1. The Problem of Present Luck

The debate over whether and how luck undermines freedom often founders on conflicting intuitions with regard to whether luck is at work in a particular scenario. To avoid this kind of impasse, we do better to utilize an account of luck. Building on the important work of Duncan Pritchard^{Footnote 4} and E.J. Coffman,^{Footnote 5} I have put forward such an account.^{Footnote 6} According to my account, there are in fact two kinds of luck. An event is subject to present luck when it is significant for an agent, that agent lacks direct control over its occurrence and it is chancy. An event is chancy when it occurs in the actual world but fails to occur in a high enough proportion of nearby possible worlds that differ from the actual world in at most trivial respects (how high a proportion is high enough depends on its significance: the lower the significance of the event for the agent, the higher the threshold). If determinism is false, then events may be chancy across possible worlds that are, up until the instant of the event, identical. If determinism is true, events may nevertheless be chancy, and therefore lucky. I don’t think we need to wait on the abstruse musing of physicists to discover whether someone who was hit by lightning (in some area of the world in which lighting strikes are infrequent) was a victim of bad luck. He is the victim of bad luck because had the world been different in trivial respects (had he been standing there and not here—assuming that he had no special reason to be here—or had he walked out of his front door a scant few seconds later—again, assuming that he might easily have done so) he would not have been hit. Similarly, we don’t need to know whether determinism is true to answer the question of whether Megan is lucky to win the lottery: had the balls rotated for microseconds longer, or if they had microscopic differences in curvature or roughness, different numbers would have been drawn.

It is easy to show that the putatively directly free decisions of agents are subject to responsibility-threatening luck if standard event-causal accounts of libertarianism are true (it is harder to show that the same holds true with regards to agent-causal libertarianism, though, like Haji himself,^{Footnote 7} I believe that this is the case).^{Footnote 8} When such agents make up their minds, and their so deciding is directly free, the decision is obviously significant for them (or rather, we only care about those that are significant for them). It takes very little work to show that in addition the agent lacks direct control over it and that it is chancy. It is chancy because (again, on standard event-causal libertarian accounts) when actions are directly free there is a significant chance that the agent makes a different and conflicting decision to the one she actually makes. She lacks direct control over it because any events, states or processes that might constitute her exercising control are identical across possible worlds in which she decides one way or another. If luck is a threat to responsibility-level freedom, then libertarians of this stripe have a major problem on their hands.

The account of luck offered above is neutral with regard to causal determinism. That raises the possibility that the problem of chancy luck might be a threat to compatibilist accounts of free will as well. Pérez de Calleja and I have each independently argued that in fact compatibilism does face such a problem. Let’s assume that the significance condition with regard to a decision made by an agent who satisfies standard compatibilist accounts of free will is satisfied. It takes a little more work to show that the other two conditions are satisfied. The two conditions are tightly linked: agents lack direct control over their decisions when how they decide is subject to the chancy occurrence of events, such that if an event occurs they will decide one way and if it fails to occur they will decide another. If the agent lacks control over whether such an event occurs (either because its occurrence is not sensitive to her actions or because she fails to satisfy the epistemic conditions of control with regard to its occurrence), and it makes a difference to how she decides, she lacks direct control over her decision. If its occurrence is chancy (that is, it occurs in the actual world but fails to occur in a high enough proportion of possible worlds that differ in at most trivial respects from the actual world), then her decision is lucky. This is present luck: luck at or near the time of decision. So compatibilists, too, face a problem of present luck.

