In what follows, I discuss two issues that are often treated as separate in discussions of moral

What do I Owe? Moral responsibility and circumstantial luck 1 Hallvard Lillehammer, Birkbeck, University of London. <h.lillehammer@bbk.ac.uk> 1. Two ways of thinking about moral responsibility In what follows, I discuss two issues that are often treated as separate in discussions of moral responsibility. The first is that there is something problematic about the attribution of moral responsibility for aspects of our situation that are beyond our control. The second is that there are situations in which the recognition of responsibility involves responding to vulnerability or risk involving morally significant others. The aim of discussing these two issues together is threefold. First, I want to explore the question whether there is any interesting relationship between the conceptions of moral responsibility at work in each case, and if so what that relationship is. (This is emphatically not to deny that they are, in ways I go on to discuss, also significantly different.) Second, I want to explore the wider significance of a kind of moral responsibility (more familiar to some than to others) that is sometimes felt in the face of misfortune or disaster involving morally significant others; where those others may or may 1 Parts of this material have been previously presented at a Birkbeck Guilt Colloquium in 2014; at an Ethics at the Intersection of Philosophy and Anthropology workshop on Rethinking Responsibility at Birkbeck, in 2015; at a CRASSH conference on Hierarchy, Egalitarianism and Responsibility at Cambridge in 2016; and at a colloquium on Modalities of the Good, organized by the Czech Academy of Sciences and Charles University, Prague in 2016. I am grateful to members of the audience on each of these occasions for their comments and reactions to the material presented here, and to Neil Carrier for many conversations about the moral significance of non-voluntarily received benefits. 1

not stand in any interesting relationship to some previous exercise of our own will, but where the misfortune or disaster in question is not in any interesting sense of our making (e.g. situations where I may be said to have non-voluntarily benefited from some past harm done by one third party to another). Third, I want to explore the connections between situations of this kind and a set of concepts by which individuals and groups of people are described in morally significant terms (e.g. as fellow humans, EU Citizens, middle aged white men, etc.); but where the moral significance of those descriptions is not obviously reducible either to considerations of valuable states of affairs impartially considered (such as the Utilitarian invocation of a point of view of the Universe ) nor to volitional considerations of individual agency (such as the Kantian invocation of a free and rational will ). The ultimate aim (but one that goes beyond the aims of this paper) is to give an account of this set of concepts as ones in terms of which morally significant descriptions can be grounded in a conception of individuals and groups of people as socially situated persons whose moral status is partly defined by the relations in which they stand to other individuals and groups as parts of ongoing histories, contingent practices or institutions, and socially defined goals and expectations that may often cut across the recommendations of any act of autonomous willing (either on their own part or on the part of others), as well as the impartial requirements made visible (if at all) from the point of view of the Universe. To that extent, I will be taking the discussion of moral responsibility and moral luck in a somewhat different direction than that which has been taken in other recent discussions of how luck affects what Williams called agent regret (see Nagel 1979: Williams 1981). The line of thought I pursue here will therefore be orthogonal to the concerns of many of those who have recently made insightful contributions to the literature on moral luck, where the main focus has tended to be on how we should think about cases where prior acts of individual volition are in some sense interestingly in play (e.g. Williams s case of Gaugin, who leaves his family behind in order 2

to pursue the uncertain path of pursuing his artistic ambitions (c.f. Slote 1994; Zagzebski 1996; Athanassoulis 2005; Raz 2009; Hwang 2013)). 2. Moral responsibility and the problem of moral luck According to one standard formulation, the problem of moral luck arises from the allegedly i) questionable legitimacy, of ii) genuinely moral assessments, iii) of individual agents, in cases where iv) significant aspects of what the agents are assessed for, depend on v) factors beyond their individual control (Nelkin 2014; my italics). Thus understood, the problem arises from a conflict between what is taken to be a fact ; namely that people do morally assess each other for things that depend on factors beyond their control (e.g. in the case of a careful driver who is unfortunate enough to run over a pedestrian), and a moral principle ; namely that people are morally assessable only to the extent that what they are assessed for depends on factors within their control. Thus understood, the problem of moral luck depends on the joint plausibility of a fact and a principle. Yet the mere fact that people morally assess each other for things that depend on factors beyond their control does not entail that it makes any sense for them to do so. I therefore prefer to interpret the fact in question as implicitly giving expression to a second principle, namely that genuine moral assessment of people can depend on facts beyond their control; in which case what we are faced with is a case of normative (in this case, moral) conflict, and the question is whether (and if so, how) this can be resolved. To be clear from the start, I assume throughout that the first principle on which the apparent conflict depends does express at least one genuine insight, namely that the presence of luck is (always, and at least all else being equal) a morally significant factor in a situation; hence a genuine moral difference maker ; hence that the presence of luck in a 3

