Moral Responsibility, Luck, and Compatibilism

Abstract

In this paper, I defend a version of compatibilism (about determinism and moral responsibility) against luck-related objections. After introducing the types of luck that some take to be problematic for moral responsibility, I consider and respond to two recent attempts to show that compatibilism faces the same problem of luck that libertarianism faces—present (or cross-world) luck. I then consider a different type of luck—constitutive luck—and provide a new solution to this problem. One upshot of the present discussion is a reason to prefer a history-sensitive compatibilist account over a purely nonhistorical structuralist account.

1 Introduction

Because of its requirement that there be indeterminacy in the world in order for there to be free actions and actions for which agents are morally responsible, libertarianism appears to be subject to luck-related worries. The very indeterminacy that libertarianism requires can apparently mitigate an agent’s control over what she does. For example, some libertarians have argued that, in the case of non-derivatively free actions (or actions for which agents are non-derivatively morally responsible), it must be undetermined at the time of action what the agent will do; but if it is undetermined what the agent will do, then the agent does not determine (or ensure, or settle) what she does, and so her control over what she does is called into question. This is sometimes called the problem of present (or cross-world) luck, since the indeterminacy required by (some versions of) libertarianism introduces problematic luck for the agent at the time of her action, and what she does at that time varies across possible worlds that have the same laws of nature and that have the same past up to the time of the action.

Because compatibilism does not require indeterminacy in order for agents to be morally responsible, it is typically assumed that compatibilism, unlike libertarianism, is not challenged by the problem of present luck.¹ Now, as it turns out, most contemporary compatibilists take moral responsibility to be compatible with both determinism and indeterminism, so, in a sense, these compatibilists inherit the libertarian’s problem.² I will not focus on this challenge to contemporary compatibilists here.³ There is another problem of luck, however, sometimes called the problem of constitutive luck, which clearly threatens compatibilism.⁴ Constitutive luck concerns the aspects of agents that make them who they are, such as their traits and dispositions, and the problem of constitutive luck says that it is a matter of luck for agents that they have the traits, dispositions, etc. that they in fact have. And if it is a matter of luck for an agent that she is who she is, so the problem goes, then she does not possess sufficient control over who she is to be responsible for the kind of person she is and for the actions that stem from her character.

Recently, Levy (2009, 2011) and Pérez de Calleja (2014) have argued that compatibilism is not only vulnerable to the problem of constitutive luck but also to a problem of present (or cross-world) luck. These approaches vary in the details, but both maintain that, even in a deterministic world, agents are subject to luck at or around the time of their actions. An agent in a deterministic world may perform some action in that world but behave differently in a nearby possible world as a result of circumstances over which she lacks control.⁵ The first aim of this paper is to show that each of these approaches (to arguing that compatibilism is challenged by a problem of present luck) does not succeed.

Even if compatibilism is not challenged by a problem of present luck, the problem of constitutive luck remains, so the second aim of this paper is to develop a model for dealing with the problem of constitutive luck. The model I propose suggests that moral responsibility comes in degrees and that the degree to which one is morally responsible is a function of the degree to which one meets the various conditions on moral responsibility. One upshot of my discussion of luck and compatibilism is that we find a reason to prefer a history-sensitive account (whether it be a historicist or structuralist one) over the traditional, purely nonhistorical structuralist account. More on this to come.

In the next section, I discuss relevant aspects of the concept of luck, including the distinction between present (or cross-world) and constitutive luck. I then, in Sect. 3, consider and reply to some recent attempts to show that compatibilism faces the problem of present luck. In Sect. 4, I provide a response to the problem of constitutive luck for compatibilism, and I claim, in Sect. 5, that one upshot of the present discussion is a reason to prefer a history-sensitive compatibilist account over a purely nonhistorical structuralist account.

2 What’s Luck Got to Do With It?

The concept of luck has featured in many arguments in debates about free will and moral responsibility.⁶ In most cases, such arguments employ an intuitive understanding of what luck is, and an analysis of the concept of luck is not provided.⁷ That said, an important aspect of the concept of luck is often highlighted, namely that a lucky agent lacks some kind or degree of control over what occurs. Alfred Mele nicely articulates this point: “Agents’ control is the yardstick by which the bearing of luck on their freedom and moral responsibility is measured. When luck (good or bad) is problematic, that is because it seems significantly to impede agents’ control over themselves or to highlight important gaps or shortcomings in such control” (2006: 7). Christopher Evan Franklin extends this point: “Moreover, there seems to be an inverse relation between luck and control: the more an action is subject to luck, the less it is under our control, and the more it is under our control, the less it is subject to luck. Luck and control thus appear to exclude each other: an action cannot be both wholly a matter of luck and wholly under our control” (2011: 200). Thus, in the case of an agent’s action, if the action is lucky for the agent in a way that is problematic for her moral responsibility for that action (and here I follow Mele in using ‘lucky’ to refer to both good and bad states of affairs), then she lacked some type or degree of control over her action.⁸

This feature of the concept of luck (that is, lack-of-control) is arguably the most important feature of the concept, since it is precisely this lack of control that is troubling for various accounts of moral responsibility. As far back as Aristotle (Nicomachean Ethics 1109b30–1111b5), it has been widely accepted that in order for an agent to be morally responsible for some action she must control her action in some sense.⁹ But if there is an inverse relation between luck and control, as Franklin suggests, and if one of the conditions for morally responsible action is that the agent exercised control over that action, then luck presents a problem for morally responsible action.

