II. Frankfurt-Style Cases and Flickers of Freedom

A genuine counterexample to PAP will be a case in which an agent is morally responsible for performing some action she could not have done otherwise than perform.^{Footnote 4} Because PAP is generally understood to be a thesis about basic (that is, direct, or nonderivative) moral responsibility, the relevant actions in these cases tend to be mental acts of deciding, which are commonly taken to be the things for which we are basically morally responsible and from which our responsibility for overt bodily actions (and their consequences) derives.^{Footnote 5} To ensure that agents cannot do otherwise than decide to act as they actually do, typical Frankfurt-style cases include the presence of a counterfactual intervener, someone who stands ready and able to force the agent to decide to act in the desired way in the event that the agent is not going to do so on her own. Crucially, however, the counterfactual intervener’s presence is unknown to the agent and plays no role in the agent’s actual behavior, lest it undermine the intuition that the agent is morally responsible for what she does.

The tricky part for cases of this sort comes in identifying the precise conditions under which the counterfactual intervener would intervene. If the counterfactual intervener has to wait to see how the agent actually decides, then the agent retains the ability to do otherwise, and PAP is unthreatened. In order to guarantee that the agent will decide to act in the desired way without actually compelling the agent’s behavior, the counterfactual intervener must have some way of knowing in advance what the agent will do if left to act on her own. Toward this end, traditional Frankfurt-style cases posit the existence of a prior sign—something beyond the agent’s voluntary control (such as a blush or a cluster of neural firings) that indicates how the agent would freely act, were no intervention to occur.

The trouble with including indicators of this sort is that they are capable of doing the work set out for them in traditional Frankfurt-style cases only by presupposing a deterministic relation between an agent’s free action and its causal antecedents—a relation incompatibilists flatly reject.^{Footnote 6} This problem can be seen as posing the following dilemma. Either the occurrences of these prior signs (or that which they indicate) make it merely probable that agents in these cases will decide to act in the desired ways, or they deterministically lead to the agents so deciding. If the occurrences of these signs make it merely probable that the agents will decide to act in the desired ways, then it remains causally open to them not to decide to act in this way after all. Insofar as this is the case, however, the agents will continue to possess alternative possibilities with respect to their actions, and so their bearing moral responsibility for these actions is perfectly compatible with PAP. If, on the other hand, the occurrences of these signs (or that which they indicate) will deterministically lead to the agents deciding to act in the desired ways, then it will no longer remain causally open to the agents not to decide to act as they do. But since having one’s actions causally determined precludes one from being morally responsible for them, according to incompatibilists, the claim that agents in these cases are nevertheless morally responsible for what they do cannot be asserted without begging any questions.^{Footnote 7} Either way, then, traditional Frankfurt-style scenarios fail to constitute genuine counterexamples to PAP. ^{Footnote 8}

This particular line of objection has prompted a number of attempts to modify traditional Frankfurt-style cases in ways that will enable them to do the work set out for them by Frankfurt without employing the use of (determinative) prior signs.^{Footnote 9} Here is a modified Frankfurt-style case offered by Alfred Mele and David Robb:

Rather than rely on the presence of some prior sign concerning what Bob will freely do if left to act on his own, Mele and Robb employ a particular kind of causal preemption to guarantee that an intervention will take place if and only if Bob does not decide on his own to steal Ann’s car (and is capable of making a decision at that time). The significance of the preemption’s occurring precisely at t2 (the moment of action) is that it ensures that at no point leading up to and including t2 is it possible for Bob to do otherwise than decide at t2 to steal Ann’s car. Because the process Black initiates in Bob’s brain is insensitive to anything Bob does prior to t2, Mele and Robb avoid the need to controversially presuppose a deterministic relation between Bob’s action and any of its causal antecedents. And because the process initiated by Black is causally preempted by Bob at t2, and so plays no causal role in Bob’s actual decision to steal Ann’s car, there appears to be no reason to deny that Bob can be genuinely morally responsible for this decision.

