Philosophy and Phenomenological Research Vol. LXXII, No. 2 , March 2006

Prkcis of Justice, Luck, and Knowledge’

S . L. HURLEY

University of Bristol

Justice, Luck, and Knowledge (JLK) contributes to recent developments in two areas, moral responsibility and distributive justice. Prominent luck- neutralizing approaches to distributive justice, exemplified in work by Cohen and by Roemer, argue that justice requires equal distribution of goods for which people aren’t responsible. Such views of justice haven’t focused atten- tion on responsibility itself. Meanwhile, responsibility has been illuminat- ingly articulated in work including, and influenced by, Frankfurt’s seminal essays. My book brings these separate lines of work, on justice and on responsibility, into contact, examining how the new articulation of responsi- bility constrains the roles responsibility can play in distributive justice. Part I focuses on responsibility and its inverse correlate, luck; Part I1 assesses responsibility-based approaches to justice in light of preceding arguments about responsibility. (See JLK, 4-5 on responsibility, reactive attitudes, and accountability.)

Chapter 1 provides a philosophical landscape against which subsequent arguments develop. Issues about responsibility are reconfigured by the impor- tant distinction, deriving from Frankfurt’s arguments, between actual- sequence and alternate-sequence conditions of responsibility. This distinction has been further articulated by others and has especially influenced the devel- opment of reason-based views of responsibility. The new articulation of responsibility is explained by reference to work by Klein, Wolf, and Fischer. While exploring this territory, I explain why two putative requirements for responsibility are independent of one another: the alternate-sequence ability- to-do-otherwise requirement, and the actual-sequence regression requirement (which says that responsibility for something requires responsibility for its causes). Causal conceptions of moral responsibility, e.g. involving actual choice or control, are distinguished from noncausal conceptions, e.g. hypo- thetical choice conceptions; the latter leave responsibility indeterminate. The

* S. L. Hurley, Justice, Lurk, and Knowlcdgc (Cambridge, Massachusetts: Harvard University Press, 2003).

418 s. L. HIJRLEY

regression requirement is distinguished from other actual-sequence require- ments, including various reason-responsiveness requirements. I explain why requiring both regression and choice or control makes responsibility impossi- ble. I consider why the major threat to responsibility is the regression requirement rather than determinism, and how reason-based accounts can respond to it.

Chapter 2 argues that alternate sequences are irrelevant to responsibility. My argument is tangentially related to Frankfurt’s. A recipe is developed for eliciting the irrelevant-alternative intuition: What matters for responsibility is not the outright possibility of doing otherwise, all else constant. If the agent wouldrz’t have done otherwise whether or not she could have, then it’s irrelevant whether or not she could have. This intuition significantly gener- alizes the intuition evoked by Frankfurt-style cases, and applies to more cases; it’s illustrated by cases involving weak will, evil, and medical prob- lems. I explain the irrelevance in terms of variable realization of a disposition across deterministic and indeterministic worlds.

Chapter 3 assesses the view that responsibility requires regressive choice or control, and is therefore essentially impossible. General issues are raised about what warrants elimination of an entity or property, as opposed to revi- sion of beliefs about it, and how this question connects with the distinction between context-driven and theory-driven views of reference and essence. Con- text-driven views tend to be less hospitable to eliminativism than theory- driven views, but this tendency shouldn’t be overstated. Since both types of view give essences explanatory depth, eliminativism associated with sup- posed impossible essences are problematic for both. These general consid- erations are applied to show that the impossibility of regressive choice or control doesn’t support eliminativism about responsibility, on either context- or theory-driven views of responsibility. To avoid indeterminacy, responsi- bility needs a causal component, e.g. a requirement of actual choice or control rather than purely hypothetical choice; such causal requirements are incompatible with a regression requirement. From this incompatibility we shouldn’t conclude that responsibility is impossible, but that responsibility needn’t be regressive.