Control is arguably not the only relevant aspect of the concept of luck, however. In order for some event or state of affairs to count as lucky for an agent, it must also be significant for her. I currently lack control over whether there is heavy traffic in Los Angeles, but, since I am not driving anywhere near Los Angeles, this state of affairs is insignificant for me; it would be strange to say that it is lucky for me (good or bad) that there is heavy traffic in Los Angeles, since this is not even remotely significant for me.¹⁰ As I understand the notion of significance that is relevant here, there will be an objective fact of the matter about what counts as significant for an agent (and the agent might be wrong about what is significant for her), but such facts depend on the agent’s aims, interests, character, etc. So it would be possible for a bird’s flying overhead to be significant for one agent and not another if, for example, the two agents are alike in every way except that one enjoys bird-watching and the other does not. Significance, then, is agent-dependent. This aspect of the concept of luck also features in the intuitive notion employed in the arguments mentioned above, though not emphasized to the degree that the lack-of-control feature is, perhaps because most (if not all) candidate morally responsible actions will be significant for agents in virtue of their being morally significant.

Unlike most others, Neil Levy attempts to provide an analysis of the concept of luck. Drawing from the work of such epistemologists as Pritchard (2005) and Coffman (2007), Levy incorporates both the lack-of-control feature and the significance feature into his analysis, as well as a third, modal feature.¹¹ According to Levy, for an event or state of affairs to be lucky for an agent (in particular, “chancy lucky” for an agent, which just is the kind of luck at issue in the present discussion), “that event or state of affairs [must fail] to occur in many nearby worlds; the proportion of nearby worlds that is large enough for the event to be chancy lucky is inverse to the significance of the event for the agent” (2011: 36).¹² If I am driving in Los Angeles and there is light traffic, this counts as lucky for me since the following three conditions are met: (1) I lack control over whether there is light traffic in Los Angeles, (2) the amount of traffic is significant for me (since I am driving in Los Angeles), and (3) that there is light traffic in Los Angeles fails to occur in a large proportion of nearby worlds. For the purposes of this paper, I will adopt Levy’s modal analysis of luck.¹³

To see how luck might be problematic for moral responsibility, and as setup for the following sections, let us consider two types of luck: present luck and constitutive luck. Present luck, on the one hand, is found at or around the time of action and, according to Levy, “significantly influences that action” (2011: 90). This type of luck, which is typically taken to be uniquely problematic for libertarian accounts of free will and moral responsibility, is also sometimes called cross-world luck, and an example illustrates the reason for this.¹⁴ Imagine that the actual world is indeterministic and that an agent, John, performs an action A at time t. According to some libertarians, in order for A to have been an action for which John is directly (non-derivatively) morally responsible, it must have been the case that, holding fixed the laws of nature and the past up to t, John could have done otherwise than A at t. (And, it is worth noting, on standard libertarian accounts, without directly morally responsible actions there cannot be indirectly moral responsible actions.) In other words, there is a possible world (call it W*) that is exactly like the actual world (call it W) right up to t and in which John does otherwise than A at t. Given that there is no difference between these two worlds right up to t, and in particular that there is no difference in John right up to t, nothing about John (including his powers, abilities, character, motives, etc.) accounts for the difference between his A-ing in W and his doing otherwise in W*. Since nothing about John accounts for this cross-world difference, so the problem goes, John lacks control over whether he does A or otherwise at t. So, if whether John does A or otherwise at t is significant for John, and if John does otherwise in a large proportion of nearby worlds, then John’s A-ing at t is lucky for him.

Constitutive luck, on the other hand, concerns the aspects of agents that make them who they are, such as their traits and dispositions. Saul Smilansky argues that, if determinism is true, we are constitutively lucky in a responsibility-undermining way; if determinism is true, “people cannot ultimately create themselves, and their choices, including their choices to change themselves, and anything they do, can only follow from factors ultimately beyond their control” (2000: 284). The way that people are is a matter of luck, since people cannot ultimately create themselves. And if it is a matter of luck for an agent that she is who she is, so the problem goes, then she does not possess sufficient control over who she is to be responsible for the kind of person she is. This is a challenge for compatibilist accounts of morally responsible action, since compatibilists say that moral responsibility is compatible with determinism. Before addressing this challenge, however, let us consider the more recent attempts to show that compatibilism faces a problem of present luck.

3 Present (or Cross-World) Luck

Present luck occurs at or around the time of an action and significantly influences that action, and an action is lucky if and only if it is significant for the agent, the agent lacks some kind or degree of control over the action, and the action fails to occur in a large proportion of nearby worlds. Recently, both Neil Levy and Mirja Pérez de Calleja have argued that present (or cross-world) luck poses a challenge to compatibilism as well as to libertarianism. Levy argues that luck undermines all accounts of moral responsibility, and essential to his argument is that compatibilism cannot avoid problems associated with constitutive luck without encountering problems associated with present luck. Pérez de Calleja argues that cross-world luck at the time of the agent’s decision is a challenge for compatibilism as well as for libertarianism. I will argue here, however, that neither of these has shown that present luck undermines all compatibilist accounts of moral responsibility. To begin, I will argue that each author fails to show that present luck is sufficiently pervasive to undermine compatibilist accounts of moral responsibility. I will then argue that even if it were pervasive, present luck would nevertheless fail to undermine compatibilist accounts of moral responsibility.¹⁵

3.1 Levy’s Argument

Levy argues that the various types of luck jointly undermine all accounts of moral responsibility. I am only concerned here with part of Levy’s project—the part in which he argues that luck undermines all compatibilist accounts of moral responsibility, including certain “history-sensitive” accounts (what he sometimes calls “historical compatibilism”). Before we turn to his argument, let me say something briefly about Levy’s intended target.