2. The Threat from Luck

From what I can tell, Haji does not deny that cases like the one described occur. If I understand him correctly, his first objection to what I have said so far is that it is too broad; my description captures both cases of luck in which luck is a (prima facie) threat to free will and those in which it is no threat at all. Consider a case in which a certain consideration comes, by chance, to an agent’s mind and because that condition comes to mind the agent makes a different and conflicting decision to the one she would have made had it failed to come to mind.^{Footnote 9} In such a case, Haji suggests, neither compatibilists nor libertarians should think that the luck involved threatens the agent’s moral responsibility. The reason is this: though it is true that what considerations come to the agent’s mind is lucky, “both libertarians and compatibilists should agree that it may well be within an agent’s control to attend to them.”^{Footnote 10} Mele^{Footnote 11} makes a closely related point, in defence of his own version of libertarianism. On Mele’s account, it is undetermined what considerations come to mind during deliberation; hence conflicting decisions are metaphysically possible for the agent. But since indeterminism affects the inputs into deliberation, leaving the agent free to assess the inputs, luck in the causal pathway is compatible with responsibility-level control.

I have responded to Mele elsewhere.^{Footnote 12} But my response targets Mele’s claim that luck affects only the inputs into deliberation, and therefore isn’t responsibility-undermining. In reply, I pointed out that this luck may be decisive in the sense that, had the lucky event of a consideration coming to mind failed to occur (as it easily might), the agent would have made a different decision. Haji, however, accepts that the luck may be decisive but insists that, because the agent exercises control in deliberation, the luck may be decisive and yet fail to be responsibility-undermining. He therefore suggests that only when the lucky event bypasses the agent’s capacities of deliberative control is there a (prima facie) problem of present luck for the compatibilist.

I am not convinced. The problem with Haji’s claim is best brought by considering a case in which present luck in the considerations which occurred to an agent have caused her to perform a morally bad action. Haji would have us blame her. But she seems to have a defence available. She can point out the following conditional is true: had consideration c not occurred to me—a chancy event over which I failed to exercise direct control—I would not have performed the morally bad action. It is, I think, reasonable to see the truth of this conditional as excusing: her counterpart in a nearby possible world is not blameworthy—may, indeed, be praiseworthy—and yet the difference between her and counterpart is not due to her, her mental states, character, will, and so on, but rather due to luck.

Reasonable people may disagree about whether the excuse offered succeeds. While, again, I am not convinced, for the rest of this paper I will assume that the excuse fails. For the sake of argument, at least, I will accept Haji’s claim: there is a prima facie problem of present luck for compatibilists only when present luck is decisive and its influence bypasses agents’ capacities of deliberative control. Will that ensure that the problem of present luck is immediately greatly reduced? I don’t think so: rather, it will have its force undiminished. To show that, however, it is necessary to spend some time showing why the problem of luck is so threatening for the compatibilist. It is time to restore the missing half of the luck problem: the problem of constitutive luck.

It was no part of my brief to argue that the problem of present luck, the problem that is widely held to beset standard versions of event-causal libertarianism, is as severe for compatibilists as it is for libertarians. Rather, I pointed out that the problem is smaller for the compatibilist than for the libertarian because, whereas the latter claims that no action is directly free unless it is metaphysically open to the agent to perform more than one incompatible action, the compatibilist does not require any equivalent openness for freedom.^{Footnote 13} It is therefore compatible with her view that agents are rarely subject to present luck. Perhaps the typical agent—or perhaps a lucky few—get through life rarely encountering situations in which present luck is decisive for the decision-making (such agents may be subject to present luck, but it is not decisive when it fails to make a difference to the kind or valence of the action which it brings about; an agent doesn’t have any sort of excuse, typically, at any rate, if she stole money at 10.01 pm rather than 10.00 pm). Nevertheless, I claimed, luck entails the falsity of all compatibilist accounts.^{Footnote 14} Though I can’t attempt to demonstrate that here, I will try to show how luck entails the falsity of the members of a particularly popular family of compatibilist views: those that accept a historical condition on moral responsibility.