Chapter 4 probes the concept of luck that features in moral and political philosophy, distinguishing a thin concept of luck from various thick con- ceptions. Thin luck is simply the inverse correlate of responsibility; what is a matter of thin luck for someone is what he is not responsible for. On alter- native thick readings, luck has more specific implications, concerning e.g. aspects of lotteries, or lack of control, or lack of choice. I argue that moral and political philosophy should use ‘luck’ in the thin sense. Applying the thirdthick distinction to ‘constitutive luck’, I criticize the thick conception of constitutive luck expressed by the idea of a ‘natural lottery’. The concept of a

BOOK SYMPOSIUM 419

lottery has specific implications, about the identities of agents who enter lotteries and the role of chance in lotteries. Both implications are misleading applied to constitutive luck. The identity-dependence of lotteries generates a bare self problem; moreover, the relationship of chance to responsibility is not what the lottery metaphor suggests, for either metaphysical or epistemic chance. The natural lottery metaphor is positively unhelpful in understanding constitutive luck; we do better to avoid it and to understand constitutive luck thinly, simply as lack of responsibility for one’s constitution.

To close Part I, I highlight certain of its arguments and preview how Part I1 will relate them to responsibility-based theories of justice. Egalitarians believe distributive justice favors equality across persons. But equality of what‘? Answers have given luck and responsibility increasingly explicit and central roles in recent decades. Yet luck and responsibility have tended to operate as black boxes within accounts of justice. When they’re opened, I argue, the roles of responsibility in justice need reconfiguring and relocating. Luck-neutralizing views focus on intuitions about justice and reverse into the role of responsibility in justice. Instead, I focus on various conditions of responsibility, their relations to control, choice, etc., and assess the possible roles in justice of responsibility thus articulated.

Part 11 begins with another philosophical landscape in Chapter 5 , which surveys the luck-neutralizing approach’s development, from Rawls, through Sen and Dworkin, to Cohen and Roemer.

Cohen has claimed that a large part of the fundamental egalitarian aim is to neutralize the influence of luck on distribution (1989, 602). Various inter- pretations of this claim are possible, but I don’t address these exegetical issues. Rather, Chapter 6 argues on the merits that the luck-neutralizing aim cannot provide a basis for egalitarianism in either of two senses: it can nei- ther specify nor justify an egalitarian pattern of distribution. Luck and responsibility can play roles in determining whar justice requires to be redis- tributed, but from this we cannot derive how to distribute: from the currency of distributive justice, we cannot derive a pattern of distribution-let alone an equal pattern. Nor does responsibility provide a basis for taking equality as normative default, departures from which require justification. In the luck neutralizer’s dilemma, bad luck can be understood either interpersonally or counterfactually. Aiming to neutralize interpersonal bad luck, on the one hand, specifies an egalitarian distributive pattern only trivially, since interper- sonal inequality defines bad luck; for the same reason, this aim presupposes rather than provides reason for egalitarianism. Aiming to neutralize counter- fucrual bad luck, on the other hand, neither specifies nor justifies an egalitar- ian pattern. Judgments about different individuals’ responsibility for choices and their outcomes cannot generally determine how to distribute between people, since such judgements aren’t about interpersonal relations and don’t

420 S . L. HURLEY

generally extend determinately to counterfactual situations. When actual dis- tribution is a matter of luck, it’s usually simply indeterminate what counter- factual distribution would not be a matter of luck. This point also threatens the implicit but implausible assumption, made by appeals to hypothetical luck-free choices, that responsible choice and luck are separable.

If no one were responsible for anything and everything were a matter of luck, would this support an egalitarian pattern? No. Consider the aim to neu- tralize the effects, on a given outcome, of luck occurring anywhere back in the causal chain leading to that outcome. This aim operationalizes the regres- sion requirement, which threatens to make responsibility impossible, by effectively treating everything as among the effects of luck: Neutralizing the efsects of luck on X is logically equivalent to neutralizing the effects of luck on X of any causes of X that were matters of luck. Any X must have such causes (whether deterministic or indeterministic), far enough back. But even if everything is an effect of luck, there’s still no reason to think equality would neutralize luck’s effects. Moreover, if responsibility is impossible, then it’s impossible to neutralize the effects of luck. Can identical autonomous bare selves-supposedly remaining when everything that’s a matter of luck is stripped away from the self-support equality? Unlikely; there are powerful reasons to doubt such bare selves exist.