It is common to distinguish between two main types of compatibilist accounts: historicist accounts and structuralist accounts. According to the latter, whether an agent is directly morally responsible for an action depends only on her psychological structure at the time of action.¹⁶ According to historicism, by contrast, how an agent came to have her psychological structure can make a difference as to whether or not she is morally responsible.¹⁷ Historicists posit historical conditions on moral responsibility that specify when an agent’s history is or isn’t conducive to her being morally responsible. Levy’s argument targets views according to which agents can “take responsibility” for their values and dispositions and thereby avoid being constitutively lucky at later times. While this claim is typically associated with historicism (the view that whether an agent is responsible depends on her history), it is worth noting that there is conceptual space for a history-sensitive structuralist position that accepts it (by maintaining that only the degree of an agent’s responsibility may depend on her history).¹⁸ Levy’s target is any sort of history-sensitive compatibilist position.

Levy begins by distinguishing between constitutive luck and present luck, and he argues that, taken together, the two types of luck undermine history-sensitive compatibilism. Levy claims that current responses to the problem of constitutive luck expose history-sensitive compatibilism to the problem of present luck.¹⁹ I argue that Levy’s claim about history-sensitive compatibilism and present luck is false. Later (in Sect. 4), I consider whether constitutive luck alone might be problematic for history-sensitive compatibilism, and I argue that it too does not undermine history-sensitive compatibilism.

Let us follow Levy in calling what an agent inherits via constitutive luck her “endowment” (2011: 88). For example, an agent’s set of values and dispositions are part of her endowment insofar as the agent is constitutively lucky in having them—perhaps if she were born in another era or geographical region her values and dispositions would be different. On some versions of history-sensitive compatibilism, agents’ endowments are a matter of luck for them at first, and yet they are able to become responsible for their endowments by taking ownership of their values, dispositions, etc.²⁰ Mele illustrates this possibility by introducing a hypothetical agent, Chuck, who, though subject to bad constitutive luck in the form of cruel desires, decides to overcome the guilt he feels when torturing animals by hardening himself to it (2006: 171). Agents like Chuck are not mere subjects of constitutive luck but are also partly responsible for becoming the way they are since they take ownership of their endowments and bring about changes to their endowments, thereby becoming responsible. Levy accepts this response as a way to see off the problem of constitutive luck; however, he claims that it “succeeds only if compatibilism is not subject to present luck” (2011: 89).

In order to show that compatibilism is subject to present luck, Levy considers a few prima facie ways in which putatively free agents in deterministic worlds are subject to present luck: which considerations come to the agent’s mind, what mood the agent is in, whether or not the agent’s attention wanders, and the way in which an agent’s deliberation is primed by her environment (2011: 90). The list is not meant to be exhaustive but rather to display several common features of the lives of putatively free agents in deterministic worlds that give rise to present luck. Consider the example of the considerations that come to mind, an example which Levy borrows from Mele (1995, 2006). During the deliberative process, the considerations that come to mind are inputs in the agent’s decision-making. These inputs, however, appear to be lucky for the agent, since they satisfy all three conditions from our analysis of the concept of luck: the inputs are significant for her, she lacks control over which come to mind, and there are many nearby possible worlds in which different considerations come to mind. Moreover, these inputs significantly influence the agent’s action, and thus count as presently lucky for her. And as in the deliberative process, so also the agent is presently lucky in other aspects of her life. Therefore, even history-sensitive compatibilism is vulnerable to the problem of present luck.

But it is not the case that agents in deterministic worlds would be presently lucky in every case of taking ownership of their endowments.²¹ Given an agent’s moral education, her self-discipline, and her habits, it is not the case that what considerations come to mind (or whether the agent’s attention wanders, or what mood she is in, etc.) are always a matter of luck for her. Furthermore, there will be many such cases in which an agent’s training, habituation, etc. preclude present luck at the time of her taking ownership of her endowment. Consider the case of Charles:

Charles: Charles is a young agent in a deterministic universe whose endowment includes an appreciation of the value of dogs and a disposition to make sacrifices of his own time in order to benefit friendly dogs. While riding his bicycle home from school, Charles hears a dog yelp in pain. In the actual world, it occurs to Charles that the yelp might have come from his neighbor’s dog, Odie. After a few moments of reflection, Charles evaluates the relevant part of his endowment; Charles decides that he would prefer to be the type of person who is disposed to help friendly dogs even when it requires a small sacrifice in his own time, so he hurries to find the source of the yelp. Importantly, the strength of Charles’s relevant values and dispositions guarantees that in a very high proportion of nearby worlds Charles decides that he wants to be a certain type of person, disposed to help friendly dogs in need, and he goes and finds the source of the yelp. Given the strength of Charles’s values and dispositions, it is not the case that any of the following had an impact on the outcome in a large proportion of nearby worlds: the considerations that came to mind, Charles’s mood, the fact that Charles’s attention did not wander, and the way in which Charles’s deliberation was primed by his environment. Given Charles’s endowment, he would have, in most cases, come to the same decision and gone looking for the source of the yelp no matter what came to mind, no matter his mood, etc.