given situation does, in itself, give grounds for evaluating people in that situation differently than one otherwise would (e.g. by mitigation or excuse). My question is therefore not if luck makes a moral difference to the moral evaluation of people. It is whether it makes all the difference, thus excluding the possibility of moral evaluation of people where, and to the extent that, luck obtains. In his original response to Williams s seminal paper on this topic, Thomas Nagel distinguishes between four kinds of moral luck (Nagel 1979; Williams 1981). The first kind of moral luck identified by Nagel is resultant luck, or luck with respect to how things turn out (e.g. drink driving with or without disastrous consequences, or successfully or unsuccessfully pursuing an artistic career). This is a kind of luck on which many recent philosophical discussions of moral luck have been focused. Central to these discussions of moral luck is the idea of an autonomous individual who sets herself an end and pursues it rationally, knowingly, and to the best of her ability, with moral assessment being focused mainly on the quality of the end set and the volitional efforts made by the individual in realizing it. The main issue in contention here is that given that the world may not oblige (and often does not), the contingent consequences of our actions might be thought to fall outside the domain of all moral assessment. Hence the problem of moral luck. The second kind of luck identified by Nagel is causal luck, or luck in how the workings of our agency are determined by antecedent circumstances (e.g. the alcoholic dispositions I have acquired by a combination of genetic inheritance and a faulty upbringing). Because of the obvious connections between this issue and the perennial issue of free will, this kind of luck is also one on which many philosophical discussions of moral luck have focused. Kant apparently believed that this kind of luck can be overcome in virtue of our possession of a rational and autonomous will, and that our possession of such a will is a transcendental 4

precondition of all genuine moral assessment (Kant 1981). More recent discussions of moral luck have picked up on this thought in connection with the ethics of blame, punishment and reform, if often on more naturalistic terms. The third kind of luck identified by Nagel is constitutive luck, or luck with respect to who we, intrinsically, are (e.g. being a white chromosomally male member of the species homo sapiens). The issue of constitutive luck is intimately connected both with the issue of causal luck (e.g. our morally relevant features as being determined by prior causes) and with the issue of resultant luck (e.g. our morally relevant features developing as a result of what we do and what happens to us). What we say about these issues will depend to a large extent on what we say about the first two kinds of luck identified by Nagel. In what follows, I have comparatively little to say about causal and constitutive luck. I have a bit more to say about consequential luck, but only after I introduce the fourth kind of luck that Nagel identifies. The fourth kind of luck identified by Nagel is circumstantial luck, or luck with respect to the circumstances in which we find ourselves (e.g. as net beneficiary of an illegitimate war conducted by our ancestors, or a net contributor to global warming). This fourth kind of moral luck has received comparatively little attention in recent discussions of the subject. (Two exceptions are Urban Walker 1991 and Hanna 2014). In what follows I address the issue of moral luck by focusing specifically on circumstantial luck as it applies in conditions where individuals are faced with misfortune or disaster involving others, and where the misfortune or disaster in question is not (or not primarily) of their own making. The point of focusing on this case is not that it is the only case worth discussing. The point is that this is a case where the issues I discuss here arise in a particularly vivid way. Four preliminaries before moving on. First, there is nothing especially suspicious about focusing on the issue of circumstantial luck when addressing questions about moral 5

responsibility (c.f. Hanna 2014). True, circumstantial luck may involve mere contingencies of my situation in a way that constitutive luck does not, but so does resultant luck. True, significant circumstantial features of my situation might be possible to alter in the future without great cost in a way that significant consequences already produced (such as another person s death) are not, but the same is true of causal luck. (This will become relevant later when I discuss the issue of our responsibilities to respond in certain ways after the fact.) And so on. Second, I shall be talking freely throughout about the facts of someone s situation calling for a response. The facts in question may include both facts within and outside an agent s control. Among the former are facts about what we may or may not be able to autonomously, rationally, or voluntarily bring about. As I use the terms, these are all facts of our situation, and for each of them we can ask if, and if so how, they affect some attribution of moral responsibility. In some of its formulations the alleged problem of moral luck is a product of drawing a sharp distinction between these facts of an agent s situation and others, as far as their moral significance is concerned. The point of the discussion that follows is partly to explore the nature and basis of this distinction (for example, whether it is sharp or not, or whether it signifies a significant shift in moral evaluation with respect to what we say that individuals or groups of people owe by way of a response, and if so, why). Third, there are ways of interpreting the problem of moral luck that might be thought to make the following discussion redundant. These might include: i) defining moral assessment exclusively in terms of a backward-looking (or retributivist) conception of praise and blame; ii) defining moral assessment exclusively in terms of a narrow set of deontic notions, such as permissibility or obligation; iii) defining moral assessment exclusively in individualist (as opposed to collective or corporate) terms; or iv) defining moral assessment so as to exclude consideration of the kind of thick or relational moral concepts that are sometimes 6