Elsewhere I have argued at length that even if the kind of modified Frankfurt-style case just described successfully avoids the dilemma facing traditional prior-sign cases, it nevertheless fails to falsify PAP. I will not rehearse all that I said there, but the core of my argument can be put as follows.^{Footnote 11} It is generally accepted by those on both sides of this debate that PAP is properly understood as a thesis about basic (that is, direct, or nonderivative) moral responsibility. Indeed, it is only when PAP is understood in this way that the “need” for Frankfurt-style examples arises in the first place. Were it taken to apply to instances of derivative (or indirect) moral responsibility as well, falsifying this principle would not require so much talk of nefarious neuroscientists and their science-fictional devices; standard drunk-driver cases would do the trick.^{Footnote 12} So understood, then, PAP claims that a person is basically morally responsible for what he has done only if he could have done otherwise. Importantly, it is the referent of the antecedent which fixes that of the consequent, not the other way around. That is to say, that than which a person must have been able to do otherwise, according to PAP, is determined by that (whatever it is) for which he is basically morally responsible.

In order for a Frankfurt-style case to count as a genuine counterexample to PAP, it must be one in which an agent could not have done otherwise than that for which he is basically morally responsible. The example provided by Mele and Robb is not such a case. Even supposing it is true both that Bob could not have done otherwise than decide to steal Ann’s car and that the deterministic process initiated by Black played no causal role in bringing about Bob’s decision, it is not deciding to steal Ann’s car for which Bob is basically morally responsible.^{Footnote 13} This is not to deny, I hasten to add, that Bob is genuinely morally responsible for deciding to steal Ann’s car, only that he is basically morally responsible for this. It is deciding on his own to steal Ann’s car for which Bob is basically morally responsible and from which his moral responsibility for so deciding simpliciter derives.^{Footnote 14}

One reason for thinking that Bob is basically morally responsible not for deciding simpliciter to steal Ann’s car but for so deciding on his own is implicit in Mele and Robb’s case itself. Like virtually all other Frankfurt-style scenarios (both in their traditional and modified forms), Mele and Robb’s case focuses on a mental act of deciding rather than an overt bodily action. This is no accident. Instead, it is the product of two widely shared assumptions. The first is that PAP is properly understood as a thesis about basic moral responsibility. The second is that responsibility for overt bodily actions (such as raising one’s arm) derives from responsibility for certain mental actions (such as deciding to raise one’s arm) much in the same way that responsibility for the consequences of one’s actions derives from one’s responsibility for the actions themselves. On this view, if one is morally responsible both for deciding to raise one’s arm and for raising one’s arm, it is the former for which one is basically morally responsible, whereas one is only derivatively responsible for the latter. This variation in the kind of responsibility agents bear for certain mental actions as compared to overt bodily actions is best explained by a difference in control. The successful performance of even simple overt actions like raising one’s arm is much more vulnerable to environmental impediments than the successful performance of simple mental actions like deciding to raise one’s arm. It is because agents have increased control over certain mental actions, such as deciding, that they are basically morally responsible for these, rather than for overt bodily actions.^{Footnote 15}

Things are a bit different in Frankfurt-style scenarios. Looking at Mele and Robb’s case, we can see that Bob has the same degree of control over whether he will steal Ann’s car as he does over whether he will decide to steal Ann’s car—namely, none. It is guaranteed that Bob will decide to steal Ann’s car and that he will steal it, and there is nothing he can do to prevent either of these things from occurring. Even in this case, however, whether Bob will decide on his own to steal Ann’s car, and whether he will steal Ann’s car on his own, remains entirely up to him. Thus, the same reasons typically taken to support the view, implicit in Frankfurt-style examples, that one’s responsibility for performing some overt action A derives from one’s responsibility for performing some mental action, such as deciding to A, equally suggest that, at least in Frankfurt-style cases, one’s responsibility for deciding to A derives from one’s responsibility for so deciding on one’s own, for which one is basically morally responsible.^{Footnote 16}

So, even if Bob is genuinely morally responsible for deciding to steal Ann’s car, his responsibility for this derives from his responsibility for deciding on his own to steal Ann’s car. Of course, since deciding on his own to steal Ann’s car is something it was within Bob’s power not to have done, the judgment that he is (basically) morally responsible for doing this is consistent with the truth of PAP.^{Footnote 17}

It is important to note that none of the foregoing hinges upon anything peculiar to Mele and Robb’s particular example. Rather, it capitalizes on two features that are essential to any Frankfurt-style scenario: the agent performs the relevant action on his own (without being forced), and it was possible for the agent not to have done so. The fact that agents in these cases act on their own, just as they would have if no ensuring mechanism had been in place, is what is supposed to drive the intuition that they are morally responsible for what they do, without which there is no threat to PAP. Furthermore, omitting to perform the relevant action on one’s own is a genuine alternative possibility in all Frankfurt-style scenarios (and one which in modified cases lies safely within the scope of the agents’ direct voluntary control). Indeed, it is precisely because it is possible for agents in these cases not to perform the relevant actions on their own that the ensuring mechanisms are there to begin with; otherwise these mechanisms would be superfluous. Since this alternative possibility is ineliminable, the particular version of the flicker strategy employed here can be brought to bear against any modified Frankfurt-style example. Insofar as it is successful, the upshot is not only that PAP has yet to be falsified by any of the Frankfurt-style cases currently on offer but that cases of this kind are theoretically incapable of falsifying PAP.