Even if we grant that Charles performs another action in some nearby worlds, clearly (given his values, dispositions, etc.) he will not do otherwise in sufficiently large proportion of nearby worlds for his action to count as lucky. Even if Charles lacks control over his action (though, as I will argue below, we should not accept this claim), and even if his action is significant for him, his action is not lucky, since the third condition on lucky action is that it fail to occur in a large proportion of nearby worlds, which Charles’s action does not. And since Charles takes ownership of (and brings about a change to) his endowment without being subject to present luck, we have a case in which both of the following obtain: the agent takes ownership of her endowment (in the way that Levy thinks gets around the problem of constitutive luck), and the agent’s taking ownership of her endowment does not fail to occur in a large proportion of nearby worlds. Levy’s argument does not succeed.

Now, even if Charles’s action did not technically count as lucky, if it turned out that, in this sort of case, the agent nevertheless lacked sufficient control over her action to be morally responsible for it, we would not yet have a vindication of compatibilism.²² But I do not think that we have reason to take Charles (and relevantly similar agents) to lack control. After all, Charles is rationally competent, succeeds in bringing about what he wants to bring about, and, given that the third condition on luck is not satisfied, his action’s occurrence is entirely dissimilar from paradigmatic cases of lucky outcomes (such as a coin’s landing heads rather than tails) that are not under anyone’s control. For these reasons, though we have focused on the fact that Charles’s action does not satisfy the third condition on luck, I do not think it satisfies the lack-of-control condition either.²³

Let us consider two potential objections to my reply to Levy’s argument. One might think that I have cheated by considering a case in which an agent (Charles) had certain values, dispositions, etc. that predisposed him to opt for a certain course of action, thereby making it the case that Charles would not do otherwise in a large proportion of nearby worlds; what I need to show, one might think, is that there is no present luck in cases in which agents are torn between two (or more) options. I have two comments. First, it is sufficient for my purposes to show that it is not the case that all instances of taking ownership of one’s endowment will be cases affected by present luck, and I have shown exactly this in the case of Charles. Second, typical compatibilist accounts do not require, as some libertarian accounts do, that there be cases of self-forming actions that arise from torn decisions.²⁴ If one is worried about how compatibilists can account for agents’ responsibility for having the reasons, character traits, etc. that they have, then one is worried about constitutive luck, not present luck, and constitutive luck is the subject of Sect. 4.

A second potential objection is that the proportion of nearby worlds in which an agent does some action to nearby worlds in which she fails to do that action is itself a matter of luck for the agent, even if the agent’s performing that action is not a matter of luck. This objection fails, however, to take account of the fact that the proportion of worlds will be determined in part by the agent’s constitution at the time of the action. If an agent enjoys bird-watching, for example, then the proportion of worlds in which it occurs to that agent that a potential destination is (say) on a certain migratory path (and in which the agent does otherwise for this reason) is going to be a much different proportion than one concerning an agent who does not care about birds. Now, one might worry that these aspects of a person’s character that determine the proportion of worlds in which she does otherwise are themselves a matter of luck, but this too is a worry about constitutive luck.

3.2 Pérez de Calleja’s Argument

Pérez de Calleja argues that cross-world luck at the time of decision is not uniquely problematic for libertarians; if cross-world luck is a problem at all, she thinks, then cross-world luck is a problem for compatibilists as well. In her view, it is not indeterminacy in particular that gives rise to cross-world luck, but rather other features of agents (such as their being motivationally split)—features that are compatible with determinism. As it turns out, Pérez de Calleja also argues that cross-world luck at the time of decision does not preclude free and morally responsible action, and, as I discuss below, I agree. Still, given what I have said in response to Levy’s argument, and in particular that causally determined agents (in contrast to agents satisfying libertarian conditions) need not always be presently lucky, I will argue that the problem of cross-world luck is not a problem for compatibilists in the first place.²⁵

Pérez de Calleja argues that there can be cross-world luck at the time of decision even in deterministic worlds:

My claim is that a decision performed at a time at which the agent is psychologically able (and suitably skilled and placed) to refrain from deciding that way in the circumstance is subject to cross-world luck, whether the world is deterministic or indeterministic. Bob [an agent in an indeterministic world], in particular, is lucky that he decides as he does rather than otherwise because, in a nearby possible world where the salient causes and background conditions which are relevant to his deciding one way or another in the circumstance are the same (including his reasons, his character traits and even his way of deliberating), he decides otherwise instead. If we make Bob’s world deterministic, we don’t thereby eliminate the nearby possible worlds where Bob’s counterpart does otherwise in conditions which don’t significantly differ from Bob’s. (2014: 114–115)

Imagine that Bob* is an agent who is in a deterministic world, who does some action A at t in that world, and who does otherwise some other world; to make things easier, let us call him Bob** in the nearest world in which he does otherwise than A at t.²⁶ On Pérez de Calleja’s view, there is no variance between Bob* and Bob** with respect to their reasons, character traits, way of deliberating, etc. Furthermore, there is no significant difference, she thinks, between the external conditions of Bob* and Bob**; in the nearest world in which Bob* does otherwise, “the salient causes and background conditions that are relevant to his deciding one way or another in the circumstance are the same” (2014: 115). But since everything about Bob* and Bob** is the same, and since there is no significant difference in the salient causes and background conditions in their worlds, both Bob* and Bob** are subject to cross-world luck.