According to what I shall call history-sensitive compatibilists, responsibility is a partially historical concept, like sunburn. Nothing counts as sunburn unless it has the right kind of history; similarly, no one is morally responsible unless she has the right kind of history.^{Footnote 15} History-sensitivity is typically motivated by reflection on certain kinds of manipulation cases: would we blame a formerly upright person who has her brain ‘rewired’ without her knowledge, so that she possesses desires and beliefs that cause her immediately to go out and commit a horrendous act? I have argued that these thought experiments are best conceived of as turning on what Thomas Nagel^{Footnote 16} called constitutive luck;^{Footnote 17} luck in the traits, dispositions, values and commitments (states and traits, for short) with which we find ourselves. The intuition that the wrong kind of history undermines moral responsibility is the intuition that constitutive luck may undermine responsibility.

Constitutive luck is clearly a species of the same genus as present luck. Constitutive luck is clearly significant for agents and, just as clearly, agents lack direct control over it. It is, however, controversial that it is chancy: some philosophers have denied that there is any such thing as constitutive luck, on the grounds, roughly, that there are no nearby possible worlds in which agents lack the traits constitutive of them.^{Footnote 18} I think that this claim is false: I don’t think we have any of our traits essentially. Nevertheless, I think that the chanciness condition needs to be modified to generate an account of constitutive luck: a trait or disposition is constitutively lucky for an agent when it is significant for her, she lacks direct control over it and the trait in question in relatively unusual in the relevant reference group. Thus, finding that one has a great talent for basketball is constitutively lucky for an agent, as is being born with a disability, but having two legs is not constitutively lucky for her, since having two legs is not unusual in her reference group. What is the relevant reference group? That is fixed by conversational context, but for most purposes we can use those members of her social group who are roughly contemporaneous with her as a reference group and yield the correct results.

The manipulated agents who feature in the thought experiments that motivate history-sensitivity are victims of good or bad constitutive luck. They suddenly find themselves with traits or dispositions over which they lack direct control, which are significant for them and which are relatively unusual in their reference group.^{Footnote 19} History-sensitive compatibilists see these kinds of manipulations as threatening to moral responsibility, but they also see themselves as having a solution to the problem they pose. Though manipulated agents are not responsible for actions that they perform immediately after the manipulation, they come to be responsible for their actions. They take responsibility for their dispositions and traits. Ordinary life offers them opportunities to learn about their dispositions, reflect on them and modify them in the light of reflection. The identical response serves as well as an answer to the problem of constitutive luck (as well it should, since, again, they are the same problem): though we are endowed, through genetics and formative environment, with varying states and traits, agents typically come to take responsibility for these states and traits as part and parcel of the normal process of maturation.

We now have the resources to see why the problem of luck is so significant for history-sensitive compatibilists. Haji wonders if Pérez de Calleja and I worry that the bypassing of agential capacities for deliberative control is much more common than philosophers have thought. Actually, I have no view on how frequently bypassing factors make a decisive difference to how agents decide (where a bypassing factor makes a decisive difference if it is true that, in its absence, the agent would have made a conflicting decision). Rather, my claim is that every decision made by the agent is infected by responsibility-undermining luck: either the decision is infected by lucky bypassing or it is infected by constitutive luck.

Whenever a decision is easy for an agent, its easiness is explained by the states and traits constitutive of her. It is a ‘no brainer’ for her, given what she desires, given what she believes, given what she values. When it is not easy for her, this fact, too, is explained by reference to the traits constitutive of her: they are silent on the matter, or they pull in opposing ways. It is these kinds of cases, I claim, in which the agent is vulnerable to responsibility-undermining present luck via bypassing. Bypassing-style factors—chance influences on our cognition that go unrecognized by us and which either we cannot deliberate about or which have causal effects on cognition regarding which we are ignorant (and about which we cannot deliberate)—are absolutely pervasive. We are prone to a range of heuristics and biases, to the kind of priming studied by social psychology, to implicit biases, and so on.^{Footnote 20} Often, though (again, I take no position on how often), these influences make no appreciable difference to how we decide, because they are overwhelmed by our deliberation. The fact that I am primed in a way that makes me somewhat more favourable to stealing, say, will almost never actually result in my stealing, because too many of the states constitutive of me pull against stealing (I value honesty, I fear the shame of exposure much more than I desire the extra money, especially given that I am relatively comfortably off, and so on). Much the same holds for all of us with regard to a great many of our decisions: if bypassing factors influence what we decide, then the influence is not decisive (it affects the way in which we do something, the time at which we do it, and so on; not whether we do it).