said to be constitutive of the ethical (Williams 1985; Margalit 2002). Given that the best way to understand the problem of moral luck is part of what is at issue, I leave these questions open at this point. Fourth, I have so far proceeded without saying anything general about what moral responsibility is. To repeat: I do not assume at this (or any) point that we are dealing with a single notion of moral responsibility that can be stipulated at the outset without implicitly loading the question in any problematic way. Having said that, when talking about moral responsibility In what follows, I will always be talking about some form of accountability with respect to what people do, are, or the situations in which they find themselves; and where someone actually being accountable (to some extent, or at all) in the relevant way is a matter of what they can be justifiably demanded, asked or advised to do, say, think or feel (e.g. by being said to have a duty to pay, to apologize, to show remorse, or just to remember). Quite what is involved in holding someone accountable in a distinctively moral way is not something I have anything conclusive to say about here (e.g. whether we should favour a content-based or an attitude-based definition of morality or moralities, or some combination of both); although I do address the issue in passing. 3. Finding a place for circumstantial moral luck In some contemporary jurisdictions, failure to respond to acute vulnerability or risk involving others has been sanctioned by law, in the form of so-called Good Samaritan legislation, at least since the middle of the Twentieth Century. Other jurisdictions have been more reluctant to engage in this kind of enforcement of altruism or legislation of morality (Feldbrugge 7

1965). Yet whatever the legal situation may be, it is widely thought that actively responding to acute vulnerability or risk, either by providing assistance or at least reporting it, is a good, admirable or virtuous thing to do; and that responding to acute vulnerability or risk with indifference or hostility is bad, vicious, wrong, or impermissible (Lillehammer 2014). Hence it is widely thought that when faced with a situation involving acute vulnerability or risk a morally responsible agent will respond accordingly, even if it is often hard to tell what the best way to address the vulnerability or risk in question is (e.g. who, if anyone, is responsible for taking the initiative to address it; how the responsibilities to address it can be most effectively exercised; how the responsibility to address it is best distributed; whether enough is already being done by others to address it, and so on). However uncertain we may be about the best way to respond in any given situation, either individually or collectively, most of us are quite familiar with norms addressing these kinds of cases, either in legalistic or more informal ways; norms which presuppose that we are being called upon to make some significant response (e.g. what to do when observing an accident on the motorway; when witnessing a medical emergency on the underground; or how best to respond to a crisis involving refugees or migrants, and so on). Events like these normally involve a substantial element of circumstantial luck (c.f. Zimmerman 1987; Urban Walker 1991; Silcox 2006; Hanna 2014). Consider, by way of example, the following two, purely schematic, cases of people (for reasons that will hopefully have become clear by the end of the paper, I label them citizens ) with respect to actual or possible situations involving the acute vulnerability or risk of others, and where those situations are in no way of their own making. Consider first the case of Citizen A, who would offer protection or help to an endangered other; who is confronted with an endangered other; and who does offer protection or help. Consider next the case of Citizen B, who would offer protection or help to an endangered other; who is not confronted with an endangered other; 8

and who therefore does not offer protection or help. The way things actually go for Citizens A and B, Citizen A is someone whom others may subsequently come to hold in particularly high esteem. Yet the only relevant difference between Citizens A and B is that Citizen A (but not B) was actually faced with an extreme situation in a way that was beyond their control, and therefore had the fortune (if that is the word) of being able to display their admirable disposition to respond the way the extreme situation called for. In at least one morally relevant respect that many readers will find quite familiar, holding Citizen A in relatively higher esteem compared to Citizen B is not unfair to B, given the very real difference between what A and B actually did. Of course, there is another respect in which Citizens A and B are on a par: they both exemplify a morally admirable disposition. Yet there is a further respect in which they are importantly different: only one of them (i.e. Citizen A) actually responded to the acute vulnerability and risk of a morally significant other; thus setting a good moral example of a certain kind. This is one respect in which the moral evaluation can be affected by circumstantial luck. Now consider two further schematic cases of people with respect to actual or possible situations involving the acute vulnerability or risk of others, and where those situations are in no way of their own making. Consider first the cold-hearted Citizen C, who would not offer protection or help to an endangered other; who is confronted with an endangered other; and who does not offer protection or help. Consider next the case of Citizen D, who would not offer protection or help to an endangered other; who is not confronted with an endangered other; and who does not offer protection or help. The way things actually go for Citizens C and D, Citizen C is someone whom others may subsequently come to hold in particularly low esteem. Yet the only relevant difference between Citizens C and D is that Citizen C (but not D) was actually faced with an extreme situation in a way that was beyond their control, and therefore had the misfortune (if that is the word) of being able to display their non- 9