Pérez de Calleja’s argument is similar to Levy’s. Since there can be nearby worlds in which, holding fixed everything about an agent that is under her control, the agent does otherwise than she does in the actual world, this cross-world difference is a matter of luck for the agent. Pérez de Calleja compares worlds in which there is no significant difference in the salient causes and background conditions, and Levy shows that such things as the agent’s mood and what comes to mind, though themselves not a significant difference between worlds, can result in the agent’s performing different actions. According to both Levy and Pérez de Calleja, what makes the cross-world difference a matter of luck is that minor (and insignificant) changes to the salient causes and background conditions in an agent’s world can result in a different outcome, and, since the agent lacks control over those salient causes and background conditions, the cross-world difference in outcomes is a matter of luck for her.

But Pérez de Calleja has not shown that compatibilism is subject to a problem of cross-world luck, for the compatibilist can show that there will be many cases in which Pérez de Calleja’s alleged cross-world luck will not in fact be cases of luck. One way to do this is to show that some of these cases are not cases in which the agent does otherwise in a large proportion of nearby worlds. Imagine that Bob* (who, remember, is in a deterministic world) has a set of reasons, character traits, and deliberating methods such that it would be unthinkable for him to turn down his dream job, were it offered to him. If Bob* is offered his dream job, the vast majority of nearby worlds will contain Bob* accepting the dream job. In such a case, Bob* is not subject to cross-world luck, since it is not the case that in a large proportion of nearby worlds Bob* does otherwise than accept the job offer.

3.3 Would Present Luck Be so Bad?

So far I have objected to Levy’s and Pérez de Calleja’s arguments on account of their failure to establish that present luck in deterministic worlds would be sufficiently pervasive to undermine the moral responsibility of causally determined agents. The cases of Charles and Bob* demonstrated that determined agents could be free of present luck, and neither Levy nor Pérez de Calleja has given us reason to think that relevantly similar cases are uncommon. Still, given that these authors point out ways in which it is possible for determined agents to be presently lucky, it is worth considering whether the sort of present luck at issue really precludes the control necessary for moral responsibility. Drawing resources from the work of Pérez de Calleja herself, as well as from Fischer (2014), I argue that it does not.

First, consider two cases. Suppose that, in the first case, you are a causally determined agent who is motivationally divided between two options, between (say) giving some bad news to your friend now or after the road trip you’ve just begun has ended—a case Pérez de Calleja introduces (2014: 117). It may be that, in this case, what you do is cross-world lucky for just the reasons that Levy and Pérez de Calleja think many determined actions are (had something else come to mind, you would have done otherwise). But suppose that, in another version of the case, what you do is not cross-world lucky (perhaps you give the news to your friend right away in a large proportion of nearby worlds), as in the cases that I have discussed in responding to Levy and Pérez de Calleja. With respect to these cases, I agree with Pérez de Calleja: “It sounds to me rather arbitrary and hence unfair to say that, in the original version of the case, cross-world luck at the time of decision precludes free will, but in this new version you may act freely in the way required for moral responsibility, since you are not subject to cross-world luck at the time of decision” (2014: 120). Given that, in these two cases, the agent performs the same action for exactly the same reasons (and from the same character, values, preferences, and so on), there does not seem to be a principled reason for taking the agent to be free and morally responsible in only the second version of the case.²⁷

One might object to the line of thought expressed in the last paragraph by accepting Pérez de Calleja’s (and my) claim that there is no relevant difference between the agents in the two types of cases but nevertheless inferring from this that agents in the second type of case are not free and morally responsible.²⁸ A problem for this objection, however, is that if one took the present luck in the first type of case to undermine the agent’s control, then one should admit that there is an important (and arguably relevant) difference between the two types of cases, since the agent in the second type of case is not subject to present luck. More importantly, though, the compatibilist who takes such things as the agent’s sensitivity to reasons (or her hierarchy of values, or…) to be constitutive of the control required for moral responsibility has a principled reason for her judgment that the agent in the second type of case is morally responsible, which is to say that, assuming the no-relevant-difference claim is correct, she has a principled reason for inferring that present luck would not undermine the agent’s control.