Chapter 6 doesn’t argue against egalitarianism, or even against taking equality as default when people aren’t responsible for their holdings. Rather, it argues that responsibility and luck provide no basis for these views. But there may be another basis for them.

Roemer gives an illuminating account of what it would be to neutralize luck. Chapter 7 examines whether Roemer’s account shows how the luck- neutralizing aim could provide a basis for egalitarianism (though this may not have been Roemer’s intention). I argue Roemer’s account cannot do this job. Rather, it shows how to reward people equally who make equal efforts to behave in ways we regard as deserving. But giving people what they deserve isn’t the same as giving them what they are responsible for ( J L K , 198). Nor i s equalizing what people are not responsible for the same as giving them what they are responsible for. Depending on what behavior we view as deserving, a Roemerian reward system may or may not favor equal distribu- tions. Treating like cases alike in Roemer’s way doesn’t guarantee that rela- tively equal patterns of distribution are favored over relatively unequal ones, other things equal. Nevertheless, applied to specific, democratically adopted policies rather than treated as a general account of distributive justice, Roemer’s reward scheme may be a useful tool of policy implementation.

Chapters 6 and 7 emphasize the distinction between what to distribute and how to distribute, and why responsibility cannot tell us how to distribute -or to distribute equally in particular. For responsibility to play the ‘how’

BOOK SYMPOSIUM 421

role, it would have to determine, when the level of goods someone actually enjoys is a matter of luck for him, what other, counterfactual level of goods would not be. But such questions of counterfactual responsibility don’t generally have determinate answers. However, these arguments against the ‘how’ role leave it open that responsibility could play the ‘what’ role, determining what goods to redistribute. It could contribute to defining the currency of distributive justice in theories about which goods are exempt from redistribution to the extent people are responsible for them, since justice is concerned with redistributing only goods that are a matter of luck for people. Responsibility could thus operate as a filter on some independently specified category of goods (e.g. welfare or resources). Chapters 8 and 9 assess this goods-filtering, currency-defining role for responsibility in distributive justice.

Chapter 8 explores issues raised by Cohen’s arguments about incentive inequality (these issues are pursued further in Hurley 2006). Cohen asks whether the normative and supposedly descriptive premises of the familiar maximin argument for incentive inequality are inconsistent: is it inconsistent for the talented to aim to make the worst off as well off as possible, while also choosing to demand incentives and refusing to work harder without additional return? By making the latter choices, the talented fail to do what they could to make the worst off still better off. However, I consider the danger of a different inconsistency: is it inconsistent to aim to redistribute to the worst off only what is not down to choice, while also holding that the talented can and should choose to work harder for less, to benefit the worst off? If goods for which people are responsible as a result of their choices are (at least partly) exempt from redistribution, why aren’t the extra goods result- ing from choices by the talented to work harder (at least partly) exempt? Chapter 8 relates Cohen’s arguments about the currency of distributive justice to his separate arguments against incentive inequality, under various assumptions about responsibility. (These arguments are compared on the merits, not in criticism of Cohen; he makes his currency and incentive inequality arguments in two different contexts.)

By now I’ve argued responsibility cannot play a patterning role, determin- ing how to distribute, and considered how it could have a currency role, in determining what to redistribute. Chapter 9 brings into focus two further roles responsibility can play in distributive justice, the incentive-parameter and well-being roles. These, I argue, are the most important roles for respon- sibility in justice. When their relations to the ‘what’ role are examined, problems emerge for the latter, which I conclude is problematic and better dispensed with.

