarianism’s acknowledgement of positive duties of justice to protect persons’ basic interests, together with its conception of the moral point of view as that of impartial concern for the life prospects of all, is that there can be circumstances in which agents on the ground cannot avoid facing awful trade-­offs. An agent might be required to violate one person’s right in order to prevent more serious or numerous violations. Amartya Sen gives the example of a requirement to violate one person’s property rights in order to prevent a vicious racist assault—which, we can suppose, would result in the victim’s brain damage or paralysis, and thereby mar their life. An agent might

10  Paul Farmer, ‘Landmine Boy’ and Stupid Deaths’, in Partner to the Poor: A Paul Farmer Reader, ed. Haun Saussey (Berkeley and Los Angeles: University of California Press, 2010), 409. 11  Shue offers a compelling defence of taking both rights to be basic in Basic Rights: Subsistence, Affluence, and U.S. Foreign Policy (Princeton: Princeton University Press, 1996).

Individualist Utilitarianism  179 even be required to kill one person if this really were the only way of preventing a number of others from being killed. It should be emphasized, though, that if the status quo were just, then such trade-­offs would be very rare. Under a just status quo, each person’s access to the objects of their fundamental human rights (including physical security, together with the means of subsistence and a reasonable level of medical care) would be socially guaranteed to a reasonable degree of security. Agents would face trade-­ offs between rights only as a result of egregious actions by other agents, that occurred despite reasonable social protection against such actions, through a well-­functioning juridical system and so on. Barring such egregious actions, and barring rare and tragic events or circumstances, the essentials of well-­being would be protected for each person; no one’s life would be gratuitously blighted or destroyed. As I discuss in the final section, when the status quo is fundamentally unjust, then agents on the ground may face awful trade-­ offs even under standard circumstances. However, as I shall argue, individualist utilitarianism—together with Kantianism—highlights the moral urgency of reforming the fundamentally unjust socio-­economic structures that that lead to such trade-­offs.

2.2  The debate over trade-­offs The opposing view is that rights impose constraints against inflicting severe harm even if this is the only way of preventing greater harm to others. Nagel and Nozick take rights to constitute side-­constraints on action, which impose only negative correlative duties, in order to avoid the ‘utilitarianism of rights’ I have been describing. If rights impose only negative duties then agents cannot be required to violate a right in order to fulfil a positive duty to prevent more serious or numerous violations. One rationale for such constraints appeals to agent-­ centred restrictions, according to which agents are especially responsible for harms they themselves inflict. Another type of rationale for such constraints is victim-­centred, and appeals to persons’ moral status as inviolable.12 Utilitarians, in turn, reply that such constraints give inadequate weight to the interests of all the others who would foreseeably be harmed as a result of the agent’s compliance with the constraint.13 And in response to a focus on the moral status of inviolability, they appeal to the persons’ moral status as savable, or unignorable.14

12  Frances Kamm, Morality, Mortality, ii. Rights, Duties, and Status (New York: Oxford University Press, 2001), 358–61. 13 Kamm, Morality, Mortality, ii.362. 14 Kamm, Morality, Mortality, ii.276.

180  Elizabeth Ashford If we focus on individualist utilitarianism rather than classical utilitarianism, then the debate between the utilitarian and the Kantian accounts of rights is over the appropriate response to the moral status of individual persons. An individualist utilitarian account of rights (unlike that of classical utilitarianism) cannot be dismissed on the ground that it ignores the separateness of persons. The debate over the issue of trade-­offs is principally shifted to two key questions: how much moral weight should be assigned to the distinction between acts and omissions; and to the number of violations that would occur in various outcomes and to the importance of the interests at stake. Individualist utilitarianism takes rights to impose both positive and negative correlative duties. Kantianism, on the other hand, is commonly characterized as drawing a sharp dichotomy between general positive and negative duties. On this view, general negative duties are strict, exceptionless, perfect duties of justice, which each agent can and should always completely fulfil, at all times and in their conduct to all. By contrast, general positive duties are taken to be imperfect duties of virtue, which are different in kind from duties of justice (with the important exception of duties of easy rescue, to which I return). Accordingly, as Onora O’Neill argues, ‘the central demand of Kantian justice is negative’.15 I now argue that there is in fact considerably more convergence between these two impartial moral theories on the nature of general duties of justice than there is commonly taken to be. Kantianism, as well as individualist utilitarianism, should take certain general positive duties to be duties of basic justice correlative to general human rights to the provision of goods and services. I focus on the positive duties correlative to the right to subsistence, which are, I argue, in fact close in nature to its negative correlative duties. I further argue that both theories should take the right to subsistence to be a basic human right, fulfilment of which is essential to the avoidance of intrapersonal trade-­offs between the interests of the individual right-­holder in virtue of which other human rights cannot be adequately fulfilled, or in certain cases even respected.

3.  Towards Convergence I begin by critiquing the claim that there is a rigid dichotomy between general positive and negative duties per se, such that only the latter can be perfect duties of justice correlative to general human rights. Arguments for such a dichotomy, I contend, tend to implicitly assume a conception of duties of justice that is arbitrarily restricted to small-­ scale interactional social contexts. I focus on O’Neill’s important and influential version of a Kantian argument against the 15 O’Neill, Faces of Hunger: An Essay on Poverty, Justice and Development (London: Allen & Unwin, 1986), 141.

Individualist Utilitarianism  181 claim that there can be genuine general human rights to the provision of goods and services.

3.1  O’Neill’s interactional argument O’Neill’s argument implicitly assumes an interactional conception of human rights, as claims against the conduct of each agent, considered one by one. Paradigm interactional negative duties are duties not to torture, assault, and so on. These duties are held by each agent, seriatim. They are specific proscriptions on acts of a certain type, which each agent can, individually, completely discharge, at all times and in her conduct towards all. These duties are, therefore, completely enactable to all by each, considered seriatim. And if an agent fails to comply with the duty on any occasion, she will be individually responsible for the infliction of a drastic harm on a particular victim, and can be identified as the perpetrator of the violation of the victim’s right (against torture, and so on). Accordingly, particular agents can be identified as specifically responsible for the interactional violations suffered by particular victims. (The agents can be collective, as in the case of state crimes.) O’Neill’s argument against the claim that there can be general positive human rights to the provision of goods and services appeals to a contrast between the general positive interactional duties which, she assumes, would correspond to such rights, and the interactional negative duties correlative to negative human rights such as the right against torture. It is clearly not the case that each agent, considered seriatim, can discharge to all a general positive duty to provide them with object of a general positive human right. O’Neill infers from this that the general positive duties correlative to purported general positive human rights are not enactable to all. She concludes that they cannot in fact be general duties of justice, fulfilment of which is owed to all. She further argues that the content of individual agents’ general positive duties is not determinate; they must choose whom to help, and what kind of help to give. She then argues that since the content of the duty is indeterminate, the content of the right itself is indeterminate, and therefore cannot be a genuine, claimable human right. Positive duties are delimited, determinate, and individually enactable ­requirements only if they are special. O’Neill concludes that genuine rights to the provision of goods and services can only be special rights, rather than general human rights. However, as I now argue, the interactional account of the general duties of justice correlative to fundamental human rights is too narrow. I begin by showing that in our contemporary complex, large-­scale social context, the interactional account fails to accommodate many of the most urgent negative duties of justice, including those correlative to widely affirmed human rights.

182  Elizabeth Ashford

3.2  The nature of our general negative duties of justice concerning what we together do In Parfit’s ‘harmless torturers’ example, the torturers move away from the ‘bad old days’, in which they were each directly inflicting agonizing electric shocks on particular victims.16 Instead, they introduce a mechanized system in which the electric shock each agent administers flows into a central pool, and is then spread out over a huge number of prisoners, so that it now causes only a very small incremental amount of electric shock to each of the prisoners, which is barely perceptible or even imperceptible. It is now the case that none of the agents, considered seriatim, is responsible for inflicting severe pain on any of the victims, although the total quantity of pain suffered by each prisoner remains the same as under the previous system. Clearly, the agents’ behaviour is just as bad as before, and should still be classified as a human rights violation. Moreover, it remains a violation even under a range of variations of the example, in which the agents are not intending the harm, and are not deliberately coordinating their behaviour to cause it. Let us consider a variation in which the agents have not set up the arrangements themselves, and work in isolation from each other. They are each allocated responsibility for manufacturing or maintaining different components of the torture equipment, and the actual infliction of the pain is done by robots. Moreover, the manufacture of different components is done in different factories. The agents each focus their attention on the technical details of the functioning of a particular component part, and remain wilfully ignorant of the overall harm the equipment causes. Although none of the agents are directly inflicting any pain at all, and they are not intentionally acting as a coordinated group in order to inflict torture, it does not seem plausible to deny that this counts as a human rights violation. It is enough that they are each participating in an ongoing practice that foreseeably leads to the infliction of agony on the victims, and that they each ought to be aware of the harm they are together causing. The harmless torturers example therefore shows that in certain contexts, our obligations can only be adequately grasped by shifting our focus to what we are together doing. This has significant implications for our understanding of the nature of our duties of basic justice in our contemporary social context, of complex large-­scale global interaction. Many of the most severe and prevalent contemporary harms are the combined effects of the behaviour of a vast number of agents. A notable example of this is anthropogenic climate change and environmental degradation, which is jeopardizing the means of subsistence of a vast number of current poor and future generations.17 16  Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), 80–1. 17  The Intergovernmental Panel on Climate Change has identified what is climatically known as the Intertropical Convergence Zone as the area in which the most severe threats to subsistence are

Individualist Utilitarianism  183 Let us therefore consider the nature of the negative duties correlative to the human right to subsistence. Following Henry Shue, I take the object of the right to subsistence to be access to the means of subsistence to a reasonable degree of security. Accordingly, the negative correlative duty, not to deprive persons of the object of the right, is the duty not to plunder, jeopardize, or destroy persons’ means of subsistence. There is, in fact, broad agreement that the plunder, jeopardizing, and destruction of persons’ means of subsistence is widespread.18 However, it rarely takes the form of an interactional violation. Interactional violations of the right to subsistence would be specific acts involving the plundering, jeopardizing, or destruction of persons’ means of subsistence. Generally, however, the physical destruction or degradation of persons’ means of subsistence (such as access to fertile land and uncontaminated drinking water) is the combined effect of processes involving the activities of a vast number of agents. And the plunder of persons’ access to the means of subsistence normally involves complex large-­scale processes, with deep historical roots, under which world’s resources come to be partitioned and possessed by a small fraction of the world’s population, while the global poor are deprived of access to even enough resources for subsistence. It is generally hard or impossible to identify particular individual or collective actions that can be held to constitute the plundering, jeopardizing, or destruction of particular persons’ means of subsistence, and there are generally no clear lines of responsibility by which to identify particular individual or collective agents as responsible for depriving particular victims of the means of subsistence. There is also broad agreement that given the overall level of global natural, technological, and social resources, there are feasible alternative structures that would avoid depriving persons of secure access to the means of subsistence—and moreover, that this could be achieved without imposing immoderate cost to any duty-­bearer. Thus, a vast number are deprived of the object of the right as a predictable and avoidable result of human behaviour, and yet most of this behaviour cannot be identified as an interactional violation. Therefore the assumption that duties of justice correlative to human rights must take an interactional form would imply that not only the positive right to the provision of subsistence, but in addition, the negative right not to be deprived of the means of subsistence—a right against behaviour that predictably and

already taking place and predicted to rapidly worsen as greenhouse gas emissions continue to rise. This area embraces the tropics, where the vast majority of the remaining subsistence societies are located. Of course, the threat to clean air is not principally confined to the global poor. IPPC 5th Assessment Report (AR5), 2014. https://www.ipcc.ch/publications_and_data/publications_and_data_ reports.shtml (accessed 11 May 2014). 18  As Kaushik Basu (recently chief economist of the World Bank) puts it: ‘Exploitation, conquest and property grabbing are alive and well. . . . Whole nations, groups and masses of people are being continuously outwitted and impoverished . . . through . . . plunder’ (Basu, Beyond the Invisible Hand: Groundwork for a New Economics (Princeton: Princeton University Press, 2011), 5).

184  Elizabeth Ashford avoidably causes physical deterioration or death—is not a genuine human right. This indicates that our traditional interactional conception of our negative duties of justice is too narrow. Clearly it is important to avoid conceiving our duties of justice in such a way as to arbitrarily preclude ex ante their applicability to harms that are the combined effect of the interaction of the behaviour of a vast number of agents. The danger of a moral blind-­spot is especially pervasive in this context, given that our moral phenomenology of agency is geared to small-­scale interactional social contexts, such that we tend to feel morally responsible only for harms for which we are individually responsible, and in particular for harms we directly inflict.19 We should therefore critically examine our paradigm interactional conception of our perfect negative duties of justice, and consider which features of this conception are essential to their special, peremptory moral force, and which features reflect our moral phenomenology of agency that is ill-­suited to our contemporary complex and large-­scale social context. We should, I contend, acknowledge, in addition to interactional negative duties, a shared duty of basic justice to structure our behaviour in such a way as to avoid, between us, depriving others of the objects of their fundamental human rights. Accordingly, we should take the right to subsistence to impose, in addition to interactional negative correlative duties, a shared duty of basic justice to structure our behaviour in such a way as to avoid, between us, plundering, jeopardizing, or destroying persons’ means of subsistence. As I now argue, while this shared duty significantly differs from our traditional paradigm interactional conception of our negative duties of justice, it has the features of such duties that ground their special, peremptory moral force.

3.3  The shared negative duty correlative to the right to subsistence By the term ‘shared duty’, I mean a duty that is held by a plurality of agents. The fundamental duty-­bearers are individual agents. However, unlike interactional duties, the agents are under the duty together: They are under the duty not in virtue of the foreseeable effects of their individual actions, but in virtue of the foreseeable combined effects of the practices, processes, and structures in which they participate; and the capacity that is relevant to the fulfilment of a shared duty is what the agents could achieve together, through restructuring their behaviour. The features of shared correlative duties can only be seen by shifting our focus away from the impact of our individual actions, and zooming out to consider the combined effect of the structures in which we participate, and of feasible 19  Samuel Scheffler, Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought (New York: Oxford University Press, 2002), 39.

Individualist Utilitarianism  185 alternative structures. These duties therefore significantly differ from our traditional paradigm interactional conception of our negative duties of justice. However, when we do zoom out, we can see that it has the features of such duties that ground their special, peremptory moral force. Let us consider, first, the urgency of the shared negative duty correlative to the right to subsistence, to structure our behaviour in such a way as to avoid, between us, plundering, jeopardizing, or destroying others’ access to the means of subsistence. Our paradigm interactional negative duties (not to torture and so on) are strict proscriptions on acts of a certain type, and if any agent fails to fulfil such a duty, on any one occasion, she will be individually responsible for a drastic harm to the victim. The special moral force of the duty is obvious to the agent, given the impact on our moral phenomenology of agency of harms that we individually and directly inflict. By contrast, the agents who, between them, fail to fulfil the shared negative duty correlative to the right to subsistence are not thereby individually causing drastic harm to anyone. Therefore, it is easy to fail to grasp the moral urgency of the shared duty; doing so requires transcending our moral phenomenology of agency. Nevertheless, the shared duty is at least as urgent as interactional duties. Ongoing shared failure to fulfil the duty predictably deprives a vast number of reasonably secure access to subsistence. Inadequate subsistence can cause permanent debility and physical deterioration or death, in just the same way as the most vicious assault, and causes drastically stunted lives. It can also prevent people from realizing their most fundamental goals and commitments, such as that of nurturing their children. Second, let us consider the specificity of the duty. The right to subsistence imposes a minimal moral constraint on our social organization: that however we structure our behaviour (our economies, laws, norms, and so on), we avoid plundering, jeopardizing, or destroying persons’ means of subsistence. There are indefinitely many schemas of requirements on agents that could fulfil this constraint. However, each of these schemas constitutes a different way of fulfilling the same overarching shared duty, the content of which is determinate, and given by the content of the right: to avoid depriving persons of the object of the right (that is, reasonably secure access to the means of subsistence). Whereas interactional duties are proscriptions on acts of a certain type, there is indeterminacy at the level of requirements on individual agents, in two respects. First, there are indefinitely many ways of implementing the shared duty. Second, prior to the implementation of structural reform, there is indeterminacy over what steps individual agents ought to take towards bringing it about. However, the content of the shared duty is both determinate and correlative. Third, the shared duty is enactable to all and owed to all. The duty can only be enacted to all through structural reform. Prior to such reform, individual agents cannot feasibly completely avoid participating in the harmful structures, and

186  Elizabeth Ashford neither can they individually achieve structural reform. Therefore the duty to avoid the harm is not completely enactable to all by each, considered seriatim. Rather, the right imposes on the rest of humanity a shared duty of basic justice to restructure our behaviour in such a way as to avoid, between us, depriving others of the means of subsistence. Fourth, the shared duty is rightfully enforced. This is essential to protecting right-­holders and empowering them to demand that their rights are respected. Conversely, it is essential to challenging the structures of power that threaten to deprive people of their most fundamental moral entitlements. In light of this conception of shared duties of basic justice, we can now return to the positive duties correlative to a general human right to subsistence.

3.4  The nature of the positive duty correlative to the right to subsistence O’Neill’s argument against a general positive right to the provision of the means of subsistence, as we have seen, is based on an interactional conception of the general positive duties that would correspond to such a right, according to which these duties must be enactable to all by each, seriatim; her argument appeals to an inference from the fact that these duties are not enactable to all by each, seriatim, to the conclusion that they are not enactable to all. However, this inference overlooks the moral significance of the respect in which such positive duties could be enacted to all, through an institutionally coordinated division of moral labour. Given the scale and complexity of the threat to persons’ interests posed by inadequate subsistence, avoiding this threat requires coordinated structural reform. But given the overall level of global resources, there are feasible and feasibly achievable alternative socio-­economic structures under which each person’s access to the basic necessities of life could be socially guaranteed to a reasonable degree of security—moreover, without immoderate cost to any duty-­bearer. The problem does not lie with a shortage of global resources, but with current social, political, economic, and legal structures that govern entitlements to them. (By the term ‘structure’ I simply mean any ongoing patterns of behaviour.) It would be arbitrary to take the scale and complexity of the threat to persons’ basic interests posed by inadequate subsistence, and of the measures needed to avert it, to mean that the general duties correlative to a general human right to subsistence, unlike duties of easy rescue, cannot be general perfect duties of justice owed to all. The general positive duty correlative to general positive human rights to the provision of certain goods and services has to be understood as a shared duty, held by the rest of humanity, to structure our behaviour in such a way as to provide each person with access to the means of subsistence, to a reasonable

Individualist Utilitarianism  187 degree of security. This shared duty does have the two features that O’Neill argues are lacked by general positive duties. First, it is enactable to all. Second, the content of the shared duty is determinate, and correlative to the right: to provide each person with the object of the right, which is reasonably secure access to the means of subsistence. While there are indefinitely many ways of specifying and allocating the shared duty, they each constitute a different way of realizing the same overarching shared duty. The duty is, moreover, of the utmost moral urgency. As I now contend, the Kantian rationale for taking general, positive interactional duties of easy rescue to be strict, perfect duties, also constitutes a rationale for acknowledging the shared positive duty correlative to the right to subsistence.

3.5  A Kantian rationale for general positive duties of justice correlative to the right to subsistence There is a widely held intuition, affirmed by several Kantian moral theorists, that in contexts of easy rescue, agents are under general, perfect, positive requirements.20 The general positive duty of beneficence is a duty to adopt the maxim of taking others’ happiness as an end, which generally allows latitude over what and how much to do to fulfil it and on what occasions (by contrast with perfect duties, which are determinate proscriptions on acts of a certain type that strictly bind agents’ conduct at all times, and each agent, seriatim, can and should completely enact to all). Nevertheless, there are certain circumstances in which failing to help a particular individual would be fundamentally incompatible with genuinely adopting the maxim of taking others’ happiness as an end. The failure to perform an easy rescue is one such case.21 Missing out on a particular occasion to promote someone’s happiness by helping them in the pursuit of one of their ends is perfectly compatible with genuinely adopting the maxim of taking others’ happiness as an end. Failure to help the person would not, in Barbara Herman’s phrase, ‘gravely diminish her life’.22 For this reason, as Thomas Hill argues, ‘Kant . . . thought that judgment guided by the Categorical Imperative would leave substantial room, in non-­emergency situations, for moral agents to pursue their own happiness’.23 20 Thomas  E.  Hill, ‘Meeting Needs and Doing Favours’, in Human Welfare and Moral Worth (Oxford: Clarendon Press, 2002), 201–43; Barbara Herman, ‘Mutual Aid and Respect for Persons’, in The Practice of Moral Judgement (Cambridge, MA: Harvard University Press, 1993); Karen Stohr, ‘Kantian Beneficence and the Problem of Obligatory Aid’, Journal of Moral Philosophy 8 (2011): 45–67. 21  As Hill argues, ‘A proper regard for the humanity of each person as an end would leave us the least latitude in cases where sacrifices of our minor pleasures can satisfy others’ basic needs’ (‘Meeting Needs and Doing Favours’, 213). 22  Herman, ‘Mutual Aid and Respect for Persons’, 71. 23  Hill, ‘Meeting Needs and Doing Favors’, 201; my italics.

188  Elizabeth Ashford By contrast, in an easy rescue emergency situation, the person in need of rescue stands to die or suffer some drastic, irrevocable harm unless she is helped. Given that the agent can prevent that harm at small cost, failing to do so would show a grossly inadequate regard for the moral importance of that person’s happiness and for her moral status as an end-­setter. Easy rescue cases therefore demonstrate that there are circumstances in which general positive duties can be urgent enough to constitute strict, perfect duties, owed to all simply in virtue of the moral significance of each person’s happiness. The rationale for taking agents to be under a perfect duty of easy rescue, I contend, also constitutes a rationale for taking affluent agents to be under a shared, general perfect positive duty to implement the structural reforms needed to ensure that each person has reasonably secure access to the means of subsistence. In just the same way as with easy rescue emergency scenarios, inadequate subsistence poses a drastic threat to persons’ most basic interests, liable to blight or altogether destroy their lives. And given the overall level of global resources, this threat could be prevented at small cost to the individual duty-­bearer, through structural reform. In both cases, therefore, the failure to avert the threat is incompatible with a minimally adequate recognition of the moral value of the lives that are marred or altogether destroyed as a result. I contend, then, that Kantianism as well as individualist utilitarianism should acknowledge a shared general positive duty to implement a coordinated global schema of duties, compliance with which would provide each person with reasonably secure access to the means of subsistence. It is not plausible to take general positive duties to provide persons with the means of subsistence to be different in kind from duties of easy rescue, such that the former are imperfect duties of beneficence and the latter are strict, perfect duties. Rather, the rationale for general perfect duties of easy rescue also constitutes a rationale for acknowledging a shared general positive duty, owed to all by the rest of humanity, to structure our behaviour in such a way as to make sure that, between us, we provide each person with the means of subsistence. Both the general interactional duty of easy rescue, and the shared general positive duty correlative to the right to subsistence, are strict, urgent duties that are enactable to all and owed to all. I conclude that both Kantianism and utilitarianism converge on taking the right to subsistence as imposing on the rest of humanity a shared positive and negative correlative duty of basic justice to implement a coordinated global schema of requirements, compliance with which would provide each person with, and avoid depriving anyone of, reasonably secure access to the means of subsistence. O’Neill is right to point out that there is a deep inconsistency in affirming a fundamental general human right, without acknowledging concrete and legally binding obligations, owed to all, compliance with which would provide each person with, and avoid depriving them of the object of the right. The failure to acknowledge such obligations does indeed rob the right of claimability and

Individualist Utilitarianism  189 actionable teeth for many right-­holders. It fails to empower many right-­holders to demand that their right is respected, and, conversely, fails to challenge the structures of power that threaten to deprive a vast number of the object of their right. Laws and norms that fail to classify as a violation or prohibit much of the behaviour that causes or allows subsistence deprivations are indeed incompatible with genuine recognition of the right to subsistence; they pay mere lip-­service to it. But what we should infer, I suggest, is that such laws and norms are fundamentally unjust, insofar as they fail to adequately recognize what is a fundamental human right. Ongoing failure to fulfil the shared duties correlative to the right to subsistence should, I suggest, be classified as a structural violation, which presents a fundamental challenge to the moral legitimacy of prevalent norms and laws themselves. As I now argue, both individualist utilitarianism and Kantianism should acknowledge this structural violation.

3.6  Structural violations of the right to subsistence A structural violation, I suggest, consists in ongoing patterns of behaviour of a large number of agents that predictably and avoidably deprive a large number of victims of the object of a human right. Accordingly, structural violations of the right to subsistence consist in ongoing patterns of behaviour that predictably and avoidably deprive a vast number of reasonably secure access to the means of subsistence. The respect in which the deprivations are avoidable is that they could be avoided under feasible and feasibly achievable alternative social, economic, political and legal structures. The notion of a structural violation brings moral and juridical scrutiny to bear on the combined effects of agents’ ongoing patterns of behaviour, and on what could be made possible under alternative structures. At the core of affirmations of universal human rights is the claim that each person has moral status, and can therefore justifiably demand not to be treated in ways that are fundamentally incompatible with recognition of that moral status. The notion of a human rights violation demarcates treatment of persons that is fundamentally incompatible with such recognition. On both a Kantian and an individualist utilitarian account of persons’ moral status, our ongoing shared failure to implement the structural reforms needed to provide each person with, and avoid depriving them of, reasonably secure access to the means of subsistence is incompatible with minimally adequate recognition of that status. Utilitarianism and Kantianism are historically the most important versions of the two main accounts of the role of fundamental human rights, instrumentalist accounts and status-­based accounts. Instrumentalist accounts take their role to be that of protecting persons’ basic interests against standard threats. Status-­based accounts take rights to be expressive of persons’ moral status and inviolability.

190  Elizabeth Ashford Let us consider, first, their instrumental role, of protecting the vulnerable against standard threats to their basic interests. The structures that predictably and avoidably result in widespread subsistence deprivations constitute an actual and ongoing threat to the basic interests of a vast number. Many of those who lack secure access to the means of subsistence are from the outset precluded from any realistic chance of a minimally decent life (indeed, from before they were born in the case of persons whose mothers suffered severe malnutrition during pregnancy, resulting in cognitive deficits in the developing foetus). And their vulnerability is compounded by the danger that the devastating harm to which they are vulnerable may make little or no impact on the moral psychology of the agents who, together, allow or inflict it, because of the complexity of the causal chains involved. The role of human rights in protecting the vulnerable against treatment liable to unjustifiably blight or destroy their lives, by institutionalizing extraordinary strong protection against such treatment, is especially important in this context. Moreover, this harm could be avoided under alternative structures without imposing immoderate cost on any duty-­bearers, and there are no countervailing moral considerations that might justify the infliction of the harm. Therefore, the ongoing shared failure to implement this structural reform amounts to the unjustifiable infliction of ruinous harm to a vast number. Clearly, minimally adequate recognition of the moral value of each person’s life demands fulfilment of the shared duty. Let us turn to the role of human rights of expressing and affirming persons’ universal moral status. While structural violations of the right to subsistence do not involve the intentional targeting or degradation of certain individuals, the lives of those who end up deprived of the means of subsistence effectively do not count. I suggest that one of the most pervasive threats to human dignity in our contemporary social context is the threat of being invisible to, and unnoticed by, the agents who are together, unwittingly, allowing or inflicting devastating harm. I conclude that both individualist utilitarianism and Kantianism should acknowledge ongoing structural violations of the right to subsistence, which urgently demand abolition. This structural violation instantiates a trade-­off (both passive and active) that is utterly gratuitous, and incompatible with a minimally adequate recognition of the moral value of the lives that are blighted or destroyed by subsistence deprivations. Moreover, as I now argue, Kantianism as well as individualist utilitarianism should take the right to subsistence to be a basic human right, the enjoyment of which is necessary to the adequate enjoyment of any other right. This has central implications for another core feature of a Kantian conception of perfect duties: that they are exceptionless. As we have seen, this is central to the debate between Kantianism and utilitarianism over the nature of duties of justice. Whereas Kantianism takes rights to impose strict, exceptionless negative duties,

Individualist Utilitarianism  191 utilitarianism holds that an agent might be required to violate one person’s rights in order to fulfil the positive duty to prevent more serious or numerous violations. The debate thus standardly focuses on the question of interpersonal trade-­offs. I now shift the focus to the issue of intrapersonal trade-­offs between the interests of the individual right-­holder, which, I argue, Kantianism as well as utilitarianism should acknowledge, since not doing so fails to give adequate weight to the overall interests and autonomous choices of the right-­holder. As I now argue, if the right to subsistence is not fulfilled, then intrapersonal trade-­offs between the interests of the individual right-­holder cannot be avoided; and in virtue of these intrapersonal trade-­offs, other human rights cannot be adequately enforced, or in some cases even respected, unless the right to subsistence is fulfilled.

3.7  The right to subsistence as a basic right The interest in subsistence is so important that, first, lacking it is completely unsustainable, and, second, the interest in subsistence is liable to outweigh the interest in the object of any other right, whenever the two interests come into conflict. For those who lack secure access to the means of subsistence, the status quo is unsustainable. They are desperate for any opportunity to obtain the means of subsistence, even if this means forgoing another right; treatment that is severely harmful, relative to an uncontroversial moral baseline, is liable to be less bad for them than the only available alternative, of being unable to obtain the means of subsistence. For example, sweatshop labour and child labour are widely considered to be violations of negative human rights against harmful treatment. A common response is to simply seek to enforce a ban on them, without addressing the background severe poverty. But an implication of the intrapersonal trade-­offs that drive persons to sweatshop labour and child labour is that simply enforcing a ban, without providing other realistic opportunities to obtain the means of subsistence, is liable to be severely detrimental to the overall interests of the individual right-­ holders. It also reduces even further their range of options. Even if consider the right against torture, we can well imagine, say, a mother agreeing to such treatment in exchange for enough money to prevent her children from dying of malnutrition, since she reasonably judges the mental agony of watching her children preventably die, and the marring of her life this would cause, to be worse than the physical agony and lasting psychological damage inflicted by torture. (Many of the women who are abused in the making of films for the most violent end of the pornography industry are extremely poor prostitutes, often from Latin America.) If the mother were legally prevented from accepting such treatment in exchange for a subsistence income, and her children died as a result, she might

192  Elizabeth Ashford reasonably complain that the legislators had worsened her suffering and degradation. She might also reasonably complain that they could not claim to have enforced her right against the treatment out of respect for her dignity as a rational autonomous agent, given that they had deprived her of the only available opportunity to realize her most important end, of nurturing her children. This indicates that these other human rights cannot be adequately fulfilled without addressing the background ongoing structural violation of the right to subsistence, which poses an ongoing threat to persons’ most basic interests, and to their most central ends. The impact on children of immediate bans on child labour, without addressing the background destitution, has been well documented.24 It has often led to the children’s starvation, or to their being driven to a worse way of earning a subsistence income, such as prostitution. Child labour has a severely detrimental impact on children’s developmental interest, and is moreover likely to leave them trapped in poverty, since they are deprived of the opportunity to have an education. But unless the right to subsistence has been fulfilled, the only available alternatives are liable to be even worse. In some cases, even the duty to respect the right against child labour may be outweighed by the children’s interest in subsistence. The right against child labour is often understood as imposing perfect duties, that is, strict requirements not to send one’s children out to work, and not to employ child labourers, that can and should always be completely fulfilled. However, if parents are unable to earn enough to provide their family with subsistence, they may face the terrible choice between supplementing their income by sending their children out to work as child labourers, or being unable to prevent their children’s physical deterioration through malnutrition. Malnutrition is likely to have an even more devastating impact on their children’s developmental interests than child labour. I contend, then, that Kantianism as well as utilitarianism should acknowledge that when the right to subsistence is not fulfilled, then individual agents on the ground may face awful intrapersonal trade-­offs, which mean that their duty to enforce or even respect certain rights may be outweighed even under standard circumstances. But both theories take the central moral question this raises to be why unjust background structures are in place that lead agents to have to face these trade-­offs. As I have argued, both theories take the intrapersonal trade-­offs to arise because of ongoing structural violations of the right to subsistence that instantiate a trade-­off that is utterly gratuitous, and incompatible with minimally adequate recognition 24  See e.g. Kaushik Basu, ‘Child Labour: Cause, Consequence and Cure, with Remarks on International Labor Standards’, Journal of Economic Literature 37 (1999): 1083–119; C.  Grootaerd and Ravi Kanbur, ‘Child Labour: an Economic Perspective’, International Labour Review 134 (1995): 187–203.

Individualist Utilitarianism  193 of the moral value of the lives that are blighted or altogether destroyed as a predictable result. Thus, both theories should acknowledge intrapersonal trade-­offs in virtue of which negative human rights cannot be adequately enforced or even respected even under standard circumstances, and both theories highlight the moral urgency of ending the structural violations that lead to such trade-­offs.

Bibliography Ashford, Elizabeth, ‘The Alleged Dichotomy Between Positive and Negative Rights and Duties’, in Charles Beitz and Robert Goodin (eds.), Global Basic Rights (Oxford: Oxford University Press, 2009). Ashford, Elizabeth, ‘The Nature of Violations of the Human Right to Subsistence’, in Adam Etinson (ed.), Human Rights: Moral or Political? (Oxford: Oxford University Press, 2018). Basu, Kaushik, Beyond the Invisible Hand: Groundwork for a New Economics (Princeton: Princeton University Press, 2011). Basu, Kaushik, ‘Child Labour: Cause, Consequence and Cure, with Remarks on International Labor Standards,’ Journal of Economic Literature 37 (1999): 1083–1119. Björnnsson, Gunnar, ‘Essentially Shared Obligations’, Midwest Studies in Philosophy 38 (2014), 103–20. Brink, David, ‘Mill’s Deliberative Utilitarianism’, Philosophy & Public Affairs 21 (1992): 67–103. Broome, John, Weighing Goods (Oxford: Wiley-Blackwell, 1995). Crisp, Roger, ‘Equality, Priority, and Compassion’, Ethics 113/4 (2003): 745–63. Darwall, Stephen, Impartial Reason (Ithaca, NY: Cornell University Press, 1983). Draper, Kai, ‘Personal and Impersonal Dimensions of Benevolence’, Noûs 36/2 (2002): 201–27. Farmer, Paul, ‘ “Landmine Boy” and Stupid Deaths’, in Haun Saussey (ed.), Partner to the Poor: A Paul Farmer Reader (Berkeley and Los Angeles: University of California Press, 2010). Frankfurt, Harry, ‘Equality and Respect’, Social Research 64/1 (1997): 3–15. Griffin, James, Well-Being: Its Meaning, Measurement and Moral Importance (Oxford: Clarendon Press, 1986). Griffin, James, On Human Rights (Oxford: Oxford University Press, 2008). Grootaerd, C., and Kanbur, Ravi, ‘Child Labour: an Economic Perspective’, International Labour Review 134 (1995): 187–203. Herman, Barbara, The Practice of Moral Judgement (Cambridge, MA: Harvard University Press, 1993).

194  Elizabeth Ashford Hill, Thomas E., Human Welfare and Moral Worth (Oxford: Clarendon Press, 2002). Hope, Simon, ‘Kantian Imperfect Duties and Modern Debates over Human Rights’, Journal of Political Philosophy 22 (2014): 396–415. Kamm, Frances, Morality, Mortality, ii. Rights, Duties, and Status (New York: Oxford University Press, 2001). Narveson, Jan, ‘Utilitarianism and New Generations’, Mind 76/301 (1967): 62–72. Nozick, Robert, Anarchy, State, and Utopia (New York: Basic Books, 1977). O’Neill, Onora, Faces of Hunger: An Essay on Poverty, Justice and Development (London: Allen & Unwin, 1986). O’Neill, Onora, Bounds of Justice (Cambridge: Cambridge University Press, 2000). Parfit, Derek, Reasons and Persons (Oxford: Clarendon Press, 1984). Rawls, John, A Theory of Justice (rev. edn., Cambridge, MA: Belknap Press, 1999). Scanlon, T. M., What We Owe To Each Other (Cambridge, MA: Belknap Press, 1998). Scheffler, Samuel, Boundaries and Allegiances: Problems of Justice and Responsibility in Liberal Thought (New York: Oxford University Press, 2002). Shue, Henry, Basic Rights: Subsistence, Affluence, and U.S. Foreign Policy (Princeton: Princeton University Press, 1996). Stohr, Karen, ‘Kantian Beneficence and the Problem of Obligatory Aid’, Journal of Moral Philosophy 8 (2011): 45–67. Wood, Allen,‘Humanity as End in Itself ’, in Derek Parfit, On What Matters (Oxford: Oxford University Press, 2011), ii. 58–82.

8

Parfit’s Reorientation From Revisionism to Conciliationalism Ingmar Persson

1. Introduction A few years after the publication of Reasons and Persons (RP), Derek Parfit said to me as we walked along the High Street in Oxford: ‘I don’t want to become like Hare.’ He went on to explain that what he had done in his book was to present various arguments rather than to defend any particular ethical positions. By contrast, what Richard Hare was known for was precisely his vigorous defence of two ethical positions: universal prescriptivism in metaethics, which he argued entails a form of rule-­consequentialism in normative ethics. The normative framework of Reasons and Persons is overall, if not utilitarian, at least consequentialist, and already at the time of writing this book Parfit’s favoured metaethical view was a non-­naturalist theory of reasons, according to which normative reasons are irreducible to natural facts. But he was anxious to stress that many of his arguments did not presuppose this objectivist theory as opposed to subjectivist theories of reasons. And in this book he did not defend consequentialism, though he is obviously most at home in this tradition, a tradition that he inherited from Henry Sidgwick, just as he inherited non-­ naturalism about reasons from him. Thus, at the time he wrote Reasons and Persons, Parfit’s way of doing moral philosophy differed from Hare’s in that he did not try to argue for any particular normative position on the basis of a particular metaethical position. He did not take what he called the ‘High Road’ (RP, 447). At that time, Hare had not yet argued in print that Kant could consistently have been a utilitarian, but he was to do so in the paper ‘Could Kant Have been a  Utilitarian?’ (1997). There is an obvious similarity between Hare’s idea of the  universalizability of moral judgments and Kant’s first formulation of the ­categorical imperative, which requires that we act only on maxims that we could simultaneously will to be universal laws (see Hare 1997: 153–4, 161). So, Kant

I am most grateful to Jeff McMahan for careful and penetrating comments on an earlier draft of this paper. Ingmar Persson, Parfit’s Reorientation: From Revisionism to Conciliationalism In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0009

196  Ingmar Persson eventually came to occupy a more prominent place in Hare’s moral philosophy— just as he did in Parfit’s moral philosophy. Whilst Kant is barely mentioned in Reasons and Persons, he undeniably looms large in On What Matters (OWM). But the only positive claim Parfit succeeds in extracting from Kant seems to come from his first formulation of the categorical imperative, which was what Hare also thought important and, in some form, defensible. The formula ‘everyone ought to follow the principles whose universal acceptance everyone could rationally will’ is the core claim of what Parfit calls Kantian contractualism. A main objective of the first two volumes of On What Matters is to show that Kantian contractualism could be aligned not only with rule-­ consequentialism, as Hare thought, but with a Scanlonian form of contractualism as well, making up what he calls the ‘Triple Theory’. This normative unification project is carried further in the third volume of On What Matters, in which Parfit also argues (ch. 58) that his Triple Theory supports the acceptance of an improved version of common-­sense morality. This, too, is in line with Hare’s ‘two-­ level’ moral theory in which common-­ sense morality corresponds to an ‘intuitive’ level that is underpinned by a consequentialist ‘critical’ level (Hare 1981). Moreover, Parfit’s normative unification project, like Hare’s, is based on a certain metaethical position, albeit of a non-­naturalist kind, which is diametrically opposed to Hare’s prescriptivism. Nevertheless, for both of them it was central that there are moral judgements that are objective to the extent that all rational subjects in possession of all relevant empirical facts would agree about them (see Hare 1997: 134; and Section 3 for Parfit). Consequently, Parfit ended up with a way of doing moral philosophy which parallels Hare’s to a much greater extent than one could have anticipated thirty years ago. They were both hoping to show that there is a ‘single true morality’ (OWM ii. 155), which could take the shape of rule-­consequentialism. This is of course compatible with there being huge differences in both their procedure and conclusions. And their personalities—well, they were as different as their hairstyles. In the introduction to Reasons and Persons, Parfit refers to Peter Strawson’s ­distinction between ‘two kinds of philosophy, descriptive and revisionary’ (RP, p. x). (Actually, Strawson’s distinction is between two types of ‘metaphysics’.) Parfit describes himself as a revisionist ‘by temperament’, and the tenor of Reasons and Persons is indeed revisionist. In these terms, On What Matters is by contrast ‘descriptive’ or conservative in matters of normative substance. Its aim is not so much to challenge commonsensical moral claims as to defend them by showing that, with a bit of revision, they can be supported by leading ethical theories once  we get the metaethics right. His fear is that ‘if we cannot resolve our ­disagreements . . . morality might be an illusion’ (OWM ii. 155). Parfit is celebrated for developing revisionist views about personal identity in Reasons and Persons, and there is evidence that he never repudiated this

Parfit ’ s Reorientation  197 revisionism. But it receded into the background in On What Matters, as I shall try to demonstrate in Section 2. The reason why this revisionism receded is probably that it is in tension with the ‘conciliationalist’ project of this book, as we shall see. Of course, revisionism and conciliationalism are not inconsistent; it is just that a combination of them is unlikely to succeed, since revisionist views are likely to stir up disagreement. In Section  3 I shall critically review his extension of this conciliationalist project into metaethics, and in Section 4 his attempt to reconcile parts of common-­ sense morality and consequentialism to buttress the Triple Theory.

2.  The Omission of Reductionism about Personal Identity in On What Matters In On What Matters (vol. i, sect. 19) Parfit develops a version of Sidgwick’s ‘dualism of practical reason’ which consists in there being two kinds of reasons to care about the well-­being of individuals: self-­interested or personal reasons to care about our own well-­being, and impartial reasons to care about everyone else’s well-­being.1 He concedes to Sidgwick that he ‘rightly claims that we have reasons to be specially concerned about our own future well-­being’. But, Parfit continues, ‘of our reasons to care about our future, many are provided, not by the fact that this future will be ours, but by various psychological relations between ourselves as we are now and our future selves’ (OWM i. 136). This suggests that some of our reasons to care about our future are provided by the fact that it is ours. Similarly, he writes that Sidgwick ‘overstates’ the rational importance of personal identity (OWM i. 136), which again suggests that it has some rational importance. These statements suggest a denial or modification of his famous claim in part III of Reasons and Persons that ‘personal identity is not what matters’. It is not that he has abandoned this claim about personal identity, however, for in his paper ‘We are not Human Beings’, published the year after On What Matters, volumes i and ii, he reaffirms his allegiance to it. After having argued that the animalist or biological view of our identity—according to which are identical to our human organisms—is not true, he confesses that he has ‘a reason to wish that Animalism were true’ (Parfit 2012: 27), since this would make it easier for him to vindicate his claim that our identity does not matter. Contradicting what is suggested in On What Matters he writes that ‘though we  have reasons for special concern about our future, these reasons are not given . . . by the fact that this will be our future’ (Parfit 2012: 27). So, he 1  Parfit writes ‘everyone’s well-­being’ (OWM i. 136)—which would include our own well-­being— but it is doubtful that this is consistent with his claim (see below) that we are permitted to give ‘even greater weight to some stranger’s well-­being’ (i. 139) than to our own.

198  Ingmar Persson apparently exhibits some double-­thinking or ambivalence about the importance of personal identity. As regards the psychological relations that provide us with reasons to care about ourselves at other times, in particular in the future, we have ‘partly similar relations to some other people, such as our close relatives, and those we love’ (OWM i. 136). Thus, these relations provide us with ‘personal and partial reasons to care about the well-­being of ourselves and those to whom we have close ties’ (OWM i. 136). These reasons are ‘only very imprecisely comparable’ (OWM i. 137) to impartial reasons. They are only very imprecisely comparable in the sense that, though we can tell, for instance, that one is permitted to save one’s own life rather than the lives of at least two strangers, we cannot give anything like a precise answer to how many strangers one is allowed to sacrifice to save oneself. According to Parfit, this imprecision is due to the fact that, whereas impartial reasons are person-­neutral, self-­interested and partial reasons are person-­relative in the sense that they ‘are provided by facts whose description must refer to us’ (OWM i. 138), either because these facts concern our own well-­being, or the well-­ being of people to whom we have close ties. Although self-­interested reasons permit us to give somewhat greater weight to  our own well-­being than to that of strangers, Parfit thinks—surely rightly— that it would be ‘too egoistic’ (OWM i. 139) to maintain that they require us to give greater weight to our own well-­being: we are permitted to ‘give equal or even greater weight to some stranger’s well-­being’ (OWM i. 139) than to our own. However, as he notes, there is in this respect a difference between our reasons to care about our own well-­being and our reasons to care about the well-­being of others to whom we have close ties. For in a case in which I could save either my own child or the child of some stranger ‘I ought morally to give priority to my child’ (OWM i. 141).2 This difference seems sufficient for holding that our self-­interested reasons to care about our own well-­being are a different kind of reason from partial reasons to care about the well-­being of others to whom we have close ties. While we seem to have especially strong moral reasons—reasons that are stronger than impartial reasons—to care about the well-­being of others to whom we have close ties, our reasons to care about our own well-­being cannot likewise be especially strong moral reasons. If they were, we could not be morally permitted to sacrifice our

2  But even though we are not allowed to sacrifice a close friend to save a stranger, we are apparently allowed to sacrifice the friend to save ourselves and to sacrifice ourselves to save the stranger, which would yield the same outcome. For instance, when chased by a hungry lion, we are allowed to outrun a friend, foreseeing that this will mean that the friend is killed by the lion, even though we follow up this by choosing not to outrun a stranger who is chased by another hungry lion. However, we are hardly allowed by common-­sense morality to save ourselves, e.g. by pushing our friend into the paws of the lion.

Parfit ’ s Reorientation  199 own greater well-­being for the sake of a smaller benefit to a stranger.3 So, keeping apart self-­interested and partial reasons seems called for. Thus, whereas Parfit lumps together self-­interested and partial reasons as person-­relative reasons and talks about two kinds of reasons (e.g. OWM i. 138), a case can be made for distinguishing between three different categories of reasons of beneficence: self-­ interested, partial, and impartial reasons. I shall now try to show that both the claim that self-­interested and partial reasons are based on the distinction between oneself and others and the claim that both of them are only very imprecisely comparable to impartial reasons are in conflict with his reductionism about personal identity. According to this reductionism, our identity consists in the holding of psychological and/or physical relations that can hold to a greater or lesser degree, and there are cases in which it is indeterminate whether or not they hold to such a degree that it can truly be said that we persist. This is a claim about the analysis of our identity. But his reductionism also features a claim about its importance, expressed by the slogan that ‘personal identity is not what matters’.4 What matters is rather ‘psychological connectedness and/or psychological continuity, with the right kind of cause’ (RP, 214). The right kind of cause is normally the persistence of one and the same brain. Psychological connectedness consists in the holding of psychological connections like persisting memories and desires, while psychological continuity consists in the existence of a sufficient number of such connections, or ‘chains’ of such connections, which could stretch over periods of a person’s life between which there are few or none psychological connections. But a substantial reduction of psychological connectedness which is abrupt could go along with such a marked reduction of the number of chains of connectedness making up psychological continuity that it could be either indeterminate whether or not the same person persists, or definitely false. To illustrate such changes, we might need to have recourse to science fiction examples, like those in which a nefarious neurosurgeon manipulates somebody’s brain by replacing major portions of their memories and desires with different memories and desires. 3  Jeff McMahan (in correspondence) suggests that we could have rights against ourselves that protect our own future well-­being, but rights that are special in that we have the power to waive them. However, the other side of rights is duties, and duties that we can release ourselves from whenever it pleases us are not really duties that can bind us as duties must do. Therefore, the other side of them would be ‘rights’ against ourselves that would not offer us any protection against ourselves and, thus, would not be genuine rights. 4  Parfit adds that reductionists also claim that ‘we could give a complete description of reality without claiming that persons exist’ (RP, 212), but this is not necessary to show that personal identity consists in something that does not matter. As remarked above, this would be easier to show if animalism were true. But animalists take human organisms to be persons, i.e. the referents of personal pronouns, and reference to our organisms need not be seen as eliminable in order for reductionism to be true, since this reference does not imply that there is anything over and above organisms and their psychology (cf. Persson 2005: 256–7).

200  Ingmar Persson But even if there is enough psychological connectedness to constitute psychological continuity, there might not be identity. Parfit maintains that this could in principle happen if the two hemispheres of a brain each are capable of sustaining the psychology of a person, and they are separated and successfully transplanted into two different bodies (RP, ch. 12). In such cases of branching psychological continuity, personal identity is disrupted, since the original person cannot be identical to both of the resulting persons, who are clearly distinct from each other, and it would arbitrary to identify the original person with either one of them. Accordingly, personal identity consists in non-­branching psychological continuity, with the right kind of cause. But Parfit maintains that the occurrence of such a division is not worse for us than survival as the same person with the same degree of psychological connectedness and continuity. Thus, it is the latter relations that matter for us rather than personal identity. If this is correct, it raises the question how it could be morally permissible to sacrifice oneself for a smaller benefit to some stranger, but not to sacrifice someone else to whom one is closely related. For surely it must be morally permissible to sacrifice the people who come into being when one’s psychological continuity branches, just as it is permissible to sacrifice oneself, even though they are people distinct from oneself to whom one is very closely related. The relations that matter to us are the same to the ‘branches’ as when identity is preserved for, according to Parfit, branching is as good as ordinary survival. Thus, the alleged difference between self-­interested and partial reasons is undercut. Consider now the proposed imprecise comparability between these person-­ relative reasons and person-­neutral or impartial reasons. Imagine a spectrum of cases in which psychological continuity is increasingly attenuated, by successively greater parts of the underlying brain being replaced by parts supporting different memories and so on. As the amount of psychological connectedness involved in continuity is gradually reduced, we first enter a zone in which it is indeterminate whether there is still identity, then a zone in which there is definitely not identity, but psychological connectedness is still strong enough for the non-­identical person being close to us, and then a zone in which psychological connectedness is so weak that the non-­identical person is not even close to us. But it would seem that the difference in strength between a case (1) in which there is definitely enough (non-­ branching) psychological connectedness for the continuity of identity and a case (2) in which there is too little for identity, and barely enough for someone being close to us could be as great as between the case (2) and a case (3) in which psychological connectedness is just about so weak that there would somebody else who is definitely not close to us. If so, and reasons for concern are based on psychological connectedness and/or continuity, it might be wondered why their comparability must be less precise in the latter comparison between (2) and (3) when person-­relative reasons are compared to impartial reasons than in

Parfit ’ s Reorientation  201 the former comparison between (1) and (2) when only person-­relative reasons are involved. Suppose, however, that we instead adopt the non-­reductionist view that Parfit in Reasons and Persons attributes to common sense and Sidgwick, according to which the difference between ourselves and others involves ‘a further fact’ beyond psychophysical continuities, a fact that is either/or rather than a matter of degree (RP, 138–9, 329). Then it appears more comprehensible how self-­interested and partial reasons, on the one hand, and impartial reasons, on the other, could be thought to be only very imprecisely comparable, given that the former are person-­ relative in the sense that they are provided at least partly by facts that refer to what is irreducibly ourselves. However, as already observed, Parfit’s reductionist approach to our identity is difficult to reconcile not only with the alleged imprecise comparability between person-­relative and impartial reasons, but also with the claim that we are morally permitted to ‘give equal or even greater weight to some stranger’s well-­being’ than to our own well-­being, but not than to the well-­being of someone else to whom we have close ties. For it seems that one is morally permitted to treat the two people who result when one’s psychological continuity branches, who are not oneself, but closely related to oneself, just as if they were oneself. To compound this difficulty, Parfit’s reductionist campaign in Reasons and Persons also encompasses a proposal to extend morality into the intrapersonal sphere which implies that we are not morally permitted to treat ourselves differently from others. He considers ‘a boy who starts to smoke, knowing and hardly caring that this may cause him to suffer greatly fifty years later’ (RP, 319–20). In such cases in which there is a considerable loss of psychological connectedness, but enough for there to be chains of it requisite for psychological continuity to preserve identity, he proposes that such great imprudence may be prohibited by importing moral reasons into the intrapersonal sphere, with the result that ‘we ought not to do to our future selves what it would be wrong to do to others’ (RP, 320). But this is incompatible with the claim in On What Matters that we are morally permitted to give ‘even greater weight to some stranger’s well-­ being’ than our own. This latter claim permits us to agree to suffer greater future pain to prevent a stranger from suffering lesser pain, even though we predict that our future self will resist this, and even though we are obviously not permitted to  make one stranger suffer greater pain to prevent another from suffering lesser pain. However, it cannot be right that prudential matters fall outside the province of morality simply because they concern the self rather than somebody else. For, as noted, we would naturally regard ourselves as permitted to deal with each of two branches of ourselves as we deal with ourselves, although they are not identical to us. The reason is, I hypothesize, that we are likely to anticipate—and, as a result,

202  Ingmar Persson sympathize with—the pain that any of the branches will feel in the near future as we anticipate the pain we ourselves will feel in the near future. I am inclined to think that such anticipation involves projecting into the future the continuity of consciousness that we experience when our consciousness flows on without interruption. On the basis of remembering this continuous flow of our consciousness in the past, we imagine it running on until it contains the future pain. Our imagination then approaches the future pain ‘from the inside’, by having our consciousness so to speak furrowing a tunnel into the future. I submit that just as this imagining does not require that we believe that our consciousness that will run on without interruption (by dreamless sleep, say), so it also does not require that we believe that it will not divide. Still, it is conducive to identification with the target, since it is the imaginative approach we customarily adopt to ourselves in the future.5 By contrast, when we imagine feeling the pain that we believe that someone else before our eyes is feeling, our imagining this is elicited by our observing outward signs of pain, such as writhing, grimacing, and screaming. This is also the method that one of the branches would naturally employ to imagine a pain that the other branch is supposed to feel and, on this basis, sympathizing with it. My conjecture is, then, that we should take our relationship to the well-­being of someone distinct from ourselves to be moral if our imagination reaches out to their well-­being from the outside and prudential if it reaches out to it from the inside. The outside approach faces more hurdles. Others may have overt properties that make it hard to imagine what they feel and sympathize with it: they may be physically repulsive, dirty, and bad-­smelling, or belong to a race or species markedly different from our own. We must then make a voluntary effort to imagine as vividly what it is like to be these individuals as we would spontaneously make were we ourselves to be at the receiving end, and ask ourselves whether we could then accept such unsympathetic treatment. When we approach the well-­being of individuals from the outside, there is also space for moral relations concerning justice and rights to develop because, if individuals exist simultaneously with us, it is possible for us to act on each other and create conflicts. Conflicts are likely because their wills, or capacities for making decisions, are more independent of ours than those of our future selves or our ‘successors’, such as branch-­persons. This creates a need for appeals to justice and rights. For instance, if others harm us, we have reason to regard them as deserving harm in return. Or we can place ourselves under obligations to treat them in ways they want to be treated which provide them with rights to be treated in these ways.6 Reasons of justice and rights can be either partial, such as reasons 5  I discuss this sort of anticipation more fully in Persson (2005: 314–20). 6  These rights cannot be waived by ourselves in contrast to the alleged rights considered in n. 3.

Parfit ’ s Reorientation  203 deriving from the special rights of those to whom we have made promises, or impartial, such as those deriving from rights that everyone has against us. We do not have such reasons with respect to our selves and their successors because their wills lack the requisite independence. The approach to someone’s well-­being from the inside could malfunction, too, with the result that people fail to satisfy the conditions of autonomy, of being free, well informed, and rational. But it is hard for us to be sure that we know better and are more concerned about what is good for others than these people themselves, unless they exhibit ‘great imprudence’. Then paternalistic restraint of them is morally justified. In other words, the domain of prudence is sealed off from moral intrusion only within the lives of autonomous people who arguably know better than others what is best for themselves. Such intrusion by other people is, however, the only sort of intrusion morality could make into the domain of prudence. I have sketched a ‘dualism’ of reasons to care about well-­being as viewed from the inside and outside. This dualism is based not on the boundary between ourselves and others, but on the boundary between, on the one hand, ourselves and our successors and, on the other hand, others who do not ‘succeed’ us, but rather ‘coexist’ along with us. Our successors need not replace us, but they spring from us in some fashion that makes them closely resemble us. Our ‘coexisters’ may also spring from us—like our children—and they may resemble us closely, like our monozygotic twins, but then it is not due to their manner of springing from us. In everyday life, other individuals we come across are coexisters. This dualism is consistent with Parfit’s reductionism about personal identity in Reasons and Persons, but it does not imply that morality could be extended to the prudential sphere. For in addition to reasons of beneficence from the outside, the boundary proposed gives rise to moral reasons of rights and justice with respect to our coexisters to deal with potential conflicts of interests between us and them. These reasons could produce the noted difference with respect to reasons concerning ourselves (and successors) and partial reasons concerning close coexisters. The difference between partial and impartial reasons consists in that both reasons of justice and rights and reasons of beneficence can be stronger when they are partial. It may be that Parfit avoids bringing his reductionism into On What Matters because, as he confesses, he is ‘deeply worried by disagreements with people who seem as likely as I am to be getting things right’ (OWM, vol. iii, p. xiii). In his introduction to On What Matters Samuel Scheffler concurs with this diagnosis: ‘The drive to eliminate disagreement . . . is a defining feature of Parfit’s work’ (OWM, vol. i, p. xxxi). If Parfit was a revisionist by temperament, he was a conciliationalist as well. The more embedded he became in the academic establishment, the stronger he might have felt the pressure to include more people

204  Ingmar Persson in the agreement. Radical revisionary ideas inevitably kick up dust, as he would have experienced not least with respect to his views about personal identity. Feeling that he could not convince his adversaries of the truth of these ideas, he omitted them in a unification project like that of On What Matters.

3.  Extending Conciliationalism into Metaethics It is also important to Parfit’s pivotal belief that we have an intuitive ability to recognize irreducibly normative truths, both rational and moral, that disagreement about these truths not be too extensive. He writes: If we had strong reasons to believe that, even in ideal conditions, we and others would have deeply conflicting normative beliefs, it would be hard to defend the view that we have the intuitive ability to recognize some normative truths. (OWM ii. 546)

Those who defend the view that we are equipped with this intuitive ability need to make plausible ‘the prediction that, in ideal conditions, we would nearly all have sufficiently similar normative beliefs’ (OWM ii. 547). But it seems clear that if attention is drawn to fundamental normative issues which deeply divide philosophers, the prospect of sufficiently strong normative convergence in ideal conditions (whatever that precisely means) diminishes. There is therefore reason to bypass in silence such divisive issues if the aim is to bolster optimism about normative convergence. It is less obvious that the fact that we can be rationally certain about some normative claims is also problematic for the view that these claims are irreducibly normative. For instance, when we feel an intense pain that we strongly dislike for its own sake, we can be rationally certain that it is intrinsically bad for us to be feeling this pain. But it seems that this claim cannot then entail that we have an irreducibly normative or non-­natural reason to dislike or be averse this pain for its own sake, as it does on Parfit’s view, for we cannot be rationally certain that there are such reasons—after all, many philosophers who have carefully scrutinized the issue doubt this. In other words, it is doubtful not only whether Parfit’s metaethical or metanormative view is compatible with a realistic estimate of our disagreement about which normative claims are correct, but also whether it is compatible with our certainty—and thus absence of disagreement—that some of them are correct.7

7  See Persson (2017: ch. 1.1) for a fuller statement of this argument.

Parfit ’ s Reorientation  205 In On What Matter, volume iii, Parfit responds to an apparently similar argument by Allan Gibbard,8 by claiming that Gibbard’s expressivist theory—and indeed any metathical theory—is vulnerable to the same objection as his theory, since it is not certain that it is correct, either: ‘It is not certain that anything matters in Gibbard’s expressivist sense’ (OWM iii. 186). But here we must distinguish between: (1) It is not certain that ‘something matters’ means what Gibbard’s expressivist theory says that it means, and (2) It is not certain that anything matters if this means what Gibbard’s expressivist theory says that it means.

(1) is indeed true, but the claim that has to be true if Gibbard’s theory is to be vulnerable to Parfit’s objection is (2), and this claim is not true: we can be certain that something—for example, that our pleasure and pain matter—if ‘matters’ means what it does on Gibbard’s theory. This is because, according to Gibbard, the meaning of the claim that something matters is to express attitudes that we as speakers can be certain that we have. In contrast, we cannot be certain that anything matters if this means what Parfit’s theory asserts because we cannot be certain that there are reasons in his sense. Furthermore, Parfit’s characterization of our alleged ‘intuitive ability’ to recognize irreducible normative truths is not sufficiently informative because it is purely negative: this ability is said not to be based on perception, empirical evidence, or inferences from the meaning of these truths. Parfit does not say what is left for it to be based on. He repeatedly compares fundamental normative truths to logical and mathematical truths. However, there is, as he notes (OWM ii. 275), a distinction between substantive and conceptual normative truths—a distinction which apparently has no counterpart in the case of logical and mathematical truths. It is clear how conceptual normative truths—for example, if something ought to be done, there is a reason to do it—can be similar to logical and mathematical truths: they are conceptual truths, for example, because ‘ought’ entails the availability of reasons. But it is unclear how substantive normative truths—such as that there is a reason to be averse to pain for its own sake—could be significantly similar to their conceptual kin, unless the result is a form of

8  Larry Temkin also puts forward an argument similar to Gibbard’s (2017: sect. 4), to which Parfit seems to give the same sort of reply as to Gibbard (OWM iii. 45–6). Temkin replies that ‘Parfit’s sense is not the only possible sense of “not mattering” that is meaningful and significant’ (Temkin 2017: 27); there is a range of normative notions of differing strength or robustness, with Parfit’s notion at the top. I am not sure how to understand this multi-­sense proposal. Suppose I am certain that it is very bad for me to be feeling the excruciating pain that I am now feeling. Then I cannot see how the truth of Parfit’s metaethical theory could make the fact I am feeling this pain worse for me, though it might perhaps be held to raise its moral badness.

206  Ingmar Persson naturalism, according to which this reason-­claim is conceptually true because something’s being painful entails a reason to be averse to it. By contrast to the two first volumes of On What Matters in which Parfit tries to refute rival metaethical theories, in volume iii he rather extends the conciliationalist strategy from normative ethics to metaethics. He argues, for instance, that his and Gibbard’s ‘main claims don’t conflict’ (OWM iii. 225). But Gibbard seems in fact to stick to his expressivism, maintaining, for example, that ‘Parfit suggests that the notion of a reason can’t be further explained . . . but I think that it can be given . . . a sort of explanation—an expressivist explanation that reveals what’s crucial to normative concepts’ (OWM iii. 209). Gibbard is indeed prepared to say that normative claims can be true, but he explicitly states that this is truth only of ‘a minimalist sort’ (Gibbard 2017: 221), according to which ‘ “It’s true that suffering is bad” just means that suffering is bad’ (205). He adds: ‘As for truth of a more robust sort, I suspend judgment pending some satisfactory explanation of what this more-­than-­minimal truth consists in’ (221–2). So, contrary to Parfit’s wishful belief (OWM iii. 226), Gibbard explicitly refrains from committing himself to truth in the more robust ‘descriptive sense’ (OWM iii. 226) required by Parfit’s non-­naturalism. The fact that there is no agreement about there being reasons in Parfit’s sense obviously reduces the significance of building an agreement in substantive, normative ethics by appeal to these putative reasons. I suspect that Parfit is carried away by his intense eagerness to come to agreement with Gibbard. For my own part, I think it likely that the opposition in metaethics between externalism/objectivism, on the one hand, and internalism/ subjectivism, on the other, is too fundamental to be resolvable, as is the opposition between consequentialism and deontology in normative ethics, and that between compatibilism and incompatibilism in the free will debate. Let us next look at how Parfit tries to resolve a tension between consequentialism and deontology.

4.  Reconciling Parts of Common Sense and Act-­Consequentialism In On What Matters, volume iii, Parfit tries to resolve partly the disagreement between common-­sense morality and act-­consequentialism by arguing that ‘we ought to reject or at least greatly weaken’ some deontological principles of the former that ‘conflict most strongly with Act-­Consequentialism’ (OWM iii. 394). His hope is that a revision of these principles might yield something that ‘can make them parts of a wider Unified Theory’ (OWM iii. 435), such as the Triple Theory. Among the relevant principles of common-­sense morality is the one which says that our duty not to harm (for example, not to kill) is in general stronger than our duty to benefit (for example, to save lives).

Parfit ’ s Reorientation  207 Parfit starts by distinguishing between two senses in which one duty can be ‘stronger’ than another: One of two duties would be stronger in the cost-­requiring sense if we would be morally required to bear greater burdens, if that were necessary, to fulfil this duty, and stronger in the conflict-­of-­duty sense if this duty would outweigh the other when these duties conflict (OWM iii. 369). ‘This distinction’, he continues, ‘might be unimportant if duties that were much stronger in one sense would also be much stronger in the other sense. But we should not assume’ that this is so (OWM iii. 370). He then argues that the duty not to harm is not—or at least not much—stronger than the duty to benefit in the conflict-­ of-­ duty sense. He does not contend that the duty not to harm is nevertheless much stronger in the cost-­requiring sense but, as he concedes that his distinction might be unimportant if this is implausible, it seems that he must think that it is a possibility. Also, as emerged in Section 2, he holds that one could permissibly save oneself, though one would thereby fail to save two or more innocent strangers, but he would presumably not hold that one could permissibly save oneself if this required killing two such strangers. Parfit observes that disagreement about the stringency of our duties in the cost-­requiring sense ‘has great moral and practical importance’ (OWM iii. 371): it  is important whether we are morally required or merely supererogatorily ­permitted to undergo some sacrifices to fulfil our duty to benefit others. But this disagreement does not involve a conflict between two views in his sense because it  is not the case that one view ‘requires certain acts which the other view claims to be wrong’ (OWM iii. 370). Nonetheless, the two views involved in this disagreement contradict each other: one view—the commonsensical one—implies that we are morally permitted to refrain from shouldering some burdens that the other—act-­consequentialist—view implies that we are not permitted to refrain from undertaking. If these views are in tension with each other, they cannot both be parts of the same unifying theory, the Triple Theory. As already remarked, Parfit’s discussion of the dualism of practical reason in Section  2 suggests that he rejects the act-­ consequentialist view that our duty to benefit is as strong as our duty not to harm in the cost-­requiring sense. But, as will now be seen, Parfit wants to drop the commonsensical view that our duty not to harm is (much) stronger in the conflict-­of-­duty sense. I shall argue, however, that this combination of views is implausible, that our duty to benefit cannot be more or less as strong as our duty not to harm in the conflict-­of-­duty sense without this being so in the cost-­ requiring sense as well. Many people who champion the view that our duty not to kill is, as a rule, much stronger in the conflict-­of-­duty sense than our duty to save lives admit that there are exceptions. For instance, in a much-­discussed kind of case they hold that we are permitted to redirect a runaway train which is on its way to kill five

208  Ingmar Persson people onto a side track where it will kill only one person. If so, our duty not to kill one cannot here be stronger than our duty to save five. Those who accept this exception typically hypothesize that the duty not to kill is significantly weakened by the fact that we are here merely redirecting an existing threat of death, the train, rather than initiating it. Parfit compares this case, Side Track, to another case, Hand Grenade, in which we have to save the five from being killed by the runaway train by throwing a hand grenade at it which would blow it up so that a bystander is killed by ‘flying fragments of this train’ (OWM iii. 389). Many would say that this is impermissible according to common-­sense morality because it involves initiating the cause of the single person’s death, the explosion. In opposition to this, Parfit maintains that those who hold that redirecting the train is permissible in Side Track for the reason given must also regard the throwing of the grenade as permissible because ‘In both cases, you would be redirecting an existing threat, the runaway train’ (OWM iii. 389). Granted, the bystander’s being killed by the fragments is a side effect of saving the five by blowing up the train. But Hand Grenade is incontestably a case of blowing the train to pieces, which is surely incompatible with redirecting it. The reason that the train is not redirected to the bystander is not just that the fragments hitting him are only small parts of the train. It is also that the source of the fragments’ motion is no longer the same as the source of their motion when they were parts of the moving train. The explosion is the source of the fragments’ motion when they hit the bystander, and this is not what originally set the train in motion. Therefore, in Hand Grenade we have initiated a new threat of death rather than redirected an existing one. That the train is not redirected is shown by the fact that the flying fragments could have had the same trajectory and lethal effect on the bystander, even though the explosion was insufficient to redirect the train, and it—that is, what still counts as the same thing, the train itself—had carried on to kill the five. The following variation of Hand Grenade would, however, feature something more like a redirection of the train. Hand Grenade Two: suppose that the exploding grenade causes the train to derail, so that it runs over a bystander. Here the explosion might just change the direction of the initial motion of the train, by forcing the train off the rail, which is why this case is like redirecting the train. Its running over the bystander is a side effect of saving the five, as in Side Track. So, it might seem that it cannot make any moral difference whether you save the five as in Hand Grenade Two or as in Side Track. On the other hand, it may seem that it cannot make any moral difference whether the grenade you throw blows up the bystander along with the train or causes it to derail and run over this person. (The death of the victim is assumed to be equally painful and bad in all cases.) But if this is so, it follows that there is no moral difference between Side Track and Hand Grenade. Thus, those who want to stand by the view that redirecting the train in

Parfit ’ s Reorientation  209 Side Track is permissible must admit that it is also permissible to cause the bystander to die by throwing a grenade in Hand Grenade, though it involves initiating a new cause of death. Consequently, the fact that a lethal threat is redirected rather than initiated does not appear to make it noticeably easier to justify. The fact that the killing of the one in these cases is not done as a means to the saving of the five but is a foreseen side effect of it is also commonly thought to be necessary to weaken the duty not to kill. In a 2014 version of what was to become chapter 56 of On What Matter, volume iii, Parfit discusses two variants of redirecting a runaway train which differ only in this respect. In both variants, one can redirect the train onto a curved side track on which a single person is placed, but they differ in that in Rejoins Before the side track rejoins the track on which the five sit in front of them, whereas in Rejoins Beyond it rejoins this track behind them. In Rejoins Before the fact that the body of the one stops the trolley is a causal means of saving the five, while in Rejoins Beyond it is merely a foreseen side effect. According to Parfit’s intuition—and mine—this makes no moral difference: in both variants redirecting the train to save the five is permitted. If these arguments are correct, it follows that the fact that a killing would be an instance of killing as a means or an instance of initiating a cause of death does not strengthen the duty not to kill so much that it outweighs the saving of four additional lives. At least for the sake of the argument, let us follow Parfit in thinking that it would not even outweigh one more life being saved, so that in the conflict-­of-­duty sense there is no significant difference in stringency between the duty not to kill and the duty to save lives. The question is then whether this is compatible with there being a significant difference in strength in the cost-­ requiring sense. I doubt it. Imagine that my self-­interested reason not to make a sacrifice is just about strong enough to outweigh my moral duty or reason to save somebody. If my moral reason to save a life is equally strong as my reason to refrain from killing or taking a life when they are compared to each other, it is hard to see how the reason to refrain from killing can be stronger than the reason to save when each of them is compared to a self-­interested reason not to make a certain sacrifice, and outweigh this self-­interested reason when the reason to save does not. If we have a right to spare ourselves a certain sacrifice that outweighs our duty to save a life, and this duty is not stronger than the duty to refrain from taking a life when they conflict, how could the right fail to outweigh the latter duty as well? The following case illustrates that there is something peculiar about denying the equivalence in stringency of the two senses. Suppose a runaway train is heading towards a number of people so large that my redirecting it onto a single person on a side track to the right would be not merely morally permissible but required if this were the only alternative available to me. It is, however, not my only alternative to letting the greater number die, for to the left there is another

210  Ingmar Persson side track on which I am sitting. But the number of lives saved is not so large that I would have been required to redirect the train towards myself had this been my only alternative to letting the greater number die. This is possible if the duty to benefit is weaker than the duty not to harm in the cost-­requiring than in the conflict-­of-­duty sense, so that I am not required to sacrifice as much to benefit as not to harm. But, as things stand, there is both the alternative of sacrificing myself to benefit by redirecting the train to the left and the alternative of harming someone else by directing it to the right, and this makes it unclear what is required of me. For I am not permitted to redirect the train to the right, killing the one, rather than to the left, killing myself. Nor am I permitted to let it kill the greater number rather than redirecting it to the one. Am I then required to redirect the train to myself on the left, though I would have been permitted to let the greater number die instead had not the side track to the right with the one existed? But it seems odd that the presence of the one to the right could require me to sacrifice myself in order to save the greater number of lives when this number by itself could not do this. It then seems unclear what I should do if there is not the equivalence of strength between the two senses. On the other hand, if stringency in the cost-­ requiring sense were equivalent to stringency in the conflict-­of-­duty sense, it would be clear that I should sacrifice myself by redirecting the train to the left, irrespective of whether there is the option of redirecting it to the right. This argument might not be sufficient to show that it is a mistake to reject the equivalence of strength, but it is an argument for it being a mistake, and we have been given no argument in favour of the rejection. Moreover, if we are uncertain where the truth in this matter lies, it is a better bet to think that it lies with the equivalence of strength, since otherwise we would risk ignoring sacrifices we ought to make in fact. Accepting this equivalence of strength would, however, mean that if Parfit’s claim that there is no significant difference in strength between the duties in the conflict-­of-­duty sense is right, there would, as act-­ consequentialism implies, be a requirement to save lives even at a cost that is more or less as great as the cost we are required to incur to refrain from killing the same number. As Parfit points out (OWM iii. 371), this is of greater practical importance than the fact that the duty to refrain from killing is at most only a little stronger in the conflict-­of-­duty sense than the duty to save lives because, whilst we seldom have to kill others to save lives, we residents in affluent countries seem daily to be in a position to save many lives in poorer countries by making donations to welfare organizations that we could bear with ease. In Section 2 I argued that if he had stood by his reductionist belief that personal identity does not matter, he would be lacking justification for the dualism of practical reason, which is his warrant for maintaining that our self-­interested reasons could offer sturdy opposition to our duties to benefit (though this leaves unexplained how they could do so without offering equally sturdy opposition to

Parfit ’ s Reorientation  211 our duties not to harm). But it seems that the Triple Theory—which includes rule-­consequentialism—could not recommend a moral rule that requires us to sacrifice more or less as much to save life as to avoid killing. This would presumably demand too much of people to save life (or too little to abstain from killing). Yet it could well recommend rules that presuppose that there is at most a small difference between the duty not to kill and the duty to save lives in the conflict-­of-­duty sense, since the acceptance of such rules would have the good consequence of saving more lives, and would not be very hard to comply with. Additionally, a morality with such a stringent duty to benefit in the cost-­requiring sense would have little semblance to common-­sense morality. On the other hand, a morality which entails that our duty to benefit is much weaker than our duty not to harm in the cost-­requiring sense than in the conflict-­of-­duty sense would be dubious if my argument has been correct. I have adduced the claim that personal identity does not matter as an example of a revisionary aspect of Reasons and Persons that is bypassed in On What Matters but it is not the only example. A leitmotif of part IV of Reasons and Persons is the search for ‘a new theory of beneficence’ which can cope with problems in population ethics such as the non-­ identity problem and the repugnant conclusion (RP, 443). In Parfit’s view, in order to show that any moral ‘theory could be objectively the best theory’, ‘we must find a theory which resolves our disagreements’ about these matters (RP, 452). But in On What Matters there is no attempt to show how a morality underpinned by the Triple Theory could resolve these disagreements. The task of finding ‘a new theory of beneficence’, ‘Theory X’, has disappeared from the horizon. The non-­identity problem and the repugnant conclusion are together with reductionism about personal identity the topics in Reasons and Persons that have generated most controversy, but they have symptomatically vanished in On What Matters.9

5.  Conclusion Between Reasons and Persons and On What Matters then, there has been a marked change in the orientation of Parfit’s moral philosophy, from a revisionist approach in Reasons and Persons to a conciliationalist approach in On What Matters consisting in an attempt to base conciliatory normative claims on a common metaethical foundation. To this extent, we can speak of an ‘earlier’ and ‘later’ Parfit, though there is undeniably a considerable overlap between the two. Perhaps it would be more apt to described him as having branched, since the 9  Parfit was also a revisionist about our attitudes to time, but this is more decisively argued in some unpublished writings subsequent to pt. II of RP, and a revisionist about desert which is the topic only of the brief ch. 11 of OWM.

212  Ingmar Persson earlier Parfit of Reasons and Persons continued to exist alongside the branch of Parfit of On What Matters that emerged later. Anyway, the fact is that in On What Matters Parfit leaves out his most hotly debated views in Reasons and Persons, such as his reductionism about personal identity and prominent problems of population ethics, though his subsequent work shows that these issues did not cease to be of great concern to him. I have also argued that the attempt in On What Matters, volume iii to strengthen the conciliationalist approach by extending it to metaethics and to common-­sense morality faces difficulties. I think, though, that this failure of the conciliatory project is not Parfit’s fault. I regard him as the greatest philosopher I have met, and I am glad that I got around to telling him so before he died. It is proper for a moral philosopher to aim, as he did, to establish a rational consensus about what is morally right and wrong, and what the ground for this is. But for my own part I am inclined to think that, unfortunately, this is an unattainable goal because we have fundamentally conflicting intuitions in both normative ethics and metaethics.

Bibliography Gibbard, Allan (2017). ‘Gibbard’s Commentary.’ In Parfit, On What Matters, volume ii, 205–24. Hare, R. M. (1981). Moral Thinking. Oxford: Clarendon Press. Hare, R. M. (1997). ‘Could Kant Have Been a Utilitarian?’ Repr. in R. M. Hare, Sorting out Ethics, 147–65. Oxford: Clarendon Press. Parfit, Derek (1987). Reasons and Persons. Reprinted with corrections. Oxford: Clarendon Press. References are to RP and page no. unless otherwise stated. Parfit, Derek (2011). On What Matters. Volumes i and ii. Oxford: Oxford University Press. References are to OWM and volume and page no. unless otherwise stated. Parfit, Derek (2012). ‘We Are Not Human Beings.’ Philosophy 87: 5–28. Parfit, Derek (2017). On What Matters. Volume iii. Oxford: Oxford University Press. References are to OWM and volume and page no. unless otherwise stated. Persson, Ingmar (2005). The Retreat of Reason. Oxford: Clarendon Press. Persson, Ingmar (2017). Inclusive Ethics. Oxford: Oxford University Press. Temkin, Larry (2017). ‘Has Parfit’s Life been Wasted?’ In Does Anything Really Matter?, edited by P. Singer, 1–34. Oxford: Oxford University Press.

9

Parfit’s Final Arguments in Normative Ethics Brad Hooker

One of the most influential books in normative ethics has been Derek Parfit’s Reasons and Persons. The moral theory dominating that book was Act Consequentialism. The first two volumes of Parfit’s On What Matters, published twenty-­seven years after Reasons and Persons, gave Act Consequentialism only passing attention. However, volume iii of On What Matters, published six years later and in the week of Parfit’s death, contains a long discussion that develops Act Consequentialism in ways that help it circumvent familiar objections.1 Yet then Parfit argues that Rule Consequentialism is a more plausible moral theory. In the sections that follow, I will provide more detail about the extent to which Parfit’s normative ethics changed, I will explain how the reflective-­ equilibrium methodology he endorsed in the earlier volumes of On What Matters can lead to conclusions he did not there consider, and I will assess the kind of Act Consequentialism he discusses in the final volume of On What Matters. Finally, I will consider the arguments for Rule Consequentialism that Parfit puts forward near the end of his final book.

1.  Act Consequentialism in Reasons and Persons The Act Consequentialism in Reasons and Persons was a criterion of moral ­rightness, holding that acts are morally right if and only if and because they produce at least as good consequences as any alternative acts. Act Consequentialism This paper has been presented at conferences at York, Reading, and Karlsruhe. The paper has been revised to deal with insightful remarks from Marco Bernardini, James Brown, Jamie Draper, Johan Gustafsson, Matthias Hoesch, Chris Jay, Todd Karhu, Kacper Kowalczyk, Katarzyna de Lazari-­Radek, Giuliana Manca, Tim Mulgan, Taylor Paytas, Adam Pearce, Chris Ryder, Christian Seidel, José de Sousa  E.  Brito, Rob Truman, Elizabeth Ventham, Tatjana Visak, Hayden Wilkinson, and Michael Zimmerman. I am grateful to all these people. For extremely helpful written comments, I thank Karhu, Rob Jubb, Roger Crisp, Aart van Gils, Charlotte Newey, Jeff McMahan, and Peter Vallentyne. 1  On What Matters (OWM), i and ii (Oxford: Oxford University Press, 2011); iii (Oxford: Oxford University Press, 2017). References to volume and page of OWM will be given in the text.

Brad Hooker, Parfit's Final Arguments in Normative Ethics In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0010

214  Brad Hooker was not taken in that book to be a decision procedure appropriate for people to use in everyday moral decision making. Reasons and Persons defended Act Consequentialism against the charge that better consequences would come not only from people’s not trying to employ Act Consequentialism as their procedure for making moral decisions but also from most people’s not believing in Act Consequentialism. Here Parfit was following his hero Henry Sidgwick.2 Reasons and Persons also contained influential discussions of what Sidgwick had called Common Sense Morality. Common Sense Morality has an Act-­ Consequentialist component. The Act-­Consequentialist component in Common Sense Morality is (a) a general duty to do good. Common Sense Morality also contains many deontological components. These are (b) prohibitions on physically harming others, stealing, promise-­breaking, and lying, (c) personal obligations to close family, friends, and others with whom one has personal connections, and (d) a permission to give one’s own good somewhat more or less weight in one’s decision making than the good of others.3 In Reasons and Persons, Parfit didn’t dismiss Common Sense Morality. Instead, he argued that we could better achieve some of its aims if we focused on what we could achieve together, rather than individually.

2.  Volume I of On What Matters Reasons and Persons was followed twenty-­seven years later by On What Matters, volumes i and ii. On What Matters, volume i argued for the Triple Theory. The Triple Theory holds that (1) a revised version of Kant’s Universal Law Formula, (2) Scanlonian Contractualism, and (3) Rule Consequentialism converge on the same set of moral rules. Act Consequentialism receives some attention in On What Matters, volumes i and ii, but Parfit’s criticisms of it cut very deep. Moreover, Act Consequentialism does not form part of Parfit’s recommended Triple Theory. The fundamental goal of On What Matters, according to Parfit, is to resolve our disagree­ments about ethics, to show that the best ethical theories converge, to reveal that in normative ethics we are ‘climbing the same mountain from different sides’, with some prospect of reaching the same summit, from which we will be able to share the same views. In order to achieve this goal, Parfit thought he had to persuade us of certain things about each of Kantianism, Rule Consequentialism, and Scanlonian Contractualism. 2 H. Sidgwick, Methods of Ethics (7th edn, London: Macmillan, 1907). See also R. E. Bales, ‘Act-­ Utilitarianism: Account of Right-­making Characteristics or Decision-­making Procedure?’, American Philosophical Quarterly 8/3 (1971): 257–65. 3  For a sustained campaign to incorporate many different deontological elements into consequentialism, see Douglas Portmore’s Commonsense Consequentialism: Wherein Morality Meets Rationality (Oxford: Oxford University Press, 2011).