A second (but related) reason for thinking that a causally determined action’s being presently lucky would not preclude the agent’s having the control necessary for moral responsibility for that action is that what Fischer calls the “requisite glue” connecting actions and their causes could still be in place even if our actions are presently lucky.²⁹ Suppose that, for some causally determined agent’s action, the required connections between the agent’s action and the causes of that action are in place. On Fischer’s view, this “requisite glue” would not vanish if there were a chance that, as the result of a genuinely random machine over which the agent had no control, the agent’s action might be preempted by an alternative sequence—a case in which the indeterminacy of the agent’s action would render it presently lucky.³⁰ If the requisite glue is present in the case without the random machine, why think that it could not be present in the case in which it was possible that the machine preempted the agent’s action but did not in fact preempt it? As Fischer says, “Indeed, it should be intuitively obvious that the mere existence and operation of the machine in [the second case] is irrelevant to whatever it is that makes it the case that the responsibility-grounding relationship obtains in the sequence flowing through [the agent]” (2014: 61). On Fischer’s view, an agent’s action’s being presently lucky as a result of a random machine would not preclude the agent’s being morally responsible for that action, since the relevant connection (the “requisite glue”) between it and its causes would remain in place. In the same vein, if a causally determined agent’s action is presently lucky in the way that Levy and Pérez de Calleja discuss, we should not think that the fact that another action would have occurred had there been certain differences (such as a different thought’s occurring to the agent) undermines the agent’s moral responsibility (since the relevant connection between that action and its causes remains in place even when the agent’s action is presently lucky). So even if a causally determined agent’s action is presently (or cross-world) lucky, we should not think that this luck precludes the control necessary for moral responsibility.

4 Constitutive Luck

We have considered both recent attempts to show that compatibilism is subject to the problem of present (or cross-world) luck, and I have argued that neither of these attempts succeeds. The failure of these attempts would not be much consolation for compatibilists if constitutive luck remained a problem. In this section, I argue that constitutive luck does not undermine compatibilism. In the next section, I discuss the implications of my response to the problem of constitutive luck for a certain dispute between historicists and structuralists.

4.1 The Problem

Recall (from the discussion of Levy’s argument) the compatibilist’s strategy to show that agents can become morally responsible despite their constitutive luck: agents are able, even if determinism is true, to evaluate their values, dispositions, etc. and to adjust them upon evaluation, thereby taking ownership of their endowments and becoming partly responsible for the way they are. We must consider, however, how it is that an agent can be responsible for the evaluation and adjusting of her endowment, when the toolkit, so to speak, is a part of her very endowment. If the agent is responsible for taking ownership of her endowment because her history includes previous modifications to her endowment, then the problem is not solved but merely moved back in time to prior uses of the toolkit with which she has been endowed. If we were to continue to pursue the regress, we would eventually come to the agent’s first evaluation of her endowment. But if the values and dispositions with which she evaluates her endowment are themselves a part of her endowment, how can she be responsible for her evaluation and modification of her endowment? Since agents’ histories do not extend infinitely into the past, it seems impossible for an agent to act freely for the first time.

It is for precisely this reason that many skeptics about moral responsibility think that the conditions on moral responsibility could not possibly be satisfied.³¹ For the sake of brevity, I will only mention one: Waller (2011).³² Waller asks us to consider two agents: Betty, adapted from Mele (1995), and Benji, Betty’s twin brother, both of whom suffer from fear of their basement (2011: 30–34). The two are much alike but differ in how their respective responses to the fear of the basement. Later in life, Betty becomes a civil rights campaigner and Benji becomes “a racist who acquiesces in the racist status quo” (2011: 34). We might be inclined to praise Betty and blame Benji, but, Waller thinks, once we see that the difference in what the two become stems from a tiny difference in their starting points (their endowments), for which they are not responsible, we should agree that the two are not responsible for what they become (nor for the actions that issue from their developed characters). Both agents apparently take ownership of what they become, but it is a matter of luck for Benji that he started with a slightly different endowment than Betty and, as a result, developed into a racist. (And likewise it is a matter of luck that Betty developed into someone who cares about civil rights.)

4.2 A Solution

One response to the problem of constitutive luck is to “stare down” the luck.³³ One way to stare down the luck is, first, to acknowledge that, at the time of an agent’s first potentially free and morally responsible action, the agent is entirely constitutively lucky, and, second, to maintain that she can nevertheless be slightly responsible for her action. In other words, on this proposal, there are degrees of responsibility, and it is possible for an agent to be slightly responsible for an action that stems from a character with respect to which she is entirely constitutively lucky.³⁴ After a brief discussion of a “little agent” who is entirely constitutively lucky, I argue that it is plausible that the agent is nevertheless morally responsible to a slight degree, and I then go on to argue that part of the reason for this agent’s only being morally responsible to a slight degree is because of the agent’s constitutive luck. And, as I discuss in the next section, this bears on a certain recent dispute between historicists and structuralists.

Mele has proposed a solution to a worry about becoming responsible agents (though in the context of a discussion of libertarianism, not compatibilism) by introducing degrees of responsibility, and he illustrates this concept by describing a typical 4-year-old, Tony, and the way that 4-year-olds are typically treated:

In some cases, four-year-olds may have an urge to snatch a toy from a younger sibling and nonactionally acquire an intention to do so…In others, they may have an urge to snatch it, think (very briefly) about whether to do so, and decide to take it. Consider the first time a normal child, Tony, makes a decision about whether to snatch a toy from his younger sister. He has occasionally acted on nonactionally acquired intentions to grab his sister’s toys, but this time he gives the matter some thought and makes a decision. Tony knows that his father is nearby; and, on the basis of some unpleasant experiences, he associates taking the toy with his sister’s screaming and his father’s scolding him. He decides not to snatch it and feels a little frustrated. Imagine that Tony’s father saw that he was tempted to take the toy and was inconspicuously watching his son to see what he would do. When he saw Tony move away from his sister and pick up something else to play with, he praised him for his good behavior. The father was not simply trying to reinforce the good behavior; he believed that Tony really deserved some credit for it. (2006: 129–130)

We can imagine that Tony’s decision to refrain from snatching his sister’s toy is his first time taking ownership of his endowment. He values playing with the toy in his sister’s hand, and he is disposed to snatch toys that he values playing with, yet he decides that those values should be relegated to his value of keeping his sister from screaming and his disposition to avoid his father’s scolding.