Beliefs about responsibility can play important roles in motivating pro- ductive activity, including as parameters on which the range of possible

422 S . L. HURLEY

incentive-seeking behavior depends. Cohen suggests that whether the talented can work harder for less may depend on alterable normative expectations. The latter, I argue, include expectations based on prevalent views about responsi- bility and corresponding beliefs by the talented about their own responsibility for what they do. Altering beliefs about responsibility could alter such expectations, and so alter the range of possible levels of incentive-seeking; beliefs about responsibility could thus act as parameters on which the range of possible incentive-seeking behavior by the talented depends. From the per- spective of a luck-neutralizing egalitarian social engineer, how should such parameters be fixed? Cohen makes the important point that until this range is determined, a maximin principle of justice has indeterminate implications.

Moreover, facts and beliefs about responsibility can constitute and influ- ence people’s well-being in important ways. How should the egalitarian social engineer factor this well-being role into the parameter-setting and currency-filtering issues? I consider how the incentive-parameter and well- being roles of responsibility interact with the currency role. Do they conflict? Constrain one another? Create unstable interdependencies? Admit of stable equilibria?

Chapter 9 concludes by viewing incentive inequality issues from an alter- native perspective, in terms of aiming to neutralize bias rather than luck. Aiming at knowledge provides reason to adopt a perspective of ignorance in thinking about how goods should be distributed: self-ignorance would rule out many biasing influences, including those deriving from self-interest. A veil of ignorance helps avoid biasing influences when addressing issues about what it’s possible for the talented to do. A bias- (rather than luck-) neutraliz- ing approach to justice retains the incentive-parameter and well-being roles for responsibility, but dispenses with the ‘what’ role as well as the ‘how’ role. In this approach responsibility has essential but different, less demand- ing roles, better suited to its nature. Implausible assumptions are avoided: that responsible choice and luck are separable, or that it’s generally determi- nate what people would be responsible for under counterfactual hypotheses about luck.

Finally, Chapter 10 pursues the bias-neutralizing approach to justice, which illustrates a cognitivist perspective in political philosophy. Political liberals like Rawls deny that aiming at knowledge can provide a basis for justice in a pluralistic democratic society. But I argue that a cognitivist approach to justice needn’t threaten the liberal values of pluralist democracies. Moreover, the bias-neutralizing aim can provide a stronger basis for egalitarianism than the luck-neutralizing aim, since the latter puts responsi- bility in roles it is not suited to play, while the former doesn’t. The bias- neutralizing aim provides an argument for a maximin principle of distribution that avoids standard objections to luck-related arguments for maximin. I tease

BOOK SYMPOSNJM 423

apart luck-neutralizing and bias-neutralizing strands in familiar arguments for maximin, and relate them to the decision-theoretic distinction between risk aversion and uncertainty aversion; the bias-neutralizing argument for maximin assumes weak uncertainty aversion, but not risk aversion. The normative significance of a veil of ignorance is found in the aim to avoid biasing influences that inevitably go with information, even probabilistic information, about who you are and what you’re like, your talents and dis- abilities. Moreover, weak aversion to uncertainty can be justified as part of the minimal rationality required for intentional agency, while risk aversion cannot. By taking a cognitivist perspective and dissociating justice from the luck-neutralizing aim and related ideas, we can reclaim egalitarian results.

JLK demonstrates how issues about responsibility interact with issues about distributive justice and with the roles responsibility can play in justice. It shows why it’s important for theories of distributive justice to distinguish responsibility from other values, e.g. deservingness or fairness, to distinguish different conceptions of responsibility, and to distinguish different roles responsibility might play in justice.

References Cohen, Gerald (1989). ‘On the currency of egalitarian justice’. Ethics 99(4):

906-944. Hurley, Susan (2006). ‘Choice and incentive inequality’. In The Egalitarian

Conscience, ed. Christine Synopwich. Oxford: Oxford University Press, 130- 153.

424 s. L. HURLEY