admirable disposition to respond the way the extreme situation called for. Once more, in at least one morally relevant respect that many readers will find quite familiar, holding Citizen C in relatively lower esteem compared to Citizen D is not unfair to B, given the very real difference between what C and D actually did. Of course, there is another respect in which Citizens C and D are on a par: they both exemplify a morally non-admirable disposition. Yet there is a further respect in which they are importantly different: only one of them (i.e. Citizen C) actually failed to respond to the acute vulnerability and risk of a morally significant other; thus setting a bad moral example of a certain kind. Once more, this is one respect in which the moral evaluation can be affected by circumstantial luck. Citizens A and B can be distinguished from Citizens C and D by their possession of an admirable disposition that C and D lack. This, I am assuming for the sake of argument, is not a matter of luck. Yet Citizens A and C can also be distinguished from Citizens B and D by actually being presented with a situation in which they go on to manifest the relevant (admirable in the case of A, non-admirable in the case of C) disposition. (Not so for B and D.) This, I am assuming for the sake of argument, is a matter of luck. Hence, there are two separate dimensions of moral evaluation along which Citizens A, B, C and D can be compared, namely i) that of having (or not having) some admirable disposition, and ii) that of how someone actually acts (or does not act). Alongside dimension i), Citizen B is on a par with A. Yet alongside dimension ii), Citizen B is on a par with C and D. Yet Citizen B is not on a par with C and D all things considered. Hence, there can be no reasonable suggestion that Citizen B is a suitable object of censure in the way that C is, in virtue of how C s coldhearted dispositions actually does prevents him from acting in the way the situation calls for when he is confronted with an extreme situation. Nor can there be a reasonable suggestion that Citizen B is a suitable object of censure in the way D is, even if D, just like B, has the fortune (if that is the word) of not actually having the (poor) moral quality of his disposition 10

tested. If so, a reasonable all things considered moral assessment of the schematically described cases of A, B, C and D will allow us to make sense of a comparative judgement in which B is held in lower esteem than A; C is held in lower esteem than D; and D is held in lower esteem than B, even though none of B, C or D have actually responded to the acute vulnerability or risk of a morally significant other. If genuine moral assessment excludes all elements of circumstantial luck, then this comparative judgement would not make sense. Suppose, then, it makes sense to applaud Citizen A for actually displaying what could, in some circumstances, amount to an act of moral heroism. Suppose further that it makes no sense to applaud Citizen B in the same way (and, by parity of reasoning, that it makes sense to negatively evaluate Citizen C for his or her actually displayed callousness in a way it does not make sense to negatively evaluate Citizen D for merely having the same disposition). If so, then some moral assessments are compatible with elements of circumstantial luck. If the presence of some element of circumstantial luck is what explains why someone (e.g. Citizen A) rather than someone else (e.g. Citizen B) can be held responsible for either responding or not responding to a morally salient aspect of their situation, then some coherent attributions of moral responsibility are not exhausted by what is within a person s individual control. The claim that there are some situations that call for a response in virtue of the vulnerability or risk they involve to others therefore conflicts with the claim that the coherent attribution of moral responsibility is fully constrained by what is within a person s individual control. If so, are forced to choose between either accepting the existence of circumstantial moral luck, or being left without anything morally substantial to say about what people actually do on the basis of their (admirable or not so admirable) dispositions. A brief pause to clarify what has been said so far; and what this implies about attributions of moral responsibility. First, I have claimed that moral evaluations of people and their actions can be affected by the quality of the dispositions they possess in any given situation (e.g. the 11

disposition to assist vulnerable others). I take this to be relatively uncontroversial in the present context. Second, I have claimed that moral evaluations of people and their actions can be affected by whether or not the relevant dispositions are actually manifested in action in any given situation (e.g. whether they actually assist any vulnerable others). Given some ways in which the control condition has historically been interpreted (e.g. by some Kantians), this is more controversial in the present context. Third, I have claimed that moral evaluation of people and their actions can be affected by facts about the situation in which they find themselves, quite independently of the moral quality of those people s independently describable dispositions (or, indeed, any other feature of those persons that is somehow within their control ). This is the real point of controversy in the present context. I make this claim because I think it offers a good explanation of the truth of the second claim, as follows. Assume that facts about a person s situation beyond the quality of their dispositions have no implications for their moral standing. If so, it will make no difference to the moral evaluation of that person whether they actually manifest that disposition on a given occasion or not. Yet it does sometimes make a difference to the moral evaluation of a person whether they actually manifest that disposition on a given occasion or not. So facts about a person s situation beyond the quality of their dispositions do have implications for their moral standing. 2 Yet what, exactly, does it mean to say that the moral evaluation of people and their actions can be affected by facts about the situation in which they find themselves, independently of moral quality of those people s dispositions? I distinguish between two interpretations of this claim of different strength, both of which I think are worth taking seriously (and the second 2 The argument in this paragraph is consistent with the independently plausible claim that we should be very reluctant to attribute a disposition to someone if that disposition (or some developmental predecessor of that disposition) were not actually manifested on a significant number of occasions. 12