Parfit ’ s Final Arguments  215 The moral theory on which Parfit spends the greatest attention in volume i of On What Matters is Kant’s. The Universal Law Formula is, in Parfit’s view, the most intriguing of Kant’s various versions of the Categorical Imperative. Kant’s Universal Law Formula of the Categorical Imperative tells each to act on maxims that he or she can will to be universal laws. Parfit suggested Kant meant to refer not merely to laws that the agent could will but instead to laws that ‘everyone could rationally will that everyone accept and follow’ (OWM i. 340). Whether or not that is what Kant meant, Parfit argued that Kant’s Universal Law Formula should be revised to become the foundational principle stating that everyone ought to comply with the principles that everyone could rationally will that everyone accept. The moral theory that consists of this foundational principle is dubbed by Parfit Kantian Contractualism. Since Kantian Contractualism holds that what is morally required, morally optional, or morally prohibited depends on what principles everyone can rationally will that everyone accept, we must ask what can be rationally willed. Parfit’s answer was that what one can rationally will depends on what normative reasons one has. He distinguished four kinds of normative reasons relevant to what an agent could rationally will. These are (1) self-­interested reasons for the agent to will what is beneficial to herself, (2) ‘partial’ altruistic reasons for the agent to will what is beneficial to people personally connected to her (e.g., her friends and family), (3) impartial reasons, and (4) ‘deontic’ reasons for the agent to will whatever is morally required. Parfit explains that ‘deontic reasons’ should be the output of our moral reasoning, not the input, on danger of circularity. So, if deontic reasons cannot be appealed to in determining what everyone can rationally will that everyone accepts, then what can be appealed are only self-­ interested, ‘partial’ altruistic reasons, and impartial reasons. These three kinds of reasons are then to be weighed against one another in order to determine what principles everyone can rationally will that everyone accept. As indicated, Kantian Contractualism consists of the foundational principle that everyone ought to comply with the principles that everyone can rationally will that everyone accept. This foundational principle refers to further principles, which are to be selected on the basis that everyone can rationally will that everyone accepts them. Confusion can arise from using the term ‘principle’ to refer both to the foundational principle and to the further principles selected in accordance with the foundational principle. To avoid this confusion, I will use the term ‘rules’ to refer to the further principles. With that terminological point made, I return to explaining the contentions that comprise Parfit’s Kantian Contractualism. Parfit contended that the impartial reasons to will the universal acceptance of the rules whose universal acceptance would make the world go impartially best are sufficient for everyone to rationally will universal acceptance of these rules. Can that contention be right? After all, to be weighed up are not only impartial reasons but also reasons of self-­interest and

216  Brad Hooker ‘partial altruism’. Rules that might be impartially advantageous might be disadvantageous for a given agent and her loved ones. Nevertheless, since what any given agent can will for acceptance by everyone (everywhere and evermore, OWM i. 382) will have immense ramifications across time for others outside the circle of the agent and her loved ones, the impartial reasons are strong enough not to be outweighed by the combination of reasons of self-­interest and partial altruism.4 Parfit also contended that the only rules that everyone would have sufficient reason to will, and could therefore rationally will, are the rules whose universal acceptance is favoured by the strongest impartial reasons (OWM i. 379–88). So, according to Kantian Contractualism, these impartially favoured rules determine right and wrong. Parfit’s Kantian theory does not hold that acts are right or wrong depending on whether everyone has sufficient reason to will that everyone do or not do these acts. The ‘sufficient reason’ test is about rules, not directly about acts. Acts are morally required, morally optional, or morally prohibited depending on what the impartially favoured rules are. In this respect, Kantian Contractualism is like Rule Consequentialism. The question for proponents of Rule Consequentialism is not whether an act would achieve what the strongest impartial reasons favour. Rather the question is whether the act is permitted by the rules whose widespread acceptance would have the best consequences impartially considered. Parfit, like others before him, refers to such rules as the optimific rules.5 One of the points Parfit makes about Rule Consequentialism, especially in volume iii of On What Matters (417), is that Rule Consequentialism should be formulated in terms of the acceptance of rules, rather than compliance with rules. Imperfect beings like us might be incapable of perfect compliance. Even if achieving perfect compliance with rules would be possible given enough training, practice, and reinforcement, the psychological costs of getting people to the point where successful compliance with the rule is secured might be very substantial. The costs to the teachers and the students of securing acceptance of rules that are more complicated would be higher than securing the acceptance of rules that are less complicated. Likewise, rules that demand more self-­sacrifice have higher 4  This line of thought is not the same as the line of thought in Reasons and Persons that, ex ante, in conditions of uncertainty, each person could expect those to whom she is specially related to do better if everyone regularly chooses what impartial reasons support rather than what reasons of partiality support. For related discussion, see Johann Frick, ‘Contractualism and Social Risk’, Philosophy and Public Affairs 43/3 (2015): 175–223; Korbinian Rüger, ‘On Ex Ante Contractualism’, Journal of Ethics and Social Philosophy 13/3 (2018): 240–58. 5  As Michael Zimmerman reminded me, C.  D.  Broad used the term ‘optimific’ and formulated some of the innovations incorporated into what Parfit called ‘act-­involving Act Consequentialism’, on which see Section 3. Broad’s Cambridge lectures on ethics can be found in his Ethics, edited by C. Lewy (Dordrecht: Nijhoff Publishers, 1985).

Parfit ’ s Final Arguments  217 ‘teaching and internalisation costs’ than less demanding rules. At some point, the higher costs involved in getting more complicated and more demanding rules accepted will outweigh the behavioural advantages of compliance with more complicated or more demanding rules.6 Just as Parfit contended both that, if Kantianism and Rule Consequentialism are to be acceptable and to be incorporated into a favoured convergence then Kantianism has to take a particular form and that Rule Consequentialism has to be framed in terms of acceptance, he contended that Scanlonian Contractualism needs to be modified. Scanlonian Contractualism holds that an act is wrong if it would be ‘disallowed by any set of principles for the general regulation of behaviour that no one could reasonably reject as a basis for informed, unforced general agreement’.7 Scanlon himself placed an ‘individualist restriction’ on what can count as a reasonable objection to a proposed principle. This restriction is that reasonable rejection must be based on the effects on a single individual, not on the effects on an aggregate of individuals. Parfit argued that the individualist restriction has implausible implications (OWM ii. 191–212). Separately, since Parfit’s Kantian Contratualism makes the justification of rules depend upon impartial reasons to favour greater aggregate net benefit over lesser aggregate net benefit, Scanlonian Contractualism will have to drop its rejection of aggregate considerations if it is to converge with Parfit’s Kantian Contractualism on the selection of moral rules. One thing Parfit’s Kantian Contratualism shares with Rule Consequentialism is the appeal to a kind of impartiality that gives full weight to aggregative considerations. And so, if Scanlonian Contractualism is to converge with Rule Consequentialism on the selection of moral rules, again Scanlonian Contractualism will have to drop its rejection of aggregate considerations. Rather than trace through the arguments Parfit put forward for the convergence of his Kantial Contractualism, Rule Consequentialism, and Scanlonian Contractualism, I want to step back and observe that woven into On What Matters, volumes i and ii was the presumption that our methodology in normative moral theory is the search for reflective equilibrium between all our beliefs, including our intuitions about kinds of cases and our most general moral ­principles.8 Commitment to such a methodology crops up in volume i of On What Matters (OWM i. 185, 352, 367, 370, 401, 415). Parfit reiterates his commitment to the reflective equilibrium methodology in volume ii (e.g. OWM ii. 154), and very emphatically in volume iii (OWM iii. 433). 6 See my Ideal Code, Real World: A Rule-­consequentialist Theory of Morality (Oxford: Oxford University Press, 2000), 76–80, 89–90, 96–99, 138–141 and Parfit’s OWM iii. 414, 417–18, 420–1. 7 T.  M.  Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998), 153. 8  I pointed this out in ‘Must Kantian Contractualisn and Rule-­Consequentialism Converge?’, in Mark Timmons (ed.), Oxford Studies in Normative Ethics 4 (2014): 34–52, at 51–2.

218  Brad Hooker Now, if our methodology is the search for reflective equilibrium, then the variety of consequentialism that comes into view is definitely Rule Consequentialism, since it coheres far better with our intuitions about what kinds of acts are right or wrong than familiar varieties of Act Consequentialism do. This observation about Rule Consequentialism is hardly new. Roy Harrod’s 1936 paper makes the observation, as does James Urmson’s 1953 paper.9 The observation has been made countless times since. But, from about the middle 1960s until the middle 1990s, most philosophers thought that there are objections to Rule Consequentialism that are utterly ­devastating. The most influential objection was that Rule Consequentialism faces a dilemma: it either collapses into extensional equivalence with Act Consequentialism or is incoherent. That objection does not rely on a commitment to seeking reflective equilibrium between our moral theory and our intuitions about what kinds of acts are right or wrong. The objection was that Rule Consequentialism’s internal workings are fatally defective. However, in the middle 1990s, arguments were developed that Rule Consequentialism can be formulated so as to avoid both collapse into extensional equivalence with Act Consequentialism and incoherence. These arguments were convincing enough to revive interest in Rule Consequentialism.10 If Rule Consequentialism is guilty of neither collapse nor incoherence, then its fit with our intuitions about right and wrong might endear it to us. The search for reflective equilibrium between our moral intuitions and our moral theories pushes us to favour whatever versions of our theories have implications that cohere with our intuitions. Thus, it can be no surprise that what seem to us the best versions of moral theories (such as the best versions of  Kantianism Contractualism, Rule Consequentialism, and Scanlonian Contractualism) have very similar implications.11 In light of that point, there is a  missed opportunity in On What Matters, volumes i and ii, which I will now identify. In the first two volumes, Parfit wrote little directly about Common Sense Morality, although he invoked common-­sense moral intuitions to reject Act Consequentialism (OWM i. 144, 191, 362, 404–5, 417, 482) and to reject various 9  Harrod, ‘Utilitarianism Revised’, Mind 45/178 (1936): 137–56; Urmson, ‘On the Interpretation of the Moral Philosophy of J. S. Mill’, Philosophical Quarterly 3/10 (1953): 33–9. 10  I first addressed the objections in ‘Rule-­Consequentialism, Incoherence, Fairness’, Proceedings of the Aristotelian Society 95/1 (1995): 19–35. There I proposed that a rule-­consequentialist agent’s most basic moral motivation could be a concern for impartial defensibility rather than concern for the impartial good. See also my Ideal Code, Real World, ch. 4; and ‘Rule-­Consequentialism and Internal Consistency: A Reply to Card’, Utilitas 19/4 (2007), 514–19. More recent discussions of the topic include Susan Wolf ’s ‘Two Concepts of Rule Utilitarianism’, in M. Timmons (ed.), Oxford Studies in Normative Ethics 6 (2016): 123–44 and David Copp’s ‘The Rule Worship and Idealization Objections Revisited and Resisted’, in M. Timmons (ed.), Oxford Studies in Normative Ethics 10 (2020): 131–55. 11  I made this point and the points below about extending this line of thought in ‘Must Kantian Contractualism and Rule-­consequentialism Converge?’, at 52, including n. 10.

Parfit ’ s Final Arguments  219 Kantian theses. As I mentioned earlier, Common Sense Morality is a pluralistic theory consisting of (a) a general duty to do good, (b) prohibitions on physically harming others, stealing, promise-­breaking, and lying, (c) personal obligations to close family, friends, and others with whom one has personal connections, and (d) a permission to give one’s own good somewhat more or less weight in one’s decision making than the good of others. Common Sense Morality consists of the same general and specific obligations, prohibitions, and permissions that the best versions of Kantian Contractualism, Rule Consequentialism, and Scanlonian Contractualism endorse. Hence, when in volumes i and ii of On What Matters Parfit put forward his Triple Theory of Kantian Contractualism, Rule Consequentialism, and Scanlonian Contractualism, he should have included Common Sense Morality, thereby producing the Quadruple Theory. Common Sense Morality is unlike the other three theories just mentioned in that Common Sense Morality, I take it, is agnostic on whether there is a deeper moral principle providing a foundational and unifying justification for the general and special duties, the prohibitions, and the permissions. Such agnosticism about a foundational unifying justificatory principle contrasts with the commitment to such a foundational unifying justificatory principle made respectively by Kantian Contractualism, Rule Consequentialism, and Scanlonian Contractualism. These three theories disagree with one another about what the one unifying foundational priniciple is. But these three theories’ disagreement about what the unifying foundational principle is doesn’t keep the three theories from being the elements of Parfit’s Triple Theory. If disagreement at the foundational level doesn’t preclude theories’ being parts of the Triple theory, then a theory’s agnosticism about whether there is such a correct unifying foundational principle cannot be a bar.12 Parfit himself later, in effect, admits that Common Sense Morality should be included in the convergence (OWM iii. 434). This line of thought might be extended even further. We should be open-­ minded, I think, that the best version of Virtue Ethics will have implications that converge with the prohibitions, duties, and permissions endorsed by the Quadruple Theory (cf. OWM iii. 418). If we find such a Virtue Ethics, then we have the Quintuple Theory—consisting of Parfit’s Kantian Contractualism, Rule Consequentialism, Scanlonian Contractualism, Common Sense Morality, and Virtue Ethics.13

12  I distinguish between Common Sense Morality, which I take to be agnostic about a correct unifying foundational principle, and deontological pluralism, which, as its name suggests, denies there is a correct unifying foundational principle. But whether or not I am mistaken about Common Sense Morality’s agnosticism, Parfit should have included Common Sense Morality in a Quadruple Theory. 13  Does a commitment to moral realism militate in favour of, or against, believing that various normative ethical theories converge? See Marius Baumann, ‘Parfit, Convergence, and Underdetermination’, Journal of Ethics and Social Philosophy 13/3 (2018): 191–221.

220  Brad Hooker

3.  Act Consequentialism in On What Matters, Volume III Can Act Consequentialism be refined to a point where it can be included in the convergence? Having spent much of Reasons and Persons on Act Consequentialism but little of volumes i and ii of On What Matters on this theory, Parfit devotes more of volume iii to discussing Act Consequentialism than he does to discussing any other normative moral theory. I find that surprising, though I guess that in epic journeys the protagonist often circles round past earlier sites. In Reasons and Persons, Parfit commented that Act Consequentialism can appeal to non-­utilitarian principles. He wrote, According to three such principles, it is bad if people are deceived, coerced, and betrayed. And some of these principles may refer to past events. Two such principles appeal to past entitlements, and to just deserts. . . . If our moral theory contains such principles, we are not concerned only with consequences in the narrow sense: with what happens after we act.14

In On What Matters, vol. iii, Parfit expands on the extent to which Act Consequentialism’s counter-­intuitive implications can be avoided if the theory postulates that acts can have ‘intrinsic moral badness’ or ‘intrinsic moral value’ (OWM iii. 396–406). For example, if acts of promise-­breaking have intrinsic moral disvalue, then we could explain in Act-­ Consequentialist terms how a particular act of promise-­breaking can be morally wrong even if this act of promise-­breaking benefits people slightly more than would the alternative act of promise-­keeping. In this particular example, the intrinsic moral badness in the act of promise-­breaking outweighs the slightly greater benefit to people that is produced by promise-­breaking. Thus, here the act of promise-­breaking would not be the available act with the greatest overall goodness. The intrinsic moral badness here is, in Parfit’s terminology, ‘non-­deontic’, meaning that the badness is not wrongness. Parfit wants to allow that, though a kind of act has intrinsic moral badness, and this often will be part of why an instance of the act is morally wrong, there can be instances of the very same kind of act which are not morally wrong, though, like all other instances of the kind, still intrinsically morally bad. An example might be one in which an act of deceit or coercion or promise-­breaking or even harming an innocent person, though intrinsically morally bad, would not be morally wrong because the intrinsic badness would be outweighed by tremendous instrumental value to something good. Suppose, for example, an intrinsically bad act of deceit or coercion or promise-­breaking or even harming an innocent person were necessary to save 14  Reading and Persons (RP) (Oxford: Oxford University Press, 1984), 26. Further references to pages of RP will be given in the text.

Parfit ’ s Final Arguments  221 very many innocent lives (OWM iii. 398–401). In such an example, the act of deceit or coercion or promise-­breaking or even harming an innocent person would be intrinsically bad but would not be morally wrong. Parfit uses the term ‘Act-­involving Act Consequentialism’ to refer to the theory that the rightness or wrongness of an act is determined exclusively by whether its consequences are at least as good as those of any alternative act, where the goodness of consequences includes not only the impartial value or disvalue in states of affairs but also intrinsic moral value or intrinsic moral disvalue of acts. Acts with intrinsic moral value might be acts of gratitude or loyalty or treating people as they deserve (OWM iii. 406). Treating people as they deserve, for example, would be the right thing to do, according to Act-­ involving Act Consequentialism, even if the consequences would contain a little less aggregate well-­being than not treating them as they deserve would. But in cases where treating people as they deserve would result in far less aggregate well-­being, the Act-­involving Act-­Consequentialist calculation would come down in favour of not treating people as they deserve. Act-­involving Act Consequentialism offers an explanation that is different from but compatible with the deontological explanation of the wrongness of these acts offered by W. D. Ross in terms of prima facie duties. In both Act-­involving Act Consequentialism and Ross’s pluralist deontology, there are various moral properties of acts that always count with the same moral polarity whenever these properties are instantiated. For example, any act with the property of being an act of promise breaking has some intrinsic moral disvalue and there is a prima facie duty weighing against it. According to both Act-­involving Act Consequentialism and Rossian deontology, the fact that an act has the property of being an act of promise breaking is always morally negative, but this negative doesn’t always outweigh other morally relevant considerations (OWM iii. 404–5). Pointing out how Act-­involving Act Consequentialism can operate in parallel fashion to Ross’s theory of prima facie duties raises two possible objections to Act-­ Involving Act Consequentialism. The first of these objections to Act-­ involving Act Consequentialism Parfit addresses. The second he does not. The first objection is that Act-­involving Act Consequentialism seems inferior to Ross’s theory of prima facie duties when a distinction is made between duties in general and duties owed to people with whom one has special connections.15 Ross stressed this distinction. In contrast, the Act-­involving Act Consequentialism that Parfit discusses focuses on the intrinsic goodness or badness of kinds of act and on consequences, assessed agent-­neutrally.16

15  For an influential discussion, see W. D. Ross, The Right and the Good (Oxford: Oxford University Press, 1930), 25. 16  In contrast, see Portmore, Commonsense Consequentialism.

222  Brad Hooker Here is a standard example. Suppose there is intrinsic moral goodness in ­ eople’s helping their own children. There might thus be an agent-­neutral duty to p promote the extent to which people in general help their own children. Such a duty can be distinguished from an agent-­relative duty that parents have to help their own children. In a situation in which somehow my helping my children will prevent other people helping their children, I am under pressure from the agent-­ relative duty to help my own children. This pressure might be opposed by a duty to promote the welfare of others in general and would be opposed by any agent-­neutral duty to promote the extent to which parents in general help their children. Where there is such a conflict of duties, depending on what is at stake for the various effected parties, my duty to my own children might outweigh my agent-­neutral duty to promote the extent to which people in general help their children, and outweigh even the duty to promote the welfare of others in general.17 In other possible cases, the agent-­relative duty to my own children is outweighed by agent-­neutral duties. Other examples of agent-­relative duties zero in on the relation of agents to the acts they do, as opposed to acts they promote or minimize. We have an agent-­ neutral duty to try to prevent the harming of innocent people, but we have a stronger agent-­relative duty not to harm people. We have a duty to oppose the coercion of people, but we have a stronger duty not to coerce people ourselves. And so on. Ross’s theory of prima facie duties stresses agent-­relative duties. This is also true of what I earlier referred to as Common Sense Morality. In contrast, Parfit’s Act-­involving Act Consequentialism is unsympathetic to agent-­relative duties (OWM iii. 406–12). He distinguishes between cases in which others are behaving wrongly and cases in which others are not behaving wrongly. He proposes most of us, and perhaps even Common Sense Morality, would agree that, in cases where others are not behaving wrongly, our duty is to minimize harm, deceit, or coercion, rather than to avoid being the one who perpetrates the harm, deceit, or coercion. The idea is that, at least in these cases where others are not behaving wrongly, we should focus on minimizing the amount of bad, rather than focus on avoiding doing anything with intrinsic moral badness in it. I am doubtful that most people will agree with him about that. Parfit thinks cases in which others are behaving wrongly are different. Suppose we are threatened that, unless we kill some others, many more others will be killed by other agents. About such a case, Parfit writes, ‘we may plausibly believe that it would be wrong to give in to such threats, since that would encourage future threats, in ways that might then lead to more people being killed’ (OWM iii. 412). This response remains completely agent-­neutral Act Consequentialist. In so far as we think morality is at least partly agent-­relative, 17  For an especially widely discussed attack on agent-­relative duties, see Samuel Scheffler’s The Rejection of Consequentialism (Oxford: Oxford University Press, 1982).

Parfit ’ s Final Arguments  223 the complete agent-­neutrality of this kind of Act-­involving Act Consequentialism will strike us as counter-­intuitive. Even more importantly, the earlier example in which I was weighing up my agent-­relative duty to help my own children against my agent-­neutral duty to promote the extent to which people in general help their own children was not one in which my helping my children will somehow lead to other parents’ wrongly failing to help their children. The situation might well be such that my helping my children will merely result in other people’s not having the opportunity to help theirs. To the extent that our intuition remains that parents can have a duty to help their own children even when their helping their children would result in other people not having the opportunity to help theirs, Parfit’s arguments against agent-­relative duty are unpersuasive. A different objection to Act-­ involving Act Consequentialism is utterly independent of the above arguments. Suppose Parfit were to respond to the above line of objection by conceding that Act-­involving Act Consequentialism should incorporate agent-­relative duties just as much as necessary in order to make the theory have implications that converge with our intuitions. Even if he responded in such a concessionary way, the objection I am about to offer would work just as well, perhaps even better. This objection is that, just as Ross’s pluralist deontology relies on a lot of postulates about prima facie duties, Act-­Involving Act Consequentialism contains a lot of postulates about the intrinsic moral value or disvalue of acts. Admittedly, these postulates enable Ross’s pluralist deontology and Act-­ Involving Act Consequentialism to have intuitively plausible implications about what is right or wrong in various circumstances. However, if some other moral theories are equally able to generate intuitively plausible results and these other theories do so on the basis of fewer postulates that are at least as attractive, then these other theories have more explanatory power. Parfit’s Kantian Contractualism, Rule Consequetialism, Scanlonian Contractualism, and Virtue Ethics may or may not  be otherwise defensible, but at least they offer (albeit different) unified ­explanations of why morality generally opposes harming others, stealing, coercion, promise breaking, failing to give special weight to one’s family and friends, etc. As long as Parfit’s Kantian Contractualism, Rule Consequentialism, Scanlonian Contractualism, and Virtue Ethics have implications that are either the same as or at least as intuitively attractive as the implications of Ross's pluralist deontology and Act-­Involving Act Consequentialism, the greater explanatory power of Parfit’s Kantian Contractualism, Rule Consequentialism, Scanlonian Contractualism, and Virtue Ethics gives them important advantage over both Ross’s pluralist deontology and Act-­Involving Act Consequentialism. Parfit did not end up endorsing Act-­Involving Act Consequentialism. He accepted that it loses out to Rule Consequentialism. Section 4 puts his arguments for Rule Consequentialism under the spotlight.

224  Brad Hooker

4.  The Arguments for Rule Consequentialism in On What Matters, Volume III The end of On What Matters contains three arguments for Rule Consequentialism. One of the arguments for Rule Consequentialism that Parfit explicitly makes is the familiar argument that Rule Consequentialism slots into reflective equilibrium with our other moral beliefs. Here is his most explicit statement of the argument: We may start by accepting what seem to us the most plausible principles of Common Sense Morality. We then ask whether these principles can all be given  some further justification, which may appeal to some feature that these principles have in common. On one plausible hypothesis, the best principles of Common Sense Morality are also the principles whose acceptance would on the whole make things go best. We might justifiably accept this hypothesis. These beliefs would then support a wider theory which combined this version of Common Sense Morality with Rule Consequentialist justification. These two parts of this wider theory would achieve more by being combined. Rule Consequentialism would be strengthened if this theory supports what seems to be the best version of Common Sense Morality. The version of Common Sense Morality would be similarly strengthened if it can be plausibly supported in this Rule Consequentialist way.  (OWM iii. 433; see also 421–2, 450)

A familiar line of objection to thzis argument is that, even if Rule Consequentialism has practical implications that are extensionally equivalent with Common Sense Morality’s requirements and permissions, Rule Consequentialism’s explanations of these requirements and permissions strikes us as counter-­intuitive. This line of  objection rejects Parfit’s ‘On one plausible hypothesis, the best principles of Common Sense Morality are also the principles whose acceptance would on the whole make things go best. We might justifiably accept this hypothesis.’ I will not pause to consider the line of objection, since I have tried to answer it elsewhere.18 Parfit’s second argument for Rule Consequentialism begins with the observation that humanity now faces many challenges, and if we go on behaving as we have done, the cumulative effects will be terrible for others, especially future generations. In particular, Parfit is considering cases in which an individual’s contributing to global warming or polluting the environment is morally wrong but this one individual’s action has very small or even imperceptible effects on anyone else. Parfit explores the possibility that we can make progress with such problems

18  See my ‘Ross-­style Pluralism versus Rule-­Consequentialism’, Mind 105/420 (1996): 531–52; Ideal Code, Real World, ch. 1; and ‘Reflective Equilibrium and Rule Consequentialism’, in B.  Hooker, E.  Mason, and D.  Miller (eds.), Morality, Rules, and Consequences (Edinburgh University Press, 2000), 221–38.

Parfit ’ s Final Arguments  225 by thinking in more careful act-­consequentialist ways. In this discussion, Parfit is developing earlier lines of thought he explored in chapter 3 of Reasons and Persons (pp. 67–86). In volume iii of On What Matters, however, when Parfit concludes this discussion of the wrongness of acts that harm very many but harm each to only a very small or even imperceptible degree, he admits that the explanation that points to the aggregate of many small harms caused by the individual’s act is inferior to an explanation that appeals, ‘not to the separate effects of particular acts, but to the combined effects of what we and others together do. Some act would be wrong, we believe, if all optimific rules would condemn such acts’ (OWM iii. 432). Parfit continues, the wrongness of these acts can be best explained in this Rule Consequentialist way. We are considering acts that would be very slightly worse for each of very many people. . . . Imperceptible amounts of pain, and other such harms, seem to most of us to be below any plausible threshold of moral significance. If we are Rule Consequentialists, however, we deny that each of these acts is made to be wrong by this act’s effects. These acts are wrong, we believe, when and because they are condemned by optimific rules. Whether some rule is optimific depends on whether things would on the whole go better if most of us, or many of us, accepted and tried to follow this rule. Because these claims are not about the effects of single acts they are not challenged by the fact that, in the cases we have been considering, no single act would have perceptibly bad effects. (OWM iii. 432)

This argument that the wrongness of acts that are very slightly worse for each of very many people is better explained by Rule Consequentialism than by Act Consequentialism was foreshadowed by Parfit’s remark in Reasons and Persons, ‘These are the cases where we naturally say, “What if everyone did that?” ’ (RP, 66). And he pointed out in Reasons and Persons that most views about the nature of morality, including Kantianism and Rule Consequentialism, share the idea ‘that morality is essentially a collective code’ (RP, 106). If moral requirements, prohibitions, and permissions are part of a collective code, we should consider the consequences of collective acceptance of and compliance with this code. Furthermore, Reasons and Persons contained an arresting discussion of the attractions of ‘Collective Consequentialism’ (RP, 30–1). These attractions not only were taken up in work by Liam Murphy, Tim Mulgan, and me but also retained their appeal to Parfit to the end.19 19  See e.g. Murphy, Moral Demands in Nonideal Theory (Oxford: Oxford University Press, 2000); Mulgan, Future People: A Moderate Consequentialist Account of Our Obligations to Future Generations (Oxford: Oxford University Press, 2003); and my Ideal Code, Real World.

226  Brad Hooker I will conclude this section by pointing to a third argument for Rule Consequentialism that Parfit presents at the end of volume iii of On What Matters. This argument starts by considering which patterns of motives and dispositions would make the world go best. Naively, we might think that the world would go best if everyone were most keenly concerned to maximize the impartial good. However, as Parfit writes, If most of us were pure Act Consequentialists who were most strongly motivated to do whatever would make things go best, our acts would have many good effects. But our lives would on the whole go better if most of us had some other strong motives and try to follow some other policies or rules. It is good, for example, that most of us strongly love our close relatives and some friends. Having such love and being loved are some of the greatest goods in most people’s lives. If instead we cared equally about everyone’s well-­being, we would have no strong love for anyone, but only what Aristotle called ‘watery kindness’. This would not be how things could go best.  (OWM iii. 420)20

The conclusion here is that, if the choice is limited to either (a) most people’s accepting Act Consequentialism and having the motivation that matches it or (b) most people’s accepting some improved version of Common Sense Morality along with the motivations it allows, then things would go better if most people accept some improved version of Common Sense Morality. And, as we have seen, Parfit maintains that the motives and rules that some improved version of Common Sense Morality would endorse would also be the optimific motives and rules—that is the ones whose acceptance would make things go best. And so these would also be the motives and rules endorsed by Rule Consequentialism. Hence, Parfit’s succinct formulation of his third argument for Rule Consequentialism: What matters most is how well things go. Things would on the whole go best if we have optimific motives and we accept and tried to follow optimific rules. Therefore, We ought to have such motives and we ought to try to follow such rules. (OWM iii. 432)

20  Here Parfit is drawing on Sidgwick, Methods of Ethics, 434. The point is one that Parfit has long accepted—see RP, 27–8, 30. Worth comparing is Garrett Cullity’s The Moral Demands of Affluence (Oxford: Oxford University Press, 2004), ch. 7.

Parfit ’ s Final Arguments  227 Parfit goes on to consider a possible objection to the second premise of this argument. This premise considers only two possible scenarios. In the first scenario, most people accept Act Consequentialism and have the corresponding motivations. In the second scenario, most people accept Common Sense Morality and Rule Consequentialism and have the corresponding motivations. The objection is that there is a third scenario that should have been considered. This is a scenario in which there is a ‘mixed population’, a population in which some people accept some improved version of Common Sense Morality and Rule Consequentialism while other people are pure Act Consequentialists. Thinking about a mixed population is a thought experiment that was given its classic formulation over 150 years ago by Sidgwick.21 It is also a thought experiment periodically posed in Parfit’s work (RP, 30; OWM iii. 414). I do not mean to suggest that the only possible ‘mixed-­world’ scenario worth considering is the one in which some people accept some improved version of Common Sense Morality and Rule Consequentialism while other people are pure Act Consequentialists. Which mixed-­ world scenarios are worth considering seems to me too large a question to be addressed in a paper focused on Parfit’s arguments. The only mixed-­world scenario on which Parfit comments is the one in which some people accept some improved version of Common Sense Morality and Rule Consequentialism while other people are pure Act Consequentialists. And so this is the mixed-­world scenario targeted here. The justification Parfit offers late in volume iii of On What Matters for rejecting this scenario is as follows: I am assuming that . . . everyone ought to have the same moral beliefs. Moral truths are not true only for certain people.  (OWM iii. 420)

The conjunction of these two sentences is surprising. Parfit is one of the people who has been most influential in arguing that, for Act Consequentialists, which moral beliefs are true is one question and which moral beliefs people ought to have is a different question (RP, 40–1; OWM iii. 415–16). Nevertheless, near the end of On What Matters, Parfit affirms that everyone ought to have the same moral beliefs, and he connects that idea with the idea that moral truths are universal. What exactly is this connection between moral truth and sameness of belief? Parfit’s first remark, ‘everyone ought to have the same moral beliefs’, is about moral belief. His second remark, ‘moral truths are not true only for certain people’, is about moral truth. Presumably, the remark about universal moral truth is meant as justification for the remark about everyone’s sharing the same beliefs. 21 Sidgwick, Methods of Ethics, 489–90.

228  Brad Hooker On that presumption, I offer the following reconstruction of the reasoning behind the idea that everyone ought to have the same moral beliefs because moral truths are not true only for certain people: (a)  Moral truths are not true only for certain people. (b)  So moral truths are true for everyone. (c)  If moral truths are true for everyone and if moral beliefs of some people conflict with those of other people, then at least some people’s moral beliefs must be false. (d)  So, if moral truths are true for everyone, then at least some people’s moral beliefs must be false unless either everyone’s moral beliefs are the same or different people’s moral beliefs are different but do not conflict. (e)  If moral truths are true for everyone, the only way for different people’s moral beliefs to be different and yet not conflict is for the different sets of moral beliefs to be incomplete. (For example, you have moral beliefs about A, B, C, D, E, F, G, but not about H, I, or J, and I have moral beliefs about A, B, C, D, H, I, J, but not about E, F, and G.) (f)  Although everyone cannot reasonably be required to have a complete set of true moral beliefs, everyone ought to have a complete set of true moral beliefs about what the correct moral rules are (and so about what properties of acts are morally positive or morally negative).22 (g)  If everyone has a complete set of true moral beliefs about what the correct moral rules are, then everyone has the same moral beliefs about what the correct moral rules are. (h)  So everyone ought to have the same moral beliefs about what the correct moral rules are.

This argument is one I am suggesting might have somehow been in the background of Parfit’s ‘I am assuming everyone ought to have the same moral beliefs. Moral truths are not true only for certain people.’ Let me explain why premise (f) focuses on ‘moral beliefs about what the correct moral rules are’. As (f) begins by acknowledging, any finite being cannot reasonably be required to have a complete set of true moral beliefs. A complete set of moral beliefs would have to include not only beliefs about the correct moral rules but also beliefs about all the more fundamental principles that singly or jointly ground the correct moral rules, or alternatively the belief that there is nothing more fundamental than the correct moral rules (the belief that is a large part of what I’m calling Ross’s pluralist deontology). Parfit would have difficulty relying 22  Here I have been helped by a suggestion from Todd Karhu.

Parfit ’ s Final Arguments  229 on a premise that includes the idea that everyone should have true moral beliefs about fundamental principles. That idea is in tension with Parfit’s apparent acceptance that different people can accept different fundamental moral principles as long as there is convergence at the level of moral rules. Kantian Contractualists have one fundamental moral principle. Rule Consequentialists have a different one. Scanlonian Contractualists have yet another. Likewise for Virtue Ethicists. Rossian Pluralists deny there is any such single founational principle. Parfit does not seem especially bothered by disagreement at the fundamental level. That was an ad hominem argument against Parfit’s having as a premise that everyone ought to have a complete set of true moral beliefs. Now I want to offer a second objection to the idea that everyone should have a complete set of true moral beliefs. This second objection should persuade everyone, not just Parfit. The objection is: (P1)  If everyone ought to have a complete set of true moral beliefs, then everyone ought to have a complete set of true moral beliefs about which actions are all-­things-­considered right or wrong in all particular cases. (P2)  It is not true that everyone ought to have a complete set of true moral beliefs about which actions are all-­things-­considered right or wrong in all particular cases. So, it is not true that everyone ought to have a complete set of true moral beliefs.

This argument is logically valid. (P1) must be true, since, unless one has true beliefs about which specific acts are all-­things-­considered morally right in all particular cases, one does not have a fully complete set of true moral beliefs. (P2) is also difficult not to accept. A complete set of true moral beliefs must include not only true moral beliefs about what is all-­things-­considered right in simple and familiar particular cases, but also true moral beliefs about what is all-­ things-­considered right in complicated and unfamiliar cases. Complexity can arise from, among other things, complicated causal connections, conceptual subtleties, and conflicts between moral considerations neither of which obviously trumps the other. Cases might be unfamiliar because they occurred in the distant past or will occur in the changed circumstances of the future, or because they occur in possible worlds that haven’t yet been considered. There is effectively no end of such actual or possible situations. Only an omniscient being could have a complete set of true moral beliefs about which actions are all-­things-­considered right or wrong in all particular cases. Since it is not true that everyone is omniscient, it cannot be true that everyone ought to have a complete set of true moral beliefs about which actions are all-­things-­considered right or wrong in all particular cases.

230  Brad Hooker In short, because of the variety of possible situations, the complexities that can be morally pivotal, and the difficulty of judging which moral reasons outweigh other conflicting reasons in many cases, it is well beyond human capacity to have a complete set of true moral beliefs about what is all-­things-­considered right or wrong in all particular cases. So it is well beyond human capacity to have a complete set of true moral beliefs. Since this is something beyond human capacity, requiring people to achieve it would be preposterous. Remember that Parfit’s view that everyone ought to have the same moral beliefs is relevant to his rejection of the possibility of a mixed population in which some people have one set of moral beliefs (e.g. Act-­Consequentialist ones) and other people have a different set of moral beliefs (e.g. the beliefs that make up Common Sense Morality or Rule Consequentialism). A world with such a mix might be one with greater aggregate welfare and other important goods. But, by the end of volume iii of On What Matters, Parfit dismisses such a world. He continues, ‘If everyone ought to have the same moral beliefs, we should ask whether things would on the whole go better if most of us were pure Act Consequentialists, or we accepted and tried to follow certain other moral rules’ (OWM iii. 420). His answer is that things would on the whole go better if we accept the rules of some improved version of Common Sense Morality. And these are also the rules that Rule Consequentialism underwrites (OWM iii. 421, 433).

5. Conclusion In Reasons and Persons, Parfit appeared to think Act Consequentialism the best ethical theory. In volume i of On What Matters, in contrast, Parfit argued for a Triple Theory combining Kantian Contractualism, Rule Consequentialism, and Scanlonian Contractualism. I have argued here that the reflective equilibrium methodology employed in assessing moral theories militates not merely in favour of a Triple Theory of Kantian Contractualism, Rule Consequentialism, and Scanlonian Contractualism, but in favour of a Quintuple Theory comprised of the best versions of Kantian Contractualism, Rule Consequentialism, Scanlonian Contractualism, Common Sense Morality, and Virtue Ethics. In volume iii of On What Matters, Parfit devotes considerable space to exploring the capacity of what he called Act-­involving Act Consequentialism to accord with our intuitions about the wrongness of some acts that would maximize utility. However, I have here marshalled some arguments against Act-­involving Act Consequentialism. One of Parfit’s arguments for Rule Consequentialism is a reflective equilibrium argument. A second is that, with respect to a very important class of cases, Rule Consequentialism’s explanation of wrongness in terms of ‘the combined effects of

Parfit ’ s Final Arguments  231 what we and others together do’ is superior to Act Consequentialism’s explanation in terms of the effects of an individual’s act. I emphatically endorse these arguments, but I have not here tried to buttress them. A third argument Parfit offers for Rule Consequentialism is surprising. This argument compares a world in which everyone accepts Act Consequentialism with a world in which everyone accepts an improved version of Common Sense Morality or Rule Consequentialism. Why not add to the comparison a world in which some people accept Act Consequentialism and other people accept Common Sense Morality or Rule Consequentialism? Parfit’s answer is that everyone ought to have the same moral beliefs since moral truths are the same for everyone. I have tried to reconstruct what might be the thinking behind that answer.

Bibliography Bales, R. E., ‘Act-Utilitarianism: Account of Right-making Characteristics or Decisionmaking Procedure?’, American Philosophical Quarterly 8/3 (1971): 257–65. Baumann, M., ‘Parfit, Convergence, and Underdetermination’, Journal of Ethics and Social Philosophy 13/3 (2018): 191–221. Broad, C. D., Ethics, ed. C. Lewy (Nijhoff Publishers, 1985). Copp, D., ‘The Rule Worship and Idealization Objections Revisited and Resisted’, in M. Timmons (ed.), Oxford Studies in Normative Ethics 10 (2020), 131–55. Cullity, G., The Moral Demands of Affluence (Oxford: Oxford University Press, 2004). Frick, J., ‘Contractualism and Social Risk’, Philosophy and Public Affairs 43/3 (2015): 175–223. Harrod, R. F., ‘Utilitarianism Revised’, Mind 45/178 (1936): 137–56. Hooker, B., ‘Rule-consequentialism, Incoherence, Fairness’, Proceedings of the Aristotelian Society 95/1 (1995): 19–35. Hooker, B., ‘Ross-style Pluralism versus Rule-Consequentialism’, Mind 105/420 (1996): 531–52. Hooker, B., ‘Reflective Equilibrium and Rule Consequentialism’, in B.  Hooker, E.  Mason, and D.  Miller (eds.), Morality, Rules, and Consequences (Edinburgh University Press, 2000), 221–38. Hooker, B., Ideal Code, Real World: A Rule-consequentialist Theory of Morality (Oxford: Oxford University Press, 2000). Hooker, B., ‘Rule-Consequentialism and Internal Consistency: A Reply to Card’, Utilitas 19/4 (2007), 514­–19. Hooker, B., ‘Must Kantian Contractualisn and Rule-Consequentialism Converge?’, in Mark Timmons, (ed.), Oxford Studies in Normative Ethics 4 (2014): 34–52.

232  Brad Hooker Mulgan, T., Future People: A Moderate Consequentialist Account of Our Obligations to Future Generations (Oxford: Oxford University Press, 2003). Murphy, L., Moral Demands in Nonideal Theory (Oxford: Oxford University Press, 2000). Parfit, D., Reasons and Persons (Oxford: Oxford University Press, 1984). Parfit, D., On What Matters, volumes i and ii (Oxford: Oxford University Press, 2011). Parfit, D., On What Matters, volume iii (Oxford: Oxford University Press, 2017). Portmore, D., Commonsense Consequentialism: Wherein Morality Meets Rationality (Oxford: Oxford University Press, 2011). Ross, W. D., The Right and the Good (Oxford: Oxford University Press, 1930). Rüger, K., ‘On Ex Ante Contractualism’, Journal of Ethics and Social Philosophy 13/3 (2018): 240–58. Scanlon, T.  M., What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998). Scheffler, S., The Rejection of Consequentialism (Oxford: Oxford University Press, 1982). Sidgwick, H., Methods of Ethics (7th edn, London: Macmillan, 1907). Urmson, J.  O., ‘On the Interpretation of the Moral Philosophy of J.  S.  Mill’, Philosophical Quarterly 3/10 (1953): 33–9. Wolf, S., ‘Two Concepts of Rule Utilitarianism’, in M. Timmons (ed.), Oxford Studies in Normative Ethics 6 (2016): 123–44.

10

Parfit on Act Consequentialism Katarzyna de Lazari-Radek and Peter Singer

1.  Introduction: How Parfit Did Not Reject Act Consequentialism In the preface to the first volume of On What Matters, Derek Parfit states that his two masters are Sidgwick and Kant, an act consequentialist and a deontologist. Though Parfit is clearly influenced by Sidgwick, the most philosophically sophisticated of the classical utilitarians, and mentions him a few times in volume i (especially with reference to the dualism of practical reason), act consequentialism plays little role in the discussion of normative ethics in the first two volumes of On What Matters. The normative thread of these volumes is Parfit’s argument for what he calls the Triple Theory. He examines three major normative the­or­ies— Kantianism, Contractualism, and Consequentialism—and seeks, by ex­amples and arguments, to show that in order to be defensible, Kantian and Contractualist theories need to be revised. He then turns to Consequentialism, focusing on the Rule Consequentialist principle: ‘Everyone ought to follow the principles whose universal acceptance would make things go best.’ Parfit argues that Rule Consequentialism, properly understood, identifies as wrong the same set of acts that are so identified by Kantianism and Contractualism. These are acts that are disallowed by principles that (a) would make things go best, (b) are uniquely universally willable, and (c) cannot reasonably be rejected. At the end of volume i of On What Matters, Parfit indicates that the fact that these three theories converge is important, because it is widely believed that there are deep disagreements between Kantians, Contractualists, and Consequentialists. If there were such deep disagreements, adherents of each view would have grounds to doubt their own view, and even to doubt whether ‘any of us could be right’—that is, whether there are objective truths in ethics. But, Parfit tells us, the disagreements are not deep: ‘These people are climbing the same mountain on different sides.’1

We thank two anonymous reviewers for Utilitas for comments that have improved this paper. 1  Derek Parfit, On What Matters (Oxford: Oxford University Press, 2011), i. 419.

Katarzyna de Lazari-Radek and Peter Singer, Parfit on Act Consequentialism In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0011

234  Katarzyna de Lazari-Radek & Peter Singer It would be easy to conclude from all this that Parfit believes that the most defensible form of consequentialism is Rule Consequentialism. That is in fact what we concluded on first reading volumes i and ii of On What Matters. The omission of any serious discussion of Act Consequentialism disappointed us. But when Peter Singer sent Parfit a draft of his review of On What Matters for the Times Literary Supplement, we learned that our belief that Parfit thought Rule Consequentialism to be the most defensible form of consequentialism was in error. Commenting on a passage in the review, Parfit replied: ‘I don’t in fact argue that the most defensible form of Consequentialism is Rule Consequentialism.’ Parfit pointed to a passage in On What Matters, volume i, in which he raised an objection to the Rule Consequentialist view that acts are wrong when they violate the rules that, if universally accepted, would make things go best. In that passage, Parfit had written: Some Rule Consequentialists appeal to the claim that (Q) all that ultimately matters is how well things go. This claim is in itself very plausible, and is not challenged by any of the arguments I have given . . . But if Rule Consequentialists appeal to (Q), their view faces a strong objection . . . We can plausibly object that, if all that ultimately matters is how well things go, it could not be wrong to do what we knew would make things go best.2

In his response to Singer, Parfit also commented that ‘on page 418, I deliberately refrain from saying that we ought to accept the Triple Theory. I write that we ought to accept (S) and (T), but only that we have strong reasons to accept (R) and (U).’3 The principles Parfit refers to are as follows: (S)  Everyone ought to regard everyone with respect, and never merely as a means. Even the morally worst people have as much dignity or worth as anyone else. (T)  If all of our decisions are merely events in time, we cannot be responsible for our acts in any way that could make us deserve to suffer, or to be less happy. (R)  Everyone ought to treat everyone only in ways to which they could rationally consent.

2 Parfit, On What Matters, i. 417. 3  This and the following passages are quoted from an attachment to an email sent by Parfit to Singer, 9 May 2011. Note also that, as explained below, in On What Matters, volume iii Parfit returns to the passage just quoted, saying that in it he made ‘a bad mistake’ (Parfit, On What Matters (Oxford: Oxford University Press, 2017), iii. 419).

Parfit on Act Consequentialism  235 Finally, (U) is the important principle for the Triple Theory. It reads: (U)  Everyone ought to follow the principles, whose being universal laws would make things go best, because these are the only principles whose being universal laws everyone could rationally will.

Summarizing what he was doing, Parfit wrote to Singer: ‘What I argue is that, if Kantianism and Contractualism are revised in ways that are clearly needed, these theories coincide with some forms of Consequentialism.’ Parfit made another comment on the draft review that is important for understanding what he is doing when he discusses Act Consequentialism in the third volume of On What Matters. Singer wrote that Parfit seemed to assume that the answer to the question ‘What ought I to do?’ is to be found by asking what prin­ ciples would, if universally accepted, make things go best. Against this, Singer drew on work we had done earlier defending Sidgwick’s idea of esoteric morality, pointing out that whether it would sometimes be right to torture (for example when torturing a terrorist is the only way to get him to reveal the location of a ticking nuclear time-bomb hidden in the midst of a city), and whether it is desirable for everyone to believe that torture is sometimes right, are two different questions. Singer then wrote: We may suspect that Parfit runs these issues together because accepting a form of consequentialism that requires us to consider the consequences of individual acts, rather than of rules or universally accepted principles, would make it much more difficult for him to achieve convergence between consequentialism and its rivals.

To this, Parfit commented: That’s true. I might reply that this convergence can take a broader form. If Kantian and Contractualist theories, in their best forms, imply some version of Rule Consequentialism, we don’t have three kinds of conflicting systematic theory. We have importantly different versions of the same kind of theory. If this argument shows that Kantian and Contractualist theories imply Rule Consequentialism, Act Consequentialists can then argue that we should make a further move.

Parfit’s discussion of Act Consequentialism in the third volume of On What Matters should be seen as developing the ‘broader form’ of convergence hinted at in this paragraph—a form of convergence between Rule Consequentialism and Act Consequentialism. We shall begin our account of it by discussing the distinction Parfit makes between Impartial Rationality and Morality.

236  Katarzyna de Lazari-Radek & Peter Singer

2.  Impartial Reason as Morality’s Rival The conflict between egoism and morality is familiar to all of us. Sidgwick used the term ‘ethics’ in a broad sense to refer to any rational method of deciding what we have most reason to do. In this sense of ‘ethics’ egoism is a method of ethics, and the conflict between egoism and universal benevolence is a conflict within ethics. Parfit takes a narrower view of morality, and regards egoism as an external rival to morality. There is nothing new about that. But in the third volume of On What Matters, Parfit suggests that it is not only personal reasons that may be in conflict with morality. There can also be conflicts, he says, between moral and impartial reasons. ‘We should ask’, he writes, ‘whether in such cases we might sometimes have sufficient reasons or decisive impartial reasons to act wrongly.’4 He answers that this view is plausible, not in its strong form, in which we always have decisive reasons to do whatever would make things go best, but in the form of what he calls: Weak High Stakes Rational Impartialism:  When some wrong act would make things go much better than all our other possible acts would do, we would sometimes have sufficient or even decisive impartial reasons to act wrongly in this way.

Before we go on to consider an example, we pause to ask why Parfit raises this possibility of a conflict between moral and impartial reasons. The answer, we believe, is that it serves his goal of reducing disagreements between different normative theories. Sometimes we think we are disagreeing, but the disagreement is only in appearance, because we are giving different answers to different questions. So the egoist who offers an answer to: ‘What is the rational thing for me to do?’ may not be disagreeing with the moralist who is asking: ‘What is the right thing for me to do?’ But the same thing may be true of the Impartial Rationalist, who does not ask ‘What is the right thing for me to do?’ but rather ‘What do I have impartial reasons for doing?’ As Parfit writes, ‘Impartial Rationalism may be an external rival to our moral view, since these Rationalists may believe that, in some cases, we would have sufficient or decisive reasons to act wrongly.’ The advantage of this move, for Parfit, is that it reduces the scope of moral disagreements. If we think of Act Consequentialism as a moral theory, then it is in disagreement with any moral view that holds that it is sometimes wrong to do what makes things go best. This disagreement is internal to morality. Parfit will subsequently provide some considerations for modifying Act Consequentialism so as to reduce the extent of this disagreement. We will examine these con­sid­er­ ations shortly. But Parfit’s other strategy for reducing moral disagreement is to

4 Parfit, On What Matters, iii. 344.

Parfit on Act Consequentialism  237 distinguish Impartial Rationalism from Morality, in the same way as we commonly distinguish egoism from morality. Impartial Rationalists may, he says, ‘agree with [those who are not Act Consequentialists] about which acts are right or wrong’. They are answering a different question, the question of what we have impartial reasons to do. In the ticking bomb scenario, for example, an Impartial Rationalist might agree that torture is always wrong, but believe that there are sufficient, or perhaps even decisive, impartial reasons for torturing the terrorist. If the separation of Morality from Impartial Rationalism has this advantage for Parfit, however, it has a corresponding serious disadvantage, of which he must have been well aware. It risks reducing the significance of morality. Parfit distinguishes what he calls ‘The Moralist’s Question’ (‘What ought I morally to do?’) from ‘The Rationalist’s Question’ (‘What do I have most reason to do?’) and then writes: The Rationalist's Question is, I claimed, more fundamental. That can be shown by supposing that we have decisive reasons to act wrongly, or that we have no reason to care whether our acts are wrong. Morality would then have a status that is like that of etiquette, or some code of honour. If we had no reason to do what we ought morally to do, that would undermine morality. There could be no such truth the other way round. If it would be wrong for us to do what we have decisive reasons to do, that would not undermine these decisive reasons.5

For morality to be ‘undermined’, we assume that Parfit means that its importance as a guide to conduct would be diminished. In the case of personal or partial ­reasons, Parfit seems to think that morality would not be seriously undermined if we always have sufficient reason to do what morality requires, even if we did not always have decisive reason to do it. We were critical of this in The Point of View of the Universe, arguing that morality is still seriously undermined if I often have sufficient reasons to act morally and sufficient reasons to act self-interestedly, for I may then often choose to act self-interestedly, and I am not acting contrary to any decisive reasons.6 Now the move to separate impartial reasons from morality undermines the position of morality even more seriously, for I may reject morality on the basis of impartial reasons as well as on the basis of personal and partial reasons, and in doing so, I am still not acting contrary to decisive reason. In addition, we can doubt whether, if Parfit were to succeed in distinguishing a sphere of impartial reasons, separate from morality, this would help his overall project of reducing the scope of moral disagreement. Many moral theorists take the fact that morality provides overriding reasons for action as an important or 5 Parfit, On What Matters, iii. 177. 6  Katarzyna de Lazari-Radek and Peter Singer, The Point of View of the Universe (Oxford: Oxford University Press, 2014), 162–3.

238  Katarzyna de Lazari-Radek & Peter Singer even essential element of the concept of morality. Unless Parfit can convince these theorists that they are wrong, his separation of morality from impartial reasons seems doomed to—at best—close one area of disagreement while opening up another.7 Another troubling question can also be asked: when morality is opposed by impartial reason, why do we even have sufficient reason to act morally? Parfit offers this example of a conflict between morality and impartial reason: The US President, we can suppose, can choose between two ways of ending the Second World War. The President has true beliefs about what would be the main effects of these two policies. In what we can call the Nuclear Policy, an atomic bomb would be dropped on one city in Japan, where it would kill about 100,000 civilians, most of them immediately or very soon. This policy would swiftly end the war. The Japanese Government would soon surrender, because the use of this vastly destructive new weapon would give the generals in the Japanese Government what they would believe to be an ­honourable way to admit defeat. In what we can call the Conventional Policy, the US armed forces would instead invade Japan and would win the war by fighting in conventional ways. These methods would involve much bombing and fighting in highly populated areas, whose un­avoid­able side effects would be to kill at least 300,000 civilians before Japan surrendered.8

The Nuclear Policy involves terror bombing, deliberately killing civilians. Their deaths are aimed at, as a means of persuading the Japanese Government to surrender. In the Conventional Policy, the deaths of the civilians are foreseen but unintended side-effects. The president therefore has a distinctive deontological reason to choose the Conventional Policy, but he has an impartial reason to cause 200,000 fewer people to be killed. Parfit acknowledges the difficulty of comparing such different reasons, but adds: This impartial reason, we may believe, would outweigh this deontological ­reason. On this view, the President would have a decisive impartial reason to act in a way that was very wrong. That conclusion would be disturbing.9

We agree that this conclusion should be disturbing. To acknowledge that, on an issue as important as this, one can have a decisive impartial reason to act in a way 7  We thank Ketan Ramakrishnan for this point.    8 Parfit, On What Matters, iii. 346–7. 9 Parfit, On What Matters, iii. 348.

Parfit on Act Consequentialism  239 that is very wrong must diminish the importance of morality. We would argue that what this move gains in making it easier to reconcile conflicting moral views, is lost in making morality less important as a guide to conduct. In a section called ‘Wrongness and Reasons’, Parfit indicates how he would have responded to this point. He repeats his earlier statement that the question about reasons is more fundamental than the question about morality, but adds that we have good reasons to go on asking the Moralist’s Question, saying: We have distinctive beliefs about which acts are wrong, and about what makes them wrong, and these aren’t beliefs about which facts give us decisive reasons.10

He goes on to say that we can ask, first, whether there are ‘morally decisive r­ easons not to act in some way’. If there are, we can then ask the further question ‘whether these morally decisive reasons are decisive all things considered, by outweighing any conflicting non-moral reasons’. That may be true, but it invites the question whether these distinctive beliefs about which acts are wrong and what makes them wrong, and the morally de­cisive reasons to which they give rise, are beliefs we ought to continue to hold, if they do not give us decisive reasons to act. Parfit concludes the section by writing: It is better, I believe, to use the word ‘wrong’ . . . so that our beliefs about which acts are wrong retain their independence from our beliefs about what we have decisive or sufficient reasons to do. I shall therefore ask whether there are other ways in which we could avoid or weaken the conclusion that we often have sufficient or decisive reasons to act wrongly. We reach these disturbing conclusions when it would be best if we or others act wrongly. I shall ask how often that is true.11

This remark leads to our next section.

3.  Deontic Badness We now turn to Parfit’s discussion of Act Consequentialism as a moral theory, that is, as an internal rival to all moral views that hold that, in some cases, it would be wrong to do what would make things go best. This discussion is in the final four chapters (55–8) of On What Matters, volume iii.

10 Parfit, On What Matters, iii. 367.

11 Parfit, On What Matters, iii. 368.

240  Katarzyna de Lazari-Radek & Peter Singer Parfit’s overarching aim is to show that we are able to ‘resolve at least some of the disagreements between Act Consequentialism and the pluralist moral views that most of us accept’.12 This would, of course, bring Act Consequentialism closer to our ordinary conceptions of morality, and therefore, to the Triple Theory. One way in which Parfit seeks to advance this aim is through the distinction between deontic and non-deontic badness: ‘Acts are deontically bad’, he suggests Act Consequentialists might say, ‘when and because these acts are wrong.’ All other kinds of badness are non-deontic.13 This view requires Act Consequentialists to hold that it is intrinsically—and non-deontically—bad to treat people in certain ways, such as deceiving or co­er­ cing them, and when these acts are wrong, because the intrinsic badness of the deceit or coercion is not justified by other good effects of the act, ‘their wrongness makes them also deontically bad’.14 We accept that this is a possible position for a pluralist consequentialist to hold. But how plausible is it? Parfit offers two reasons for thinking it plausible. The first, which he claims even a utilitarian could accept, involves denying that well-being needs to be limited to conscious states like happiness or pleasure, or to the satisfaction of desires. It can, Parfit suggests, include our achievements, and also our moral goodness or lack thereof. The second invites us to imagine two different worlds in which everyone has the same overall level of wellbeing, but everyone in the first world is morally good, and everyone in the second world is morally bad. If we accept that moral badness makes someone’s life go worse, we can spe­cify that, independently of that, the well-being of those in the second world is sufficiently better in non-moral ways to compensate for this. Parfit, following Ross, suggests that we might believe that the existence of the morally bad people would be less good than the existence of the good people. We might think this to be the case even if the morally bad people had a higher level of well-being.15 These reasons do not persuade us that consequentialists should include de­ontic badness or goodness as intrinsic values. We hold that well-being is limited to conscious states like happiness or pleasure, or to the satisfaction of desires. In fact, we agree with Sidgwick that it is ‘Desirable Consciousness which we must regard as ultimate [or as we prefer to say, intrinsic] Good’.16 Deontic norms are instrumentally valuable when they conduce to this intrinsic good, and otherwise not. On Parfit’s second reason for holding that moral goodness and badness are intrinsic values, we share the intuition on which his example relies. That is, if we 12 Parfit, On What Matters, iii. 366. 13 Parfit, On What Matters, iii. 400–1. 14 Parfit, On What Matters, iii. 401. 15 Parfit, On What Matters, iii. 402. 16  Henry Sidgwick, The Methods of Ethics (7th edn., London: Macmillan, 1907), 397.

Parfit on Act Consequentialism  241 are asked to say whether we feel more positively about the world in which everyone is morally good than we do about the world in which everyone is morally bad, we acknowledge that this is the way we feel, even though the level of well-being is the same. But we regard this intuitive response as a tenuous thread on which to hang a judgement about what is of intrinsic value. Our moral intuitions are shaped by, among other things, the way in which we are taught what is right and what is wrong. As we will see in Section  4, consequentialists have long acknowledged that the consequences of teaching people to follow simple moral rules may be better than the consequences of teaching them to calculate, with each choice they face, which act will have the better consequences. One way to teach and motivate people to follow moral rules is to inculcate in everyone, from an early age, the belief that certain acts are right and people who do them deserve to be rewarded, whereas other acts are wrong and people who do them deserve to be punished. That teaching, we suggest, is why most of us have the intuition that a world with morally good people is intrinsically better than a world with morally bad people in it, and why raising the well-being of those who are morally bad beyond the level of those who are morally good does nothing to assuage the feeling that the world with morally good people is a better world. (At least, this is likely to be our intuition if we assume that well-being in these worlds is positive.) Note, though, that this is an explanation of our intuitions, not a justification of them. The explanation debunks the intuition that there is intrinsic goodness or badness in certain acts, irrespective of the consequences of those acts.17

4.  Act Consequentialism as Self-Effacing In the final chapter of the third volume of On What Matters, Parfit returns to the possibility of a normative theory being self-effacing, which he first discussed in part I of Reasons and Persons. He points out that many Act Consequentialists assume that things would go best if we always tried to do what could be expected to make things best, but this assumption may be false: Our attempts to make things go best would often fail. We would misjudge or overlook the effects of many possible acts. It is also easy to believe falsely that we could make things go better by doing what would be better for ourselves, or for those we love.

17  For more on debunking explanations in ethics, see Lazari-Radek and Singer, The Point of View of the Universe, 179–96.

242  Katarzyna de Lazari-Radek & Peter Singer Even just being motivated along the lines that pure Act Consequentialism implies might cut us off from some of the greatest sources of happiness, for ourselves and others: If we cared equally about everyone’s well-being, we could not strongly love any particular people, and we would have to lose many of the aims, concerns, and ambitions on which much of our happiness depends.18

Parfit admits that this is not universally true: if rich people, who are hundreds or thousands of times wealthier than the poorest people in the world, were to follow Act Consequentialism, and give away most of their wealth, that would make things go best. But for many other people, he thinks, acceptance of Act Consequentialism would, on the whole, make things worse.19 These ideas are not new. Sidgwick thought that the Utilitarian conclusion, carefully stated, would seem to be this; that the opinion that secrecy may render an action right which would not otherwise be so should itself be kept comparatively secret; and similarly it seems expedient that the doctrine that esoteric morality is expedient should itself be kept esoteric. Or if this concealment be difficult to maintain, it may be desirable that Common Sense should repudiate the doctrines which it is expedient to confine to an enlightened few.20

This was, he thought, a ‘necessary evil’ given the ‘seductive impulses’ that lead us away from morality, and the varying degrees of moral and intellectual development to be found in the world. Parfit quotes most of the above passage from Sidgwick, and notes that Bernard Williams objected to Sidgwick’s view on the grounds that it is elitist and paternalist. He responds that this objection might be justified if, as Williams believed, there are no moral truths. On that point, Parfit’s claim can be questioned, given that Williams’s objection seems to be a moral one. If there are no moral truths, what significance is there in objecting to a view on the grounds that it is elitist and paternalist?21 On the other hand, if there are moral truths as Sidgwick believed, then Parfit’s further comment is relevant: It is not elitist or paternalistic to believe that the true moral view implies that it would be better if most people did not accept this view. This view would be depressing. But we cannot rationally assume that depressing moral beliefs ­cannot be true. 18 Parfit, On What Matters, iii. 414. 19 Parfit, On What Matters, iii. 414–15. 20 Sidgwick, The Methods of Ethics, 490. 21  We thank Roger Crisp and Tyler Paytas, each of whom independently suggested this point to us.

Parfit on Act Consequentialism  243 We agree, at least to the extent of accepting that there can be true or justified moral views entailing that it is better if some moral truths are not widely known, and such views are not necessarily elitist or paternalistic. Parfit immediately follows the statement just quoted with the suggestion that if Act Consequentialism is self-effacing in this way, that counts against it: Though Act Consequentialism could not be directly shown to be false by the fact that it would be better if most people rejected this view, this fact might indirectly help to show that some other moral view is true.22

As this sentence occurs at the end of a section, it isn’t immediately clear in what way Parfit thinks that Act Consequentialism’s partial self-effacement might help to show that some other moral view is true. In the following section (183), Parfit introduces Rule and Motive Consequentialism: RMC: We ought to have optimific motives and we ought to accept and try to follow optimific rules.23

Parfit argues that, in part for reasons already mentioned, the optimific motive is not that of always wanting and trying to act in optimific ways. Instead, ‘things would go on the whole better if, rather than always wanting and trying to act in optimific ways, we had certain other motives and tried to follow certain other rules’. Thus Act Consequentialism and Rule and Motive Consequentialism are incompatible. If things would on the whole go best if most of us had these other motives—that is, motives other than the motive of trying to do the most good— and tried to follow rules other than the rule of doing what can be expected to bring about the most good, then Rule and Motive Consequentialism would be the optimific moral view. Parfit then concludes this section with the comment: ‘This fact would give us some reasons to accept Rule and Motive Consequentialism, but it would not show that Rule and Motive Consequentialism is true. We should not assume that an optimific view must be true.’24 An initial point to make here is that Parfit writes of reasons to accept Rule and Motive Consequentialism, but accepting a theory is not the same as adhering to it, and it is possible that we could accept Rule and Motive Consequentialism, in the sense of believing it to be the true moral theory, without having optimific motives and without trying to follow optimific moral rules. So from the fact that if we did have optimific motives we would do more good than if we were Act Consequentialists, it does not follow that we have some reason to accept Rule and Motive Consequentialism. This would only follow if the link could be made 22 Parfit, On What Matters, iii. 416. 24 Parfit, On What Matters, iii. 418.

23 Parfit, On What Matters, iii. 418.

244  Katarzyna de Lazari-Radek & Peter Singer between believing the theory to be true, and adhering to it. Similarly, it does not follow from the fact that we accept that we ought to try to follow optimific rules, that we will actually follow them. So the reasons that Parfit has offered for accepting Rule and Motive Consequentialism are no stronger than the reasons for believing that these links hold. If we focus not on what moral view would have optimific consequence if everyone accepted it in the sense of believing it to be true, but rather on what moral view would have optimific consequences if everyone succeeded in adhering to it, then it is not clear that that view would be Rule and Motive Consequentialism. It too might be partially self-effacing. Perhaps adherence to Common Sense Morality would have better consequences, overall, than adherence to Rule and Motive Consequentialism, because if everyone accepted the latter, many of them might persuade themselves that convenient violations of the common-sense rules are justifiable, even when those violations are not in fact optimific. If Rule and Motive Consequentialism may also be self-effacing, however, then the fact that Act Consequentialism may be self-effacing cannot give us any reasons to accept Rule and Motive Consequentialism rather than Act Consequentialism.25 Moreover, given the distinction Parfit himself makes between a moral theory being optimific and it being true, we are still lacking an explanation of the final sentence of the preceding section (182), where Parfit said that the fact that it might be better if most people rejected Act Consequentialism ‘might indirectly help to show that some other moral view is true’. Parfit begins the next section (184) by referring to the passage in the first volume of On What Matters that he quoted in his comments on Peter Singer’s draft review, and which we have already cited, but will cite again here: Some Rule Consequentialists appeal to the claim that (Q) all that ultimately matters is how well things go. This claim is in itself very plausible, and is not challenged by any of the ­arguments I have given . . . But if Rule Consequentialists appeal to (Q), their view faces a strong objection . . . We can plausibly object that, if all that ultimately matters is how well things go, it could not be wrong to do what we knew would make things go best.

Parfit now says: ‘These claims made a bad mistake. What I called this “strong objection” applies only to some versions of Rule Consequentialism.’26 Parfit then refers to other familiar reasons why one might follow Rule Consequentialism rather than try to calculate what would make things go best—for example, de­tect­

25  We owe points in the preceding paragraph to Jake Nebel (personal communication). 26 Parfit, On What Matters, iii. 419.

Parfit on Act Consequentialism  245 ives in some counter-terrorist police force might have good reasons to f­ ollow the rule ‘Never use torture’ because they are too ready to believe that the benefits of using torture outweigh the costs. This is still compatible with Sidgwick’s defence of esoteric morality, for example with the idea that it would be best if most people accept some improved version of Common Sense Morality, while a few accept Act Consequentialism. But Parfit now introduces a new assumption: Everyone ought to have the same moral beliefs. Moral truths are not true only for certain people.27

This enables him to ask whether things would on the whole go better if most of us were pure Act Consequentialists, or we accepted and tried to follow certain other moral rules. From here, a path opens towards the latter, Rule Consequentialist, choice. We will not follow that path in detail. Some of it, for example the valuing of deontic goodness and badness, we have already discussed. Other aspects, such as the ability of Rule Consequentialism to overcome the problem of many imperceptibly small effects leading to great harms, are worthy of more detailed discussion than is appropriate here. Instead, we want to question the assumption that everyone ought to have the same moral beliefs. Parfit supports this assumption by saying that moral truths are not true only for certain people. Indeed, no truths are ‘true only for certain people’. This common phrase is just a misleading way of putt­ing certain facts. I may say, for example, that it is true for me (Katarzyna de Lazari-Radek, a philosopher working at the University of Lodz in 2020) that strawberries are more desirable than raspberries, and true for you (Peter Singer, a philosopher working at Princeton University in 2020) that raspberries are more desirable than strawberries. But it is true for everyone that the specific Katarzyna identified above likes strawberries better than raspberries and true for everyone that the specific Peter identified above likes raspberries better than strawberries. Still, what does this have to do with whether everyone ought to have the same moral beliefs? Parfit has just been arguing that if most people believe that Rule and Motive Consequentialism, or perhaps some form of improved Common Sense Morality (which might turn out to be identical with Rule and Motive Consequentialism), is the true moral view, this belief is likely to have better consequences than if most people believe that Act Consequentialism is the true moral view. But we take it that the ought in ‘Everyone ought to have the same moral beliefs’ is a practical ‘ought’, and a consequentialist one at that. It is not an epistemic ‘ought’, and it is not supported by the claim that ‘Moral truths are not true only for certain people’. So we think that the path that opens for Parfit when we makes the assumption that everyone ought to have the same moral beliefs is a

27 Parfit, On What Matters, iii. 420.

246  Katarzyna de Lazari-Radek & Peter Singer path that leads him away from a discussion of which is the true moral theory, and into a discussion of which moral theory it would be optimific for everyone to accept. If his claim is that this is an improved form of Common Sense Morality, and that this coincides with some form of Rule Consequentialism, we are not concerned to oppose this claim. But this does not ‘help to show that some other moral view is true’. Against that claim, we need only to repeat what Parfit has himself said: ‘We should not assume that an optimific view must be true.’

Bibliography Lazari-Radek, K. and Singer, P., The Point of View of the Universe (Oxford: Oxford University Press, 2014). Parfit, D., On What Matters, vols. i–ii (Oxford: Clarendon Press, 2011). Parfit, D., On What Matters, vol. iii (Oxford: Clarendon Press, 2017). Sidgwick, H., The Methods of Ethics (7th edn., London: Macmillan, 1907).

11

Nonlegislative Justification Liam Murphy

1.  Reductionist Nonconsequentialism If we take for granted that familiar moral constraints on the treatment of other people cannot be explained away in terms of the beneficial consequences of adopting certain standing dispositions and deliberative rules of thumb, the question remains as to what kind of explanation and justification of them is available. Can more be said than that it is self-­evident to ‘thoughtful and well-­educated ­people’ that certain ways of treating people are wrong?1 As most moral philosophers are in this respect Rawlsians now, most answer yes. Moral theory starts but does not end with a statement of considered judgments; it aims to provide a set of principles or a moral conception that matches our considered judgments in reflective equilibrium. Even philosophers such as Judith Jarvis Thomson and F.  M.  Kamm, who, unlike Rawls, insist that some considered judgments about particular cases must be treated as fixed, nonetheless see the point of moral theory as providing explanation and justification for those beliefs. Thomson writes, invoking Socrates, that ‘while a cluster of beliefs may be a cluster of true beliefs . . . ,

This paper derives from part of my 2017 Herbert Morris Lecture in Law and Philosophy at the University of California, Los Angeles (UCLA). For discussion I am grateful to the audience at UCLA, as well as to audiences at Binghamton University, Toronto University, the Free University, Berlin, the University of Texas, Austin, Union College, York University, Ontario, and New York University. My understanding of Scanlon’s contractualism benefitted greatly from discussion with students in my graduate seminar at New York University’s Department of Philosophy, Spring 2019. I have benefitted particularly from feedback from Stefan Gosepath, Moshe Halbertal, Paul Horwich, Rahul Kumar, Sophia Moreau, Robert Myers, Thomas Nagel, Ketan Ramakrishnan, Joseph Raz, Lawrence Sager, Samuel Scheffler, Seana Shiffrin, Jeremy Waldron, and Jake Zuehl. I am especially grateful to T. M. Scanlon for extensive correspondence. I started out in moral philosophy engaging with part I of Derek Parfit’s Reasons and Persons, with a particular interest in ‘what we together do’; without his generous mentorship later on I doubt I would have gone on with the subject. My first ideas on the themes of this paper emerged when co-­teaching with Derek at NYU in 2006. I greatly regret that I was not able to discuss the paper itself with him. The financial assistance of the Filomen D’Agostino and Max  E.  Greenberg Research Fund of the New York University School of Law is gratefully acknowledged. 1 W. D. Ross, The Right and the Good (Oxford: Oxford University Press, 1930), 41.

Liam Murphy, Nonlegislative Justification In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0012

248  Liam Murphy knowledge that they are true requires knowledge of what makes them true’.2 And Kamm, who insists that the first step in moral theory is the formulation of intricate moral principles as generalizations of ‘as many case-­based judgments . . . as prove necessary’, holds that we cannot conclude that any principle is correct unless we find that it ‘expresses some plausible value or conception of the person or relations between persons’.3 Thomas Nagel well expresses the standard view: ‘Common sense doesn’t have the last word in ethics or anywhere else, but it has, as J. L. Austin said about ordinary language, the first word: it should be examined before it is discarded.’4 A rather different approach has been explored by T. M. Scanlon since ‘Rights, Goals, and Fairness’ in 1975.5 His ‘reductive’ strategy does not start from con­ sidered judgments about particular cases or proposed principles as data. The project has been to find an intuitively compelling unified account of the domain of nonconsequentialist principles as a whole, reasoning to particular conclusions about moral principles from within the terms of that account. In a recent paper, Scanlon writes: In developing my contractualist view, I was following Aristotle’s model. I was using the method of reflective equilibrium to identify the contractualist pro­ced­ ure of justification, which I thought gave the best account of (at least a portion of) morality, and then taking this procedure to be a way of reasoning ‘from first principles’ about what the content of this morality is and why.6

Scanlon does indicate that the overall plausibility of his contractualist method will depend not just on its appeal—in reflective equilibrium—as a general account of what makes certain ways of treating people wrong, but also on the intuitive acceptability of its outputs. Some principles may just seem obviously reasonably rejectable, and the task would then be to figure out what the grounds for this could be. Moreover, in some instances the structure of the contractualist method itself might have to be tweaked if it is the only way to block unacceptable results for particular cases.7 Still, the primary direction of argument is to judgments of right and wrong rather than from them. 2  The Realm of Rights (Cambridge, MA: Harvard University Press, 1990), 31. 3 F. M. Kamm, Intricate Ethics (Oxford: Oxford University Press, 2007), 5, 469. 4  The View from Nowhere (Oxford: Oxford University Press, 1986), 166. 5  Repr. in The Difficulty of Tolerance (Cambridge: Cambridge University Press, 2003), 26–41. 6 ‘Contractualism and Justification’, presented at the 2016 Lauener Conference, ‘Themes from Thomas  M.  Scanlon’, MS available at https://www.academia.edu/35861169/Contractualism_and_ Justification_pdf, 5. 7  See e.g. What We Owe To Each Other (Cambridge, MA: Harvard University Press, 1998), 242–6. An example of changing the structure of the contractualist framework would be moving from an ex post to an ex ante perspective on the rejectability of principles. Such a revision is argued for in Johann Frick, ‘Contractualism and Social Risk’, Philosophy & Public Affairs 43 (2015): 175–223; Scanlon indicates his openness to the revision in ‘Contractualism and Justification’, 31.

Nonlegislative Justification  249 One advantage of this direction of argument is that a moral theory that shows how principles of right and wrong—‘deontic’ principles—can be derived from normative considerations that are not themselves deontic will potentially explain more, be more illuminating. The more familiar approach, seeking principles together with considered judgments about particular cases in reflective equilibrium, may in the end suggest to us a compelling nondeontic normative account of why those are the right principles, but it also may not. The reductive approach starts with such an account: there needs to be a master idea to identify and organize the nondeontic considerations that lead us to de­ontic conclusions. This is the sense in which the account is reductive, explaining all deontic principles by appeal to the same core considerations. In Scanlon’s contractualism, the master idea is justifiability of our actions to those they affect.8 If the idea that wrongness is best understood and explained in terms of justifiability is not appealing, some other master idea is required for a reductive approach to be viable. A plausible master idea will have to navigate what Scanlon calls Pritchard’s Dilemma:9 the idea must not be too external to morality, for example reducing it to enlightened self-­interest, but nor must it simply state deontic prin­ ciples seriatim, thus not producing a reductive account at all. Scanlon has long used the example of welfarist consequentialism as a possible competing master idea: the idea that wrongness of action must always in the end be explained in terms of the impact on people’s welfare. Scanlon mentions a second important appeal of a reductive approach: it provides ‘a way of explaining and interpreting incompletely specified exceptions’10 to familiar nonconsequentialist principles. This certainly would be good to have. Especially the most important nonconsequentialist commitments, such as a right not to be killed, turn out to be extremely difficult to pin down at the level of casuistry. Kamm’s method of as many case-­based judgments as prove necessary leaves many of us feeling that our ability to judge or even consider has reached its limit and that the only way to continue is to appeal to some principle or general method for thinking about such cases. On the face it, then, Scanlon’s reductive strategy is very appealing. It would be good to have something more to say than that treating someone in such-­and-­such a way is obviously wrong, and it would be good to have a method for resolving puzzling cases other than that of appeal to ‘intuitions’ about cases and principles that, for many of us, either don’t exist or can only be understood as the outputs of some imperfectly understood theory. I find Scanlon’s master reductive idea appealing as 8  Rainer Forst plausibly suggests that (at least as a moral theorist) Scanlon is better characterized as a ‘justification fundamentalist’ rather than a ‘reasons fundamentalist’—as Scanlon characterized himself in Being Realistic about Reasons (Oxford: Oxford University Press, 2014); see Forst, ‘Justification Fundamentalism: A Discourse-­Theoretical Interpretation of Scanlon’s Contractualism’, presented at the 2016 Lauener Conference, ‘Themes from Thomas M. Scanlon’. 9  What We Owe to Each Other, 150. 10  ‘Contractualism and Justification’, 8.

250  Liam Murphy well: with justification as the core idea, we reason to nonconsequentialist moral principles by thinking about what kinds of reasons we could offer others in justification for our actions, taking into account our recognition of the reasons they also have to ‘object to being affected in certain ways’.11 By contrast, the consequentialist reduction seems dogmatic and too substantive to serve as a starting point for thinking about how we may treat each other. And, looking in the other direction, Kant’s ambition in the Groundwork to derive the content of the moral law from nothing more substantive than the conditions of rational agency seems clearly hopeless.12 So the thought is that treating people rightly means being able to justify my actions to them, by appeal to the reasons I have for acting in that way and taking into account the reasons they have to object. But it is a crucial part of Scanlon’s contractualism that this justification must not just be interpersonal or relational, a conversation between the affected parties only. Justification requires that there is a principle allowing my act that can be reasonably rejected by no one, not even those unaffected by it. The reason for this wide field of justification, as I will call it, is that we are not just looking for a principle appropriate for cases like this in circumstances like these, but one ‘for the general regulation of behavior’. What this means for Scanlon is that we are not looking just for a general principle, applying to any relevantly similar case, but for a principle that we imagine is generally followed, or accepted. So in addition to justifying my action to the persons immediately affected, I must also consider objections currently unaffected persons may have to the prospect of my proposed principle being generally accepted or followed.13 But why exactly should justification run via a principle for the general regulation of behavior, so understood? As Scanlon explains, any answer to the question of how I may treat you now will take the form of a principle: To justify an action to others is to offer reasons supporting it and to claim that they are sufficient to defeat any objections that others may have. To do this, however, is also to defend a principle, namely one claiming that such reasons are sufficient grounds for so acting under the prevailing circumstances.14

A principle in this sense is general, but it is not yet a principle for the general regulation of behavior, and defending it does not require a wide field of justification. The wide field of justification is forced upon us only if we accept that we are looking for principles for the general regulation of behavior rather than a general

11  ‘Contractualism and Justification’, 7. 12  On this, see Scanlon, ‘How I Am Not a Kantian’, in Parfit, On What Matters (Oxford: Oxford University Press, 2011), ii. 116–39. 13  What We Owe to Each Other, 202–4. 14  What We Owe to Each Other, 197.

Nonlegislative Justification  251 principle that states reasons that are sufficient grounds for acting in a certain way in the prevailing circumstances. As Scanlon writes, to justify a principle for the general regulation of behavior, we must take into account not only the consequences of particular actions, but also the consequences of general performance or nonperformance of such actions and of the other implications (for both agents and others) of having agents be licensed and directed to think in the way that that principle requires.15

We must think about the consequences of general performance or nonperformance and the other implications of ‘having agents be licensed and directed to think in the way that that principle requires’ because we are considering whether anyone would have grounds to object if that principle were generally followed or accepted. Relevant consequences of general performance would include, for instance, the anxiety that would be caused if behavior were regulated by a prin­ ciple that allowed one to be sacrificed, by force, for the sake of many others, even if any one person’s chances of being sacrificed were very low.16 Derek Parfit brings out the dramatic significance of this feature of contractualism: ‘in the thought-­ experiment to which the Kantian Formula appeals, I would have the power to choose which principles everyone would accept, both now and in all future centuries. The principles I chose would be accepted by many billions of people.’17 Principles for the general regulation of behavior require a wide field of justification; general principles for cases like this in circumstances like these do not. Principles for the general regulation of behavior are also subject to design constraints that do not apply in the same way to principles for cases like this in circumstances like these. For the latter, my principle can be as complex as the case requires in the circumstances. Not so principles that can serve for the general regulation of behavior. These have to be tractable and thus suitably broad. ‘[F]iner-­grained principles will create more uncertainty and require those in other positions to gather more information in order to know what a principle gives to and requires of them.’18 A general principle for cases like this in circumstances like these just is a statement of my justification for treating you in a certain way.19 By contrast, it is not immediately clear why a defense of a principle for the general regulation of behavior which allows me to treat you in that way should be thought of as a justification for my treatment of you at all. When thinking about principles for the 15  What We Owe to Each Other, 203. 16  See Parfit, On What Matters, ii. 210. 17 Parfit, On What Matters, i. 382. 18  What We Owe to Each Other, 205. 19  The description of the circumstances might include information about me, such as my strength or wealth, so that a person with different strength or wealth might be subject to a different principle. These facts about me might affect the reasons I can offer another for or against being required or permitted to act in a certain way.

252  Liam Murphy general regulation of behavior we are thinking as legislators do, about what would, on balance, be the best rule for everyone to follow. We take into account all possible effects of general compliance, and we fashion the norms in such a way that they can efficiently guide behavior. None of that seems to be relevant when one person is trying to justify treating another in a particular way in particular circumstances. Scanlon writes that, ‘According to contractualism, thinking about right and wrong is in one respect like thinking about the civil and criminal law: it involves thinking about how there is reason to want people in general to go about deciding what to do.’20 This is precisely the aspect of Scanlon’s contractualism that I want to resist. For my case, then, the appeal of Scanlon’s master idea—as the core of a reductionist way of reasoning about nonconsequentialist constraints about how people may be treated—depends on us thinking about general principles for cases like this in circumstances like these. The introduction of a mediating level of prin­ ciples for the general regulation of behavior changes the subject and introduces factors that seem on the face of it to be irrelevant to the issue of how one person may treat another in given circumstances.21 The same point applies to R. Jay Wallace’s embrace of Scanlon’s contractualism as illuminating Wallace’s own relational conception of morality.22 Wallace’s relational conception has it that we should understand morality, or at least the morality of right and wrong, in terms of directed duties, the violation of which wrongs persons who have moral claims not to be treated in certain ways. I myself strongly doubt that the morality of right and wrong can all be accounted for in relational terms. But the relational account does seem to be appealing for the specific case of nonconsequentialist constraints on how people may be treated. The affinity between the relational account and Scanlon’s view turns on Scanlon’s master idea, that of justifiability. In Wallace’s account, relational moral norms identify situations where people have claims as against others to be treated in certain ways. The process of contractualist justification seems precisely tailored to the identification of such claims: ‘An individual has a claim against the agent that the agent should comply with a candidate moral principle, just in case the personal interests of that individual make it reasonable for someone in his or her position to reject alternative principles for the general regulation of behavior.’23 Wallace remarks that, in responding to Parfit’s insistence that Scanlon should drop his requirement that principles can be rejected only by appealing to personal

20  What We Owe to Each Other, 153. 21  Here, as elsewhere in this paper, I am largely in agreement with Adam Hosein, ‘Contractualism, Politics, and Morality’, Acta Analytica 28 (2012): 495–508, and Hanoch Sheinman, ‘Act and Principle Contractualism’, Utilitas 23 (2011): 288–315; many of the themes in this paper were also raised in Thomas Pogge, ‘What We Can Reasonably Reject’, Philosophical Issues 11 (2001): 118–47. 22  The Moral Nexus (Princeton: Princeton University Press, 2019), 178–89. 23  The Moral Nexus, 180.