Now, it is quite plausible, I contend, that little agents like Tony can be slightly morally responsible for making adjustments to their endowments despite being entirely constitutively lucky at that time. After all, while Tony clearly does not have the same level of impulse-control, or the same understanding of pertinent morally relevant facts, as does an ordinary adult, several facts about Tony speak in favor of his being somewhat morally responsible: he is sensitive to pertinent moral reasons, he is not compelled to refrain, he refrains because he does not want to cause his sister to scream or his father to be upset. (And, it should be noted, if you think that there are other possible features of agency that Tony lacks but that would allow for a greater degree of control in acting, we could discuss a case at another point along the continuum that would more clearly be a case of morally responsible agency.) As I understand the dialectic, the skeptic enters at this stage and reminds us that Tony is constitutively lucky, and that, as we all know, luck precludes control and thus precludes moral responsibility. But given the facts about Tony mentioned above, the skeptic’s appeal to constitutive luck only suffices to show that Tony’s control (and thus his moral responsibility) has been mitigated, not that it has been eliminated. While I agree with the skeptic that taking into account an agent’s constitutive luck can (and often should) lead us to see that the agent is less morally responsible than we had previously thought, it does not follow from this that we should take the agent’s moral responsibility to have been undermined.³⁵

If I am right that it is plausible to attribute a small degree of moral responsibility to little agents like Tony, why is this so plausible? I want to suggest a way of answering this question according to which the degree to which agents are responsible is a function of the degree to which they meet the various conditions on moral responsibility. It is widely accepted that in order to be morally responsible for an action an agent must meet, at the very least, a control condition (sometimes called the “freedom-relevant” condition, though I will simply refer to it as the control condition) on moral responsibility with respect to that action.³⁶ I have already mentioned this condition, since the various problems of luck seek to show that luck precludes the agential control required for morally responsible action. In my view, the control agents have over their actions comes in degrees, and often the degree to which agents control their actions is diminished by present and constitutive luck. Let us consider each of these claims in turn.

I take it that few would deny that agential control comes in degrees. We are often merely moderately reasons-responsive, not as strongly reasons-responsive as is possible, and this is just one way in which our agential control can come in degrees.³⁷ When it comes to certain overt actions and outcomes, luck in circumstances and results can mitigate our control.³⁸ Not only do we typically only meet the control condition on moral responsibility to some degree, however, but we often take people to be less responsible for actions when we learn that they did not meet this condition to the degree to which we had previously supposed. Imagine that your colleague, Smith, makes a remark about you that you find cruel, and you are disposed to blame Smith for hurting you by the remark. You then discover (upon meeting Smith’s mother, say) that, given Smith’s childhood, especially the way that such hurtful remarks were regarded in Smith’s family, the type of remark Smith made is one that comes second-nature for Smith, and he has made similar remarks to other colleagues in the past (though no one has told Smith that the remarks were hurtful). Upon learning that Smith’s control over making the remark was mitigated by his past circumstances, you would likely be inclined to take Smith to be less responsible for the hurtful remark (even if you did not think that he was fully exculpated). One explanation for this tendency to take people to be less responsible when we learn that they did not meet the control condition on moral responsibility to the degree that we had previously supposed is that such agents are in fact less responsible for what they have done.

In the case of little Tony, too, it is plausible to suppose that the reason we take him only to be slightly morally responsible for deciding not to snatch the toy is that he only slightly meets the control condition on moral responsibility. That is, Tony meets the control condition on moral responsibility, but he meets it only to a slight degree. Tony has some control over his decision not to snatch the toy, but, since this is his first time taking ownership of the desires and traits with which he has been endowed, his constitutive luck diminishes the degree to which he meets the control condition. One explanation for Tony’s being only slightly morally responsible for his decision, then, is that he meets the control condition (and presumably any other condition there is) on moral responsibility, but only to a slight degree.³⁹

Is it true that Tony’s control is mitigated by his constitutive luck, or is it rather the case that Tony satisfies the control condition to a slight degree for other reasons, such as that his impulse control is much less mature than an ordinary adult’s?⁴⁰ In my view, while Tony’s control may be mitigated for a variety of reasons, his constitutive luck is one of them, and this can be seen by comparing two agents who are exactly alike in every way except with respect to their constitutive luck. Suppose that Anthony is a time-slice duplicate of Tony but has, at some point in the past, endorsed the relevant aspect of his endowment. He has been in relevantly similar circumstances as Tony finds himself now and has chosen to show kindness to his sister and to seek his father’s approval. Now Anthony finds himself in exactly Tony’s circumstances (and with all of Tony’s time-slice properties) and refrains from taking his sister’s toy. Intuitively, Anthony exercises more control than Tony in these circumstances, and given that the only difference between them is a difference in their constitutive luck, the reason for this difference in control is that Anthony is not as constitutively lucky as Tony. (In the next section, I will argue that the same applies to mature, adult agents.)