of which will be important in what follows). On the first (and weaker) interpretation, being affected by implies that some fact about a person s situation makes a difference to whether or not that person has some moral standing, in the sense that it is one among a larger set if facts that make it possible for the person to acquire that standing in that situation. It is, in one sense, among the enablers of that moral standing without making that person acquire that standing on its own. (This is the sense in which actually being presented with a case of acute vulnerability or risk can arguably be an enabler of someone s moral heroism.) On the second (and stronger) interpretation, being affected by that some fact about a person s situation makes a difference to whether or not that person has some moral standing, but in the sense that it makes it the case that the person acquires that standing in that situation. It is, in one sense, what constitutes that moral standing hence it makes that person acquire that standing on its own. (This the sense in which actually being the non-voluntary beneficiary of some grave injustice can arguably constitute someone as morally compromised by the actions of others.) In order to establish the existence of circumstantial luck strictly speaking, it is only weaker of these interpretations that is required. (And indeed, it is only the weaker interpretation that is required to follow the argument given in the previous paragraphs.) Having said that, it is arguably the stronger interpretation that is required in order to make sense of the connection between the two ways of thinking about moral responsibility that I outlined in the first section of this paper. It is therefore the stronger interpretation that most interests me, and that will be the focus of much of the discussion that follows. I am now in a position to draw a direct connection between the foregoing discussion of circumstantial luck and standard discussions of (normally resultant) luck. For if what has been said above about circumstantial luck is plausible, then it is arguably even more plausible for resultant luck. The kinds of luck that allegedly conflict with the principle in the case of resultant luck stand in an obvious causal relation to ends the agent has actually set herself and 13

her diligence in taking due account of risk and contingent obstacles to their realization (e.g. in the case of a driver who unfortunately runs over a pedestrian). The kinds of luck that allegedly conflict with the principle in the case of circumstantial luck need not bear any interesting relation to ends the agent has previously set herself (although how we may want them to subsequently behave obviously presupposes the prior development of dispositions that make this possible). If the possibility of genuine moral assessment obtains in the latter case (circumstantial luck), it should obtain in the former case (resultant luck); which, from this perspective, would seem to be the easy case, given the prior facts of individual volition resulting in the problematic (and in many cases, previously imagined) consequence. Thus, if we accept the claim that we can be faced with situations that morally call for a response regardless of the ends we have voluntarily set ourselves and regardless of aspects of the situations that are within our control, we should also accept that we can be faced with situations that call for a response when our personal projects turn out in ways that are beyond our control. If so, the principle on which the problem of moral luck depends is not compelling in its standard formulation. There is a kind of moral assessment that transcends the limits of what is beyond our control. The scope of this kind of moral assessment extends to cases where individuals and groups of people are intelligibly describable as morally responsible for responding to the situation in which they find themselves. Any plausible account of the principle on which the problem of moral luck depends should therefore construe it as (at least, but no more than) one relevant factor in moral assessment among others. 14

4. Response 1: dimensions of moral assessment There are several reasonable sources of suspicion about my attempt to connect the two ways of thinking about moral responsibility in the way outlined in the previous section. Here I shall mention three. The most obvious source of suspicion is that the kind of judgement made about someone who responds inappropriately to a situation that calls for a response fails to properly connect with the real problem of moral luck, which is one about whether we can intelligibly hold someone to account for something that is beyond their control, and not a problem of whether we can intelligibly hold someone to account for how they go on to respond to something that is beyond their control - where the latter response is clearly something that could be within their control (insofar as anything ever is). If so, the apparent conflict between the two ideas identified is based on an equivocation. This response quickly removes one potential source of misunderstanding, although not the one at issue here. The present challenge is not about the intelligibility of holding someone morally accountable for how they conduct themselves once they find themselves in a given situation. The challenge is to make sense of the idea that someone can intelligibly be held morally accountable merely in virtue of finding themselves in that situation where the fact that they find themselves in that situation is a matter beyond their control. The answer to this challenge is obviously one that could have serious implications how it makes sense to assess the person for how they go on to conduct themselves in that situation. Yet it is also an answer that it must be possible to arrive at in some sense prior to that assessment. 15