Nonlegislative Justification  253 reasons, Scanlon seems ‘surprisingly open’ to the idea of allowing aggregation and appeal to impersonal reasons. Even Scanlon’s recent idea24 that while the number of people affected by a principle cannot itself be a reason for rejecting a principle, it might be relevant to whether rejection is reasonable, seems to Wallace to weaken contractualism’s appeal: We were looking for an interpretation of moral reasoning that would elucidate the idea that moral obligations are owed to particular individuals, who have claims against the agent to compliance with the principles that determine them. Those elements are most firmly in place if what resolves the question of whether a given individual can reasonably reject a principle are the comprehensive implications of the principle for the life of that very individual, as compared to the similar effects of the alternative principles on the lives of other individuals.25

But if that is what we were looking for, then we surely do not want what resolves the question of what someone can reasonably reject to include considerations of how general compliance with the principle would shape the social world, or considerations of (as it were) moral drafting, having to do with limits to how fine-­ grained a principle for the general regulation of behavior can usefully be. The appeal of contractualism, I agree with Wallace, is that it focuses our thinking, when we are thinking about the directed obligations we have to others in virtue of their rights or claims to be treated in certain ways, to what seems to be at stake—a comparison of my reasons for acting in a certain way and their reasons to object. But if we are told that what we are doing is engaging in hypothetical moral legislation, that appeal is lost.26 Consider this striking statement of Scanlon, connecting the contractualist idea to the value of human life: respecting the value of human (rational) life requires us to treat rational creatures only in ways that would be allowed by principles that they could not rea­ son­ably reject insofar as they, too, were seeking principles of mutual governance which other rational creatures could not reasonably reject.27

24  In ‘Contractualism and Justification’. 25  The Moral Nexus, 180. 26  Clara Lingle, ‘Generic Objections, Actual Claims: A Problem for Contractualism as Wronging’, unpublished MS 2020, points out that if what I am calling a wide field of justification is used in a theory such as Wallace’s, which appeals to contractualism to provide an account of wronging, or relational morality, it has the consequence that failing to follow certain principles wrongs everyone, not just the particular victim. As Lingle notes, Wallace acknowledges this consequence (The Moral Nexus, 207 and 219) but feels that it is compatible with the relational account. I agree with Lingle that if we accept that consequence the distinctive explanatory potential of the relational account is lost. If anyone who might benefit from compliance by others with a principle has a relational claim against those who fail to comply, then utilitarianism can be given a relational characterization. 27  What We Owe to Each Other, 106.

254  Liam Murphy As Kamm notes, this is Scanlon at his most Kantian.28 Scanlon later aligns ­himself with Kant’s claim that the first and second formulations of the categorical imperative come to the same thing. Interpreting the ‘universal law formula’ along contractualist lines (‘What general principles of action could we all will?’), treating others as ends in themselves must be understood to be just the same thing as justifying our actions by appeal to general principles of action. It seems very plaus­ible that to treat rational agents with proper respect, as ends in themselves, just is to recognize what Rainer Forst calls their right to justification.29 But to require that that justification take the form of figuring out a legislative code of moral rules for all of humanity to live by, taking into account all kinds of secondary effects and drafting issues—that, I think, drains the association between human value and the claim to justification of its plausibility. It might be thought, however, that principles for cases like this in circumstances like these have no explanatory power; that they offer us little help in our quest to say more about nonconsequentialist constraints than that they seem clearly correct. For perhaps no two cases and sets of circumstances are sufficiently alike for this approach to yield any general principles at all.30 But it is extravagant to claim that for the purposes of justification—the consideration of reasons for and against the permissibility of acting in some way—no two cases and sets of circumstances are relevantly alike. Just as in common-­law reasoning about the application of precedent, some differences in cases and circumstances do justify distinguishing the cases and thus not applying a principle, but other differences in cases and circumstances are not relevant. It is therefore appropriate, when offering reasons for and against a principle, to consider the possible cumulative ‘intrapersonal’ burdens it would entail.31 After all, if I am to accept the principle for this case and circumstances, I should be prepared to follow it when relevantly similar cases and circumstances arise in the future. It is also relevant to consider hypothetical variations on the case and the circumstances to see how a suggested principle would fare in such a situation. This is not because we are looking for a principle to cover both situations—a prin­ ciple of sufficient breadth to be workable, teachable, and enforceable that will necessarily have to smooth over some relevant differences. The point of considering hypothetical variations, once again akin to common-­law reasoning, is to help us sharpen our understanding of the reasons we are bringing to bear in the actual

28  ‘Owing, Justifying, and Rejecting’, Mind 111 (2002): 325–54, at 327. 29  The Right to Justification (New York: Columbia University Press, 2012). 30  See Jonathan Dancy, Ethics without Principles (Oxford: Oxford University Press, 2004). Though I reject Dancy’s particularism, I wholeheartedly agree that it is a mistake to ‘think of morality as a sort of social device, a human institution that has got set up for a purpose’ (133) and that ‘the model of an antecedently accepted legal system . . . should not be applied to ethics’ (134). 31  See Scanlon, What We Owe to Each Other, 237; ‘Contractualism and Justification’.

Nonlegislative Justification  255 case and circumstances.32 What is not relevant is the impact of the principle in counterfactual circumstances of its general acceptance. General principles for acts like this in circumstances like these do have ex­plana­tory power. They help us identify the reasons that people may reasonably advance in favor of treating others in some way, and the reasons the others may reasonably advance in objection. As an example, take the issue of responsibility for reliance that one has invited or encouraged. A possible principle, R, is that if one person invites another to rely on their stated intentions, and the other person does rely, then the first person must do what they can to prevent that reliance coming at a loss.33 To some a principle like this is very plausible; to others, it seems surely false. Considering the matter by way of Scanlon’s reductionist method is illuminating and offers the prospect of advancing beyond a crude battle of intuitions. The relying person has obvious grounds to reject a principle permitting the reliance-­inviter to walk away from the induced losses. And rejection of R by the inviter would not be appear to reasonable: Yes, it is now inconvenient and perhaps costly for the inviter to take responsibility for the losses, but the relier has the losses just because of what the inviter did, and the inviter got themselves into this by doing something they didn’t have to do.34 The great merit of Scanlon’s reductionist project is to have us focus on this kind of comparative evaluation of reasons, rather than going directly—blindly—to a verdict on the wrongness or not of the action or the soundness of the principle.35 Removing the factor of the impact of general acceptance of a principle on the social world will likely mean that thicker or more moralized reasons than those having just to do with relative burdens will have to be invoked.36 But this appears likely to be necessary even under the model of hypothetical legislation.37 This is not the place to try to establish the fruitfulness of the method of nonlegislative justification across the board, but it is not obvious that it will fail. What I do

32  This point, and the earlier one about the application of precedent, shows that there is a sense in which this nonlegislative version of Scanlon’s contractualism is nonetheless not antilegalist—the weighing of parties’ reasons for and against a principle is precisely what common-­law judges do, at least when they are considering matters of principle rather than policy—in Ronald Dworkin’s sense of those terms (see Taking Rights Seriously (London: Duckworth, 1977) ). 33  This is a more restrictive and, I think, more plausible principle than Scanlon’s Principle L, What We Owe to Each Other, 300–1. 34 Wallace, The Moral Nexus, 181, goes through this argument, but wrongly takes it to explain why promises bind. It cannot do this, as promises bind even in the absence of reliance. 35  As indicated earlier in this section, there is room in Scanlon’s method for taking intuitive plausibility judgments into account as a kind of secondary check on possible principles. 36  For example, Rahul Kumar offers as a reason someone might reject a principle requiring sacrifice whenever this will benefit another more, the reason that this will greatly reduce the control people have over their lives. ‘Defending the Moral Moderate’, Philosophy & Public Affairs 28 (1999): 275–309, 298. Lingle argues that if appeal to what the social world would be like under full compliance is not part of the process, the contractualist method of justification will simply rehearse existing de­onto­ logic­al intuitions. I am not as pessimistic as Lingle, but the danger is clearly real. 37  See Scanlon, ‘Contractualism and Justification’.

256  Liam Murphy believe to be clear is that the appeal of the method of justification is lost if it is understood to involve hypothetical legislation. In embracing a nonlegislative form of Scanlon’s method of justification, I have so far been discussing deontological constraints, or what Thomson calls the realm of rights. In What We Owe to Each Other, Scanlon clearly has in mind a theory with broader scope, something like an account of impartial moral rules governing interactions among humans; some morally significant matters such as friendship and the treatment of animals and nature are excluded, but the domain of imperfect impartial duties, such as beneficence, is included.38 There is some reason to object that the strategy of interpersonal justification is appropriate only in the domain of perfect right and directed duty—that it is best taken as an account of wronging.39 It might be thought to follow from the fact that violations of imperfect duties do not wrong anyone in particular that no particular person has grounds to object to any proposed principle of imperfect duty. This objection does not apply to the legislative version of Scanlon’s method, however, because if we are asking about the effects of general compliance with a principle of beneficence, more or less everyone will have a stake. In discussing his principle of rescue, Scanlon writes that ‘we have to consider the general costs (and benefits) of its acceptance’.40 But the objection would seem to apply to the nonlegislative version. It would be regrettable to conclude that some moral questions about what ­people owe to others are illuminated by the method of justification while for others that method is entirely inapt. It is true that the domains of deontology and beneficence are distinct. The former is the home of perfect agent-­relative duties, the latter of imperfect agent-­neutral duty. But it would be good if those distinctions emerged in the process of justification itself, rather than serving as meth­odo­logic­al­ly and morally prior ground rules that we have to make sense of in some other way.41 The challenge then is to see if the nonlegislative method of justification can accommodate a principle of beneficence after all. Here is a rough sketch of how the distinctive features of a principle of beneficence might emerge naturally in the course of a nonlegislative justificatory back and forth. Suppose A accepts a moderately demanding imperfect duty of beneficence. B may object that this doesn’t do him any good, because it allows A to do nothing to benefit B in particular. A better principle for B, no more burdensome on A, would require A to direct resources to B alone. But, of course, everyone else that might benefit from A’s resources can make the same complaint as B, and could also reasonably reject a principle that favors B alone. It might next be said that, in that case, everyone should be benefitted equally. But that principle is 38 See What We Owe to Each Other, 171–87, 218–23. 39  See Kamm, ‘Owing, Justifying, and Rejecting’, 333–6. 40  What We Owe to Each Other, 225. 41 In Nagel’s discussion of agent-­relativity and agent-­neutrality, both agent-­relative and agent-­ neutral reasons are recognizable from the impartial point of view. See The View from Nowhere, ch. 9.

Nonlegislative Justification  257 reasonably rejectable by worse-­off people on the ground that they have stronger claims. A principle that calls for A to seek out the worst-­off people and direct her resources to them could be proposed; but that principle is impracticable (she is an individual, not a state) and therefore wasteful, and could be reasonably rejected by anyone on those grounds. What seems likely to emerge from the process of justification is a principle that requires A to direct her resources in one of the many ways that, given what she knows or has reason to know, will produce the most expected weighted benefit for the dollar. This is a principle of imperfect duty. Now B may grant that the principle of beneficence should state an imperfect duty but object next to the moderate demands of A’s favored principle. After all, B’s chances of receiving a benefit will be greater, the greater the extent to which A is required to transfer resources to worse-­off people. B may say that a reasonable principle would require A to give up resources until the point where further transfers would do more harm than good in terms of expected weighted benefit. Contractualists have resisted this conclusion in a number of ways.42 Nagel concludes that there may be a standoff here: it can be reasonable for the better-­off to reject a principle for being too demanding, and yet reasonable for the worse-­off to reject any less demanding principle.43 Scanlon, in saying that ‘we have to consider the general costs (and benefits)’ of the acceptance of his rescue principle, clearly has in mind intrapersonal aggregation, since he believes that the principle that you should prevent dire harm to others if it can be done for only slight or moderate sacrifice would have to take into account previous contributions to block unlimited sacrifice over time.44 But it also suggests that we evaluate the demandingness of a proposed principle on the assumption of its general acceptance. This would allow the demands to be set reasonably low, as is the case for his rescue principle, since if everyone better-­off complied, the need for the beneficence of each of them would be so much less than in the actual world of minimal compliance. If we are not entitled to imagine general acceptance when evaluating prin­ ciples, B’s latest objection to A’s moderate principle, and to Scanlon’s moderate rescue principle, seems powerful.45 But A does have a reply within a nonlegislative approach. She could concede that better-­off people as a group should continue to transfer resources up to the margin, but object to a principle that requires her to do this as an individual in a situation where other better-­off people are contributing less than their part. Just 42 See Elizabeth Ashford, ‘The Demandingness of Scanlon’s Contractualism’, Ethics 113 (2003): 273–302. 43  Thomas Nagel, Equality and Partiality (Oxford: Oxford University Press, 1991), 48–51. 44  What We Owe to Each Other, 224. 45  See Ashford, ‘The Demandingness of Scanlon’s Contractualism’; Kamm, ‘Owing, Justifying and Rejecting’, 338–9. Rahul Kumar defends a moderate contractualist duty of aid in part by insisting on the fact that we are discussing principles for the general regulation of behavior that will shape our social world, in ‘Defending the Moral Moderate’, 296.

258  Liam Murphy as A is not responsible for B’s welfare in particular, but rather for that of worse-­off people in general, B has no claim on A in particular, but a claim on better-­off people collectively. If B has needs, it makes no moral difference if they are met by A or some other well-­off person. If for some reason more good will be done by others if A stands aside and does nothing, A acts rightly by standing aside. Any potential beneficiary could object to an agent-­relative principle of beneficence on the ground that it will tend to do him less good, and for no reason. So a principle of beneficence must be agent-­neutral, applying to well-­off people collectively rather than individually. This has implications for the level of sacrifice that can reasonably be required of each member of the collective. If the collective is not doing all it can do, a member of the collective that is doing her part can rea­son­ ably reject a principle that requires her to take up the slack left by others. A reasonable principle of beneficence will not require a person to sustain greater losses than she would do if everyone were doing their part. In the course of this reply by A, an appeal is made to hypothetical general compliance. But this is not to fall back into the legislative approach. First, the appeal to general compliance is grounded in the recognition that agent-­neutral duties of beneficence are collective duties. No reference to general compliance is warranted in the case of justification of agent-­relative duties. Second, the idea is not that we should follow the principle that, were it generally accepted, no one could rea­son­ ably reject. It is rather that, when the issue of the reasonableness of costs comes up in the context of collective duties, it is appropriate to point out that the responsibility of each is to do her part in the collective endeavor, but not more.46

2.  The Wrong Mountain: Legislation, Real and Imagined Scanlon’s contractualism is of course not the only moral theory that employs the model of morality as hypothetical legislation; the other main example is rule consequentialism. Defenders of that view are comfortable, vocal even, about the le­gis­la­tive analogy: we are asking what moral code would be best.47 What Parfit calls Kantian contractualism, a view based on Scanlon’s contractualist reading of the universal law formula of Kant’s categorical imperative, also fits the bill.48 And of course, whether or not that is the right way to interpret the universal law formula, Kant’s language would seem to cast him as the modern philosophical father of the whole school: always moral law (Gesetz—a statutory provision) rather than prin­ciple or rule or norm; legislating for a kingdom of ends. The Kant-­inspired 46 For defense of the position sketched here, see Murphy, Moral Demands in Nonideal Theory (Oxford: Oxford University Press, 2000). 47  See Richard Brandt, A Theory of the Good and the Right (Oxford: Oxford University Press, 1979); Brad Hooker Ideal Code, Real World (Oxford: Oxford University Press, 2000). 48  On What Matters, vol. i, sect. 52.

Nonlegislative Justification  259 discourse-­theoretic theories of Habermas and Forst work with the same model. Habermas’s discourse principle is a general principle for the justification of all norms, be they legal or moral: ‘Only those norms of action are valid to which all possibly affected persons could assent as participants in rational discourses.’49 As Forst elaborates, norms from different ‘contexts of justification’—ethical, moral, political, legal—will require different kinds of justification, but the result is the same: valid general norms.50 Nicholas Southwood’s Habermas-­inspired ‘deliberative’ contractualism is cast as method for determining a ‘common code’ to live by, thus appropriating the explicitly legislative language of rule consequentialists.51 Not all these theorists are offering constructivist or reductive theories in Scanlon’s sense. Brad Hooker, for example, falls in the more traditional camp of offering rule consequentialism as a moral conception that matches our con­ sidered judgments in reflective equilibrium.52 And on the level of substantive detail too there will be considerable divergence. But in so far as they all share a conception of moral theorizing as a search for an ideal set of moral principles or rules for the general regulation of behavior, then there is a sense in which they are all, as Parfit says, climbing the same mountain.53 In my view, all of these moral theories are fundamentally flawed. The model of morality as hypothetical legislation should be rejected wherever it is found. When a legislature makes law, it should consider the reasons for and against the provisions it enacts. Those reasons relate to the effects of the provisions on people—the benefits and burdens and forms of treatment they involve. Attention needs to be paid not just to direct effects but also to secondary or systemic effects. Legislation obviously requires a wide field of justification. Legislation setting up a scheme of random selection of healthy people for compelled organ donation burdens not just the unlucky few, but all healthy people. Discriminatory treatment of some people on the grounds of membership in some group affects the social standing of all members of the group, even those never treated in that way. Different kinds of reasons are relevant to different kinds of provision. Often the main issues at stake are social welfare and its distribution. Examples would include legislation concerning public transport or consumer contracting. But sometimes matters of what we might naturally call individual right are involved. Such matters can come before a legislature as ordinary business, or before a legislature or a constitutional convention drafting a bill of rights to serve as a constraint on the permissible content of ordinary legislation. Where legal enshrinement of rights to bodily integrity, liberty of movement, liberty of expression, and so on

49  Between Facts and Norms, trans. William Rehg (Cambridge, MA: MIT Press, 1996), 459. 50  Contexts of Justice (Berkeley and Los Angeles: University of California Press, 2002); The Right to Justification. 51  Contractualism and The Foundations of Morality (Oxford: Oxford University Press, 2010), 102. 52  Ideal Code, Real World. 53  On What Matters, vol. i, sect. 64.

260  Liam Murphy is the topic, the relevant reasons seem less likely to be social welfare and its ­distribution. The lawmakers may be inclined to appeal directly to their beliefs about natural moral rights. Another option, however, would be to think about the justifiability of the enactment and enforcement of certain legal rights by asking questions about whether anyone could reasonably reject proposed provisions on the grounds of undue relative burdens or unfair treatment. It is clear what lawmakers taking the latter approach would be doing. They would be looking at the way society would be if different possible bills of rights, including none at all, were adopted and enforced, and checking to see whether anyone in each society would have grounds for rejecting the relevant list of rights. The legal rules or principles that enshrine the rights need to be well drafted, of course, with an eye to ease of use by their intended audiences (legislatures, courts, officials). The wide field of justification and the importance of drafting are obvious in this case. It is arguable that the legislative approach is very well suited to the task of figuring out how to put rights into law—how to figure out what legal rights there should be. Scanlon notes that his reductionist approach to moral rights developed out of his reflection on the right to freedom of expression, a political-­legal right.54 The crucial point is that when considering both direct and systemic effects lawmakers can usually assume a high level of compliance with rules they enact. Most of the time, and for most people, both private citizens and state officials, the existence of sanctions for noncompliance will incentivize compliance.55 In some cases, where the person involved is an official high up in the executive branch of government, direct coercive enforcement may not be present; but even then, if institutions are well structured, there will be other pressures encouraging compliance. When thinking about direct and systemic effects, then, legislators usually have good reason to assume a high level of general compliance. The systemic effects must be considered because those effects will be real. Now consider the hypothetical legislation of moral principles or rules. We must think about what the world would be like if everyone followed or accepted a  certain principle or rule and consider how well off people would be, or the grounds individuals would have to reject principles.56 It is not obvious what the general relevance of the effects of compliance with or acceptance of a principle is if people will not, in fact, comply with or accept the principle. Thompson asks, during a discussion of property rights: Is it so much as within reason to suppose that the fact that adopting a certain set of rules would be efficient in one or another of the ways I indicated, or in any

54  The Difficulty of Tolerance, p. xx; ‘Contractualism and Justification,’ 9. 55  On occasion it may be predictable that the level of compliance with a legal rule would be low. This would count against the rule. 56  Like Scanlon, I will from now on focus on these two possibilities.

Nonlegislative Justification  261 way at all, makes it the case that those rules among them that govern property rights are already in force?57

Let us start with an even more extreme example. It is obvious that zoning law in New York City does not enforce independently existing moral rules. You can’t make sense of the rules other than by seeing them as part of a scheme of urban planning that will be good or bad because of its effects—does it make life in the city more affordable, efficient, beautiful, integrated, and so on. Real legislators can reason as follows: It would be good if there were a set of legal rules that required such-­ and-­ such behavior, because such-­ and-­ such behavior would have good effects; therefore, we now enact those rules. They are entitled to reason in that way because they are making law. When you make law you aim to change the world by affecting behavior. Legislators in well-­functioning polities have the power to affect behavior because for the most part people generally comply with the rules they make. Not so the hypothetical moral legislator. Suppose I rightly conclude that a certain zoning code of my devising would be best in terms of consequences, or justification; suppose also that it is compatible with but more restrictive than the existing legal code. It would be mad to point out to my neighbors that the modifications they are making to their home, though legal, are morally prohibited. It is obviously not within reason to think that these best moral zoning rules are already in force in the sense of being generally observed or accepted, since moral legislators have no reason at all to expect that. Hypothetical legislation does not end with the practical acts of enactment, publication, and enforcement. But the idea that that they are in force in the sense of being valid, binding moral rules seems just as mad. This is not just because we think that morality does not speak to such matters as zoning—perhaps because the issues involved are too local. Thomson’s incredulous reaction to the idea that the best property rules are ipso facto already in force—in the sense of being valid moral rules—seems equally ­warranted, and morality does speak to property. But is it perhaps significant that no one believes that there are natural property rights, binding independently of social convention—any more than that there are natural rules of zoning? Perhaps legislative moral theorizing is plausible only for the natural, and not the artificial or convention-­parasitic parts of morality? It is noteworthy in this connection that Scanlon’s illustration of his contractualist method for the case of promising is at the same time an argument against conventionalist theories of promising.58 For one might have thought that there is a rather straightforward contractualist case for a principle of promissory

57  The Realm of Rights, 332–3.

58  What We Owe to Each Other, ch. 7.

262  Liam Murphy fidelity: a rule requiring that promises be kept would, if generally accepted, be beneficial to individuals in creating new opportunities for cooperative behavior, and would not impose greater burdens on anyone than would a principle of promissory permissiveness. But that straightforward argument, like familiar rule-­ consequentialist arguments about promising, relies entirely on the systemic effects of general acceptance of the fidelity principle. The benefits—greater opportunities for co­oper­ation—are prospective; no particular broken promise affects those benefits. There may be reliance losses in a particular case, but that is a different matter. The trouble with the straightforward argument is that it is not within reason to think that if good effects would flow if the principle of fidelity were generally accepted, then I have reason to reject a principle of promissory permissiveness whether or not the fidelity principle is in fact generally accepted.59 Scanlon’s own argument for a principle of promissory fidelity attempts to show that promisees have reason, based in the value of assurance, to reject promissory permissiveness even without considering the systemic benefits that would flow from general acceptance of the fidelity principle.60 There is something slightly paradoxical about the idea that the legislative method can only be employed to account for aspects of morality that are not para­sit­ic on actual social convention; conventionalism and reductionism would seem to go hand in hand. But what consideration of moral phenomena parasitic on actual social convention brings home is the oddness of appealing in moral argument to the effects of counterfactual compliance. If the only thing that can be said in favor of a principle is that general acceptance would bring benefits, the fact that we have no reason to expect general acceptance seems immediately fatal to the argument. Where, on the other hand, reference is made to systemic effects merely to supplement strong reasons individuals have to reject a principle whatever the level of compliance, the objection does not seem so important since it does not undermine the entire case for the principle. But the appeal to the effects of counterfactual levels of acceptance is no more  justified when it is only part of the case for a rule or principle. Take a ­principle permitting forcible but not (in the normal case) fatal harvesting of organs (one kidney, say) to save another’s life. A comparison of burdens on the two people most affected seems to favor this principle, but consideration of the systemic effects of general acceptance changes the calculus dramatically.61 With such a principle generally accepted, we would all be anxious and mistrustful all

59  The example of a conventionalist account of promising shows that the reason why the method of hypothetical legislation seems absurd in the case of property cannot be explained by noting that there are multiple justifiable property schemes. 60  As it happens, I do not believe that the value of assurance can be explained without reference to systemic effects of general compliance with the fidelity principle; see Murphy, ‘The Artificial Morality of Private Law: The Persistence of an Illusion’, University of Toronto Law Journal 70 (2020): 453–88. 61  See discussion in Parfit, On What Matters, ii. 209–10.

Nonlegislative Justification  263 the time; no one would prefer, at least not ex ante, a world in which that principle were generally accepted.62 But the question is just how that fact is at all relevant to assessment of the rightness or wrongness of forcible organ harvesting in a world where most people do not accept the principle allowing this.63 And if the appeal to what things would be like in a counterfactual world where the principle is accepted is valid in this case, why is it not in the case of property, (convention-­ dependent) promise, or zoning? It may be wondered why I am so sure that we should not evaluate principles based on an assumption of general acceptance or compliance. After all, isn’t ‘What if everyone did that?’ a very familiar question in moral argument? It is, but that question generally has force precisely in a context where most people are not doing that. If your child wants to throw the paper out the car window onto generally unlittered streets, the question ‘What if everyone did that?’ is meant to get them to consider the unfairness of free-­riding on others’ beneficial compliance with the no-­littering principle. Compare a situation in which everyone is littering. We could point to the benefits that general acceptance of the no-­littering prin­ ciple would bring, but not by asking ‘What if everyone did that?’ Everyone already does do that and considerations of fair play offer no support for the principle. In Parfit’s terms, the question ‘What if everybody did that?’ is most appropriate in cases where an ‘each-­we dilemma’ has been solved by a social practice. If the each-­we dilemma is unsolved—everyone is littering—we can’t ask ‘What if everyone did that?’, since they are doing that.64 But we could ask instead, ‘Wouldn’t it be better if no one did that?’ This question is most pressing where each person’s act does not by itself cause detectable harm. So instead of littering, we may consider releasing some trace amount of a toxic chemical into a stream. Suppose that if only one person did that, the minimal contamination of the water would do no harm to anyone or anything. But if a hundred people did that, serious harm would result. Here it might be thought that each should refrain from polluting, whatever the others do, and that we can see this by noting that this is what a reasonable principle, assumed to be generally accepted, would require. We do need to be able to explain why each of a hundred polluters acts wrongly, even though the effect of each act alone is not harmful. Parfit once proposed that 62  For a rule consequentialist, the systemic bad effects are obviously decisive. For Scanlon’s contractualism to reach the intuitively right result, it would perhaps have to adopt either an ex ante perspective, or allow the numbers to count. Parfit, On What Matters, vol. ii, sect. 76, discusses aggregation. A nonlegislative Scanlonian justification could be this: ‘It is not unreasonable to refuse to regard one’s own life and body as “on call” to be sacrificed whenever it is needed to save others who are at risk’; Scanlon, ‘Thomson on Self-­Defense’, in Alexander Byrne, Robert Stalnaker, and Ralph Wedgewood (eds.), Fact and Value (Cambridge, MA: MIT Press, 2001), discussed in Parfit, On What Matters, ii. 209. 63  Parfit goes so far as to imagine legal enforcement, invoking the reasons we all have ‘to want not to live in a world in which the police hunt some people down and take their organs by force’, On What Matters, ii. 211. 64 See On What Matters, i. 306.

264  Liam Murphy we appeal here to the idea of ‘what we together do’.65 Each polluter is not acting wrongly, but they together are. I believe that this is the best way to understand these cases, though it is very difficult to come up with a convincing account of when I should foreground the effects of my act considered alone, and when, by contrast, I should foreground the effects of my act considered as one of the acts of a group acting together. In arriving at the right principle for cases like this, it will be appropriate and necessary to take note of what would happen if everybody acted in certain ways. Each polluter needs to see that, though one of them refraining will make no difference, it will make a big difference if they all do. And each needs to consider the moral significance for themselves of that fact. But, as in the case of the discussion of the demands of beneficence above, this reflection on the moral significance of what things would be like if everyone acted in a certain way is quite different from saying that the right principle is the one it would be best for everyone to follow (or the one that, assumed to be generally accepted, could not be reasonably rejected). If the right principle for the pollution case were the principle it would be best for everyone to accept, a single isolated act of harmless pollution by one person would be wrong, even if no one else ever considered polluting—and so there is no free-­riding involved.66 This does not seem plausible. And if instead the rest of the hundred are polluting, it is not obvious that the right decision for a single polluter is simply to stop, since that will not in itself lessen the harm. Perhaps it would by force of example? But suppose it would not. And suppose that not polluting would cost this single polluter some resources. Wouldn’t it be better to keep polluting and spend the saved resources trying to persuade the group to agree to stop polluting?67 Perhaps not. Perhaps the right thing for each member of a group whose acts together cause harm to do is to stop acting in that way, so long as enough others continue to act in that way for the group together to cause harm. Either way, we would miss the complexity of the case if we approached it by asking what prin­ciple it would be best for everyone to follow. The way in which it makes sense in these cases to think about what would happen if everybody did or did not do something gives us no reason to reconsider the rejection of legislative moral theory. * * * The objection I am raising is related to but distinct from the problem of partial compliance as traditionally discussed by rule consequentialists. Rule consequentialists note that it won’t do to have rules optimific only in circumstances of

65  Reasons and Persons (Oxford: Oxford University Press, 1984), 75–86. 66  We can imagine that though the others could pollute, they have no reason to do so. 67  This is compatible with the analysis in Virginia Held, ‘Can a Random Collection of Individuals be Morally Responsible?’, Journal of Philosophy 67 (1970): 471–81.

Nonlegislative Justification  265 general acceptance. We need principles for the nonideal case as well. A general prohibition on the use of force would work well in the ideal case of universal acceptance, but rules are clearly needed to deal with the legitimate use of force in cases of self-­defense, and so on. So canonical rule consequentialists will favor a code that includes rules that are only relevant in situations of less than full compliance or acceptance. One well-­known approach assesses moral codes on the assumption of acceptance by ‘the overwhelming majority of everyone’, which turns out to be 90 percent.68 The eligible codes will include rules governing interactions with the 10 percent. A  familiar objection to this approach is that the chosen level of acceptance is arbitrary. A more important objection is that such a code will not necessarily ­provide the right rules for situations of less than, or more than, 90 percent acceptance; at 60 per cent acceptance the rules designed for the 90 percent case may require pointless or harmful actions.69 But the fundamental objection, the one I have been raising, is this: Whatever level of counterfactual acceptance is chosen, the effects of that level of acceptance do not determine what would be the right way to act in the actual world. Michael Ridge suggests that we adopt the code with the highest average welfare score across all possible acceptance levels.70 This proposal too has been criticized as arbitrary (why not the median?) and for not taking into account the likelihood of particular levels of acceptance. A flurry of further proposals have recently been made to address these issues.71 But the objection to moral reasoning that appeals to the effects of counterfactual levels of compliance will remain unless rule consequentialism offers rules tailored to each possible level of compliance. Parfit suggests the following formulation: ‘Everyone ought to follow the rules whose being followed by any number of people rather than by no one would make things go best.’ As he indicates, such a set of rules would include ones that tell us to ‘act in the ways that would make thing go best, given the number or proportion of ­people who are following these rules’.72 Until recently, the partial compliance problem as it applies to contractualism was largely ignored.73 It is clear enough that Scanlon envisages the need for prin­ 68 Hooker, Ideal Code, Ideal World, 80–5. 69  See Michael Ridge, ‘Introducing Variable-­Rate Rule-­Utilitarianism’, Philosophical Quarterly 56 (2006): 242–53, 245–6. 70  Ridge, ‘Introducing Variable-­Rate Rule-­Utilitarianism’. 71  For a helpful overview, and defense of what might be the best response to the traditional question of partial compliance, see Kevin  P.  Tobia, ‘Rule-­Consequentialism’s Assumptions’, Utilitas 30 (2018): 458–71. 72  On What Matters, i. 319. For discussion, see Michael Ridge, ‘Climb Every Mountain?’, Ratio 22 (2009): 59–77. 73 The exception, among early commentators, is Pogge. More recently, see Jussi Suikkanen, ‘Contractualism and the Conditional Fallacy’, Oxford Studies in Normative Ethics 4 (2014): 113–37, and ‘Contractualism and the Counter-­Culture Challenge’, Oxford Studies in Normative Ethics 7 (2017): 184–206; Nicholas Southwood, ‘Contractualism for Us as We Are’, Philosophy and Phenomenological

266  Liam Murphy ciples appropriate for situations of less than full compliance.74 Beyond that, ­however, the same issues that face rule consequentialism arise, and the only option that does not involve thinking about what reasons people would have to object to principles under counterfactual levels of compliance would be, again, to consider what principles would be reasonably rejectable, ‘given the number or proportion of people who are following’ these principles. It seems, then, that moral theories that use the model of hypothetical legislations must offer a rule or principle for every possible level of acceptance. For some moral issues, the rules or principles may be the same for many or even all levels of acceptance. But for many issues, this will not be the case. The resulting complexity is obviously in itself a ground for skepticism about the overall approach. But there are more serious problems.75 If we are to offer a rule or principle for every level of acceptance, we will need to imagine what the world will be like at that level. But the effect on people of some level of acceptance of a rule or prin­ ciple is entirely indeterminate unless we know what the nonaccepters accept.76 There is no nonarbitrary way to fix the moral outlook of the nonaccepters—are they partial accepters, psychopaths, act utilitarians, Kantians, or utterly without moral compass?77 So it seems we are limited to considering our actual world and actual rules or principles that we can determine are to some degree accepted. We have to ask, for example: Given the actual level of acceptance of principle P, is it a principle that no one could reasonably reject for that level of acceptance in these circumstances? To answer this question, we will have to consider how things would be different in the actual world if the people who accept P accepted differResearch 99 (2019): 529–47. The problem has long been acknowledged in the context of Kant’s cat­ egor­ic­al imperative. See Thomas E. Hill, ‘The Kingdom of Ends’, Proceedings of the Third International Kant Congress: Held at the University of Rochester, March 30–April 4, 1970 (Dordrecht: Springer Netherlands, 1972) 307–15, repr. in Dignity and Practical Reason in Kant’s Moral Theory (Ithaca, NY: Cornell University Press, 1992), 58–66; Christine Korsgaard, ‘The Right to Lie: Kant on Dealing with Evil’, Philosophy & Public Affairs 15 (1986): 325–49, repr. in Creating the Kingdom of Ends (Cambridge: Cambridge University Press, 1996), 133–58. 74 See ‘Contractualism and Utilitarianism’, in Amartya Sen and Bernard Williams (eds.), Utilitarianism and Beyond (Cambridge: Cambridge University Press, 1982), 111; ‘Contractualism and Justification’. 75  For general discussion of the problems of assessing the effects of differing levels of acceptance of different moral theories, see Murphy, Moral Demands in Nonideal Theory, ch. 3. 76  See Holly Smith, ‘Measuring the Consequences of Rules’, Utilitas 22 (2010): 413–33. 77  In ‘Contractualism and the Counter-­Culture Challenge’, Jussi Suikkanen suggests the following solution. We look to our actual world and sort it into sheep (those who accept ‘conventional morality’) and goats (those who do not). Now we take a close look at the goats and see what they are like; they will fall into several categories. For all possible compliance worlds, we assume that the noncompliers are just like the actual goats, with numbers adjusted to keep the proportions of different types of goats fixed. In other words, in each possible compliance world, the noncompliers are just like the noncompliers with conventional morality are now. But why should we assume that if someone complying with conventional morality in the actual world were not to comply with some proposed code, they would become a goat, rather than stick with conventional morality or choose some third option? Moreover, what moral view a person takes is affected by the moral outlook of others around them.

Nonlegislative Justification  267 ent principles. P may or may not turn out to be justified by this test. Suppose that it is not: some people could reasonably reject P on the ground that Q would impose lesser burdens on them and not burden anyone else more. Would this justify Q? It would not, because we do not know whether all the current P-­accepters will in fact convert to Q. Q may simply not be a feasible principle at the level of acceptance current for P.78 There is a further problem. I have said we are limited to considering our actual world and actual rules or principles that we can determine are to some degree accepted. But this is in fact an impossible task.79 If we are comparing entire moral codes, it seems unlikely that any of them are fully accepted by more than a few theorists, if that. As different moral views overlap considerably in their content we could say that many of the codes are partly accepted by many people. But this leaves us unable to say anything nonarbitrary about the acceptance level of particular codes. This problem is alleviated somewhat if we work at the level of prin­ ciples, but not entirely, since not many people actually accept carefully formulated moral principles. There will again be much overlap, and we may be able to say that there is a certain level of acceptance of some broad and vague principle. But the reductive approach is attractive in part for allowing us to test more refined principles, for the promise of offering, as Scanlon says, ‘a way of explaining and interpreting incompletely specified exceptions’. It seems that the only way we can talk sensibly about the effects of compliance with rules or principles is, after all, to stipulate some level of compliance and stipulate also what the noncompliers will be doing. But this is to bring us back face to face with the fundamental objection to legislative moral theory: that the effects on people of counterfactual levels of compliance with some principle does not determine the merits of that principle for us now.

3.  Political Theory on the Legislative Model The fundamental objection to legislative moral theory does not carry over to contractualist or consequentialist theories of justice or legitimacy, or contractualist or consequentialist accounts of what the law should be. Rawls’s contractualism concerns the best justified principles of justice. His theory of justice is not a moral theory, providing rules for individuals. Rawls did flirt with the idea of rightness as fairness, the possibility that all of morality could be modelled as principles chosen behind a veil of ignorance.80 But he dropped that idea, and his theory of justice is

78  This point is related to the issue of circularity mentioned by Ridge, ‘Introducing Variable-­Rate Rule-­Utilitarianism’, 247–8. 79 Murphy, Moral Demands in Nonideal Theory, 50–9. 80  A Theory of Justice (rev. edn., Cambridge, MA: Harvard University Press, 1999), 95.

268  Liam Murphy a set of principles for use by constitution-­framers and legislators.81 Those ­prin­ciples are not morally fundamental, they are shaped or constructed out of fundamental moral considerations for use in a particular task, the task of making just law.82 They are carefully drafted for that purpose. Rawls counts in favor of the  difference principle that it is easier to interpret and apply than some other pos­sible principles.83 The element of drafting is entirely appropriate, given the purpose. So Rawls’s theory of justice and contractualism as a political theory generally— including Nagel’s application of Scanlon’s contractualism to the political realm84— is not to be faulted for appealing to counterfactual circumstances in order to determine what is to be done.85 The theory provides a blueprint for constitutional design and legislation, it proposes that we do something—build these institutions and write these codes. Of course, one might criticize Rawls for presenting us with a theory for the ideal case, rather than a theory of how to make things incrementally better, given where we are now.86 It might be thought that knowing what the ideal is doesn’t help us that much when we are trying to make things better. Whether or not that objection is well founded, it has nothing to do with the objection that I have been making to legislative moral theory. It is clear that there is no justice, for Rawls, other than in virtue of the actual operations of actual legal and political institutions.87 In this respect the social contract tradition in political theory is relevantly similar to classical utilitarianism. Bentham was deeply interested in legislation—real

81  This is evident from his account of the four-­stage sequence for the implementation of the theory of justice: the original position, the constitutional stage, the legislative stage, and the judicial and administrative stage. See A Theory of Justice, sect. 31. Another strand in Rawls suggests that the ‘basic structure’ includes non-­legal institutions. For discussion, see G.  A.  Cohen, Rescuing Justice and Equality (Cambridge, MA: Harvard University Press, 2008), ch. 3. 82  Since the principles of justice are not morally fundamental, but constructed for a particular task, there is no special puzzle about why they apply only to the basic structure, and not to individual choice. This understanding is compatible with both Liam Murphy, ‘Institutions and the Demands of Justice’, Philosophy & Public Affairs 27 (1991): 251–91, and Seana Shiffrin, ‘Incentives, Motives, and Talents’, Philosophy & Public Affairs 38 (2010): 111–42. 83  A Theory of Justice, 281. 84  Equality and Partiality (Oxford: Oxford University Press, 1991). There is also no objection to Scanlon’s use of the reasonable rejection test to think about what legal rights there should be (discussed above) or to evaluate whether certain principles can justifiably be legally enforced—see ‘Promises and Contracts’, in Peter Benson (ed.), The Theory of Contract Law: New Essays (Cambridge: Cambridge University Press, 2001), 86–117. 85  See Hosein, ‘Contractualism, Politics, and Morality’. 86  See e.g. Amartya Sen, The Idea of Justice (Cambridge, MA: Harvard University Press, 2009). 87 Cohen’s argument in Rescuing Justice and Equality that Rawls goes astray in offering as an account of justice a theory of how to choose principles for social regulation has structurally a lot in common with my argument that we should not confuse interpersonal right with ideal legislation. But Cohen is motivated by a view about the content of the concept of justice, and my argument is not at all concerned with conceptual content. Once we leave conceptual issues aside, moreover, the substance of my argument against legislative moral theory does not apply to Rawls’s political theory, which, as indicated, just is a theory of institutional design.

Nonlegislative Justification  269 legislation.88 He did not assimilate the question of how people should act to the question of what would be required by ideal hypothetical legislation. He didn’t actually have much to say about how people should act, morally speaking, since he was a psychological hedonist who thought that all of us were hard-­wired to pursue our own pleasure. So Bentham, the moralist, was concerned with actual legislation, since legislation could lead people to act in ways that make things go better. Mill was not a psychological hedonist, but he too was mostly concerned with the moral evaluation of law and customary practices, since he thought it would be rare for a good utilitarian to be able to calculate the consequences of particular actions.89 Chapter  5 of Mill’s Utilitarianism is often interpreted as defending a rule-­ utilitarian account of rights. I do not read it that way. Rather, the doctrine that rights are things that society ought to defend me in the possession of is very much of a piece with Bentham’s account of legal rights, extended to take into account non-­legal normative practices, in particular that of socially-­enforced conventional morality. Taking up this suggestion from Mill, might we understand contractualism and rule consequentialism as proposals for the actual inculcation of a code of conventional morality, one that would be best justified or optimific?90 If we did, the objection of Sections 1 and 2 would not apply, just as they do not apply to Mill. But that is not the aim of these theories. Scanlon does indicate that if there is a social practice already in effect, and its norms are among those that could not be reasonably rejected, then, under the Principle of Established Practices, it would be wrong to violate those norms.91 But it is not a condition of a non-­reasonably rejectable principle identifying what acts are wrong that the principle is already generally accepted. And though Scanlon’s account of moral motivation stresses the value of living in a relation of mutual recognition with others—living, in Mill’s words, ‘in unity with our fellow creatures’—he also makes it clear that he does not have in mind actual social harmony in the sense that everyone accepts the same principles as justified and for the same reasons.92 Rule consequentialists are a little less clear on this point. Hooker often seems to be arguing that acts are wrong if they violate an ideal code that may not be actual. On the other hand, a chapter of Ideal Code, Real World is titled ‘What are the Rules to Promote?’ and his canonical statement of what makes acts wrong uses the indicative rather than the subjunctive mood.93 Moreover Hooker, like other defenders of rule consequentialism, takes into account the ‘cost’ of inculcating or 88  Jeremy Bentham, An Introduction to the Principles of Morals and Legislation (1781). 89  John Stuart Mill, Utilitarianism (1863), On Liberty (1859). 90  See Pogge, ‘What We Can Reasonably Reject’, 135–8, for discussion of this ‘political’ in­ter­pret­ ation of Scanlon. 91  What We Owe to Each Other, 339. 92  What We Owe to Each Other, 154, 162–3. 93  Ideal Code, Real World, 32.

270  Liam Murphy internalizing the ideal moral code, which makes most sense if what is on offer is a blueprint for reform of actual social practices. I suspect the solution to this ­question of interpretation is that Hooker has in mind a case where the actual code generally accepted—’conventional morality’—happens to coincide to at least a good extent with the ideal code. This is supported by his embrace of ‘wary rule-­ consequentialism’, which implies that attempts ‘at moral reform should begin with existing practices’.94

4.  The Justification and Reform of Existing Social Practices This way of thinking about legislative moral theory gives it a purpose entirely different from Scanlon’s reductive method of justification. We are now looking not for greater depth and detail in our understanding of limits to how we can treat each other, but for an evaluation of existing social practices. This different project is certainly worthwhile. As Gerald Gaus has compellingly argued, the legitimacy of informal social enforcement of what he calls ‘social morality’ is an important and neglected issue.95 And even beyond legitimacy, there is a question worth asking whether these practices are as good, or well justified, as they might be, and if not how they might be improved. So if the model of hypothetical moral legislation were investigating the good consequences, or reasonableness from each person’s point of view, of possible reforms to conventional normative orders, there would be no objection. But we have to remember that to argue that a normative social practice of ­following and informally enforcing conventional moral norms is legitimate or even best justified is not to give an account of right and wrong at all; it is to give an account of the justification or otherwise of a practice. Rawls once argued to the contrary. His starting point in ‘Two Concepts of Rules’ is act utilitarianism, which, as he points out, implies that the conventional moral rules of promising and punishment are not moral rules at all; they are simply positive rules of a conventional social practice that it may or may not make good moral sense to comply with in particular cases depending on whether that does any good. To get to a duty to follow the rules of the practice, Rawls did not offer the rule-­utilitarian idea that since it is or would be better if all follow those rules, they are binding on each. His response, instead, was that the proper object of utilitarian assessment was the overall practice, and that if somebody raised the

94  Ideal Code, Real World, 114. Here Hooker follows Brandt, e.g. A Theory of the Good and the Right, 290. 95  The Order of Public Reason (Cambridge: Cambridge University Press, 2011). In Gaus’s language, the issue is the authority of social morality.

Nonlegislative Justification  271 question of whether a good utilitarian should comply in a particular case with the rules of the practice, he would simply show that he didn't understand the situation in which he was acting. If one wants to perform an action specified by a practice, the only le­git­ im­ate question concerns the nature of the practice itself (‘How do I go about making a will?’).96

It is interesting that the example Rawls gives here is a legal one. Of course, if I want to make a will, I want to make a legally valid will; so I have to follow the formal legal requirements to the letter. But what is the moral analogy? ‘If one wants to perform an action specified by a practice,’ Rawls writes, ‘the only le­git­ im­ate question concerns the nature of the practice itself ’—which we must take to mean: if I want to participate in a fully compliant way with the practice, I must follow the rules. But the question always remains whether there are good moral reasons to participate in a fully compliant way with any given social practice. The goodness or justifiability of a social practice, such as informally enforced conventional morality, does not in itself establish that the norms of that practice are valid or true moral principles. Scanlon, of course, never suggests anything to the contrary. The satisfaction of the test of reasonable rejectability would give conventional norms the status of principles of right and wrong only to the extent that that test is generally successful as an account of what makes acts wrong. A view which holds that acts are wrong if they are held to be so by norms of an actual social practice that could, as a practice, be justified, would be a very different view. There may be arguments to support such a position, turning on the normative significance of convention, but they are not provided by legislative moral theory.

5. Conclusion I have argued that a version of Scanlon’s reductive approach that seeks justification for conduct of such-­and-­such a kind in such-­and-­such circumstances by comparing the reasons in favor and the reasons others have to object is a very attractive way to think about moral constraints on how we may treat one another. Of course, as decades of discussion of Scanlon’s contractualism have shown, there is much to think about when considering the kinds of reasons that are appropriately appealed to in this thought process. I have not begun to investigate that here.

96  ‘Two Concepts of Rules’ (1955), repr. in Collected Papers (Cambridge, MA: Harvard University Press, 1999), 38.

272  Liam Murphy But the main argument of this paper has been negative. It is that Scanlon’s reductive method should not be oriented to justification by way of principles for the general regulation of behavior. This model of hypothetical legislation, which is embraced also by rule consequentialism, is unmotivated at its foundation. The natural law tradition, extending back for millennia, and of course also Kant’s explicit invocation of the moral law and legislation for a kingdom of ends, have made the association between morality and legislation seem natural. But it is just a mistake to set about constructing morality as if we were legislators making law.

Bibliography Ashford, Elizabeth, ‘The Demandingness of Scanlon’s Contractualism’, Ethics 113 (2003): 273–302. Bentham, Jeremy, An Introduction to the Principles of Morals and Legislation (1781). Brandt, Richard, A Theory of the Good and the Right (Oxford: Oxford University Press, 1979). Cohen, G.  A., Rescuing Justice and Equality (Cambridge, MA: Harvard University Press, 2008). Dancy, Jonathan, Ethics Without Principles (Oxford: Oxford University Press, 2004). Dworkin, Ronald, Taking Rights Seriously (London: Duckworth, 1977). Forst, Rainer, Contexts of Justice (Berkeley and Los Angeles: University of California Press, 2002). Forst, Rainer, The Right to Justification (New York: Columbia University Press, 2012). Forst, Rainer, ‘Justification Fundamentalism: A Discourse-Theoretical Interpretation of Scanlon’s Contractualism’, presented at the 2016 Lauener Conference, ‘Themes from Thomas M. Scanlon’. Frick, Johann, ‘Contractualism and Social Risk’, Philosophy & Public Affairs 43 (2015): 175–223. Gaus, Gerald, The Order of Public Reason (Cambridge: Cambridge University Press, 2011). Habermas, Jürgen, Between Facts and Norms, trans. William Rehg (Cambridge, MA: MIT Press, 1996). Held, Virginia, ‘Can a Random Collection of Individuals be Morally Responsible?’, Journal of Philosophy 67 (1970): 471–81. Hill, Thomas E., ‘The Kingdom of Ends’, Proceedings of the Third International Kant Congress: Held at the University of Rochester, March 30–April 4, 1970 (Dordrecht: Springer Netherlands, 1972), 307–15; repr. in Dignity and Practical Reason in Kant’s Moral Theory (Ithaca, NY: Cornell University Press, 1992), 58–66. Hooker, Brad, Ideal Code, Real World (Oxford: Oxford University Press, 2000).

Nonlegislative Justification  273 Hosein, Adam, ‘Contractualism, Politics, and Morality’, Acta Analytica 28 (2012): 495–508. Kamm, F. M., ‘Owing, Justifying, and Rejecting’, Mind 111 (2002): 323–54. Kamm, F. M., Intricate Ethics (Oxford: Oxford University Press, 2007). Korsgaard, Christine, ‘The Right to Lie: Kant on Dealing with Evil’, Philosophy & Public Affairs 15 (1986): 325–49; repr. in Creating the Kingdom of Ends (Cambridge: Cambridge University Press, 1996), 133–58. Kumar, Rahul, ‘Defending the Moral Moderate: Contractualism and Common Sense’, Philosophy & Public Affairs 28 (1999): 275–309. Lingle, Clara, ‘Generic Objections, Actual Claims: A Problem for Contractualism as Wronging’, unpublished MS, 2020. Mill, John Stuart, On Liberty (1859). Mill, John Stuart, Utilitarianism (1863). Murphy, Liam, ‘Institutions and the Demands of Justice’, Philosophy & Public Affairs 27 (1991): 251–91. Murphy, Liam, Moral Demands in Nonideal Theory (Oxford: Oxford University Press, 2000). Murphy, Liam, ‘The Artificial Morality of Private Law: The Persistence of an Illusion’, University of Toronto Law Journal 70 (2020): 453–88. Nagel, Thomas, The View from Nowhere (Oxford: Oxford University Press, 1986). Nagel, Thomas, Equality and Impartiality (Oxford: Oxford University Press, 1991). Parfit, Derek, On What Matters, vols. i and ii (Oxford: Oxford University Press, 2011). Pogge, Thomas, ‘What We Can Reasonably Reject’, Philosophical Issues 11 (2001): 118–47. Rawls, John, ‘Two Concepts of Rules’ (1955); repr. in Collected Papers (Cambridge, MA: Harvard University Press, 1999). Rawls, John, A Theory of Justice (rev. edn., Cambridge, MA: Harvard University Press, 1999). Ridge, Michael, ‘Introducing Variable-Rate Rule-Utilitarianism’, Philosophical Quarterly 56 (2006): 242–53. Ridge, Michael, ‘Climb Every Mountain?’, Ratio 22 (2009): 59–77. Ross, W. D., The Right and the Good (Oxford: Oxford University Press, 1930). Scanlon, T. M., ‘Rights, Goals, and Fairness’ (1975); repr. in The Difficulty of Tolerance (Cambridge: Cambridge University Press, 2003), 26–41. Scanlon, T.  M., ‘Contractualism and Utilitarianism’, in Amartya Sen and Bernard Williams (eds.), Utilitarianism and Beyond (Cambridge University Press, 1982), 103–28. Scanlon, T. M., The View from Nowhere (Oxford: Oxford University Press, 1986). Scanlon, T. M., What We Owe To Each Other (Cambridge, MA: Harvard University Press, 1998).

274  Liam Murphy Scanlon, T. M., ‘Promises and Contracts’, in Peter Benson (ed.), The Theory of Contract Law: New Essays (Cambridge: Cambridge University Press, 2001), 86–117. Scanlon, T. M., ‘Thomson on Self-Defense’, in Alexander Byrne, Robert Stalnaker, and Ralph Wedgwood (eds.), Fact and Value: Essays on Metaphysics for Judith Jarvis Thomson (Cambridge, MA: MIT Press, 2001), 199–214. Scanlon, T.  M., The Difficulty of Tolerance (Cambridge: Cambridge University Press, 2003). Scanlon, T.  M., ‘How I Am Not a Kantian’, in D.  Parfit, On What Matters (Oxford: Oxford University Press, 2011), ii. 116–39. Scanlon, T. M., Being Realistic About Reasons (Oxford: Oxford University Press, 2014). Scanlon, T.  M., ‘Contractualism and Justification’, presented at the 2016 Lauener Conference, ‘Themes from Thomas M. Scanlon’, available at https://www.academia. edu/35861169/Contractualism_and_Justification_pdf. Sen, Amartya, The Idea of Justice (Cambridge, MA: Harvard University Press, 2009). Sheinman, Hanoch, ‘Act and Principle Contractualism’, Utilitas 23 (2011): 288–315. Shiffrin, Seana, ‘Incentives, Motives, and Talents’, Philosophy & Public Affairs 38 (2010): 111–42. Smith, Holly, ‘Measuring the Consequences of Rules’, Utilitas 22 (2010): 413–33. Southwood, Nicholas, Contractualism and The Foundations of Morality (Oxford: Oxford University Press, 2010). Southwood, Nicholas, ‘Contractualism for Us as We Are’, Philosophy and Phenomenological Research 99 (2019): 529–47. Suikkanen, Jussi, ‘Contractualism and the Conditional Fallacy’, Oxford Studies in Normative Ethics 4 (2014): 113–37. Suikkanen, Jussi, ‘Contractualism and the Counter-Culture Challenge’, Oxford Studies in Normative Ethics 7 (2017): 184–206. Thomson, J. J., The Realm of Rights (Cambridge, MA: Harvard University Press, 1990). Tobia, Kevin P., ‘Rule-Consequentialism’s Assumptions’, Utilitas 30 (2018): 458–71. Wallace, R. J., The Moral Nexus (Princeton: Princeton University Press, 2019).

12

Doing Right by Wrong Stephen Darwall

In On What Matters, Derek Parfit commits himself to the position that T. M. Scanlon calls ‘reasons fundamentalism’—in Scanlon’s formulation: ‘reasons are . . . the only fundamental elements of the normative domain, other normative notions such as good and ought being analyzable in terms of reasons’ (Scanlon 2014: 2). Parfit agrees: ‘Reasons are . . . fundamental. Something matters only if we or others have some reason to care about that thing’.1 Parfit distinguishes two ‘conceptions of normativity’, only one of which is normative in the sense with which Scanlon and other metanormative theorists are usually concerned.2 ‘On the reason-­involving conception,’ Parfit says, ‘normativity involves reasons or apparent reasons.’ By contrast, ‘on the rule-­ involving conception, normativity involves requirements, or rules, that distinguish between what is correct and incorrect, or what is allowed or disallowed’ (OWM i. 144). ‘Rules’ here are the sort that can be socially actualized, for example, in law, custom, or social mores, or even in a game, independently of the holding of normative reasons, hence, independently of normativity in the reason-­involving sense. Something’s being incorrect or disallowed in the rule-­involving sense conceptually entails nothing about the existence of (normative) reasons not to do it or reasons to take any attitude toward someone’s doing it. Reason-­involving normativity, on the other hand, is not so much realized in the actual world as lived up to or conformed to there. Whether any social rule or law conforms to reason-­involving norms is always an ‘open question’.3 ‘Reasons’ in reasons fundamentalism and the reason-­involving conception of normativity are normative reasons, that is, reasons to have some attitude or perform some act, rather than motivating reasons. The latter are considerations on the basis of, or for, which some agent actually performs some action (or holds some attitude). Alternatively, it consists in the psychological state (desire in the 1  On What Matters (OWM), i. 148. References to Parfit (2011 and 2017) will be to OWM and volume and page numbers. See also Raz: ‘the normativity of all that is normative consists in the way it is, or provides, or is otherwise related to reasons’ (Raz 1999: 67). 2  See also MacPherson and Plunkett (2017: 16). 3  The reference is to G.  E.  Moore’s ‘open question’ argument (Moore 1903/1993). However, see Frankena (1942) for an argument that what underlay Moore’s idea was normativity in something like the reason-­involving sense. Stephen Darwall, Doing Right by Wrong In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0013

278  Stephen Darwall case of action) that involves the agent’s taking something as a reason and being moved by it. In the latter case, the agent regards the consideration as a normative reason. They can do that, however, whether or not the consideration is genuinely a normative reason.4 W. D. Ross made a distinction between acts and actions that is helpful here and that will also be of use later when we try to come to terms with quite different things that Parfit says about wrongful acts and actions in Reasons and Persons and On What Matters respectively. Ross reserved ‘act’ for things we might do (or might have done) as we consider these in practical reasoning (prospectively or retrospectively). By contrast, he used ‘action’ to refer to actual doings (including their actual motives). In this usage, which I hereby adopt, normative reasons to do things are for reasons for acts, and motivating reasons explain, or are included in, actual actions. When it comes to moral assessment, Ross says, ‘we should . . . talk of a right act but not of a right action, of a morally good action, but not of a morally good act’ (Ross 1930/2002: 7). Parfit remarks on the fundamentality of reasons in On What Matters in the course of discussing the relationship between reasons for acts that derive from morality and practical reasons of other kinds. In particular, he is concerned with what he calls ‘the profoundest problem’: Sidgwick’s quandary of what reason dictates when ‘duty and self-­interest conflict’ (OWM i. 142). What, in other words, is the normativity of deontic morality—of something’s being morally obligatory or required and therefore wrong not to do? How do reasons figure in facts about moral right and wrong, and how do these relate to reasons of other kinds? Philippa Foot famously argued that the fact that an act is ruled out by morality may not be fundamentally different from its being prohibited by etiquette (Foot 1972). Both seem to issue categorical imperatives in the sense that their requirements are independent of agents’ desires and ends. However, this may only concern normativity in the rule-­involving sense. Parfit thinks, like most of us, that morality has a de jure normativity that goes beyond any de facto, rule-­involving normativity that can be realized in actual social rules, customs, and law (OWM i. 148). To be sure, ‘morality’ can function as a count noun, referring to (actual) moralities of societies, groups, or even individuals having rule-­involving normativity. But in the sense with which moral philosophers are generally concerned, ‘morality’ is not a count noun. It refers not to any group or individual’s morality, but to standards that transcend these and that cannot, by their nature, be realized in the actual world as rules of moralities are. Morality’s requirements, in this different de jure sense, purport to bind us in ways that actual social moralities and laws cannot.

4  When I use ‘reason’ or ‘reasons’ in what follows, I will generally be referring to normative reason (as in ‘there is reason’) or reasons, unless context makes clear otherwise.

Doing Right by Wrong  279 But how are we to explain or account for this de jure normativity? To pursue this question, I propose we accept Parfit’s reasons fundamentalism. The question will then be: what kind of reason-­involving normativity do morality and its requirements have? I shall claim that deontic morality does indeed having reason-­ involving normativity, although not in the first instance for acts, but for blame and other holding-­accountable, ‘reactive’ attitudes. I also claim that once we appreciate the conceptual connection between deontic moral concepts and accountability, we can then explain the kind of normativity these concepts have for acts. My claims, I emphasize, concern deontic moral concepts, like those of obligation, right, and wrong. Parfit is also concerned with deontic concepts like wrong in On What Matters in a way he was not in Reasons and Persons and that helps explain otherwise puzzling differences in the normative moral theories he defends in those works, as we shall see. However, Parfit holds in On What Matters that the ‘ordinary’ concept of wrongness is indefinable, and, therefore, apparently lacks any reason-­ involving conceptual entailments. But if that is so, it will follow that Parfit’s ‘ordinary’ concept of wrongness is not a normative concept. Like Parfit, I reject any attempt to define wrongness in terms of reasons for acts, even moral reasons. I argue that unless we hold that there is a conceptual connection between wrongness and reasons for holding-­accountable attitudes like moral blame we cannot plausibly understand wrongness to be a normative concept, as Parfit claims it to be (OWM i. 174). Of course, reasons fundamentalism may not be correct. Some philosophers, like Broome and Sidgwick, argue that ought is more fundamental than reasons (Broome 2018). In The Methods of Ethics, Sidgwick says that all ethical judgments contain ‘the fundamental notion represented by the word ought’ (Sidgwick 1907/1967: 25). A metanormative theorist might extend this claim, like Broome, to normative judgments more generally. Yet other philosophers argue that the idea of fittingness or fit is the most fundamental normative concept (Chappell 2012; McHugh and Way 2016; Howard 2019).5 These differences will not matter for our purposes. All three approaches agree that (de jure) normativity always concerns some relation between an agent (understood broadly to include the holder of an attitude) and some presenting deliberative context. What they disagree about is which relation is normatively fundamental. Reasons and ought fundamentalists hold that it is the deliberative 5  The case for this proposal primarily rests on the ‘wrong kind of reason’ problem. This can be more easily seen with reasons for attitudes of various kinds. For example, if someone offers you a prize if you believe a manifest falsehood, then even if, in some sense, that gives a reason to believe it, it doesn’t give you a reason of the kind that has to do with whether belief is fitting to its epistemic context, to its being credible. It is said, therefore, a ‘reason of the wrong kind’ for belief. Parfit denies that such cases can arise, since he holds that in such a case, the prospect of the prize is not a reason to believe the falsehood, but a reason to desire that one could so believe. See his discussion of ‘object-­ given’ and ‘state-­given’ reasons (OWM i. 50–1, 420–32).

280  Stephen Darwall context’s grounding reasons and oughts (of the right kind) respectively. And fit fundamentalists hold that it is the fittingness of some attitude or action to its deliberative context. It will simplify things if we just assume that reasons fundamentalism is correct and ask whether morality and its requirements have reason-­involving normativity and, if so, of what kind? The arguments to follow could as well be formulated in terms of oughts or fittingness.

1.  Discussion of C as a ‘Moral Theory’ in Reasons and Persons Presently, I will be pointing to substantial differences between Parfit’s approach to these issues in On What Matters and the main lines he pursued in Reasons and Persons. First, however, note how Reasons and Persons begins in a reasons fundamentalist key: What do we have most reason to do? Several theories answer this question. Some of these are moral theories; others are theories about rationality. When applied to our decisions, different theories give different answers. We must then decide which is the best theory.6

This implies that morality is reason-­ involving in the sense that morality necessarily bears on what we have reason to do, or at least that ‘moral theories’ claim this to be so. Part I of Reasons and Persons is an extended discussion of two theories, S and C, that are, Parfit tells us, theories of rationality and morality respectively. S is the Self-­Interest Theory of rationality: ‘S gives to each person this aim: the outcomes that would be best for himself, and that would make his life as good as possible’ (RP, 3). C (Consequentialism) is, by contrast, a moral theory. C holds that ‘there is one ultimate aim: that outcomes be as good possible’. As applied to acts, C holds that ‘what each of us ought to do is whatever would make the outcome best’ (RP, 24). Both S and C make claims about normative reasons, hence, about what we should do in a reason-­involving sense. If S is a theory of rationality and C is a moral theory, then S must tell us what we should do ‘rationally’ and C, what we should do ‘morally’. And both must bear on what we have reason to do sans phrase, on what ‘we have most reason to do’. The substantive difference between S and C is that whereas S (rationally) counsels doing whatever will best promote the agent’s own interest or good (agent-­ relative value), C (morally) counsels acts that best promote what is impersonally or impartially best (agent-­neutral value). However, both S and C 6  Reasons and Persons (RP), 3. References to Parfit (1984) will be to RP and page numbers.

Doing Right by Wrong  281 both make claims about what there is reason for agents to do sans phrase. Sidgwick’s ‘profoundest problem’ looms. Parfit’s analyses of S and C are substantially similar. He finds both likely to be indirectly self-­defeating in the sense that if agents accept each and try to follow them in their deliberative reasoning, then their respective aims will not be maximally met. Someone may better promote their own interest by being somewhat ‘self-­denying’, aiming at things other than their own interest or good. This is sometimes called the ‘paradox of egoism’. Similarly, a number of people may better promote impersonal or impartial good collectively by not being ‘do-­ gooders’ who aim at maximizing agent-­neutral value directly. The end C gives them may be better promoted indirectly when they are moved by other things also, say, by personal concerns, or by common-­sense moral duties, or by both. We might call this the ‘paradox of consequentialism’. Reasons and Persons forcefully argues that these and similar considerations are not decisive objections to either theory. Neither S nor C are thereby shown to fail ‘in their own terms’ (RP, 28). The so-­ called ‘paradoxes’ of egoism and consequentialism are not really paradoxical; we just have to keep straight what is being assessed. We can put the point in terms of Ross’s distinction between acts and actions. S can sensibly recommend all and only acts that are self-­interested while recommending against actions being motivated solely by self-­interest. Its theory of acts and actions will understandably diverge in this way. Similarly, C can recommend optimific acts without being committed thereby to recommending only actions that are motivated by maximal agent-­neutral value. S and C’s theory of normative reasons (for acts) will naturally come apart from their theories of optimal motivating reasons (for actions). Reasons and Persons’s arguments were thought by many to provide substantial support for Act Consequentialism (AC) as a theory of morally right conduct. As I once put it, Reasons and Persons ‘wrote the game plan for how consequentialists might maintain an act-­consequentialist theory of right while nonetheless agreeing that it might maximize the good for agents to accept and be guided by some . . . theory closer to common-­ sense morality like rule consequentialism’ (Darwall 2017: 271). Consequentialists, Parfit seemed to be arguing, would do best to hold AC as a theory of right acts, while recommending actions that are motivated by something more like an acceptance of RC (Rule Consequentialism) or ‘Common-­Sense Morality’ (Reasons and Persons’s M). Also important to the reception of C as a moral theory was Reasons and Persons’s famous attack on personal identity as a ‘deep further fact’. After Rawls’s A Theory of Justice, deontological moral theory seemed to have the wind at its back, owing partly to its influential claim that utilitarianism failed adequately to respect the ‘distinction between’ or ‘separateness’ of persons (Rawls 1971: 27, 167). In Reasons and Persons, however, Parfit argued that if there is no ‘further

282  Stephen Darwall fact’ of personal identity, then this has important implications for moral theory that put any case for deontology based on the separateness of persons in jeopardy. Taken together, these two dialectical prongs helped to stem the trend toward deontology following Rawls, and to dramatically raise the prospects of Act Utilitarianism (AU) and AC as normative moral theories. Even if Parfit did not argue for AC directly in Reasons and Persons, readers might have been forgiven if they took from the book that that was where his moral theoretic heart lay (OWM iii. 291; Darwall 2017: 271).

2.  Consequentialist Moral Theory and Moral Wrong in On What Matters In On What Matters, however, all seems changed. Gone is any defense of C’s direct consequentialist theory of right acts. And in its place is a defense of RC as one of three elements of the ‘Triple Theory’, along with Kantian Contractualism and Scanlonian Contractualism, all three of which Parfit claims to be equivalent. RC is a consequentialist theory, of course, but most philosophers take its theory of right acts to differ sharply from AC’s and to be, extensionally at least, much closer to Common-­Sense Morality (Reasons and Persons’s M). That, indeed, was part of Parfit’s point in Reasons and Persons in holding that C may well be indirectly self-­defeating. Defenders of AC and RC can agree that consequentialist aims are likelier to be promoted better if agents accept and are moved in their practical reasoning, not by AC, but by something closer to RC or M. What they disagree about are the implications of this fact for a theory of right acts. Defenders of AC hold that a consequentialist theory of right acts can be comfortably separated from its theory of morally good actions in the way that Reasons and Persons’s C proposes. Defenders of RC deny this. And the striking fact, again, is that On What Matters seems to be supporting RC rather than AC, quite against the tenor of Reasons and Persons. Parfit emphasized after On What Matters volumes i and ii that he was not so much arguing there for RC in the sense of trying to convince readers to accept it rather than AC. He was rather claiming that ‘we have strong reasons to accept this Triple Theory, adding that we could justifiably reject this theory if its implications conflicted too often or too deeply with our moral intuitions’ (OWM iii. 291). Since, however, Parfit himself emphasizes how much closer RC is to moral intuition than AC is, it was natural to read On What Matters as defending RC over AC. There is another respect in which Parfit’s treatment of consequentialist theories is very different in On What Matters than it was in Reasons and Persons. In On What Matters, Parfit discusses what he calls ‘Impartial-­Reason Act Consequentialism’

Doing Right by Wrong  283 (IRAC), which says that ‘we have the strongest impartial reasons’ to do whatever will promote the greatest impartial good. Parfit cites Sidgwick’s ‘axiom of Rational Benevolence’ as an expression of this idea: the good of any one individual is of no more importance, from the point of view . . . of the Universe, than the good of any other . . . And . . . as a rational being I  am bound to aim at good generally . . . not merely a particular part of it. . . . I ought not to prefer my own lesser good to the greater good of another (OWM i. 168)

Parfit then remarks that ‘this kind of Consequentialism may be better regarded, not as a moral view, but as being like Rational Egoism, an external rival to morality’ (OWM i. 168). Someone could accept IRAC without accepting any ‘distinctively moral claims, such as claims about moral wrongness, obligations, duties, or blameworthiness’, as he put it later (OWM iii. 292). IRAC is not really a moral theory. This was not, I think, a point that Parfit fully appreciated in Reasons and Persons. There, as I shall show presently, Parfit presents C’s theory applied to acts in a way that is virtually indistinguishable from IRAC, although he there calls C a ‘moral theory’. About acts, C says that what ‘each of us ought to do is whatever would make the outcome best’ (in terms of agent-­neutral value) (RP, 24). As we shall see, ‘ought’ here does not entail that it would be wrong to act otherwise in the sense of ‘wrong’ that Parfit later defends in On What Matters. It simply indicates what, from the point of view of moral theory anyway, there is most reason to do. In On What Matters, by contrast, Parfit has come to realize that the deontic moral concepts of obligation, duty, wrong, and right differ from those of normative reasons and moral reasons to act. I conjecture that appreciating this distinctive normative profile has much to do with Parfit’s later view that IRAC (and I will argue, Reasons and Persons’s C) is not really a moral theory and with his change of focus, when it comes to morality, from direct consequentialist theories like C (and IRAC) to an indirect form of consequentialism like RC (OWM i. 375). When Parfit discusses morality and moral theories in On What Matters, he does so in a way that centers around the concept of wrongness. At the end of On What Matters’s section on ‘the profoundest problem’, Parfit says that in the rest of volume i he will ‘mostly discuss morality’ (OWM i. 148). ‘If reasons are more fundamental,’ he continues, ‘it may seem that I should continue to discuss reasons. But we have sufficient reasons for turning to morality,’ since ‘we can plausibly assume that we have strong reasons to care about morality, and to avoid acting wrongly’ (OWM i. 148). This is followed by chapter  7, ‘Moral Concepts’, which

284  Stephen Darwall begins: ‘Before we start to ask which acts are wrong, it will help to discuss what we mean by “wrong”, and what we are believing when we believe that some act is wrong’ (OWM i. 150). My main focus in what follows will be on details of Parfit’s discussion of the concept of moral wrong and the question of how wrong can be a normative concept in the reason-­involving sense on Parfit’s analysis. In my view, Parfit is right to reject any attempt to analyze moral wrongness in terms of reasons for acting, even moral reasons (OWM i .166–8). We need to ask, therefore, how the concept of wrong can be reason-­involving and, therefore, a normative concept on the reasons fundamentalist assumption we have been accepting. First, however, I want to draw attention to the way the concept of wrong, as Parfit understands it in On What Matters, is a central element of the moral theories Parfit considers there, and how this differs from what he calls ‘moral theories’ in Reasons and Persons. The former point is easily seen in Parfit’s formulation of the three theories—Kantian Contractualism, Scanlonian Contractualism, and Rule Consequentialism—that together form the ‘Triple Theory’. We can begin with what Parfit calls ‘the Kantian Contractualist Formula’. Parfit initially formulates this with the same normative term ‘ought’ that features in his formulation of C’s theory of right acts in Reasons and Persons (‘what each of us ought to do is whatever would make the outcome best’) (RP, 24). the Kantian Contractualist Formula: Everyone ought to follow the principles whose universal acceptance everyone could rationally will, or choose (OWM i. 355)

However, Parfit then reformulates Kantian Contractualism in a way that makes clear that he intends ‘ought’ not just as a ‘flavorless’ practical-­reasons-­involving ought, but in the distinctively deontic sense of obligation or duty that is analytically related to wrongness:7 An act is wrong just when such acts are disallowed by one of the principles whose universal acceptance everyone could rationally will.  (OWM i. 368)

Parfit follows this same pattern with Scanlonian Contractualism, first formulating it with ‘ought’ and then with ‘wrong’: Scanlon’s Formula: Everyone ought to follow the principles that no one could reasonably reject.

7  For this use of ‘flavorless’, see Gibbard (1990: 7).

Doing Right by Wrong  285 In a fuller statement: Some act is wrong just when such acts are disallowed by some principle that no one could reasonably reject, or when any principle permitting such acts could be reasonably rejected by at least one person. (OWM i. 360)

The final element of the Triple Theory is Rule Consequentialism (RC). In On What Matters, Parfit initially defines consequentialist theories of right in general: ‘Whether our acts are right or wrong depends only on facts about how it would be best for things to go’ (OWM i. 373). He then defines Act Consequentialism and Rule Consequentialism as follows: (Act Consequentialism) Everyone ought always to do whatever would in fact make things go best  (OWM i. 374)8 (Rule Consequentialism) Everyone ought to follow the principles whose universal acceptance would make things go best.  (OWM i. 374)

Although these both feature ‘ought’, it is clear that AC and RC will not be theories of moral right in the sense Parfit now intends unless we read their oughts, not as a ­flavorless ought conceptually related to normative reasons for acting, but as expressing the deontic idea of moral duty or obligation that is analytically related to wrongness. Finally, having argued that Kantian Contractualism, Scanlonian Contractualism, and Rule Consequentialism are all equivalent, Parfit introduces the Triple Theory: An act is wrong if and only if, or just when, such acts are dis­ allowed by some principle that is (1) one of the principles whose being universal laws would make things go best, (2) one of the only principles whose being universal laws everyone could rationally will, and (3) a principle that no one could reasonably reject. More briefly, TT: An act is wrong just when such acts are disallowed by some principle that is optimific, uniquely universally willable, and not reasonably rejectable. (OWM i. 412–13)

Clearly Parfit intends all three of the theories comprising the Triple Theory to be theories of right and wrong in the distinctive sense with which he is concerned in On What Matters. 8  I ignore here complications concerning cases where agents do not know relevant facts.

286  Stephen Darwall

3.  Reasons and Persons Redux: Right Acts and Wrong Actions Presently, we shall see how Parfit explicitly distinguishes in On What Matters between the concept of moral wrong, and deontic moral concepts that are analytically tied to it like obligation and duty, on the one hand, and normative reasons to act, on the other. More specifically, Parfit rejects any attempt to analyze moral wrong in terms of reasons to act, however decisive or ‘morally decisive’ (OWM i. 164–74). In Reasons and Persons, however, Parfit’s approach is very different. To begin with, Parfit characterizes ‘moral theories’ as telling us ‘what we have most reason to do’, as do theories of ‘rationality’ (RP, 3). Moreover, it is clear that when Parfit formulates C in Reasons and Persons, he does not do so with distinctively deontic moral concepts of right and wrong as these feature in the moral theories of right he discusses in On What Matters. ‘C’s central claim,’ Parfit tells us, is: (C1) There is one ultimate moral aim: that outcomes be as good as possible. C applies to everything. Applied to acts, C claims both (C2) What each of us ought to do is whatever would make the outcome best, and (C3) If someone does what he believes will make the outcome worse, he is acting wrongly. (RP, 24)

Parfit does not say whether ‘ought’ in C2 should be read as ‘morally ought’ or in an unqualified way. Whichever, Parfit is clear that although C is a moral theory, it nonetheless pronounces, like theories of ‘rationality’, on what ‘we have most reason to do’ (RP, 3). Having given these formulations, Parfit then distinguishes subjective and objective rightness and wrongness and gives C’s theories of each (RP, 25). The objective rightness of an act depends on the facts as they actually are, whereas subjective rightness depends on what the agent believes or ought to have believed the relevant facts to be. C2 states C’s theory of objectively right acts: acts are objectively right if, and only if, they make the outcome best. Subjective rightness, according to C, turns on whether the agent believed or ought to have believed that the act’s outcome would be best. ‘We need a theory of subjective rightness for two reasons,’ Parfit says. ‘We often do not know what the effects of our acts would be. And we ought to be blamed for  doing what is subjectively wrong’ (RP, 25). (C3) gives C’s theory subjective wrongness: an agent’s action is subjectively wrong if they do what they believe will have a worse outcome than they could have achieved by another available act. Parfit then adds that he will generally ‘use right, ought, good, and bad in the objective sense. But wrong will usually mean subjectively wrong, or blameworthy’ (RP, 25).

Doing Right by Wrong  287 Several things follow from this. First, in On What Matters, Parfit holds that the idea of moral wrong (and its deontic corollaries duty, obligation, and the like) are  essential to moral theories of right acts. A normative practical theory like Impartial-­Reason Act Consequentialism, which says that everyone has the weighti­ est impartial reasons to do whatever act would make outcomes best, is not, Parfit there says, so much a moral theory as an ‘external rival to morality’ (OWM i. 169). The reason, again, is that someone could accept IRAC and ‘be doubtful whether any acts are duties’ in the distinctively deontic moral sense in terms of which the moral theories of conduct comprising the Triple Theory are formulated (OWM i. 169). The sense of objective right and wrong that is in play in C in Reasons and Persons, however, is substantially the same as that of impartial ­practical reasons involved in IRAC. C’s theory of conduct concerns what there is reason to do. This means that C’s theory of objective right and wrong would not count as a moral theory by Parfit’s standards in On What Matters. Second, if Parfit were to stick by his policy of reserving ‘wrong’ for subjective wrongness in Reasons and Persons, C would then not really be proffering a theory of wrong acts at all. (C3) says that when someone does something that they believe will make the outcome worse (than it would have been had they done something else), they are ‘acting wrongly’. But that is not a claim about wrong acts in Ross’s sense; it is a claim about wrongful action. What is subjectively wrong in Reasons and Persons’s sense of being blameworthy was their doing what they did believing that the chosen act’s outcome was suboptimal. It is actions and not acts, in Ross’s helpful senses, that are blameworthy. Below I will defend a conceptual claim about moral wrongness of acts that does connect that concept to blameworthiness. But this must be done carefully to  respect the fact that only actions, and not acts, can be blameworthy.9 What I propose as conceptually necessary is that an act is wrong if, and only if, it is an act of a kind that it would be blameworthy to perform without excuse. It is the action of performing the act that is blameworthy, not the act itself. In Parfit’s terms in Reasons and Persons, therefore, there is thus no such thing as a wrong act (in the Rossian sense). Finally, connecting subjective wrongness to blameworthiness (of actions) as Parfit does in Reasons and Persons creates a tension, given C’s hegemonic reach. ‘C applies to everything’, Parfit says (RP, 25). Consequently, C propounds a consequentialist theory of blameworthiness that is independent of (C3)’s claim that doing what one believes will have suboptimal outcomes is blameworthy. C includes a ‘reaction theory’ that ‘claims that we ought to feel remorse, and to blame others, when this would make the outcome best’ (RP, 113). There is

9  Strictly speaking, it is the person who is to blame for having acted as they did (say, believing that the consequences of their so doing would be suboptimal).

288  Stephen Darwall obviously no guarantee that having the attitude of blame toward actions done in the belief they will have suboptimal outcomes will itself have optimal outcomes. In any case, it is a widely accepted consequence of Strawson’s arguments in ‘Freedom and Resentment’ that the good consequences of having reactive attitudes like blame are reasons of the wrong kind to support the claim that their attitudes’ objects are blameworthy or culpable (Strawson 1968). Moreover, this is a point to which Parfit himself is committed in On What Matters by his rejection of ‘state-­given’ reasons. If all reasons are ‘object-­given’ rather than state-­given, as he there argues, then the blameworthiness of some action must be grounded in features of that action rather than in aspects of outcomes of holding any attitude toward it (OWM i. 50–1, 420–32). It is no exaggeration, I think, to say that Reasons and Persons tends to treat blame, not as an attitude having its own normative standards, conceptually related to the wrongness of acts, but as an instrument that can be used for good or ill. Moreover, the uses of this instrument are to be assessed by the same general normative concepts (oughts) that are involved in C1’s claim that we ought always to do whatever will be for the best (impersonally or impartially conceived). So formulated, C and C1 are not moral claims as Parfit understands the moral in On What Matters. C1 is not intended to provide a theory of right and wrong acts as the theories comprising the Triple Theory—Kantian Contractualism, Scanlonian Contractualism, and Rule Consequentialism—attempt to do.

4.  On What Matters on Moral Wrong ‘Wrong’ can obviously have senses that have nothing to do with morality. Answers on an algebra quiz, fashion choices, or faux pas of etiquette can be wrong in ways that do not imply wrongdoing in the morally relevant sense. But what is that sense? That is the question on which On What Matters’s chapter  7 (‘Moral Concepts’) focuses. Parfit begins with the assumption that ‘ “wrong” has only one moral sense’, which he calls the ‘ordinary sense’ (OWM i. 150). This assumption is most plausible, he says, with objective right and wrong, when agents know all the morally relevant facts. Senses of subjective rightness can then be defined in terms of objective rightness, conditional on the facts being as the agent believes them to be or as the evidence warrants believing them to be (OWM i. 150–1). What is most important for our purposes is Parfit’s investigation of ‘whether we can explain [the] ordinary sense [of moral wrong], and whether there is more than one such sense’ (OWM i. 164). Parfit first shows how once we have the concept of moral wrong, we can then define its deontic moral correlates—moral requirement, obligation, duty, and permission—in obvious ways. What is morally permissible is what it would not be wrong to do. And what is morally obligatory is

Doing Right by Wrong  289 what it would be wrong not to do (OWM i. 165). This much is uncontroversial. What is in question is whether the concept of moral wrong can be defined itself. Parfit’s position in On What Matters is that it cannot be. The ‘ordinary’ concept of moral wrong is, ‘like the concept of a reason’, ‘indefinable’ (OWM i. 165). Parfit’s point is not that the concept of moral wrong is more fundamental than its deontic moral correlates—duty (obligation, requirement) or permission. He realizes that just as we can define duty in terms of wrong, so also can we define wrong as failing to do one’s duty or as what is not morally permitted. Rather, the ‘group of [deontic] concepts all have a common element which we cannot helpfully explain merely by using words’ (OWM i. 165). The circle of deontic moral concepts can only be defined in terms of one another.10 ‘Like the concept of a reason, and the decisive-­reason-­implying concept should or ought, these moral concepts must be explained in other ways, by getting people to think certain thoughts’ (OWM i. 165). Now our aim, recall, is to understand morality’s normativity. On the reasons fundamentalist assumption we have been accepting, our question is how moral wrong can be normative in a reasons-­involving sense. Parfit uses the phrase ‘mustn’t-­be-­done’ to express the indefinable ‘ordinary’ concept of moral wrong (OWM i. 165). Even here, of course, we must understand this phrase in a distinctively moral way. We can easily say that I must not wear this tie with that coat, or give that answer on the algebra exam, or respond in the first person to an invitation addressed in the third person in violation of a norm of etiquette, without saying anything about what it would be morally wrong for me to do. These are all examples of rule-­involving rather than reason-­involving normativity. Our aim, again, is to understand how moral wrong (and its correlate deontic moral concepts) can be normative in the reason-­involving sense. This is made more urgent by the fact that Parfit, in my view, completely correctly, rejects any attempt to define the ordinary sense of ‘wrong’ in terms of normative reasons against acting, even moral reasons. He considers three such attempts: Decisive-­Reasons sense: ‘wrong’ means ‘what we have decisive reasons not to do’. Decisive-­Moral-­Reasons sense: ‘what we morally ought to do’ means ‘what we have decisive moral reasons to do’, and ‘wrong’ means ‘what we have such ­reasons not to do’. Morally-­Decisive-­Reasons sense: ‘what we ought morally to do’ means ‘what we have morally decisive reasons to do’, and ‘wrong’ means ‘what we have such ­reasons not to do’.  (OWM i. 166–7) 10  Compare my claim that second-­personal concepts, which I argue deontic moral concepts to be, constitute a mutually definable circle that cannot be defined with concepts that are not second-­ personal (Darwall 2006: 12).