Now we are in a position to reconsider our judgments concerning Waller’s (2011) Betty and Benji. The apparent trouble with praising Betty for becoming a civil rights campaigner and with blaming Benji for becoming a racist is that the difference in what the two become stems from a tiny difference in their endowments, for which they are not responsible. But, on the model that I am proposing, it is not completely a matter of luck that Betty ended up praiseworthy and Benji blameworthy. Consider Benji: when confronting his fear of the basement, he was (we are supposing) entirely constitutively lucky to have his endowment, yet, if he is indeed responsible for becoming a racist, there were intermediate stages at which he had opportunities to reflect on his endowment, take ownership of it, and become more responsible for it over time. At first, he might have been responsible for his endowment to only to a slight degree, but there was no immediate connection between his response to his fear of the basement (with respect to which he was entirely constitutively lucky) and his racism (with respect to which he was not entirely constitutively lucky). Over the years, Benji made modifications to his endowment, sometimes reaffirming aspects of his character that he had already owned. Provided that Benji has such opportunities, as we are supposing, there is no reason to think that it was completely a matter of luck for him that he became a racist rather than a civil rights campaigner, and so there is no reason to think that constitutive luck undermines Benji’s responsibility for what he has become.

5 Historicism and Structuralism

So far I have argued that compatibilism is not undermined by the challenges associated with present and constitutive luck. As it turns out, my replies to these challenges are consistent with both historicism and structuralism—the two types of compatibilist account that I mentioned in Sect. 3. Again, structuralists claim (and historicists deny) that whether an agent is morally responsible depends only on the structure of an agent’s psychology at the time of her action. While my replies are consistent with each type of account, the proposal that I sketched in Sect. 4 (in response to the problem of constitutive luck) does entail that degrees of moral responsibility (though not moral responsibility itself) is history-sensitive. And this is because whether a person is constitutively lucky or not (and how constitutively lucky a person is) depends upon how she came to have the character she has. So, if constitutive luck mitigates control (and thus moral responsibility), then the degree to which one exercises control (and thus the degree to which one is morally responsible) can depend upon one’s history.

Since all of this is consistent with genuine historicism, according to which whether an agent is morally responsible at all depends upon her having or not having a certain kind of history, and also consistent with structuralism (albeit a history-sensitive structuralism), what I have said so far does not settle the dispute between historicists and structuralists.⁴¹ However, my response to the problem of constitutive luck does bear on a certain recent dispute between historicists and structuralists concerning certain magical agents. Michael McKenna has recently defended structuralism (though he himself remains agnostic with respect to this debate) by appealing to “instant agents” who are created to be psychological duplicates of normal adults.⁴² I claim that the present discussion of luck not only has implications for McKenna’s instant agents but also gives us a reason to prefer a history-sensitive structuralism (or genuine historicism) over McKenna’s proposed version of structuralism.

McKenna attempts to motivate structuralism by introducing a hypothetical instant agent, Suzie Instant, who was created by a god “to be a psychologically healthy woman indistinguishable from any other normally functioning thirty-year-old person” (2004: 180). In fact, Suzie Instant is created with a psychological condition that is identical to the psychological condition of Suzie Normal, who came into the world in the usual way.⁴³ McKenna argues that there is no non-arbitrary difference between Suzies with respect to freedom and moral responsibility, and this fact, he thinks, supports structuralism.⁴⁴

Given the model for dealing with the problem of constitutive luck that I have presented here, however, there is reason to think, contra McKenna, that there is a non-arbitrary difference between Suzies with respect to responsibility, namely that Suzie Instant’s responsibility is mitigated by her constitutive luck. Like little Tony, Suzie Instant is entirely constitutively lucky at the time of her action. Suzie Normal, on the other hand, has presumably had many opportunities to take ownership of her endowment, thereby overcoming her constitutive luck. Even if McKenna is right that Suzie Instant is morally responsible, given her psychological condition, she is not as responsible as Suzie Normal. In fact, Suzie Instant would be like little Tony with respect to her constitutive luck, and despite Suzie Instant’s satisfaction of the various conditions on moral responsibility (which might be met to a greater degree in her case than in Tony’s), her responsibility is mitigated by her constitutive luck. The type of response to the problem of constitutive luck that I defend gives the history-sensitive structuralist (or the historicist) the resources to point to a non-arbitrary difference between the Suzies with respect to responsibility. Furthermore, since a history-insensitive structuralist account (like McKenna’s proposal) denies that an agent’s history makes a difference to her moral responsibility, such accounts will, in some cases (such as the Suzies’ cases), render constitutively lucky agents just as responsible as non-lucky ones. Since this leaves the problem of constitutive luck intact, this counts as a reason to prefer a history-sensitive account (whether a historicist or structuralist one) to a history-insensitive account.⁴⁵

6 Conclusion

I have argued that neither the problem of present (or cross-world) luck nor the problem of constitutive luck undermines compatibilism. I first responded to two recent attempts to show that compatibilist accounts are vulnerable to the problem of present (or cross-world) luck, and I then gave a sketch of a compatibilist response to the problem constitutive luck according to which moral responsibility comes in degrees and constitutive luck can diminish the degree to which agents are morally responsible. Essential to this response to the problem of constitutive luck is that the degree to which one is morally responsible can depend upon one’s history, and I argued that an upshot of this reflection on constitutive luck is a reason to prefer a history-sensitive account (whether historicist or structuralist) rather than a history-insensitive structuralist account.