A different (but potentially more controversial) way of putting the challenge is this: suppose I find myself faced with acute vulnerability or risk involving a morally significant other, and I am able to respond to it. What do I owe by way of a response in virtue of that fact? Whether more controversial or not, putting the challenge in this way helps to remove another potential source of misunderstanding, namely the thought that the very appearance that there is a problem here is purely an effect of a misguided insistence on thinking about the case along the wrong normative dimension (c.f. Andre 1983; c.f. Aristotle 1925/1980; Nussbaum 1986). 3 Thus, one source of resistance to the idea of moral luck is the commonly held view that it makes no sense to specifically praise or blame someone for facts of their situation beyond their control. Thus, we can agree that, prior to responding appropriately or not to a situation involving acute vulnerability or risk, the person facing that situation is not a paradigmatic object of blame. (In this sense, we might be said to attribute moral responsibility in the absence of fault ). 4 When we focus our attention on the evaluative dimension of assigning or withdrawing admiration or esteem, the source of resistance to moral luck may be harder to pin down. Thus, I might come to admire a fortuitously successful person, even if I would admire them even more were they to moderate their self-congratulatory impulses by reflecting on the fact that said success has been achieved in an environment of widespread and underserved misery on the part of others - misery on which their own fortune for 3 In his (1995), Williams moderates his initial discussion in Williams (1981) along these lines, when he writes: morality does try to resist luck but not every ethical outlook is equally devoted to doing so. I entirely agree that an Aristotelian emphasis in ethics, for instance, need not run into the same difficulties (Williams 1995, 241). Explicating Williams s distinction between Aristotelian ethics and the morality system would take me too far afield here. It is arguable that her omission to consider either this, or some analogous, distinction subtracts from her otherwise compelling critique of the problem of moral luck as originally posed by Williams and Nagel in Urban Walker 1991. 4 The practice of blaming people can have a variety of functions in virtue of i) having certain consequences (e.g. protection, deterrence, reform, reconciliation); ii) expressing certain attitudes (e.g. disappointment, anger); or iii) just signaling that behavior has fallen below some minimally acceptable threshold (e.g. norms or ideals of mutual recognition). Tracing the implications of these different ways of thinking about blame would take me too far afield here (For further discussion, see e.g. Smith 1983; Adams 1985; Scanlon 2008; Fricker 2014). 16

historical reasons depends. Be that as it may, I doubt if this issue is one that can be simply resolved by attributing it to some form of semantic mischief making, whereby only the most narrowly deontic dimensions of moral assessment are classified as moral (Rosebury 1995, 517; c.f. Margalit 2002). When asking what someone owes in virtue of finding themselves in a certain situation I am describing that situation in recognizably deontic, not in paradigmatically evaluative, terms. I am therefore not obviously best interpreted as asking what it would be good, better, or even best for me to do; as opposed to what it is my duty (in this case a positive duty) to do. 5 5. Response 2: being responsible or taking responsibility? Yet is the moral responsibility that arises from the idea of being faced with a situation that calls for a response really of a relevantly similar kind to that which arises from the idea of an autonomous exercise of agency over which we can be said to have control? And if it is not, why is that? The answer is not at all obvious. To put the point in terms of the distinction drawn earlier, it is one thing to say that finding oneself faced with acute vulnerability or risk that one is well placed to do something about may give rise to some form of liability and that someone who fails to accept responsibility on that account could coherently be subject to 5 One issue arising here, but one that I will not go on to discuss any further, is whether the deontic (self-) description of owing something, or having a duty in these cases, is itself compulsory, or whether it is merely one intelligible way of making moral sense of the situation among others (such as potentially more permissive descriptions in terms of good or admirable character traits). This question (which is itself a substantially moral one) is not one that needs to be definitely answered here. It will more than suffice for my purposes here that the deontic description is one coherent way of making sense of these cases among others, this being all that is required to show that genuinely moral assessments can coherently transcend the domain defined by the control condition on moral responsibility attributions. 17