290  Stephen Darwall The decisive-­reasons sense is easily rejected since ‘we often believe that we have decisive reasons to act in some way, though we do not believe that we ought morally to act in this way’ (OWM i. 166). The decisive-­moral-­reasons sense overcomes this objection, but faces two problems. One is an issue that is common also to the morally-­decisive-­reasons sense, namely, that it is unclear what makes a reason for acting a moral reason. I argue elsewhere that there is a plausible way of understanding the concept of a moral reason for acting, but in terms of deontic moral concepts like moral wrong, namely, as pro tanto moral obligation-­making considerations (Darwall 2019). Obviously, that is no help here, since we are looking for an account of deontic moral concepts in terms of moral reasons, not vice versa. The other problem is ‘that we already have the concept of what we have decisive reason to do, and it adds little to claim that some of these reasons are moral reasons’ (OWM i. 167). This leaves the morally-­ decisive-­ reasons sense, according to which being wrong is being something there are morally decisive reasons not to do in the sense that there is more moral reason not to do them than any alternative course of action that involves doing them (that is, there is more moral reason to do something else).11 We have already seen one problem this definition faces, namely, that it is not clear that there exists an account of moral reasons that both makes it a normatively significant concept and is independent of deontic moral concepts (like wrongness). It looks as if all the normative work is being done by the concept of a reason for acting with nothing of normative significance being added by the fact that a reason for acting is a moral one. We can of course stipulate a meaning for ‘moral reason’, say, in other-­regarding terms, but no normative consequences can follow from this stipulation. What is more, to attempt to understand wrong and moral obligation in terms of morally decisive reasons is to collapse any distinction between the morally choiceworthy and the morally obligatory. It would make supererogation conceptually impossible and render unintelligible moral disagreements that seem not only intelligible but perfectly ordinary. Consider, for example, a disagreement between an act consequentialist and someone like Scheffler in The Rejection of Consequentialism (Scheffler 1982). Both could agree that morality always most recommends the optimific act, that that is always what there is most moral reason to do. Consistently with that, however, they could disagree about whether especially burdensome costs to the agent can render an action that would otherwise be morally obligatory morally permissible. The act consequentialist thinks that non-­optimific acts are always morally wrong, whereas the Schefflerian holds that non-­optimific acts are sometimes not wrong when they are covered by

11  For Parfit’s account of ‘decisive reason’, see OWM i. 32–3.

Doing Right by Wrong  291 an agent-­relative prerogative or permission that is operative when personal costs are sufficiently high in relation to the impartial benefits. Suppose, for example, that an agent is in a position to save someone drowning in turbulent seas at great personal risk or cost. The Schefflerian might think that the personal risk is not so much a moral reason against so acting (assuming that they do not make it obligatory not so to act), as a consideration that defeats what would otherwise be a moral obligation. For such a disagreement to be so much as possible, the parties must distinguish the distinctively deontic concept of moral obligation from any ‘moral ought’ of moral reasons. They would be agreeing about what should be done in the morally-­decisive-­reasons sense, but disagreeing about whether failing to perform that act, in the circumstances, would be ­morally wrong. However, in this disagreement about moral wrongness what normative issue would be in dispute? What is the normativity of wrongness, if normativity is fundamentally reason-­involving, and wrongness cannot be defined in terms of morally decisive reasons? The most promising direction to answering this question, as I see it, is provided by two further ‘definable’ ‘versions’ of the concept of moral wrong that Parfit considers, which he calls the blameworthiness and reactive-­attitude senses respectively: in the blameworthiness sense ‘wrong’ means ‘blameworthy’, in the reactive-­attitude sense, ‘wrong’ means ‘an act of a kind that gives its agent reasons to feel remorse or guilt, and gives others reasons for indignation and resentment.   (OWM i. 165)

An obvious problem with Parfit’s definition of wrong in terms of blame­worthi­ ness is that, as we noted above, blameworthiness is a property of actions, whereas wrongness as Parfit is considering it in On What Matters is a property of acts. The normative theories of right and wrong that comprise the Triple Theory are accounts of right and wrong acts in Ross’s sense of ‘things to do’. What is blameworthy, by contrast, are doings of such things: actions, in Ross’s sense. That problem can be fixed by saying, similarly to Parfit’s reactive-­attitude definition, that ‘wrong’ means ‘act of a kind that it would be blameworthy to perform’. But there is a further problem. Performances of wrongful acts are not always blameworthy, since they may be covered by some excuse. An excuse does not make the act performed any less wrong, but it does affect the blameworthiness of the action consisting in the act’s performance. The reactive-­attitude definition of wrongness avoids the first problem, but not the second. There is, however, an obvious way of fixing both definitions that brings them  together and very close to the kind of claim that I have defended in my

292  Stephen Darwall work (Darwall 2006, 2013a, 2013b, 2016, 2019). Strictly speaking, what I have been defending has been less a definition than the conceptually necessary ­equivalence claim: An act is wrong if, and only if, it is an act of a kind that it would be blameworthy to perform without excuse (see e.g. Darwall 2016: 266).

I take blameworthiness to be the normative concept of what justifies or warrants blame. On a reasons fundamentalist approach, it is the concept of what there is reason to blame. Blame, in this formulation, is not any blaming activity where the relevant reasons might concern, among other things, the beneficial effects of engaging in it. It is rather an attitude, so the relevant reasons (reasons ‘of the right kind’, as it is said) must be, as Parfit argues is true of the attitudes of desire and belief, ‘object given’ rather than ‘state given’ (OWM i. 50–1, 420–32). Blame is, I argue, a Strawsonian reactive attitude that has the same ‘interpersonal’, or, as I put it, second-­personal, character as do all reactive attitudes (Darwall 2006, 2013a, 2013b). In contrast with ‘objective attitudes’, reactive attitudes like blame are felt from the perspective of implicit relating to their objects (Strawson 1968: 79). Blame sees its object from a second-­ person perspective and bids for acknowledgment of the authority to hold, and to hold themselves, accountable. It comes with an implicit RSVP. I argue that blame is fundamentally the same ‘holding accountable’ attitude whether it is felt from the perspective of a victim (resentment), a perpetrator (guilt), or a disinterested third party (blame). Guilt, for example, is self-­blame— felt as if from the perspective of a representative person or member of the moral community for oneself as an accountable individual. It follows that blame is warranted in the object-­given sense that amounts to an action’s culpability, if, and only if, resentment and guilt are also. Once proper account has been taken of the act/action distinction and a phrase has been added to account for the possibility of nonculpable wrongdoing, therefore, properly modified blameworthiness and reactive attitude senses come to the same thing: In the modified reactive-­attitude/blameworthiness sense, ‘wrong’ means an act of  a  kind that would warrant guilt, resentment, or blame in the reasons-­ involving sense. On a reasons fundamentalist account, normative notions like desirable, credible, and blameworthy are to be understood in terms of there being reason (of the right kind) to have the attitude of desire, belief, and moral blame respectively.12

12  Brandt (1946) refers to these as Y-­able notions.

Doing Right by Wrong  293 Understanding wrongness in terms of what there would be reason to blame, ­lacking excuse, has the great benefit of revealing moral wrong to be a normative concept according to reasons fundamentalism. If the ordinary sense of ‘moral wrong’ is indefinable, as Parfit says, it remains mysterious how it can be a normative concept, since it can be, according to reasons fundamentalism, only if it entails propositions about normative reasons. The modified reactive-­ attitude/blame­ worthi­ness account agrees with Parfit in rejecting any attempt to reduce deontic moral concepts like wrongness to normative reasons for acts, even to moral ­reasons. No such account, it agrees, can capture the deontic moral concepts’ dis­ tinct­ive normative profile. But neither can holding that an act’s being wrong entails nothing about normative reasons of any kind as an indefinability account requires. It might be objected to a modified reactive-­attitude/blameworthiness def­in­ ition of ‘wrong’ that if it is accepted, we can no longer say that an action is blameworthy because it is wrong.13 But that is not strictly true, since the definition does not say that blameworthiness and being wrong are the same thing. It must also be the case that there is no excuse. If we add that in, we might still say that an action is blameworthy because the agent performed a wrong act without excuse, realizing that we are just listing blameworthiness’s conceptual prerequisites. It is true that according to the definition, if we stipulate that an agent has performed a wrong act without excuse, it is no longer a conceptually open question whether the action is blameworthy. But that just seems obviously true. It seems incoherent to claim that an action was inexcusably wrong but not blameworthy.14 If blame were simply an assertion of unexcused wrongdoing, then ‘wrong’ might be indefinable with it nonetheless being true that the blameworthiness of unexcused wrongdoing is conceptually guaranteed, since it is true quite generally that what is known to be true is warrantedly assertible or believed. But blame is not simply the assertion of or belief in unexcused wrongdoing; it is a reactive attitude with its distinctive second-­personal profile (Strawson 1968; Darwall 2006). So if the blameworthiness of unexcused wrongdoing is conceptually guaranteed, it must be guaranteed in some other way. The modified reactive attitude/blameworthiness definition accounts for that. Despite steadfastly maintaining that the ‘ordinary’ sense of moral wrong is indefinable, Parfit nonetheless sums up his discussion of moral concepts by saying, somewhat remarkably: We need not choose between theses senses of ‘wrong’, and the concepts they  express. It is worth using several of these concepts, asking for example, 13  I am indebted here to Ketan Ramikrishnan. 14  It might be thought that Parfit can evade the criticism that on his account of the ‘ordinary’ sense of wrong, it is not a normative notion (on a reasoning-­involving conception of normativity), by simply denying that wrong is a normative notion. But if that were true, it would not be incoherent to claim that an action was wrong and inexcusable but not blameworthy.

294  Stephen Darwall which acts are wrong in the indefinable, justifiabilist, reactive-­ attitude, or ­blameworthiness senses. In the rest of this book, I shall use ‘ought morally’ and ‘wrong’ vaguely, in some combination of these senses.  (OWM i. 174)15

This suggests that Parfit is implicitly relying on a conceptual connection between wrongness and accountability when he defends the theories that comprise the Triple Theory. And that, I believe, goes a long way toward explaining what can otherwise seem such a puzzling divergence between his approaches to consequentialist theories in Reasons and Persons and On What Matters, defending AC in Reasons and Persons and defending RC in On What Matters. A similar movement of thought can be found in Mill’s Utilitarianism. Early in chapter 2, Mill writes that ‘actions are right in proportion as they tend to promote happiness; wrong as they tend to produce the reverse of happiness (Mill 1998: ch. 2, para. 2). This can sound close to AU, although Mill’s ‘tendency’ may suggest something somewhat different. In chapter  5, however, Mill focuses on some conceptual features of the concept of justice and, more generally but relatedly, the concepts of moral obligation and wrong. There Mill says that ‘the idea of justice supposes two things; a rule of conduct, and a sentiment which sanctions the rule’ (ch. 5, para. 23). The sentiment is ‘resentment’, although Mill’s account of the attitude is more retaliatory than Strawson’s.16 This account ties justice to reactive attitudes conceptually. Mill remarks that his account of justice ‘contains, as yet, nothing to distinguish that obligation [of justice] from moral obligation in general’ (ch. 5, para. 14). And then he famously remarks, ‘‘we do not call anything wrong, unless we mean to imply that a person ought to be punished in some way or other for doing it; if not by law, by the opinion of his fellow creatures; if not by opinion, by the reproaches of his own conscience’ (ch. 5, para. 14). We may think there are strong moral reasons for people to do something and ‘dislike or despise them for not doing’ it; but unless we think ‘blame’ or some other form of ‘punishment’ is warranted, ­perhaps just the feeling of guilt, we do not think it ‘a case of moral obligation’ (ch. 5, para. 14). These conceptual reflections do not lead Mill to give up utilitarianism. But they seem to lead him to a different form than AU. Justice is, as a conceptual matter, aptly protected by sanction-­backed rules that enshrine rights, where rights are ‘something that society ought to defend me in the possession of ’ through sanction-­backed rules (ch. 5, para. 25). To the question why we ought to be protected by such rules, Mill replies, ‘I can give . . . no other answer than general 15 ‘In the justifiabilist sense, “wrong” means “could not be justified to others” ’ (OWM i. 166). In Darwall (2006), I argue that the contractualist approach underlying this proposal is best grounded in a second-­personal framework for which I there argue and argue conceptually underpins a modified reactive-­attitude/blameworthiness account of deontic moral concepts. 16  On this, see Darwall (2013b: 50–72).

Doing Right by Wrong  295 utility (ch. 5, para. 25). This is a rule-­utilitarian account of the obligation of justice, which is easily extended to a rule-­utilitarian account of moral obligation in general, or RU. I conjecture that something similar is going on in Parfit’s thought. Once Parfit focused on the distinctive features of deontic moral concepts like moral wrong, especially, their conceptual connection to accountability, C, as he formulated it in Reasons and Persons, no longer seemed to provide a (deontic) moral theory at all. It is one thing to think, as C2 held, that what we have reason to do is ‘whatever would make the outcome best’. But it is quite another to think that failure so to act would be wrong in the sense of something we are justifiably held accountable for, something that is blameworthy lacking excuse.

5.  Conclusion: ‘The Profoundest Problem’ I have argued that the normativity of deontic moral concepts, like wrongness, is to be accounted for by their conceptual connection to justified accountability-­ seeking attitudes, that it is a conceptual truth that an act is wrong if, and only if, it is an act of a kind that it would be blameworthy to perform without excuse. This can seem puzzling, however. After all, what is morally obligatory or wrong not to do is an act. How can the normativity of the moral obligation to do something be accounted for in terms of reasons to have an attitude (blame) toward someone who fails to do it without excuse? Nothing seems to be said about what anyone ought to do in a reasons-­involving sense. Nothing seems to follow about any normative reason to do what we are obligated to do. We can put the issue in terms of Sidgwick’s ‘profoundest problem’ of what to do when ‘duty and interest’ conflict. The basic problem does not require a conflict with self-­interest, although it does on Sidgwick’s formulation.17 The underlying concern is to explain how the fact that an act is morally obligatory ensures that there is some reason to do it, indeed that there is decisive reason, practical reasons of whatever kind to the contrary notwithstanding. In this concluding section, I will show how an account of deontic moral concepts that ties them to accountability conceptually in the way the modified reactive-­ attitude/blameworthiness account does can solve these apparent problems. The key to the solution, as I have argued in earlier work, is to see that although deontic moral claims are not themselves reason-­involving practical claims, they nonetheless entail reason-­involving practical claims. What makes this so, I claim, is a presupposition of blame as a reactive attitude (Darwall 2016).

17  Since it concerns, for Sidgwick, the equally fundamental ‘axioms’ of prudence and universal benevolence.

296  Stephen Darwall Bernard Williams, among others, noticed that it is incoherent to blame someone for doing something one thinks they had sufficient normative reason to do (Williams 1985: 193; see also Gibbard 1990: 299–300; Skorupski 1999: 42–3; Shafer-­Landau 2003: 181–3; Darwall 2006; 2016; Portmore 2011: 44). Some kind of inconsistency seems to be involved in having the attitude of blame toward someone while simultaneously fully accepting that they had sufficient reason for doing what they did. After all, if someone can establish that they had sufficient reason, then they will have adequately accounted for themselves. They will have justified what they did and shown, therefore, that it was not actually wrong. Blame must assume that they cannot, that there was insufficient reason for doing what they did. Now, if an act is morally obligatory, all things considered, it is conceptually necessary that it is an act of a kind that, in the circumstances, it would be blameworthy to fail to perform without excuse. But if that is so, it follows that the act must be one there cannot be sufficient reason to omit, because if there were such reason, that would show that blame would be unjustified (Darwall 2016). On the modified reactive-­attitude approach I favor, the distinctive normativity of deontic morality consists in the fact that deontic moral judgments are normative, in the first instance, for reactive attitudes rather than for acts. The deliberative question they distinctively concern is whether to blame someone for an act of the relevant kind in the relevant circumstances, assuming the agent lacks an excuse. This gives us the conceptual space to hold that an act might be recommended by moral reasons to act, but nonetheless not be morally required or obligatory. If the act would be wrong, that is, culpable lacking excuse, then it is conceptually guaranteed that it is not one the agent could have sufficient reason to choose. The choiceworthiness of obligatory conduct in the reason-­involving sense is not the same thing as their being obligatory, but is rather a conceptual consequence of presuppositions of blame, which is the attitude for which deontic moral judgments are normative in the first instance. A modified reactive-­attitude/blameworthy account of deontic moral concepts like moral wrong can thus both account for their normativity (in the first, instance, for accountability-­ seeking attitudes) and explain their decisive normativity for acts, thereby solving the ‘profoundest problem’. These are conceptual benefits of doing right by wrong.

Bibliography Brandt, Richard (1946). ‘Moral Valuation.’ Ethics 56: 106–21. Broome, John (2018). ‘Reasons Fundamentalism and What is Wrong With It.’ In The Oxford Handbook of Reasons and Normativity, edited by Daniel Star. Oxford: Oxford University Press.

Doing Right by Wrong  297 Chappell, Richard (2012). ‘Fittingness: The Sole Normative Primitive.’ Philosophical Quarterly 62: 684–704. Darwall, Stephen (2006). The Second-Person Standpoint: Morality, Respect, and Accountability. Cambridge, MA: Harvard University Press. Darwall, Stephen (2013a). Morality, Authority, and Law: Essays in Second-Personal Ethics I. Oxford: Oxford University Press. Darwall, Stephen (2013b). Honor, History, and Relationship: Essays in Second-Personal Ethics II. Oxford: Oxford University Press. Darwall, Stephen (2016). ‘Making the Hard Problem of Moral Normativity Easier.’ In Weighing Reasons, edited by Errol Lord and Barry Maguire. Oxford: Oxford University Press. Darwall, Stephen (2017). ‘Morality, Blame, and Internal Reasons.’ In Does Anything Really Matter?: Essays on Parfit on Objectivity, edited by Peter Singer. Oxford: Oxford University Press. Darwall, Stephen (2019). ‘What Are Moral Reasons?’ The 2017 Amherst Lecture in Philosophy . Foot, Philippa (1972). ‘Morality as a System of Hypothetical Imperatives.’ Philosophical Review 81: 305–16. Frankena, William (1942). ‘Obligation and Value in the Ethics of G. E. Moore.’ In The Philosophy of G. E. Moore, edited by Paul A. Schilpp. La Salle, IL: Open Court. Gibbard, Allan (1990). Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press. Howard, Christopher (2019). ‘The Fundamentality of Fit.’ In Oxford Studies in Metaethics, Vol. xiv. Oxford: Oxford University Press. MacPherson, Tristram, and Plunkett, David (2017). ‘The Nature and Explanatory Ambitions of Metaethics.’ In The Routledge Handbook of Metaethics, edited by Tristram MacPherson and David Plunkett. New York: Routledge. McHugh, Conor and Way, Jonathan (2016). “Fittingness First.” Ethics 126: 575–606. Mill, John Stuart (1998). Utilitarianism, edited by Roger Crisp. Oxford: Oxford University Press. Moore, G.  E. (1903/1993). Principia Ethica. Rev. edn. with the preface to the (projected) 2nd edn. and other papers. Edited with an introduction by Thomas Baldwin. Cambridge: Cambridge University Press. Parfit, Derek (1986). Reasons and Persons. Oxford: Oxford University Press. Parfit, Derek (2011). On What Matters. Vols. i and ii. Oxford: Oxford University Press. Parfit, Derek (2017). On What Matters. Vol. iii. Oxford: Oxford University Press. Portmore, Douglas  W. (2011). Commonsense Consequentialism: Wherein Morality Meets Rationality. New York: Oxford University Press. Rawls, John (1971). A Theory of Justice. Cambridge, MA: Belknap Press of Harvard University Press. Raz, Joseph (1999). Engaging Reason. Oxford: Oxford University Press.

298  Stephen Darwall Ross, David (1930/2002). The Right and the Good. Oxford: Clarendon Press. Scanlon, T. M. (2014). Being Realistic About Reasons. Oxford: Oxford University Press. Scheffler, Samuel (1982). The Rejection of Consequentialism: A Philosophical Investigation of the Considerations Underlying Rival Moral Conceptions. Oxford: Oxford University Press. Shafer-Landau, Russ (2003). Moral Realism: A Defense. New York: Oxford University Press. Sidgwick, Henry (1907/1967). The Methods of Ethics. 7th ed. London: Macmillan. Skorupski, John (1999). Ethical Explorations. Oxford: Oxford University Press. Strawson, P.  F. (1968). ‘Freedom and Resentment.’ In Studies in the Philosophy of Thought and Action. London: Oxford University Press. Williams, Bernard (1985). Ethics and the Limits of Philosophy. Cambridge, MA: Harvard University Press.

13

Giving Reasons and Given Reasons John Broome

1.  Introduction: Giving Reasons and Given Reasons Derek Parfit says that the concept of a reason is fundamental and indefinable.1 This chapter analyses Parfit’s concept of a reason, and evaluates his claim. His concept of a reason differs from most philosophers’. Most philosophers apply the concept of a reason to particular facts. For example, some of us think the fact that walnuts will kill you is a reason for you not to eat walnuts. But Parfit thinks differently. He says that the fact that eating walnuts will kill you gives you a reason not to eat walnuts (OWM i. 32). He does not say it is a reason. Parfit’s use of ‘gives’ in this way is deliberate and explicit. He openly contrasts his practice with other authors’: ‘Rather than saying that certain facts give us reasons, some people say that these facts are reasons for us’ (OWM i. 32). He sticks to his practice consistently throughout On What Matters. It is not a whim; it must be significant. This is why I say his concept of a reason differs from the prevalent one, and I take its implications seriously.2 Just after the sentence I quoted in the previous paragraph, Parfit says ‘But these people’s claims do not conflict with mine, since these are merely different ways of saying the same things.’ At first, you might think Parfit is here asserting that his meaning of ‘a reason’ is the same as other people’s. But actually he is asserting the opposite. He asserts that the phrase ‘gives you a reason’ means the same as ‘is a reason for you’. But since ‘gives you’ has a different meaning from ‘is . . . for you’, this implies that ‘a reason’ has different meanings in the two phrases. Similarly, ‘Put the horse in the stable!’ means the same as ‘Stable the horse!’, and this implies that ‘stable’ has different meanings in the two commands. My thanks to Jeff McMahan for valuable comments. Research for this paper was supported by ARC Discovery Grants DP140102468 and DP180100355. 1  On What Matters (Oxford: Oxford University Press, 2011), i. 31. Further references will be to OWM with volume and page numbers in the text. 2  Ruth Chang tells me I should attach no particular significance to Parfit’s usage: I should assume he means the same by ‘a reason’ as most people. She reports that this conclusion emerges from discussions she had with Parfit about his usage. It puts me in a quandary. Chang’s report is authoritative. But the text is also authoritative, and what Parfit says in the text is clear, deliberate, and consistent. I have chosen to take the text at face value. This is the choice that gives Parfit the greater credit, since this chapter argues that Parfit did well to shift attention from giving reasons to given reasons. John Broome, Giving Reasons and Given Reasons In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0014

300  John Broome ‘Gives’ is not a mere predicating or identifying word like ‘is’. In this context, giving can only be understood as a sort of explaining or making the case. When Parfit says The fact that eating walnuts will kill you gives you a reason not to eat walnuts.

he could have said, with much the same meaning, any of The fact that eating walnuts will kill you makes it the case that you have a reason not to eat walnuts.

or The fact that eating walnuts will kill you explains why you have a reason not to eat walnuts.

or You have a reason not to eat walnuts because eating walnuts will kill you.

The giving relation holds between two different things: what gives and what is given. What gives in the example is the fact that eating walnuts will kill you. This is a reason in the common terminology of philosophy. Let us call it a ‘giving reason’. What is given is a reason in Parfit’s terminology. Let us call it a ‘given reason’. Parfit’s concept of a reason is the concept of a given reason rather than a giving reason.

2.  Favouring and Owning The two concepts share some features. Both are monadic in the sense that they apply to individual things. However, both are relational in the sense that they apply to something only because that thing stands in particular relations to other things. Compare the concept of being a sister. This is a monadic concept. It applies to individual things, in this case individual people. Only an individual person can be a sister. But the concept applies to someone only because she stands in a particular relation to another thing, in this case another person. She is a sister of someone else. From a monadic relational concept such as the concept of a sister, a narrower monadic concept can be constructed that embeds a relatum. An example is the concept of a sister of the president. The concept of a reason, whether giving or given, is monadic, but it is doubly relational. The concept applies to something because of the relations that thing stands in to two other things: on the one hand to a state of affairs, and on the other hand to a person or other agent. It is a reason for some state of affairs and it

Giving Reasons and Given Reasons  301 is a reason for someone or some agent. Both relationships can be expressed by the same preposition ‘for’, but ‘for’ has different meanings in the two cases. The relationship to the state of affairs may be called ‘favouring’, and ‘in favour of ’ can replace ‘for’. Suppose there is a reason for you not to eat walnuts. This reason is in favour of—for—the state of affairs of your not eating walnuts. In this chapter, I shall most often deal with concepts of a reason in which a favoured state of affairs is embedded, such as the concept of a reason for your not eating walnuts. That is, I shall most often deal with the narrower concept of a reason for N’s Fing or more colloquially a reason for N to F, rather than the broader concept of a reason. The narrower concept is monadic like the broader one. I use schematic letters for the sake of generality. ‘N’ denotes a person or other agent and ‘F’ is a verb phrase that may describe acting, believing, hoping for or anything else. N’s Fing is a state of affairs. The relationship of a reason to the person or agent may be called ‘ownership’.3 It is difficult to describe. Various phrases express it. The reason is for you. You have the reason. The reason applies to you. It is agent-­relative to you. It is your responsibility. It is best explained by means of an example. Suppose Alex has committed a crime. Then there is a reason for—in favour of—Alex’s going to prison. But many of us think this reason is not owned by Alex: Alex does not have a reason to go to prison; it is not his responsibility to get himself to prison. Perhaps the judicial system owns the reason. In The Possibility of Altruism,4 Thomas Nagel argues that every reason is owned by everyone. If he is right, it follows that, if there is a reason for Alex’s going to prison, Alex owns this reason just as everyone else does. Then Alex does have a reason to go to prison. I do not dispute Nagel’s claim, but in the example I set it aside for the sake of illustrating the idea of ownership. Nagel recognizes ownership, and makes the particular claim about it that every reason is owned by everyone. If that is true, ownership exists but is harder to illustrate. The sentence ‘Alex has a reason to go to prison’ makes it explicit that the reason is owned by Alex. In general, ‘N has a reason to F’ says both that the reason is in favour of N’s Fing and that the reason is owned by N. Many philosophers use the sentence ‘N has a reason to F’ to say that N stands in some epistemic relation to the reason—for instance, that N knows the reason obtains. This is not correct English. Even if Alex knows that he has committed a crime so there is a reason for him to go to prison, many of us still think that Alex does not have a reason to go to prison. It is unfortunate that philosophers misappropriate the expression ‘has a

3  Ownership of reasons is investigated in my Rationality Through Reasoning (Oxford: Blackwell, 2013), 65–9. 4  (Oxford: Oxford University Press, 1970), esp. ch. 10. On p. 91, Nagel declines to distinguish the claim that all reasons are owned by everyone from the claim that no reason is owned by anyone. But I explain in my Rationality Through Reasoning, 66–8, that his argument supports the former claim and not the latter one.

302  John Broome reason’, because in doing so they obscure the ownership of reasons. In English, ‘has’ expresses ownership. On the other hand, the sentence ‘There is a reason for Alex to go to prison’ is ambiguous. In general, ‘a reason for N to F’ is ambiguous. This phrase may be parsed ‘(a reason for) (N to F)’. In that case ‘for’ has is favouring meaning and the phrase has the same meaning as ‘a reason for N’s Fing’. Or the phrase may be parsed ‘(a reason for N) (to F)’. In that case ‘for’ has its ownership meaning and the phrase ascribes ownership of the reason to N. The reason’s favouring relation to N’s Fing is implicit in the grammar. So, in one of its meanings ‘There is a reason for Alex to go to prison’ ascribes ownership to Alex, and in another it does not. However, from this point on I fix the meaning of ‘a reason for N to F’. I use it always with the implication that N owns the reason. I do so because Parfit does the same.5

3.  Reasons Primitivism When Parfit says that the concept of a reason is fundamental and indefinable, he is stating the doctrine I call reasons primitivism. This is the doctrine that the concept of a reason cannot be defined in terms of other concepts. It is one element of the ­reasons first movement, which has swept over the philosophy of normativity in recent decades. Parfit is a leader of this movement, and reasons primitivism lies at its core. It is closely connected with another element of the reasons first movement: the metaphysical doctrine that may be called reasons fundamentalism or reason fundamentalism. This doctrine is about the property, rather than the concept, of being a reason. It claims that this property is an irreducible feature of normativity.6 Parfit concentrates on the concept rather than the property, and for that reason I do the same. Other philosophers besides Parfit are reasons primitivists,7 but Parfit’s version of the doctrine differs from most because he claims the concept of a given reason is primitive whereas they claim that the concept of a giving reason is primitive. I shall argue that his is the better version. I do not accept reasons primitivism myself,8 5  I also assume that every reason has an owner. At the beginning of this section, I said this is a feature of the concept of a reason. That is debatable; possibly there are unowned reasons. In ‘Ought and Moral Obligation’, in his Moral Luck (Cambridge: Cambridge University Press, 1981), 114–23, Bernard Williams claims there are unowned oughts—at least according to his own interpretation of that paper in a later lecture. See my ‘Williams on Ought’, in Ulrike Heuer and Gerald Lang (eds.), Luck, Value and Commitment: Themes from the Ethics of Bernard Williams (Oxford: Oxford University Press, 2012), 247–65. If there are unowned oughts there must be unowned reasons. However, since Parfit assumes every reason has an owner, I need not engage in the debate about whether there are unowned reasons. 6  See my ‘Reason Fundamentalism and What Is Wrong With It’, in Daniel Star (ed.), The Oxford Handbook of Reasons and Normativity (Oxford: Oxford University Press, 2018), 297–318. 7  e.g. T.  M.  Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998), 17. 8  See my ‘Reason Fundamentalism and What Is Wrong with It’.

Giving Reasons and Given Reasons  303 but I think Parfit’s is the best version of it. If the doctrine is to be refuted, it is important to identify its best version. The fact that eating walnuts will kill you is a giving reason for you not to eat walnuts. It explains why you have a given reason not to eat walnuts. In general, a giving reason for N to F explains why N has a given reason to F. This is a conceptual connection between a giving reason and a given reason. It can be made into a definition of a giving reason in terms of a given reason: A giving reason for N to F is something that explains why N has a given reason to F.

More accurately: The concept of a giving reason for N to F is the concept of something that explains why N has a given reason to F.

There is no reciprocal definition of a given reason in terms of a giving reason. You might think a given reason for N to F could be defined as something whose existence is explained by a giving reason for N to F. But that is not so. A giving reason for N to F can explain all sorts of things, and only one of them is the existence of a given reason for N to F. For instance, the fact that eating walnuts will kill you is a giving reason for you not to eat walnuts. It explains why you have a given reason not to eat walnuts. But it may also explain why you are frightened of eating walnuts, why you have a given reason to banish walnuts from your house, why you have taken out life insurance, and other things too. The concept of a given reason is therefore more primitive than the concept of a giving reason. The latter can be defined in terms of the former, but not vice versa. This may come as a surprise. We normally think that what explains is more fundamental that what is explained. I do not deny that. In the walnut example, what explains is a giving reason, which is the fact that eating walnuts will kill you. I am not saying that this fact is less fundamental than what is explained, namely the fact that you have a given reason not to eat walnuts. I am talking about the primitiveness of concepts, not the primitiveness of things in the world such as facts. I am saying that the concept of a giving reason is less primitive than the concept of a given reason. Similarly, the fact that something is a magnet explains why ferrous objects are attracted to it. But the concept of a magnet is less fundamental than the concept of attraction; the concept of a magnet can be defined as the concept of something that attracts ferrous objects. Most philosophers use ‘a reason’ to refer to a giving reason. For them, reasons primitivism is the view that the concept of a giving reason is a primitive, undefinable concept. Reasons primitivism of this sort is easy to refute. I have just defined the concept of a giving reason in terms of the concept of a given reason, together with the concept of explaining.

304  John Broome Since Parfit uses ‘a reason’ to refer to a given reason, reasons primitivism for him is the view that the concept of a given reason is a primitive, undefinable concept. This is a better version of reasons primitivism; it is closer to the truth than the commoner one. The concept of a given reason is not so easily defined.

4.  What Is a Given Reason? The problem is that given reasons may not exist. A giving reason is typically thought to be a fact. It may alternatively be something else such as a view—the view from the top of Everest is a reason to climb it. But at any rate, there are things that fall under the concept of a giving reason. On the other hand, Parfit does not tell us what a given reason is. It would be nice to know what sorts of thing fall under the concept of a given reason— what sorts of thing Parfit’s term ‘a reason’ refers to—but we are not told. We are told when a person has a reason, and we are told what reasons are given by or provided by, but we are never told what a reason is. On What Matters contains many examples of facts that give reasons, but no examples of the reasons that are given. One thing is clear from the text: a given reason is not a giving reason. What is given is not the same as what gives. Whereas a giving reason is typically a fact, Parfit makes it perfectly clear that a given reason is not a fact. It seems that a given reason, if it exists, would have to be some sort of normative force. Plausibly, the fact that eating walnuts will kill you creates a normative force or push towards your not eating walnuts. I do not reject this idea; it is a credible interpretation of Parfit’s concept of a given reason. However, it leaves us with the metaphysical task of making sense of normative forces. And it is easy to be sceptical about such things. You might find them such mysterious entities that you are unwilling to believe they exist. They would be roughly analogous to physical forces, and there are grounds for being sceptical about the existence even of physical forces as entities.9 It is therefore unsafe simply to assume that given reasons are normative forces and leave it at that. For safety, from here on I shall take a sceptical stance towards them. I shall offer an alternative account of what Parfit refers to using the term ‘a reason’. To understand this alternative, start by remembering that we sometimes say ‘You have reason not to eat walnuts’, using the mass noun ‘reason’ rather than the count noun ‘a reason’. This sentence raises the question of what the mass noun 9 See John Bigelow, Brian Ellis, and Robert Pargetter, ‘Forces’, Philosophy of Science 55 (1988), 614–30.

Giving Reasons and Given Reasons  305 ‘reason’ refers to, which is parallel to the question of what Parfit’s term ‘a reason’ refers to. There is no temptation in this case to think ‘reason’ refers to a giving reason. A giving reason is a particular thing, and a mass noun cannot refer to a particular thing. The referent of a mass noun, if it has one, is not a particular but stuff of some sort. The mass noun ‘water’ refers to watery stuff. Notice next that—whatever ‘reason’ refers to—the sentence ‘You have reason not to eat walnuts’ says that a particular normative relation holds between you and the state of affairs of your not eating walnuts. Let us call it the owning-­reason relation. Compare this sentence with ‘You ought not to eat walnuts’. This also says that a particular normative relation holds between you and your not eating walnuts. Let us call it the owning-­ought relation. The two normative relations are closely parallel to each other. The owning-­reason relation is parallel to the owning-­ought relation, but weaker. The expression ‘you ought’ is not quantified, so it does not imply the existence of anything. On the other hand, grammatically, the expression ‘you have reason’ is implicitly quantified. It is equivalent to ‘there is some stuff that is reason, and it belongs to you’. So it implies the existence of reason-­stuff. But we do not have to accept this implication of grammar. We can instead think that the expression ‘you have reason’ is simply the means we have in English of saying that the owning-­ reason relation holds between you and a state of affairs. We can deny the existence of reason-­stuff and deny that the mass noun ‘reason’ refers to anything. We could even assert the existence of an owning-­reason relation by means of a sentence that is not implicitly quantified. Using an artificial terminology, we could say ‘You pro tanto ought not to eat walnuts’, meaning exactly the same as ‘You have reason not to eat walnuts’. The artificial sentence does not even suggest there is reason-­ stuff. So the mass noun ‘reason’ need not have a referent. The answer to the question of what it refers to may be ‘nothing’. Now back to Parfit’s count noun ‘a reason’. My alternative interpretation of this term is that he uses it to mean exactly what we mean by the mass noun ‘reason’. Parfit’s ‘You have a reason not to eat walnuts’ means the same as ‘You have reason not to eat walnuts’. It says that the owning-­reason relation holds between you and your not eating walnuts. Using a count noun in this way is grammatically defensible. For instance, we might say ‘She has sharp intelligence’ or ‘She has a sharp intelligence’, meaning the same thing either way. We might say ‘The message gave her hope of success’ or ‘The message gave her a hope of success’. How come? I think the explanation is that a count noun can refer to a piece or parcel of what is referred to by the corresponding mass noun, and a mass noun itself usually refers only to a piece or parcel. If you buy beer at the bar, you do no buy all beer, but only some beer, which may be referred to as a beer or several beers. The message did not give her all of hope, but only some hope, and this piece of hope may be called ‘a hope’.

306  John Broome Likewise, you have reason not to eat walnuts, but you do not have the whole of reason not to eat them. You have only a part, and that part may be called ‘a reason’. True, there may be metaphysical doubt about what reason is, but this usage is impervious to metaphysical doubt. Parfit himself sometimes reverts to the mass noun in quantified expressions such as ‘more reason’ and ‘most reason’ (OWM i. 32). One example is: If we see dark grey clouds . . . that gives us some reason to believe that it will soon rain. If we know that gold weighs more than lead, which weighs more than iron, these facts give us a decisive reason to believe that gold weighs more than iron. (OWM i. 47)

Here, the quantified mass noun ‘some reason’ in the first sentence is matched with the qualified count noun ‘a decisive reason’ in the second. So Parfit is not distinguishing the count noun from the mass noun in this context. This adds evidence to my suggestion that Parfit uses ‘a reason’ in place of ‘reason’. My interpretation does encounter a further problem. Suppose you have two giving reasons not to eat walnuts. Let one be that eating walnuts will kill you and the other that you hate the taste of walnuts. In Parfit’s terminology, the fact that eating walnuts will kill you gives you a reason not to eat walnuts, and the fact that you hate the taste of walnuts gives you a reason not to eat walnuts. He could say these are different reasons, so you have two given reasons. Because he uses the count noun, he could make this claim. But all we can say with the mass noun is that each giving reason gives you reason not to eat walnuts. We cannot distinguish two given reasons. If the separate individuation of given reasons plays an essential role in Parfit’s account of reasons, my interpretation will therefore fail. I do not know whether separate individuation is essential; I have not gone through all Parfit’s arguments to check whether or not they can be formulated without individuation. If some of them cannot, Parfit’s account is still up against the metaphysical problem of making sense of given reasons. They would have to be normative forces, which are subject to scepticism. For safety, I adopted the sceptical stance and I am now following up its consequences. So I shall continue to assume that Parfit uses ‘a reason’ in place of ‘reason’, which means I have to assume that separate individuation is not essential. If my interpretation is correct, just as we can deny that the mass noun ‘reason’ has reference, we can deny that the count noun as Parfit uses it has reference. ‘You have a reason’ is simply Parfit’s way of saying that you stand in the owning-­reason relation to a state of affairs. On my interpretation, when Parfit writes of ‘the concept of a reason’, he means to refer to the concept of having a given reason for. This is the same as the concept of having reason for. It is a dyadic concept that corresponds to the owning-­reason

Giving Reasons and Given Reasons  307 relation, and it applies to whatever satisfies this relation. Like any dyadic relation, the owning-­reason relation is satisfied by pairs. Specifically, it is satisfied by any pair consisting of a person and a state of affairs when the person has reason for the state of affairs. The concept of having a given reason for applies to just these pairs. This is the way a sceptic about normative forces can treat the concept of a given reason. She can replace it with the concept of having a given reason for. This is a perfectly good dyadic concept that has application; it is not subject to the sceptical doubt that afflicts the concept of a given reason.

5.  Reasons Primitivism Again Given scepticism about normative forces, we need to reinterpret the doctrine of reasons primitivism as Parfit sees it. When Parfit says the concept of a reason is indefinable, we must interpret him as saying that the concept of having a given reason for is indefinable. The conclusions I reached about reasons primitivism in Section 3 are unaltered. The concept of a giving reason can be defined as before: The concept of a giving reason for N to F is the concept of something that explains why N has a given reason to F.

As I explained in Section 3, there is no reciprocal definition of the concept of having a given reason for in terms of the concept of a giving reason. The concept of having a given reason for is therefore more primitive than the concept of a giving reason. The latter can be defined in terms of the former, but not vice versa. Thomas Nagel’s definition of a giving reason in The Possibility of Altruism is a precedent for the one I have just provided. Nagel defines a giving reason in terms of having reason as follows: Every reason is a predicate R such that for all persons p and events A, if R is true of A, then p has prima facie reason to promote A.10

This definition has some features that are peculiar to Nagel’s own thinking: for instance, he assumes reasons are predicates and he inserts the qualification ‘prima facie’ for his own purposes. He also does not recognize that a giving reason explains why p has reason to promote A. But the bones of the definition are the same. Remember that having reason for is the same as having a given reason for.

10  The Possibility of Altruism, 47. On p. 48 this is explicitly said to be a definition.

308  John Broome

6.  Conclusion The common version of reasons primitivism—the doctrine that the concept of a giving reason is undefinable—is plainly false. A giving reason can easily be defined. Parfit uses ‘a reason’ to refer to a given reason. For him, reasons primitivism is the view that the concept of a given reason is undefinable. I have explained that there may be a metaphysical doubt about the existence of given reasons. It is safer to replace the concept of a given reason with the concept of having a given reason for. Then reasons primitivism is the view that the concept of having a given reason for is undefinable. This is the best interpretation of Parfit’s view. It is also the best version of reasons primitivism. It is closer to the truth than the common version. The concept of having a given reason for is not so easily defined. As it happens, I think it can be defined—though not so easily—in terms of the concept of ought. So even this version of reasons primitivism can be refuted. But that is another story altogether.11

Bibliography Bigelow, John, Ellis, Brian, and Pargetter, Robert, ‘Forces’, Philosophy of Science 55 (1988), 614–30. Broome, John, Rationality Through Reasoning (Oxford: Blackwell, 2013). Broome, John, ‘Reason Fundamentalism and What Is Wrong with It’, in Daniel Star (ed.), The Oxford Handbook of Reasons and Normativity (Oxford: Oxford University Press, 2018), 297–318. Broome, John, ‘Williams on Ought’, in Ulrike Heuer and Gerald Lang (eds.), Luck, Value and Commitment: Themes from the Ethics of Bernard Williams (Oxford: Oxford University Press, 2012), 247–65. Nagel, Thomas, The Possibility of Altruism (Oxford: Oxford University Press, 1970). Parfit, Derek, On What Matters, vols. i–ii (Oxford: Oxford University Press, 2011). Scanlon, T.  M., What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998). Williams, Bernard, ‘Ought and Moral Obligation’, in his Moral Luck (Cambridge: Cambridge University Press, 1981), 114–23.

11  See my ‘Reasons Fundamentalism and What Is Wrong with It’.

14

Reply to Parfit’s ‘Innumerate Ethics’ John Taurek

Derek Parfit and I seem to think about human pain and suffering in very different ways. Perhaps we can clarify these differences. We often compare in terms of more or less, greater or less, the pain we experience on one occasion with the pain we felt on another. Even with such comparisons there are problems. But I will ignore them. Often we are confident in passing such judgments. Just as often we will compare one person’s pain with another’s. With such comparisons the problems are even more notorious. I will ignore these too. As Parfit points out, my problem is not the problem of interpersonal comparisons. My problem lies in the attempt to compare the magnitude of one person’s pain with the summed or added pain of several. Let me explain what I mean with the aid of a thought experiment. I am contemplating alternative outcomes. In the one case a certain person, Z, will suffer pain. In the other a different person, B, will suffer. I am contemplating these alternatives with a view either to deciding which, if either, would be a worse outcome, a greater evil, or, imagining myself in a position to prevent either the one or the other but not both, I am asking myself whom I should spare. Of course there is something absurdly abstract about putting these questions to myself in the face of such woefully minimal information. In any actual case there is so much that could be relevant. Who are these people? What are their circumstances? What meaning might such pain have for them in their lives? How are they connected with me? In the absence of any such information who could say which would be a worse thing, a greater evil? Who could make up his mind that he ought to spare the one person rather than the other? Who could do anything but shrug? Still in my thought experiment I do want to abstract from all such matters. It is to be just a question of Z’s suffering pain for twenty-five seconds, or B’s being in pain for that long. I want to abstract from all this other information because I do recognize in general the moral relevance of the question: Who will suffer more? Whose pain will be greater? And I want to focus attention on this question. As I understand it this question calls for a matter-of-fact comparison between the magnitude of Z’s pain and that of B’s. I may go on to justify or explain my evaluational judgment by reference to such a comparison. But the comparative John Taurek, Reply to Parfit’s ‘Innumerate Ethics’ In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0015

312  John Taurek judgment that Z will suffer more pain or greater pain than B is not itself a moral or evaluative judgment. When I tell myself, in my thought experiment, that Z’s pain would be very much greater than B’s, when I think of Z suffering agonizing pain while supposing that B would suffer only minor pain, I am inclined to judge that, other considerations apart, it would be a worse thing, a greater evil, were Z to suffer than it would be if B does. In such a case were the choice mine to make, I would be inclined to spare Z rather than B. On the other hand, when I tell myself that Z’s pain would only be slightly greater than B’s I am not so inclined to this evaluative judgment about these alternatives. Nor am I so inclined to think I should choose to spare Z rather than B. I am not trying to justify my taking this factual comparison between the magnitude of Z’s pain and the magnitude of B’s as relevant in reaching an evaluative decision on these alternatives. I am simply acknowledging that for me it would be relevant. I believe that many others will find that it is relevant for them as well. Now let me alter the conditions in this thought experiment. As an alternative to Z’s suffering an agonizing pain I now contemplate not only B’s suffering some relatively minor pain, but C’s suffering pain as well. Again, before reaching a comparative evaluation of these outcomes, before deciding which I should prevent supposing the choice is mine to make, I will want to know how much C will suffer in comparison to Z. Will his pain be as great, nearly as great, perhaps even greater than Z’s? I accept the relevance of this new factual comparison between the magnitude of Z’s pain and the magnitude of C’s. Now if I imagine that Z’s pain will be much greater than C’s, that C too will suffer but twenty-five seconds of relatively minor pain, my evaluational attitudes towards these outcomes do not change. I am still inclined to think that it would be a worse thing, a greater evil, if Z were to suffer than it would be if both B and C were to suffer. And if the choice were mine to make I would spare Z rather than B and C, other considerations apart. As I continue with my thought experiment, I make a jump. I recognize that it won’t make any difference to me how many individuals I add to the side of B and C. If no one of them suffers anything more than twenty-five seconds of relatively minor pain, and the choice were mine to make, I would spare Z her agonizing pain, not the many, however many they might be. I am equally inclined to judge that other considerations apart it would be, morally speaking, preferable were Z spared twenty-five seconds of agony. But at this point I can hear the Parfitian caution: ‘Do not rush to judgment here. Indeed it is true that Z would suffer greater pain, more pain, than either B or C or any one of these many. Before you reach a decision, however, you should ask yourself whether Z’s pain would be greater than the summed, added, or combined pains of these many. After all two people in pain, when taken together, between them suffer more pain than either suffers separately. And so, though Z

Reply to Parfit ’ s ‘ Innumerate Ethics ’   313 suffers more than either B or C or any one of these others, she may not suffer as much as they do when considered together.’ This I do not understand. It is here that I suspect the presence of a meta­phys­ ic­al fiction. To me pain and suffering are magnitudes that cannot be added or summed across individuals. They are like physical beauty, boxing skill, or artistic talent. I can compare the boxing skill of Marvin Hagler to the skills of each of these 250 individuals who work as sparring partners in the Middleweight Division. He has greater boxing skill or more boxing skill than any one of them. If someone were to say that when taken or considered together these 250 have greater or more boxing skill than he does I would not know what is being said. What am I to focus on when I am asked to consider, in comparison to Hagler’s skill, the sum of boxing skill obtained by adding together the boxing skills of these 250 individuals? If I call your attention to the great beauty in this woman’s face, and having acknowledged it, you reply ‘But her beauty pales when compared to the awesome beauty we contemplate when we add together or sum the beauty found in each of a sea of ordinary faces,’ I will not know what to make of this. And how many individuals with relatively modest artistic talent does it take before the sum of their individual talents surpasses the artistic talent of Leonardo da Vinci? To me pain and suffering are magnitudes of this kind. I can compare the pain or suffering of this one person with the pain or suffering of any one of these many. But when I am asked to compare the magnitude of her pain or suffering with the magnitude of the pain or suffering obtained by adding or summing the pains of these many I am at a loss to know what it is I am to focus my attention on. Sometimes it seems to me that Parfit regards pain and suffering as magnitudes more like weight than boxing skill or beauty. We can compare Hagler’s weight with the weight of any one of these ten lightweights. He has more weight, greater weight than any one of them. But taken together, their combined or summed weight is much greater than his. Their combined or summed weight is no meta­ phys­ic­al fiction. It is the same magnitude that I have before my mind when I compare Hagler’s weight with the weight of just one of these ten. It is simply much greater. The move from comparing the magnitude of this one person’s pain with the pain of any one of these ten to comparing her pain with the added or summed pain of each of these ten is not the same. Their combined or added pain is a meta­ phys­ic­al fiction in a way that their combined or added weight is not. Let me put it this way. When we move from saying ‘He has greater weight than any of these ten’, to saying ‘But taken together they have greater weight, or more weight, than he has’, the expressions ‘more weight’ and ‘greater weight’ do not undergo a change in sense or meaning. They have the same meaning in both comparisons. However, when we move from saying ‘She will suffer greater pain, or more pain, than any of these many’ to saying ‘But when taken together the many will suffer more pain, or greater pain, than she will’, the expressions ‘more

314  John Taurek pain’ and ‘greater pain’ do undergo a change in sense or meaning. I understand their meaning in the first comparison. I do not understand their meaning in the second. Parfit does not explicitly acknowledge this shift in the meaning of these expressions. And I am puzzled as to what he might mean by them when he says, in this case, that the many will suffer greater pain, more pain, than Z will. At one point Parfit suggests that in denying this claim—a claim that is in his mind a straightforward factual claim—I may really have intended to assert the valuational claim that the pains of these many cannot be ‘morally summed, that they cannot together make the outcome worse’. I do believe this. But this is not what I intended to affirm when I denied that the many ‘taken together’ suffer greater pain, or more pain, than Z. I deny this because I can’t see that it expresses any relevant fact in the situation. When it is said that taken together these many will suffer more pain, or greater pain, than she will, I can no more see what fact is being alluded to than I can when it is said that these 250 journeymen, none of whom holds a candle to Marvin Hagler, possess, when taken together, more boxing skill, or greater boxing skill, than Hagler does. What are the facts concerning boxing skill in this situation? Here we have these 250 journeymen sparring partners on this side of the room and over here stands Marvin Hagler. There is first of all the fact that Hagler has far greater boxing skill than any one of these 250. Second, there is the fact that on this side of the room we have a far greater number of individuals who possess a significant degree of boxing skill than we have on the other, where stands only one. To me these are the facts, all the facts. Were someone to maintain that there is yet a third fact, namely, that the magnitude of the boxing skill on this side of the room obtained by adding together the boxing skill of each of these 250, is greater than Hagler’s boxing skill, I would be inclined to say that he has introduced a pure metaphysical fiction into this situation. Similarly, what are the facts concerning pain in the situation we have before us? First, there is the fact that this woman’s pain is far greater than is the pain suffered by any one of these 10,000. Second, there is the fact that on the one side of the ‘equation’ we have a far greater number of individuals who suffer a degree of pain than we have on the other, where stands only this one. To me these are the facts, all the facts. When I am told that there is yet a third fact, namely, that the magnitude of her pain is not nearly as great as the magnitude of the pain obtained by adding together the relatively minor pains of these many, my response again is to say that a metaphysical fiction has been introduced into our thinking about this situation. At one point Parfit suggests that what he and others mean to say when they say that the minor pains of these many sum to, or add to, more pain, or greater pain, than this one person’s agony is that these many minor pains together would be

Reply to Parfit ’ s ‘ Innumerate Ethics ’   315 worse than her pain. But what does this mean? In explanation of its meaning he suggests that I imagine having to choose between suffering her twenty-five se­conds of agonizing pain on the one hand, or as the alternative, suffering, let us say, 10,000 twenty-five-second periods of relatively minor pain occurring with some suitable intervals between them. He points out that I might prefer the former to the latter. And well I might, especially if the intervals between them are made rather brief. But I can’t see how or why this should be thought relevant to my evaluation or choice between the alternatives in question. I ask myself whom I would spare, if God gave me the power, Z her one ag­on­iz­ ing pain or these 10,000 each his or her relatively minor pain. Parfit suggests that I ask myself whether I would prefer to undergo twenty-five seconds of agonizing pain or 10,000 twenty-five-second periods of relatively minor pain spaced, say, at two-minute intervals. Of course I would prefer to undergo the former pain and to be done with it. But I am not weighing Z’s suffering twenty-five seconds of ag­on­ iz­ing pain against some other person’s spending the next year of his life suffering twenty-five-second episodes of relatively minor pain at two-minute intervals. So I am afraid that I have failed to glean from Parfit’s paper a clear understanding of the alleged fact that is being alluded to when in this case he says that these many will suffer more pain, or greater pain, than Z will. On one point concerning this case Parfit and I are in agreement. This surprised me when I came to it in his paper. He too thinks that in this case we should spare the one person her agonizing pain rather than any number of others each his or her relatively minor pain. He recognizes that this raises a problem for him. For it strongly suggests if it does not imply that the pains of these many cannot be meaningfully summed in a way that has moral significance for preference and choice. You wouldn’t expect someone who sees the magnitude of pain involved increasing as we add more people who are in pain to take the view that we ought to spare this one person in this case rather than any number of the many. Nor would you expect him to take the view that the outcome in which Z suffers her twenty-five seconds of agonizing pain would be worse from a moral point of view than its alternative where these many, as many as you like, each suffers his or her relatively minor pain. (Parfit rejects this comparative evaluation of these two alternatives. I think this creates an even more formidable problem for him. He must now explain how it is that in this case he thinks that one alternative involves far less pain, far less suffering, than does the other; that it is, from a moral point of view the worse alternative; and yet that we should if given the choice choose it.) Let me try to state the problem Parfit faces by contrasting the current case with another. Again we contemplate alternatives. Either this one person will suffer twenty-five seconds of agonizing pain or these ten others will each suffer twentyfive seconds of agonizing pain. Here Parfit will say that, given the choice, virtually anyone with the exception perhaps of this one person or those who find themselves particularly pained by her pain, ought to choose to spare the ten rather

316  John Taurek than the one. Why? Because the magnitude of the pain involved when we sum the agonizing pains of each of these ten far exceeds the magnitude of this one person’s pain. But if the pains of different people can be meaningfully summed in this way, then the lesser pains of these many must sum to more pain, or greater human suffering, than the agonizing pain of this one, provided only that there are enough of them. We should then be able to make the magnitude of the summed pain of the many exceed the magnitude of the suffering of this one by at least as much as the magnitude of the summed agonizing pains of these ten was thought to exceed the agonized suffering of the one. Of course precision is not to be had in such matters. So let us just say that 10,000 people each suffering some relatively minor pain will do it. The problem for Parfit is to explain how in this case he avoids the conclusion that we ought to spare the many rather than the one. He considers several ways in which he might reconcile his judgment in this case that we ought to spare the one rather than the many with his view that the separate pains of different people are summable in a meaningful way that has moral significance for choice and preference. He considers several ways and rejects them all as implausible, which leaves me thinking still that they are not reconcilable. He first suggests that he might explain, that is justify, his judgment in this case by claiming that Z would be the ‘worst-off ’ individual in a certain sense, and then appeal to a well-known principle of justice he calls ‘Maximin’. I do not understand the principle of Maximin. But the sense in which Z must be the ‘worst-off ’ individual if we are to appeal to it implies a more general assessment of her condition relative to theirs. It is not sufficient that her pain will be worse than any one of theirs. ‘In deciding who will be worst off,’ he writes, ‘we must think in terms of lives—we must ask whose life would have gone worst.’ Understanding the appeal in this sense, he rejects it, or perhaps we should say he simply makes it unavailable by altering or adding to the imagined situation. Let us suppose, he suggests, that yesterday each of these many suffered twentyfive seconds of agonizing pain, while this one person suffered only twenty-five seconds of minor pain. In this case she would not be the worst-off individual in the sense required if an appeal to ‘Maximin’ is to be made. Still in this case he thinks we should spare her the agony rather than spare the many the minor pain. So the question remains. Why should we spare the one rather than the many? We might, he says, claim that (Q) We would be preventing the greater suffering. But is this true? he asks. Of course I would say this and defend it by saying that their individual minor pains cannot be meaningfully summed. However, Parfit writes, ‘We have rejected this view. If more people are in pain, there is more pain.’ He continues, ‘There is another way of defending (Q)’ – that is, the claim that in sparing the one rather than the many in this case, we would be preventing the greater suffering. ‘We might claim (R), “Agony is infinitely worse than minor pain.” ’ I don’t understand this claim, nor do I see how it could be a way of

Reply to Parfit ’ s ‘ Innumerate Ethics ’   317 defending (Q). In any case Parfit concludes that the agony of this one person would not be infinitely worse than the pains of the many. ‘So,’ he writes, ‘we must abandon (Q). We cannot claim that, in sparing this one person we would be preventing the greater suffering.’ So again the question remains. Why should we spare the one rather than the many? He writes: ‘We might claim (S) We would be preventing the worse of two outcomes.’ Again he asks whether this is true. He argues that it is not. He suggests that someone might think (S) true in this case because he is imagining the minor pain suffered by each of the many to be so minor as to be without moral significance. He remarks that the assumption here that there is some threshold below which pain has utterly no moral significance could be challenged. But it’s not ne­ces­sary that we do so. We need only make sure that in our example the lesser pain suffered by each of the many is imagined to be above this threshold. Let it verge in your mind toward the ‘moderate’ band of the spectrum of pain at least far enough to give it some moral significance. Perhaps someone’s suffering such a pain would not be a really great evil, but it would be a minor evil. So the question remains. How can (S) be defended? ‘We might claim,’ he writes, ‘(T) No number of these lesser evils could together be as great an evil.’ But he discounts this as pretty plainly implausible. So, on his view, he and we must abandon (S). We cannot claim that in sparing this one person her agonizing pain we would be preventing the worse of two outcomes. So again the question still remains. How does he explain or justify his judgment in this case that we should spare the one rather than the many? How is this to be explained, especially in light of the fact that he thinks it would be a worse thing, or a greater evil, were so many people each to suffer a relatively minor to moderate pain than it would be were this one person to suffer her agonizing pain? Parfit makes no further attempt to explain it. He writes: ‘Why should we help Z rather than any number of the many? Taurek might say: “You must now accept my explanation. Pains in different lives cannot be morally summed.” ’ And he continues, ‘If we are consequentialists, we may have to agree. We must then accept (S). We must think that, in helping Z, we would be preventing the worse of two outcomes. How could this be true? If they came within one life, 1,000 of the lesser pains would be a worse outcome. How could there be no such number when they come in different lives? We may have to accept Taurek’s view. Perhaps pains in different lives cannot be morally summed.’ Yet he asserts, ‘There is an alternative. We need not be consequentialists. We might say: “We ought to prevent one much greater harm rather than any number of much smaller harms. But this is not because we should be preventing the worse of two outcomes. The urgency of moral claims [that is, what we ought to do] does not always correspond to the badness of outcomes.” On this alternative we avoid Taurek’s view.’ And he concludes with the rhetorical question: ‘Which alternative is the more plausible?’

318  John Taurek My problem is that I don’t understand what the alternative is to my explanation of our judgment that we should rather spare the one person her agonizing pain than each of the many his or her relatively minor pain. I am very confused about this passage. As I see it, Parfit faces a rather formidable problem. He must explain why he thinks we should in this case spare this one person rather than the many, however many they might be. The task is formidable because on Parfit’s view the many, because they are so many, will suffer more pain, or greater pain, than she will. Of course in virtue of this fact Parfit feels compelled to admit that the outcome in which the many would suffer while the one would be spared is, from a moral point of view, the worse alternative. After all it involves far greater, or far more, human suffering. Yet he thinks we should choose the alternative which is admittedly worse. He seems to think that the question: ‘Why should we?’ can be dismissed by saying that in general what we ought to do is not always determined by the badness of outcomes. We needn’t be consequentialists, he remarks. But consequentialism has little to do with this problem. If we think of a consequentialist as someone who thinks that what one ought to do is always determined by a comparison of the various outcomes possible, and we imagine him confronting these alternatives with a view to choosing between them, then we know that for him the crucial question will be ‘Which would be the better outcome?’ or, alternatively, ‘Which would be the worse?’ If he is an unusual consequentialist and thinks that pain and suffering cannot be meaningfully summed across individuals, then he will no doubt agree with us that the alternative in which this one person will suffer an agonizing pain is or would be worse, and so agree with us that we should spare the one rather than the many. If he is a more typical consequentialist who thinks that the pains of different people can be meaningfully summed, then he will not agree with us that we ought to spare the one rather than these many. For he will think that the outcome in which the many suffer would be worse. A problem exists only for those who, like Parfit, think both that we ought to spare the one rather than the many, and that the separate pains of the many sum to greater pain, or more pain, than would be suffered by the one. I don’t think Parfit has responded to this problem. If Parfit retains his view that in the case in question we should spare the one rather than the many, however many there are, then he should abandon his view that the outcome in which the one is spared and the many suffer is, morally speaking, the worse alternative. But if he abandons this view, then he must explain how this is so despite the fact that, on his view, it involves far greater, or far more, human suffering. I believe that he should abandon his claim that the suffering, or the pain, of the many would be greater than the pain of this one. But perhaps he would rather abandon his claim that in this case we should spare the one person her agonizing pain rather than each of the many his or her relatively minor pain.

Reply to Parfit ’ s ‘ Innumerate Ethics ’   319 There is another point about which Parfit and I seem to be in disagreement when we contemplate the alternatives in these thought experiments. This has to do with the connection between the thought that one thing, or one alternative, is, from a moral point of view, in itself preferable to another, and the thought that one, or anyone, should prefer it. When I bring to mind situations in which I judge that one alternative would be, morally speaking, preferable to another, I see that in every such situation this thought is inextricably bound up with the thought that it is to be preferred. I should prefer it. You should prefer it. Anyone should prefer it. For example, I really do think it would be, morally speaking, a better thing if this one person were to suffer some substantial pain if these many others could each thereby be spared an agonizing pain. The one alternative is to be preferred to the other. With these words I already express a kind of preference between these alternatives, a reflective preference if you will. It becomes for me a standard against which to measure my actual preference. It doesn’t matter what role I imagine myself to occupy in this situation. When I imagine myself in the position of the one, and I imagine myself hoping, wishing, and preferring that I might be spared, I put this down to fear battling against my reflective judgment. Perhaps I cannot abolish the fear or banish the desire, but if the power were given to me to choose I would hope that I would follow my conscience and not the promptings of fear. When I imagine myself failing to do this I see myself as having failed to live up to my values, my ideals, and my convictions about what would be better and what would be worse from a moral point of view. Parfit doesn’t see it this way. He thinks that I am mistaken in chiding myself as I do in this situation. He tells me that I have an agent-relative permission to prefer the greater evil, or the worse outcome, since I, personally, will be substantially better off should the greater evil prevail. Such a preference bespeaks no moral shortcoming in me in his view. And if the power were given to me to choose in such a situation there would be nothing amiss in my choosing the greater evil. I have no difficulty with the general idea that there are agent-relative permissions, any more than I have with the idea that there are agent-relative obligations. Another may be permitted to do things I would not be permitted to do, just as he may be obligated to do things I am not obligated to do. It is with the idea that I have an agent-relative permission not to prefer what I recognize would be, from a moral point of view, preferable, that I am uncomfortable. In ‘Should the Numbers Count?’ I traded on what for me is an inextricable connection between the thought of one thing’s being, from a moral point of view, preferable to its alternative, and the thought that one should prefer it. Anyone should, for it is, morally speaking, preferable. I felt that in the minds of most these two thoughts would be similarly connected. And I sought to make use of this.

320  John Taurek In these thought experiments, where we put to ourselves, as detached third parties the alternatives—sparing this one person or these few some pain, suffering, or loss, or sparing some yet larger number of people a similar pain, suffering, or loss—many respond with the thought that they should spare the greater number. Often this claim is made to rest on an impersonal comparative evaluation of the alternatives. The reason given for the preference shown is that it would be, from a moral point of view, a preferable outcome, or the better alternative. But if these two thoughts are as inextricably bound up in the minds of others as they are in mine, I thought that others too would see that when one alternative, seen from this impersonal perspective, is judged morally preferable to another, then that is what we should prefer; that’s what anyone contemplating these same alternatives should prefer. So if you now drop the detached perspective and cast yourself in one of the roles you would still have to acknowledge that the one alternative is, from a moral point of view, preferable to the other. So you should prefer it. It doesn’t matter in which role you place yourself. There is no permission not to prefer it. It is on your own reflective judgment the preferable alternative. If you cast yourself in the role of one who might be worse off in some way, you will see yourself as unfortunate. But you know what you should prefer, because you know which is preferable from that detached, impersonal, moral point of view. Now in thought experiments in which the alternatives are such that the one person stands to suffer or lose at least as much as, if not more than, would be suffered or lost by any one of these other five, let us say, this thinking causes acute discomfort in me. I cannot imagine what I might say to David to convince him that it would be in itself a better thing if he died and these other five survived than it would be if he survived and they died. Why should he judge the former alternative preferable to the latter? In this case, David’s imagined response ‘Better? Better for whom?’ carries great force with me. I had hoped that it would with others. That he prefers the alternative in which he survives manifests to me no moral deficiency in him. I cannot expect that he should prefer the alternative in which they survive and he dies. Hence I cannot give as my reason for sparing the five instead of him the impersonal evaluational comparison that the one outcome is, from a moral point of view, preferable to the other. This doesn’t work with Parfit because he believes that we all have agent-relative permissions not to prefer what is morally preferable in the event that we would be substantially better off should the greater evil prevail. There seems to be some important difference in the way we think here. Let me return to the case in which we contemplate as alternatives this one person suffering twenty-five seconds of agonizing pain or these many each suffering some relatively minor to moderate pain for twenty-five seconds. If my decision as to whom to spare is to be taken in light of some matter of fact, it must be taken in light of the fact that not one of these many would suffer nearly as much as she, or in light of the fact that they are many and she is only one. There is no third matter

Reply to Parfit ’ s ‘ Innumerate Ethics ’   321 of fact that is relevant here. I would spare her rather than the many because her pain would be much greater than theirs. Now I can imagine someone choosing to spare each of the many his or her minor to moderate pain, and citing as the fact that moves him the bare fact that there are more of them. I want to emphasize that for this person it is the bare fact that there are so many of them while she is only one that moves him, just as for me it is the bare fact that she will suffer more pain than any of them that moves me. For him the transition from the thought that they are many and she but one to the evaluational decision is direct, unmediated by the reasoning: ‘Since they are so many, their pain, or their suffering, will be greater than hers.’ He is under no illusion that the pains of different people can be summed in any meaningful way that has relevance for choice and preference. He has no doubt that she will suffer more pain, or greater pain, than they will. It is the bare fact that there are so many of them that moves him to think that it would be worse for the many to suffer. His judgment that it would be worse if the many were to suffer is not based on the alleged fact that their suffering would ‘greater than hers’. In this case this is simply not a fact. Their suffering would not be greater than hers. It would not be as great. It’s not that the many would suffer more than she would that moves him, it is simply their great number. Suppose he is quite clear about this. He understands that in this situation there are only two relevant facts concerning pain. She will suffer greater pain, or more pain, than they would. And they are very many and she but one. He doesn’t think that the suffering of the many increases as their numbers increase. However many they are, it will never be the case that they would suffer as much as she will. It’s just that for him how many would suffer is as relevant as who would suffer more, or whose suffering would be greater. In this case these two independent features pull in his mind in opposite directions. ‘Look,’ he says, ‘it’s a fact that her suffering would be greater than the suffering of the many. When “the many” were only two it was a fact that her suffering would be greater than theirs. When the many number 10,000 this remains as much a fact. Her suffering, or her pain, would be greater than theirs. “The many” suffer no more pain, no greater pain now that they number 10,000 than “the many” suffered when they numbered only two. The difference in the magnitude of her suffering and that of the suffering of the many remains constant throughout my thought experiments. But for me how many will suffer is important as well as how much they will suffer. In the case in which we compare her suffering twenty-five seconds of agonizing pain with the alternative in which two others each suffer some relatively minor to moderate pain for twenty-five seconds, it’s not as if the fact that they are two and she is only one counts for nothing at all with me. It just does not yet outweigh, in my mind, the fact that in sparing her we would be preventing the greater human suffering. But at some point, when I make the numbers large enough, the fact that they are many and she is only one outweigh, in my mind, the fact that in sparing her we

322  John Taurek would be preventing the greater suffering. For me the numbers just do count. If you ask me why? well, I cannot say why. It just does. ‘But really are you any better off? For you the fact that they would not suffer nearly as much as she would is decisive. If I ask why this fact weighs with you at all, will you be able to give me a satisfactory answer? This fact weighs with me too. I can’t say why. It just does. But for me the particular number of people involved weighs as well. I can’t say why. It just does. And when I make it large enough it outweighs, in my mind, the fact that they would not suffer as much as she would. When I say that the one fact “outweighs” in my mind the other I only mean to report that, in view of it, I come down on the side of the many. I declare in favor of the many as against the one.’ When I think of a person who would take this, if I may so describe it, ‘logically pure line’ on how the numbers count, I cannot see that I have anything very compelling to say to him that might change his mind. He seems to be free of any metaphysical illusion concerning the ‘summed’ or ‘added’ pains of different individuals. The difference between us is simply this. He attributes significance to something that for me has no significance. For me the only question is whose suffering would be greater, who will suffer more, the few or the many? For him how many will suffer is also of significance. For me it is only the magnitude of the suffering of the few or the one in comparison to the magnitude of the suffering of the many that counts. For him the disparity in numbers also counts. But not because he thinks that this alters the magnitude of the suffering of the many relative to the magnitude of the suffering of the one or the few. The numbers simply count for him.

15

Defence Against Parfit’s Torturers Jeff McMahan

1.  Torturers, Harmful and Harmless One of Derek Parfit’s concerns in the section of Reasons and Persons on ‘mistakes in moral mathematics’ is whether there can be imperceptible harms and benefits.1 My concern here is quite different: it is with what it can be permissible to do in defence against people who will otherwise inflict a perceptible, though quite small, harm on each of a large number of victims. Although our topics are quite different, there are nevertheless two ways in which Parfit’s discussion provides the basis for mine. First, slightly modified versions of his core examples are ideal for illustrating the issues I will discuss. Second, each of the rival explanations he distinguishes of why it can be wrong to cause imperceptible increases in a person’s pain might also partly explain why it might be permissible to engage in harmful defensive action against a person who threatens to cause a barely perceptible harm to each of a large number of victims. The summary of Parfit’s discussion in this first section will therefore serve to introduce my own discussion. Parfit invites his readers to consider a group of people I will call The Traditional Torturers Each of a thousand people has a button that, if pressed, will cause an innocent person to experience a thousand distinct but simultaneous electric shocks for eight hours. Each shock on its own would be imperceptible but a thousand such shocks together constitute torture. Each of these thousand Traditional Torturers presses his button, thereby causing a thousand victims each to suffer agony for eight hours.2

I have been presenting and gradually refining the material in this essay in lectures given in various places since 2013. I have been benefited from comments from audiences at the University of Leeds, the Rutgers-­Camden School of Law, the University of Manchester, the University of Stockholm, the University of Oslo, Boston College, and the University of Southern California. I am also greatly indebted for written comments on earlier drafts to Ben Bronner, Derek Parfit, Victor Tadros, and Patrick Tomlin, and for illuminating discussion to Fiona Clarke, Johann Frick, Helen Frowe, and, especially, Tim Campbell, Shelly Kagan, Frances Kamm, and Larry Temkin. 1  Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1987 repr.), ch. 3, sects. 28 and 29. 2 Parfit, Reasons and Persons, 80. My statements of Parfit’s examples are modified in several trivial ways. Jeff McMahan, Defence Against Parfit’s Torturers In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0016

324  Jeff M c Mahan It is obvious that what each Traditional Torturer does is wrong. And the obvious explanation is that, although each shock on its own is imperceptible, ‘the total effect of what each torturer does’ is to inflict great suffering on a single victim.3 Parfit next asks us to consider The Harmless Torturers Each of a thousand people has a button that, if pressed, will administer an imperceptible electric shock to each of a thousand innocent people for eight hours. Each Harmless Torturer’s thousand victims are the same individuals as every other Harmless Torturer’s victims. (We can suppose, indeed, that the thousand button pressers and the thousand recipients of shocks in this case are the same people as those in the corresponding roles in the case of the Traditional Torturers.) All the Harmless Torturers press their buttons at the same time. The result is that the thousand victims each suffer agony for eight hours.

There are, I will assume, no morally relevant differences among the Harmless Torturers. All are equally culpable, none would be more seriously harmed by being killed than any other, and so on. Parfit claims, correctly in my view, that each Harmless Torturer is ‘acting just as wrongly as’ each Traditional Torturer. Yet none of the Harmless Torturers on his own causes a perceptible effect on any one of the victims. Parfit suggests two explanations of how what the Harmless Torturers do is nevertheless wrong. The first presupposes that one’s pain or suffering can be made imperceptibly worse and that one can be harmed by an imperceptible increase in one’s pain or suffering. In that case, Parfit says, ‘since each torturer adds to the suffering of a thousand victims, each torturer imposes a great total sum of suffering’.4 This explanation is similar to that which applies to the action of the Traditional Torturers. The difference is that the total sum of suffering inflicted by each Traditional Torturer is concentrated in one period in the experience of one victim, whereas the total inflicted by each Harmless Torturer is dispersed equally among a thousand victims. Parfit concedes, however, that many people believe that any increase or worsen­ ing of pain or suffering must be perceptible. If this belief is correct, no Harmless Torturer worsens the condition of, or harms, anyone. Thus, Parfit writes, ‘if we cannot appeal to the effects of what each torturer does, we must appeal to what the torturers together do. Even if none of them causes any pain, they together impose great suffering on a thousand victims.’5 This is the second explanation of why what the Harmless Torturers do is wrong. 3 Parfit, Reasons and Persons, 80; italics added. 4 Parfit, Reasons and Persons, 80; italics added.

5 Parfit, Reasons and Persons, 80.

Defence Against Parfit ’ s Torturers  325 Parfit next presents a third example: The Single Torturer One person has a button that, if pressed, will administer an imperceptible shock to each of a thousand innocent people for eight hours. This person knows that each of these thousand people is about to experience eight hours of pain from natural causes that will be equivalent in intensity to the aggregate pain from 999 of the individually imperceptible shocks. If the Single Torturer presses his button, the result will thus be the same as it would be if he were adding his shocks to those of 999 other Harmless Torturers—that is, all thousand victims will ex­peri­ence the same pain that the victims suffer in the case of the Harmless Torturers. The Single Torturer presses his button.6

In the first printing of Reasons and Persons, Parfit indicates that he believes that the Single Torturer acts wrongly and that the first explanation of why the Harmless Torturers act wrongly provides a plausible explanation of why the  Single Torturer acts wrongly as well. He acts wrongly because each of the thousand imperceptible increases in pain he causes harms, or worsens the ­condition of, the victim of that increase. He is, therefore, ‘imposing . . . a great total sum of suffering’ on his thousand victims. In this same first printing, however, Parfit concedes that many people believe that the Single Torturer does not act wrongly. These people, he writes, believe that it cannot be wrong for someone to affect others in a certain way, if this person knows both (1) that these effects will be imperceptible, and (2) that they will not be part of a set of effects that, together, are perceptible. Since this belief is widely held, and not implausible, it is better not to appeal to the effects of what each torturer does. Even if we believe that there can be imperceptible harms and benefits, as I do, it is better to appeal to what groups together do. This appeal is less controversial.7

In short, if the fact that the Single Torturer’s act causes a great total sum of suffering is insufficient to make his act wrong, the fact that each Harmless Torturer’s act causes a great total sum of suffering cannot be the explanation of why his act is wrong. The relevant difference between the acts of the Harmless Torturers and 6  Parfit’s example is in Parfit, Reasons and Persons, 81. My statement of the example differs from his in two ways. First, the thousand victims in his example have already been suffering pain for some unspecified period when the Single Torturer presses his button. Second, the suffering they are already experiencing is equivalent to the sum of only five hundred of the individually imperceptible pains. Their suffering is thus only half as bad as that of the victims of the Harmless Torturers. 7  Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), 82.

326  Jeff M c Mahan the act of the Single Torturer is that the effects of each Harmless Torturer’s act are ‘part of a set of effects that, together, are perceptible’, whereas that is not true of the effects of the Single Torturer’s act. What Parfit must mean here by ‘effects’ is ‘effects that are the products of agency’. The imperceptible effects caused by a single Harmless Torturer are wrong because they are additions to other effects of the same sort caused by other Harmless Torturers with whom he is acting together, as a group, or with whom he is collaborating. Although the imperceptible effects caused by the Single Torturer are additions to other effects of the same sort, these other effects are not caused by people with whom the Single Torturer is col­lab­or­ at­ing. Understood in this way, Parfit’s second explanation of why what the Harmless Torturers do is wrong applies to their acts but not to the act of the Single Torturer. Thus, if we believe that the Single Torturer does not act wrongly, we can accept the second explanation of why the Harmless Torturers act wrongly, even though we cannot accept the first. In a later reprinting of Reasons and Persons, however, Parfit slightly modified his view. There he writes that I believe that the Single Torturer is acting wrongly. How can it make a moral difference whether he produces bad effects jointly with other agents, or with Nature? I therefore prefer, in both cases, to appeal to the effects of single acts. Some people disagree. Even if we believe that there can be imperceptible harms and benefits, it may thus be better to appeal to what groups together do. This is less controversial.8

In this passage, Parfit expresses scepticism about the view that adding bad effects to those produced by others with whom one is collaborating is wrong in a way that adding equivalent bad effects to equally bad effects produced by natural causes is not. But if there really is no moral difference between these two ways of adding to bad effects from other sources, then both the first and second of Parfit’s explanations of the wrongness of each Harmless Torturer’s act also imply that the Single Torturer’s act is wrong as well, and presumably equally wrong. But Parfit nevertheless suggests that, because an explanation that appeals to the distinct wrongness of collaborative harming implies that the acts of the Harmless Torturers are wrong without implying that the act of the Single Torturer is wrong, it is perhaps better, because less controversial, to adopt that explanation rather than the one that appeals to the effects of single acts. I think, however, that Parfit is right to doubt that there is a distinctive wrongness to collaborative harming that is absent from acts that add or contribute to harms from sources other than the acts of collaborators. I will try to reinforce his scepticism by appealing to some further examples in Section 3. If I am right, we 8  Reasons and Persons (1987 repr.), 82.

Defence Against Parfit ’ s Torturers  327 should reject rather than embrace his second explanation of the wrongness of the acts of the Harmless Torturers. This conclusion would not have unduly troubled Parfit, for, as he indicates, his only reason for endorsing this explanation is that it is likely to be acceptable to more people than his first explanation, which is the explanation that he favours on its merits. I think, however, that the first ex­plan­ ation is also inadequate, for reasons I will present in Section 2. The best ex­plan­ ation is a variant of Parfit’s second explanation, purged of the collectivist component of which Parfit was rightly sceptical.

2.  Why, and to What Extent, a Harmless Torturer’s Act Is Wrong My concern in this essay is, as I noted, not with whether there can be im­per­cept­ ible harms and benefits or whether it can be wrong to cause effects that are im­per­ cept­ible. In the remainder of this essay, therefore, I will assume that the individual shocks inflicted by the Traditional Torturers, the Harmless Torturers, and the Single Torturer all cause perceptible, though only barely perceptible, pains. Even though it lasts for eight hours, each of these pains is only the mildest, almost negligible, annoyance. But, as with the imperceptible effects in Parfit’s original ex­ amples, a thousand simultaneous annoyances amount to agony. (On the assumption that the pains they cause are perceptible, the Harmless Torturers are not, strictly speaking, harmless. But I will continue to use this label nonetheless.) When Parfit’s examples are understood in this way, his first suggested ex­plan­ ation of why what the Harmless Torturers do is wrong clearly applies. Because each inflicts a thousand perceptible pains, each inflicts a great total sum of pain, or suffering. The question remains, however, whether this best explains why what they do is wrong, or whether it explains why what they do is as seriously wrong as it is—that is, as wrong as what the Traditional Torturers do. We can, I believe, see that this explanation does not account for how seriously wrong what the Harmless Torturers do is by considering a further group of p ­ eople who also inflict tiny harms on a large number of people. The Inflictors Each of a thousand people has a button that, if pressed, will administer a barely perceptible electric shock to each of a thousand innocent people for eight hours. There is, however, no overlap among each of these people’s victims. All of these Inflictors of tiny harms press their buttons at the same time. The result is that each of a million innocent people experiences a barely perceptible pain for eight hours.9

9  This example first appears in my review of Helen Frowe’s book, Defensive Killing, in Ethics 126 (2016): 825–31, at 826. The discussion there anticipates some of what I say in this section.

328  Jeff M c Mahan What the thousand Inflictors do is wrong. But it is not as seriously wrong as what the Traditional Torturers do, and thus not as seriously wrong as what the Harmless Torturers do. Yet in certain respects what each Inflictor does is the same as what each Harmless Torturer does: each inflicts a barely perceptible pain for eight hours on each of a thousand innocent people. And each Inflictor causes the same total sum of suffering that each Harmless Torturer causes. The Inflictors might, moreover, act in collaboration, as a group, so that the pains caused by each are ‘part of a set of effects’—a million tiny pains—that they together produce. And each Inflictor might be just as culpable as each Harmless Torturer (each might, for example, mistakenly believe that she is a Harmless Torturer).10 But it would still not be true that each Inflictor’s act is as wrong as that of a Harmless Torturer. Indeed, even if we were to stipulate that each Inflictor has five thousand victims, and so causes a total sum of suffering that is five times as great as that caused by each Harmless Torturer, it would still be difficult to believe that her act is as wrong as that of a Harmless Torturer. So Parfit’s first explanation that appeals to the total sum of harm that an act causes cannot account for the intuitive difference between what the Inflictors do and what the Harmless Torturers do. What, then, is the correct explanation? The degree to which what these different people do is wrong is, I believe, a function not only of the magnitude of the total harm they each cause, and of the magnitude of the harms they each cause to particular individuals, but also of the magnitude of the harms to particular individuals to which they each contribute. Thus, whereas each Inflictor does not contribute to a greater harm than that which she alone inflicts, each Harmless Torturer contributes to torture—or, more precisely, to the experience of eight hours of agony by a thousand individuals. It is, I think, the fact that each Harmless Torturer’s act makes a contribution to very great harms to individuals, and to very many individuals, that makes the acts of the Harmless Torturers more seriously wrong than those of the Inflictors. And this same fact makes the Single Torturer’s act more seriously wrong as well (on the assumption, one will recall, that the bad effects caused by the Inflictors and the Single Torturer are perceptible pains). There can be groups of people who inflict many tiny harms that are intermediate between the Harmless Torturers and the Inflictors. One example is The Contributors Each of a thousand people has a button that, if pressed, will deliver two simultaneous, barely perceptible shocks to each of five hundred innocent 10  Parfit, writing in 1984, referred to all his torturers using male-­gendered pronouns. For consistency, I have done the same. Because it is important in this essay to distinguish clearly between what each individual of a certain type does and what all the individuals of a type together do, I do not use plural pronouns to refer to individuals. For variety, I use female-­gendered pronouns to refer to Inflictors, to those I will call ‘Contributors’, and to third-­party defenders.