III. Why Must Flickers Be Robust?

The chief objection to the flicker strategy comes from John Martin Fischer. The problem with the various versions of this approach, he argues, is that the alternative possibilities on which they focus are not sufficiently robust:

According to Fischer, neutralizing the threat to PAP from Frankfurt-style cases requires not only that agents in these cases could have done otherwise than that for which they are (basically) morally responsible but that the existence of these alternatives is able to help account for why agents would, in virtue of possessing these alternatives, be morally responsible for what they actually do. As he puts it, “the flicker theorist must claim that the addition of the sort of possibility he has identified would transform a case of lack of responsibility into one of responsibility.”^{Footnote 19}

Fischer grants that it remains open to agents in Frankfurt-style cases not to decide on their own to act as they do. The reason this alternative is not sufficiently robust to figure into an account of why these agents are morally responsible for what they do, he maintains, is that whether this alternative is realized depends on whether the agents display some prior sign indicating that they are not going to so decide on their own, thereby prompting the ensuring mechanism to kick in and force them to decide to act in the desired way. By hypothesis, however, the display of these signs (such as a blush or a neurological pattern in the brain) is involuntary. Since it is involuntary, it is implausible to suppose that the addition of the possibility of displaying one of these signs is itself sufficient to transform a case of lack of responsibility into one of responsibility. This, Fischer says, would “appear to involve a kind of alchemy, and it is just as incredible.”^{Footnote 20}

One thing to notice about this line of response is just how dependent it is upon the presence of some prior sign indicating how agents in these scenarios will “freely” behave if left to act on their own. It is this very feature of traditional Frankfurt-style cases, however, that has been the target of heavy criticism in recent years, prompting the creation of modified Frankfurt-style examples (such as Mele and Robb’s) that do not rely on prior signs of this sort. As noted earlier, one of the effects of eliminating determinative prior signs from Frankfurt-style examples is that, right up until the moment of choice (the locus of free action), it remains within the agents’ direct voluntary control whether they decide on their own to perform the relevant action. Thus, whatever the merits of this particular attempt to undermine the flicker defense of PAP against traditional Frankfurt-style cases, in the context of modified Frankfurt-style scenarios it has no force. Those wishing to preserve this line of objection to the version of the flicker strategy defended here will need to show why omitting to decide on one’s own is not a robust alternative possibility even when it lies within the scope of an agent’s direct voluntary control.

It is also important to see that, since PAP makes no mention of the necessity of robustness (nor of anything that implies the necessity of robustness), merely defending PAP against falsification by Frankfurt-style cases does not require that agents in these cases possess distinctly robust alternative possibilities.^{Footnote 21} There is, however, another way that these cases can be used to challenge PAP, one for which the demand for robustness would be appropriate. Here I make a distinction between two kinds of Frankfurt-style attack, each of which corresponds to a different way of conceiving the role of Frankfurt-style examples. In the first kind of Frankfurt-style attack, which I shall call direct Frankfurt-style attack, the role of Frankfurt scenarios is to act as counterexamples that falsify PAP. In the second kind of attack, which I shall call indirect Frankfurt-style attack, the aim of these cases is not to falsify PAP but to falsify some other principle to which it is connected in some important way.^{Footnote 22}

This second kind of Frankfurt-style attack is suggested by something Fischer says. He writes, “Of course, I do not accept the alternative-possibilities control model of moral responsibility. But my contention is that, if you do buy into this traditional picture, then you should also accept that the alternative possibilities must be of a certain sort—they must be sufficiently robust.”^{Footnote 23} According to Fischer, then, one who endorses PAP should also endorse something like the following general principle (call it the Principle of Robust Alternative Possibilities [PRAP]):

Since PAP does not entail PRAP, proponents of the former are not forced to accept the latter upon pain of contradiction. Why, then, should proponents of PAP endorse PRAP? An earlier passage seems to provide the answer:

In Fischer’s view, it is the intuition that moral responsibility requires a certain sort of control—namely, the type of control that involves possessing not just any alternative possibilities but ones that are distinctly robust—which lies behind the demand for alternative possibilities. It is PRAP that is the fundamental intuition, not PAP. Of course, PRAP entails PAP, so one who accepts the former should also accept the latter; but, again, PRAP is fundamental.