negative moral assessment for that reason. 6 It is quite another thing to say that finding oneself faced with acute vulnerability or risk that one is well placed to do something about in itself affects one s moral standing in such a way as to render one a suitable object of moral assessment. Perhaps the former is perfectly coherent, but it does not prove the existence of moral luck in any controversial sense. Perhaps the latter would prove the existence of moral luck in a controversial sense, but it is hardly coherent. Insisting on this point might be thought to be necessary in order to resist what might otherwise be a natural temptation, namely to let the contingent circumstances in which someone finds themselves somehow stain our image of them, as when a negative assessment is made of people who are doing quite well, but who happen to be located near terrible poverty ( such cruel people ), even if they are not doing any better than those who judge them, and are not obviously better placed to effectively address the poverty in question than are those who judge them negatively in this way. It can be hard (but no less important for that reason) to distinguish this kind of illusory stain (or taint ) from the genuine moral stain (or taint ) that can be acquired through active or passive complicity (e.g. when does benefiting from advantages initially received non-voluntarily turn into an exercise of opportunism? (see e.g. Klosko 2005; Carrier 2014)) The focus of moral assessment in cases of moral liability of this kind is arguably located causally downstream, as when someone callously refuses to respond to a situation that calls for a response, or when someone responds to their situation in a morally problematic way (e.g. by responding to it grudgingly, or by conceptualising it in a tendentious manner). 6 The legal notion of strict liability does not, I think, have the same moral connotations as talk of moral responsibility in the sense in which I am interested here. Mapping the relationship between different conceptions of moral and legal liability would clearly be a useful exercise, although it is not one I propose to pursue here. For further discussion of the notion of liability as it features in cases of (primarily resultant) luck, see e.g. Thomson 1986. 18

In a recent paper, David Enoch has proposed a diagnosis of the problem of moral luck that promises to deliver this result; a diagnosis that delivers two subtly different ways of thinking about moral responsibility. On Enoch s analysis, the crucial distinction is one between being responsible for something and taking responsibility for it (Enoch 2011). 7 Being responsible is essentially backward looking, and involves acknowledging responsibility that is already there (such as having taken a bribe). It includes cases close to what Enoch calls the core of our agency, possibly including (but clearly not only) the kinds of cases that are sometimes affected by consequential luck. Taking responsibility is essentially forwardlooking, and involves making it the case that one is responsible (such as being a beneficiary of an institutional culture in which bribes are widely offered and taken, and then responding appropriately to this fact). Enoch argues, quite plausibly, that in some cases we can have a moral duty to take responsibility where no prior responsibility exists. These are cases involving what he calls penumbral agency ; cases where the relevant actions and events are not within the scope of our core agency, but where they are nevertheless not too far removed from it to e reasonable to expect that we take personal responsibility for them. Enoch s examples of cases where taking responsibility may be called for include the actions of one s country and the actions of one s young children. Cases where it is not called for include the unjust actions of past generations or distant others (such as past ownership of other persons either captured or raised for that purpose). 8 According to Enoch, the distinction 7 It might also be natural to talk of accepting responsibility in some of cases at issue in the following paragraphs. To avoid confusion, I interpret accepting responsibility as presupposing the prior existence of responsibility and therefore being responsible in Enoch s sense. 8 One case that Enoch does not discuss involves a person having a voluntary association (with associated benefits) with a particular institution (such as a University) the continued existence of which is contingent on its past involvement (e.g. by way of endowment) in what is currently deemed to be serious injustices (such as slavery). The question for Enoch s account is whether, and if so why, their relationship to the relevant injustices should be counted as falling within that person s penumbral or impure (see Urban Walker 1991) agency. I am grateful to Alice Crary and Alison MacIntyre for raising this question, which may or may not 19

between being responsible and taking responsibility allows us to say that although there is no such thing as moral luck strictly speaking (as in being responsible for the uncontrolled consequences of our actions), we can still make sense of a duty to take responsibility for those consequences when they occur, provided that their being the consequences of our action puts them within the domain of our penumbral agency. In this way, Enoch plausibly argues that we can make sense of the idea that there is a distinctive form of agent regret for the harms we cause ourselves that is often appropriate even in cases where the harm we regret was outside our control (c.f. Nagel 1979; Williams 1981). (I shall return to this issue briefly in the final section below.) It is natural to think that Enoch s distinction between being responsible and taking responsibility is applicable also to cases of circumstantial luck, and that the examples of being presented with acute vulnerability or risk that I have described in the previous sections of this paper are best understood as cases where someone has a duty to take responsibility; not as cases where someone strictly speaking is responsible. If so, the two ideas I started with can be treated separately, and the principles they embody can be shown to be mutually consistent in a way that would arguably resolve the problem of moral luck as posed at the outset. I think Enoch s distinction between being responsible and taking responsible is useful and illuminating. I also think it serves to pick out one morally significant difference among others, in these and similar cases. (Enoch s own example of duties acquired by promising is a good case in point.) I am less confident about the notion of penumbral agency that Enoch invokes to make sense of this distinction, and hence about his success in explaining away apparent cases of moral (including circumstantial) luck on these terms. I have two be a problem for Enoch s account, but which helps to bring out what is at stake in the insistence on constraining attributions of moral responsibility either by some control condition, or by other facts about volition or agency. 20