Defence Against Parfit ’ s Torturers  329 people for eight hours. There are two thousand potential victims. All the Contributors press their buttons. Two hundred and fifty of them inflict shocks  on five hundred of the victims, another two hundred and fifty inflict shocks on a different five hundred, and so on. The result is that each of the two ­thousand  victims suffers five hundred simultaneous tiny pains for eight hours.

Each Contributor inflicts the same total sum of suffering that the Single Torturer and each Traditional Torturer, Harmless Torturer, and Inflictor causes. But the pain that each Contributor inflicts on each of her five hundred victims is twice as intense, and therefore presumably twice as bad, as that which any of the others inflicts on any individual victim. This makes what each Contributor does in one important respect more seriously wrong. And each Contributor’s act is in another respect more seriously wrong than that of each Inflictor: for each Contributor adds her two simultaneous pains to 498 equivalent pains that her victims suffer at the same time from other sources, whereas the lesser pains caused by each Inflictor are not contributions to greater pain. Each Contributor’s act is, however, less seriously wrong, at least in this one respect, than the act of a Harmless Torturer; for the pains each Contributor inflicts are contributions to suffering by individuals that is only half as intense as that to which each Harmless Torturer contributes. Given these various similarities and differences between the effects of the Contributors’ acts and the effects of the acts of the other inflictors of tiny harms, it is perhaps difficult to determine whether what each Contributor does is as ser­ ious­ly wrong as what each Harmless Torturer does. There may be reasonable dis­ agree­ment about whether one of the considerations just noted is more important than another, and thus reasonable disagreement about whether the act of a Contributor is more or less seriously wrong than the act of a Harmless Torturer. In my view, however, it is not implausible to suppose that what each Contributor does is less seriously wrong, at least slightly. Although each Contributor contributes more to the suffering of each of her victims, it may be more important to the determination of how seriously wrong her act is that it contributes to the suffering of victims whose suffering is only half as bad as that of the victims of the Harmless Torturers.

3.  The Irrelevance of Collaboration My claim, then, is that it is a crucial part of the explanation of why what each Harmless Torturer does is wrong, and seriously wrong, that his act contributes to, or increases, individual suffering (or, more generally, harm) that is extremely bad. It is also important, of course, that it does this to a very large number of victims.

330  Jeff M c Mahan The great suffering to which each Harmless Torturer contributes is, of course, the product of the acts of the other Harmless Torturers. Each one’s act is wrong, in other words, because it is a contribution to what they together do, which is to cause great suffering to a large number of victims. It may therefore seem that my claim is no different from Parfit’s second explanation. But, as I indicated earlier, I share Parfit’s scepticism, stated in the later reprinting of Reasons and Persons, about the moral significance of collaboration. What matters, in my view, is that each Harmless Torturer is knowingly and intentionally adding another tiny pain to the very great suffering that each victim experiences from other sources. With Parfit, I believe that it does not matter to the wrongness of the Harmless Torturer’s act whether those sources are agents with whom he is collaborating or something else. I believe, for example, that there is no moral difference between what a Harmless Torturer does and what the Single Torturer does. If the Single Torturer knows that each of a thousand people is about to experience 999 simultaneous insect bites, each of which will cause pain of the same intensity and duration as the pain from a single Harmless Torturer’s shock, and if the Single Torturer chooses then to press his button, what he does is just as wrong, and for the same reasons, as what a Harmless Torturer does. Yet I know of other moral philosophers who believe that it is essential to the evaluation of each Harmless Torturer, and to understanding what may per­mis­ sibly be done to him (for example, in defence, as punishment, or to compel him to compensate his victims), that he is collaborating with others in the infliction of wrongful harms. Some may believe this because they believe that collaboration can affect the degree of a wrongdoer’s culpability. Others may believe it because they believe that, when one collaborates with others, one bears some responsibility for what those others do.11 There are obvious ways in which this latter belief can be true. If, for example, each Harmless Torturer is motivated to press his button in part by pledges from the others that they will also press theirs at the same time, or if his conscience is silenced by the thought that a great many others will also do exactly what he is tempted to do, then each Harmless Torturer does indeed bear some responsibility for the acts of all the others. Yet we can assume, or stipulate, that nothing of this sort occurs in the case of the Harmless Torturers. Let us assume, in the remainder of this essay, that there is an independent reason why each Harmless Torturer wants the thousand victims to experience pain at a specific time. Perhaps each torturer hates the victims because they together constitute a group to which the torturer is averse. And suppose that there is some precise time at which the victims plan to celebrate an event that is of significance to them as members of the group. It is then that each Harmless Torturer wants them to suffer. So each Harmless Torturer is strongly motivated entirely independently of the others to inflict as much pain on the 11  This is Frances Kamm’s view, expressed to me in discussion.

Defence Against Parfit ’ s Torturers  331 victims as he can at that exact time. Of course, each can inflict only a tiny pain on each victim. So each is pleased to discover that the others share his antipathy and plan to do exactly what he plans to do. Each is delighted to learn that, because of the others, he will not just be causing a tiny pain to each victim but will be contributing to each victim’s agony. Perhaps they all rejoice in this together and in this way are collaborators. But each would have pressed his button in the absence of the collaboration and each will press his button at the relevant time whatever the others might do. Given these assumptions, it does not seem that any one Harmless Torturer bears responsibility for what any of the others does. Another problem for the view that collaboration in the infliction of wrongful harm is a distinct wrong that makes what each Harmless Torturer does more ser­ ious­ly wrong than what the Single Torturer does is that it may be unclear what exactly collaboration is, or what it involves. An examination of some examples in the grey area supports the view that collaboration is morally irrelevant, apart from any bearing it might have on an agent’s culpability, or any influence it might have on what people actually do. We might, for example, reflect on these vari­ ations on the case of the Harmless Torturers. The Aspiring Collaborator A man knows that 999 people will, at a precise time, collaborate in causing agony for eight hours to a thousand innocent victims by each inflicting a tiny pain on each victim. This man is not known to any of the collaborators and cannot communicate with them but he is able to inflict an equivalent tiny pain on each of the thousand victims at the same time that the 999 inflict theirs, and does so. The Hopeful Torturers Each of a thousand people has a button that, if pressed, will administer a barely perceptible electric shock to each of a thousand innocent people for eight hours. Each presses his button at the same time in the hope, based on reports he has heard, that many others will also inflict shocks on these same victims at that same time. None of these people can communicate with any of the others; indeed, none knows the identity of any of the others or even whether anyone other than himself has the ability to inflict shocks on the same thousand victims. Each one never­the­ less makes a tiny contribution to the suffering of agony by a thousand victims.

It seems to me that, although the Aspiring Collaborator would like to collaborate with the others, he does not do so. Yet it seems clear that what he does is no less wrong than what each Harmless Torturer does, and that he is no less culpable than any of those who do collaborate. One could, of course, contend that there is a sense in which he does act together with the others even though they do not know it, or a sense in which he is a member of the group that together inflicts the

332  Jeff M c Mahan great suffering even though the others do not know he is. But one must then ask: what difference does this make, morally? What matters is not the proper description of his relation to the others but that he deliberately makes a contribution to the suffering of agony by a thousand innocent people. Similar remarks apply to the Hopeful Torturers. It seems to me that, although they would like to, they do not collaborate or act together. They are not a group and certainly do not constitute a collective agent. But again this seems not to matter. Whether they satisfy some condition for acting together seems irrelevant. What matters is that each one successfully tries to make a contribution to the suffering of agony by a thousand innocent victims. Hence each one’s act is as ser­ ious­ly wrong as that of an individual Harmless Torturer. This is compatible with the possibility that each individual Hopeful Torturer is less culpable than each Harmless Torturer. In the absence of collaboration, each Hopeful Torturer cannot act with the assurance that he will contribute to the infliction of agony on the thousand victims. For all he knows, his act may be like that of an Inflictor rather than that of a Harmless Torturer. If this might diminish his culpability, it illustrates the way in which collaboration may be relevant to culpability. If it is correct that collaboration is irrelevant apart from its effects on the mo­tiv­ ations and acts of those who collaborate, this supports the claim that there is no morally significant difference between any one Harmless Torturer and the Single Torturer. Each deliberately adds his tiny pains to the agonizing pains that he knows his victims will suffer simultaneously from causes that are independent of his action. My claim that the wrongness of what each Harmless Torturer does is partly explained by the magnitude of the individual suffering to which he contributes can thus be restated in prioritarian terms. According to prioritarianism, any increase or decrease in an individual’s well-­being matters more the worse off that individual is at the time. (There is also a version of prioritarianism that is concerned with lifelong well-­being but it is not relevant here.) Although the pain that any Harmless Torturer inflicts on any victim is tiny, each victim is, at the time of the individual Harmless Torturer’s action, extremely badly off in absolute terms because of the action of the other 999. This is the core element of the explanation of why each Harmless Torturer’s act is so much more seriously wrong than each Inflictor’s act. For each of the victims of any of the Inflictors is, we are implicitly assuming, reasonably well off, in the way most people normally are, at the time the Inflictor inflicts his tiny pains. The prioritarian view is also the core part of the explanation of why the act of the Single Torturer is as seriously wrong as that of any Harmless Torturer.12

12  I should be clear that I am not here embracing any particular version of prioritarianism. I am simply appealing to a consideration that is an element of different versions of the larger view—namely, that the worse off an individual is independently of one’s action, the more seriously objectionable it is to inflict on that individual a fixed amount of pain, if other things are equal.

Defence Against Parfit ’ s Torturers  333

4.  Liability to Defensive Harming The considerations that explain why an act of harming is wrong may also help to explain why the person who will otherwise do the act may be morally liable to be harmed in defence of the potential victim or victims of the act. And these same considerations also contribute to determining how much defensive harm the person may be liable to suffer. The degree of harm to which he is liable is a matter of what is known as narrow proportionality. Narrow proportionality is a constraint on a liability justification for harming a person. If the degree of harm inflicted on a person as a means of preventing the harm he will otherwise inflict on others exceeds the amount that can be justified on grounds of moral liability, that harm is disproportionate in the narrow sense and, according to many writers on defensive harming, wrongs the potential wrongdoer. Narrow proportionality contrasts with wide proportionality, which is a constraint on a lesser-­evil justification for the infliction of harm to which the victim is not liable.13 I have thus far discussed four types of agent: Traditional Torturers, Harmless Torturers (including the Single Torturer), Inflictors, and Contributors. I will refer to agents of all four types as ‘harmers’. I have also noted four factors that I believe contribute to determining the degree to which what these various harmers do is wrong. These are (1) the magnitude of the harm that each inflicts on individual victims, (2) the total sum of harm that each inflicts (which is a function of the magnitude of the harm to individual victims and the number of those victims), (3) the magnitude of the total harm to each victim to which the harmer contributes, and (4) the degree of the harmer’s responsibility or culpability.

One might question whether the degree of a harmer’s responsibility or culp­ abil­ity can affect the degree to which that harmer’s action is wrong. The wrongness of an act and the culpability of the agent are in general distinct; hence, an act can be seriously wrong and yet the one who does it may not be culpable at all. This is what is meant when it is said that a person is excused for wrongful action. Yet I believe it is not unreasonable to suppose that the wrongness of an act—even the objective or fact-­relative wrongness of that act—is exacerbated if the agent is highly culpable rather than merely responsible or altogether excused in the doing of the act. I will, in any event, assume that this is so, though it is not essential to the points I will make. 13  For elucidation and further discussion, see Jeff McMahan, ‘Proportionate Defence’ (revised and expanded version), in Jens Ohlin, Larry May, and Claire Finkelstein (eds.), Weighing Lives in War (Oxford: Oxford University Press, 2017): 131–54.

334  Jeff M c Mahan The same four factors all seem relevant as well to the degree of harm to which the different types of harmer may be liable in defence of their victims. In one of the cases I have discussed—the Traditional Torturers—the amount of harm that each harmer inflicts on his individual victim is very great: agony for eight hours. This, combined with the assumption that each Traditional Torturer is culpable, makes each one liable to a defensive harm significantly greater than that which he would otherwise inflict. Let us assume that each is morally liable to be killed to prevent him from torturing his single victim. This, I believe, is the common-­sense view and is not implausible. By contrast with a Traditional Torturer, each harmer of the other three types inflicts no more than a very slight harm on any individual victim. Yet we may assume that each of these harmers is culpable to some degree. Each, we may suppose, acts maliciously without justification or excuse. If these were the only rele­ vant factors, these harmers might be liable to a defensive harm somewhat, though not substantially, greater than the slight harm they would inflict on their victims. But of course all three of these types of harmer have a great many individual victims. Because of this, the total sum of harm that each inflicts is, as Parfit notes, as great as that which a Traditional Torturer inflicts. This seems to increase the amount of harm to which each is liable as a matter of defence. Yet the amount of harm to which an individual Inflictor is liable to prevent her from pressing her button seems intuitively not to be very great. I believe, for example, that it would be disproportionate to cause her to suffer a hundred simultaneous pains, each of the same intensity and duration as of one of the thousand pains she would other­ wise inflict. It seems, therefore, that the total sum of harm that an act inflicts is not a source of liability to substantial defensive harm, provided the act neither causes nor contributes to a significant harm to any individual. Although no Harmless Torturer causes more than a trivial harm to any one individual victim, each does contribute, albeit in a comparatively trivial way, to the infliction of great harm on each of his victims. Because of this, the total sum of harm caused by each Harmless Torturer has greater significance than that caused by each Inflictor, though less significance than that caused by a Traditional Torturer. But, again, the total sum of harm that a person causes seems to have little significance except insofar as it is the product of significant harms to many individuals or of contributions to significant harms to many individuals. This seems true of the significance that the total sum of harm has both in determining the degree to which an act of harming is wrong and in determining the amount of harm to which the harmer is liable to prevent him from harming others. Assuming, at least for the sake of argument, that each Traditional Torturer is liable to be killed if that is necessary to prevent him from pressing his button, and assuming that each Inflictor is liable only to a comparatively small harm to prevent her from pressing her button, we can now ask how much harm a Harmless Torturer is liable to as a means of preventing him from pressing his button. This, I will

Defence Against Parfit ’ s Torturers  335 argue, depends on and varies with certain wholly circumstantial conditions of liability. The circumstantial conditions of liability are conditions over which a potentially liable person has no control. They contrast with the agential conditions of liability, such as what the person intends and why she intends to do it, over which she does have some control. In the case of the Harmless Torturers, one rele­vant circumstantial condition is how many other Harmless Torturers will simultaneously be prevented from pressing their buttons, either by the same defending agent or by others. If the only possibility is preventing one Harmless Torturer from pressing his button, this Harmless Torturer may be liable to only a relatively small defensive harm—one greater than that to which an Inflictor is liable but substantially smaller than that to which a Traditional Torturer is liable. It might, for example, be proportionate to break the Harmless Torturer’s fingers, or his wrists, as a means of preventing him from pressing his button; though because the prevention of his contribution alone would make very little difference to the magnitude of any individual’s suffering, even that might be disproportionate in the narrow sense. Certainly, at least in my view, it would be disproportionate to kill him. Suppose next that five hundred of the Harmless Torturers have already been defensively incapacitated. The only possibility now is to prevent one more from pressing his button. The amount of harm to which this Harmless Torturer is liable is, I think, less than that to which he would be liable if the five hundred others had not been incapacitated. This is because, while the magnitude of the harm he would inflict on each of the thousand victims is the same in each case, the magnitude of the harm to each individual to which he would contribute when only 499  others are contributing is less than half the magnitude of that to which he would contribute if all 999 others were contributing as well. If 999 of the Harmless Torturers had already been incapacitated, the remaining one would be liable to the same amount of harm to which an individual Inflictor is liable. And the amount of harm to which an Inflictor is liable does not depend on how many others would also be prevented from pressing their buttons. The one remaining Harmless Torturer would have effectively become a mere Inflictor rather than a contributor to torture. There is another way in which the amount of defensive harm to which a Harmless Torturer is liable might vary. Suppose that in conditions in which it is impossible to prevent 999 of the Harmless Torturers from pressing their buttons, the maximum amount of harm that it would be proportionate to inflict on the remaining Harmless Torturer to prevent him from pressing his button is n. Next suppose that it becomes possible to prevent two hundred of the thousand Harmless Torturers from pressing their buttons, though only by inflicting on each of them a harm several times greater than n (n × 3). It seems plausible to suppose that each of the two hundred Harmless Torturers could be liable to that amount of harm—that is, that it would not be disproportionate to inflict a harm of n × 3

336  Jeff M c Mahan on each. This is because the elimination of the contribution of any one Harmless Torturer matters more when it is part of a substantial reduction of great suffering than when it occurs alone or is part of only a relatively insignificant reduction of great suffering. Suppose next that 750 of the Harmless Torturers have already been incapacitated. The only further defensive option is to prevent two hundred more from pressing their buttons. This can be done only by inflicting a harm of n × 3 on each of them. We are assuming that n × 3 is the maximum harm to which each of two hundred Harmless Torturers could be liable as a means of reducing the victims’ suffering from agony (a thousand simultaneous small pains) to great but not ag­on­iz­ing suffering (eight hundred simultaneous small pains). Given this assumption, I think it would be disproportionate to inflict n × 3 on each of two hundred Harmless Torturers when that would reduce their victims’ suffering by the same amount, though from moderate suffering (250 simultaneous small pains) to very mild suffering (fifty simultaneous small pains). This is because the elimination of the contribution of any one Harmless Torturer matters more when it is a part of a substantial reduction of very great suffering than when it is part of an equally substantial reduction of suffering that is significantly less bad.

5.  Defensive Killing I have suggested that the prevention of one Harmless Torturer’s contribution to the suffering of the thousand victims matters more when it would be part of a substantial reduction in the suffering of many victims, each of whom would other­ wise experience very great suffering. The prevention of one Harmless Torturer’s contribution therefore matters most when it would be accompanied by the prevention of the contributions of all the other 999. In this case, the harm to which a Harmless Torturer would be liable should be greater than that to which he would be liable if the prevention of his contribution were part of a less substantial reduction of suffering that was already less bad. Suppose that a third party has only two options: she can either do nothing, thereby allowing all thousand Harmless Torturers to press their buttons, or she can simultaneously painlessly kill all of them, thereby preventing them from pressing their buttons. Given the assumption that it would be permissible for a third party to kill all thousand Traditional Torturers, I think we should accept that it would be permissible, in these conditions, for a third party to kill all thousand Harmless Torturers. Indeed, if we accept that it would be obligatory for a third party to kill all thousand Traditional Torturers, provided that this could be done without excessive personal cost, then I think we should accept, with the same proviso, that it would also be obligatory for a third party to kill all the Harmless Torturers rather than not kill any. For the morally relevant effects would

Defence Against Parfit ’ s Torturers  337 be the same in both cases: the same thousand culpable people would be killed to prevent them from doing what would make them causally and morally re­spon­ sible for the agonizing suffering of the same thousand innocent people. The justification for the killing of each Traditional Torturer is a liability justification. Each is morally liable to be killed because killing him is a necessary and proportionate means of preventing him from torturing an innocent person. Similarly, in a choice between killing all the Harmless Torturers and allowing them to press their buttons, the justification for killing them is also, I believe, a liability justification. Assuming that it is in fact permissible to kill them, one might argue that the justification must be a liability justification by default, for there does not seem to be any other form of justification that applies: the Harmless Torturers do not deserve to die, killing them is not the lesser evil (as eight hours of agony is not worse than death, when death involves the loss of many years of good life), the third party (we may assume) is not specially related to any of the victims, and so on. Some philosophers with whom I have discussed the case of the Harmless Torturers say, however, that they think that no Harmless Torturer is liable to be killed, irrespective of how many it is possible for a third party to kill. This is because each Harmless Torturer, like each Inflictor, causes no more than a trivial pain to any individual. Although each Harmless Torturer, again like each Inflictor, causes a great many such pains, these pains do not aggregate to become morally significant. To kill a Harmless Torturer to prevent him from inflicting these trivial pains would therefore, these philosophers claim, be disproportionate. This understanding of the grounds of liability raises a problem to which Philip Pettit has called attention in another context. Pettit observes that the failure to impose a regime of corporate responsibility can expose individuals to a perverse incentive. Let human beings operate outside such a regime, and they will be able to incorporate, so as to achieve a certain bad and self-­serving effect, while arranging things so that none of them can be held fully responsible for what is done.14

Applied to Parfit’s cases, the concern is that the claim that no Harmless Torturer is liable to be killed has a highly implausible implication—namely, that the Traditional Torturers, each of whom is probably liable to be killed and certainly liable to a defensive harm significantly greater than eight hours of agony, could achieve exactly the same bad effects while altogether escaping liability to serious defensive harm simply by reprogramming their buttons to operate in a way that would make them Harmless Torturers rather than Traditional Torturers.

14  Philip Pettit, ‘Responsibility Incorporated’, Ethics 117 (2007): 171–201, at 196.

338  Jeff M c Mahan This is not just a concern about the practical consequences of the acceptance of a view that would exempt the Harmless Torturers from liability to serious harm. It is an objection to the truth of the view itself. But is the implication actually unacceptable? It is clearly possible for a person to arrange his causal environment in such a way as to enable him to engage in serious wrongdoing while ensuring that others have no proportionate means, and indeed no permissible means, of preventing his action—for example, by surrounding himself with a large number of ‘innocent shields’, thereby making it impossible for others to prevent him from committing a murder without killing all of his innocent captives as a side effect. There is, however, an important difference between the example of the murderer with many innocent shields and the case of the Harmless Torturers. Defensive action that would kill the murderer’s innocent shields would be disproportionate in the wide sense. The murderer is himself liable to be killed but the liability justification for killing him is overridden by the rights of others who would be killed if one were to act on that justification. But the claim of those who deny that any Harmless Torturer is liable to be killed is that killing any of them would be disproportionate in the narrow sense. Because harm that is disproportionate in the narrow sense is, by definition, harm in excess of that to which a person is liable, no one can be liable to harm that is disproportionate in that sense. The question, therefore, is simply whether a Harmless Torturer can be li­able to be killed in the circumstances. It is important, in considering this question, not to confuse the issue of liability with that of desert. It is obvious that, even if each Harmless Torturer deserves some degree of harm, the harm of death would be vastly disproportionate as a matter of desert. But the harm to which a person can be liable is sensitive to morally arbitrary circumstances (again, the ‘circumstantial conditions of liability’) in ways that the harm a person deserves is not. In particular, the harm to which a threatening person can be liable as a matter of defence is highly sensitive to who else will unavoidably be harmed, and by how much, if the threatening person is not harmed. The harm to which a person is liable is thus determined by comparisons among possible courses of action, whereas the harm a person deserves depends only on what that person has done or is doing. Because of this, the harm to which a person can be liable may be much greater or much less than the harm, if any, that he deserves. Because of this comparative dimension to liability, there is good reason to accept that, if the choice is between killing all and killing none, each of the Harmless Torturers is morally liable to be killed. In the circumstances, it is un­avoid­able that either a thousand potential victims will each suffer a grave harm or a thousand different people—the Harmless Torturers—will each suffer a significantly greater harm. But none of the potential victims bears any responsibility for this situation. Each of them is entirely morally innocent. Each Harmless Torturer, by contrast, is fully responsible for a small part of the grave harm that

Defence Against Parfit ’ s Torturers  339 each of the potential victims may suffer. Because each Harmless Torturer would be an intentional contributor to these grave harms, each is culpable and bears a share of the responsibility for the fact that a third party must choose between inflicting a greater harm on each of them and allowing them together to inflict grave harms on each of the potential victims. Even though death is a significantly greater harm than eight hours of agony, it is not a disproportionate harm in these circumstances, in which each Harmless Torturer could have chosen, without personal cost, not to press his button, thereby avoiding becoming culpably responsible for a situation in which it is unavoidable that each of a thousand people will suffer a very great harm. It is thus a matter of justice that each of the innocent people should be defended from being caused to suffer a very great harm even if each of the culpably responsible people must be caused to suffer an even greater harm as a means of defence. It is essential to each Harmless Torturer’s liability to be killed in certain conditions of choice not only that the instances of individual suffering to which he intentionally contributes are extremely severe but also that he contributes to the suffering of many victims. We can imagine a case in which this latter feature is absent. The Harmless Torturers with a Single Victim Each of a thousand people has a button that, if pressed, will administer a tiny electric shock to a single innocent person for eight hours. Each knows that 999 other people will press a button at the same time that will have the same effect on the same victim. The result of their combined acts will be that a single victim will experience agonizing suffering for eight hours.

Suppose that the only way that the potential victim or a third party can prevent any of these Harmless Torturers from pressing their buttons is to kill them. Although some moral philosophers with whom I have discussed this example believe that it would be permissible for the potential victim or a third party to kill any or all of these Harmless Torturers, I believe that killing any number of them would be disproportionate and therefore that none of them is morally liable to be killed.15 Killing only one of them would make only a trivial difference to the suf15  For a defence of a view that implies that, if a Traditional Torturer is liable to be killed, any or all of the Harmless Torturers with a single victim are liable to be killed, see Helen Frowe, Defensive Killing (Oxford: Oxford University Press, 2014), 78 and 175. For objections to this view, see my review of the book, cited in n. 9. Frowe’s view combines the idea that collaboration is itself morally significant with the idea that the defensive harm to which a threatener is liable is affected by the magnitude of the harm to individuals to which he contributes. According to her view, the defensive harm to which a threatener is liable is, when he acts alone, a function of the magnitude of the harm he would himself cause, whereas when he acts in collaboration with others, the defensive harm to which he is liable is instead a function of the magnitude of the harm that he and his collaborators would together cause, to which he would only contribute. This view implies that all the Harmless Torturers, and even all the

340  Jeff M c Mahan fering of the victim but would, we may suppose, deprive the Harmless Torturer of as many as thirty to fifty years of good life. Killing a second one as well would also make only a trivial difference and the suffering that would be reduced by the second killing would be slightly less bad than that which would be reduced by killing the first. Killing all of them would, of course, prevent the innocent victim from suffering eight hours of agony, but the cost of that would be that a thousand ­people, each of whom would otherwise have inflicted only a  tiny pain on one person, would together be deprived of a total of 30,000–50,000 years of good life. I find it impossible to believe that this could be proportionate. If I am right about this, this example is an instance in which culpable people can evade liability to defensive harm by acting together in the way identified by Pettit. Suppose that each of these thousand Harmless Torturers wants the innocent victim to suffer agony for eight hours. Each would act on his own as a Traditional Torturer to make this happen, were that necessary. But each understands that if he were to attempt to inflict the eight hours of agony by himself, he would become liable to be killed, as killing him would be a necessary and proportionate means of defending the victim. Given that each knows this, and that each knows that there are 999 others who would also be willing to be a Traditional Torturer to make the victim suffer, these thousand people decide to achieve their shared aim by becoming Harmless Torturers, thereby, I believe, avoiding becoming liable to be killed. I have told this story in a way that makes these Harmless Torturers col­lab­or­ ators. And the form of their collaboration has clear moral significance, in that it is intended by each to enable him and the others to achieve an aim they share that is seriously wrong, and to do so without exposing themselves to liability to the only form of defensive action that is possible in the circumstances. Because of this, each is arguably more culpable, and in principle liable to greater harm, than each would be if they were all only Hopeful Torturers with a single victim. Still, because each of these Harmless Torturers with a single victim is willing to be a Traditional Torturer, they are all highly culpable even in the absence of the collaboration. But given that all that each intends to do is to inflict a single barely perceptible pain for eight hours on a single innocent person (rather than single-­handedly inflict eight hours of extreme agony on that person), it remains impossible for me to believe that it could be proportionate to kill any of them in defence of the victim. It is worth mentioning, if only in passing, that Pettit’s implied suggestion that the problem he cites can be avoided by appealing to a doctrine of collective responsibility is of no help in understanding what harm these Harmless Torturers with a single victim might be liable to suffer as a means of defending their victim. Harmless Torturers with a single victim, are liable to be killed, but the Single Torturer (who, like the Harmless Torturers, has a thousand victims) is not.

Defence Against Parfit ’ s Torturers  341 It may well be that, because of their planning and coordination, they together constitute a group agent. And there is certainly a sense in which, were they to act in coordination, they would be collectively responsible for the agony their victim would suffer. But we cannot, I think, infer anything about their individual liability from these claims that is not already implied by the fact that each acts in col­lab­or­ ation with the others to achieve an immoral aim that they share in a way that enables each to avoid becoming morally liable to be killed. Recognizing that they constitute a group agent that is responsible for torturing an innocent person would not make it any less implausible to suppose that any of them are liable to be killed. Similarly, claims about group agency and collective responsibility are unlikely to help in understanding how much harm the Harmless Torturers with a thousand victims might be liable to suffer in defence of those victims. As I hope the subsequent discussion will show, each Harmless Torturer’s liability can vary with the options available to the defender, and there seems to be scope for reasonable disagreement about individual liability in many of the possible cases. And because I have stipulated that, although there is a sense in which the Harmless Torturers collaborate, what each does is causally independent of what any other does, claims about group agency and collective responsibility are substantially less plausible in this case. In these conditions, I suggested, each Harmless Torturer is relevantly like the Single Torturer, who simply adds his pains to many others with causes (such as insect bites) that are entirely beyond his control. There can be no collective responsibility in such a case.

6.  Liability in Varying Conditions of Choice In Section 5 I sought to explain why, when a third-­party defender can only kill all the Harmless Torturers or else kill none, every Harmless Torturer is liable to be killed. Perhaps the most salient circumstantial condition of their liability is that either all the innocent victims who bear no responsibility for the unavoidability of great harm must suffer grave harms or all the Harmless Torturers who are culp­ ably responsible for the unavoidability of great harm must suffer even graver harms. In these conditions, proportionality is not assessed by reference to the magnitude of the individual harms that each Harmless Torturer threatens to inflict. The assessment of proportionality instead requires weighing the difference in moral responsibility between the Harmless Torturers and the innocent victims against the difference in the magnitudes of the individual harms that the members of each group might suffer, given that the members of one group must suffer those harms. On the assumption that it is proportionate to kill a Traditional Torturer rather than allow him to inflict eight hours of agony on an innocent victim, it seems that it is also proportionate to kill the Harmless Torturers rather than allow them to inflict eight hours of agony on their innocent victims.

342  Jeff M c Mahan The assessment of liability is quite different, however, in conditions in which harm is unavoidable but can be divided or shared between those who are morally responsible for this and those who are not. Suppose, for example, that a third party could either kill a Traditional Torturer, thereby preventing him from causing any pain to his victim, or cause him to be permanently quadriplegic, thereby preventing him from inflicting nine hundred of the thousand tiny pains he would otherwise inflict on his victim. In the latter case, the victim would have to suffer a hundred tiny pains—relatively mild but still significant pain—for eight hours. Assuming that continued life in a quadriplegic condition is very substantially better than death, it might be that in these conditions of choice the Traditional Torturer would not be liable to be killed but only to being made quadriplegic. This is not, however, a matter of proportionality but is instead a matter of a different constraint on a liability justification—namely, necessity. Although it is generally assumed that the necessity constraint simply requires that a defender cause the least harm necessary to achieve a particular defensive aim, I believe that this is an oversimplification. Suppose, for example, that a third party has two options for defending an innocent victim from a Traditional Torturer. She can prevent the Traditional Torturer from inflicting all thousand tiny pains, but only by killing him; or she can prevent him from inflicting 999 of the tiny pains by giving him a sharp pinch. If she had only the first option, the killing would be both necessary and proportionate. But given that she has the second option, killing the Traditional Torturer is unnecessary. Killing him is of course physically necessary for the full defence of the victim, but it is morally unnecessary because necessity is sensitive to trade-­offs between the defensive harm inflicted and the wrongful harm to be prevented. The same is true, though less obviously so, when a third party can choose between killing a Traditional Torturer to spare the victim all thousand tiny pains and causing the Traditional Torturer to become quadriplegic to spare the victim nine hundred of the tiny pains. The first of these options is wrong because it is morally unnecessary. I believe, moreover, that a person cannot be liable to defensive harm that is morally unnecessary—or, as this claim is sometimes expressed, that necessity is ‘internal’ to liability.16 If this is true, and if killing the Traditional Torturer is unnecessary when he could be caused to be quadriplegic instead, it follows that in these conditions he is not morally liable to be killed. He is liable only to be made quadriplegic and his innocent victim is morally required to endure a hundred tiny pains for eight hours (though not as a matter of liability). This may seem unfair to the victim but in fact the requirements of necessity and proportionality often require innocent victims to endure harms from culpable aggressors (for 16  For a defence of this claim, see Jeff McMahan, ‘The Limits of Self-­Defence’, in Christian Coons and Michael Weber (eds.), The Ethics of Self-­Defence (New York: Oxford University Press, 2016): 185–210, at 195–7.

Defence Against Parfit ’ s Torturers  343 example, when the only way to prevent a malicious person from giving an innocent person a vicious pinch is to kill him). Harms can be divided or shared by culpable threateners and innocent victims either individually or collectively. Suppose that a third party could (1) kill every Traditional Torturer, thereby preventing each from inflicting any pain on his intended victim, (2) cause each Traditional Torturer to become quadriplegic, thereby preventing him from inflicting nine hundred of the thousand tiny pains he would otherwise inflict, or (3) allow all the Traditional Torturers to inflict agony on their victims. This is a case in which individual sharing of harms is possible. Either every victim but no Traditional Torturers will be harmed, or each torturer but no victims will be harmed, or each torturer and each victim will suffer a lesser harm than he or she would suffer in one of the other possible outcomes. In these conditions, in which some harm is unavoidable, the third option offers a way for each victim and each torturer to suffer a lesser harm than either would suffer in one of the other options, thereby sharing the unavoidable harm, though of course they do not literally divide the same harm between them. Suppose next that a third party could (1) kill all of the Traditional Torturers, (2) kill five hundred of them, or (3) kill none of them, thus allowing each to inflict agony on his victim. Suppose the third party chooses, impermissibly in my view, to kill only five hundred of the Traditional Torturers. No individual torturer is forced to share a harm with his victim. Rather, the five hundred who are not killed suffer no harm while their victims each suffer all the harm that, in conditions in which individual sharing was possible, each might have shared with her torturer. But there is a collective sharing of harms. Half the members of the group of victims are spared any harm while each of the Traditional Torturers suffers the full harm that, in other conditions, he might have shared with his victim. In the remainder of this section, I will discuss only cases in which collective sharing of harms is possible. In particular, I will assume that the only means of preventing any Harmless Torturer from pressing his button is to kill him, but that a third-­party defender has, in various possible cases, options other than simply killing all or killing none. In these conditions, the liabilities of the Harmless Torturers can be quite complicated. This is an important difference between the Harmless Torturers and the Traditional Torturers (in the original example, in which the only option is to kill them)—a difference that does not emerge in Parfit’s discussion, which is limited to understanding whether, and if so why and to what extent, what the Harmless Torturers do is wrong. Any Traditional Torturer is liable to be killed no matter how many others might be prevented from pressing their buttons. If a third party has a choice between killing one Traditional Torturer and killing none, it is permissible, and arguably obligatory, for her to kill the one. And, given the choice, it is always permissible, and perhaps obligatory, for a third party to kill more Traditional Torturers rather than fewer. But if a third party has a choice between killing one Harmless Torturer and killing

344  Jeff M c Mahan none, it would be impermissible for her to kill the one. Although the one Harmless Torturer would be liable to a certain amount of harm to prevent him from pressing his button, it would be disproportionate to kill him; for killing him would, in these circumstances, make almost no difference to the suffering of any victim. And because one cannot be morally liable to harm that is disproportionate in the narrow sense, no one Harmless Torturer can be liable to be killed on his own. It is, moreover, not always permissible for a third party, if given the choice, to kill more Harmless Torturers rather than fewer. Indeed, as we will see, the truth seems to be the other way around. If these claims about the liability of individual Harmless Torturers in certain conditions of choice are true, it seems that at least some of the Traditional Torturers could achieve their aims, or at least have an equal probability of achieving the aims, while evading liability to be killed. They could do this by reprogramming their buttons, thereby becoming Harmless Torturers, when a defender is able to prevent them from pressing their buttons only by killing them, but can kill varying numbers of them between, and including, all thousand and none. I will give examples of this in the remainder of this section. These will be examples involving only Harmless Torturers, but one can understand them as Harmless Torturers who were formerly Traditional Torturers but have reprogrammed their buttons. In these examples, all the Harmless Torturers are conditionally liable to be killed. They would, for example, all be liable to be killed if the defender’s options were restricted to killing all and killing none. And even in cases in which the defender has many options for killing varying numbers, each is initially conditionally liable to be killed because, if it is permissible for the defender to kill only some number fewer than all, it makes no difference which particular Harmless Torturers she kills. But once she kills that number, the remaining Harmless Torturers cease to be even conditionally liable to be killed. Suppose next that the only options were killing five hundred Harmless Torturers and killing none. My intuition is that five hundred of them might well be liable to be killed in these circumstances—that is, that it would be proportionate to kill them. (More precisely, all thousand are conditionally liable to be killed; it would not matter, if other things are equal, which five hundred one killed.) I am, however, less confident about this judgement than I am about the judgement that all thousand are liable to be killed if the choice is between killing all and killing none. This is because, as I suggested earlier, the prevention of any one Harmless Torturer’s contribution matters more—so that he is liable to greater harm—when it is part of a larger reduction in suffering that is itself greater or more intense. One could continue to explore a great many other possibilities, including choices between killing some and killing none, sequential choices, and ‘paired’ choices, such as a choice between killing some and killing none when some o ­ thers will be prevented from pressing their buttons by other causes. But there is a more

Defence Against Parfit ’ s Torturers  345 difficult question. Suppose that the only way to prevent any Harmless Torturer from pressing his button is to kill him. And suppose that a third party could, at no personal cost, kill any number of them. How many would it be permissible for her to kill? It would certainly be impermissible for her to kill all thousand rather than 999; for if 999 were killed, the remaining one would not, as I noted, be a Harmless Torturer but an Inflictor, a minor nuisance to a thousand people. Killing him in addition to killing the other 999 would be disproportionate. And the same seems to apply to killing 999 rather than 998. If 998 were killed, both the remaining ones, even acting together, would be nuisances rather than contributors to torture. Surely, however, there is some number (though not a number that can be identified with precision) whose killing would make a sufficient difference to the suffering of each victim that killing them would be proportionate. Killing a hundred Harmless Torturers, for example, would substantially reduce the suffering of all thousand victims. It would still leave them in terrible pain, but the reduction would be far from trivial. The problem, however, is that we are not comparing killing a hundred only with killing all thousand, or only with killing none. We are assuming that it is possible to kill any number, from zero to a thousand. We know that to kill only one is impermissible, whether the alternative is to kill none or to kill some greater number. And we know that killing all thousand is impermissible when killing somewhat fewer is an option. Is there an optimal number between zero and a thousand that it would be permissible to kill? If there were, it might be impermissible to kill more than that, or fewer. There is a general problem here that the work of certain philosophers, above all that of Larry Temkin, has made familiar.17 It would, as we have seen, be wrong to kill only one Harmless Torturer. This is because the reduction in suffering it would bring to any one victim is so slight that, even though there are a thousand victims, it would be disproportionate to kill a person, even a culpable person, to bring about those barely perceptible reductions. But killing two of the thousand Harmless Torturers would seem to be even more seriously wrong than killing one. The killing of the second one would, like the killing of the first, achieve only a trivial reduction in the suffering of any victim; but it would also be wrong for an additional reason—namely, that the trivial reduction in each victim’s suffering that would be achieved by killing the second Harmless Torturer would be a reduction in suffering that would be less bad, albeit only slightly, than the suffering that would be reduced by the killing of only one. That is, while the suffering reduced by killing only one Harmless Torturer would measure 1,000 in intensity, that which would be reduced by killing a second would be 999 in intensity. (This is 17  Larry Temkin, Rethinking the Good: Moral Ideals and the Nature of Practical Reasoning (New York: Oxford University Press, 2012).

346  Jeff M c Mahan true not just when the reductions would be temporally sequential but also, as I am imagining, when they would be simultaneous. The suffering that each victim would endure if one Harmless Torturer were killed, and would thus be reduced if a second were killed as well, is 999.) So the all-­things-­considered reason not to kill a second Harmless Torturer is marginally stronger than the all-­ things-­ considered reason not to kill the first; thus, while killing one would be wrong, killing two would be slightly more seriously wrong. It might seem, indeed, that the extent to which killing two would be more wrong than killing one would be greater than the extent to which killing one would be more wrong than killing none. There is, however, another relevant consideration. I have thus far identified two reasons why killing x + 1 Harmless Torturers (x being some number between 0 and 999) is more seriously wrong than killing x. These are (1) that killing the additional Harmless Torturer would make only a trivial difference to the suffering of any victim, so that there would be one more death that would produce only a trivial good effect, and (2) the suffering of each victim that would be trivially lessened by killing x + 1 would be slightly less bad than that which would be lessened by killing only x. A further relevant consideration is (3) that the killing of the additional Harmless Torturer in the killing of x + 1 would be part of a greater reduction in the suffering of each victim than would be achieved by killing only x. It is difficult to assess the relative importance of factors 2 and 3. To test our intuitions, we might ask whether reducing the suffering that a person must endure from 600 to 590 matters more than, less than, or imprecisely just as much as reducing the suffering that a person must endure from 300 to 280. My admittedly weak intuition in this comparison is that it matters more to reduce the greater suffering by less than to reduce the lesser suffering by more. This comparison, which echoes the earlier brief discussion of the Contributors, may suggest that the prioritarian consideration—the magnitude of the individual suffering to which an individual Harmless Torturer would contribute, and thus the magnitude of the suffering that would be reduced by killing him—matters at least slightly more. Yet there may well be no general truth about the relative im­port­ ance of these two factors. Suppose one could reduce the suffering that a person must endure from 1,000 to 999 or reduce that which another person must endure from 999 to 997. In this case my weak intuition is that one ought to produce the slightly greater reduction of the slightly less intense suffering. The issues here are subtle. If a third party kills fewer Harmless Torturers, the intensity of the suffering that would be reduced by the killing of each one would be greater than the intensity of the suffering that would be reduced by killing more in addition. But by killing the additional ones as well, the third party would produce a greater reduction in the suffering of each victim. Whereas the first consideration favours killing fewer, the second favours killing more. I will simply assume that, in a comparison between killing a greater number and killing a

Defence Against Parfit ’ s Torturers  347 lesser number, in conditions in which one could kill any number, these opposing considerations simply cancel each other out. This still leaves the fact that, in killing x + 1 rather than only x, one would be causing one more person to die as a means of achieving no more than a trivial reduction in any victim’s suffering. In this respect, killing one more is always worse than killing one fewer. If this consideration is not outweighed by some countervailing consideration, and killing only one Harmless Torturer would be wrong (for exactly this reason), then killing two would be more wrong than killing one, killing three more wrong than killing two, . . . killing 999 more seriously wrong than killing 998, and killing all thousand more seriously wrong than killing 999. But we have now reached the conclusion that killing all thousand would be wrong, and not just wrong when the only alternative is killing 999. Assuming that ‘more seriously wrong than’ is a transitive relation, the foregoing abbreviated chain of reasoning entails that killing all thousand is wrong even when the alternative is killing none. Indeed, it seems to entail that, when it is possible to kill any number, killing any number is wrong; for it is always better to kill one fewer, all the way down to none. But these conclusions are, I think, clearly mistaken. If it is permissible to kill all the Traditional Torturers rather than killing none, it should also be permissible to kill all the Harmless Torturers when the alternative is killing none. This is an example of what Temkin calls a ‘spectrum argument’. We begin with an act that kills one person and judge that it would be wrong. We then consider a sequence of acts (a notional rather than temporal sequence) that are like the first except that each succeeding act kills one person more than is killed by the preceding act. We judge that each act that kills one more is more seriously wrong than the act that kills one fewer. An act that kills substantially more people than another should therefore be substantially more seriously wrong. But this does not seem to be true. Killing five hundred does not seem to be substantially more wrong than killing only one. Indeed, killing five hundred seems permissible— unless there is some other number that would be optimal so that one ought to choose that number rather than five hundred. It is, I believe, morally unacceptable to conclude that, although killing all thousand Traditional Torturers is permissible, and better than killing any fewer than that, killing all thousand Harmless Torturers is impermissible when one could kill fewer than that, and indeed more seriously wrong than killing any lesser number. While it is true that killing all thousand would be worse than killing 999, and is impermissible when killing 999 is an option, it is not true that killing all thousand would be worse than killing none, or one, or two, or fifty. Killing all thousand would be permissible if these were the only other options. It may be, as Temkin argues, that these arguments compel us to accept that various relations that we have assumed must be transitive are not in fact

348  Jeff M c Mahan transitive. This may be true, for example, of ‘more seriously wrong than’. But I will not speculate about general responses to the problems posed by spectrum arguments. I have nothing to add to Temkin’s masterly discussions of these arguments.18 I will instead return to the question I posed earlier—namely, whether, if it were possible to kill any number of the Harmless Torturers, there is some number that it would be optimal to kill. One reason to think there is not is that, as we saw, it seems that for any number one might kill, it would always be better to kill one fewer. But we must somehow resist the regress argument to the conclusion that it is not permissible to kill any. What follows is one suggestion for how we might think about the difficult problem of defence against the Harmless Torturers. We should, I think, rule out the permissibility of killing only a small number of Harmless Torturers: one, or two, or ten, or probably even twenty. This would, I will say, be ‘too few’. Even though the magnitude of the individual suffering that would be reduced by killing too few would be very great, the amount by which it would be reduced would be too insignificant to justify killing people, even people who are culpable to whatever extent the Harmless Torturers are culpable. Killing too few would, moreover, be impermissible irrespective of how many other Harmless Torturers it would be possible to kill instead. It would, for example, be impermissible to kill too few rather than kill none and it would be impermissible to kill too few rather than any number greater than too few. (I do not claim to know where the boundary is between too few and enough. It is, in any case, likely to be imprecise. It might be, for example, that any number up to n is too few, and any number over n +10 is enough for killing that number to be potentially permissible. But it may be indeterminate whether killing some number between n and n +10 would be too few or enough.) When it would be possible to kill any number, we should also rule out the permissibility of killing too many. In these conditions, it would be impermissible to kill all thousand, or 990, or even 950. For, as I noted earlier, killing a thousand rather than 999 would be to kill someone who, in the circumstances, would be only a trivial nuisance to a thousand people. To kill 980 rather than 950 would be  to kill thirty people, each of whom would make only a tiny contribution to relatively mild pain that would be suffered by a thousand people. This too would be disproportionate. (Because we are assuming that the defensive killings would be simultaneous rather than temporally sequential, these thirty could be any of the thousand. Their identities do not matter. If one were to kill 980, at least thirty of those killed would not have been liable to be killed, as one cannot be liable to disproportionate harm. At least thirty would therefore have been wronged, but it is arbitrary which thirty these are. If by contrast, the killings were sequential, 18  Theron Pummer, however, has much to add. See his ‘Sorites on What Matters’, in Jeff McMahan et al. (eds.), Ethics and Existence: The Legacy of Derek Parfit (Oxford: Oxford University Press, forthcoming).

Defence Against Parfit ’ s Torturers  349 those wronged might be the thirty killed after 950 had already been killed. Those thirty would not have been liable to be killed, though only because of a morally arbitrary circumstantial condition—namely, their position in the queue.) Killing too many—that is, killing some number in the upper range (which also has an imprecise or indeterminate lower boundary)—differs in at least four important ways from killing too few. First, killing too few is always wrong whatever the relevant alternative might be. But killing some number in the upper range may be permissible if only certain other options are possible. For example, whereas killing 980 is impermissible if it is possible instead to kill 960, or perhaps 940, it is permissible to kill 980 if the only alternative is to kill none, or to kill some number that would be too few (or even, perhaps, some number not too far above the upper boundary of the ‘too few’ range). The second difference is that the numbers in the ‘too many’ range seem more extensive than those in the ‘too few’ range. This is because the magnitude of the suffering that would be reduced by killing some number in the ‘too few’ range is very great, so that reducing it by some fixed amount matters much more than a reduction by the same amount of suffering that is much less severe, such as the suffering that would be reduced by some of the killings if one were to kill some number of Harmless Torturers in the ‘too many’ range. Suppose a third party can kill either eighty of the Harmless Torturers or kill none. To kill eighty, she would have to make a significant personal sacrifice. Suppose further that she would be acting alone, so that her action would reduce the suffering of the victims from 1,000 to 920. Let us assume that she would be morally required to make the sacrifice because this would be a substantial reduction of extremely great suffering. But now suppose that she could act in conjunction with another third party who will kill eight hundred Harmless Torturers whatever she does. Even if it would be permissible for her to kill an additional eighty Harmless Torturers in these circumstances, she might not be required to make the same sacrifice to bring about a reduction of eighty when it would be a reduction of only comparatively mild suffering. The third difference is that, when it is possible to kill any number of Harmless Torturers, killing some number in the ‘too many’ range rather than some number just outside that range is more seriously wrong than killing some number in the ‘too few’ range rather than some number just outside that range. This is because the suffering that would be reduced by killing some number within the ‘too many’ range is relatively mild, whereas that which would be reduced by killing some number within the ‘too few’ range is severe. Yet—and this is the fourth difference— killing some number within the ‘too many’ range would be better than killing some number within the ‘too few’ range precisely because it would make the difference between excruciating suffering and quite mild suffering, whereas killing some number within the ‘too few’ range would make the difference only between more excruciating and less excruciating suffering.

350  Jeff M c Mahan This same consideration suggests that, when it is possible to kill any number, it would be better to kill some number just outside the ‘too many’ range rather than some number just outside the ‘too few’ range. So even within the large range of options between killing too few and killing too many, some options are better or worse than others. Even if it is true that, in a pairwise comparison between adjacent options (such as killing 830 and killing 829), it is always better to kill one few rather than one more, there are other pairwise comparisons in which killing more is better than killing fewer (for example, killing 830 seems better than killing only sixty, even if sixty is outside the ‘too few’ range). Killing n more just outside the ‘too few’ range seems to matter more than killing n fewer just outside the ‘too many’ range because the suffering that would be reduced by killing some number just outside the ‘too few’ range is very great; thus reducing the suffering of the innocent victims just outside that range matters more than, or has priority over, sparing the lives of culpable Harmless Torturers, whereas the opposite may be true in comparisons between reducing suffering and sparing lives just outside the ‘too many’ range. The view that emerges from these reflections is complex and in various respects indeterminate. Killing too few Harmless Torturers is always wrong provided that killing none is an option. To kill ten is to kill too few. Twenty may be too few as well, though I am not sure. There is, however, some number, perhaps fewer than fifty, the killing of whom would make a noticeable, significant difference to the suffering that the victims would experience. It would, I think, be permissible to kill that number rather than kill none. Suppose, for the purpose of illustration, that that number is forty. That might constitute a genuine threshold, at least in conditions of restricted choice. It might, for example, be permissible to kill forty but not only not better to kill thirty-­nine but impermissible to kill thirty-­nine, as thirty-­nine would be too few, or in the zone of indeterminacy between too few and enough. Beyond forty, it might be permissible to kill any number short of too many. Within this broad range between too few and too many, there is no number that it would be optimal to kill. But killing some number might be better or worse than killing some other number. This would depend on how various relevant considerations weigh against one another in the particular comparison. Among the rele­ vant considerations are that it is worse to kill a greater number of culpable people rather than fewer, that it is more important to reduce greater suffering by some fixed amount than to reduce lesser suffering by the same amount, that a greater reduction of suffering of some rough magnitude matters more than a lesser reduction of suffering of the same rough magnitude, and so on. When it is possible to kill any number of Harmless Torturers, there is, I suspect, considerable ineliminable arbitrariness in the choice among options between too few and too many. For almost any number one might kill within that broad range, there is some number that it would be better to kill and some

Defence Against Parfit ’ s Torturers  351 number that it would be worse to kill. (It would not be better to kill fewer than the lowest number that is clearly outside the ‘too few’ range and it would not be better to kill more than the lowest number outside the ‘too many’ range. But it might be much better to kill significantly more than the first of these numbers and much better to kill significantly fewer than the second.) There is no number that it would be optimal to kill, yet it would be much better, and therefore permissible, to kill any number between too few and too many than to kill too few (including none) or too many (including all). Although there are cogent objections to any choice one might make within that range, any of those choices are permissible.

7.  The Relevance of Collaboration to Liability I claimed earlier that the fact that a Harmless Torturer is collaborating with others in the infliction of agony on their thousand victims does not itself affect the wrongness of his action (though it may be correlated with other factors that do). We can see this, I think, when we recognize that the Single Torturer acts just as wrongly as any of the Harmless Torturers when he deliberately adds his tiny pain to the 999 equivalent pains that each of his thousand victims will suffer at the same time from insect bites. I have also claimed that, just as collaboration makes no difference to the wrongness of a Harmless Torturer’s act, so it makes no difference to his liability to defensive harming. I will conclude by briefly considering a challenge to this last claim. Thus far I have considered only cases in which killing a Harmless Torturer would prevent only the effects of his own action. But we can imagine cases in which the killing of one Harmless Torturer would not only prevent him from pressing his button but would also prevent 99 others from pressing theirs, without harming them. In this case, the killing would function both eliminatively and opportunistically: it would eliminate the threat from the Harmless Torturer but also use him as a means of eliminating the threats posed by ninety-­nine others.19 Suppose a third party has only two options: kill only this one Harmless Torturer or kill none. It seems to me, intuitively, that it would be permissible, and indeed obligatory in the absence of significant personal cost, for the third party to kill this one Harmless Torturer. Although this Harmless Torturer is not morally liable to be killed to prevent only his own contribution to the suffering of the victims, the fact that he culpably intends to contribute to the great suffering of these innocent victims makes him morally liable to be killed as a means of achieving a 19 The distinction between eliminative and opportunistic harming was introduced by Warren Quinn in ‘Actions, Intentions, and Consequences: The Doctrine of Double Effect’, Philosophy & Public Affairs 18 (1989): 334–51, at 344.

352  Jeff M c Mahan substantial reduction in their suffering. Assuming, as I have suggested, that to kill a hundred Harmless Torturers would not be to kill too few, killing any hundred of the Harmless Torturers would be proportionate and permissible. Each of the hundred would be liable to be killed as a means of achieving a reduction in every victim’s suffering from a thousand to nine hundred. Assuming that this is correct, it seems that killing only one must be a proportionate means of achieving the same reduction. And, in the circumstances, it would also satisfy the necessity condition. This is so even though the third party would be harmfully using the one Harmless Torturer as a means. Given that the third party must choose between killing the Harmless Torturer as a means and allowing him and ninety-­ nine others to increase the thousand innocent victims’ suffering from severe to agonizing, it seems clear that the Harmless Torturer is liable to be op­por­tun­is­tic­ al­ly used in this way.20 But now return to the case of the Single Torturer, who knows that each of a thousand people will suffer 999 tiny pains from natural causes, such as insect bites, for eight hours. He is about to inflict one more tiny pain for eight hours on each of these victims, bringing their suffering to the same intensity as that of the victims of the Harmless Torturers. I have claimed that what he is about to do is just as wrong as what each Harmless Torturer does, and for the same reason. Yet it may seem, intuitively, that he is not liable to be killed even if killing him would not only prevent his own contribution to the suffering of the thousand innocent victims but also prevent 99 of the 999 insect bites that each victim will suffer independently of his action. How, one might ask, could it be proportionate to kill someone when all he would otherwise do is inflict a tiny, trivial pain on each of a thousand people, even when each of them will be in great pain quite in­de­pend­ ent­ly of his action? But this is, of course, also a description of what each Harmless Torturer does. And, for the reasons I have given above, it seems intuitively plausible to suppose that it would be proportionate, and permissible, to kill one Harmless Torturer to reduce the suffering of each of the thousand victims from 1,000 to 900. Is there a morally significant difference between an individual Harmless Torturer and the Single Torturer that can account for the difference in our intuitions about the opportunistic killing of a Harmless Torturer and the opportunistic killing of the Single Torturer? One might, as I suggested earlier in Section 3, argue that the relevant difference between them is that, whereas the Single Torturer merely inflicts a tiny pain on each of a thousand victims, each Harmless Torturer is collaborating with others in the torture of a thousand victims. Because each Harmless Torturer is 20  One might, alternatively, justify the killing of the one Harmless Torturer by appealing to a variant of Frances Kamm’s Principle of Secondary Permissibility. See Kamm, Intricate Ethics (New York: Oxford University Press, 2006), 170–73.

Defence Against Parfit ’ s Torturers  353 col­lab­or­at­ing with the others, he bears some responsibility for what they do, and this could make him liable to be killed opportunistically to prevent some of the others from pressing their buttons. The Single Torturer, by contrast, bears no responsibility for what the insects do. I argued earlier, however, that if we assume that each Harmless Torturer would have acted in the same way even in the absence of collaboration with the others, there is then no reason to suppose that any one of them is responsible for the action of any of the others, and therefore no reason to think that their collaborating affects their individual liability. One might also think, as I also suggested earlier, that the collaboration of the Harmless Torturers makes each more culpable than the Single Torturer and that this affects their liability. But again we can simply stipulate that the Single Torturer and all the Harmless Torturers are culpable to the same degree. We might, for example, stipulate that the Single Torturer mistakenly believes that he is acting, not on his own, but together with 999 Harmless Torturers. Yet we might still be reluctant to accept that he is liable to be killed opportunistically as a means of preventing each of the victims in his case from suffering 99 insect bites. There remain three broad options. One is to accept that it is permissible to kill a single Harmless Torturer as a means of preventing him and 99 others from pressing their buttons, that there is no relevant difference between the Single Torturer and a Harmless Torturer, and therefore that it is permissible to kill the Single Torturer as a means of reducing the suffering of the victims in that case from 1,000 to 900. Another is to reverse this reasoning and accept that it is impermissible to kill a single Harmless Torturer as a means of preventing him and ninety-­nine others from pressing their buttons. The third option is to find a rele­vant difference between the Single Torturer and the Harmless Torturers. While I believe that we ought to accept the first of these options, there is one difference between the Harmless Torturers and the Single Torturer that could explain why killing an Harmless Torturer opportunistically is permissible but killing the Single Torturer opportunistically is not. This is that what would be prevented by killing the Harmless Torturer are harms caused by wrongdoing, while what would be prevented by killing the Single Torturer are harms resulting from natural causes. I am sceptical of the claim that the reason to prevent a harm caused by wrongdoing is stronger—or significantly stronger—than the reason to prevent an equivalent harm produced by natural causes.21 But many people disagree and the fact that the opportunistic killing of a Harmless Torturer would prevent a hundred acts of serious wrongdoing may well explain the difference in our intuitions about the permissibility of opportunistically killing the Harmless Torturer and the permissibility of opportunistically killing the Single Torturer. 21  See Jeff McMahan, ‘Humanitarian Intervention, Consent, and Proportionality’, in N. Ann Davis, Richard Keshen, and Jeff McMahan (eds.), Ethics and Humanity: Themes from the Philosophy of Jonathan Glover (New York: Oxford University Press, 2010): 44–72, at 60–2.

354  Jeff M c Mahan

Bibliography Frowe, Helen, Defensive Killing (Oxford: Oxford University Press, 2014). Kamm, Frances, Intricate Ethics (New York: Oxford University Press, 2006). McMahan, Jeff, ‘Humanitarian Intervention, Consent, and Proportionality’, in N. Ann Davis, Richard Keshen, and Jeff McMahan (eds.), Ethics and Humanity: Themes from the Philosophy of Jonathan Glover (New York: Oxford University Press, 2010): 44–72. McMahan, Jeff, Review of Frowe, Defensive Killing, in Ethics 126 (2016): 825–31. McMahan, Jeff, ‘The Limits of Self-Defense’, in Christian Coons and Michael Weber (eds.), The Ethics of Self-Defense (New York: Oxford University Press, 2016): 185–210. McMahan, Jeff, ‘Proportionate Defense’ (revised and expanded version), in Jens Ohlin, Larry May, and Claire Finkelstein (eds.), Weighing Lives in War (Oxford: Oxford University Press, 2017): 131–54. Parfit, Derek, Reasons and Persons (Oxford: Clarendon Press, 1984). Parfit, Derek, Reasons and Persons (Oxford: Clarendon Press, 1987 repr.). Pettit, Philip, ‘Responsibility Incorporated’, Ethics 117 (2007): 171–201. Pummer, Theron, ‘Sorites on What Matters’, in Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan (eds.), Ethics and Existence: The Legacy of Derek Parfit (Oxford: Oxford University Press, forthcoming). Quinn, Warren, ‘Actions, Intentions, and Consequences: The Doctrine of Double Effect’, Philosophy & Public Affairs 18 (1989): 334–51. Temkin, Larry, Rethinking the Good: Moral Ideals and the Nature of Practical Reasoning (New York: Oxford University Press, 2012).

16

Overdetermination and Obligation Victor Tadros

Sometimes my act, taken by itself, makes no difference to anyone’s well-­being overall. However, my act is one of a group of acts that either has been performed or would be performed were I not to act. These acts or potential acts together make a very significant negative difference to the well-­being of others overall. Were my act alone not performed, no one would be any better off. Were my act and all of these other acts not performed, someone would be much better off. Where these things are true, is my act morally wrong? My aim is to demonstrate that the best answer to this question calls into doubt two well-­known moral views. One is the view that the wrongness of an act depends either solely on the consequences of that act or solely on the consequences of groups of acts of which that act is a member. These are two versions of act-­ consequentialism. The other is the view that intentions are irrelevant to permissibility. This is the view that if a person with good intentions may perform some act in some circumstances, a person with bad intentions may perform the same act in the same circumstances. It follows, it is argued, that the intentions of the person performing an act cannot make that act wrong. The wrongness of forming certain intentions derives from the wrongness of acts and not the other way around. I will argue against both views. Let us begin with a little conceptual work. Where my act causes some bad result that would have been caused by some other act that has been performed or would have been performed had my act not been performed, the bad result of my act is overdetermined. For example, suppose that X breaks V’s toe. Had he not done so Y would have broken V’s toe. The breaking of V’s toe is overdetermined because it is determined both by X’s act and by the act that Y would have performed had X not performed his act. A bad result might also be overdetermined because of events other than acts that have occurred or that would have occurred were I not to have acted. For example, suppose that it is not Y’s act but rather a falling boulder that would have broken V’s toe. The breaking of V’s toe is overdetermined: it is determined both by X’s act and by the falling of the boulder that would have occurred had X not acted. And, of course, a bad result can be overdetermined by two events neither of which are acts. This is true if a boulder falls on my toe and had it not done so Victor Tadros, Overdetermination and Obligation In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0017

356  Victor Tadros another boulder would have fallen on my toe. This last case is not of interest to us, though. We are interested in the permissibility of acts the bad result of which is overdetermined, either by other real or potential events, including other acts. There are two kinds of case of overdetermination. The first kind, the one that we have just been focusing on, involves pre-­emption. In such a case, the act or event that is pre-­empted has no causal power. In the case where X breaks V’s toe pre-­empting Y’s breaking of V’s toe, for example, only X’s act directly causes the bad result. If Y had no bearing on X’s act Y is not causally responsible for the breaking of the toe. The second kind of case involves what is called simultaneous overdetermination. In such cases two acts are simultaneously performed which together cause the bad result. Each of these acts would independently have been sufficient to produce that result alone. Consider: Simultaneous Killers:  X and Y simultaneously shoot and kill V. Either shot, by itself, would have killed V.

X and Y together kill V. What are we to say about X’s act though? David Lewis once thought that in simultaneous overdetermination cases such as Simultaneous Killers neither event independently causes the outcome.1 This view is revisionist.2 Though it is controversial, I will say that X kills V and that Y kills V and that X and Y together kill V.  This is just a linguistic choice, though, and nothing substantive turns on it. My interest is in the normative implications of overdetermination. First, I consider act-­consequentialist approaches to the problem. I follow Derek Parfit’s view that Simultaneous Killers suggests the falsity of one version of act-­ consequentialism. I then evaluate Parfit’s proposal to revise act-­consequentialism better to handle problems of overdetermination. I offer some criticisms of that proposal that might be met by amending it. Others, though, undermine the whole approach. These criticisms provide some reason to doubt many versions of act-­ consequentialism. They point in the direction of a view that distinguishes ­permissible and wrongful actions on the basis of the motivations with which people act that is at odds with many versions of act-­consequentialism. But they also challenge arguments by non-­consequentialists for the view that intentions are irrelevant to permissibility. The only plausible solution to the problem of overdetermination, I will suggest, is a stone that kills (or at least damages) two families of bird.3 1  See ‘Causation’, Journal of Philosophy 70 (1973): 556–67. 2  See J. Woodward, ‘Psychological Studies of Causal and Counterfactual Reasoning’, in C. Hoerl, T. McCormack, and S. R. Beck (eds.), Understanding Counterfactuals, Understanding Causation: Issues in Philosophy and Psychology (Oxford: Oxford University Press, 2011), 16–53, 35. 3  The stone kills other families of bird too. It really is an impressive stone. But I will focus only on these two.

Overdetermination and Obligation  357 Here is a roadmap of the paper. In Section 1, I introduce Parfit’s proposal to revise act-­consequentialism better to handle the problem of overdetermination. In Section 2, I stipulate what I mean by harm, criticizing Parfit’s revisionist conception. This helpfully clarifies pre-­emption and simultaneous overdetermination problems. In Section 3, I spell out Parfit’s proposal in more detail. In Section 4, I mount two criticisms of Parfit’s proposal that aim at its focus on groups of acts that have actually been performed. In Section  5, I demonstrate that Parfit’s proposal has absurd implications, even when amended. These sections demonstrate that Parfit’s proposal fails. Section 6 explores two other considerations that might distinguish permissible from wrongful acts in cases of overdetermination. I consider the view that harming a person is permitted when, first, that person is not harmed overall because the harm is overdetermined, and, second, that person or another person is benefited. Acting in such circumstances may be Pareto optimal. I demonstrate that Pareto optimal acts of this kind are sometimes wrong. Whether such an act is wrong, I show, sometimes depends on the intentions of the person performing it. The argument in Section  6 supports the view that intentions are relevant to permissibility. Section  7 tentatively explores some more complex motivational states to assess which intentions are relevant to permissibility and under what conditions. I thus make some progress in developing an intention-­based view.

1.  Introduction to Parfit’s View In section 26 of his seminal work Reasons and Persons,4 Parfit explored the implications of Simultaneous Killers for versions of act-­consequentialism. Act-­ consequentialism, as I will understand it, is the view that the permissibility of an act or set of acts is determined solely by the outcome of that act or some set of acts of which that act is a member. Outcomes will be understood agent-­ neutrally—it does not matter whether the outcome is produced by the person performing the act or by another person.5 Standard act-­consequentialism requires people to maximize the good: a person must perform an act if the outcome of that act is better than the outcome of any other act available to the person. The argument in this section poses an equally powerful challenge to more modest consequentialist views that allow for options—for example the view that a person is unconditionally permitted, but is

4 Parfit, Reasons and Persons (Oxford: Oxford University Press, 1984), section 26. 5  Some deny that this is the best way to characterize consequentialism. See e.g. J. Drier, ‘In Defense of Consequentialising’, in M Timmons (ed.), Oxford Studies in Normative Ethics, vol. i (Oxford: Oxford University Press, 2011), 97–119. Associating consequentialism with agent-­neutralism is common, however.

358  Victor Tadros not always required, to perform an act that has the better consequences than any other act available to the person. Here is one version of act-­consequentialism: Single Act-­Consequentialism:  the rightness or wrongness of an act depends solely on the difference that that particular act makes to an outcome.

Simultaneous Killers, Parfit plausibly argued, demonstrates that Single Act-­ Consequentialism is false. X’s act makes no difference to the outcome. V would be dead whether or not X shot him. Yet, Parfit suggested, the verdict that X does not act wrongly in virtue of having killed V is absurd. He is surely right to think this verdict absurd. This holds even if no other act that X could have performed would have had better consequences overall. X acts wrongly even if his only alternative is doing nothing. Perhaps consequentialists might argue that X acts wrongly because there is an extra murderous act, X’s act, that exists if X acts but not if X does not act. The character of X’s act might be thought a ‘constitutive consequence’6 of X acting that makes the outcome worse in agent-­neutral terms. It is difficult to believe that this is why X acts wrongly, though. Furthermore, this idea will not explain what is wrong in cases of pre-­emption. Suppose that X kills V and pre-­empts Y killing V. X does not add to the number of murderous acts performed: one act will be performed if he acts, one will be performed if he refrains. Yet pre-­emptive killing is typically wrong. This is an example of a more general familiar idea that constitutive consequences are of limited value in showing that consequentialism yields intuitive verdicts.7 Parfit did not think that Simultaneous Killers provides an objection to act-­ consequentialism more generally. He thought that act-­consequentialism could be revised to meet the challenge of overdetermination. His revision tied the wrongfulness of an act not to the overall consequences of that act but to the overall consequences of the group of acts of which that act is a member. This seems, at first blush, a natural way to respond to Simultaneous Killers. The acts of X and Y are wrong, on this view, because they are members of a small group of similar acts that together harm V. As Parfit was aware, this approach faces the following challenge. A person is sometimes permitted to perform an act that is a member of a group of acts that 6  This phrase is from Drier, ‘In Defense of Consequentialising’. 7  For the limits of this approach in other contexts, see S. Kagan, Normative Ethics (Boulder, CO: Westview, 1998), 215–19; J. McMahan, ‘Humanitarian Intervention, Consent, and Proportionality’, in N. Ann Davis, R. Keshen, and J. McMahan (eds.), Ethics and Humanity: Themes from the Philosophy of Jonathan Glover (New York: Oxford University Press, 2010), 44–72; and V. Tadros, The Ends of Harm: The Moral Foundations of Criminal Law (Oxford: Oxford University Press, 2011), ch. 7. Kagan also mentions (Normative Ethics, 221–3) the possibility that act-­consequentialists might endorse respecting rather than maximizing or satisficing the good. I leave this version of act-­consequentialism aside as well.

Overdetermination and Obligation  359 are together harmful if that person’s own act harms no one overall. This is sometimes permitted if acting benefits a person overall. For example, suppose that being shot by both X and Y renders V much better off than being shot just by X but much worse off than being shot by no one. Suppose also that Y cannot prevent X shooting V. What is Y to do? He cannot prevent V from being shot. He can either shoot V as well, benefiting V overall, or refrain from doing so. Y seems permitted to shoot V, or even required to shoot V, at least in some circumstances. But if Y shoots V his act is a member of a group of acts which together harm V. Why doesn’t he act wrongly? Parfit developed a sophisticated revision of act-­consequentialism to meet this challenge. Crudely, Parfit thought that the best way to explain why Y’s act is not wrong is that Y’s act would not have occurred had it not been for X’s act. One way to put Parfit’s view, though this is not how he puts it, is as follows: X could independently have averted the threat to V by refraining from acting. This differentiates X’s act from Y’s act. Y could not independently have averted the threat to V. He could not have prevented X from acting. This fact renders Y’s act permissible. The fact that Y’s act is a member of a group of acts which together harms V is not, on this view, decisive. If Y’s act is a member of such a group only as a consequence of X’s act, Y’s act is permissible.

2.  What Is Harm? Before outlining Parfit’s proposal in more detail, I will briefly stipulate what I mean by ‘harm’. This will lead me to present Parfit’s proposal in terms that he does not use; the difference is only verbal, but the stipulation allows us to describe cases in a more natural way. Parfit suggested that neither X nor Y harms V in Simultaneous Killers. X and Y together harm V. Parfit’s is a simple counterfactual account of harm: E harms V iff E makes V worse off than he would have been had E not occurred. As X’s act does not make V worse off than he would have been had X not acted, X’s act does not harm V. As Parfit acknowledged, this account of harm conflicts with common intuitions about Simultaneous Killers. As Parfit also acknowledges, his conception of harm is even more revisionist in pre-­emption cases. Consider: Pre-­emptive Killer:  X tricks V into drinking poison, of a kind that causes a painful death within a few minutes. Before this poison has any effect, Y kills V painlessly.

Parfit’s conception of harm implies, counter-­intuitively, that Y does not harm V. I will not adopt Parfit’s revisionist conception of harm. We are better, at the outset at least, distinguishing between whether some act harms a person and whether that act harms the person overall. The best way to do so is to compare the

360  Victor Tadros actual outcome with the outcome in more than one counterfactual.8 Return to Pre-­emptive Killer. It is intuitive that Y harms V. That verdict compares V’s actual state with the state that he would have been in had neither X nor Y acted. Y doesn’t harm V overall, though. Whether E harms V overall depends on whether V is worse off than she would have been had E not occurred. I will use ‘harm overall’ to refer to what Parfit calls ‘harm’. It is useful to be able to refer to Y as having harmed V in Pre-­emptive Killer. We can then ask: is it permissible for Y to harm V given that Y does not harm V overall? This is a good question. It is distinct from the question whether Y is permitted to act in a way that does not harm V at all. Similarly, we can ask: does Y have a duty to compensate V for the harm that he caused even though he did not harm V overall? And: does Y have a duty to compensate V for the full magnitude of the harm caused even though he did not harm V overall? These questions are difficult to answer, but they are good questions. Our account of harm should make them easy to ask. Of course, there are many counterfactuals that we might consider, which compare V’s state with other states that he might have been in. Some must be set aside. Otherwise every act is harmful, for every act renders a person worse off than she would have been in some possible world. We need reasons to restrict the possible worlds to compare this world with. Spelling out those reasons is a difficult and important task that I leave aside.9

3.  Parfit’s Collective Wrongdoing Principle Parfit thinks that the idea that X and Y each act permissibly in Simultaneous Killers is absurd. He also thinks that neither X nor Y harm V even though X and Y together harm V.  This drives him to conclude that wrongdoing and harming come apart. X and Y each acts wrongly in Simultaneous Killers even though neither harms V. Using our account of harm, we express this view as follows: X and Y each wrongfully harm V even though neither X nor Y harms V overall. 8 Some writers think that pre-­emption and overdetermination problems (and other problems) should lead us to abandon counterfactual accounts of harm altogether. See e.g. S. Schiffrin, ‘Wrongful Life, Procreative Responsibility, and the Significance of Harm’, Legal Theory 5 (1999): 117–48; J. D. Velleman, ‘Persons in Prospect’, Philosophy and Public Affairs 36 (2008): 221–44; and M. Hanser, ‘The Metaphysics of Harm’, Philosophy and Phenomenological Research 67 (2008): 429. I argue against these views, and for the view that many counterfactuals are relevant in determining whether someone has been harmed, and how much, in Wrongs and Crimes (Oxford: Oxford University Press, 2016), ch. 10. 9 This is a general problem for counterfactual analysis. Compare counterfactual approaches to ­causation, which must also pick out some counterfactuals rather than others from an infinite range of possibilities. For recent work, see C. Hitchcock, ‘Counterfactual Availability and Causal Judgements’, and P.  Menzies, ‘The Role of Counterfactual Dependence in Causal Judgements’, both in Hoerl, MacCormack, and Beck (eds.), Understanding Counterfactuals, Understanding Causation, 171–185 and 186–207 respectively. Difficulties in picking out counterfactuals relevant to the badness of death are explored in depth in J. McMahan, The Ethics of Killing: Problems at the Margins of Life (Oxford: Oxford University Press, 2002), ch. 2.

Overdetermination and Obligation  361 Parfit also believes that Y acts wrongfully in Pre-­emptive Killer. His conception of harm implies that Y does not harm V. The account of harm offered above results in the following description of the same idea: Y wrongfully harms V even if he does not harm V overall. With these revisions, Parfit’s conclusions are highly intuitive. Let us accept them. He then considers a third case: Pre-­emptive Killer (Save Variation):  X tricks V into drinking poison of a kind that causes a painful death within a few minutes. Y knows that he can save V2’s life if he acts in a way whose inevitable side-­effect is V’s immediate and painless death. Because Y also knows that V is about to die, Y acts in this way.

Parfit claimed that Y acts permissibly. This view is also highly intuitive. Here is a central component of the explanation why Y acts permissibly on any plausible view. V is doomed to die whatever Y does. Y’s actions do not harm V overall. They save V2 from a great harm. Y’s reasons to protect V2 from this harm are much stronger than any reason that Y has to ensure that V is not killed by Y rather than X. It is Pareto optimal for Y to kill V if his only other option is to do nothing: if Y kills V, V2 is much better off than he would have been had Y not killed V and no one is worse off. Even if Pareto optimal action is not always permitted,10 the fact that an act is Pareto optimal is an important part of a case in favour of its being permitted. Parfit’s moral verdict about Pre-­emptive Killer (Save Variation) led him to: Parfit’s Collective Wrongdoing Principle (Parfit’s Principle):  Even if an act harms no one, this act may be wrong because the person performing the act is a member of a group of people whose acts together harm people. A person is a member of such a group if she is a member of the smallest group of whom it is true that, had these people all acted differently, the other people would not have been harmed.11

Parfit’s Principle implies Parfit’s intuitive verdict that X but not Y acts wrongly in Pre-­emptive Killer (Save Variation). According to this principle, whether a person acts wrongly depends on whether she is a member of the smallest group of whom it is true that, if they had acted differently, the other people would not have been harmed. In Pre-­emptive killer (Save Variation) this group is restricted to X. If X had acted differently, V would not have been harmed as Y would not have harmed 10  Doubts about whether Pareto optimality is always decisive in cases of pre-­emptive harm go back to Bernard Williams’s famous Jim and the Indians case outlined in ‘A Critique of Utilitarianism’, in J.  J.  C.  Smart and B.  Williams, Utilitarianism: For and Against (Cambridge: Cambridge University Press, 1973), 98–9. 11  Parfit’s Principle is not stated in this way in Reasons and Persons. I have composed it from C7 and C8, at 70 and 71 respectively.

362  Victor Tadros X to save V2. If Y had acted differently, X would nevertheless have killed V. Hence, Parfit’s Principle implies that X but not Y acts wrongly.

4.  Acts and Groups of Acts In this section I make two specific criticisms of Parfit’s Principle. Before making them, it is worth noting at the outset that it is difficult to support the view that the permissibility of an act depends on the characteristics of other acts in the way that Parfit’s Principle suggests. Why should the wrongness of my act depend on whether the non-­ performance of some other act could independently have averted the harm that I cause? Beyond exploring its ability to explain intuitive verdicts in cases of overdetermination, Parfit did not defend this idea. This basic concern should already make us doubt that Parfit’s proposal provides an adequate response to the problem of overdetermination. It is worth bearing this general concern in mind when evaluating the criticisms mounted both here and in Section 5. The criticisms that I make in this section are both directed at the same target. As I have just suggested, Parfit’s Principle focuses our attention on the possibility that the non-­performance of some other act that has been performed could independently have averted the harm caused. This, I will argue, is doubly mistaken.

4.1.  Pre-­empting the Consequences of Acts that Haven’t Yet Been Performed The first criticism arises from the fact that whether pre-­ emptive action is permissible depends on the inevitability of the performance of a similar act rather than on whether some similar act has actually been performed. Parfit’s Principle implies that in order to determine whether some act is wrong we must consider whether the harm caused would have occurred had some other acts that have been performed not been performed. This leads to the wrong verdict if pre-­ emptive action is motivated not by an act that has already been performed whose result is pre-­empted, as in Pre-­emptive Killer, but by the dispositions of another person to act if pre-­emptive action is not taken. Consider: Pre-­emptive Killer (Disposition Variation):  X is disposed to trick V into drinking poison of a kind that causes a painful death within a few minutes. Y knows that he can save V2’s life if he acts in a way whose inevitable side-­effect is V’s immediate and painless death. This is the only way to prevent X from poisoning V. Because Y also knows that V is about to die, Y acts in this way.

Overdetermination and Obligation  363 The only salient difference between Save Variation and Disposition Variation is that X has not yet acted in Disposition Variation. An implication of this is that Y is in the smallest group of people of whom it is true that had they all acted differently, V would not have been harmed. Remove Y’s act and keep all of the other acts performed constant. V is not harmed. Yet if Y acts permissibly in Save Variation he also acts permissibly in Disposition Variation. What matters is not whether X has acted, but rather whether X will kill V.12 Parfit’s Principle is defective in restricting pre-­emptive action to the acts that have been performed. Parfit could amend his view to incorporate not only acts that have been done but also acts that would have been done had the relevant person not acted. We can call these ‘potential acts’.

4.2.  Pre-­empting Harm by Natural Causes A second problem is more serious. It is also a result of the fact that Parfit’s Principle refers only to acts. Parfit’s Principle falsely implies that we should starkly discriminate between cases where the overdetermined result would have been caused by an action and those where it would have been caused by another event. The amendment suggested in Section 4.1 will not solve this problem. Consider: Pre-­emptive Killer (Bee Variation):  A highly poisonous bee has stung V.  The ­poison will kill V painfully in a few minutes. Before this poison has any effect X poisons V killing him painlessly.

It is tempting to conclude that if X acts wrongly in Pre-­emptive Killer, X also acts wrongly in Bee Variation. On this view, whether an act pre-­empts harm does not depend on whether the harm pre-­empted would be the product of an act or a natural event. I am not sure whether this view is right. Perhaps the permissibility of pre-­empting harm does depend to some extent on how harm would otherwise have come about. What Parfit needs to vindicate Parfit’s Principle, though, is the implausible proposition that it is always permissible to harm a person when doing so pre-­ empts the same harm arising naturally. Suppose that X in Bee Variation has been 12  Frances Kamm denies this. In Intricate Ethics: Rights, Responsibilities, and Permissible Harm (Oxford: Oxford University Press, 2011), ch. 10, she distinguishes cases where X has a disposition and cases where X makes an offer to Y that if Y kills V, X will refrain from killing both V and V2. The offer makes a difference, she thinks, because it renders X responsible for the death. This counts in favour of the permissibility of Y’s act. I doubt that this difference is sufficiently significant to make Y’s act wrong in Disposition Variation. See also J. McMahan, ‘Responsibility, Permissibility, and Vicarious Agency’, Philosophy and Phenomenological Research 80 (2010): 673–80. Even if Kamm is correct, Parfit’s Principle would not be rescued. Disposition Variation could be revised by introducing an offer.

364  Victor Tadros hired as a hitman to kill V. He will not get paid if the bee’s poison does its work. The intuition that he would act wrongly in killing V for the money is powerful— similarly powerful to the intuition that X acts wrongly in Pre-­emptive Killer. To bolster this conclusion, imagine that a highly poisonous bee has stung V. X does not know whether the bee was released by a wrongdoer or was just flying about. It would be surprising if X had a reason to discover whether it was wrongfully released on the grounds that he could permissibly kill V for the money if he discovers that it has not been.13 And our decisions whether to blame a person for what she has done, and to impose social and legal duties on her, depend on whether she has acted wrongly. It is hard to believe that these responses depend on whether the bee was wrongly released or not. The same verdict is warranted in variations on our other cases. Consider: Simultaneous Killers (Spasm Variation): As Simultaneous Killers but X fires his gun as a result of a muscle spasm rather than an action.