My claim is that bypassing factors are not decisive when, and only when, the states constitutive of us insulate us against them. When matters are more evenly poised, from the deliberative perspective, we are vulnerable to these factors and present luck tips the balance.^{Footnote 21} When matters are not evenly poised, we are insulated against these factors, but then, I claim, constitutive luck explains our decision.

This latter claim needs some defence. I have argued that, whenever we are insulated against present luck, this fact is explained by the states and traits constitutive of the agent. But, obviously, the states and traits constitutive of the agent need not be the states and traits that she has as the result of luck in her genes and formative environment. That fact is central to the history-sensitive compatibilists’ response to the problem of constitutive luck. They claim that agents may take responsibility for their endowment: if they are not satisfied with the relevant traits, they may alter them, and when they do not alter them they may nevertheless endorse them. These claims—that agents may deliberately (or otherwise) alter their traits and that they may endorse them without altering them—are both obviously true. But the truth of these claims doesn’t offer a solution to the problem of luck. The very actions whereby agents are supposed to take responsibility for their endowment of constitutive luck are themselves either vulnerable to present luck (when their initial endowment doesn’t insulate them from it) or they are insulated from it by that endowment (or by states constitutive of the agent which have been produced by the interaction of present luck and the initial endowment). Though agents may alter themselves, sometimes quite dramatically, they don’t exercise control over how they alter themselves: rather, the actions which alter their states, or which endorse them, are the product of present luck and the initial endowment as it has been modulated by present luck.

This being the case, the frequency of decisions produced by deliberation bypassing present luck doesn’t matter. While our responsibility-level freedom is threatened by each such episode, we are no freer in the absence of such luck, because it is our endowment of constitutive luck, directly or as modified by present luck, that explains our resistance to present luck. Whether Haji is correct in thinking that luck in the considerations that come to mind doesn’t undermine our moral responsibility by posing a problem of present luck, because we remain able to assess the considerations (say) that luckily come to mind is also irrelevant to whether we possess responsibility-level freedom: if in deliberating we are not subject to the problem of present luck, our resistance to that problem is explained by our constitutive luck. Between them, bypassing present luck and constitutive present luck play a decisive role in all our decisions. I call this the luck pincer. On their own, the threat from present luck and the threat from constitutive luck might each be addressed. But we do not face them alone, and our responses to each are vitiated by the other.^{Footnote 22}

3. Manipulation and Luck: Is the Threat Novel?

Haji claims that the threat that luck poses to compatibilism is not novel; rather, it is the very same threat that is at issue in manipulation cases. Furthermore, he suggests, compatibilists (and libertarians) have the resources to address the threat. It is neither new, nor especially threatening. In setting out the nature of the threat from luck, I have had occasion to utilize the comparison with manipulation cases myself. That suggests that Haji might be on to something. Is the threat from luck just the threat from manipulation, and, if so, might the available responses to the latter suffice to see it off?

The threat from manipulation cases is the threat from constitutive luck. But the threat on which Haji focuses in his paper is not the threat from constitutive luck; it is present luck that is his focus (it is because he ignores constitutive luck that he drastically underestimates how serious the threat to history-sensitive compatibilism is). So, while Haji is right that there is a threat from luck that is identical to the threat from manipulation, he is wrong in thinking that the threat he identifies is the same as the threat from manipulation.^{Footnote 23}