reservations in particular about Enoch s account. First, although there is clearly a difference between already being responsible for something on the one hand and becoming responsible by taking responsibility for it on the other (Enoch s own example of making a promise and then keeping it once again provides a helpful illustration), I am doubtful whether interpreting this distinction in terms of a voluntaristic notion such as penumbral agency is able to fully capture what goes on in all cases where one may coherently say that some particular individual has a duty to take responsibility. There is a plurality of possible cases here, of which I shall mention three. 9 In the first case, someone within a sub-class of people ought to take responsibility, and it makes a crucial difference exactly who does so (e.g. it would be more fitting in the circumstances to be told the bad news by the person who actually caused the mess in the first place). In the second case, someone within a sub-class of people ought to take responsibility, but it makes no difference exactly who within that class actually does so (although it could make a difference how, and by whom, the responsibility is distributed, as in a case of harm caused by the activities of a large organization). In the third case, someone ought to take responsibility, but it makes no difference at all who actually does so (as in the case where someone is faced with the presence of acute need or vulnerability of some morally significant, but otherwise arbitrary, other). There are obvious moral differences between these three cases, at least the third of which allows that individuals can legitimately be counted as having duties to take responsibility for events and states of affairs that involve members of a moral community with which they may never previously have interacted (and to which they have certainly never applied for admission). The issue is whether what makes 9 There is a historical connection between these cases and Kant s distinction between perfect and imperfect duties, where imperfect duties are said to admit of exceptions and/or discretion in who should fulfill them, or how. (Indeed, a standard example of imperfect duties is the duty to help vulnerable others.) Further pursuit of this connection would take me too far afield (see Kant 1981). 21

the moral difference between these cases is most helpfully explained in terms of some voluntaristic notion such as penumbral agency. My view is that it is not. In at least in some of these cases where a situation can be said to call for a response, there is no interesting sense in which the events in question can be said to fall within a domain of penumbral (or any other form of) agency with respect to each and every person who could reasonably consider themselves either to be responsible, or to otherwise have a duty to take responsibility in that situation. The facts they are confronted with could include the acute vulnerability or risk to a complete stranger, such as an injured passer-by (c.f. Lillehammer 2014). It is primarily for this reason that I am doubtful whether the domains of penumbral agency and duties to take responsibility overlap in the way required for Enoch s account to speak to all the cases I am interested in here (which I dogmatically assume that it should). 10 Enoch does briefly mention the kind of cases I have in mind when he writes that it is important to distinguish between (purported) cases of moral luck on one side, and cases of plain luck that has moral implications, or morally significant plain luck on the other. (Enoch 2009, 131). With respect to the latter, he writes that whether you are morally required to give to famine relief depends on whether you find yourself in a situation in which people suffer from hunger (Enoch 2009, 130). The existence of this kind of morally significant plain luck is said to be a trivial consequence of the fact that moral judgement should be sensitive to context. In a footnote, Enoch points out that cases like these can in principle be accommodated on his account if the category of penumbral agency is understood so as to include the entire human (or otherwise moral) community (Enoch 2009, 131). I suspect the 10 For an earlier treatment that I take to be in many ways sympathetic to my own, see Urban Walker 1991, who writes that the match between choice and action on the one hand, and accountability and desert on the other is mediated by complex social understandings which these agents are expected to appreciate and to share (Urban Walker 1991, 22). 22

fact that this is worth pointing this out suggests that how we should understand such cases is not so trivial. To put it in Enoch s terms, it might be said to be desirable (in some cases indispensable) to engineer people (biologically, educationally, politically) to take responsibility in these kinds of cases, partly on the grounds that it is morally valuable that people take moral responsibility for the acute vulnerability and risk faced by morally significant others, whoever they may be. To do so might be the expression of a highly valuable (even if supererogatory) sense of solidarity or humanism. For similar reasons it might be valuable that some people (e.g. future generations or descendants) atone for, or just remember, the wrongs done by others (e.g. previous generations or ancestors). 11 Either way, the distinction between what does and what does not fall within the boundaries of someone s penumbral agency is not obviously what we need in order to make sense of cases where someone finds themselves thrown into a situation of morally critical choice, regardless of the usefulness of drawing a moral distinction between being responsible and taking responsibility for morally significant events or states of affairs, and the classification of the person thrown as falling into the second of these two categories. 6. Response 3: moral predation, individual freedom and the demandingness problem Responding to circumstantial luck involves moral risk. Thus, it might be objected that rejecting the principle on the grounds that a moral response can be called for merely in virtue of the fact that we find ourselves in a situation calling for a response leaves our moral 11 The obvious flipside of this is the intelligibility of adopting a questioning attitude towards the behaviour of someone who actively cultivates ignorance of acute vulnerability or risks faced by others, thereby making it the case that they never even consider themselves as called upon to make any kind of response. There is much more to be said here than I can begin to do justice to here. For further discussion, see e.g. Lillehammer 2014. 23