Y, even if he knows that X will fire as a result of a muscle spasm, acts wrongly in killing V.  Perhaps his wrongdoing is not as serious as Y’s wrongdoing in Simultaneous Killers itself, but it is wrongful nevertheless. This has important implications for Parfit’s way of tackling overdetermination problems. Parfit thinks that act-­consequentialism can be saved from implausible implications in overdetermination cases by amending it to focus not on the consequences of single acts but on the consequences of groups of acts. We can see from this subsection that this approach saddles act-­ consequentialism with counter-­intuitive implications in cases where the event pre-­empted is not an act. If Single Act-­Consequentialism yields the wrong results in these cases, revising consequentialism to focus on groups of acts will not correct the problem.

5.  The Mechanisms of Pre-­emption Let us leave these criticisms aside. Even if the second criticism can somehow be met by amending Parfit’s Principle, it faces an even more serious difficulty: it does not discriminate sufficiently between different ways in which one act in the relevant group of acts or events influences other acts. Take some group of acts (v1 and v2), each of which has been performed and each of which would have been sufficient on its own to produce some harm H. Each of these acts is performed and v2, either independently or together with v1, causes H.  Parfit’s Principle implies that if the non-­performance of v1 alone would have resulted in H not 13  For a somewhat similar objection to Kamm’s handling of pre-­emptive killing, see McMahan, ‘Responsibility, Permissibility, and Vicarious Agency’.

Overdetermination and Obligation  365 occurring, v2 is permitted. This is too permissive. v2 is permitted in only a subset of such cases. Before spelling out this criticism, it is worth mentioning that whether Parfit’s Principle implies that Y acts wrongly in Pre-­emptive Killer depends on it being true that Y would have acted had X not acted. Otherwise, Y acts permissibly. Y would not be a member of the smallest group of which it is true that, had these acts not occurred, V would not have been harmed. Hence, Parfit’s Principle implies that Y acts wrongly in Pre-­emptive Killer only if more detail is added to the case—the detail is that X’s act does not cause Y to act. The same thing is true of X. Suppose that X poisons V only because he knows that Y will do so whatever he does. Parfit’s Principle (at least if the revision suggested in Section 4.1 is accepted) implies that X acts permissibly. For then, X is not a member of the smallest group of whom it is true that had they not acted (or potentially acted) in this way the harm would not have occurred. Only Y is a member of that group. Here is one kind of case that illustrates the problem that this idea creates. Suppose that X’s actions cause Y’s actions because X’s actions corrupt Y. The fact that Y has been corrupted by X may make it true that Y’s act would not have been performed were it not for X’s act. Parfit’s Principle implausibly implies that the influence of X’s act on Y makes Y’s act permissible. Consider: Pre-­emptive Killer (Corruption Variation):  X is set on poisoning V. He begins to prepare the poison. Y looks up to X and is very influenced by what X does. Seeing that X is about to poison V, Y prepares his own poison and swiftly poisons V. Had Y not poisoned V, X would have poisoned V in just the same way.

Y is not in the smallest group of whom it is true that, if they had all acted ­differently, V would not have been harmed. If X had not prepared the poison, Y would not have been influenced, and Y would not have poisoned V. If X had not prepared the poison, X would not have poisoned V (if this is in doubt, we can stipulate it). Parfit’s Principle implausibly implies that Y acts permissibly. Even worse, consider: Kicking:  X and Y hate V. V is tough, but X is tougher. X breaks V’s leg. Y is a wimp who would never normally have dreamed of attempting to harm V. Now that V is on the floor, Y takes his opportunity to kick V in the head. Had X not broken V’s leg, Y would not have kicked V. Had Y not kicked V in the head, X would have kicked V in the head in just the same way.

Y is not a member of the smallest group of whom it is true that, if they had all acted differently, other people would not have been harmed. Only X is in that

366  Victor Tadros group. If X had not broken V’s leg, neither X nor Y would have kicked V in the head. Parfit’s Principle implausibly implies that Y acts permissibly in Kicking. The general lesson is that Parfit’s Principle fails to condemn pre-­ emptive harming in the following circumstances. Suppose that X wrongfully harms V. This makes it possible for Y to harm V more. However, had Y not harmed V, X or X’s original act would have harmed V more. Because Y’s act is possible only in virtue of X’s original act, Y’s act would not have occurred but for X’s act. It follows that Y is not a member of the smallest group of whom it is true that had they acted differently V would not have been harmed. X is the only member of that group. Parfit’s Principle implausibly implies that the pre-­emptive action is not wrongful in these circumstances. Parfit might offer the following response. In Pre-­emptive Killer (Corruption Variation) and Kicking, X performs two acts at different times. In Pre-­Emptive Killer (Corruption Variation), these are ‘preparing the poison’ and ‘poisoning V’. In Kicking, they are ‘breaking V’s leg’ and ‘kicking V in the head’. If we exclude the first act from consideration, Parfit’s Principle has intuitive implications. Parfit’s Principle could be revised to yield intuitive results in these cases by making it time-­specific. The second sentence of Parfit’s Principle could be amended as follows: A person is a member of such a group if she is a member of the smallest group of whom it is true that there is some time period at which, had these people all acted differently, the other people would not have been harmed.

This amendment yields the right result in Pre-­ emptive Killer (Corruption Variation) and Kicking. In Pre-­emptive Killer (Corruption Variation) the relevant time period is after X has prepared the poison. In Kicking it is after X has broken V’s leg. In these time periods Y is a member of the smallest group of whom it is true that, had these people all acted differently, V would not have been harmed. This amendment also yields the right verdict in Pre-­ emptive Killer (Save Variation). There is no time period in Save Variation at which Y would harm V had X not acted. However, not only is this amendment hard to justify on principled grounds, it has dubious implications. One problem is that it proves too much. Corruption Variation and Kicking can be amended to make them more like Pre-­emptive Killer (Save Variation). If they are so amended, Y’s act will be permitted. This can be done adding the feature of Pre-­emptive Killer (Save Variation) that makes Y’s action permissible in that case. Amend Pre-­emptive Killer (Corruption Variation) so that Y acts not because he is corrupted by X but rather for the following ­reasons. After seeing X preparing the poison, Y realizes that there is no way that he can save V from being poisoned, either by himself or by X. If he poisons V himself the life of a further person, V2, will be saved (Pre-­emptive Killer (Another Save)).

Overdetermination and Obligation  367 If Y acts permissibly in Pre-­emptive Killer (Save Variation), he acts permissibly in Pre-­emptive Killer (Another Save). A similar amendment to Kicking would make Y’s act permissible in the circumstances of that case. The time-­amended version of Parfit’s Principle yields the wrong result in these cases. Furthermore, the distracting feature of the two acts can be removed from Pre-­ emptive Killer (Corruption Variation). This will not alter the intuitive verdict. Suppose that the facts of Pre-­emptive Killer (Corruption Variation) are similar to those of Pre-­emptive Killer in the following respect. X poisons V. When he does this, Y, who looks up to X, is corrupted. Y then poisons V before X’s poison takes effect. The fact that Y has been corrupted still does not make his act permissible. But now, just as in Pre-­emptive Killer, there is no time at which X refrains from acting but Y still acts. Furthermore, X performs only one act. It is difficult to see how to amend Parfit’s Principle to have intuitive implications in these cases. The verdict in Pre-­emptive Killer (Save Variation) is best explained by the fact that Y acts in order to save V2, or some particular aspect of that fact, rather than the fact that X’s act causes Y’s act. Parfit’s Principle picks out the wrong difference between Pre-­emptive Killer and Pre-­emptive Killer (Save Variation).

6.  Intentions and Permissibility At the end of Section 5, I pointed to a complex fact that makes Y’s act permissible in Pre-­emptive Killer (Save Variation) and Pre-­emptive Killer (Another Save): Y acts in order to save V2. This is a complex fact because it has two components: that V2 will be saved and that Y acts for this reason. We can now contrast two views that make reference to this fact. An objective view is that what makes Y’s act permissible in Pre-­emptive Killer (Save Variation) is the fact that Y’s act saves V2. A view which has subjective and objective components is that what makes Y’s act permissible is that Y’s act saves V2 and that Y is influenced (in some way to be specified) by this fact. I will argue that a purely objective view is mistaken. Only a view that refers to Y’s motivations has intuitive implications. This provides part of a defence of the view that the intentions of the person performing the pre-­emptive act are relevant to the question of permissibility. I explore that idea in the light of arguments mounted by, inter alia, Judith Jarvis Thomson,14 Frances Kamm,15 and T. M. Scanlon16 that a person’s intentions do not typically render her acts wrongful. The views of these scholars differ in many 14  ‘Self-­Defense’, Philosophy and Public Affairs 20 (1991): 283–310, and ‘Physician-­Assisted Suicide: Two Moral Arguments’, Ethics 109 (1999): 497–518. 15  Intricate Ethics: Rights, Responsibility, and Permissible Harm, ch. 5. 16  Moral Dimensions: Meaning, Permissibility, Blame (Cambridge, MA: Harvard University Press, 2008).

368  Victor Tadros respects. But their main idea, which we are focused on here, can be simply explained. The features that are primary in distinguishing between wrongful and permissible acts are independent of the intentions of the person acting. If an act is permitted because of these features, the intentions with which a person acts cannot typically make the act wrong. For example, if it is permissible to divert a trolley away from five people towards one person intending to save the five, knowing that the one will be killed, it is also permissible to divert the trolley away from the five towards the one with the intention of killing the one. The bad intentions of the trolley driver may alter our moral judgement in some way—about the meaning of the act, or the character of the person who performs it—but they do not make the act wrong. This view is consistent with the idea that it is wrong to form certain intentions. But the wrongfulness of doing so derives from the other wrong-­making features of the act intended. If an act, v, is wrong in virtue of some objective feature, f, it is also wrong to form an intention to v in virtue of f. But the intentions of the person performing the act do not typically play a more fundamental role in making the act wrong.

6.1.  Is Pareto Optimal Action Always Permitted? Pre-­emptive Killer (Save Variation) is distinct from Pre-­emptive Killer Corruption Variation and Kicking in virtue of the fact that Y’s action not only fails to harm anyone overall, it also prevents V2 from being harmed. Y benefits no one in Corruption Variation and Kicking. The fact that Y benefits no one in the latter case may be sufficient on its own to make these acts wrongful. Causing harm must be justified, even when no harm overall is caused. As Y’s act does no good to anyone, it cannot be justified. If this is true, pre-­emptive harming is distinct from other acts which do not cause harm overall—those which are not harmful. Scratching my nose makes no difference to anyone overall. Scratching my nose is nevertheless permissible. The contrast between scratching my nose and pre-­emptive harming is that whilst no one is harmed overall in either case, someone is harmed in the latter case but not the former. We should conclude that the fact that no one is harmed overall in Pre-­ emptive Killer (Corruption Variation) and Kicking does not put the acts in these cases on a par with non-­harmful acts.17 This further vindicates the decision to draw a distinction between harming and harming overall defended in Section 2. The question is whether the fact that someone is benefited always justifies ­pre-­emptive harming. Some acts of this kind are Pareto optimal. Even where this 17  It is also worth noting the contrast between killing and letting die in cases of overdetermination, an issue nicely explored in M.  Hanser, ‘Why Are Killing and Letting Die Wrong?’, Philosophy and Public Affairs 24 (1995): 175–201.

Overdetermination and Obligation  369 is so, Parfit implicitly rejects the view that these acts are always permitted. To see this, note that he argues that Y acts wrongly in Pre-­emptive Killer. But if Y can do nothing else to improve V’s lot in Pre-­emptive Killer and his only alternative is to do nothing, Y’s act is Pareto optimal (at least if we don’t count a person’s good intentions as improving the situation). Because of Y’s act, V suffers a painless rather than a painful death. No one else is harmed. Yet Parfit’s view that Y acts wrongly is highly intuitive. Pre-­emptive Killer is different from Pre-­ emptive Killer (Save Variation) in another respect. In the latter case another person is benefited rather than the person who is pre-­emptively harmed. Could that explain the difference between them? It could not. If anything, this feature of Pre-­emptive Killer counts in favour of, rather than against, the permissibility of Y’s act when compared with Y’s act in Pre-­emptive Killer (Save Variation). Here’s why. In Pre-­emptive Killer and Pre-­ emptive Killer (Save Variation), the person who is pre-­ emptively harmed is harmed without her consent. She is not harmed overall, but she is harmed. Harming a person, even pre-­emptively, needs to be justified to her. This is easier to do if she is the person benefited. Better to be harmed pre-­emptively for one’s own sake than for the sake of others. Perhaps, on reflection, some will be inclined to permit Y’s act in Pre-­emptive Killer because although Y harms V, Y pre-­empts a greater harm to V. With Parfit, I find this view hard to believe, or at least I find it hard to believe if the case is specified in the following way. Suppose that Y hates V. Y would ideally like V to suffer a painful death. What he wants even more is to be the one that ends V’s life. This will give Y great satisfaction. Y surely acts wrongly if these things are true. The view that Pareto optimality is insufficient to justify harming a person seems even more difficult to question in cases of simultaneous overdetermination. That is so even if the person who is harmed is also benefited overall. Here’s a slightly modified version of a case I prepared earlier:18 Poisoned Pipe.  Boss offers a reward of £1,000 to anyone who kills Victim. Two henchmen, X and Y, independently find different points in the water pipe leading to Victim’s home. X puts sufficient poison in the pipe to kill Victim. At the same time, Y puts sufficient poison in the pipe to kill Victim. The poisons mix. Each sees what the other is doing. Neither can influence the behaviour of the other. X’s poison alone or Y’s poison alone would have caused Victim to suffer a very slow and painful death. Together, their poison kills Victim swiftly. When X and Y both act, V’s death occurs at exactly the same time as it would have occurred had either X or Y acted alone.19 X and Y each act in order to secure the reward from boss.

18  The Ends of Harm, 159–60. I analyse it more deeply here. 19  This feature of the case is potentially significant because it might wrong for V to kill him earlier than he would have been killed without his consent even when his remaining life will involve only pain and suffering.

370  Victor Tadros Poisoned Pipe is a case of simultaneous overdetermination. Here is an unusual feature of the case that I want to highlight: V is benefited by the fact that X and Y act together when compared with either X or Y acting alone. If X refrained from putting poison in the pipe, Victim would have died a slow and painful death as a result of Y’s act. If Y had refrained from putting poison in the pipe, Victim would have suffered an identical fate. Hence, V is benefited overall by X’s act, when that act is considered alone. The same thing is true of Y’s act. Together, these acts make V much worse off overall than he would have been had neither of them acted. The view that Pareto optimal action is always permitted implies that X and Y act permissibly in Poisoned Pipe when their only alternative is to do nothing. This view is powerfully counter-­intuitive.

6.2.  Liability and Overdetermination I have claimed that it is intuitive that X and Y each act wrongly in Poisoned Pipe if they each act for the money. This view is further supported by intuitive judgements about when X and Y are liable to various different kinds of harming because of their acts.20 I consider three kinds of liability: liability to preventive harm, liability to be forced to pay compensation, and liability to be punished. The best justification of these liabilities is, of course, highly contested and I will propose no particular theory of them. I will simply outline a set of highly intuitive views in this case. I will suppose, though, that liability to preventive harm, to pay compensation, and to be punished is normally warranted only on condition that the liable person has acted wrongly. This admittedly may not true in every case— it is intuitively true in the cases that we are considering. First consider preventive harm. X is liable to be harmed to avert the threat that he, together with Y, poses. Perhaps it might be argued that if it is possible to avert the threat by harming X, X’s act will necessarily make V worse off overall. But this is not necessarily so. The possibility of preventing the harm that Y would otherwise cause may emerge only as a result of X’s act. To see this, consider: Poisoned Pipe (Preventive Variation). As Poisoned Pipe but when X puts poison in the pipe, unknown to him, he alerts Cop to the existence of the poison. This is the only way of alerting Cop. Cop will then be able to prevent V from drinking the water, but only by blowing up the sewer, harming both X and Y. There is no other way of averting the threat to V. 20  Considerations of liability also help to demonstrate that intermediate positions that condemn a person only for forming and acting on illicit intentions rather than for the acts that they perform are problematic. A view of this kind is developed in A.  Walen, ‘The Doctrine of Illicit Intentions’, Philosophy and Public Affairs 34 (2006): 39–67.

Overdetermination and Obligation  371 X’s act renders V better off overall than he would have been had he not put poison in the pipe. This is so either if Cop refrains from acting (V will then die a swift rather than a painful death), or if Cop saves V’s life harming X and Y. Nevertheless, if X and Y act for the money, it is permissible for Cop to harm X and Y to avert the threat. It may well even be permissible for Cop to kill both X and Y to serve this end. That is because, through his wrongful act, X makes himself liable to preventive harm. Second, consider liability to pay compensation.21 Intuitively, X owes compensation to V, and the same is true of Y. This is so even though, taken independently, X has benefited V overall, and so has Y. Here is the intuitive way to calculate compensation. The magnitude of compensation owed overall is calculated by comparing V’s current condition (speedy death) with the condition that V would have been in had neither X nor Y have acted. X and Y are jointly and severally liable to pay this amount of compensation into V’s estate. If X and Y can both be discovered each must pay half the amount of compensation owed. If only one can be discovered, he must pay full compensation. It is very difficult to believe that these results are unwarranted because X’s action, taken independently, does not harm V overall. The alternative is that no compensation is owed by either X or Y for V’s death. That is hard to believe. Finally, if X and Y act for the money it is permissible to punish them both for murdering V.  The fact that V is not harmed overall by X’s conduct does not prevent X from being liable to punishment. This view is even more clearly intuitive in more standard pre-­emption cases. Consider Pre-­emptive Hitman.  Boss hires Hitman 1 to kill V. Hitman 1 succeeds. Had he not succeeded, Boss would have hired Hitman 2 to do the job. Hitman 2 employs an especially cruel method of killing and he would certainly have succeeded in doing so.

If Hitman 1 acts only for the money and not to benefit V, it is permissible to punish Hitman 1 for murder. Any other view is very hard to believe.22

6.3.  What if X Is Well Intentioned? In Sections 6.1 and 6.2, I argued that X and Y each act wrongly in Poisoned Pipe. One feature of Poisoned Pipe is that X and Y each kill V for the money. This is not 21  Those who reject a counterfactual account of harm that I briefly defended in Section 2 often overlook how difficult it is for other accounts to provide a plausible approach to the duty to compensate. 22  This need not be for reasons of desert. I believe that liability to preventive, compensatory, and punitive harm are more closely related than most punishment theorists. See, further, The Ends of Harm.

372  Victor Tadros the only fact that might motivate X, however. Suppose that X acts not for the money but only in order to benefit V (Poisoned Pipe (Euthanasia Variation)). Does his good intention make his act permissible? Suppose that Thomson, Kamm, and Scanlon are right that intentions are irrelevant to permissibility. Given the verdict reached in Sections 6.1 and  6.2 about Poisoned Pipe, we should conclude that X acts wrongly in Poisoned Pipe (Euthanasia Variation). The only difference between Poisoned Pipe and Poisoned Pipe (Euthanasia Variation) is X’s intentions. If a person’s intentions cannot make a permitted act wrongful, either X must act wrongly in both cases or X must act permissibly in both cases. I have provided powerful support for the claim that X acts wrongly in Poisoned Pipe. This view, when combined with the view that intentions are irrelevant to permissibility, implies that X acts wrongly in Poisoned Pipe (Euthanasia Variation), which is hard to believe. X has two choices: to poison V or not to poison V. If he poisons V, X prevents V suffering excruciating pain. Otherwise, his act makes no difference to V overall. He is surely permitted to avert the pain that V would suffer. Administering the poison, where it is done in order to relieve V’s pain, is morally equivalent to giving V other kinds of pain relief, which is not wrong. Rejecting the significance of intentions thus gives rise to the following dilemma: either the verdict that X and Y each act wrongly in Poisoned Pipe must be rejected or the verdict that X acts permissibly in Poisoned Pipe (Euthanasia Variation) must be rejected. Neither option is attractive. It may have seemed that Parfit’s Principle provides a way out of this dilemma. According to Parfit’s Principle, X’s act, but not Y’s act, is permitted in Poisoned Pipe (Euthanasia Variation) because the absence of Y’s act alone would avert the harm to V. That is not true of X’s act. But the reasons that I offered to reject Parfit’s Principle can be extended to this scenario. To see this, suppose that X acts in Poisoned Pipe only because Y’s potential act corrupts him. Parfit’s Principle yields the unintuitive verdict that X acts permissibly. The view that X’s intentions distinguish between these cases is far more plausible. If X acts only in order to get the money from Boss he acts wrongly. If X acts only in order to benefit V he acts permissibly. Hence, X’s intentions determine whether he acts permissibly. The judgement that X acts permissibly in Euthanasia is further supported by intuitive verdicts about liability. Recall that X and Y are each liable to preventive harm, to pay compensation, and to be punished in Poisoned Pipe. Things are different in Poisoned Pipe (Euthanasia Variation). Consider liability to preventive harm first by considering: Poisoned Pipe (Euthanasia and Prevention): As Poisoned Pipe (Euthanasia Variation) except that when X puts poison in the pipe, he alerts Cop who can

Overdetermination and Obligation  373 then blow up the pipe saving V. Cop is only alerted by X’s action and X knows that he will be alerted. If Cop blows up the pipe, X will be harmed.

X is not liable to be harmed by Cop to save V. If the harm done to X will be significant, Cop may not harm X. To see why, consider X’s situation prior to putting poison in the pipe. If he puts poison in the pipe in order to benefit V, he has ­benefited V at some small cost to himself. Furthermore, the opportunity for Cop to save V arises only because of X’s act. So, in acting in this way, X has not deprived V of an opportunity that he would otherwise have had to be saved. Helping V in this way cannot make X liable to be killed to save V. Only Y is liable to preventive harm to rescue V. Now consider compensation. It is powerfully intuitive that X is not liable to pay compensation to V’s estate for V’s death. X has acted only in order to benefit V. Y owes any compensation that is owed to V’s estate for V’s death. The same conclusion should be drawn about punishment. It would be wrong to punish X, but not wrong to punish Y, for killing V in Poisoned Pipe (Euthanasia Variation). Intentions, we should conclude, are relevant to a person’s liabilities in cases of overdetermination. Given the plausible assumption that liability to preventive harm, to pay compensation, and to be punished at least typically track wrongdoing, there is good reason to believe that intentions are relevant to permissibility. Of course, this discussion alone is not sufficient to demonstrate that Thomson, Kamm, and Scanlon are mistaken in believing that intentions are generally irrelevant to permissibility. Their views might be powerfully supported by other considerations, considerations powerful enough to drive us to impale ourselves on one or other of the horns of the dilemma I have been exploring. Alternatively, overdetermination cases may provide an exception to the general rule that intentions are irrelevant to permissibility. Scanlon, for one, accepts that intentions are relevant to permissibility in some cases such as wrongful discrimination or wrongful threats.23 It is difficult to see how to extend his reasoning about those cases to cases of overdetermination. It is also more generally difficult to see why cases of overdetermination should be exceptional. If intentions are relevant to permissibility in this case, why should they not be relevant in other cases where a person does the ‘right thing’ for the ‘wrong reason’? This discussion is at least an important part of a more general case to be made for the traditional view that the intentions with which a person acts determine whether their acts are permitted.24 23  Moral Dimensions, 69–87. 24 For further defence of the traditional view against Thomson, Kamm, and Scanlon, see e.g. W.  Fitzpatrick, ‘Acts, Intentions, and Moral Permissibility: In Defence of the Doctrine of Double Effect’, Analysis 63 (2003): 317–21; J.  McMahan, ‘Intention, Permissibility, Terrorism, and War’, Philosophical Perspectives 23 (2009): 345–72; Tadros, The Ends of Harm, ch. 7; Wrongs and Crimes, ch. 16; To Do, To Die, To Reason Why (Oxford: Oxford University Press, 2020), ch. 6.

374  Victor Tadros

7.  Complex Motivations So far, we have considered variations on Poisoned Pipe where X acts on the bad intention to kill V for the money and where X acts on the good intention to relieve V from pain. This leaves open our evaluation of cases where X’s motivations are more complex. There is a broad range of issues and I will mention only three.

7.1.  Acting as an Agent It may seem that it is always wrong to kill a person for the money, and that explains what is wrong in Poisoned Pipe. Money, it might be argued, is an ‘irrelevant utility’ (to use Kamm’s phrase)25 in matters of life and death. If this is understood as the claim that it is always wrong to kill a person for money, this claim is false. Consider: Poisoned Pipe (Hire Variation):  V’s mother knows that Y will poison V and that V will die a slow and painful death if nothing is done. The only thing that can be done to benefit V is to ensure that additional poison is added to the pipe. Only X can do this. However, doing it will be burdensome to X. X is unwilling to help unless V’s mother pays him.

X is permitted to act for the money (or at least on condition that he gets the money). One way to contrast this case with Poisoned Pipe is that in Poisoned Pipe it is the harm that X does to V that is the means by which he will secure money. In Hire Variation, in contrast, it is the benefit that X provides to V that is the means by which he will secure money. Perhaps it might be argued that Hire Variation calls into question the idea that it is X’s intentions that determine whether his actions are permissible in cases of this kind. It might be thought that the wrong-­making feature of Poisoned Pipe is that Boss’s bad intentions are furthered if X follows his instructions. This is false for two reasons. First, what distinguishes X and Y in Hire Variation is that X is acting as V’s mother’s agent whereas Y acts as Boss’s agent. Whether a person is acting as an agent of another person itself depends on that person’s intentions. Secondly, X would act wrongly in Poisoned Pipe even if he were not acting as an agent for Boss, but rather in order to secure his own bad ends. Exclude Boss from Poisoned Pipe. Suppose that X and Y each act only out of hatred of V. Their hatred is so strong that not only does each of them want V dead, each of them wants to cause V’s death. Each acts wrongly. It is wrong for X to act on his own bad intention. It is wrong for X to act as Boss’s agent. It is permissible for X to act 25 From Morality, Mortality, i. Death and Whom to Save from It (Oxford: Oxford University Press, 1993), ch. 8.

Overdetermination and Obligation  375 as V’s mother’s agent. It is permissible for X to act on his own good intentions. I doubt that a similarly plausible set of verdicts can be yielded without referring to X’s intentions.

7.2.  Intending to Intend A second kind of complex case involves people who can further the plans that they fundamentally endorse only by forming intentions that are at odds with those plans.26 There are two variations of this idea to consider. In one variation, X has a good intention that he finds difficult to execute. Poisoning is just not in his nature. For this reason, he causes himself to form a bad intention. He undergoes hypnosis to make him hate V. This will lead him to form the intention to poison V because he wants him dead. Even though X will act with the intention to poison V because he wants him dead, X’s overall course of conduct is, in this case, permitted. He is permitted to cause himself to act on a bad intention. In a second variation X has a bad intention, but he will secure his purposes more effectively if he can cause himself to form a good intention. Suppose that X wants to poison V because he hates V. He knows that if he acts on this intention, he is more likely to be held criminally liable for killing V. He undergoes hypnosis to make himself care about V. This causes him to act with the intention to relieve V’s pain. Although X acts with the good intention of relieving V’s pain, his overall course of conduct is nevertheless wrongful. We can conclude that it is not simply the direct intention with which the person acts that determines whether he acts permissibly. The plans that are more fundamentally connected with the person determine whether the person acts permissibly or wrongly.

7.3.  Conditional Intentions A third more complex motivational state involves agents who are willing to harm others only on condition that the harm that they cause is overdetermined. Kamm helpfully distinguishes between a person acting for a reason and a person acting only on condition that some fact obtains. Kamm’s well-­known Party Case nicely illustrates this distinction.27 If I have a party it will create a mess. I am willing to have the party only if others will help me to clear up the mess. Suppose that others will help me and I have the party. I do not have the party for the reason that others 26  I owe my discussion of this problem to the Rutgers Ethics Discussion Group, especially to Holly Smith and Larry Temkin. 27  Intricate Ethics: Rights, Responsibilities, and Permissible Harm (Oxford: Oxford University Press, 2007), 95–6.

376  Victor Tadros will help clear up the mess. The reason why I have the party is that it will be fun. That others will help to clear up the mess is a condition on which I will have a party. Now return to Poisoned Pipe. Suppose that X acts only for the money in Poisoned Pipe. He is willing to do this only on condition that Y also acts. As Kamm’s distinction makes clear, Y’s action need not provide the reason for X’s action. It might provide a condition that he treats as necessary to act for another reason: say to get Boss’s money. Kamm, it will be remembered, thinks that if a person is permitted to do some act with good intentions, she is permitted to do the same act with bad intentions. She also thinks that a person with bad intentions acts permissibly even if that person does not treat the properties of the action that render it permissible as necessary conditions for her to act. If the objective properties of the action make the action permissible, that act may be performed with any intention and regardless of whether she would be willing to act were the relevant conditions not present.28 In contrast with Kamm, I have argued that the intentions of the person acting discriminate between permissible and wrongful action in overdetermination cases. Good intentions are sufficient to render X’s act in Euthanasia Variation permissible. Bad intentions, acted on unconditionally, are sufficient to render X’s act in Poisoned Pipe wrong. This is so even though these cases are identical with respect to their objective features. What is the proper verdict where a person acts on bad intentions but acts on condition that the good-­making property (in this case, saving V from pain) is present? Is conditioning one’s action on the features of the action that render it permissible sufficient to render the actions of a person who acts with bad intentions permissible? I cannot develop a full answer to this question here. Here is a brief suggestion: different explanations for a person’s decision to condition her action on the relevant facts might warrant different verdicts. Return to the facts of Poisoned Pipe. Suppose that X is willing to act only on condition that he benefits V overall. Here is an explanation why X conditions his action in this way that puts X in a worse light. X likes money but is unwilling to take too great a risk for it. If Y acts as well, X will be able to escape criminal liability by pretending that he was acting only to relieve V’s pain and suffering. X would not act were this not true. In this case, X acts wrongly. The fact that he is willing to act only on condition that he is less likely to be held criminally liable for doing so does not make his act permissible. This conclusion is bolstered by the following consideration. It is implausible that X and Y each act permissibly if they both condition their actions in this way. Here is an explanation that puts X’s act in a relatively good light. X likes money but very much dislikes harm. He is willing to poison V only on condition not only that he will not cause V to suffer any harm beyond that which he was already 28  Her view about this is developed in Ethics for Enemies: Terror, Torture, and War (Oxford: Oxford University Press, 2011), esp. in ch. 3.

Overdetermination and Obligation  377 doomed to suffer, but also that V will be benefited by his action overall. This is not the reason for which he acts—he acts for the money—but it is a necessary condition for him to act. This variation is more difficult to evaluate. I doubt that X’s act is necessarily permissible in this case. Here is a troubling implication of the view that X is permitted to act if he satisfies this condition. I bring this implication to light without having the space fully to evaluate it. Suppose that it is permissible for X to act if he acts only on condition that Y also acts. The following scenario is then possible: X and Y will together kill V even though each of them acts rationally (in a sense to be specified in a moment) and each is fully motivated to conform to what morality requires. This implication seems perverse. To see this implication, consider: Conditional Poison: As Poisoned Pipe except for the following. X is willing to act only if Y acts. Y is willing to act only if X acts. They know this about each other and neither can influence the action of the other. X and Y would each prefer a world in which V is poisoned by both X and Y and they are rewarded to a world in which V survives and neither X nor Y is rewarded. Each is willing to act only on condition that the other acts in virtue of the fact that this ensures that he does not harm V overall.

What is it rational for X to do? Let us suppose, for the sake of argument, that what it is rational for a person to do is determined only by her preferences rather than by moral considerations. I don’t claim that this is the best understanding of rationality; I stipulate it to clarify the exposition. Suppose also that X knows that Y will act rationally. He also knows that Y knows that X will act rationally and vice versa. Each also knows the preferences of the other and each knows that the other knows his preferences. It would be rational for X to poison the pipe. It is rational for him to do this on the assumption that he ought to make that Y, like X, will recognize that both get their preferred outcome if each poisons the pipe.Furthermore, as X cannot influence Y, he should also assume that, as Y will reason in this way, V is doomed. It follows that if X and Y act rationally in the sense outlined above, V will be poisoned by X and Y. If X and Y are each permitted to act on condition that each does no harm overall, each is permitted to do what it is rational for him to do. Hence, V will be poisoned even if X and Y are unwilling to act wrongly. This perverse implication counts against the view that it is permissible for X to act on condition that he does no harm. It does so on the basis of the following idea. It counts against a person being permitted to v if that permission leads groups who are badly motivated but who are rational actors who wish to conform to their moral duties to cause very serious harm to others. It is worth noting that common-­sense morality does often lead to bad results due to lack of information about what others will do. In section 36 of Reasons and

378  Victor Tadros Persons Parfit showed that the common-­sense moral view that provides people with special duties to rescue their own children rather than the children of others leads to bad results overall. To use his example, suppose that A and B each have two children who are drowning. A can rescue only one of his children but two of B’s. B can rescue only one of his children and two of A’s. A principle permitting or requiring the rescue of one’s own children rather than a greater number of the children of other people has the implication that not all of the children will be rescued in this case. Parfit thought that common-­sense morality is self-­defeating for this reason. But this implication of common-­sense morality is less troubling than the implication outlined in Conditional Poison for the reason that each person is well motivated but lacks information about what the other person will do if he rescues the other’s children. If A could rely on B rescuing A’s children were A to rescue B’s children, A would be required to rescue B’s children. The bad results in Parfit’s example can be attributed to the fact that A and B cannot rely on each other. The fact that complying with a duty leads well-­motivated actors to bring about bad results in circumstances where people cannot rely on each other is somewhat troubling. It is much more troubling that a group of people who are, in the sense just stipulated, rational, but who are badly motivated can permissibly kill if each acts individually but knowing how the others will act. If X and Y are permitted to act in Conditional Poison, we imply that X and Y can each treat poisoning as an option that may be taken if the other acts. But there is a decisive reason against X and Y treating this action as an option: their attitudes will result in V being poisoned. This seems at least to count against the idea that X is permitted to act if the condition that he does not harm V overall is satisfied. No similar problem arises if X and Y prefer the state of affairs where V is not poisoned to the state of affairs where V is poisoned by both X and Y. In that case it would be rational for X and Y not to poison the pipe. What seems to matter, in that case, is that X and Y are suitably constrained in their motivations by concern for V. If the consideration that I have been pointing to is decisive, we should at least qualify the idea that X is permitted to act if he acts on condition that he does not harm V overall. He is permitted to act only in a subset of these cases: where, out of concern for V, he would prefer the world in which V is not poisoned and X does not benefit to the world where V is poisoned and X benefits.

8.  Conclusion Overdetermination is interesting for a wide range of reasons. It has advanced our understanding of, inter alia, the metaphysics of causation, free will, and harm. I have attempted to make some progress in assessing its implications for normative theories.

Overdetermination and Obligation  379 Of course, a single issue does not decisively demonstrate that any general moral view is right. It is, at most, part of a case for a moral view. Any well-­worked-­ out moral view is likely to have some counter-­intuitive implications. Perhaps agent-­ neutral act-­ consequentialists and those who think that intentions are irrelevant to permissibility will accept the counter-­intuitive implication of their preferred view in the cases I have been discussing. Or perhaps they can find other ways of explaining the intuitive verdicts that I have outlined without giving up significant commitments. To my mind, though, referring to the motivational states of agents is essential to yield plausible verdicts in cases of overdetermination. The view to be preferred, at least on these grounds, takes the significance of intentions over and above the quality of outcomes seriously. There is a great deal more work to do in defending this view—especially in developing a deeper justification of it that shows not only that it is intuitive but also that it can be given the kind of deep explanation that we should wish for in moral theory. That project is left for another day.

Bibliography Drier, J., ‘In Defense of Consequentialising’, in M. Timmons (ed.), Oxford Studies in Normative Ethics, vol. i (Oxford: Oxford University Press, 2011): 97–119. Fitzpatrick, W. ‘Acts, Intentions, and Moral Permissibility: In Defence of the Doctrine of Double Effect’, Analysis 63 (2003): 317–21. Hanser, M., ‘Why Are Killing and Letting Die Wrong?’, Philosophy and Public Affairs 24 (1995): 175–201. Hanser, M., ‘The Metaphysics of Harm’, Philosophy and Phenomenological Research 67 (2008): 421–50. Hitchcock, C., ‘Counterfactual Availability and Causal Judgements’, in C.  Hoerl, T. McCormack, and S. R. Beck (eds.), Understanding Counterfactuals, Understanding Causation: Issues in Philosophy and Psychology (Oxford: Oxford University Press, 2011): 171–185. Kagan, S., Normative Ethics (Boulder, CO: Westview, 1998). Kamm, F., Morality, Mortality, i. Death and Whom to Save from It (Oxford: Oxford University Press, 1993). Kamm, F., Intricate Ethics: Rights, Responsibilities, and Permissible Harm (Oxford: Oxford University Press, 2007). Kamm, F., Ethics for Enemies: Terror, Torture, and War (Oxford: Oxford University Press, 2011). Lewis, D., ‘Causation’, Journal of Philosophy 70 (1973): 556–67. McMahan, J., The Ethics of Killing: Problems at the Margins of Life (Oxford: Oxford University Press, 2002).

380  Victor Tadros McMahan, J., ‘Intention, Permissibility, Terrorism, and War’, Philosophical Perspectives 23 (2009): 345–72. McMahan, J., ‘Humanitarian Intervention, Consent, and Proportionality’, in N. Ann Davis, R. Keshen, and J. McMahan (eds.), Ethics and Humanity: Themes from the Philosophy of Jonathan Glover (New York: Oxford University Press, 2010): 44–72. McMahan, J., ‘Responsibility, Permissibility, and Vicarious Agency’, Philosophy and Phenomenological Research 80 (2010): 673–80. Menzies, P., ‘The Role of Counterfactual Dependence in Causal Judgements’, in C.  Hoerl, T.  McCormack, and S.  R.  Beck (eds.), Understanding Counterfactuals, Understanding Causation: Issues in Philosophy and Psychology (Oxford: Oxford University Press, 2011): 186–207. Parfit, D., Reasons and Persons (Oxford: Oxford University Press, 1984). Scanlon, T. M., Moral Dimensions: Meaning, Permissibility, Blame (Cambridge, MA: Harvard University Press, 2008). Schiffrin, S., ‘Wrongful Life, Procreative Responsibility, and the Significance of Harm’, Legal Theory 5 (1999): 117–48. Tadros V., The Ends of Harm: The Moral Foundations of Criminal Law (Oxford: Oxford University Press, 2011). Tadros V., Wrongs and Crimes (Oxford: Oxford University Press, 2016). Tadros V., To Do, To Die, To Reason Why (Oxford: Oxford University Press, 2020). Thomson, J. J., ‘Self-Defense’, Philosophy and Public Affairs 20 (1991): 283–310. Thomson, J. J., ‘Physician-Assisted Suicide: Two Moral Arguments’ Ethics 109 (1999): 497–518. Velleman, J. D., ‘Persons in Prospect’, Philosophy and Public Affairs 36 (2008): 221–44. Walen, A., ‘The Doctrine of Illicit Intentions’, Philosophy and Public Affairs 34 (2006): 39–67. Williams, B., ‘A Critique of Utilitarianism’, in J.  J.  C.  Smart and B.  Williams Utilitarianism: For and Against (Cambridge: Cambridge University Press, 1973): 77–150. Woodward, J. ‘Psychological Studies of Causal and Counterfactual Reasoning’, in C.  Hoerl, T.  McCormack, and S.  R.  Beck (eds.), Understanding Counterfactuals, Understanding Causation: Issues in Philosophy and Psychology (Oxford: Oxford University Press, 2011): 16–53.

17

What Is Harming? Molly Gardner

1.  Introduction In some cases, it is easy to tell whether one individual has harmed another. For example, when Smith breaks Jones’s knees with a baseball bat, Smith harms Jones. In other cases, whether an act is harmful can be more difficult to discern. Can we harm people with speech? Can we harm someone who is dead? What about someone who hasn’t yet been born? A theory of harming can help us articulate and justify answers to these and other questions. A complete theory of harming must have both a substantive component and a formal component. The substantive component of the theory tells us what it is about you or your life that I interfere with when I harm you. Is a harm to you just any impediment to your well-­ being, or does harm occur only when the impediment is to some interest of yours that has special significance? Can you suffer a harm even while your well-­being remains unaffected? These are questions about what Victor Tadros (2014) calls the ‘currency’ of harm, and the issue of whether we can harm others with offensive or hateful speech turns upon what we take the currency of harm to be. Whether there are posthumous harms is also partly a question about currency. In addition to the currency component, a complete theory of harming will also have a formal component, which Tadros refers to as the ‘measure’ of harm. This component of the theory tells us how a particular harm might be related to an action or event. In harming you, do I make you worse off (with respect to whatever the currency of harm might be) than you were, worse off than you otherwise would have been, or worse off than you should have been? Must my harmful action cause the harm you suffer, or is it enough that the harm you suffer depends counterfactually on the action I perform? What we say about the measure of harm can bear on some important questions in ethics, such as the question of whether it is worse to do harm than to allow it; the question of whether we can harm future generations; and, as before, the question of whether there are posthumous harms. The literature on harm has not always been careful to distinguish between the currency and the measure of harm. This is unfortunate, for when the distinction Molly Gardner, What Is Harming? In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0018

382  Molly Gardner is not drawn clearly, it is very easy for work on the currency of harm to reinforce unjustified assumptions about the measure of harm, or vice versa. In what follows I hope to root out these unjustified assumptions and bridge the gap between work on the currency and measure of harm. In the course of doing so, I will also argue for substantive accounts of both. Specifically, I will argue that well-­being is the currency of harm and that the measure of harm is best captured by a causal account of harming.

2.  The Currency of Harm Work on the currency of harm is interdisciplinary, appearing in law journals as well as philosophy journals. Consequently, substantive accounts of harm are sometimes couched in legal terms that obscure the relationship between the concept of harm and other important philosophical concepts, such as the concepts of well-being, rights, desires, and interests. This kind of terminological disconnect is unfortunate. If the relationship between harm and other philosophical concepts were made clearer, harm theorists could more easily avail themselves of innovations in, for example, the well-­being literature; likewise, well-­being theorists could more easily avail themselves of innovations in the harm literature. In this section, I aim to clarify the relationship between harm and well-­being and then argue that the currency of harm is well-­being. But since harm is so frequently analyzed in terms of interests, I will begin with a brief survey of interest-­ based views. One prominent line of thought begins with the notion that harm is a setback to an individual’s interests. Many who take this starting point also hold that not just any setback to an individual’s interests qualifies as a harm. Instead, there is a privileged class of interests that matter more than others, and only setbacks to these important interests count as harms. For example, John Kleinig (1978: 31) argues that the interests that must be set back are welfare interests, or interests that are ‘indispensable to the pursuit and fulfilment of characteristically human interests, whatever those interests might be’. Others argue that the relevant class of  interests is delineated by an individual’s moral claims. Thus, Joel Feinberg (1984: 36) argues that according to the sense of ‘harm’ in which he is interested, ‘only setbacks of interests that are wrongs, and wrongs that are setbacks to interest, are to count as harms’. Stephen Perry (2003: 1306) affirms both a moral criterion and a ‘core’ interest criterion, noting that in cases where a person’s core interests are unimpeded, ‘a right can be violated, and an interest set back, without any harm occurring’. Philosophers who write about the currency of harm often acknowledge a relationship between the concept of someone’s interests and the concept of her well-being, but they characterize this relationship in different ways. Feinberg

What Is Harming?  383 (1992: 4) takes interests to be ‘distinguishable components of a person’s good or well-­being’, and he holds that ‘interests can be summed up or integrated into one emergent personal interest’. Perry (2003:1304) agrees that interests are related to well-­being, but he rejects the second of Feinberg’s claims, arguing that ‘we have no good reason to think that the myriad array of interests that are subject to harm and benefit . . . can all be reduced to a single underlying interest of any kind’. Some theorists appear to equate interests with the satisfaction of a person’s preferences or desires.1 Thus, if and only if a preference- or desire-­satisfaction theory of well-­ being is true, these theorists appear to be committed to the claim that the currency of harm is well-­being. Finally, Simester and von Hirsch (2011: 36) take interests to be distinct from well-­being, writing that ‘Our interests merely serve our well-­ being. . . . They do not determine it’. We can bring more clarity to this discussion by temporarily bypassing the concept of an interest altogether and focusing directly on the relationship between harm and well-­being. Let us distinguish between four possible accounts of this relationship. According to the sufficiency view, necessarily, any impediment to someone’s well-­being qualifies as a harm. (Here I am using ‘impediment’ as a placeholder; in Section 3, I will consider the precise way in which the currency of harm needs to be measured.) According to the non-­sufficiency view, some impediments to well-­being are not harms. Both the sufficiency view and the non-­ sufficiency view leave open the question of whether an impediment to one’s well-­ being is a necessary condition of harm, so we can round out the discussion with two additional views. According to the necessity view, all harms are impediments to well-­being. And according to the non-necessity view, some harms are not impediments to well-­being. Many people who theorize about harm in terms of ‘well-­being’ rather than ‘interests’ take the conjunction of the sufficiency view and the necessity view to be the default: they assume that all and only impediments to well-­being are harms.2 Such a claim is prima facie plausible. Common sense affirms that your well-­being is a matter of how well things are going for you and that when things go badly for you, you are harmed. Indeed, I agree with these theorists that we should accept the conjunction of the sufficiency view and the necessity view unless we

1  For example, George Pitcher (1984: 184) writes, ‘An event or state of affairs is a misfortune for someone (or harms someone) when it is contrary to one or more of his more important desires or interests’. Similarly, in a passage about harm and compensation, Nozick writes, ‘In the terminology of economists, something compensates X for Y’s act if receiving it leaves X on at least as high an indifference curve as he would have been on, without it, had Y not so acted’ (Nozick 1974: 57). As Victor Tadros (2014) observes, the latter passage suggests that Nozick takes harm to be something akin to preference frustration. 2  Some of the many philosophers who start from this assumption include Derek Parfit (1987), Alastair Norcross (2005), Ben Bradley (2009), Matthew Hanser (2009), Robert Northcott (2015), Thomas Bontly (2016), myself (Gardner 2015), Michael Rabenberg (2015), and David Boonin (2019).

384  Molly Gardner encounter sufficiently compelling arguments for either the non-­sufficiency view, the ­non-­necessity view, or both. One argument for the non-­sufficiency view appeals to the moral significance of harm. Some commonsense moral principles suggest that there is a strong reason against harming, and a prohibition on harm is built into many of our laws and codes of professional ethics. In Seana Shiffrin’s words, ‘harm commands special attention’ (Shiffrin 2012: 366). At the same time, there appear to be cases in which the reason against impeding a person’s well-­being are weak at best. Possibly for that reason, John Kleinig holds that it is not a harm to a millionaire to be shortchanged by his paper boy.3 Likewise, Seana Shiffrin holds that it is not a harm to a billionaire to accidentally lose a thousand dollars.4 Although Shiffrin and Kleinig do not explicitly endorse the non-­sufficiency view, their examples could be taken to support the following argument for it: 1.  There is always a strong moral reason against causing a harm to someone. 2. There are cases in which there is at most a weak reason against causing an impediment to someone’s well-­being. 3.  Therefore, some impediments to well-­being are not harms.

Let us call this the ‘Moral Reason Argument’. The Moral Reason Argument is not especially persuasive because premise 1 is overstated. While commonsense morality does suggest that there is always a reason against causing a harm to someone, it’s not clear that this reason must always be strong. After all, commonsense morality also tells us (a) that the reason against causing a harm is commensurate with the gravity of the harm and (b) it is possible to suffer only a minor harm. Together, (a) and (b) imply that premise 1 is false: it is possible for the reason against harming to be weak.5 In addition to the Moral Reason Argument, there is at least one other argument for the non-­sufficiency view. This second argument takes its cue from John Stuart Mill’s famous Harm Principle. According to Mill, ‘the only purpose for which 3  Referring to the case in which the millionaire is shortchanged, Kleinig observes that ‘the loss is so trivial that calling it a harm is like using a sledgehammer to drive in a pin’ (Kleinig 1978: 29). 4  She frames this as an objection to various accounts of harm, writing that such accounts ‘overidentify conditions as harm that do not merit the label. . . . A billionaire’s accidental loss of a thousand dollars will be said to be a harm to him, assuming he has a stake in his stockpile, as many billionaires do’ (Shiffrin 2012: 371). 5  Of course, one might also challenge the suggestion that the millionaire and billionaire examples support premise 2. The millionaire’s well-­being might not actually be affected by the paper boy’s short-­ changing him, and it is possible that the billionaire’s well-­being is not affected by the loss of a thousand dollars. These examples would then do nothing to establish the claim that in some cases, there is at most a weak reason against impeding someone’s well-­being. However, it would be easy enough motivate premise 2 with other examples. We can imagine a case in which the well-­being of an otherwise extremely well-­off person would drop slightly while he ate an under-­ripe banana. Plausibly, there is only a weak moral reason against offering him the banana. In any case, I believe that the main weakness of the moral reason argument lies in premise 1, rather than premise 2.

What Is Harming?  385 power can be rightfully exercised over any member of a civilized community, against his will, is to prevent harm to others’ (Mill 2002: 11). It is widely supposed that such a principle limits state interference in the affairs of individuals, especially when it comes to issues like speech and sex. But in some cases, hateful, threatening, or offensive speech can impede the well-­being of those who hear it, and even sex between consenting adults can affect the well-­being of third parties.6 We thus arrive at the following argument: 1. Neither speech nor sex between consenting adults can cause harm to third parties. 2. Both speech and sex between consenting adults do sometimes cause impediments to the well-­being of third parties. 3.  Therefore, some impediments to well-­being are not harms.

Because the first premise appears to be motivated, both by the Harm Principle and by the liberal assumption that speech and consensual sex should not be punished by the state, I will call this the ‘Liberal Argument’. The problem with the Liberal Argument is once again with the first premise. For one thing, the motivation behind the first premise is weaker than it might originally have appeared. According to Mill’s formulation, the Harm Principle establishes harm to others as a necessary condition for the justified prohibition of various activities, not a sufficient condition. So even if Mill’s Harm Principle is true, it does not entail that if hate speech and consensual sex should be permitted, then they must not cause harm to others. Some theorists have defended alternative formulations of the Harm Principle that might, if true, provide additional support for premise 1. Joel Feinberg’s (1984) formulation of the Harm Principle implies that harm to others is a positive reason for state coercion. If this positive reason for coercion has no defeaters, and if there is no reason for the state to prohibit speech and consensual sex, then it does, indeed, follow from Feinberg’s view that speech and consensual sex are harmless to others. However, it is not clear that we should accept both conjuncts of the antecedent. Perhaps it is true both that there is a prima facie reason for the state to prohibit hate speech or sex between consenting adults and that other moral considerations function as defeaters. The first premise of the Liberal Argument thus appears to lack full justification, whether we accept Mill’s original formulation of the Harm Principle or Feinberg’s (1984) formulation. And without a more persuasive argument for the first 6  Nils Holtug (2002) offers a persuasive argument for this point. He considers three prominent theories of well-­being: hedonism, desire-­satisfaction theories, and objective-­list theories. He then argues that if someone feels uneasy when other people engage in homosexual intercourse, then on any of these theories of well-­being, the person who feels uneasy will have diminished well-­being.

386  Molly Gardner premise, the Liberal Argument is not especially compelling. At least, it does not strike me as more compelling than the sufficiency view. There may be better arguments for the non-­sufficiency view, but until they are fully articulated, I think we ought to provisionally accept the sufficiency view. What about the other two views? Recall that according to the necessity view, all harms are impediments to well-­being. I suggested above that the necessity view is prima facie plausible, and that we should abandon it only if we encounter a persuasive argument for the non-­necessity view. There are at least two arguments for the non-­necessity view that we ought to consider. The first argument, which I will call the ‘Legitimate Interest Argument’, appeals to a generic version of the moralized accounts of harm that I discussed in the first part of this section. Here it is: 1.  2.  3.  4. 

All impediments to morally legitimate interests are harms. All acts of property damage are impediments to morally legitimate interests. Some acts of property damage are not impediments to well-­being. Therefore, some harms are not impediments to well-­being.

The first two premises seem to fall out of Simester and von Hirsch’s (2011) view. For example, Simester and von Hirsch endorse the Harm Principle, and they write, ‘Within the terms of the Harm Principle, interference with another’s property rights constitutes a prima facie harm’ (2011: 41). Their view is also consistent with premise 3. Simester and von Hirsch argue that there is a close connection between property rights and well-­being, such that ‘welfare and human flourishing . . . would become unattainable should [the law of property] be lost’ (41). But the claim that well-­being depends upon a system of property rights is consistent with the claim that on occasion, some particular property right could be violated without a concurrent drop in well-­being. Moreover, Simester and von Hirsch appear to endorse the conclusion of the moralized harm argument when they write, ‘when we are harmed . . . it does not follow that our lives will actually go less well’ (36).7 Nevertheless, the Legitimate Interest Argument is not compelling. Commonsense morality tells against the notion that damage to my property necessarily constitutes harm to me. It is implausible, for example, that if you scratch my car door and the scratch is never discovered, you have harmed me. Perhaps there are good policy reasons to hold you legally accountable for the scratch, as if you had harmed me. But even in that case, I think that for the sake of clarity we

7  Here I am assuming they believe that when your life goes less well, you have suffered an impediment to your well-­being. However, I will consider shortly whether well-­being and having your life go a certain way can come apart.

What Is Harming?  387 should at least say, not that the scratch is a harm, but that we are treating the scratch as if it were a harm. Even better, we can appeal to the independently plausible view that a person can be wronged (legally, morally, or both) without being harmed. Let us set the Legitimate Interest Argument aside, then, and consider a second argument for the non-­necessity view. Some philosophers of well-­being hold that there is a difference between how well you are doing and how well your life is going. For example, Shelley Kagan (1994: 321) argues for the possibility that ‘a person’s life might be going poorly, even though the person himself is well-­off ’. He illustrates this possibility with the case of a businessman who dies believing that he was well respected by his peers and loved by his family. Nevertheless, the businessman’s wife was secretly cheating on him, and his peers only pretended to respect him. Kagan’s judgment is that although the man’s well-­being was not affected by the deception, his life was worse because of it. If we combine Kagan’s distinction between the man’s well-­being and his life with the principle that making someone’s life go poorly is a way of harming them, we arrive at the following argument:8 1.  Every instance of a life going poorly is a harm. 2. At least one instance of a life going poorly is not an instance of an impediment to well-­being. 3.  Therefore, some harms are not impediments to well-­being.

I will refer to this as the ‘Going Poorly Argument’. The second premise of the Going Poorly Argument is, of course, controversial; not everyone agrees with Kagan. However, even if we accept the second premise, it’s not clear that we have much reason to accept the first. Certainly, having your life go poorly is bad for you, but if it doesn’t affect your well-­being, it’s not clear why we should classify this as a harm. One of the main reasons to distinguish between a person and a person’s life is to achieve greater conceptual clarity, especially about some of the more puzzling cases in the well-­being literature. But that clarity is largely sacrificed if, immediately after drawing the distinction, we fail to draw a corresponding distinction between what’s bad for a person and what’s bad for a person’s life. So if we accept Kagan’s distinction, we ought to draw another distinction between harm and having your life go poorly. I conclude, then, that none of the arguments for either the non-­sufficiency view or the non-­necessity view are successful. We have little reason, then, to abandon what I take to be the default view, namely, the conjunction of the 8 David Boonin (2019) also considers how this argument bears on the possibility of harming the dead.

388  Molly Gardner sufficiency view and the necessity view. It remains highly plausible that all and only impediments to well-­being are harms, or that well-­being is the currency of harm. Recall that in Section 1, I suggested that questions about whether we can harm people with speech and about whether we can harm the dead turn largely on what we take the currency of harm to be. We can now start to see how such questions might be answered. First, consider the extent to which various theories of well-­being agree about whether hateful, threatening, or offensive speech can affect a person’s well-­being. If there is general agreement that such speech can impede a person’s well-­being, then we should tentatively conclude that speech can be harmful. Whether any speech should be regulated by the state is, of course, a separate question. With respect to the question of harming the dead, we might divide theories of well-­being into those that posit an experience requirement and those that don’t. If you have to experience an impediment to your well-­being in order to suffer one, then you cannot suffer harm after you are dead. On the other hand, if you can suffer an impediment to your well-­ being without experiencing it, then our account of the currency of harm leaves it open whether we can harm the dead. To fully settle the matter, we would need to inquire more fully into the precise boundaries of an individual’s well-­ being. We would also need to consider questions about relationship between harms and harmful events or actions. I will survey some of these questions in Section 3.

3.  The Measure of Harm So far, I have been using the terms ‘harming’ and ‘causing harm’ interchangeably. However, it is time to be more precise. In the literature on the metaphysics of harm, there are three prominent accounts of harming, and only one of these accounts is explicitly causal. I will formulate the three accounts as follows, using ‘A’ to stand for an action or omission and ‘V’ to stand for a victim: The temporal account:  A harms V if and only if V is worse off after A than V was before A.9 The counterfactual account:  A harms V if and only if a consequence of A is that V is worse off than V would have been in the absence of A.10

9  I include this view because it is an illuminating contrast to the counterfactual account. However, it’s not clear that anyone would endorse this view as I have formulated it. Stephen Perry (2003) endorses a ‘historical’ account of harming, and Michael Rabenberg (2015) endorses ‘historical worsening’ as a sufficient condition for harming. However, neither Perry’s view nor Rabenberg’s view, as I understand them, are really temporal accounts of harming. I take them both to be causal accounts of harming conjoined with temporal accounts of harm. Temporal accounts of either harming or harm are also discussed but not endorsed by Holtug (2002), Norcross (2005), Hanser (2008), Thomson (2011), and Bradley (2012). 10  Proponents of this account or some modified version of it include Parfit (1987), Norcross (2005), Klocksiem (2012), Tadros (2014), Feit (2015), Hanna (2016), and Boonin (2014, 2019).

What Is Harming?  389 The causal account:  A harms V if and only if A causes a harm for V.11

In the first two formulas, the term ‘worse off ’ should be understood in terms of whatever we take the currency of harm to be. If we take it to be well-­being, for example, then V will need to have less well-­being, either after A or given A. Notice that in the first two formulas, I did not use any causal language. For example, I did not say that A makes V worse off than she was before or worse off than she otherwise would have been. ‘Makes’ is a causal verb, and to highlight the difference between the first two accounts, which are both non-­causal, and the third, which is causal, we need to formulate the first two accounts without any causal verbs.12 An important difference between non-­causal and causal accounts involves the individuation of harms. The temporal and counterfactual accounts tend not to distinguish harming from harm. Proponents of these accounts often use the terms ‘harmful event’, ‘harm’, and ‘harming’ interchangeably. On the other hand, the distinction between harming and harm is important to a causal theorist: proponents of the causal account supplement their account of what it means to harm someone with a separate account of what a harm is. Generally, a harm for a causal theorist is some sort of condition or state of affairs that has something to do with the currency of harm.13 For example, if well-­ being is the currency of harm, a hedonist causal theorist might claim that the state of affairs in which I suffer six units of pain is a harm for me. Alternatively, a causal theorist who endorses desire-­satisfaction theory might say that the state of affairs in which my desire for health is frustrated is a harm for me. Or a causal theorist who endorses an objective-­list theory of well-­being might say that the state of affairs in which I lack four units of friendship is a harm for me. Here is an illustration of this difference in how causal and non-­causal theorists think about what a harm is. Take a case in which Smith breaks Jones’s knee with a baseball bat. Proponents of all three accounts would likely agree that Smith harmed Jones. Causal theorists might insist that something like the broken knee is 11  Proponents of this account or some modified version of it include Shiffrin (1999, 2012), Harman (2004, 2009), Thomson (2011), Northcott (2015), Bontly (2016), and myself (Gardner 2015, 2017, 2019a, 2019b). As I noted in n. 9, I also take Perry (2003) and Rabenberg (2015) to be proponents of this view. 12  Nevertheless, it is almost impossible to discuss the implications of non-­causal accounts without using causal language. Therefore, I will occasionally use causal terms to discuss these accounts, but the reader should understand these causal terms as stand-­ins for the relevant temporal or counterfactual relations. 13  For example, Harman (2004, 2009) and Shiffrin (1999, 2012) favor a ‘non-­comparative’ account of harm. According to the non-­comparative account, a harm is a condition or state of affairs that is bad for someone even if it doesn’t make her worse off than she would have been in its absence. However, a proponent of the causal theory of harming need not endorse a non-­comparative account of harm. My account (Gardner 2015) includes a counterfactual comparative account of harm, even though it does not include a counterfactual comparative account of harming. Causal theorists can also endorse a temporal account of harm conjoined with a causal account of harming; indeed, Perry (2003) and Rabenberg (2015) both seem to do something like this. For more on the distinction between accounts of harming and accounts of harm, see Gardner (2017).

390  Molly Gardner the harm. More precisely, a causal theorist might say that we can refer to the harm as the ‘broken knee’, but strictly speaking, the harm is something like the state of affairs in which Smith suffers eight units of broken-­knee pain. According to the causal theorist, Jones causes that state of affairs by causing Smith’s broken knee, but this doesn’t mean that Jones’s action is a harm; Jones’s action is merely a harmful event. For the non-­causal theorist, on the other hand, Jones’s action doesn’t just cause a harm—it is a harm. So there are at least two harms in the non-­causal theorist’s picture: the event in which Jones swings the baseball bat and the event in which Smith’s knee breaks. The distinction between causal accounts and non-­causal accounts is important because some of the main problems in the harm literature turn upon whether harming is causal or non-­causal. Indeed, I will next argue that with respect to solving at least two problems involving apparent harm, the causal account has an advantage over the non-­causal accounts. These two problems are the non-­identity problem and the problem of preemption. The non-­identity problem arises in what I will call ‘non-­identity cases’, or cases where an action that results in an individual’s existence also constrains his or her prospects for well-­being. One of Derek Parfit’s (1987) famous examples of a non-­ identity case centers on a 14-­year-­old girl who wants to have a child. Because she is so young, her child—call him Alex—will have a bad start in life; this makes her action seem intuitively harmful. On the other hand, Alex’s life will still be worth living. Importantly, Alex will never exist at all if his mother waits until she is older to conceive a child. (In that case, she will conceive a different child from a different egg and sperm.) Thus, if the 14-­year-­old conceives now and has Alex, Alex will be no worse off as a consequence of her action than he would have been had she not conceived him. Likewise, Alex will be no worse off after the girl’s action than he was before she conceived him. It seems, then, that neither the counterfactual account nor the temporal account can accommodate the intuition that the girl’s act of conceiving Alex would be harmful. The causal account fares better; to vindicate the intuition that Alex would be harmed, all a causal theorist needs to establish is that Alex’s having a bad start in life corresponds to some state of affairs that would be a harm for him and that the girl’s act of conceiving Alex would cause that state of affairs. My own solution to the non-­identity problem attempts to justify these claims by appealing to the following causal theory of harming:14 Harm (def.):  An event, E, harms an individual, S, if and only if E causes a state of affairs that is a harm for S.

14 This is roughly the account I defend in Gardner 2015 and Gardner 2019b, though I have reworded the definitions slightly.

What Is Harming?  391 Harm (def.):  A state of affairs, T, is a harm for an individual, S, if and only if (i) an essential component of T is the absence of something intrinsically good for S or the presence of something intrinsically bad for S; and (ii) if S existed and T had not obtained, then S would have had more of that good or less of that bad.

To see how this account works, first consider whether the state of affairs in which Alex has a bad start in life satisfies the first condition on being a harm. This seems likely; it’s plausible that a bad start in life is or at least supervenes on something that is intrinsically bad for Alex, according to whatever we take to be the true theory of well-­being. This state of affairs also satisfies the second condition. If Alex existed and the state of affairs in which he had a bad start in life had not obtained, then Alex would have had less of the corresponding intrinsic bad. Notice that when we imagine the possibility that Alex did not have a bad start in life, we are not imagining a possible world where Alex was never born. That is the possible world we would have to consider if we were imagining away the cause of the apparent harm, but on my account we imagine away the apparent harm, itself—not its cause. Instead of imagining that Alex was never born, we are imagining a possible world where the 14-­year-­old girl conceived Alex and then later had help raising him. In this scenario, Alex was raised with love, support, and stability. He had more of the relevant intrinsic good in his life than he did in the scenario where he had a bad start in life. Thus, the state of affairs in which Alex has a bad start in life is or supervenes on a harm for Alex: this state of affairs is worse for him than the state of affairs in which he exists and does not have a bad start in life. And since the 14-­year-­old girl would cause the state of affairs in which he has a bad start in life by conceiving Alex, she would harm him. The non-­identity problem is not the only problem that a causal account of harming can help to solve. A causal account can also help with preemption cases. Suppose that Smith breaks Jones’s knees with a baseball bat, and that if Smith had not broken Jones’s knees, Brown would have. Intuitively, it seems as though Smith harms Jones. Nevertheless, it is not true that Jones is worse off, given Smith’s action, than he would have been, had Smith not acted. For if Smith had not acted, Brown would have, and Jones would have been just as badly off. On the other hand, it is true that Jones is worse off after Smith’s action than he was before Smith swung the bat. It is also true that Smith caused Jones’s broken knees. Thus, in preemption cases, temporal accounts and causal accounts both seem to have an advantage over counterfactual accounts. If we consider both the non-­identity problem and the preemption problems together, it appears that the causal account most consistently has the advantage. But if we accept the causal account, how should we answer questions about harming the dead and harming future people? First, much (though not all) of our

392  Molly Gardner puzzlement about harming future people centers on the non-­identity problem. We saw above that the causal account can vindicate the intuition that harm occurs in a non-­identity case. Thus, the answer to our question about harming future people—at least in a non-­identity case—is that we can, indeed, harm them. What about the dead? I noted in Section 2 that if we accept well-­being as the currency of harm, and if we reject the experience requirement for well-­being, then many of our remaining questions about harming the dead will turn on the metaphysics of harming. The causal account of harming can help with some of these questions. For example, if we reject backwards causation, then we must reject the view that an event that happens after I die can cause a harmful state of affairs that obtains before I die. Nevertheless, it might still be possible that an event that happens after I die can cause a subsequent state of affairs that is a harm for me. A lingering question, then, is whether a state of affairs that obtains after I die can be a harm for me.

4.  Conclusion I have argued that well-­being is the currency of harm and that causation is its measure. I also showed how the issues of whether we can harm with speech, whether we can harm the dead, and whether we can harm future people are related to what we take to be the currency or the measure of harm. Nevertheless, I did not settle all the questions we might have about these issues. It remains to be settled, for example, whether a state of affairs that obtains after you are dead can be an impediment to your well-­being. Other unresolved questions have to do with how the literature on harming relates to the literature on doing and allowing harm15 and to the literature on collective harm.16 If we accept a causal account of harming, should we hold that all and only instances of harming are instances of doing harm? Though I won’t make a full case for it here, I believe that we should. Notice, however, that if harming is doing harm, and if doing harm is causing harm, and if causes can be actions or omissions, then it is possible to do harm by action and to do harm by omission. On the view I am describing here, then, the distinction between doing and allowing harm does not correspond to the distinction between action and omission.17 Next, what does harm have to do with collective harm? Issues of collective harm arise most famously in cases where we can formulate counterfactuals of the following kind: Even if I hadn’t gone on that Sunday drive, climate change would

15  For a helpful overview of this literature, see Woollard (2012a, 2012b). 16  For an illuminating discussion of collective harm, see Kagan (2011) and Nefsky (2012). 17  There is also small and interesting literature on the question of whether enabling harm is equivalent to either doing harm or allowing harm. See e.g. Rickless (2011) and Hanser (1999).

What Is Harming?  393 still have occurred.18 Or even if I hadn’t purchased meat, animals would have suffered just as much.19 The problem is that in cases like climate change or meat eating, the actions of large numbers of individuals seem to be collectively harmful, but the action of a single individual does not seem to be harmful at all. However, if the causal account of harming is true, the truth of the above counterfactuals does not matter to whether you have caused harm. Instead, whether your individual act of driving or purchasing meat was harmful will depend upon whatever we take to be the correct theory of causation. This might solve part of the problem; it might allow us to say that the actions in question are both collectively and individually harmful. However, if causation comes in degree,20 there may still be concerns about how strong your reasons are to refrain from contributing to climate change or animal suffering. Further work needs to be done, then, to determine whether the reasons for you not to contribute to a collective harm are as strong as you take them to be.

Bibliography Bernstein, Sara (2017). ‘Causal Proportions and Moral Responsibility.’ In Oxford Studies in Agency and Responsibility, iv, edited by David Shoemaker, 165–82. Oxford: Oxford University Press. Bontly, Thomas (2016). ‘Causes, Contrasts, and the Non-Identity Problem.’ Philosophical Studies 173: 1233–51. Boonin, David (2014). The Non-Identity Problem and the Ethics of Future People. Oxford: Oxford University Press. Boonin, David (2019). Dead Wrong: The Ethics of Posthumous Harm. Oxford: Oxford University Press. Bradley, Ben (2009). Well-Being and Death. New York: Oxford University Press. Bradley, Ben (2012). ‘Doing Away with Harm.’ Philosophy and Phenomenological Research 85: 390–412. Budolfson, Mark (2016). ‘Consumer Ethics, Harm Footprints, and the Empirical Dimensions of Food Choices.’ In Philosophy Comes to Dinner: Arguments About the Ethics of Eating, edited by Andrew Chignell, Terence Cuneo, and Matthew Halteman, 163–81. New York: Routledge. Feinberg, Joel (1984). Harm to Others. New York: Oxford University Press. Feinberg, Joel (1992). Freedom and Fulfillment: Philosophical Essays. Princeton: Princeton University Press. 18  See Sinnott-­Armstrong (2005). 19  See Budolfson (2016). 20  Michael Moore (2009) makes a compelling argument for the claim that causation is scalar. Sara Bernstein (2017) shows how a view like Moore’s gives rise to an interesting puzzle about causation and moral responsibility.

394  Molly Gardner Feit, Neil (2015). ‘Plural Harm.’ Philosophy and Phenomenological Research 90/2: 361–88. Gardner, Molly (2015). ‘A Harm-Based Solution to the Non-Identity Problem.’ Ergo 2/17: 427–44. Gardner, Molly (2017). ‘On the Strength of the Reason Against Harming.’ Journal of Moral Philosophy 14/1: 73–87. Gardner, Molly (2019a). ‘When Good Things Happen to Harmed People.’ Ethical Theory and Moral Practice 22/4: 893–908. Gardner, Molly (2019b). ‘David Boonin on the Non-Identity Argument: Rejecting the Second Premise.’ LEAP 7/3. doi 10.31009. Hanna, Nathan (2016). ‘Harm: Omission, Preemption, Freedom.’ Philosophy and Phenomenological Research 93/2: 251–73. Hanser, Matthew (1999). ‘Killing, Letting Die and Preventing People from Being Saved.’ Utilitas 11/3: 277–95. Hanser, Matthew (2008). ‘The Metaphysics of Harm.’ Philosophy and Phenomenological Research 77: 421–50. Hanser, Matthew (2009). ‘Harming and Procreating.’ In Harming Future Persons: Ethics, Genetics, and the Non Identity Problem, edited by M.  Roberts and D. Wasserman, 179–99. Dordrecht: Springer. Harman, Elizabeth (2004). ‘Can We Harm and Benefit in Creating?’ Philosophical Perspectives 18: 89–113. Harman, Elizabeth (2009). ‘Harming as Causing Harm.’ In Harming Future Persons: Ethics, Genetics, and the Non Identity Problem, edited by M.  Roberts and D. Wasserman, 137–54. Dordrecht: Springer. Holtug, Nils (2002). ‘The Harm Principle.’ Ethical Theory and Moral Practice 5/4: 357–89. Kagan, Shelly (1994). ‘Me and My Life.’ Proceedings of the Aristotelian Society 94: 309–24. Kagan, Shelly (2011). ‘Do I Make a Difference?’ Philosophy and Public Affairs 39/2: 105–41. Kleinig, John (1978). ‘Crime and the Concept of Harm.’ American Philosophical Quarterly 15/1: 27–36. Klocksiem, Justin (2012). ‘A Defense of the Counterfactual Comparative Account of Harm.’ American Philosophical Quarterly 49/4: 285–300. Mill, John Stuart (2002). The Basic Writings of John Stuart Mill. New York: Modern Library. Moore, Michael (2009). Causation and Responsibility: An Essay in Law, Morals, and Metaphysics. Oxford: Oxford University Press. Nefsky, Julia (2012). ‘Consequentialism and the Problem of Collective Harm: A Reply to Kagan.’ Philosophy and Public Affairs 39/4: 364–95.

What Is Harming?  395 Norcross, Alastair (2005). ‘Harming in Context.’ Philosophical Studies 123: 149–73. Northcott, Robert (2015). ‘Harm and Causation.’ Utilitas 27/2: 147–64. Nozick, Robert (1974). Anarchy, State, and Utopia. Oxford: Blackwell. Parfit, Derek (1987). Reasons and Persons. Oxford: Oxford University Press. Perry, Stephen (2003). ‘Harm, History, and Counterfactuals.’ San Diego Law Review 40: 1283–1313. Pitcher, George (1984). ‘The Misfortunes of the Dead.’ American Philosophical Quarterly 21/2: 183–8. Rabenberg, Michael (2015). ‘Harm.’ Journal of Ethics and Social Philosophy 8/3: 1–32. Rickless, Samuel (2011). ‘The Moral Status of Enabling Harm.’ Pacific Philosophical Quarterly 92: 66–86. Roberts, Melinda (2015). ‘The Nonidentity Problem.’  The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta. Winter. https://plato.stanford.edu/archives/ win2015/entries/nonidentity-problem/. Shiffrin, Seana (1999). ‘Wrongful Life, Procreative Responsibility, and the Significance of Harm.’ Legal Theory 5: 117–48. Shiffrin, Seana (2012). ‘Harm and Its Moral Significance.’ Legal Theory 18: 357–98. Simester, A.P., and von Hirsch, Andreas (2011). Crimes, Harms, and Wrongs: On the Principles of Criminalisation Oxford: Hart Publishing. Sinnott-Armstrong, Walter (2005). ‘It’s Not My Fault: Global Warming and Individual Moral Obligations.’ In Perspectives on Climate Change: Science, Economics, Politics, and Ethics, edited by Walter Sinnott-Armstrong and Richard Howath, 285–307. Amsterdam: Elsevier. Tadros, Victor (2014). ‘What Might Have Been.’ Philosophical Foundations of the Law of Torts, edited by John Oberdiek, 171–92. Oxford: Oxford University Press. Thomson, Judith (2011). ‘More on the Metaphysics of Harm.’ Philosophy and Phenomenological Research 82: 436–58. Woollard, Fiona (2012a). ‘The Doctrine of Doing and Allowing I: Analysis of the Doing/Allowing Distinction.’ Philosophy Compass 7/7 (2012): 448–58. Woollard, Fiona (2012b). ‘The Doctrine of Doing and Allowing II: The Moral Relevance of the Doing/Allowing Distinction.’ Philosophy Compass 7/7 (2012): 459–69.

18

Prioritarianism, Risk, and the Gap Between Prudence and Morality Nils Holtug

1. Introduction Roughly, prioritarianism—or the ‘priority view’, as it was coined by Parfit (1991)—is the view that benefits to individuals matter more the worse off these individuals are. This implies that we should give priority to the worse-­off in the distribution of benefits. I shall give a more precise definition of prioritarianism below, but for now this brief characterization will suffice. Parfit proposed prioritarianism as a distinctive alternative to egalitarianism, with which it is (or was) easily conflated. In particular, he argued that, unlike at least some versions of egalitarianism, prioritarianism is not vulnerable to the levelling-­down objection. Thus, unlike these forms of egalitarianism, prioritarianism sees no value in bringing about equality when it is achieved by reducing the levels of some and increasing the levels of none. Since what prioritarians care about is bringing about benefits, and especially benefits to the worse-­off, and levelling down does not involve bringing about any benefits, not even to the worse-­off, levelling down is not in any respect a change for the better. Parfit was not the first to distinguish between egalitarianism and principles akin to prioritarianism (see, for example, McKerlie 1984), but one of his great achievements in moral and political philosophy was to sharpen the contrast and to strengthen the case for prioritarianism. Mostly, prioritarianism has been discussed in the form of an axiological prin­ ciple, and more specifically as a principle that orders outcomes in terms of betterness. This was also the version with which Parfit (1991) was mostly concerned when he first presented it.1 Furthermore, with a few notable exceptions (Broome 1991; McKerlie 1984; Rabinowicz 2001; 2002), the question of how to apply prioritarianism to choices that involve risk has not received a great deal of attention, until relatively recently. The recent debate, on the other hand, has focused quite I would like to thank Matt Adler, Timothy Campbell, Carl Knight, Michael Otsuka, Shlomi Segall, and Hlynur Orri Stefánsson for comments on an earlier version of this article. 1  Although he did distinguish between ‘telic’ (or axiological) and ‘deontic’ prioritarianism (Parfit 1991: 20). Nils Holtug, Prioritarianism, Risk, and the Gap Between Prudence and Morality In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0019

400  Nils Holtug extensively on risky choices, and in particular on whether prioritarianism is cap­able of coming up with plausible answers when facing risk. One of the ­objections to prioritarianism that has been raised is that it inappropriately implies a gap between prudence and morality, even in single-­person cases. Thus, according to prioritarianism, we should sometimes sacrifice an individual’s expected welfare in order to protect her from the risk of a worse outcome. Or, more precisely, this is an implication of some versions of prioritarianism, including ex post prioritarianism. Parfit (2012) responded to this criticism by arguing for a pluralist form of prioritarianism, which encompasses both ex post and ex ante forms. I believe this move was quite unnecessary (Holtug 2019). However, my primary aim here is not to respond to Parfit’s view but to consider, more generally, ways in which different prioritarian theories imply a gap between prudence and morality and the extent to which this poses a problem for these theories. I shall argue that to the extent these theories, in different ways, imply such a gap, this is in fact unproblematic. Let me be a bit more precise about what I mean by ‘a gap between prudence and morality’. What I have in mind is a situation in which prudence unequivocally favours A over B, where A and B are both either outcomes or acts, whereas morality favours B over A. In order for prudence to unequivocally favour A over B, it must be the case that, among the individuals included in the comparison, (1) if A and B are outcomes, everyone has a higher welfare in A than in B, and (2) if A and B are acts, performing A leads to a higher level of welfare for everyone than performing B (assuming that A and B have certain outcomes), or performing A leads to a higher level of expected welfare for everyone than performing B (assuming that A and B have uncertain outcomes). Note that therefore there can only be a gap between prudence and morality, in the sense I have specified, in the absence of interpersonal conflicts (as regards welfare in the case of comparing outcomes and comparing acts with outcomes that are certain, and as regards expected welfare in the case of comparing acts with outcomes that are uncertain). Note also that, while I shall here primarily be concerned with single-­person cases, there may also be a gap between prudence and morality in multi-­person cases (such as if, in a case involving two individuals, act A gives to both a higher level of expected welfare than act B, but B is nevertheless considered morally preferable). My reason for focusing on single-­person cases is simply that these are the cases that critics of the prioritarian gap have tended to focus on. For some of these critics, this focus reflects their belief that, in multi-­person cases, a gap between prudence and morality may in fact be justifiable because interpersonal considerations, for example of equality, enter the picture. Thus, some egalitarian critics (Otsuka and Voorhoeve 2009: 195–8) believe that in a two-­person case, an act A is to be preferred to an act B, even though B maximizes both individuals’ expected welfare, if we know that one of these individuals is going to be much worse off ex post (although we do not know whom), and act A would benefit the worse-­off

Prioritarianism and Risk  401 individual (although slightly less than act B would benefit the better-­off individual). However, at the end of the day, a justification of the gap between prudence and morality must account also for multi-­person cases, and the prioritarian justification of the gap I provide in this article is supposed to apply to single- and multi-­person cases alike. As it transpires, prioritarianism is not alone in having to account for a gap between prudence and morality. As suggested, some forms of egalitarianism also imply a gap (albeit only in multi-­person cases). Egalitarians will of course justify such a gap differently than prioritarians, appealing to the (dis)value of relations of (in)equality between individuals, but the point is that there are a number of the­ or­ies in distributive justice that imply a gap between prudence and morality and that therefore, each in their way, need to justify the existence of such a gap and to do so in light of the values they invoke. In what follows, then, I explain the moral basis in prioritarianism for the gap between prudence and morality. In Section 2, I provide a more precise account of axiological prioritarianism and what it implies for the relation between prudence and morality. In Section 3, I then provide an account of four prioritarian theories that (unlike axiological prioritarianism) have implications for risky choices, namely ex ante prioritarianism, ex post prioritarianism, pluralist prioritarianism, and my favoured version, factualist prioritarianism. In Section  4, I present the objection that prioritarianism implies a gap between prudence and morality in single-­person cases in greater detail, which includes explaining the extent to which this objection applies to different versions of prioritarianism. In Section 5, I explain why I believe the prioritarian gap between prudence and morality is unproblematic, even in single-­person cases. And in Section 6, I conclude.

2.  Axiological Prioritarianism Consider: Axiological Prioritarianism.  An outcome is non-­ instrumentally better, the larger the sum of weighted individual benefits it contains, where benefits are weighted such that they gain a greater value, the worse off the individual is to whom they accrue. (Holtug 2006: 132–6; 2010: 204–9)

Prioritarianism can be represented by:

G = f ( w1) + f ( w2 ) + . . . + f ( wn ) ,

where f is an increasing and strictly concave function of individual welfare (wi). It is instructive to consider some structural aspects of prioritarianism (Adler and

402  Nils Holtug Holtug 2019; Holtug 2015, 2017). First, it satisfies the Pareto principle, according to which (1) an outcome is better than another if it is better for at least one individual and at least as good for everyone, and (2) these outcomes are equally good if they are equally good for everyone. Thus, for example, (2, 1) is better than (1, 1), where the numbers refer to individual welfare levels. Second, axiological prioritarianism satisfies the Pigou-­Dalton principle of transfer, according to which shifting a given sum of welfare from a better-­off individual to a worse-­off individual, where this does not reverse their relative positions and holding the levels of everyone else constant, increases the value of an outcome. Thus, for example, (3, 3, 1) is better than (5, 1, 1).2 Third, axiological prioritarianism satisfies separability, according to which the outcome ordering is not impacted by individuals whose welfare is not impacted. Therefore, if (2, 2, 2) is better than (4, 1, 2), then (2, 2, 1) is better than (4, 1, 1), where in each pairwise comparison the last individual is not impacted and so her level cannot impact the ordering. Fourth, axiological prioritarianism satisfies anonymity, which is an impartiality requirement, according to which the value of an outcome is not affected by a permutation of welfare over individuals who exist in the outcome. Thus, (2, 1) and (1, 2) are equally good. Finally, axiological prioritarianism satisfies a person-­ affecting principle, according to which an outcome, x, cannot be better (worse) than another outcome, y, unless there is someone for whom x is better (worse) than y. Therefore, for example, (1, 1) cannot be better than (2, 1). In fact, axiological prioritarianism even satisfies a strong person-­affecting principle, according to which an outcome, x, cannot be in any respect better (worse) than another outcome, y, unless there is someone for whom x is better (worse) than y. So (1, 1) cannot be in any respect better than (2, 1), whereas (some) egalitarians would of course say that (1, 1) is better as regards equality, which is also what renders them vulnerable to the levelling-­down objection. Now, these person-­affecting principles have implausible implications in population ethics and I have argued elsewhere that, ultimately, they should be replaced with some more sophisticated person-­affecting principles (Holtug 2010: 184–8). However, for simplicity, I shall here assume fixed populations and as regards fixed populations, the implications of these more sophisticated principles coincide with those of the person-­affecting principles listed above. Therefore, for present purposes, I can stick to the latter. Consider first the implications of axiological prioritarianism in single-­ person cases. Here, the prioritarian ordering of outcomes coincides with the prudential ordering. That is, an outcome, x, is at least as good as another 2  In these comparisons, each individual’s place in the parenthesis stays constant, such that for example the first (same) individual is at 3 in the first distribution and 5 in the second.