To be sure, the threat from present luck could be presented in the form of a thought experiment turning on manipulation, but the thought experiment would present a decidedly non-standard manipulation case. We’ve already seen why: standard manipulation cases bring about states and traits that are significant for the agent and over which she lacks direct control, but the production of these states and traits need not be chancy. They are cases of constitutive luck, because the states themselves vary across relevant reference groups, but not of present luck. With that in mind, it is easy to conjure up a manipulation case that brings in present luck. We can modify the well-known Ann/Beth thought experiment^{Footnote 24} to make it turn on present luck. In the original case, their dean enlists new-wave brainwashing techniques to modify easy-going Beth to motivate her to work as hard as dedicated Ann. Beth is subject to a highly unusual form of constitutive luck, but not to present luck because the manipulation is not chancy: there is no reason to think that there is some nearby possible world that, up until the time of the manipulation, differed in only trivial respects from the actual world and in which Beth is not manipulated. There are, of course, possible worlds in which Beth is not manipulated, but they are worlds that differ markedly from the actual world: worlds in which Beth is already highly productive, in which the dean is less unscrupulous, or in which he suddenly dies, and so forth. We introduce present luck by stipulating details that make it the case that such a world is nearby. Here’s one such case: Beth’s transformation is the product of a lightning strike (which might easily not have occurred or which she might easily have avoided). Here’s another: Beth’s dean decides to conduct the manipulation only if he draws a face card from an ordinary deck at his first try. Now it is the case that Beth is subject to constitutive luck as well as to present luck, because her having particular states or traits is chancy.

So Haji is wrong to think that the problem of present luck is the problem that typical cases of manipulation pose. This is so even if he is right in thinking that present luck can undermine responsibility only when it proceeds by bypassing, and in claiming that standard manipulation cases also undermine responsibility via bypassing. It is, nevertheless, worth pointing out that, though compatibilist responses to constitutive luck cannot generalize to present luck (how could someone take responsibility for their present luck? Taking responsibility, in the manner required, takes time; it cannot be done instantaneously), Haji is correct in thinking that compatibilists can offer a unified explanation of why manipulation cases and present luck undermine moral responsibility: by way of bypassing agents’ capacities for deliberative control.

By offering this unified explanation of why present luck undermines moral responsibility, Haji may respond to Pérez de Calleja by conceding that present luck threatens moral responsibility, but noting that the kind of bypassing required for the threat is surely not ubiquitous. At worst, he may contend, Pérez de Calleja shows only that we are freer less often than we think. But whatever the merits of this reply, it will not see off the much more serious threat from the luck pincer. The response does not begin to address the combined responsibility-undermining effects of constitutive and present luck. Even if Haji can diagnose the threat posed by present luck, and give reasons why it ought not to trouble us overmuch, he does not have a solution to the bigger threat and therefore to what he calls “the problem of compatibilist luck.”^{Footnote 25}

With this result in hand, I am now in a position to bring out why the threat from luck is a novel one, distinct from the threat that manipulation poses to moral responsibility, even if Haji is right in claiming that present luck threatens moral responsibility only when it bypasses agents’ capacities for deliberative control and that constitutive luck (in the form of manipulation cases) threatens moral responsibility by bypassing the same capacities. It doesn’t follow from the fact that present luck threatens moral responsibility via bypassing and constitutive luck threatens moral responsibility via bypassing that the combined threat undermines moral responsibility via bypassing. The luck pincer threatens moral responsibility not via bypassing, but due to the fact that the only plausible solution to the problem of constitutive luck, agents’ taking responsibility for their initial endowment from their genes and formative environments, is vitiated by luck (since the acts whereby agents would take such responsibility themselves either express their constitutive luck, or are subject to present luck). I think this is better seen as a threat from luck, and not a threat from bypassing.

Because he focuses on the work of Pérez de Calleja, Haji overlooks the real threat from luck to compatibilism. Pérez de Calleja gives us only half of the challenge, and by itself it is a challenge that might be overcome. Moreover, he mischaracterizes the threat on which her work concentrates, assimilating it to the threat that is at work in manipulation cases. Standard manipulation cases do indeed turn on a threat from luck, but not the threat with which Pérez de Calleja is concerned. The threat from luck is indeed novel. It is quite different from the threats to which philosophers have hitherto responded. And it is, I think, much more intractable.