Prioritarianism and Risk  403 outcome, y, where x and y contain only a single individual, p, if and only if x is at least as good as y for p. This simply follows from the fact that axiological prioritarianism satisfies the Pareto principle. Furthermore, for the same reason, if everyone in a multi-­person case is at least as well off in x as in y, x is at least as good as y. Therefore, axio­logic­al prioritarianism does not imply a gap between prudence and morality. Of course, this does not mean that there is no difference between prioritarianism and prudence. Prioritarianism weights benefits according to how well off the bene­fi­ciary is, whereas prudence does not. I develop this point further below. It is worth pausing to consider the moral basis for these weights in single-­ person cases. The prioritarian weights are what allow us to give priority to the worse off in multi-­person cases. That is, if we can either give a benefit of a fixed size to a worse-­off or a better-­off individual, it is better to give it to the ­worse-­off individual, everything else being equal. Indeed, this is what allows prioritarians to respect the separateness of persons. According to Rawls (1971: 27), utilitarianism conflates ‘all persons into one through the imaginative acts of the impartial sympathetic spectator’. Prioritarianism does not do this since it is sensitive to the individual levels at which benefits fall. However, the prioritarian weighting applies not only to interpersonal trade-­offs, it also applies in single-­person cases where there is no separateness of persons to respect. Might then the prioritarian weighting be an unwarranted residue in single-­person cases from the aim of respecting the separateness of persons in multi-­person cases (Holtug 2019)? As Michael Otsuka and Alex Voorhoeve (2018: 76) put it: ‘prioritarianism . . . adopts for a single person a principle that is appropriate only for a society consisting of many separate persons whose interests clash with one another’ (cf. Porter 2012). However, I believe the prioritarian weighting is well motivated, even in single-­ person cases. To bring out the difference between axiological prioritarianism and prudence more clearly, compare two possible words, w1 and w2, that contain only one (the same) individual, p, who is very well off in w1 and very badly off in w2. Suppose now that at the end of her life in both w1 and w2, a sizeable benefit accrues to p. Perhaps she is able to go on a very enjoyable trip travelling in South America. Suppose also that, while her life has of course been very different in w1 and w2 up till then, there is nothing in her past (or present) that would suggest that she derives more welfare from the trip in one of these worlds than in the other. For example, it does not in one of them to a greater extent satisfy long-­ standing preferences or contribute to projects she has been working on. In fact, she benefits equally from the trip in w1 and w2. Nevertheless, intuitively, this equally sized benefit contributes more moral value in w2 than in w1. This reflects that, since p has been so much worse off in w2, it is more important that she receives the benefit there. Note that this is a moral rather than a prudential

404  Nils Holtug judgement. It is not that she benefits more in w2, rather, it is simply morally more important that she benefits there.3 Indeed, the reason for applying the prioritarian weighting in single-­person and multi-­person cases is the same. Just as it is more urgent to provide a benefit to a worse-­off individual than to a better-­off individual because the worse-­off individual is at a lower absolute level of welfare, it is more urgent to provide a benefit to p in w2 than in w1 because she is at a lower absolute level in w2. Now, the prioritarian weighting, which is what differentiates morality and prudence in single-­person cases, does not imply a ‘gap’ in the sense I have specified above. Thus, prudence and axiological prioritarianism generate identical orderings of w1 and w2 and, more generally (as also pointed out above), identical orderings of outcomes. This is so even though, for example, they do not ascribe the same value to a particular benefit to p in w1 and w2. Of course, prioritarianism is by no means the only moral principle that invokes non-­prudential values in single-­person cases. Consider, for example, absolute desert.4 More specifically, consider the view that there is some absolute level of welfare that people deserve, where that level is fixed on the basis of choices they have made in their lives. On this view, for example, Hitler is at a low level of desert and so does not deserve a very high level of welfare. Now compare two outcomes that contain only Hitler, where he enjoys a life of extreme luxury, fulfilment, and pleasure in one outcome but is significantly worse off in the other. While the outcome in which he is better off is prudentially better for him,5 we might believe that, on the basis of this notion of absolute desert, the outcome in which he is less well off is morally better. Hitler simply does not deserve to be that well off.6 Before I go on to consider risky choices, note that the prioritarian weighting includes an impersonal element (Holtug 2010: 213–18). Consider again the different contribution of the benefit to p to the moral value of w1 and w2. This difference in moral value is not a value for anyone. After all, the benefit has the same size in these two worlds. Therefore, prioritarians are, in a certain sense, committed to the existence of impersonal value. However, this impersonal value is linked 3  Note also, incidentally, that neither utilitarianism, egalitarianism, or some combination thereof captures this intuition about the contribution of the benefit in w1 and w2. Utilitarianism implies that the benefit is of equal moral value in these two worlds, and (at least on the way it is usually construed) egalitarianism doesn’t apply to single-­person cases. 4  Similarly, in his discussion of the prioritarian gap between prudence and morality, Porter (2012: 360–3) appeals to retributive theory as an example of a moral concern, other than prioritarianism, that applies to single-­person cases and may outweigh or cancel out prudential concerns. 5  Of course, it might be denied that an outcome in which Hitler fares much better than he deserves could be better for him, say because sensitivity to absolute desert is an element in an objective list account of welfare. However, note that desert would then have to be so weighty a factor in welfare that, however much value we fill into his life in terms of the other items on the objective list, this could not give him a life with a higher welfare than a life in which he gets what he deserves. This claim seems to me extremely implausible. 6  Incidentally, for discussions of the prospect of rendering prioritarianism sensitive to desert, see Adler (2018) and Arneson (2006).

Prioritarianism and Risk  405 to personal value (welfare) in a particularly intimate way; the former cannot exist in the absence of the latter. Furthermore, we cannot have an increase in impersonal value in the absence of an increase in personal value. In other words, impersonal value supervenes on personal value. This is captured by the person-­affecting principles. An outcome cannot be (in any respect) better (worse) than another unless it is better (worse) for someone. I return to this point about impersonal value below.

3.  Prioritarianism and Risk Axiological prioritarianism does not in itself inform us what choices we should make, let alone what decisions to make when facing risk. Therefore, I shall now outline four different prioritarian theories that do address this issue. These the­or­ ies provide criteria of rightness, that is, inform us what actions are right. Three of them are possibilist, which means that they assess risky choices in terms of the possible outcomes they may lead to and their probabilities. The last theory is fact­ ual­ist, which means that it assesses risky choices in terms of the outcomes to which they will in fact lead. For simplicity, I assume consequentialism, and more specifically that acts should be assessed simply on the basis of prioritarian value. According to ex post prioritarianism, an act is right if and only if the expected prioritarian value of its possible outcomes is as great as that of any alternative act. So, to assess an act, we first determine the prioritarian value of each of its possible outcomes (as it would be assessed on axiological prioritarianism), the probability of each outcome, and then add up probability-­weighted prioritarian outcome values. This theory differs from ex ante prioritarianism, according to which an act is right if and only if the prioritarian value of the expected individual welfare is as great as that of any alternative act. Here, we first determine the expected welfare of each individual on the basis of their welfare in the possible outcomes and the probabilities of these, apply the prioritarian function to each individual’s expected welfare, and then add up these prioritarian-­weighted expected individual welfares. A third theory combines ex ante and ex post prioritarianism, which involves settling the relative weight of each of these two concerns (as pointed out above, Parfit 2012 adopted this approach). Finally, according to factualist prioritarianism, an act is right if and only if it will in fact lead to an outcome with at least as much prioritarian value as any alternative act. Unlike the three former possibilist theories, this theory determines rightness on the basis of the outcomes to which acts will in fact lead rather than on the basis of expected values. For a variety of reasons, criteria of rightness need to be supplemented with decision procedures (Bales 1971; Hare 1981: ch. 2). Unlike criteria of rightness, decision procedures don’t tell us which acts are right, but how to make decisions when confronting choices. The decision procedure we have most reason to adopt

406  Nils Holtug is the procedure that, if adopted, will best promote the aim set out in our criterion of rightness. Once the criterion of rightness is in place, it is largely an empirical question what decision procedure we should adopt. One reason why we need a decision procedure, which applies in particular to factualist prioritarianism (and other factualist views), is that we sometimes face risky choices, which means that we don’t know which outcomes particular acts will in fact lead to. The factualist’s decision procedure will therefore include mechanisms for dealing with risky choices. But of course, ex ante, ex post, and pluralist prioritarians also need decision procedures, for example because sometimes we may not have access to reliable probabilities, do not have the time to do the prioritarian calculations, or will not be sufficiently motivated by the results of these calculations. I have elsewhere proposed that factualist prioritarianism is the most plausible of the prioritarian theories outlined in this section (Holtug 2019), but here I want to focus more generally on the gap between prudence and morality in prioritarian theories.

4.  On the Gap Between Prudence and Morality The objection that prioritarianism implies a gap between prudence and morality, even in single-­person cases, has been made by a number of philosophers. For example, Otsuka and Voorhoeve (2009, 2018; Otsuka 2012, 2015) have argued that prioritarianism is insensitive to what they call ‘prudential justification’. Suppose we know of a person, Adrian, that he will either develop a slight impairment or equiprobably a severe impairment. Both can be treated, but we need to do so now, before he develops one of these impairments. The severe impairment cannot be fully cured and, in fact, the treatment for the slight impairment will give rise to slightly greater benefit for Adrian if this is the impairment he develops than the treatment for the severe impairment will if this is the impairment he develops. The welfare levels in the possible outcomes are represented in Table 18.1. I shall assume square root as our prioritarian weighting function. The ex post prioritarian value of treating for, respectively, the severe and the slight impairment will then be: Table 18.1 Adrian develops the severe impairment A. We treat the severe impairment B. We treat the slight impairment

4 2

Adrian develops the slight impairment 7 10

Prioritarianism and Risk  407

( 0.5 × √ 4 ) + ( 0.5 × √ 7 ) = 2.323 ( 0.5 × √ 2 ) + ( 0.5 × √10 ) = 2.288

Therefore, according to ex post prioritarianism, we should treat the severe impairment. This, however, would be to ignore prudential justification, according to Otsuka and Voorhoeve. Suppose we treat Adrian for the slight impairment and, as it turns out, he develops the severe impairment. In that case, we can still justify our treating for the slight impairment to Adrian by pointing out that we did what was prudentially best for him, we maximized his expected welfare. Such a justification, on the other hand, is unavailable when we move from intra­per­ son­al to interpersonal trade-­offs. If we can either provide a benefit for a much worse-­off individual or a slightly greater benefit for a much better-­off individual, we cannot provide a prudential justification to the worse-­off individual if we decide to provide the greater benefit to the better-­off person. Thus, according to Otsuka and Voorhoeve (2018), we ignore the unity of the individual in the intra­ per­son­al trade-­off; the fact that both possible benefits fall in possible futures of Adrian’s. This is what is problematic about the ex post prioritarian gap between prudence and morality in single-­person cases. In a recent article, Otsuka (2015: 5) has elaborated on the claim that prudence requires providing the treatment for the slight impairment for Adrian. Here, he suggests that ‘ideally rational self-­interested preferences will provide the measure of a person’s well-­being’, where ‘such preferences, being ideally rational, will conform to sound axioms of expected utility theory’. Therefore, treating for the slight impairment, which if effective would give rise to the greater increase in welfare, would be to provide the treatment which it is in Adrian’s rational self-­interest to prefer. David McCarthy (2008: 19–22) also complains about the gap between prudence and morality to which prioritarianism gives rise and, more specifically, argues that it is an implausible feature of ex post prioritarianism that it violates what he calls the ‘coincidence principle’. According to this principle, assuming a population of exactly one person, a lottery over histories is at least as good as another if and only if the former lottery is at least as good for this person. Furthermore, Toby Ord (2015) presents a similar but apparently deeper challenge to ex post prioritarianism. Thus, he presents a case in which ex post prioritarianism favours one act over another, even though the former act both holds a lower total of expected welfare, at lower level of expected welfare for everyone, and is more unequal. The case is represented in Table 18.2 and includes two acts, A and B, where ‘heads’ and ‘tails’ refer to states of nature, a and b are individuals, and the numbers refer to welfare levels. A has a certain outcome of 36 welfare units for a and 49 for b, whereas B will either lead to 4 units or 100 units for each. This gives an expected value of 36 for a

408  Nils Holtug Table 18.2

Act A Act B

 

Heads

Tails

Expected welfare

Prioritarian Value

 

a

b

a

b

a

b

 

   

36 4

49 4

36 100

49 100

36 52

49 52

13 12

and 49 for b in A, and of 52 for each in B. The prioritarian weighting function assumed is again square root, which gives a prioritarian value of 13 for A and 12 for B.  So A is better than B even though both a and b have a higher level of expected welfare in B, B therefore has a higher total of expected welfare, and B is more (in fact perfectly) equal both ex ante and ex post. Ord (2015: 5) argues that therefore, ex post prioritarianism does not merely make trade-­offs between efficiency and equality, in this case it favours a reduction in both. But of course, prioritarianism was never about equality in the first place, it was proposed by Parfit (1991) as a distinctive alternative to egalitarianism, including to forms of egalitarianism that combine concerns for equality and efficiency. This is what allows prioritarians to avoid the levelling-­down objection. Therefore, the implication that ex post prioritarianism may recommend acts that sacrifice not only ex ante efficiency but also ex ante and ex post equality need not in itself worry prioritarians. The issue is with the gap between prudence and morality. Note, however, that in Ord’s example, this gap appears in a case that involves two individuals, reminding us that such gaps are not restricted to single-­person cases. Unlike ex post prioritarianism, ex ante prioritarianism does not imply a gap between prudence and morality. For illustration, consider the ex ante prioritarian value of the two possible treatments for Adrian (see Table 18.1):

√ ( 0.5 × 4 ) + ( 0.5 × 7 )  = 2.345



√ ( 0.5 × 2 ) + ( 0.5 × 10 )  = 2.449

Thus, ex ante prioritarianism favours treating the slight impairment, which is also what maximizes Adrian’s expected welfare. More generally, ex ante prioritarianism will always favour an act that maximizes everyone’s expected welfare. As noted above, Parfit (2012) ended up defending pluralist prioritarianism (but without specifying how ex ante and ex post prioritarianism are to be weighed against each other). Because it includes ex ante prioritarianism, pluralist prioritarianism will to a greater extent be sensitive to prudential justification than pure ex post prioritarianism is. Nevertheless, according to Otsuka and Voorhoeve (2018: 79), it is still insufficiently sensitive to prudential justification and in

Prioritarianism and Risk  409 particular, it still implies a gap between prudence and morality in single-­person cases (Otsuka 2015: 4–5). As regards factualist prioritarianism, it implies that the right act is the act that in fact leads to the outcome with the highest prioritarian value (as assessed by axiological prioritarianism). And we know that axiological prioritarianism does not imply a gap between prudence and morality. Nevertheless, factualist prioritarianism will sometimes favour an act even though there is another act that would maximize everyone’s expected welfare. Consider again the case of Adrian and suppose that he will in fact develop the severe impairment. In that case, fact­ ual­ist prioritarianism implies that the treatment for the severe impairment should be provided, although the treatment for the slight impairment would maximize Adrian’s expected welfare. And so factualist prioritarianism implies a gap between prudence and morality, even in single-­person cases. However, factualist prioritarians (like other factualists) should also be fact­ual­ ists about prudence. This is because the argument for factualism in morality is also an argument for factualism as regards prudence. What reason would there be to, for example, maximize expected value other than that we want to maximize the value that is in fact realized (Carlson 1995: 20–5)? And if the value that is in fact realized is what ultimately matters, then this should be reflected in our (ul­tim­ate) principles of prudence and morality. Since factualist prioritarians will therefore also be factualists about prudence, factualist prioritarianism does not imply a gap between morality and the form of prudence that ultimately matters. Any act that in fact leads to an outcome that is better for everyone than any alternative act would will also be a right act according to factualist prioritarianism. And, as regards single-­person cases in particular, an act is right if and only if this act renders the individual at least as well off as any alternative act would. (For a more elaborate discussion and defence of factualist prioritarianism against, among other things, the objection that it implies a gap between prudence and morality, see Holtug 2019.) Of course, we will often not know which outcomes particular acts lead to and in part for this reason, as stated above, factualist prioritarians need a decision procedure. Furthermore, it is at least possible that this decision procedure will include a rule, according to which we should sometimes aim to maximize expected prioritarian value (that is, ex post prioritarianism, albeit in the decision procedure rather than as a criterion of rightness). Remember that, as I said, it is largely an empirical question what decision procedure the factualist prioritarian has most reason to adopt. And it is possible that some of the cases in which this rule should be applied are cases in which we should then perform a particular act even though there is another act that would hold a higher level of expected welfare for everyone. This would be true if, for example, we should aim to maximize expected prioritarian value in the case of Adrian. And if so, we would have a case in which, arguably, our moral decision procedure recommends a different act

410  Nils Holtug than our prudential decision procedure would (assuming that our prudential decision procedure would tell us to maximize Adrian’s expected welfare). At the level of decision procedures, then, there would be a gap between prudence and morality. Therefore, in different ways and at different levels, various prioritarian prin­ ciples imply (or may imply) a gap between prudence and morality. In Section 5, I argue that to the extent this is so, it does not constitute a good objection to (these forms of) prioritarianism. Before I do so, however, I want to briefly consider a different argument prioritarians may make, that would altogether remove the gap between prudence and morality. According to this argument, prioritarianism should be considered a conflict resolution principle, that is, a principle that applies only to interpersonal conflicts of interest (cf. Williams 2012). If so, prioritarianism would not apply to single-­person cases. Which multi-­person cases it would apply to would depend on how, exactly, the conflict resolution idea is interpreted, whether it for example rules out cases where there are no interpersonal conflicts in expected welfare or cases where, in pairwise comparisons of all the possible outcomes, there are no interpersonal conflicts. My main reason for not construing prioritarianism as a conflict resolution principle is that it seems to me an unmotivated restriction of the principle. As I explained in Section  2, prioritarian weighting seems to me appropriate even in single-­person cases, where there are of course no interpersonal conflicts. A second reason is that, if prioritarianism is construed in this manner, it would seem to lose some appealing features in non-­conflict cases, such as the fact that it satisfies the Pareto principle. Thus, (2, 1) is Pareto superior to (1, 1), but since the comparison does not involve an interpersonal conflict, prioritarianism would not apply to it.

5.  Minding the Gap I have already explained in Section 2 why, even in single-­person cases, I believe that prioritarian weighting is appropriate. We should assign greater moral weight to a benefit that befalls an individual in a world in which she is badly off than an equal benefit that befalls her in a world in which she is well off. Morally speaking, this benefit is more urgent in the world in which she is badly off. And this weighting is what explains, in the case of (some versions of) prioritarianism, the gap between prudence and morality. This, of course, is equally true in the particular case of Adrian. When ex post prioritarianism implies that Adrian should be treated for the severe impairment, this is due to the prioritarian weighting. Furthermore, this weighting is motivated by the claim that it is more urgent to prevent the outcome in which Adrian is worst off, because here he develops the severe impairment and it is not treated, than to secure for him a slightly greater benefit in case it turns out he has the slight impairment.

Prioritarianism and Risk  411 Something similar may be said on behalf of Parfit’s pluralist prioritarianism. To the extent pluralist prioritarianism favours treating Adrian for the severe impairment, due to its ex post prioritarian component, this is motivated in the prioritarian weighting of his welfare and the idea that it is most urgent to prevent a state in which he develops the severe impairment and it is untreated. Of course, in pluralist prioritarianism, ex post prioritarianism is combined with ex ante prioritarianism, but the greater urgency of preventing lower levels of welfare is what motivates the inbuilt tendency towards a gap between prudence and morality, due to the former of these principles. Finally, factualist prioritarianism does not imply a gap between morality and the form of prudence that ultimately matters, namely factualist prudence. Nevertheless, as we have seen, it is possible that the decision procedure of the factualist prioritarian will include a rule according to which we should sometimes aim to maximize expected prioritarian value, which is at least potentially in tension with a prudential decision procedure according to which we should sometimes aim to maximize expected welfare. Now, insofar as we are worried about gaps between prudence and morality, it may seem less worrying if the gap is one that potentially occurs in decision procedures rather than in criteria of rightness, where decision procedures are after all somewhat crude instruments for realizing the values that ultimately matter. More importantly, however, to the extent to which the factualist prioritarian decision procedure does conflict with a prudential decision procedure, according to which we should aim to maximize expected welfare, this will again reflect the prioritarian weighting. If our decision pro­ced­ ure tells us to treat Adrian for the severe impairment, this is because the fact­ual­ist prioritarian criterion of rightness applies prioritarian weighting to individual welfare. And this reflects our concern that it is more urgent to benefit an individual in case she is worse off than in case she is better off, everything else being equal. In a critical discussion of Parfit’s response to the gap between prudence and morality, Otsuka (2015) distinguishes between defences of prioritarian weighting in single-­person cases that are based on rational self-­interest and on impersonal values. And he argues that neither is particularly plausible. I shall consider some of his objections, but first let me position my own defence of prioritarian weighting in Otsuka’s dichotomy. I do not claim that prioritarian weighting is an expression of rational self-­interest, even in single-­person cases. Rather, such weighting is an explicitly moral concern of giving priority to the worse-­off. Thus, for present purposes at least, I can simply accept that it is in Adrian’s self-­interest to receive the treatment for the slight impairment.7 However, nor is my defence of prioritarian weighting straightforwardly based on impersonal reasons. As argued above, there is an impersonal element in the prioritarian weighting in that a bene­fit of a 7  Nevertheless, I do have certain reservations about Otsuka’s account of the measure of welfare and its relation to rational self-­interest.

412  Nils Holtug fixed size gives rise to unequal increases in moral value depending on the level at which it falls. However, as also pointed out, this is an impersonal element that can only occur in the presence of a personal element, that is, welfare. Furthermore, there can only be a change in this impersonal value insofar as there is a change in welfare. Impersonal value supervenes on personal value. And as we have seen, prioritarianism satisfies the person-­affecting principles.8 In his description of a case that is similar to the case of Adrian, Otsuka (2015: 5) states that ‘there are . . . no interpersonally comparative or otherwise distributive considerations here that tell in favour of paying heed to anything other than this [person’s] rational self-­interest’. My own account, on the other hand, implies that there are such other distributive considerations. And note, as also pointed out in Section 2, that it is not prima facie implausible to suggest that con­sid­er­ations that are distinctively moral may play a role for distributive justice in single-­person cases. Consider again absolute desert. The suggestion is that we may have reason not to maximize an individual’s welfare, or expected welfare, insofar as this would give her a better life than she deserves. While I do not hold this view myself, it is fairly clear how a moral concern for desert could provide a reason not to maximize an individual’s welfare, or expected welfare, in certain single-­person cases and so could motivate a gap between prudence and morality. Such moral concerns for respectively the worse-­off and for absolute desert are not captured by Otsuka’s notion of what individuals have rational self-­interested reasons to prefer and so it is not surprising that they may imply moral judgements according to which we should sometimes act contrary to an individual’s rational self-­interest, even in single-­person cases. Of course, concern for the worse-­off and desert may impact the preferences of individuals and so to that extent what we have reason to do insofar as we want to act as they prefer. Note, however, that the moral suggestion under consideration is not that we should give priority to the worse-­off or give people what they deserve only insofar as they prefer this themselves. Furthermore, even if an individual for example prefers less enjoyment to more because she believes that this is what she deserves, it seems implausible to conclude that she therefore has a rational self-­interest in receiving less enjoyment. That would imply that it would be in an individual’s self-­interest to be horribly tortured if she preferred this because she thought she deserved it, which it does seem possible to prefer while satisfying the requirements of rationality. Otsuka (2015: 20–1) also levels three specific objections against ‘impersonal’ justifications of prioritarian weighting in single-­person cases. According to the 8  In Holtug (2010: 213–18), I further explain this impersonal element in prioritarianism. And I argue that while there is an impersonal element in prioritarianism, it does not hold intrinsic (non-­ instrumental) value. Rather, what is intrinsically valuable, according to prioritarianism, is compound states of affairs, each consisting in the state that a benefit of a certain size befalls an individual and the state that this individual is at a particular welfare level, where this intrinsic value increases as the size of the benefit increases but decreases as the welfare level increases.

Prioritarianism and Risk  413 first objection, ‘it is hard to identify any traces of purely impersonal value’ in cases similar to that of Adrian, which imply a gap between prudence and morality (21). However, as pointed out above, my account does not construe the case for prioritarian weights in purely impersonal terms; rather, my account satisfies the person-­ affecting principles. And in the case of Adrian, I appeal to the claim that there is one of the possible outcomes it is particularly urgent to prevent, namely that in which he is worst off, and that that provides a case for prioritarian weights. While one may of course disagree that this is a relevant concern, it does not involve invoking impersonal values that it is hard to identify traces of. Otsuka’s second objection pertains specifically to Parfit’s account of the gap between prudence and morality and he points out that, in single-­person cases, Parfit appeals only to what is at stake for these single individuals themselves (and so not to purely impersonal values). Thus, in a case similar to that of Adrian, Parfit (2012: 424–5) emphasizes that ‘there is a one in two chance that [the act that prevents the worst possible outcome] would actually be better for [this individual], and this act would be better for [this individual] if he is very badly off ’. On this basis, Otsuka (2015: 12) concludes that Parfit ‘appeals to the rational self-­ interest or prudence of such a choice’. However, I have carved out a moral middle way between rational self-­interest and purely impersonal values, which is distinctively moral and person-­affecting and which applies to single-­person and multi-­person cases alike. This may also be the middle way to which Parfit is appealing here. When Parfit emphasizes that treating for the more severe condition may turn out to be better for the affected individual, that there is in fact a one in two chance of that, and that this treatment would be better for him if he is very badly off, this may simply be a way of linking possible benefits and moral urgency to the favoured act. In other words, he may be emphasizing that prioritarianism satisfies the person-­ affecting principles, which implies that had there not been a possible outcome in which the treatment for the severe impairment would benefit this individual, there would not be a prioritarian case for providing this treatment, and also that, compatibly with these person-­affecting principles, we may consider it particularly urgent to benefit him in the event he is very badly off. If this is the correct interpretation of Parfit, he is invoking a person-­affecting moral perspective, which is not reducible to ‘rational self-­interest’ and which furthermore cannot plausibly be criticized for not complying with a plausible account of such interests. The final objection Otsuka (2015: 21) raises to an impersonal account of prioritarian weighting in single-­person cases invokes Joseph Raz’s (1986: 240) proposal of an alternative to egalitarianism, to which prioritarians sometimes approvingly refer (Parfit 1991: 41; Holtug 2010: 203): what makes us care about various inequalities is not the inequality but the concern identified by the underlying principle. It is the hunger of the hungry, the

414  Nils Holtug need of the needy, the suffering of the ill, and so on. The fact that they are worse off in the relevant respect than their neighbours is relevant. But it is relevant not as an independent evil of inequality. Its relevance is in showing that their hunger is greater, their need more pressing, their suffering more hurtful, and therefore our concern for the hungry, the needy, the suffering, and not our concern for equality, makes us give them priority.

Otsuka suggests that it is in its insistence on the pressing needs of the needy etc. that prioritarianism is at its strongest, where impersonal values tend to divert us from such concerns for persons. However, I have argued that prioritarianism expresses a moral person-­affecting concern that, even in single-­person cases, justifies prioritarian weighting. It is this very weighting that accommodates the needs of the needy, the suffering of the ill, and so on. Let me finally return briefly to the case of Adrian. Do I really believe that prioritarian principles that, in some way or another, favour treating him for the severe impairment and (by assumption) acting contrary to his rational self-­ interest are not to be faulted for doing so? Yes. This can be justified with reference to a prioritarian weighting that I have argued is plausible even in single-­person cases. But note that I have not said anything that commits me to the claim that Adrian’s own preferences, insofar as they are known, should be ignored when making the choice between the treatments. Thus, there may be other concerns, besides distributive justice, we may want to pay attention to, including that of personal autonomy.9 What choices we should make insofar as other values are at stake is simply a separate question that I have not considered here. Here, I have merely been concerned with the question of whether the gap between prudence and morality that some forms of prioritarianism imply can be given a (prioritarian) justification, which is compatible with the suggestion that at the end of the day, we will want to combine this concern with other concerns. What, then, about Adrian himself, what attitude should he take towards the possible treatments? Again, I have not said anything that settles this question. Adrian might adopt a prudential perspective and on that basis favour the treatment for the slight impairment. But he may also adopt a moral (prioritarian) perspective and on that basis favour the treatment for the severe impairment, emphasizing the moral urgency benefiting him in the event it turns out he is worse off. I have not made any suggestions as to how these two perspectives should or may figure in Adrian’s reasoning about the case.

9  In that case, an ‘everything else being equal’ clause should be included in prioritarianism in the definition above.

Prioritarianism and Risk  415

6. Conclusion I have considered the objection to prioritarianism that it implies an implausible gap between prudence and morality, which is especially clear in single-­person cases. More specifically, I have distinguished between different versions of prioritarianism and considered the extent to which they imply such a gap. And I have argued that while axiological prioritarianism does not imply a gap between prudence and morality, it introduces the weighting function that does imply such a gap for other prioritarian views. More specifically, in different ways, ex post and pluralist prioritarianism imply a gap, and factualist prioritarianism may perhaps do so in its decision procedure, given certain assumptions about what this pro­ ced­ure looks like. However, I have also argued that to the extent these different prioritarian views imply a gap between prudence and morality, this is simply a reflection of the (appropriate) weighting of welfare in the prioritarian function. This is the weighting that allows us to plausibly say that a benefit to p in w2, where he is very badly off, contributes more moral value than an equal benefit to p in w1, where he is much better off. It is the weighting that, even in single-­person cases, allows us to consider benefits to an individual more urgent in case he is badly off than insofar he is already quite well off. I have also argued that while the prioritarian weighting function includes an impersonal element, it nevertheless satisfies the person-­affecting principles, and so does not exemplify an impersonal approach to morality. Rather, prioritarianism steers a middle course between Otsuka’s rational self-­interest and impersonal morality. It reflects a moral concern that is basically about how individuals fare, but particularly to the extent they fare worse.

References Adler, Matthew (2018). ‘Prioritarianism. Room for Desert?’ Utilitas 30/2: 172–97. Adler, Matthew, and Holtug, Nils (2019). ‘Prioritarianism. Replies to Critics.’ Politics, Philosophy and Economics 18/2: 101–44. Arneson, Richard (2006). ‘Desert and Equality.’ In Egalitarianism. New Essays on the Nature and Value of Equality, edited by Nils Holtug and Kasper Lippert-Rasmussen, 262–93. Oxford: Clarendon Press. Bales, Eugene (1971). ‘Act-Utilitarianism: Account of Right-making Characteristics or Decision-making Procedure?’ American Philosophical Quarterly 8/3: 257–65. Broome, John (1991). Weighing Goods. Oxford: Basil Blackwell. Carlson, Erik (1995). Consequentialism Reconsidered. Dordrecht: Kluwer Academic Publishers. Hare, Richard (1981). Moral Thinking. Oxford: Clarendon Press.

416  Nils Holtug Holtug, Nils (2006). ‘Prioritarianism.’ In Egalitarianism. New Essays on the Nature and Value of Equality, edited by Nils Holtug and Kasper Lippert-Rasmussen, 125–56. Oxford: Clarendon Press. Holtug, Nils (2010). Persons, Interests, and Justice. Oxford: Oxford University Press. Holtug, Nils (2015). ‘Theories of Value Aggregation: Utilitarianism, Egalitarianism, Prioritarianism.’ In The Oxford Handbook of Value Theory, edited by I. Hirose and J. Olson, 267–84. Oxford: Oxford University Press. Holtug, Nils (2017). ‘Prioritarianism.’ In Oxford Research Encyclopedia of Politics, edited by William R. Thompson. Oxford: Oxford University Press. Holtug, Nils (2019). ‘Prioritarianism. Ex Ante, Ex Post, or Factualist Criterion of Rightness?’ Journal of Political Philosophy 27/2: 207–228. McCarthy, David (2008). ‘Utilitarianism and Prioritarianism II.’ Economics and Philosophy 24/1: 1–33. McKerlie, Dennis (1984). ‘Egalitarianism.’ Dialogue 23/2: 223–37. Ord, Toby (2015). ‘A New Counterexample to Prioritarianism.’ Utilitas 27/3: 298–302. Otsuka, Michael (2012). ‘Prioritarianism and the Separateness of Persons.’ Utilitas 24/3: 365–80. Otsuka, Michael (2015). ‘Prioritarianism and the Measure of Utility.’ Journal of Political Philosophy 23/1: 1–22. Otsuka, Michael, and Voorhoeve, Alex (2009). ‘Why It Matters that Some Are Worse off than Others: An Argument Against the Priority View.’ Philosophy and Public Affairs 37/2: 171–99. Otsuka, Michael, and Voorhoeve, Alex (2018). ‘Equality Versus Priority.’ In Oxford Handbook of Distributive Justice, edited by Serena Olsaretti, 65–85. Oxford: Oxford University Press. Parfit, Derek (1991). Equality or Priority? The Lindley Lecture. University of Kansas. Parfit, Derek (2012). ‘Another Defence of the Priority View.’ Utilitas 24/3: 399–440. Porter, Thomas (2012). ‘In Defence of the Priority View.’ Utilitas 24/3: 349–64. Rabinowicz, Wlodek (2001). ‘Prioritarianism and Uncertainty: On the Interpersonal Addition Theorem and the Priority View.’ In Exploring Practical Philosophy. From Action to Values, edited by Dan Egonsson et al., 2–21. Aldershot: Ashgate. Rabinowicz, Wlodek (2002). ‘Prioritarianism for Prospects.’ Utilitas 14/1: 2–21. Rawls, John (1971). A Theory of Justice. Oxford: Oxford University Press. Raz, Joseph (1986). The Morality of Freedom. Oxford: Clarendon Press. Williams, Andrew (2012). ‘The Priority View Bites the Dust.’ Utilitas 24/3: 315–31.

19

Relational Egalitarianism Telic and Deontic Kasper Lippert-­Rasmussen

1.  Introduction Equality was one of many objects of Derek Parfit’s wide-­ranging philosophical interests, albeit not likely among his most central ones. Nevertheless, he succeeded in writing a now-­classic piece on equality (Parfit 1995)1 in which he draws a commonly adopted distinction between telic and deontic egalitarianism.2 Parfit notes that there are many different ideals of equality, including political equality, equality before the law, and giving equal moral weight to everyone’s interests. His concern is with the ideal of distributive equality: to be an egalitarian ‘in my sense’, one must be ‘concerned with people’s being equally well off’ (Parfit 1998: 3). Accordingly, his distinction between telic and deontic egalitarianism is a distinction between two forms of distributive egalitarianism drawn within the field of the ‘ethics of distribution’ (Parfit 1995: 2). Recently, many philosophers have come to reject the view that egalitarian justice can be adequately understood in the light of distributive egalitarianism. In an early critique of the so-­called distributive paradigm of justice, Iris Marion Young (1990: 8) complains that it ‘tends to focus on the possession of material goods and social positions’.3 While she concedes that some of the friends of this paradigm extend a concern for distribution ‘to cover such goods as self-­respect, opportunity, power and honor’, such extensions result in distorted understandings of these goods ‘as identifiable things or bundles distributed in a static pattern among identifiable, separate individuals’ (8).4 In Young’s view ‘oppression and I thank Ketan Ramakrishnan for very helpful comments on a previous version of the chapter. 1  His Lindley Lecture is reprinted later in its full length as Parfit (2002: 81‒125). A shorter, slightly revised version came out in Parfit (1998: 1‒20). 2  R. Arneson (1989), G. A. Cohen (2011), and L. Temkin (1993) are generally regarded as leading proponents of (luck egalitarian) telic egalitarianism, R.  Dworkin (2000) as a leading proponent of (luck egalitarian) deontic egalitarianism. 3  The target of Young’s critique is the distributive paradigm; and while that includes distributive theories of justice, it comprises more than that, including ‘characteristic questions’ (16). 4  Pace Young, friends of the distributive paradigm are not committed to caring about the distribution of ‘identifiable things or bundles’ in Young’s concrete sense. The goods, the distribution of which

Kasper Lippert-­Rasmussen, Relational Egalitarianism: Telic and Deontic In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0020

418  Kasper Lippert-Rasmussen domination . . . should be the primary terms for conceptualizing injustice’ (8‒9). We can set aside the details of her criticism. For present purposes, what is im­port­ ant is that something like Young’s view on egalitarian justice has become increasingly influential. According to that view, which I shall refer to as ‘relational egalitarianism’, egalitarian justice is fundamentally a matter of the egalitarian nature of social relations rather than a matter of an equal distribution of well-­ being or, for that matter, any other good.5 My main concern in this chapter is to argue that we can extend the distinction between telic and deontic distributive egalitarianism to relational egalitarianism (Sections 2 and 3). I submit that doing so makes us see that, in some respect, telic relational egalitarians have a better chance of rebutting the levelling-­ down objection than do telic distributive egalitarians (Section  4); that telic relational egalitarianism is not committed to what many see as the implausibly wide scope of telic distributive egalitarianism (Section 4); but that telic relational egalitarianism is difficult to articulate in relation to choices involving different population sizes (Section  5); that deontic relational egalitarianism captures the agent-­relative shape of central egalitarian duties but is problematically silent on whom we have duties to relate to (Section 6); and, finally, that, partly in view of two problems just mentioned, relational egalitarians should accept both telic and deontic relational egalitarianism as well as telic distributive egalitarianism (Section 7). My aim, therefore, is to both develop the relational egalitarian approach, by articulating two distinct strands it ought to embrace, and to demonstrate the approach’s limitations.

2.  The Distinction between Telic and Deontic Distributive Egalitarianism According to Parfit’s account, telic egalitarians hold that It is in itself (unjust and therefore) bad if some people are worse off than others. (Parfit 1998: 3)6

is their concern, are goods in the abstract sense that they are something that we want or have reason to want. 5 Other important pieces expounding relational egalitarianism are Anderson (1999, 2010); Kolodny (2014); Miller (1998); Scanlon (2018); Scheffler (2003, 2005, 2015); Schemmel (2012); Viehoff (2014, 2019); Wolff (1998). 6  Parfit (1998: 3) notes that ‘we might add, “through no fault or choice of theirs” ’. However, making this addendum means that the state of inequality is not in itself bad (Lippert-­Rasmussen 2007: 104‒5). On the telic view so qualified, inequality is bad only if it has a certain genesis. This blurs the distinction between telic and deontic egalitarianism, because it is then false that the former is concerned only with the state of inequality itself, while the latter is concerned with how inequality was brought about (though, admittedly, telic and deontic egalitarians might be asking different questions about the

Relational Egalitarianism  419 As Parfit points out, the most plausible form of telic egalitarianism is pluralist; it is one according to which there are other features such that it is in itself good or bad if those features are realized. In addition to the indented claim above, one pluralist, telic egalitarian view claims that well-­being is in itself good, and more of it is in itself better. According to this view, a world marked by inequality and a very large sum of welfare might be better than a world containing the same people being equally well off, but all leading very bad lives.7 While deontic egalitarianism is not merely the logical complement of telic egalitarianism—the view that it is not the case that it is in itself (unjust and therefore) bad if some people are worse off than others does not amount to a distinctively egalitarian view—deontic egalitarians do not—at least not qua deontic egalitarians—think that it is in itself ‘bad if some people are worse off than others’ (Parfit 1998: 6). So, negatively, deontic egalitarians are neutral on the truth of telic egalitarianism. Positively, Parfit writes that while it is often, but presumably not always, unjust, according to deontic egalitarians, that some people are worse off than others, ‘[their] objection is not really to the inequality itself. What is unjust, and therefore bad, is not strictly the state of affairs, but the way in which it was produced’ (Parfit 1995: 9).8 Accordingly, deontic egalitarians think that ‘we should aim for equality, not to make the outcome better, but for some other moral reason. We may believe, for example, that people have rights to equal shares’ (Parfit 1995: 4; 1998: 6). In view of these remarks, I propose the following reconstruction of Parfit’s notion of deontic egalitarianism: It is often bad if some people are worse off than others, because often such in­equal­ity results from violations of distinctively egalitarian duties.9

Let me make three clarifying points about this definition. First, ‘often’ is im­port­ ant, since without it there might be nothing distinctively egalitarian about deontic egalitarianism.10 Nozick was no deontic egalitarian. Yet there could be circumstances under which even he would object to inequality, because of how it was produced (e.g., through theft; cp. Parfit 1998: 6). genesis of distributive inequality). At least this is so on the plausible assumption that when ‘we ask whether some state is in itself bad, it is irrelevant how it came about’ (Parfit 1998: 8). 7  Cp. the discussion of the levelling-­down objection later. 8  Parfit (2017: 400‒6) later introduces the term ‘deontic badness’—the badness in question results simply from the act being wrong and is not what makes the act wrong—to refer to the badness in question here. 9  Cp. ‘Egalitarianism can also be a purely deontic view, which is not about the goodness of outcomes, and claims only that, in many cases, we ought to distribute benefits more equally between different people’ (Parfit 2011: 198). I believe this formulation is too broad. 10  It would be too strong to say that inequality is always bad, because it always results from violations of distinctively egalitarian duties. If all worse-­off people consented to being made worse off, presumably, inequality might not result from any violation of a distinctively egalitarian duty.

420  Kasper Lippert-Rasmussen Second, ‘distinctively egalitarian duties’ reflects that, according to Parfit, de­ontic egalitarians object to inequality only when it reflects wrongdoing and, presumably, wrongdoing involves violating a duty. Moreover, it has to be a ‘dis­ tinct­ive­ly egalitarian duty’ for a reason similar to the one expounded under the previous point. Suppose people often promise each other to bring about equality but break their promises for reasons unrelated to incapacity. If so, one need not be an egalitarian to object to how the distributive inequality has been produced; one simply has to think that promises are morally binding, and many who are not egalitarians think so. Parfit offers some examples of distinctively egalitarian duties, such as respecting people’s rights to equal shares. For present purposes, we can leave the notion of a ‘distinctively egalitarian duty’ undefined.11 Third, ‘because’ can be understood in two different ways. It can be understood to mean that the unequal distribution is bad, say, because it is tainted by its origin.12 Or it can be understood epistemically, as a signal that something bad has taken place: namely, the violation of distinctively egalitarian duties. When a police detective thinks to herself ‘That’s bad’ upon learning that there is DNA from a missing person at a location where a witness reported hearing a gunshot and van speeding away shortly thereafter, she might simply mean that this is bad in light of the likely explanation of those facts and, thus, what those are a (natural) sign of. As articulated by Parfit and formulated here, telic and deontic egalitarianism are different, though consistent, views.13 Consider Parfit’s case of the divided world (1998: 5)—say, Europe and the Americas before the crossing of the Atlantic. Suppose that Europeans are better off than Americans. While this state of affairs is in itself bad according to telic egalitarianism, according to deontic egalitarianism it is not. Given that there has been no interaction between the parties, the in­equal­ity in question presumably did not result from any wrongdoing.14 11  One possible distinctively egalitarian duty is the duty to remedy purely natural inequalities, i.e., inequalities that have come about through no wrongdoing. The existence of such a duty does not imply—even setting aside inequalities that have come about through wrongdoing—that all in­equal­ ities are unjust, since purely natural inequalities that cannot be remedied are possible. 12  Cp. Goodin (2013). The ‘taint’ view must offer an account of whether an equal distribution would be tainted by antecedent wrongdoing if an unequal one would. 13  Cp. O’Neill (2008: 120). Admittedly, Parfit (1995: 4) does say that deontic egalitarians believe that we ‘should aim for equality, not to make the outcome better, but for some other moral reason’. But on the assumption that a state of affairs can be good even if we should not bring it about for that reason but for some other moral reason; e.g., it might be better if you get my books, but I should bring that about not for that moral reason but because I have promised you my books, this does not show that the two views are mutually exclusive. Similarly, Parfit says: ‘On the Telic View, inequality is bad; on the Deontic View, it is unjust’ (9). However, that is consistent with the deontic view being neutral on whether inequality is bad. Moreover, Parfit (4) explicitly concedes that, possibly, if something is unjust it is for that reason bad, and he acknowledges that ‘We might of course hold beliefs of both kinds [i.e., both telic and deontic egalitarian kind: KLR]’, even if such a view ‘does not need a separate discussion’. (Finally, since ‘bad’ here is not restricted to ‘bad in itself ’, surely, deontic egalitarians can agree that inequality is bad, cp. O’Neill’s (2008: 120) strong reading of the quote.) Hence, neither of the two passages to which I refer shows that Parfit thinks that the telic and deontic views are inconsistent. 14  Or at least: the inequality does not result from any wrongdoing done to Americans by Europeans; the inequality might result from wrongdoing done to some Americans by other Americans, in which case, presumably Parfit assumes, the inequality might still plausibly be unobjectionable from the point

Relational Egalitarianism  421 More generally, Parfit thought that the two views of distributive equality were naturally aligned with different views regarding the scope of equality. While telic egalitarianism naturally has a universal scope, deontic egalitarianism might plausibly have a narrower scope; for example, plausibly it can be held that relations of deontic, egalitarian justice obtain only between citizens who are subjected to a common political authority, in which case the scope of distinctively egalitarian duties do not extend beyond the borders of this political community (cf. Nagel 2005). Another way of seeing the difference between the two views is to consider the levelling-­down objection (Parfit 1998: 9–11), which involves a comparison of two states—an equal and an unequal one—in which some (or even all) are better off and no one is worse off than in the first state. According to telic egalitarianism, the former and equal state of affairs is in one respect better than the second, unequal one—it is better with respect to egalitarian justice even if it is better for no one in any respect. Partly for reasons relating to the non-­identity problem, Parfit (1995: 42n57; 1984: 351–79) does not think this feature provides basis for decisive objection to telic egalitarianism. However, he thinks that the objection exerts pressure on telic egalitarianism (especially because many of the egalitarian intuitions that lead people to become telic egalitarians can be explained instead, and in some cases are in fact better explained, by the priority principle, which is immune to the levelling-­down objection (cf. Parfit 1995: 26–7)). No such thing is true of deontic egalitarianism. Deontic egalitarianism does not imply that inequality is bad in itself and, thus, unless the unequal state of affairs involves or results from some kind of wrongdoing that the equal state of affairs does not, they have no reason to favour the equal over the unequal state of affairs (Parfit 1998: 10).15

3.  Telic and Deontic Relational Egalitarianism Having introduced a reconstructed version of Parfit’s distinction between telic and deontic distributive egalitarianism, I want to show how a similar distinction can be drawn between two forms of relational egalitarianism. But before doing so, I need to expound slightly on relational egalitarianism. The most influential statement of this view is probably Elizabeth Anderson’s ‘What Is the Point of Equality?’. While other relational egalitarians offer somewhat divergent statements, for reasons of space I simply rely on Anderson’s characterization.

of view of deontic equality. For present purposes, we can ignore unusual deontic views, which imply that the inequality is bad even in the present scenario. 15 Cp. Lippert-­Rasmussen (2007: 118–23); O’Neill (2008: 142–3). Note that according to the ‘tainted’ version of deontic equality, that ideal might also be vulnerable to the levelling-­ down objection.

422  Kasper Lippert-Rasmussen Anderson draws the following contrast between her favoured account of relational egalitarianism, democratic equality, and a (in her mind) particularly objectionable version of the ideal of distributive egalitarian justice; to wit, luck egalitarianism: democratic equality . . . is a relational theory of equality: it conceives equality as a social relationship. Equality of fortune is a distributive theory of equality: it conceives of equality as a pattern of distribution. Thus, equality of fortune regards two people as equal as long as they enjoy equal amounts of some distributable good—income, resources, opportunities for welfare, and so forth. Social relations are largely seen as instrumental to generating such patterns of distribution. By contrast, democratic equality regards two people as equal when each accepts the obligation to justify their actions by principles acceptable to the other, and in which they take mutual consultation, reciprocation, and recognition for granted. Certain patterns in the distribution of goods may be instrumental to securing such relationships, follow from them, or even be constitutive of them. But democratic egalitarians are fundamentally concerned with the relationships within which goods are distributed, not only with the distribution of goods themselves.  (Anderson 1999: 313–14)16

While the immediate target of Anderson’s critique is luck egalitarianism, as is apparent in the quoted passage, her real target is much broader; to wit, any view that, at a fundamental level, articulates justice in distributive terms, whether those terms are egalitarian or non-­egalitarian or whether they are luck-­sensitive or luck-­insensitive. Unlike the distributive ideal of equality, Anderson believes that democratic equality captures the concerns of real-­life egalitarians; concerns which, in a nutshell, she articulates as pertaining to social equality. She argues that, while ‘Inegalitarianism asserted the justice or necessity of basing social order on a hierarchy of human beings, ranked according to intrinsic worth’ and ‘inequalities in the distribution of freedoms, resources, and welfare’ lie at the core of ‘inegalitarian ideologies of racism, sexism, nationalism, caste, class, and eugenics’, Egalitarian political movements oppose such hierarchies. They assert the equal moral worth of persons . . . Negatively, the claim repudiates distinctions of moral worth based on birth or social identity—on family membership, inherited social status, race, ethnicity, gender, or genes. There are no natural slaves, plebeians, or aristocrats. Positively, the claim asserts that all competent adults are equally moral agents: everyone equally has the power to develop and exercise moral 16  Cp. Scheffler (2003: 21‒2, 31; 2015: 21–2), who draws a similar contrast, although his form of relational egalitarianism ultimately takes a different shape.

Relational Egalitarianism  423 responsibility, to cooperate with others according to principles of justice, to shape and fulfill a conception of their good.  (312)

These passages raise numerous questions (Lippert-­Rasmussen 2015: 179–207; 2018a; 2018b). Most importantly, there is nothing preventing one from adopting a pluralist account of egalitarian justice according to which egalitarian justice has a distributive as well as relational component (and possibly others too). We can set aside these complications, however, since my concern here is to show how we can distinguish between telic and deontic relational egalitarianism in the same way that Parfit thinks we can distinguish between telic and deontic distributive egalitarianism. Telic relational egalitarians maintain: It is in itself (unjust and therefore) bad if social inequality exists; i.e., that some relate to others as inferiors or superiors.

Conversely, deontic relational egalitarians affirm: It is often bad if social inequality exists, as such inequality often results from violations of distinctively egalitarian duties.17

The parallels to the definitions of telic and deontic distributive egalitarianism should be obvious. However, before proceeding to offer some examples of relational egalitarian views that seem best classified as telic or as deontic relational views, let me add a few clarifying remarks about telic and deontic relational egalitarianism. First, generally social equality is thought by relational egalitarians to be non-­ existent in cases where one person—say, someone with a dominating personality— idiosyncratically relates to others as superior and/or they to him as inferiors.18 On this view, social inequality only obtains when the infrastructure of the relevant unequal relation is provided by generally accepted social norms, practices, institutions, expectations, etc. (e.g., if the person relates as superior to others because of his and their race in a racist society). From Anderson’s examples of what relational egalitarians oppose, it becomes rather apparent that her main concern— and, so she would claim, that of real-­life egalitarians—is social inequality. When exactly a number of unequal social interactions turn into social inequality is an interesting question (Viehoff 2019). However, for present purposes we can ignore that question and simply rely on examples that are clearly one or the other. Second, as with telic and deontic distributive egalitarianism, the affirmation of one of these positions does not necessitate the denial of the other. Hence, deontic 17  Cp. n. 8. 18  As the case of one person enslaving another in a state of nature shows, possibly relational egalitarians should care about relational inequality even when it is does not amount to social inequality in the relevant sense.

424  Kasper Lippert-Rasmussen relational egalitarians might also think that social inequality, e.g. unavoidable, unintended forms of social hierarchy, is bad (because unjust) even when it does not result from violations of any egalitarian duties. Third, some might think that it is an understatement relative to the actual views of relational egalitarians to imply that they think that social inequality is ‘often’ bad because it results from violations of egalitarian duties. However, note, first, that, strictly speaking, ‘often’ also covers ‘almost always’.19 Moreover, substituting ‘always’ for ‘often’ is too strong for the same reason as why that move would not do in the case of deontic distributive egalitarianism either; for example, that, at least within certain limits, people might validly consent to implement social inequality, in which case social inequality might not result from a violation of any egalitarian duty. To see how the two views differ, consider a situation marked by social inequality that has not resulted from anyone’s wrongdoing. Suppose, for instance, that there is a social hierarchy such that beautiful people are treated as superior to not-­so-­ beautiful people in a wide range of different contexts, including many in which physical appearance makes no relevant difference; that is, to one’s ability to perform the particular job in question. Beauty is a status, as sociologists might put it. Suppose also that sociobiologists provide a compelling explanation of why such hierarchy is in our genes and why we—neither individually, nor as a society—can do nothing about it. Suppose, finally, that to violate an egalitarian duty, it must be the case that the agent could have acted in such a way that he or she did not violate it. Given those assumptions, we have a social hierarchy that telic relational egalitarians condemn as bad, even if deontic relational egalitarians do not.20 Something akin to the opposite situation is also imaginable. Suppose we have a distinctively egalitarian duty to relate to others as equals. Suppose also that while this duty is often violated, none of these violations result in or reproduce social inequality. We live in a multiethnic society consisting of several equally powerful groups, the members of which often treat the members of other groups as inferior. But because no group is dominant, none of these acts are backed up by dominant hierarchical social norms etc. and all attempts to dominate fail. Here, telic relational egalitarians will have no objection, since arguably there is no social inequality.21 Deontic relational egalitarians might object, however, since even if there is no social inequality, ex hypothesi, people often violate their distinctive egalitarian relational duties; for example, they often violate their duty to justify their actions to others, which Anderson imputes to us. 19  Some deontic, distributive egalitarians might say the same about distributive inequalities; i.e., that they—at least if they are substantial—almost always reflect wrongdoing. 20  ‘For Deontic Egalitarians, if nothing can be done, there can be no injustice’ (Parfit 1998: 7; cp. Lippert-­Rasmussen 2007: 106‒7). 21  Admittedly, this might not be so according to all relational egalitarians, cp. Viehoff (2014: 355–7).

Relational Egalitarianism  425 Before proceeding to the exploration of telic and deontic relational egalitarianism in slightly greater detail, let me offer some motivation for my Parfitian taxonomy in the form of textual evidence to the effect that the views of at least some relational egalitarians are best classified as either telic or deontic relational egalitarian views. As an example of the former, consider Martin O’Neill’s egalitarianism; he embraces the view that ‘certain kinds of egalitarian relations have a value that is not reducible to the effects on individual welfare that those social relations may have’ (O’Neill 2008: 141‒2). He thinks equality, which is good for no one in any respect, might be better than inequality, because the former state of affairs involves the ‘eradication of servility, domination, and exploitation’ (142, 146)—the existence of which are (also) impersonally bad.22 Other relational egalitarians are better classified as deontic relational egalitarians. Consider, for instance, the following quote from Christian Schemmel (2011: 366): Relational egalitarianism . . . is a view about social justice; its aim is to specify rights and duties that individuals have as members of society, and which normally override other social values  .  .  .  The objection to [inegalitarian] relationships is not merely that they are, in some sense, bad for people, but that they constitute unjust treatment: domination involves subjection to the arbitrary exercise of power on the part of somebody else; marginalization involves an unjust denial of opportunities to participate in basic social and political institutions.

According to this view, social equality need not be good (because just) in itself. However, people have certain rights and duties, the content of which can be specified without any reference to social equality and which nevertheless are such that if everyone’s rights are fulfilled and no one violates their relevant duties, then social equality obtains.23 Strictly speaking, Schemmel might not object to social inequality in itself but to the manner in which it is brought about; to wit, through certain rights-­violating acts—acts of domination, etc.—which implies the nonfulfillment of the duties of the agents.24

22  Non-­intrinsic egalitarianism is the egalitarian position favoured by O’Neill. It is different, he contends, from both Parfit’s telic and deontic egalitarianism. This is true in the sense that O’Neill’s egalitarianism is not a form of distributive egalitarianism and he does not entertain the possibility of applying the telic/deontic distinction to relational egalitarianism. Note also that, at least in the case of servility it would be difficult to argue that the relevant impersonal badness consists in the violation of distinctively egalitarian duties. 23  I am assuming here that Schemmel thinks that domination etc. amounts to unjust treatment independently of whether non-­domination is a constitutive element of social equality. 24  Admittedly, things are slightly complicated, because once social inequality is in place, individual acts violating egalitarian duties of justice might themselves be constituents of the unequal state of affairs, to which extent Schemmel objects to social inequality itself. However, plausibly there are

426  Kasper Lippert-Rasmussen In short, a distinction can be drawn between telic and deontic relational egalitarianism. In Sections  4–6, I will argue why drawing this distinction is relevant.

4.  Scope, Levelling-­Down, and Telic Relational Egalitarianism Consider, first, the two issues that, according to Parfit, separate telic and deontic egalitarianism: the issue of scope and the levelling-­down objection. As regards the former, one might think that the value of relational equality is different from that of distributive equality; that is, the value of two persons relating as equals is conditional on their being socially related. If they are not socially related, they are related neither as equals nor as unequals, which is neither bad nor good.25 However, even if two people are not socially related, one can still be worse off than the other. This is bad, according to telic distributive egalitarianism. If this is the correct view of the impersonal value of relational equality, then relational egalitarianism implies a different answer to the scope question than that offered by telic distributive egalitarianism. In Parfit’s divided world scenario, positive telic egalitarianism implies that it would be better for Europeans and Americans to start interacting socially and to do so on an equal basis. According to negative telic egalitarianism, there is nothing bad about them not relating as equals. Similarly, since we are not related socially to distant past generations, no issue about the badness of how we relate to them arises. In short, telic relational equality does not have some of telic distributive equality’s problematic implications. While Parfit views the broad scope of telic distributive equality as an attractive feature of the view, others see things differently; for example, according to them, it is not bad in any respect that we are better off than the Incas (cp. Hirose 2015: 80‒2; O’Neill 2008: 134). If they are correct, then telic relational egalitarianism is a more plausible view in relation to the scope issue than telic distributive egalitarianism. Is telic relational egalitarianism vulnerable to the levelling-­down objection? Interestingly, some relational egalitarians favour relational egalitarianism over telic, luck egalitarianism, in part because the latter is vulnerable to the levelling-­ down objection. For instance, Elizabeth Anderson (2010: 20‒1) writes: ‘ “Levelling down,” in the sense of demanding a gratuitous waste of a good that cannot be redistributed to another’s advantage, for the sole purpose of ensuring that no one has more than others, is unjust.’ inegalitarian social relations that do not consist in one agent violating a distinctively egalitarian duty owed to another, e.g. relations of (voluntary) servility. 25  This is consistent with it being worse that people are not socially related than that they are socially related and relate as equals; i.e., that which is implied by positive telic relational egalitarianism.

Relational Egalitarianism  427 When directed against telic relational egalitarianism, the levelling-­ down objection presumably goes as follows: Compare the following two states of affairs. In the first, everyone relates as equals. In the second, some people relate as superiors to inferiors etc., but some people are better off and no one is worse off. We could read ‘no one is worse off ’ in a modest sense: that is, in the sense that no one’s well-­being is lower, all things considered, under social inequality than under social equality. In the ambitious sense, ‘no one is worse off ’ means that no one is worse off under social inequality than under social equality, neither in terms of well-­being, nor, all things considered, relationally speaking.26 Interestingly, telic relational egalitarianism might be more robust against the levelling-­down objection than is telic, distributive egalitarianism, whether we read ‘worse off ’ in a modest or more ambitious sense. This is because a plausible case can be made for the claim that relating to others as equals has personal value. If so, then it appears as though the situation appealed to in the levelling objection—to wit, that equality is realized but that it is not good for anyone in any respect—is not possible. If relational equality is achieved, then, necessarily, that will be good for some people in some respect; at least, this might be so on a particular theory of well-­being. To see what I have in mind here, consider Parfit’s (1984: 493‒502) tripartite division of different theories of well-­being: mental state-­based, preference-­based, and objective-­list theories. Starting with the first theory, one cannot appeal to mental state-­based accounts of well-­being to deny the possibility of levelling down. While people’s mental states might be better (e.g., more pleasurable) if they relate as equals—the sense of humiliation one can suffer from being treated as less than an equal can be intensely unpleasant—it is also clear that, at best, it is contingently true that relational equality makes mental states more pleasurable. Appealing to a preference-­based theory involves a problem similar to this one. Many people prefer to relate to (at least some) others as equals, and for them relating to others as equals is good on a preference-­based account. However, some

26  This might not appear to make sense, as how well off people are relationally speaking is a zero-­ sum game such that if some people are superior to others then the latter must be worse off, relationally speaking, relative to a situation in which everyone related to one another as equals (cp. Rawls 2001: 131). However, thinking about relational standing in this manner is too narrow. Compare two scen­ arios: In the first, anything can be done to anyone provided this is required to maximize goodness. In this crude way, everyone has the same relational standing. In the second scenario one cannot sacrifice individuals to bring about marginally greater benefits for others, but what sacrifices one can impose on individuals to produce great benefits for others varies slightly such that people have slightly different relational standing. Arguably, in the first situation involving equal standing, people are worse off, relationally speaking, than in the second situation. We might say that everyone has greater moral worth or higher moral status in the second situation (cp. Kamm 1996: 259‒89; 2007: 227‒36). If this is so, how well off people are relationally speaking is not a zero-­sum game.

428  Kasper Lippert-Rasmussen prefer not to relate to others as equals and, presumably, levelling down relationally will not be better for them.27 In response to this observation, some might suggest that the relevant preference-­based account pertains to hypothetical rather than actual preferences. Presumably, there are people who prefer relating to others as superiors, who would instead prefer to relate as equals if they were better informed or more rational; for example, if they did not have false beliefs about the possibility of having deep personal relations to someone whom one treats as an inferior. However, that does not secure the stronger claim relational egalitarians need to defend; to wit, that no one who is suitably ideally informed and rational would prefer relational inequality to social equality. To show that this is so seems quite difficult, unless of course being relevantly informed includes being informed about what is objectively valuable for persons. This takes us to the objective-­list theory in a way that makes preferences epiphenomenal. Assuming that we endorse an objective list account of well-­being, on what grounds could relational egalitarians claim that relational equality is good for people such that levelling down will always be in one respect better for people? There are at least four ways in which relational equality might be part of what is objectively good for people. First, according to some, deep personal relations are constitutive of living well (Griffin 1986: 56‒92). Hence, if two persons live identical lives except for the fact that the one enjoys deep personal relations while the other does not, the former lives a better life. Relational egalitarians might then add that deep personal relations can—not as a matter of causality, but as a matter of the nature of deep personal relations—obtain only between persons who relate as equals (cp. Mill 1988 [1869]: 26).28 Second, relational egalitarians might say that living well is living in truth. Since people are moral equals, they live in truth only if they relate as moral equals. Hence, a slave owner who is being treated as a superior lives in untruth and, thus, lives in a way which, in one respect at least, is worse for him than if he lived together with all others as an equal. In this respect, eliminating slavery would be good for him. Third, it might be argued that freedom and autonomy are objectively valuable: ‘To be subject to another’s command threatens one’s interests, as those in 27  I set aside here the complication consisting of cases where, under favourable circumstances, two parties consent to one party dominating the other for a limited period of time. In such cases, arguably, the parties relate as equals, at a second-­order level, even if, at a first-­order level, they do not. The kind of case I speak to in the main text is one in which no such consent has been given. 28  Cp. ‘[On objective list accounts] our well-­being partly consists in . . . our having various good relations with some other people’ (Parfit 2017: 401); however, there is a complication. Suppose that everyone relates to so many significant others as equals that forming deep egalitarian, personal relations to more people would not make their lives better. Suppose also that they relate to some non-­ significant others as inferiors/superiors. If levelling down affects only the latter sort of relations, levelling down might be possible nevertheless.

Relational Egalitarianism  429 command are liable to serve themselves at the expense of their sub-­ordinates . . . Such a condition of subjection to the arbitrary wills of others is objectionable in itself, and has further objectionable consequences’ (Anderson 2008: 145‒6; cp. O’Neill 2008: 130n30; Tomlin 2015: 13n46).29 Since, by definition, deontic relational equality implies not being subject to others’ (arbitrary) commands, deontic relational equality is good for people.30 Fourth, Parfit suggests that ‘it is bad for people if they are servile or too deferential, even if this does not frustrate their desires or affect their experienced well-­being’ (Parfit 2002 86; cp. Rawls 2001: 131; O’Neill 2008: 123). If so, e.g. if it is bad for the happy servile slave to be a servile slave despite his desire to be one, the elimination of the social relations involved in such inegalitarian dispositions is good for some people in one respect at least.31 No doubt there are other items of an account of objective value for persons, which are tied by their very nature to relating to (significant) others as equals. The main attraction for relational egalitarians of an objective list account including such constituents of the good life is that, necessarily, relating as equals is good for people in some respect. Hence, levelling down might be impossible. No similar claim regarding distributive equality appears plausible.32 Since telic relational egalitarianism avoids, or at least stands a better chance of avoiding, two of the potential objections which telic distributive egalitarianism faces, this is some reason to favour telic relational over telic distributive relational egalitarianism (if we must choose). In Section 5, however, I explore a challenge to telic relational egalitarianism.

29  Cp. ‘servitude, except when it brutalizes, though corrupting to both, is less so to the slaves than to the slave-­masters. It is wholesomer for the moral nature to be restrained, even by arbitrary power, than to be allowed to exercise arbitrary power without restraint’ (Mill 1988 [1869]: 82; cp. Anderson 2012: 50). 30  Telic relational egalitarianism implies that social equality might obtain if there is constant vari­ ation in who dominates and who gets dominated such that the instability of relations of domination prevents it from ever turning into a social hierarchy. 31  Interestingly, all four reasons mentioned here for why relational inequality can be bad for people imply that relational inequality which does not reflect social inequality, i.e. does not take place on the background of hierarchical social norms, expectations, etc. but simply reflects idiosyncratic features of the individual instance of relational inequality, can be bad for people. This raises the question of whether the site of relational equality extends beyond social equality to individual inegalitarian interactions. Such inequalities might be less of a concern because they are likely not to be cumulative and systematic. This makes an important difference, since being systematically targeted means that an individual act of domination, say, is likely to be much more harmful to the dominatee than it is to the dominatee of an isolated but otherwise identical act of discrimination; e.g., because only in the former case can the dominatee understand her situation as the expression of a hostile social environment. That, however, does not imply that isolated unequal social relations merit greater concern per se than systematic unequal social relations. For present purposes, however, we can disregard the issue of the site of relational equality. 32  Cp. ‘Why would it matter if, in some far off land, and quite unknown to me, there are other people who are better off than me?’ Parfit (1998: 13; 1995: 29).

430  Kasper Lippert-Rasmussen

5.  Badness of Social Inequality and Population Size In this section, I develop a challenge to telic relational egalitarianism, to which there is no plausible analogue challenge to deontic relational egalitarianism. To do so, I first note that there are two ways to articulate telic relational egalitarianism, when understood as a view about impersonal value. One is the way I employed in Section 4, which I now relabel: Negative telic relational egalitarianism:  It is in itself (unjust and therefore) bad if social inequality exists.

The other states a closely related view: Positive telic relational egalitarianism:  It is in itself (just and therefore) good if social equality exists.33

Consider a world with no people in it. Compare the value of this world to a world that contains people and in which social equality obtains spotlessly. Suppose the two worlds are identical in all other relevant respects. According to negative telic relational egalitarians, the two worlds are equally good. According to positive telic relational egalitarians, the latter world is better than the former. We might think this speaks in favour of positive telic egalitarianism. A taxonomy merely based on these two views is probably too simplistic for the following reason. Compare a world in which many people relate and almost do so as equals with one in which there is no people. According to positive telic relational egalitarianism, the former world is no better than the latter, and according to negative telic relational egalitarianism, the former world is worse. However, one might plausibly think that the former world is better than the latter. In view of this case, consider the following: Graded positive telic relational egalitarianism: It is in itself (just and therefore) good if social equality exists. A state of affairs still contains positive relational value if it deviates somewhat from perfect social equality, but the greater the deviation the less relational value. At some threshold degree of deviation, the state of affairs imperfectly realizing social equality has zero relational value and any increase in the degree of social inequality will result in a state of affairs that  has negative relational value, the greater the deviation the more negative relational value.34

33  A similar distinction applies to telic distributive egalitarianism. 34  One could take a similar view regarding the value of distributive equality.

Relational Egalitarianism  431 This view gives us the intuitively right result regarding the relevant comparison: the world in which many people relate almost as equals is better, social relations-­ wise, than a world in which no one exists. If we accept either of the positive telic relational egalitarianisms, there is a further dimension in which we need to specify our view. For generally when relational egalitarians think about the value of relational equality, they have situations in mind in which people relate to one another socially and then ask what difference it makes, axiologically speaking, whether social equality obtains. However, there are other situations to consider. First, there are cases in which existing ­people have no social relations but can develop them. Second, there are cases in which people are brought into existence and then start relating to others as equals. Especially in the latter of the two cases, one does not intuitively make the world a better place by bringing more people into existence, whom one then relates to as an equal, if there already exist plentiful people relating as equals. No relational egalitarian has ever favoured an expansive population policy on such grounds.35 Hence, on that assumption, telic relational egalitarians face a dilemma: Either they embrace negative or they embrace positive relational egalitarianism. On the former horn of the dilemma, they are committed to the implausible view that a world with no people is better than one with lots of people, all of whom relate to each other almost as equals. On the latter horn, they are committed to the equally implausible view that by bringing more people into existence or by starting to interact with others with whom one had no prior interactions, and then relating as equals (or almost as such), one makes the world better, even in a world where many people already relate as equals. Perhaps there are ways to develop telic relational egalitarianism so that it avoids this dilemma. Or perhaps, on reflection, one of the two horns of the dilemma is not so unattractive; for example, it might be held that if the additional people’s lives are good in other ways as well, then their relating to others as equals is a constituent of the additional impersonal good realized by bringing more people into existence. But a third option also exists: to embrace deontic relational egalitarianism.

6.  Deontic Relational Egalitarianism According to deontic relational egalitarianism, social equality might or might not in itself be good. The defining feature of this view is that it holds that it is often bad if social inequality exists, because such inequality often results from (or, in

35  My thought here is not that one can never make the world a better place by bringing people into existence who have certain positive features, e.g. a high level of well-­being. My thought concerns specifically whether one can do so by bringing new people into existence who stand in egalitarian social relations to others, when the putative value of such relations does not consist in value for them.

432  Kasper Lippert-Rasmussen part at least, consists in) violations of distinctively egalitarian duties.36 One can understand such duties in two ways. Suppose I can either dominate Al or refrain from doing so, in which case five other agents will each engage in (otherwise) five morally equivalent acts of domination of other dominatees. Does deontic relational egalitarianism tell me to engage in or refrain from dominating Al? As previously formulated, this view can either be read agent-­neutrally (as offering the former answer) or agent-­relatively (the latter answer).37 A similar question does not naturally arise in relation to telic relational egalitarianism, since the value of relational equality is naturally thought of as impersonal, in which case everyone has reason to promote a maximum degree of relational equality, to wit, by performing a minimizing violation of a distinctively egalitarian duty. If so, that seems a strong reason in favour of deontic relational egalitarianism, since people naturally take an agent-­relative view of obligations of justice. For instance, people will object to instances of oppression that minimize the overall number of instances of oppression; they think that agents should not oppress others (almost full stop), not that agents should minimize the number of instances of oppression.38 Hence, the most natural version of deontic relational egalitarianism is:39 Agent-­relative deontic relational egalitarianism: it is bad if social inequality exists, because such inequality often results from violations of distinctively egalitarian duties.

To some, Parfit included, the very idea of a prohibition against minimizing the number of instances of a morally objectionable form of behaviour has an air of irrationality about it (Parfit 1984: 95‒8; Scheffler 1982: 80‒114). Hence, on Parfit’s view, agent-­relative deontic relational egalitarianism is problematic. I shall not here enter into the huge discussion of the justifiability of agent-­ relative moral constraints such as those which are involved in distinctively egalitarian duties. I simply note that the concern for relational equality intuitively takes this form. Parfit might also agree, as shown, inter alia, by the fact that the title of the section in Reasons and Persons in which he criticizes agent-­relative moral constraints is ‘How Common-­Sense Morality Is Directly Self-­Defeating’. Instead, I want to draw attention to a complication that is captured by neither telic nor deontic relational egalitarianism. Compare the following three states:

36  Cp. the formulation of deontic relational egalitarianism in Section 3. 37  Similarly, one can distinguish between time-­relative and time-­neutral understandings of dis­ tinct­ive­ly egalitarian duties. 38  ‘Almost full stop’, because deontic relational egalitarians might believe, say, that it is permissible to dominate, etc. someone when the consequences of not doing so are sufficiently bad. 39  An agent-­relative version of Parfit’s deontic distributive egalitarianism is probably the most natural version as well, although his formulation leaves this an open question.

Relational Egalitarianism  433 One-­way domination:  X dominates Y for their entire lives. Two-­way domination:  X dominates Y for half of their lives, and Y dominates X for the other half of their lives. No domination:  X and Y relate as equals at any given point in time.40

If we merely understand telic relational egalitarianism as meaning that dom­in­ ation is bad in itself, we should be indifferent between one-­way and two-­way domination; both states involve domination—and equally much of it. Intuitively, however, if we must choose between bringing about a state of one-­sided dom­in­ ation and a state of bidirectional domination, we might think that it is worse to bring about the former. If so, we should revise telic relational egalitarianism such that it says: Distribution-­sensitive telic relational egalitarianism:  It is in itself (unjust and therefore) bad if social inequality exists. It is worse if, in all instances of social in­equal­ity, the same people are treated as inferiors and the same people are treated as superiors than if people are treated as inferiors in equally many instances of unequal social relations and if they are treated as superiors in equally many instances of unequal social relations.

On distribution-­sensitive telic relational egalitarianism, relational inequality is in itself bad. However, how bad it is depends on something that is external to the mere fact that relational inequality obtains, to wit, whether people take positions in hierarchical relations in an egalitarian (i.e. everyone is at the bottom for same period of time, etc.) or in an inegalitarian manner (the same people are always at the bottom, etc.). Consider next agent-­relative, deontic relational egalitarianism. On this view, both one-­ way and two-­ way discrimination involves violations of egalitarian duties, thus leaving us with no grounds for saying that the latter is better than the former. Hence, if one-­way domination is worse than two-­way domination, this shows that, in a certain sense, agent-­relative, deontic relational egalitarianism needs to be combined with telic relational egalitarianism. In a certain abstract sense, it is a good not to be subjected to domination. It is a good in the sense that it is good for a person not to be dominated. Moreover, that is a good that can be distributed in such a way that we can rank people in terms of how much of that good they enjoy; ceteris paribus, a more equal distribution of that good is a better one. Agent-­ relative deontic relational egalitarianism would also appear to embrace distribution-­sensitive telic relational egalitarianism. Fortunately, the two views are, as I have argued, not inconsistent. 40  Cp. Temkin (1993: 235‒6).

434  Kasper Lippert-Rasmussen Combining deontic with telic egalitarianism would also address an important limitation of deontic relational egalitarianism, which shows that it cannot capture all of egalitarian justice. That limitation exists provided the following claim is true: In some circumstances, we have duties to relate socially to others and then relate to them as equals. Such cases exist. Suppose we discover some individuals living in social isolation. We could improve their appalling situation by interacting with them. Since deontic relational egalitarian norms apply only to those cases where people relate socially to one another, they do not impose such a duty on us.41 Telic relational egalitarianism could potentially ground a duty to relate to these people.42 One might make an even more radical claim and argue that deontic relational egalitarians must adopt an even more pluralistic approach to egalitarian justice. For suppose there is such a thing as intergenerational (egalitarian) justice.43 This would mean that we can act unjustly toward future generations by acting in such a manner that they are much worse off than we are. Since, in the nature of the case, we cannot be socially related to distant future generations, there is no way such a duty can be grounded in deontic relational egalitarianism (Lippert-­ Rasmussen 2018a: 123‒30). Accordingly, such a duty must be grounded elsewhere, and some kind of telic distributive view—an egalitarian one to the extent that we have egalitarian intuitions about the distribution across generations, whoever their members turn out to be—seems to be the best candidate. What this comes down to is that while agent-­relative deontic egalitarianism probably captures our concern for relational equality best, it cannot stand on its own. To capture all of our pertinent egalitarian intuitions, it must be supplemented with some form of telic egalitarianism, distributive as well as relational. Hence, the present essay is a plea for a pluralist egalitarian position.

7.  Conclusion Parfit (1998: 20) modestly ends his exciting piece on equality thus: ‘Taxonomy, though unexciting, needs to be done. Until we have a clearer view of the alternatives, we cannot hope to decide which view is true, or is the best view’. I have argued that Parfit’s taxonomy is also instructive when applied to recent relational egalitarian views, and not just in relation to what was his concern; to wit, distributive egalitarian views. However, while telic and deontic relational egalitarianism are ‘alternatives’ in the sense that they are different views, they are 41  Alternatively, deontic relational egalitarians could appeal to other egalitarian duties to fill the present gap. 42  Admittedly, there are likely to be non-­egalitarian reasons why we should initiate social relations to these people. However, what I am suggesting is that we have egalitarian reasons to do so (also). 43  A similar point applies to justice and individuals who are not persons.

Relational Egalitarianism  435 not ‘alternatives’ in the sense that it is not the case that both of them can be true. And while both face certain challenges, to account for all intuitions regarding the value of relational equality, we should probably embrace both. Moreover, I have argued that telic relational egalitarianism might be immune to the levelling-­down objection given a range of different objective list theories of well-­being, but, due to the agent-­relative nature of our concern with relational egalitarian norms, telic relational egalitarianism must be supplemented with deontic relational ­egalitarianism to capture the intuitive shape of our concern with relational equality.

Bibliography Anderson, E. (1999). ‘What Is the Point of Equality?’ Ethics 109/2: 287–337. Anderson, E. (2008). ‘Expanding the Egalitarian Toolbox: Equality and Bureaucracy.’ Aristotelian Society, supplementary vol. 82/1: 139–60. Anderson, E. (2010). ‘The Fundamental Disagreement between Luck Egalitarians and Relational Egalitarians.’ Canadian Journal of Philosophy, supplementary vol. 40: 1–23. Anderson, E. (2012). ‘Equality.’ In The Oxford Handbook of Political Philosophy, edited by D. Estlund, 40–57. Oxford: Oxford University Press. Arneson, R. (1989). ‘Equality and Equal Opportunity for Welfare.’ Philosophical Studies 56/1: 77–93. Cohen, G.  A. (2011). On the Currency of Egalitarian Justice, and Other Essays in Political Philosophy. Princeton: Princeton University Press. Dworkin, R. (2000). Sovereign Virtue. Cambridge, MA: Harvard University Press. Goodin, R. (2013). ‘Disgorging the Fruits of Historical Wrongdoing.’ American Political Science Review 107: 478–91. Griffin, J. (1986). Well-Being: Its Meaning, Measurement and Moral Importance. Oxford: Clarendon Press. Hirose, I. (2015). Egalitarianism. London: Routledge. Kamm, F. M. (1996). Mortality, Morality, ii. Rights, Duties, and Status. Oxford: Oxford University Press. Kamm, F.  M. (2007). Intricate Ethics: Rights, Responsibilities, and Permissible Harm. Oxford: Oxford University Press. Kolodny, N. (2014). ‘Rule Over None II: Social Equality and the Justification of Democracy.’ Philosophy & Public Affairs 42: 287–336. Lippert-Rasmussen, K. (2007). ‘The Insignificance of the Distinction between Telic and Deontic Egalitarianism.’ In Egalitarianism: New Essays on the Nature and Value of Equality, edited by N.  Holtug and K.  Lippert-Rasmussen, 101–24. Oxford: Clarendon Press. Lippert-Rasmussen, K. (2015). ‘Luck Egalitarianism versus Relational Egalitarianism.’ Canadian Journal of Philosophy 45/2: 220–41.

436  Kasper Lippert-Rasmussen Lippert-Rasmussen, K. (2018a). Relational Egalitarianism: Living as Equals. Cambridge: Cambridge University Press. Lippert-Rasmussen, K. (2018b). ‘(Luck and Relational) Egalitarians of the World, Unite!’ Oxford Studies in Political Philosophy 4: 81–109. Mill, J. S. (1988 [1869]). The Subjection of Women. Indianapolis: Hackett. Miller (1998). ‘Equality and Justice.’ In Ideals of Equality, edited by A. Mason, 21–36. Oxford: Blackwell. Nagel, T. (2005). ‘The Problem of Global Justice.’ Philosophy and Public Affairs 33/2: 113–47. O’Neill, M. (2008). ‘What Should Egalitarians Believe?’ Philosophy & Public Affairs 36/2: 119–56. Parfit, D. (1984). Reasons and Persons. Oxford: Clarendon Press. Parfit, D. (1995). Equality or Priority? Lindley Lecture 1991. Department of Philosophy, University of Kansas. Parfit, D. (1998). ‘Equality and Priority.’ In Ideals of Equality, edited by A.  Mason, 1–20. Oxford: Blackwell. Parfit, D. (2002). ‘Equality or Priority?’ In Ideals of Equality, edited by M. Clayton and A. Williams, 81–125. Houndmills: Palgrave Macmillan. Parfit, D. (2011). On What Matters. Vol. ii. Oxford: Oxford University Press. Parfit, D. (2017). On What Matters. Vol. iii. Oxford: Oxford University Press. Rawls, J. (2001). Justice as Fairness. Cambridge, MA: Harvard University Press. Scanlon, T. M. (2018). Why Does Inequality Matter? Oxford: Oxford University Press. Scheffler, S. (1982). The Rejection of Consequentialism. Oxford: Clarendon Press. Scheffler, S. (2003). ‘Equality as the Virtue of Sovereigns: A Reply to Ronald Dworkin.’ Philosophy and Public Affairs 31/2: 199–206. Scheffler, S. (2005). ‘Choice, Circumstance, and the Value of Equality.’ Politics, Philosophy, and Economics 4/4: 5–28. Scheffler, S. (2015). ‘The Practice of Equality.’ In Social Equality: On What It Means to Be Equals, edited by C.  Fourie, F.  Schuppert, and I.  Walliman-Helmer, 21–44. Oxford: Oxford University Press. Schemmel, C. (2011). ‘Why Relational Egalitarians Should Care about Distributions.’ Social Theory and Practice 37/3: 365–90. Schemmel, C. (2012). ‘Distributive and Relational Equality.’ Politics, Philosophy & Economics 11/2: 123–48. Temkin, L. (1993). Inequality. Oxford: Oxford University Press. Tomlin, P. (2015). ‘What Is the Point of Egalitarian Social Relationships?’ In Distributive Justice and Access to Advantage: G. A. Cohen’s Egalitarianism, edited by A. Kaufman, 151–79. Cambridge: Cambridge University Press. Viehoff, D. (2014). ‘Democratic Equality and Political Authority.’ Philosophy & Public Affairs 42: 337–75.

Relational Egalitarianism  437 Viehoff, D. (2019). ‘The Value of Political Equality.’ Oxford Studies in Political Philosophy 5. Oxford: Oxford University Press. Wolff, J. (1998). ‘Fairness, Respect, and the Egalitarian Ethos.’ Philosophy & Public Affairs 27/2: 97–122. Young, I.  M. (1990). Justice and the Politics of Difference. Princeton: Princeton University Press.

20

Duties That Become Supererogatory or Forbidden? F. M. Kamm

In the last part of volume iii of On What Matters,1 Derek Parfit explains his view that if morality is deontological, it faces the threat of being undermined by prac­ tical reason (which answers the question, ‘What do I have most reason to do?’). He tries to save morality from this threat by showing that morality is not de­onto­ logic­al in the way commonly thought. He does this in part by presenting general arguments for the conclusion that parts of Common Sense Morality are com­pat­ ible with consequentialism. In this article, I will first describe some relations between making personal sacrifices, fulfilling personal duties, and promoting the greater good that are the basis for Parfit’s general arguments. I will then focus on whether fulfilling certain duties could require us, contrary to the relations described, to make great sacrifices to promote the greater good. I will first con­ sider this question with regard to duties to aid. Then after comparing these duties with ones not to harm with which Parfit is concerned, I will consider it with regard to duties not to harm. The article concludes by considering Parfit’s view on whether a deontological permission to harm another varies with permission of other agents to also harm.2

1.  Measures of Stringency Parfit first argues against the deontological principle that not harming is a more stringent duty than aiding. He notes there are at least two possible measures of Earlier versions of this article were presented in May 2019 at the Workshop on Rights and Supererogation at MIT, in July 2019 at the Australian National University, and in February 2020 at Claremont McKenna and Pomona Colleges, and also at a conference in honour of Marshall Cohen at the University of Southern California. I am grateful to Joe Bowen, my commentator at MIT and to the audiences on all four occasions for helpful suggestions. I also had helpful discussion with Shelly Kagan, Jeff McMahan, and Larry Temkin, and helpful written comments from Ketan Ramakrishnan. 1  (Oxford: Oxford University Press, 2017). 2  I discuss other aspects of his arguments in volume iii of On What Matters against deontological distinctions in my ‘Parfit on the Irrelevance of Deontological Distinctions’, in M.  Timmons (ed.), Oxford Studies in Normative Ethics (Oxford: Oxford University Press, 2020), 9–31. F. M. Kamm, Duties That Become Supererogatory or Forbidden? In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0021

442  F. M. Kamm such proposed stringency: (a) the agent is morally required to bear greater costs to avoid causing harm than to aid; and (b) not harming takes priority over aiding when these conflict. His further claim is that it is possible that not harming can be more stringent by the first measure (cost) and yet be less stringent by the second measure (priority). That the measures of cost and priority can come apart was one major point of my article ‘Supererogation and Obligation’.3 I there argued that one might, for example, have to bear greater costs to fulfil one’s contractual business duty (which is a positive duty correlative to a positive contractual right) than to help save someone’s life, yet one might also permissibly give priority to saving someone’s life over fulfilling the contractual duty when they conflict. (Call this the Business Case.) This is true, I argued, whether saving someone’s life is what is called a simple natural duty done at low cost to oneself or if it is supererogatory because of the high cost to oneself.4 For example, one might miss an important business deadline to jump into shark-infested waters to save someone’s life. I argued that this is true even when the person to whom the contractual duty is owed would not be morally obligated to waive it so that someone else may be saved, because of the high cost to him of doing so. However, the cases that I used to show how the two measures of stringency can come apart did not involve a choice between saving some and not physically harming someone else. It is the latter sort of case, which involves a negative duty not to harm correlative to a so-called natural, non-contracted negative right not to be harmed, that concerns Parfit. Nevertheless, I did note (what Parfit empha­ sizes) that not harming being more stringent by the cost measure does not alone show it is more stringent by the priority measure.5 Parfit may emphasize the difference between these two measures of stringency because he thinks that rejecting the priority within morality of not harming over aiding does not imply that morality requires one to make great personal sacrifices for purposes of aiding. Though he does not say so, his aim seems to be to avoid the so-called Over-Demandingness Objection to Consequentialism by retaining 3  Journal of Philosophy 82/3 (1985): 118–38. A revised version appears as ch. 15 in Morality, Mortality, ii. Rights, Duties, and Status (Oxford: Oxford University Press, 1996). 4  I claimed that saving at high cost to oneself remains supererogatory and not a duty even when one is willing to make the effort. If it became a duty one could not refuse to save at the high cost to which one does not object when the opportunity, for example, to win an important contest comes along if the contest would not override the duty to save at small cost to oneself. Yet I think the contest could override saving at high cost. 5  In ‘Supererogation and Obligation’, 125 n, I said: ‘Because an act that is less stringently required by one measure may take precedence over another act that is more stringently required by that same measure, proving that negative duties are more stringently required than positive duties in the sense that we must make a greater effort to perform them, does not resolve the problem of whether we may violate a negative duty to perform a positive one. For example, Philippa Foot says that we may not kill one person to save others, because the duty not to kill is more stringent than the duty to aid (in “The Problem of Abortion and the Doctrine of Double Effect,” reprinted in Bonnie Steinbock, ed., Killing and Letting Die (Englewood Cliffs, NJ: Prentice-Hall, 1980) ). But that greater stringency may mean only that we would have to make a greater effort to avoid killing someone than to aid others, and this does not imply that if we want to we may not violate a negative duty in order to aid.’

Duties That Become Supererogatory  443 Common Sense Morality’s non-consequentialist personal prerogative not to make large personal sacrifices to achieve the greater good, thus making such sac­ rifices supererogatory.6 However, he wants to combine this with jettisoning Common Sense Morality’s deontological side constraint on not harming others to achieve the greater good. This would permit those who would not sacrifice them­ selves for the greater good to try to sacrifice another for its sake. Parfit does not rule out that these positions are compatible with costly self-sacrifices being required and not supererogatory to avoid harming others when only one’s own interests and no greater good are at stake. It might seem that if Parfit were correct that not harming has no priority over aiding, then we should not be required to bear higher personal costs to avoid harming than to aid. But I argued in ‘Supererogation and Obligation’ that it is intuitively plausible that achieving greater good could sometimes take precedence over positive moral duties such as keeping promises, and yet one may not avoid the personal sacrifices necessary to meet these minimal duties even if one is not morally required to make such sacrifices to go beyond minimal morality and achieve the greater good. The minimal duties can be weaker relative to going beyond minimal morality than are my interests and yet my interests are weaker relative to these duties. Hence, Parfit’s applying such a view to negative duties yields the same sort of (at least seeming but explicable) intransitivity that I identi­ fied in ‘Supererogation and Obligation’ in the case of positive duties. To see this, combine his and my original view to get the following set of relations: Avoiding personal sacrifices (S) could take precedence over (>) aiding (A), or (S>A); aiding some could take precedence over the negative duty not to harm (N) someone else as well as over performing a positive (P) duty (D) for someone else, or (A>DPN); and yet not harming someone else as well as performing a positive duty for some­ one else could take precedence over avoiding personal sacrifices, or – (S>DPN). I call this only a ‘seeming intransitivity’ because different factors account for precedence relations in the different steps; there is no single relation that is shown to be intransitive. That is, since it is not because one would be doing the greater good in fulfilling a personal duty that the duty can dominate personal interests in the conclusion, the fact that personal interests can dominate the greater good and greater good can dominate personal duty need not imply that personal interests can dominate personal duty. Suppose that in this system of relations between duty, personal sacrifices, and the greater good, there is a case in which one has a duty to make significant per­ sonal sacrifices rather than harm someone else and also to aid someone. Paying personal costs rather than harming someone could come about by either (i) not doing something to someone else that would prevent losses with which one is already threatened (e.g., one would not throw someone else in front of a threat to 6  Such a prerogative is argued for in Scheffler’s The Rejection of Consequentialism (Oxford: Oxford University Press, 1982).

444  F. M. Kamm prevent one’s being hit by it) or (ii) preventing harm to someone else from a threat that one would initially present to him (e.g., one would swerve to avoid running over someone). In (i) but not (ii) if one did not harm someone, one would suffer personal losses independently of trying to prevent harming them. Parfit does not consider whether the personal duty to make such large sacrifices rather than harm another might imply having a duty to make them to achieve the greater good instead of harming someone else for this purpose when this is possible.7 Similarly I did not consider whether the duty to make such large sacrifices to fulfil a personal duty to aid another might imply having a duty to make them to achieve the greater good without abandoning the person one must aid when this is pos­ sible. In what follows I will first deal with my omission and then with Parfit’s.

2.  Duty to Supererogatory Consider what I will call the Personal Lifeguard Case: Someone’s personal life­ guard has a duty to prevent his client drowning not merely at small cost to himself but even at the cost of his own life. This duty is correlative to the client’s con­ tracted positive right to be saved. (I assume it is morally possible to contract for such high sacrifice as bodyguards can be morally required to take a bullet for a client.) I shall assume that the client would have gone in the water even if he had no personal lifeguard and no one else but the actual lifeguard applied for the job. A big wave now threatens to drown the client and also one hundred other people at some distance from him. If the lifeguard saves the one person, he cannot save the hundred and vice versa. He does not have either a contracted or a simple (natural) duty to save the hundred at the cost of his life. I think it is permissible for him to save the hundred people and abandon the client whether or not the effort involved in saving them would be as great as, more than, or less than what is morally required to save the client.8 Given what I have assumed, the lifeguard’s failure to aid his client would be a case of letting his client die and not a case of killing him, since the lifeguard has not made his client worse off with respect to loss of his life than he would have been without the lifeguard’s commitment. In this case, the one person does not have a duty to waive his right against the lifeguard so that the lifeguard may save the others. The client would also not have a duty to sacrifice his life (which he will lose if the lifeguard fails him) to save the hundred other people. Hence, I think we 7  I am grateful to Ingmar Persson for pressing this point in discussion. 8  This is so even if the act of saving the hundred is not a morally exemplary act due to bad mo­tiv­ ation. For example, suppose it is only because he hates his client or only because the cost to save the one hundred is less than what he would have to pay to save his client that he saves the hundred. The properties that make doing one act rather than another permissible can still be present to justify doing the act even if an agent does not do the act for the sake of the properties that justify doing it.

Duties That Become Supererogatory  445 cannot account for the permissibility of the lifeguard not saving his client on the grounds that there is no duty still in existence to his client that he does not at least infringe.9 However, suppose that if and only if the lifeguard employed a means to save the hundred that cost him his life, he could also save his client. (Call this the Personal Lifeguard Both Case.) Suppose in this case he would not be willing to pay this cost as a supererogatory sacrifice to save the hundred alone though he would be willing to make a much smaller supererogatory sacrifice to do so. But he would be willing to sacrifice his life as morally required by his client’s right were this necessary (which it is not) to save his client alone. The question is whether he could no longer permissibly abandon his client to save the hundred at small or minimally supererogatory cost when there is a way for him to also fulfil his duty to his client by saving the hundred at a cost that is as great but no greater than what could be morally required of him to save his client alone. So must he avoid failing in his duty as he saves the hundred or else not save the hundred and fulfil his duty to his client at a low cost? It may seem that he could permissibly refuse to make the largest sacrifice and just do the minimum necessary to save his client or that he could choose to make the bigger sacrifice that will save his client and the one hundred, but that he may no longer abandon his client. An alternative view which I shall call Still Abandon Client is that in Personal Lifeguard Both it is permissible for the lifeguard to abandon his client. Here is a particular defence of it: Saving the hundred lives could be a good enough reason for the lifeguard to leave his client to die as shown by what the lifeguard may do when there is no possibility of saving everyone. Furthermore, if he would not save the hundred were he morally required to also save his client at the high cost of his life, the outcome that could have overridden the duty to his client will not come about. It may seem odd to have to sacrifice an outcome that could have overrid­ den the duty just so that he will perform the duty when he is only trying not to lose his life in both fulfilling his duty and saving the hundred whom he had no duty to save at that high cost. (Call this the Oddity Defence.) If the Still Abandon Client View were correct, it would show that in this case a duty to make an other­ wise morally-required sacrifice for one’s client becomes supererogatory. In this case, saving his client at the high cost of giving up his own life becomes necessary only because the lifeguard chooses to save the hundred which is what it is permissible (and even right) for him to do rather than save his client when they cannot both be accomplished. The cost to him becomes so great because he would be saving the one hundred though he would not be willing to pay the cost in order 9  Perhaps the number saved should be a thousand (which is the number I used in ‘Supererogation and Obligation’ in discussing a similar case) instead of a hundred to make it permissible for the life­ guard to abandon his client. I am not concerned with the specific number in this example. I am just seeking a number such that (a) the lifeguard may save them instead of his client while (b) the client has no moral duty to release the lifeguard from saving him so that he may save more people instead.

446  F. M. Kamm to save the hundred alone. By contrast, what if it would actually have been physically necessary for him to give up his life in order to save his client alone had he not chosen to save the one hundred, and he can now save both at this high cost? In this case the high cost predates his saving the one hundred. (Call this case Personal Lifeguard Both Pre-Cost.) Then it does indeed seem that he should give up his life to save both at this same cost. In this case, it is not true that he needn’t save both just because when only one option could be achieved, saving the hun­ dred even at low cost would override the contractual duty he would have been required to do even at high cost. It seems the difference between Personal Lifeguard Both Pre-Cost and Personal Lifeguard Both is how the high cost comes to be needed to save his client. In Personal Lifeguard Both Pre-Cost it is independent of, but in Personal Lifeguard Both dependent on, doing the act that could permissibly override the dutiful act to his client if only one of them could be done. However, there is another difference between the cases to consider. In Personal Lifeguard Both Pre-Cost the high cost to the lifeguard is causally needed to save his client. By contrast, in the Personal Lifeguard Both Case as described the high cost is causally needed to save the hundred if he is to also save his client in a caus­ ally non-costly way. I do not think this is a morally significant difference. After all, there could be a version of the Personal Lifeguard Both Case in which it is only because the lifeguard chooses one non-costly way rather than another to save the one hundred that he can also save his client but only at the cost of his life. I think the same question of whether he must do what allows him to also save his client arises as in the Personal Lifeguard Both Pre-Case even though the high cost would have a causal role in saving his client and not the one hundred. So we should consider further whether it could be the difference in how the high cost to save both comes about that is sufficient to make a moral difference to whether the lifeguard may abandon his client. Consider a case called Personal Lifeguard Both Post-Cost: The lifeguard abandons his client, whom he could have saved at a low cost, to save the hundred at the only time this is possible and at low cost when there is no way to help both. His client does not immediately drown and he can still save his client after he saves the one hundred. However, because he saved the hundred first, saving his client now involves sacrificing his life. In this case, even though the high cost to also save his client arises because of his saving the one hundred, it may seem that he is morally required to pay the cost. After all, many dutiful or non-dutiful but morally worthwhile things we do in life may raise the eventual cost to us of doing our contractual duties at a later time, yet this does not relieve us of the duty to pay the higher costs. If so, how could the fact that the high cost to save his client arises from his saving the one hundred be sufficient, as suggested earlier, to morally distinguish Personal Lifeguard Both and Personal Lifeguard Both Pre-Cost? Perhaps the temporal difference between Personal Lifeguard Both and Personal Lifeguard Both Post-Cost plays a role in justifying the Still Abandon Client View

Duties That Become Supererogatory  447 for the former case alone.10 That is, in both cases the lifeguard’s act of saving the hundred results in there being a high cost to him of saving both the hundred and his client, but he must pay that cost at different times in each case: In Personal Lifeguard Both, he would have to pay it concurrent with performing the act that could override his requirement to save his client, whereas that act has already occurred by the time he must pay the high cost in Personal Lifeguard Both PostCost. Hence, in Post-Cost there is nothing that overrides his duty to his client at the time he must perform the duty. But this analysis in terms of time does not seem fundamental. Rather it seems fundamental that in Personal Lifeguard Both the high cost would have to be paid to prevent a bad effect (i.e., the client drown­ ing) that would be caused by the doing of the very act that justifies that bad effect. (Call this the Causal Explanation of the Still Abandon Client View.) By contrast, in Personal Lifeguard Both Post-Cost the high cost would not be paid to prevent a bad effect caused by the doing of an act that justifies that bad effect since that act has already been done without causing the client to be drowned. What would cause the bad effect is crucial, not merely what causes the high cost to prevent the bad effect. Nevertheless, suppose that the lifeguard in Personal Lifeguard Both Post-Cost could foresee the high cost he would be morally required to pay to save his client if he first saved the one hundred. He might then reasonably refuse to save the one hundred. Even though saving them would not involve a large sacrifice in itself, he would have to lose his life eventually because he saved them. Once again, as emphasized in the Oddity Defence of the Still Abandon Client View, we would lose the good outcome that would have morally overridden the duty to save his client when both could not be done and this may seem odd. But this Oddity Defence of why Personal Lifeguard Both and Personal Lifeguard Both Post-Cost should be treated as morally alike seems too consequentialist, aiming merely to not undermine achieving the best outcome.11 By contrast the Causal Explanation of why Personal Lifeguard Both Post-Cost and Personal Lifeguard Both are mor­ ally different focuses on whether the high cost would have to be paid to prevent a bad effect caused by the doing of the very act that justifies the bad effect. This occurs only in Personal Lifeguard Both. The Causal Explanation seems to me a better justification of the Still Abandon Client View. What if it would take only a bit more effort than the lifeguard is required and prepared to make to save the hundred on their own in order to save both them and his client in another version of Personal Lifeguard Both? The total effort would still be within what could be required of the lifeguard to save his client

10  I am grateful to Casper Hare for discussion about the issue of time. 11  There could be a similar concern sometimes about justifying a course of conduct on the grounds that it would be odd to forgo an outcome that is best overall from the point of view of someone’s own self-interest.

448  F. M. Kamm alone. In this case, I think, the fact that saving the hundred is a good reason for not saving his client when both cannot be done would not relieve him of duty of also saving his client by paying the higher cost. This is so even if the higher cost and the bad effect of abandoning his client would be caused by the doing of the act that overrides the duty to his client and for which on its own the larger cost is supererogatory. After all, he will have infringed his client’s right in abandoning him to save the hundred (even if he acts permissibly and so does not violate his client’s right).12 This means, I think, that he could wrong his client even if he does not do a wrong act and he should pay some price to avoid such wronging if not the very great one required were he not saving the one hundred. Not doing so would itself be wrong. In sum, on the Still Abandon Client View deciding whether the lifeguard mor­ ally must pay more to save both parties in Personal Lifeguard Both than to only save the hundred does not involve considering what is the most the lifeguard owes to his client independent of his doing what could override his duty at that time. Rather, deciding involves considering the additional price he would have to pay to avoid the bad effect that is caused by the doing of what could override his duty to his client. If the price is low, he should pay it to also fulfil the duty to his client; it is not supererogatory with respect to his client. However, if the price is high, he needn’t pay it. The basic idea (roughly) is that when only his personal interests compete with fulfilling his lifeguard’s duty, he cannot refuse to make large sacrifices. But when achieving some greater good that overrides his duty to his client would threaten non-performance of the duty and also increase the cost of performing the duty, he should not be required to make the very large sacrifice to also save his client. Paying such a cost to save his client, which in some circum­ stances is a duty, becomes supererogatory.13 However, this is not the same as saying that he needn’t save his client at any cost to himself that would justify his not actually bringing about the greater good that overrides his duty. For the cost that could justify him in not bringing about the greater good is a supererogatory cost for that purpose. But this cost could be the result of even a small additional cost he would have a duty to bear to avoid wronging his client. Saving the one hundred will not justify refusing to pay the 12 ‘Violation’ refers to impermissible failure of a duty or transgression of a right; ‘infringing’ includes permissible transgression of a duty or right. 13  Suppose the lifeguard was morally required to lose up to two limbs to save his client but had already supererogatorily sacrificed one leg to save the hundred. Perhaps he is not still required to sac­ rifice a leg and an arm to save his client if this is possible. This would turn giving up two limbs for his client, which had been a duty, into a supererogatory act. In this case, what he does for others might affect what he is morally required to do for his client because the baseline for determining his duty to his client might vary with the physical condition he is already in. This is a different way in which what was previously a duty could become supererogatory. Adrienne Martin has pointed out that by contrast to what I have claimed about the lifeguard, a parent’s duty to sacrifice himself to save his child seems to remain in place even if it is his dong what would override that duty that threatens the abandonment of the child and increases the cost to save her. An explanation is needed if there is this contrast.

Duties That Become Supererogatory  449 small extra cost to avoid wronging his client the way saving the one hundred jus­ tified not saving the client when both could not be done. Hence someone could decide to avoid wrongfully omitting to bear the extra cost to save both by not saving the one hundred. Whether this would be a morally wise decision I leave unanswered. We might represent part of our conclusion (where ‘→S(PD)’ means ‘causes large sacrifices to be needed for positive duty’) as: Sometimes when we may avoid large personal sacrifices to achieve a great good (S>A), achieving that good over­ rides some duty to aid for whose sake we could be morally required to make large personal sacrifices and achieving the good also causes the sacrifices to perform that duty to become large ( (A>PD)→S(PD) ). When this is so, we may avoid the large personal sacrifices rather than perform the duty (S>PD). (This contrasts with the conclusion that – (S>PD) in the representation of the seeming intransi­ tivity at the end of Section 1.)

3.  Harming Versus Not Saving Before proceeding to consider Parfit’s omission, it is worth noting some ways in which the Personal Lifeguard Case differs from the cases of interest to Parfit. Personal Lifeguard is a case of (i) letting someone die (a type of not aiding) while (ii) merely foreseeing his death in order to (iii) go and help a far greater number of others not merely any additional number of people. By contrast, Parfit is inter­ ested in cases that involve killing someone (a type of harming) in order to help others. Further, while Parfit first tries to undermine the moral significance of the harming/not aiding distinction by arguing that a bystander may save five people by diverting a trolley merely foreseeing that this will kill one other person, he goes on to argue that killing as a mere means (as in toppling a fat man from a bridge to stop the trolley) and as a side effect of a means to redirecting the trolley (as in using a grenade to redirect the trolley when a part of the grenade kills a bystander) are all permissible to save the five. Furthermore, he thinks we may kill one person in each of these ways to save only two other people from an unin­ tended threat. Presumably he would also think we may kill four people to save five others. Consider the issue of numbers of people saved. Though I disagree with Parfit’s claim that all these ways of killing are morally equivalent, I do think that if an argument justifies turning the trolley towards one person to save five, it would also at least justify turning it from two people to one. If the argument could not justify this other thing, I do not think it could justify killing one to save as small a number as five. Furthermore, suppose it were permissible to turn the trolley to save only five or more. This is still a much lower number of people saved, I think, than would

450  F. M. Kamm justify the personal lifeguard leaving his client to die. This would be true even if the lifeguard weren’t morally required to do as much to save his client as giving up his life and what he was required to give up was less than he was required to give up not to kill someone when only that person’s and his own interests were at stake. It is standardly thought that the ordinary negative duty not to kill someone is more stringent than even a strong positive duty, as shown by the impermissibility of killing to achieve an end that one may abandon such a positive duty to achieve. For example, one may not kill or do what kills one’s client or anyone else to go and save the hundred. Yet in a variant of the Personal Lifeguard Case one would not be permitted to abandon a contracted positive duty to achieve the good of saving five lives though many think achieving this would justify one in killing someone at least by diverting a threat from the five. This is so even though independent of any contract or promise each person is thought to have a so-called ‘natural’ claim right against everyone that they not kill him. Might this difference between the cases exist because the death of the one in the trolley diversion case would be a result of the five people already being saved when the trolley moves away, but the death of the client would not be a result of the five people already being saved but a result of what the lifeguard does in order to save them?14 But suppose the five being saved by the personal lifeguard would itself create a barrier that alone pre­ vents him from saving his client. Then the five already being saved would also lead to the death of his client. Yet it would still be impermissible for the lifeguard to save the five rather than his client. The added strength of a positive right due to its being contracted can help explain some of the difference in the cases since being contracted can also strengthen negative rights. For example, an agreement not to kill the one other person on the side track could help make turning the trolley to save the five impermissible, as Judith Thomson pointed out.15 This suggests that while duties may be distinguished by whether they are or are not correlative to rights (e.g., duties to give charity are thought not to be correla­ tive to rights), duties correlative to rights may be distinguished by whether they are correlative to uncontracted rights (and so are uncontracted correlative duties) or correlative to contracted rights (and so are contracted correlative duties). (See Figure 20.1.) Let us now consider some ways of not fulfilling the positive duty in Personal Lifeguard that are analogous to harming as a mere means, which Parfit thinks is permissible to achieve a greater good. What if the client’s not being saved was not a mere side effect of the lifeguard going to save others or of the others being saved but played a causal role in their being saved? Here are at least three ways in which 14  I have suggested such an explanation of the permissibility of turning the trolley. See e.g. my The Trolley Problem Mysteries (Oxford: Oxford University Press, 2015). 15  See Thomson, ‘Killing, Letting Die, and the Trolley Problem’, The Monist 59/2 (1976), 204–17.

Duties That Become Supererogatory  451 Duties Not correlative to rights

Correlative to rights

uncontracted right

contracted right

Figure 20.1

this could happen: (a) Only the client’s drowning makes available a device the lifeguard needs to swim fast enough to save the other people. In this case the cli­ ent’s death is part of the lifeguard’s required means to save the others.16 This is why he cannot save the client and also the hundred and he lets the client drown so the device becomes available. (b) Everything is as in (a) except that he leaves his client to save seventy other nearby people he would not otherwise have saved (since he is prejudiced against them), merely because doing so allows his client to die so that the means he needs to save the hundred is available. (c) The lifeguard can save the hundred people merely by swimming towards them when that is incompatible with his saving his client. However, doing this would not protect those people from dying of a second threat that arises because he saves them from the wave (e.g., a shark would then sense their presence and attack). It is only because his client died—a result of the lifeguard abandoning him to save the hun­ dred from the wave—that a device is available to ward off the second threat.17 In (c) the lifeguard would not go to save the hundred from drowning unless the second threat could be stopped so he acts because he foresees that his client will die. Nevertheless, in this case he does not refrain from saving his client as a means to save the hundred from the wave which is the first threat from which he must save them if they are to survive. In (b) he does refrain from saving his client as a means to save the hundred from the wave but this does not make what he does impermissible if, regardless of his reason for leaving his client, saving the seventy people could justify his leaving his client. It is only in case (a) that his not saving his client as a means to save the hundred seems to me impermissible since there is nothing else brought about without the client’s death being a means to it (as in (b) ) that justifies his refraining from aiding the client. Hence I think the lifeguard in (a) may not fail to perform a duty for which he is morally required to pay great costs in order to do something for which he is not morally required to  pay great costs. But this is not because of the difference in costs morally required of him; it is because he omits to do his duty in order to bring about the death of his client for the sake of its causal role in saving other people. Given

16  There could be a version of the Business Case in which it is because of the loss to the client when a deadline is not met that a device is available to save a life. 17  This case has a causal structure similar to the Loop Case, which is a version of the Trolley Problem.

452  F. M. Kamm Parfit’s arguments defending killing as a mere means, he would claim that omitting to aid one’s client in (a) to save the hundred people is permissible.

4.  Redirecting a Threat As we have seen, Parfit wants to argue that it is permissible to harm, not merely not to aid, people for the sake of a greater good. In his view, the harm might be a mere causal means to the greater good or a side effect of either a mere causal means to the greater good or of redirecting a threat. He also holds that the per­ missibility of such harm is theoretically compatible with having a duty to make large personal sacrifices to avoid harming someone when no greater good is at stake and with one not having to make the same personal sacrifices to bring about the greater good that can license one in harming someone else. However, he did not consider whether one might have a duty to make large personal sacrifices to bring about a greater good rather than harm someone else for that good. We shall now consider this issue in connection only with redirecting a threat.

4.1  Two Measures of Stringency I have already agreed that there are two measures of stringency and that having to make greater personal sacrifices to avoid harming someone than to produce a greater good does not by itself show that we may not harm that same person to produce that greater good. But there are other issues to consider.

4.2  Imposing Versus Volunteering Losses In her ‘Turning the Trolley’,18 Judith Thomson argued against the view that a bystander may turn an out-of-control trolley away from killing five people when it will then kill some other innocent person. (Parfit calls this case Side Track but following Thomson I shall call it the Bystander Case.) She first argued that the bystander would not be morally required to turn the trolley away from the five when it will kill the bystander himself if this was his only way to save the five. It would be supererogatory to do this. She took this to imply that when there was no way the bystander could turn the trolley towards himself, it was wrong for him to turn the trolley to another innocent person (when it would kill him) rather than to let the five die. This was because this other person too was not morally required

18  Philosophy & Public Affairs 36/4 (2008), 359–74.

Duties That Become Supererogatory  453 and (by hypothesis) did not agree to turn the trolley on himself to save the five. So Thomson rejects the view that we may in this way harm someone for the greater good of saving five other people. She might think the same if the greater good were saving one hundred other people.19 By contrast, Thomson thinks that the trolley driver would kill five people (not merely let them die) if the trolley he started hits them and he has a duty of justice not to kill five. Hence, she thinks, he should turn the out-of-control trolley to the one person, even though that person also need not volunteer his life to either save the five or prevent the driver from unjustly killing them. Suppose we agreed that neither the bystander nor the one who would be killed by the turning trolley is required to volunteer to save the five by turning the trol­ ley away from them and towards himself. It might still be argued that volunteer­ ing to suffer a loss is morally different from having the loss imposed on one in general. If this is so, imposing the loss on someone without his consent could sometimes be permissible even though the person hit is not morally required, and does not agree, to volunteer his life and may permissibly resist (in certain ways) what is being done to him. Hence, it might be argued that it is permissible for the bystander (not only the trolley driver) to impose the loss on the other person by redirecting the trolley away from the five, at least when the bystander cannot sacrifice himself to save the five. On the other hand one might argue against the view that the trolley driver’s duty of justice not to kill the five is suffi­ cient to justify his turning the trolley to the one. After all, it is not sufficient to justify the driver toppling someone in front of the trolley as a mere means to stop it. This is an act that I, unlike Parfit, think is impermissible.

4.3  Harming Another Rather Than Oneself It is not my aim to argue these points here.20 I simply wish to suppose that the bystander is permitted to turn the trolley when it is all he can do to save the five, and inquire whether this implies that when the bystander is able to turn the trol­ ley on himself to save the five, he may instead still impose the loss on the one innocent person. An analysis like the first analysis offered for the Personal Lifeguard Both Case suggests that he may not. For according to that analysis the lifeguard must perform his duty to his client at the maximal required cost if it is possible to do so while saving the one hundred. Analogously, it might be said in (what I shall call) the Bystander Both Case that though the bystander needn’t turn

19  These views of Thomson’s leave it open whether she would think one may permissibly omit a positive contractual duty to save someone in order to instead save one hundred other people. 20  I emphasized these points and the distinction between imposing and volunteering when dis­ cussing Thomson’s view in The Trolley Problem Mysteries.

454  F. M. Kamm the trolley to himself when this is the only way to save the five, he may have to turn it to himself rather than turn the trolley from the five towards that one other person. On this view the important point is that if the duty not to kill another can take precedence over one’s self-interest to the extent that one is required to sacri­ fice one’s life, then if the bystander wants to save the five he would have to ­sacrifice himself rather than kill the one person instead for that end. This is how he can both save the five and fulfil his duty to pay personal costs rather than kill the other person. This could be true even if we (unlike Thomson) think that the bystander may harm the innocent person to save the five when the bystander cannot save the five by sacrificing himself instead of that person. If this analysis were correct, it would raise a problem in some cases for a view, to which Parfit is open, that having a duty to sacrifice oneself rather than harm someone else is compatible with not having a duty to sacrifice oneself for the greater good. The problem arises in cases where one could both do one’s duty and achieve the greater good if serving self-interest drops out. According to the first analysis of the Personal Lifeguard Both Case, one has to fulfil one’s contractual duty at the maximal morally required cost if one chooses to save the one hundred (whether or not saving them per se is so costly) since this allows one to also save one’s client. In the Bystander Both Case, saving the five people would itself involve making a great personal sacrifice (turning the trolley on oneself). However, one would not make the sacrifice to help them alone but rather to help them without failing in a duty not to harm someone else even at great cost to oneself. There is an alternative view—call it the Still Harm Another View—of the Bystander Both Case that is similar to the Still Abandon Client View about the Personal Lifeguard Both Case. One defence of this view begins by claiming (con­ trary to Thomson) that if the bystander’s only possible means of saving the five were redirecting the threat towards one other person, it would be permissible (though not required) for him to do so. This is so even though that person would not be morally required to give up his life to save the five (e.g., by diverting the trolley to himself) if this were possible, and he would not do so. The next step is to consider the case in which the bystander can both save the five and not harm the one by turning the trolley towards himself instead. This high cost to him would arise because he would be saving the five, something that on its own would (by assumption) override not killing the one other person if no other way of saving the five existed. Because the bystander predictably would not kill himself instead of one other person to save the five, a consequentialist argument for the Still Harm Another View might emphasize that the bystander should not be required to avoid harming the one person at any cost to himself that would lead to his not saving the five. But a non-consequentialist would not think the good of saving the five is always sufficient to relieve the bystander of a duty to make large sacrifices rather than kill someone else. To defend the Still Harm Another View, a nonconsequentialist might instead emphasize that a person has a duty to make big

Duties That Become Supererogatory  455 sacrifices to prevent himself killing another person when he is not doing some­ thing that justifies causing that killing. But if saving the five justifies turning the trolley though it kills another person, there is no need for the bystander to sacri­ fice his life to prevent the justified killing. This is so even in cases where it is not the bystander’s duty but is only morally permissible to turn the trolley. The following type of case might be raised as an objection to the Still Harm Another View: I should not use my neighbour’s car without his permission merely to avoid the sacrifice to me of using my own car (which wears it down a bit). I may use his car without his permission if there is no other way to get one hun­ dred people who will otherwise die to a hospital. However, if my car is available, I cannot still permissibly use my neighbour’s car to get the people to a hospital on the ground that I am not merely trying to avoid a personal sacrifice (of wearing my car down) but am saving people and the latter overrides the neighbour’s per­ sonal claim to his car. I think this case is not a good objection to the Still Harm Another View because I assume that in it both my neighbour and I have a noncontracted duty to use our car to save the one hundred people; it is not super­ eroga­tory to do so. I should not interfere with the car to which my neighbour has a right to avoid my duty even when my neighbour also would have such a duty. By contrast, in the Bystander Case (we are assuming) I have no duty to sacrifice my life to save the five though (it is assumed) I am permitted to impose loss of a life on another person by turning the trolley. (This is so even though he too has no duty to sacrifice his life to save the five.) Something similar to what the Still Harm Another View implies about the Bystander Both Case seems to be true in other cases (though in certain ways they are significantly different from turning the trolley). For example, suppose it is true, as commonly thought, that a strategic bomber in a war (at least on the just side) may sometimes permissibly bomb an opponent’s military facility even fore­ seeing that some non-combatants will be killed as a side effect. One condition that must be met for permissibility is that the deaths are at least proportionate to the military goal.21 Suppose a bomber goes on a very useful but for him optional bombing mission and it becomes possible for him to redirect the effects of the bombing so that he rather than a proportional number of non-combatants would be killed as a side effect (when there is no shortage of other soldiers to fight if he dies). Is the bomber morally required to do what kills him rather than what kills the non-combatants merely because were he not engaged in doing what made it permissible for him to cause side-effect deaths, he would have to sacrifice his life rather than kill non-combatants as a side effect of his actions? I do not think so. 21  I think it is not clear why such bombing is permissible, but I will not discuss that issue here. Notice that the deaths due to the bombing would probably not be proportionate if just one more per­ son would be saved by blowing up the factory than would be killed by blowing it up. One indication that the justification for turning the trolley differs from the justification for strategic bombing is that it seems permissible to turn the trolley to save even two people at the cost of killing one other person.

456  F. M. Kamm When the need for him to sacrifice his life to prevent harm to others arises from his doing something that is permissible despite its causing that harm to others and he has no independent duty to die in order to bomb the facility itself, this great sacrifice is not required.22 One way this Bomber Case differs from the Bystander Both Case is that the bomber’s means to his goal creates a new threat to non-combatants while the Bystander redirects a pre-existent threat. It would be particularly significant if even this difference does not alter there being implications for the Bomber Case analo­ gous to those that the Still Harm Another View has for the Bystander Both Case.23 22  Still, he may have a duty to make smaller sacrifices to avoid killing the non-combatants since he still infringes their rights not to be harmed even if he acts permissibly. Notice also that, strictly speak­ ing, bombing in a way that kills non-combatants is not necessary when the pilot could do what substi­ tutes his death for theirs. So it seems to violate the necessity condition that it is said must also be satisfied to make doing what kills non-combatants permissible. However, the bombing may satisfy Seth Lazar’s account of the necessity condition if bombing in a way that kills non-combatants is neces­ sary to give appropriate moral weight to the potential deaths of both combatants and non-combatants. (See Lazar’s ‘Necessity in Self Defense and War’, Philosophy & Public Affairs 40/1 (2012), 3–44.) 23  Jeff McMahan may disagree that the bomber does not have to do what kills himself, at least when the non-combatants killed are not morally subject to bearing risks of being killed. (He thinks their being potential beneficiaries of the military action could make them morally subject to bearing risks of being killed. I have suggested that their being citizens of an unjust country could make them mor­ ally subject to bear risks of being killed in order that their country stop being unjust.) However, McMahan does emphasize the killing/letting die distinction. So he may think the bomber should, for example, let himself die by being killed by enemy fire that prevents his dropping bombs in a way that kills non-combatants when this does not interfere with his destroying a military facility. In this case, unlike mine, the bomber would not face a choice of killing himself or killing the non-combatants but rather letting himself be killed or killing non-combatants. However, I do not think the bomber has a duty to let himself be killed either. To justify his position, McMahan emphasizes the duties of some professionals like firefighters to take risks as part of their job. But suppose a fireman was permitted to redirect a fire from killing ten people to killing only two other persons in a burning building. If he is also able to redirect the fire towards himself when this will kill him alone, must he do so? I do not think he has a professional duty to kill himself (even when there is no shortage of firefighters) rather than redirect the fire to the two other people. (I think this same would be true of someone who had no duty to fight the fire but volun­ teered to help.) Sometimes when people would carry out permissible harm to others they do not thereby make themselves liable to self-harm instead. Similarly, suppose the firefighter’s mission is to save the ten people and to do so, it would be permissible to redirect the fire to two other people instead. However, the wind is about to redirect the fire towards the firefighter instead. His only way to stop that is to immediately redirect the fire towards the two. Given that he is there on a mission to help save the ten, must he let himself be killed by the fire rather than redirect it from ten to two others? I do not think he must let himself be killed. These Firefighter Cases, like the Bystander Both Cases, involve redirecting threats not created by the Firefighter. Hence, McMahan might want to limit his view to agents like the bomber who create new threats. But suppose the firefighter were permitted to start a flood to put out the fire to the ten foreseeing the flood would create a deadly threat to two other ­people instead. (This is like Parfit’s Fire and Flood Case in volume iii of On What Matters.) I do not think the firefighter would instead have to create a different flood that will only kill him. To further support his position on the bomber, McMahan also refers to what he says is the accepted view that one may not avoid a lethal threat to oneself by means that will kill someone else even as a side effect. But cases to which he refers involve deciding what one has to suffer rather than harm someone else when the only justification proposed for harming is that one’s own interests would otherwise suffer. I have suggested that when a greater good would justify doing what harms another, what one is morally required to suffer rather than harm that person for the greater good may differ from what one has to suffer rather than harm her when only one’s interests are at stake. For McMahan’s views to which I refer, see his ‘The Just Distribution of Harm Between Combatants and Noncombatants’, Philosophy & Public Affairs 38 (2016): 342–79.

Duties That Become Supererogatory  457 What if in a different version of the Bystander Both Case there was already a high cost that the bystander had to bear so as not to harm the one other person independent of his saving others? For example, suppose the bystander’s car is going to crash into the one person and the bystander therefore has a duty to swerve towards a tree in order to avoid killing the person, though by doing this the bystander will himself be killed. It turns out that his swerving will also divert a trolley away from killing five towards killing no one. (Call this case Bystander Both Pre-Cost.) In this case, he may not omit swerving thus killing one person in order to instead do something else that will divert the trolley alone perhaps at a lower cost to himself. Omitting to swerve in this case is not justified merely because in a different hypothetical case if his diverting the trolley from five would kill that one person, he need not swerve his car in a way that will cause his own death in order to save the one person from the trolley. An explanation of the dif­ ference in duties is that in the Bystander Both Pre-Cost Case having to pay a large cost rather than harm the one person is not called for to prevent harm that is an effect of saving the five that would justify killing the one. Similarly, suppose a bomber is going to do something impermissible that will kill some non-combatant bystanders and so he should instead divert in a direc­ tion where he will be killed. If he diverts in this way, he will also crash into and destroy a munitions facility. He should not omit to divert from killing these noncombatants and instead bomb the facility in a way that does not cause any noncombatant deaths when this can be done at a low cost to himself. Omitting to divert in this case would not be justified merely because in a different hypo­thet­ ic­al case if bombing the facility would kill the same non-combatants as a propor­ tional side-effect, he need not bomb the facility in a way that kills him instead of the non-combatants. Now consider a harming case analogous to Personal Lifeguard Both Post-Cost. A bomber goes on a useful but optional mission to destroy a munitions factory in a just war. Non-combatants whose deaths would have been proportional are sur­ prisingly unhurt. It would ordinarily have been easy for the bomber to avoid harming them on his route back to his home base. However, his successful bomb­ ing of the factory caused a defect in his plane and now he would have to sacrifice his life in order not to drop new bombs onto those non-combatants. In this case, the cost to him of not harming the non-combatants goes up only because he did something that would have been permissible even if it had caused the deaths of those non-combatants. However, in the actual case the deaths do not arise in doing the act that would have justified causing that bad effect. Hence, he must sacrifice himself in order not to kill the people. This is so even if his foreseeing this course of events would have led him to not undertake the useful mission thus undermining best consequences. There is a possible further analogy between the Still Abandon Client View as applied to the Personal Lifeguard Both Case and the Still Harm Another View as

458  F. M. Kamm applied to the Bystander Both Case. It concerns what one should do if there is only a small additional cost to oneself beyond what is needed to save the many (the hundred or the five) in order to also help the one (by saving the client or by not harming the one other person on the track). The lifeguard has a standing cor­ relative duty to help his client. As already noted, it seems to me this could imply that he should pay some additional cost to fulfil the special duty to his client even if he is saving the hundred which overrides and also threatens non-performance of the duty. The situation may be similar for the bystander. He has a correlative non-contractual duty to pay high costs rather than harm someone when it is only the interests of these two people that are at stake. Even if this correlative duty would not be violated when he diverts the trolley to someone rather than to him­ self to save the five, he may still be infringing the right of the one person not to be harmed and so wrong him in the course of a permissible act.24 His paying a small additional cost to save the five without harming the one could represent the attempt to avoid even infringing a right. I have argued that Parfit did not consider the possibility that there is a duty to make large personal sacrifices as a substitute for harming someone else to bring about a greater good for others. Similarly, in ‘Supererogation and Obligation’ I did not consider the possibility that there is a duty to make large personal sacrifices as a substitute for not fulfilling a positive duty to someone else in order to bring about a greater good for others. However, I have also argued that in many cases there are no such duties to make large personal sacrifices rather than harm some­ one else or not fulfil a positive duty to her in order to bring about a greater good. Hence what would otherwise have been one’s duty to make large personal sacri­ fices becomes supererogatory. However, this will only be true when the way in which one would not fulfil the positive duty or harm someone else would be per­ missible. Suppose Parfit is wrong that harming someone else as a mere means or as a side effect of a mere means to a greater good is permissible. Then we would have to make the large personal sacrifice rather than do these impermissible things if we chose to bring about the greater good.

5.  Permissible to Forbidden? We have considered whether there is a change in the status of an act—from duti­ ful to supererogatory—based on whether we still may impose or not prevent a 24  As noted above, ‘violation’ refers to impermissible transgression; ‘infringes’ includes permissible transgression. Some evidence that the right is infringed and not merely ‘wiped out’ is the fact that it would be permissible for the one person towards whom the trolley is turned to turn it back towards the bystander and even the five originally threatened. This is so even though the bystander acts per­ missibly in turning the trolley and it would not be permissible for the one person to turn a trolley that was originally headed to him in a way that will kill the bystander or five other people.

Duties That Become Supererogatory  459 loss to someone else. Though Parfit did not consider this issue he did propose that sometimes there is a change in the status of an act—from permissible to forbidden— in a version of deontology that he rejects (in part because he thinks it involves this change). We will briefly consider whether this change occurs. Parfit believes that deontology gives each of us agent-relative duties in which the aim makes reference to the agent. For example, X’s aim is for him (X) not to kill innocents and Y’s aim is for him (Y) not to kill innocents. By contrast, nonagent-relative consequentialism gives each of us a common aim (e.g., to minimize killings of innocents) which may require us to deal with what Larry Alexander has called ‘other people’s messes’ (e.g., kill one person to stop other agents wrong­ fully killing five other people). Parfit tries to move us towards a common aim view of morality. To do so he takes a step beyond dealing only with cases where there is an unintended threat (e.g., a runaway trolley or a natural disaster) and one agent who must respond (like the bystander in the Bystander Case). In the cases he presents in the last part of volume iii of On What Matters, one agent is called on to reduce the intended yet permissible harmful acts of other agents who are also trying to stop an unin­ tended threat. In these cases, he argues, we should all have a common aim of reducing permissible (not impermissible) intended harmful acts. For example, in his Power Plant Case the blowing-up of the plant that would kill a million people could be stopped by me only if I intentionally kill one person (who would not otherwise shortly die). Parfit assumes that if there were no other alternative, this killing would be permissible according to threshold deontology which permits killing someone to stop the deaths of others once the number of people that would otherwise die is beyond a threshold. However, if I do not do this, the only alternative is that someone else will permissibly (according to threshold deontol­ ogy) kill two innocent people (who would not otherwise shortly die) to stop the plant blowing up. (Note that in one version of this case, if I refrain from killing the person, he will become one of the two killed by someone else. In a second version, if I do not kill the person he will not otherwise be killed. I will discuss the second case since I think it is the one Parfit intends and because it may be harder to justify killing a person who would not soon be killed by someone else.) Parfit thinks that the agent-relative aspect of deontology incorrectly implies that I should not kill one person to stop the plant blowing up since my primary respon­ sibility is that I not kill. Thus an act that was permissible becomes forbidden. This is so even though my killing could prevent someone else from having to permis­ sibly kill two other people instead. He thinks this implication shows that deontol­ ogy incorrectly draws a moral distinction between what I do and what others do and incorrectly tells me to give priority to my not killing over preventing others from killing. I think Parfit’s analysis involves a misunderstanding of deontology. Suppose that an act that is (by hypothesis) deontologically permissible or a duty for a

460  F. M. Kamm particular agent (e.g., killing one person to stop the plant blowing up) has a better outcome than an alternative act that would be permissible or a duty when done by another agent if the first agent did not act (e.g., killing two people to stop the plant blowing up). While I do not agree that deontology is based on an agent’s special concern for his own acts,25 I also think there is nothing in an agentrelative analysis of deontology that requires (or even permits) the first agent to ‘kick’ the act-type (in this case killing permissibly or as a duty) down the road so that another agent does it with a worse outcome (more people are killed). What may lead one to think otherwise is conceiving of the Power Plant Case as one in which I would kill one person merely in order to prevent someone else from kill­ ing two people, which would be ruled out by deontology. It is true that the numerical difference between my act (resulting in one million minus one being alive) and the other agent’s act (resulting in the same one million people minus two others being alive) is one extra life. However, my saving one million by killing one does not just save one additional person from death, it saves one million and one. That someone else would have saved the same million if I had not does not imply that they are not still under a threat at the time I act and that I don’t save those million people. And my saving those one million can help make my act permissible according to threshold deontology. Similarly, if someone will kill a million people plus two, the person who pre-emptively kills the same million plus only one still also kills the one million plus one. This is compatible with it being true that because another agent will save those same one million if I do not, I should go and save by innocent means a different group of five people who would not otherwise be saved. This is so at least if saving the different group of people is more important than preventing a killing of one additional person by the other agent. What if saving that different group of five people required me to kill someone as a mere means? If it is impermissible to do this, I should instead reduce lives lost by killing one to save the million rather than allowing another agent to kill two to save the million. This would not make sense on the view that in the latter case one would be killing one person as a mere means to save two others given that it is impermissible to kill one as a mere means to save the group of five.26 The Power Plant Case is unlike cases in which consequentialists claim that an act becomes permissible or a duty because it substitutes for another act done by someone else that would have a worse outcome (like the case where killing one person would be impermissible were it not done to stop someone else impermis­ sibly killing more people). In the Power Plant Case the agent’s act that kills one 25  I have criticized the view that deontology is based on concern for the agent’s point of view in acting in my Morality, Mortality, ii. Rights, Duties, and Status. See chapter entitled ‘Constraints and You’. 26  I thank Shelly Kagan and Larry Temkin for suggesting this last case and for discussion of issues raised in Section 5.

Duties That Become Supererogatory  461 person is (by hypothesis) initially determined to be deontologically permissible independent of considering what another agent would do and I do not think it becomes impermissible merely because the alternative is that another agent would act permissibly or dutifully with worse consequences. If the consequences were better if another agent acted, then the first agent’s act could become impermissible. Parfit presents another case involving deception. He imagines that it would be permissible for someone to deceive two people to keep some important secret. He concludes that it would be permissible for someone else to deceive one person instead, thus achieving the common aim of safeguarding the secret by deceiving as few people as possible. In this case, he does not explicitly say that it would have been permissible for the second agent to deceive one person even if no one else would permissibly deceive two. This leaves it open that the person who would deceive two people has special permission or a duty to deceive that others lack. If so, it is not true that another agent acquires a permission or duty to deceive only one person merely because this is fewer than would be permissibly deceived by someone else. If the agent who would deceive one independently has a permis­ sion or duty to deceive, then the agent who deceives fewer may act and the other agent may be forbidden to deceive more. I conclude that Parfit’s cases do not show that what he calls common sense deontology incorrectly forbids an agent to do what is initially judged to be a permissible or dutiful killing for him and shifts the performance of permissible and dutiful killings to others simply because of the differences between the agents. Furthermore, common sense deontology could be correct in not deriv­ ing the permissibility of a type of act for one agent from its permissibility with worse consequences for another agent. Hence, Parfit has not shown that there is  a status change from permissible to forbidden in these cases according to deontology.

Bibliography Foot, P., ‘The Problem of Abortion and the Doctrine of Double Effect’, repr. in B.  Steinbock (ed.), Killing and Letting Die (Englewood Cliffs, NJ: PrenticeHall, 1980). Kamm, F., Morality, Mortality, ii. Rights, Duties, and Status (Oxford: Oxford University Press, 1996). Kamm, F., ‘Parfit on the Irrelevance of Deontological Distinctions’, in M.  Timmons (ed.), Oxford Studies in Normative Ethics, 9–31 (Oxford: Oxford University Press, 2020). Kamm, F., ‘Supererogation and Obligation’, Journal of Philosophy 82/3 (1985): 118–38. Kamm, F. The Trolley Problem Mysteries (Oxford: Oxford University Press, 2015).

462  F. M. Kamm Lazar, S., ‘Necessity in Self Defense and War’, Philosophy & Public Affairs 40/1 (2012): 3–44. McMahan, J., ‘The Just Distribution of Harm Between Combatants and Non­ combatants’, Philosophy & Public Affairs 38 (2016): 342–79. Parfit, D., On What Matters, vol. iii (Oxford: Oxford University Press, 2017). Scheffler, S., The Rejection of Consequentialism (Oxford: Oxford University Press, 1982). Thomson, J. ‘Killing, Letting Die, and the Trolley Problem’, The Monist 59/2 (1976): 204–17. Thomson, J. ‘Turning the Trolley’, Philosophy & Public Affairs 36/4 (2008): 359–74.

21

More Supererogatory Thomas Hurka and Evangeline Tsagarakis

In one of its uses the concept of supererogation is binary, or doesn’t admit of degrees. If the moral duties or reasons an act would fulfil don’t outweigh the costs it would involve for you, the act is simply supererogatory; if they do outweigh those costs, it’s simply not supererogatory but instead is required. In another use, however, this concept does admit of degrees, and one of two supererogatory acts can be more supererogatory than the other. Supererogatory acts are commonly said to be ‘beyond the call of duty’; surely one act can be further beyond duty than another. These acts are associated with saintliness or heroism, and one can be more saintly or heroic than another; they also call for admiration, and can be more or less admirable. An act’s degree of saintliness, heroism, or admirability may not depend only on how supererogatory it is; other factors, such as ones relating to its motivation, may be, and we’ll later argue are, relevant. But the extent to which it has these qualities depends in part on its degree of supererogation, or on how far beyond duty it is. This paper will first give examples where one of two acts is more supererogatory than the other, and propose an initial account of what makes that so. It will then connect this account to a more general theory of supererogation, as well as to an account of degrees in other deontic concepts. And it will contrast its theory of supererogation with the one Derek Parfit defends in On What Matters; one weakness in Parfit’s theory, it will argue, is that it doesn’t allow degrees of supererogation.

1.  Degrees of Supererogation: Examples An act is supererogatory if the moral duties it will fulfil don’t outweigh the costs or sacrifices it will involve for you. But of two supererogatory acts that fulfil the same moral duty, one can impose greater costs than the other. Consider two acts The idea of a gap treatment of degrees of supererogation in Sections 1 and 2 is due to Tsagarakis; the general defence of the permission-­based theory in Section 2 and the critique of Parfit in Section 3 are due to Hurka; and Section 4 is a joint product. For discussion we’re grateful to Brendan de Kenessey, and for helpful comments to James Goodrich.

Thomas Hurka and Evangeline Tsagarakis, More Supererogatory In: Principles and Persons: The Legacy of Derek Parfit. Edited by: Jeff McMahan, Tim Campbell, James Goodrich, and Ketan Ramakrishnan, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/9780192893994.003.0022

464  Thomas Hurka & Evangeline Tsagarakis that will save another person’s life. One requires you to sacrifice one of your arms, while the other requires you to sacrifice both your arms. If the duty to save another person doesn’t outweigh the cost of one arm, both acts are supererogatory. But surely the second act, which requires you to sacrifice two arms rather than one, is more supererogatory than the first. It’s more heroic to make the larger sacrifice than to make the smaller one, and also more admirable; you go further beyond duty when you sacrifice more. And we can give an initial, relatively theory-­neutral account of what makes this so.1 The moral duty to save the other person turns on the benefit to him of being saved, or of being spared the harm of death. And the reason both acts are supererogatory is that this benefit, though greater than the costs the acts impose on you, isn’t vastly greater. To put it another way, the gap between the size of the benefit to him and the size of the cost to you isn’t large enough to make saving him a duty. But this gap isn’t the same for the two acts: it’s smaller for the second act, given its larger cost for you. And we can take this difference to explain why the second act is more supererogatory. In cases of benefiting beyond duty, we can say, the degree of supererogation depends on the size of the gap between the benefit to the other person and the size of the cost to you. More specifically, the smaller this gap is, the more supererogatory the act, and the larger the gap, the less supererogatory. A smaller gap means the act is further from being your duty; it would require a greater increase in the benefit the act gives the other to make it your duty. But when the gap is larger, or when the benefit exceeds the cost by more, the act is closer to being your duty and less far beyond duty. In this first example the two acts fulfil the same duty but differ in their costs for you. But we can also consider a case where two acts fulfil different duties but impose the same cost. Imagine two acts, both of which require you to sacrifice one of your arms but one of which will save another person’s life while the other will prevent him from losing both his arms. Here the second act seems more supererogatory, and we can again explain why by looking to the gap between the benefit it provides and the cost it imposes on you. In the first act this gap is large, because the harm of death is much greater than the harm of losing an arm. This means this act is not that far from being your duty; not that much would have to 1  We assume that among the duties that bear on these acts is a duty to promote the good of all, including yourself, impartially. This implies that if an act’s cost to you is vastly greater than its benefit to another—for example, if it involves sacrificing an arm to spare him a mild headache—it doesn’t fulfil your moral duties; it violates them and can’t be supererogatory. We also assume that in these acts the cost—either one arm or two—is the smallest that will produce the relevant benefit. If you sacrifice two arms to save another’s life when you could do so equally well by sacrificing one, your act doesn’t fulfil your impartial duty as well as the less costly alternative and may therefore be wrong. Whether the act is supererogatory then depends on whether supererogatory acts must be all-­things-­considered permitted; we leave this issue open. And even if the act is supererogatory, whether it’s more so than its alternative depends on a more complex measure we introduce in Section 2; on the simplest reading of this measure the more costly act isn’t more supererogatory.

More Supererogatory  465 be added to the benefit to make the act required. In the second act, however, the gap is smaller. And if a smaller gap between benefit and cost makes for a higher degree of supererogation, by making the act further from your duty, the second act is more supererogatory. This verdict may not seem quite as compelling as the one in the first example, because another factor, to be discussed later, may differentiate this pair of acts that didn’t distinguish the earlier pair. But if the size of the gap between benefit and cost makes one of two acts more supererogatory in the first example, it should do the same here. In both these examples, one end of the relevant gap is the same for both acts. In the first, both acts provide the same benefit to the other; in the second, both involve the same cost for you. This makes comparative judgements of supererogation relatively easy to make, since they require only a comparison of costs in the one example and of benefits in the other. These judgements are harder to make when neither end of the gap is the same for two acts, as when we compare saving another person’s life at the cost of two arms for yourself with preventing him from losing two arms at the cost of one arm for yourself. To determine whether one of these acts is more supererogatory than the other we need to compare the size of two gaps that differ at both their ends, that is, to compare the gap between a life and two arms with the gap between two arms and one arm. Because these comparisons are harder to make, our intuitions about which if either act is more supererogatory may be less firm. But if the gap between a life and two arms is greater than the gap between two arms and one, sacrificing one arm to save another person two arms is more supererogatory than sacrificing two arms to save his life. Since the second act is closer to being your duty, the first is further beyond duty. (Again another factor, to be discussed later, may complicate our judgement here.)

2.  Supererogation and Prima Facie Permissions The above account of degrees of supererogation, though persuasive as far as it goes, is in two respects incomplete. In taking one end of the gap to be the benefit an act gives another person, it assumes that the duty a supererogatory act fulfils is always the duty to promote the good, which we’re assuming is the duty to promote the good of all, including yourself, impartially. The account therefore doesn’t extend to supererogatory acts, if there are such, that fulfil other duties, say by keeping a promise, paying compensation, or refraining from directly causing harm when, given the costs of doing so, that isn’t required. And though the account is right to make the other end of the gap depend on an act’s costs for you, it doesn’t explain exactly how or why these costs have the normative effect they do. Different theories of supererogation give different explanations of this, by

466  Thomas Hurka & Evangeline Tsagarakis linking the costs to different normative factors that weigh against, and can at times outweigh, the requiring force of the duties a supererogatory act would fulfil. A complete account of degrees of supererogation should be embedded within some such theory. We’ll now propose one, which we’ll argue is both attractive in itself and well suited to accommodate degrees of supererogation. This theory takes the duties relevant to supererogation to be what W. D. Ross called prima facie duties,2 and can also be called pro tanto duties or duties other things equal. They include a prima facie duty promote the good, and perhaps also similar duties to keep promises, pay compensation, and so on. But it supplements these duties with an independent prima facie permission, or permission other things equal, to promote your own good and, as part of that, to avoid harms to yourself. This prima facie permission functions in the same way as a prima facie duty. If there’s a prima facie duty to do acts of type F, then an act’s being F tends to make it, or counts in favour its being, what Ross called your duty proper; this can also be called your duty all things considered or simpliciter. Analogously, if you have a prima facie permission to do acts of type G, an act’s being G tends to make it, or counts in favour of its being, all-­things-­considered permitted, or permitted simpliciter. In each case the prima facie or conditional factor contributes to an all-­things-­considered or unconditional one of the same type: prima facie duties to duties simpliciter, and prima facie permissions to permissions simpliciter. Just as one prima facie duty can conflict with another, so can a prima facie permission conflict with a duty. This will be so if there’s a prima facie duty to do acts of type F and a prima facie permission to omit acts of type G, and some act is both F and G. And just as prima facie duties can have differing weights, so can prima facie permissions: in each case it’s a matter of the strength of the tendency to make for duty or permission simpliciter. The prima facie duty to do F is stronger than the prima facie duty to do G if there’s some other prima facie duty, say to do H, that the duty to do F outweighs but the duty to do G does not. (A simpler test says the duty to do F is stronger than the duty to do G if, in a direct conflict between them, what you ought all things considered to do is F.) Analogously, the prima facie permission to do F is stronger than the prima facie permission to do G if there is some prima facie duty, say to do H, that the first permission outweighs but the second does not. (Here the simpler test isn’t applicable because prima facie permissions don’t conflict: it can be both all-­ things-­considered permitted to do an act and all-­things-­considered permitted not to do it.) Thus the prima facie permission not to sacrifice two arms is stronger than the prima facie permission not to sacrifice one arm if there is some duty that the first would outweigh but the second would not, for example some benefit to

2 W. D. Ross, The Right and the Good (Oxford: Clarendon Press, 1930), ch. 2.

More Supererogatory  467 another that you would be required to provide at the cost of one arm but not at the cost of two arms. There presumably is some such benefit. A theory of supererogation that uses these two prima facie factors can say the following: an act is supererogatory if the prima facie duties that favour doing it are outweighed by a prima facie permission not to do it, where that permission derives in some way from the more general permission to promote your own good. That the prima facie duties are outweighed by the prima facie permission means not doing the act is all-­things-­considered permitted, but doing the act is also all-­things-­considered permitted. The prima facie duties to do the act entail a prima facie permission to do it—‘ought’ in general entails ‘may’—and that permission isn’t undermined by the prima facie permission to avoid its costs. As noted above, prima facie permissions don’t conflict; it can be all-­ things-­ considered permitted to do an act and also all-­things-­considered permitted not to do it. So supererogatory acts are ones it’s both simply permitted to do and simply permitted not to do. And what grounds the simple permission not to do them is a prima facie permission to avoid costs to yourself that, in these cases, outweighs whatever duties the acts would fulfil.3 This theory can be extended to allow degrees of supererogation. These will again depend on a gap, now that between the strength of the prima facie permission to avoid an act’s costs and the strength of the duties to do it that the permission outweighs. But here the size of the gap has the reverse effect to the one described above: now a larger, not a smaller, gap makes for a more supererogatory act. Return to the first example, where you can save another’s life either by sacrificing one of your arms or by sacrificing two. Since the prima facie permission to avoid losing two arms is stronger than the permission to avoid losing one, the prima facie permission in the first act is stronger than the one in the second. The first permission therefore outweighs the duty to save the other’s life by more, which makes the gap in the first act larger. And that is the act that is more supererogatory. It would take a greater increase in the strength of the prima facie duty an act with this cost fulfils to make it simply your duty; maybe it would need to save ten lives, whereas only five would be needed to make sacrificing one arm your duty. The first act is therefore further from being simply your duty than the second, or further beyond duty. And in this context, where we are comparing prima facie permissions with prima facie duties rather than benefits with costs, a larger gap makes that the case. As so extended, the theory corrects the incompleteness of the initial account of degrees of supererogation. Because the factor on one side of the gap is now the strength of a prima facie duty rather than the size of benefit, it can allow supererogatory acts, if there are such, that fulfil duties other than beneficence, 3  For this theory, see T. Hurka and E. Shubert, ‘Permissions to Do Less than the Best: A Moving Band’, Oxford Studies in Normative Ethics 2 (2012): 1–27.

468  Thomas Hurka & Evangeline Tsagarakis such as promise-­keeping and compensation. And because the factor on the other side of the gap is a prima facie permission to pursue your good that’s present in contexts other than that of supererogation, it gives a more general explanation of why an act’s costs to you can make it not required. Moreover, its account of degrees of supererogation connects with accounts of degrees in other deontic concepts such as wrongness.4 In one use the concept of wrongness doesn’t admit of degrees. If the prima facie duties an act violates outweigh the prima facie duties and/or permissions it fulfils, the act is simply wrong; if they don’t, it’s simply not wrong. But in another use one act can be more wrong than another or, if we want to reserve the word ‘wrong’ for the binary concept, more seriously wrong. And the ­seriousness of a wrong again depends at least in part on a gap, now that between the strength of the prima facie duties an act violates and the strength of any duties and/or permissions it fulfils. Imagine two wrong acts, one that gives a small benefit to one person rather than a somewhat larger benefit to another, and one that gives the same small benefit to one person rather than a vastly larger benefit to another. Here the second act is more seriously wrong because the gap between the strength of the prima facie duty it violates and the strength of the one it fulfils is larger. (When an act that violates a prima facie duty fulfils no duty or permission, its seriousness as wrong depends just on the strength of the violated duty; but that is just the gap between the strength of that duty and nothing.) Derek Parfit implicitly took this view at one point in On What Matters. He argued that if killing one person to save several others was permitted only if the number saved was at least a hundred, then killing one to save ninety-­nine would be only ‘slightly wrong’ or something there’s only a ‘weak moral reason’ not to do.5 And the explanation is presumably that, when the number saved is so close to the threshold for permissibility, the gap between the strength of the duty the act violates and that of the one it fulfils is small. His claim about this kind of case is actually contestable. It can be argued that when the duty a wrong act violates is deontological, like the duty not to kill, rather than just about giving benefits, the seriousness of the wrong depends not only on the size of the gap but also on the strength of the violated duty on its own. But even on that view the gap is relevant, so killing one to save 99 is less seriously wrong than killing one to save 50 or 25 or two. An account of degrees of supererogation based on the gap in strength between a prima facie permission and the prima facie duties it outweighs parallels an attractive account of degrees of seriousness of wrong.

4 See  T.  Hurka, ‘More Seriously Wrong, More Importantly Right’, Journal of the American Philosophical Association 5 (2019): 41– 58. 5 Parfit, On What Matters (Oxford: Oxford University Press, 2017), iii. 346, 341.

More Supererogatory  469 The above theory of supererogation, which weighs prima facie duties against a prima facie permission, isn’t the only or even most common such theory. But it has several attractive features. First, it gives a straightforward explanation of why the costs to you that make an act supererogatory aren’t just those of some precise size but all those within a range. If you’re permitted not to save another’s life at the cost of one of your arms, you’re also permitted not to save it at the cost of two arms, two legs, or your life. You have the first permission because the prima facie permission to avoid losing an arm outweighs the prima facie duty to save a life, but if that is so, the prima facie permissions to avoid losing two arms, two legs, and so on also outweigh that duty, since they’re stronger. The result of weighing your general prima facie permission against the duty to benefit others isn’t an absolute permission to weigh your own good more heavily than others’ in just some precise ratio, say 5:1 and no more or no less. If you’re permitted to prefer one unit of your own good to five units of good for others, you’re also permitted to prefer it to four, three, or two units for others. Second, the theory can be extended to accommodate a related but contrary phenomenon. Standard cases of supererogation involve what’s been called an ‘agent-­favouring permission’, since they permit you to prefer a lesser good for yourself to a somewhat greater good for someone else. But common-­ sense morality also grants ‘agent-­sacrificing permissions’, since it sometimes permits you to forgo a greater good for yourself to give a somewhat smaller one to someone else.6 Just as it allows you to care somewhat more about yourself, so it allows you to care somewhat less. (It may do this most when the other person is someone close to you, like your child or spouse. But it also allows it to some degree with strangers.) The theory we’ve proposed can grant agent-­sacrificing permissions if it supplements its prima facie permission to pursue your own good with a parallel prima facie permission not to pursue it, or to omit pursuing your good, where that can involve accepting a harm to yourself. This second prima facie permission can likewise conflict with prima facie duties, such as an impartial duty to promote everyone’s good, including your own, and it can sometimes outweigh them. When it does, giving a greater good to yourself and giving a lesser one to another are both all-­things-­considered permitted. The same type of normative factor, a prima facie permission, that allows acts with less than the best overall consequences in cases of supererogation can also allow them in cases of agent-­sacrifice. Finally, the theory we’ve proposed can explain how doing a supererogatory act, rather than permissibly omitting it, is in that deontic respect morally superior. On the one hand, the absolute status of the two is the same: both doing the 6 M.  Slote, Common-­Sense Morality and Consequentialism (London: Routledge & Kegan Paul, 1985), ch. 1.

470  Thomas Hurka & Evangeline Tsagarakis supererogatory act and not doing it are all-­things-­considered permitted but not required. On the other hand, there seems to be something especially laudable about going beyond duty. This is often in part a matter of the motive from which a supererogatory act is done. Someone who sacrifices his arm to save another’s life usually has a more benevolent, and therefore a more virtuous, motive than if, in the same circumstances, he preferred to preserve his arm. But if you sacrifice your arm to save another person only because he’s a billionaire who you hope will give you a large reward, your act, even though less admirably motivated, is still beyond duty and therefore supererogatory. The concept of supererogation is deontic rather than evaluative, one in the same category as right and wrong rather than like morally good or worthy, which depends entirely on motivation. A supererogatory act should therefore be superior deontically as well as, when that is the case, motivationally, and it can be so on the permission-­based theory. Though both a supererogatory act and its alternative exercise a prima facie permission, in one case to do the act and in the other not to, the first act also fulfils one or more prima facie duties. And a crucial feature of prima facie duties is that they don’t disappear when they’re outweighed. They leave what have been called ‘traces’, such as the appropriateness of feeling what Ross called ‘compunction’ about not fulfilling an outweighed duty, as when you break a promise because you’re helping accident victims, and sometimes a further duty to compensate the person to whom that duty was owed.7 An additional trace, we can now say, is to make a supererogatory act in this respect deontically superior to its equally permissible alternative. Though no better in its absolute deontic status, and independently of any value in its motive, it’s better in relation to prima facie moral factors because it fulfils prima facie duties as well as exercising a prima facie permission.8

3.  Parfit on Supererogation Whereas this theory weighs normative factors of two different types, prima facie duties and a prima facie permission, the more common theories of supererogation see your permissions simpliciter to do and not do a supererogatory act as arising from a conflict between two factors of the same type, which they often call ‘reasons’ and take to count, like prima facie duties, in favour of an act’s being one you have most reason, or ought simply, to do. The competing reasons, however, 7 Ross, The Right and the Good, 28. 8  Imagine that one person sacrifices his arm when this is his duty, while another does so when his act is supererogatory. The first person’s motive may be just as good as the second’s; his desire to promote the good impartially may be just as strong, so he too would have done the supererogatory act had that been his situation. Even so, we think the second person’s act is all things considered more admirable, because it is beyond duty whereas the first person’s isn’t.

More Supererogatory  471 reflect different ‘points of view’9 or have different contents and sources. Parfit gives one version of this theory. He recognizes two fundamental categories of reasons: impartial reasons to care about everyone’s good equally, and partial reasons to care about your own good or that of people close to you such as your children or spouse. The two kinds of reason can conflict, and in some cases one of them outweighs the other. If you can save thousands of people’s lives at minimal cost to yourself, your impartial reason outweighs your partial one and you’re required to do the saving. But in other cases neither kind of reason outweighs the other, nor are their weights exactly equal. They then in effect cancel each other out, and you’re permitted all things considered both to do the act favoured by impartial reasons and to do the one favoured by partial reasons, so the former is in that way beyond duty. Though the two kinds of reason are to some degree comparable, they’re only imprecisely rather than precisely comparable, since sometimes no determinate judgement of their relative strengths can be made. When that’s the case, the impartially better choice isn’t required but is supererogatory.10 This is an initially attractive result, but theories of this type have difficulty matching the attractive features of the permission-­based one above. Consider first the fact that supererogatory acts occupy a range. This wouldn’t follow from Parfit’s theory if impartial and partial reasons could be compared precisely. If partial reasons were, say, exactly five times as strong as impartial ones, you would be permitted both to give five units of good to another and to give one unit to yourself. But you would be required to give six units to her rather than one to yourself, since then the impartial reason would outweigh the partial one, and you would be forbidden to give four, three, or two units to her rather than one to yourself, since then the partial reason would weigh more. Hence Parfit’s insistence that the two categories are only imprecisely comparable, which implies that a certain kind of transitivity fails. If B is exactly equal to C, then if A is greater than B it’s also greater than C. But if B and C aren’t determinately ranked, so neither is greater than the other nor are they exactly equal, A can be greater than B but also unranked with respect to C. In our context this means that if the partial reason to retain an arm and the impartial reason to save another’s life aren’t determinately ranked, the reason to retain two arms can be stronger than the reason to retain one arm but likewise unranked with respect to the reason to save a life. You can be equally permitted to choose between saving a life and keeping one arm and between saving a life and keeping two arms; in both cases the saving is supererogatory.

9 S. Scheffler, The Rejection of Consequentialism (Oxford: Clarendon Press, 1982), ch. 2. 10 Parfit, On What Matters (Oxford: Oxford University Press, 2011), i. 137–41. Parfit doesn’t use the word ‘supererogation’, but what he describes is equivalent to it.

472  Thomas Hurka & Evangeline Tsagarakis Parfit’s claim of imprecise comparability is overwhelmingly plausible if read epistemically, as a matter of what we’re able to judge. If we’re asked to specify an exact level of cost to you at which saving another switches from being permitted to being required, we can’t do so. But epistemic imprecision is compatible with real or metaphysical precision. It’s possible that the normative truth is completely determinate, with exact weights for all normative factors, but we just can’t know them. (Ross seems to have taken this view about his prima facie duties: he thought we can never know which of two duties is stronger even though there’s always an objective truth of the matter.) So if Parfit’s theory is to yield a range of objective permissions, where a series of acts are actually supererogatory rather than just ones whose determinate status we can’t know, it needs metaphysical imprecision, where the moral truth itself is only partly determinate. That is a much more contentious claim. Some philosophers may deny that any truths can lack full determinacy; reality is always completely precise. But even apart from that it seems problematic to rest the intuitively compelling claim that we’re not always required to sacrifice for the greater good, and have a range of permissions not to do so, on an abstruse and contestable thesis in moral metaphysics. Would we really have to abandon our belief in supererogation if we learned that the normative truth is completely determinate? The metaphysical thesis seems of the wrong kind to ground the moral one. At one point Parfit suggests that the imprecise comparability of his two categories of reason follows from the fact that they differ in kind, the one involving a reference to yourself and the other not.11 But while some philosophers have argued that goods distinct in kind can’t be compared at all, that’s not his view; he thinks the impartial reason to save thousands of lives outweighs the partial reason to avoid some minimal cost to you. And it’s hard to see why a difference in kind that doesn’t entirely block comparison shouldn’t allow it entirely, as it would given metaphysical precision. Parfit’s theory can allow an objective range of supererogatory acts only by assuming a thesis of metaphysical imprecision or indeterminacy that’s both some distance from our belief in supererogation and less securely grounded than that belief. Our permission-­based theory, in contrast, requires no such thesis. Even if there’s a precise point at which the prima facie permission to promote your own good is outweighed by the prima facie duty to promote another’s, say the point where her good is five times as great as yours, you’re permitted to prefer a unit of your own to four, three, or two units of hers. Second, theories like Parfit’s that weigh partial against impartial reasons have difficulty accommodating agent-­sacrificing permissions. Consider a case where you can give either one unit of good to another person or two units to yourself. Here your impartial reason tells against giving the one unit to the other, since this

11 Parfit, On What Matters, i. 138–9.

More Supererogatory  473 will result in less total good than if you had the two. Your partial reason also tells against giving the other the one, since this will result in less good for you. So both kinds of reason tell against giving the other the one, and doing so is all-­things-­ considered or simply wrong: there’s no permission to prefer another’s even slightly lesser good. Moreover, on a natural reading these theories also deny the permission to prefer another’s equal or slightly greater good, as in some cases of supererogation. If the two types of reason aren’t determinately ranked, then just as sometimes an impartial reason outweighs a partial one, so sometimes a partial reason should outweigh an impartial one. This can’t occur when an act’s costs to you are greater than its benefits to others, since then your impartial reason, too, condemns the act. Nor can it occur when the costs are much or only moderately smaller, since then the impartial reason outweighs the partial one or the two aren’t determinately ranked. But it can occur when the costs to you are just a little smaller than the benefits, for example if you can either give either five units of good to another or four units to yourself. If partial reasons can ever outweigh impartial ones, they do so in cases like this, and the upshot is that some acts of preferring another’s greater to your lesser good are not only not required but positively forbidden. Some acts that would normally be considered supererogatory are wrong. Parfit doesn’t himself draw these conclusions. He acknowledges that some theories of the same general type as his say we’re always required to give our own good more weight than a stranger’s, but he rejects this claim as ‘too egoistic’. He says that if you’re in pain and have the only dose of morphine, you’re permitted to use the morphine on yourself but are also permitted to give it to a stranger in the next bed if he’s suffering the same pain as you, and even if his pain is somewhat less bad, so the benefit to him is somewhat smaller. More generally, he expresses sympathy for a view that allows you to give somewhat more weight to a stranger’s good than to your own, though not as much more weight as you may give your own.12 He therefore allows some agent-­sacrifice and presumably also allows, as supererogatory, any act whose cost to you is a little smaller than the benefit it gives another. These are attractive claims, but it’s hard to see how Parfit can accommodate them in the theory he initially proposes, where all-­things-­considered permissions result from conflicts between normative factors of the same kind, namely reasons that count in favour of an act by tending to make it one you have most reason, or ought simply, to do. This is certainly the way he understands his impartial reasons. When your reason to save thousands of lives outweighs your partial reason to 12 Parfit, On What Matters, i. 139–40. This last claim follows in the permission-­based theory if the prima facie permission not to promote your own good is weaker than the prima facie permission to promote it. Then the first permission is more likely to be outweighed by the duty to promote the good impartially, and permissions all things considered to agent-­sacrifice more limited than ones to agent-­ favour (see Hurka and Shubert, ‘Permissions to Do Less than the Best’, 13–14).

474  Thomas Hurka & Evangeline Tsagarakis avoid some minor cost, it makes the saving positively required rather than something that’s merely allowed. And how can a partial reason of this kind, one that positively favours promoting your good, prefer an act that leaves you with less good than some alternative would? In allowing agent-­sacrifice Parfit seems tacitly to understand the relevant partial reason not as one that tends to make for a requirement, as his impartial reason does, but as having merely permissive force, like what we’ve called a prima facie permission. It’s not that, in his case of permitted agent-­sacrifice, you have a positive reason to give the stranger with the less bad pain your morphine. It’s that you lack a positive reason to keep it for yourself, or have a permission other things equal not to keep it. The all-­things-­ considered agent-­sacrificing permission isn’t grounded in a normative factor of the same kind as an impartial reason, but in one from a different, merely permissive category. And that category is needed because, given only factors with the same requiring force, acts of agent-­sacrifice are doubly condemned. Finally, Parfit’s theory has difficulty explaining why a supererogatory act is deontically superior. If you save another’s life at the cost of your arm, your act better fulfils impartial reasons than if you preserved your arm, and so is impartially or, as Parfit would say, morally better. But if you don’t sacrifice your arm and let the person die, your act better fulfils partial reasons and is therefore partially, or prudentially, better. Each is therefore better by some standard, and there’s no ground to value one kind of betterness more than the other, or to say one of the two acts is in some neutral way superior. From a neutral standpoint the two acts are on a par. The permission-­based theory, in contrast, weighs factors with different kinds of force but doesn’t distinguish them as moral and prudential. It can say a supererogatory act is, though no different in its final deontic status than the alternative, nonetheless preferable because it both exercises a prima facie permission and fulfils a prima facie duty rather than just exercising a permission. Relatedly, and returning to our initial topic, a theory like Parfit’s has difficulty explaining how one supererogatory act can be more supererogatory than another. In the permission-­based theory this difference depends on a difference in the relations between the factors that make for supererogation: an act is more supererogatory when a prima facie permission outweighs prima facie duties by more. But in Parfit’s theory there is and can be no such difference: the relation between the factors it considers is the same in all supererogatory acts. This follows from the way, in his theory, imprecise comparability blocks transitivity. In our first example, your impartial reason to save the other’s life isn’t determinately ranked against your partial reason to keep one arm, but it also isn’t ranked against your partial reason to keep both your arms. That the second partial reason is stronger than the first has no effect on the relation between the two types of reason in these acts, which is the same. And it’s in fact the same in all cases of supererogation: in all such cases the relation between the competing reasons is just that of not being determinately ranked, so no differences in these relations

More Supererogatory  475 are ever possible. The very feature Parfit needs to generate a range of supererogatory acts, namely imprecise comparability, prevents him from recognizing degrees of supererogation. If it’s intuitive that some acts are more supererogatory than others, the fact that theories like his can’t allow that is a further strike against them.

4.  Non-­Supererogatory Admirability When one of two acts is more supererogatory, they don’t differ in their final deontic status: each is one you’re all-­things-­considered permitted to do and also to omit. They differ in certain accompanying properties, so the first is further beyond duty and therefore more heroic or saintly, where these properties in turn make different responses to them appropriate. Thus a more supererogatory act calls, other things equal, for more admiration or praise and may warrant more pride on the part of the agent. This last difference can even be a test of the phenomenon: it can be evidence that one of two acts is more supererogatory that more admiration or praise for it is fitting, and we may tacitly rely on this evidence when we compare the acts. The test must, however, be used with caution, because an act’s degree of supererogation isn’t the only factor that affects how much admiration it calls for; others can too. Imagine that you sacrifice your arm to save not one but five people’s lives, where the number of lives at stake makes your doing this your duty all things considered, or that a soldier sacrifices his life to save enough of his comrades that, given his special responsibilities to them, the sacrifice is morally required. Though these acts aren’t supererogatory, they’re surely still to a considerable degree heroic and ad­mir­ able; the soldier, for example, would surely still be given a medal for bravery. This suggests that another factor relevant to an act’s admirability is just the size of the cost you accept in order to act on a moral consideration, regardless of whether that consideration is outweighed, as in supererogation, or outweighs competing considerations, as when an act is obligatory. (This factor may increase the act’s admirability either by making its motivation more virtuous or in some purely deontic way; we leave this issue open.) But this can complicate our judgements of degrees of supererogation, and may do so in particular in the three examples in Section 1. The first example involves one act of saving another’s life at the cost of one arm and another act of saving him at the cost of two arms. The second act is more supererogatory because there’s a larger gap between the strengths of the prima facie permission to avoid its costs and the prima facie duty it fulfils; on that basis it’s more admirable. But this act is also more admirable because it accepts a larger cost, independently of that cost’s role in determining the size of a gap. We suggested in Section 1 that this example may seem especially compelling, and we can now see why. We may tacitly assess the two acts’ degrees of supererogation by seeing how far each calls for admiration, and here the second act’s greater

476  Thomas Hurka & Evangeline Tsagarakis admirability has two grounds, with that due to its degree of supererogation reinforced by that of its involving a greater cost. In the second example, however, there’s no such reinforcement. Here one act saves another’s life at the cost of one arm for you while the other saves him from losing two arms at the cost of one for you. The second act is more supererogatory because it involves a larger gap between permission and duty, but the greater admirability this makes for isn’t reinforced by any greater accepting of costs; the two acts have the same cost. As a result the second act isn’t more admirable by as much as in the first example, so an intuitive test for degrees of supererogation yields a less compelling verdict here. The second act still is somewhat more admirable, because it’s further beyond duty. But a natural indicator of greater supererogation isn’t as clearly present, and the same may be true in the third example. It compares saving another’s life at the cost of two arms for you with saving him from losing two arms at the cost of one arm for you. Here gap considerations make the second act more supererogatory—since it’s further from being required—and in that respect more admirable, but considerations of accepted costs make the first more admirable, and it may be unclear which act calls on balance for more admiration. Here the second factor not only fails to reinforce the first but positively weighs against it. There may also be a third relevant factor. It’s arguably more important to care about weightier moral considerations, or stronger prima facie duties, and some may say it’s therefore more admirable to fulfil these duties. (A contrary view would say it’s less admirable to fulfil them, because doing so is more expected or unquestionably demanded.) Accepting this view wouldn’t affect our judgements in the first example, where both acts fulfil the same prima facie duty. But it would be relevant to the second, where the less supererogatory act fulfils a stronger duty; there it would generate a ground of admirability that competes with that based on degree of supererogation. It would do the same in the third example, where saving a life at the cost of two arms fulfils a significantly stronger duty than saving two arms at the cost of one. There it might combine with the factor of costs to make the first act on balance more admirable. Even then, however, the second would call in one respect for more admiration, since it goes further beyond duty. It’s just that, in this example, the other two factors would make the effect of this particular one especially hard to see.

5.  Conclusion If supererogatory acts are beyond the call of duty, it seems that some can be further beyond duty than others and therefore more supererogatory. We’ve given an account of this phenomenon within a general theory that sees supererogation occurring when a prima facie permission to promote your own good outweighs a

More Supererogatory  477 prima facie duty, for example to promote the good of all impartially. This account uses the same tools as accounts of how other deontic concepts such as wrongness admit of degrees, and therefore is continuous with those. It also strengthens the claims of the permission-­based theory of supererogation, in particular against ones like Parfit’s that seem unable to allow degrees of supererogation.

Bibliography Hurka, T., ‘More Seriously Wrong, More Importantly Right’, Journal of the American Philosophical Association 5 (2019): 41–58. Hurka, T., and Shubert, E., ‘Permissions to Do Less than the Best: A Moving Band’, Oxford Studies in Normative Ethics 2 (2012): 1–27. Parfit, D., On What Matters, vol. i (Oxford: Oxford University Press, 2011). Parfit, D., On What Matters, vol. iii (Oxford: Oxford University Press, 2017). Ross, W. D., The Right and the Good (Oxford: Clarendon Press, 1930). Scheffler, S., The Rejection of Consequentialism (Oxford: Clarendon Press, 1982). Slote, M., Common-Sense Morality and Consequentialism (London: Routledge & Kegan Paul, 1985).