Mounting an indirect Frankfurt-style attack against PAP involves two steps. The first step is to establish a connection between PAP and some other principle, in this case PRAP. This is where Fischer’s point about the basic picture underlying the traditional alternative-possibilities view of moral responsibility comes into play. I suggest that the most plausible way to think about what Fischer says concerning the relationship between PAP and robustness is that PRAP (or some other similar principle) grounds PAP in such a way that PAP is plausible only if PRAP is plausible. The second step in indirect Frankfurt-style attack is to falsify PRAP (or at least show that it is implausible). This is where Frankfurt-style cases come in. The role of these cases in indirect Frankfurt-style attack is to act as counterexamples to PRAP.

One advantage of indirect Frankfurt-style attack (over against direct Frankfurt-style attack) is that Frankfurt-style cases need not eliminate all alternative possibilities in order to do the work set out for them. In order for a case of this sort to count as a genuine counterexample to PRAP it need only be one in which an agent has no robust alternative possibilities to doing that for which he is basically morally responsible. (Like PAP, PRAP should be taken to be a thesis about basic moral responsibility.) A further advantage of using Frankfurt-style cases in this way is that, here, the demand for robust alternative possibilities is entirely fitting. Any flicker-of-freedom-type defense of PRAP against Frankfurt-style cases, if it is to be successful, will need to show both that agents in these cases continue to possess alternative possibilities and that these alternatives are sufficiently robust.

Of course, since PAP does not entail PRAP, falsifying PRAP is insufficient to demonstrate the falsity of PAP. Thus, indirect Frankfurt-style attack is not able to show that PAP is false (then again, as I have argued, neither is direct Frankfurt-style attack). Rather than falsifying PAP, the aim of indirect Frankfurt-style attack is to demotivate this principle by showing that the reasons for endorsing PAP in the first place ought not to be accepted. This in itself would be significant. And given that Frankfurt-style examples are theoretically incapable of falsifying PAP, I suggest that this is the most that opponents of PAP can reasonably hope to achieve with cases of this sort. To accomplish even this much, however, it must be shown both that PAP is plausible only if PRAP is plausible and that there are possible Frankfurt-style scenarios in which agents without any robust alternative possibilities are nevertheless (basically) morally responsible for what they do.

For present purposes, I am willing to grant that PAP is plausible only if PRAP is also plausible, and that falsifying PRAP would have the result of demotivating PAP in the way suggested above. In what follows, I want to consider whether there could be Frankfurt-style scenarios in which agents without any robust alternatives are nevertheless (basically) morally responsible for what they do. First, however, we must get clearer on the nature of robustness.

IV. What Does Robustness Require?

The robustness challenge, originally offered by Fischer, has been echoed by a number of prominent defenders of Frankfurt to address the flicker strategy. It is generally agreed that a robust alternative is one that is capable of grounding ascriptions of moral responsibility—that is, one that is relevant in and of itself to explaining why an agent is morally responsible for what she has actually done. There is some variation, however, when it comes to spelling out just what is required for an agent to have alternatives that are robust in this sense. Michael McKenna maintains that, in addition to being within an agent’s voluntary control, a robust alternative must be such that “It would tell us something (different from what we are told in the actual world) about the moral quality of the agent’s conduct were she to have so acted in this alternative scenario.”^{Footnote 25} To do this, he says, an alternative must be both morally significant (that is, morally better or worse than the action the agent actually performs) and deliberatively significant (that is, reasonable to expect the agent to have considered as an alternative to what she actually does, given her agent-relative perspective). According to Derk Pereboom, an agent’s having a robust alternative requires that the agent “could have willed something other than what she actually willed such that she understood that by willing it she would thereby have been precluded from the moral responsibility she actually has for the action.”^{Footnote 26}

For present purposes, I want to grant that robustness requires everything that Fischer, McKenna, and Pereboom say it does—with one possible exception. At certain points, some of their characterizations of robustness seem open to being interpreted as claiming that a robust alternative must involve an action (even if only a mental act of deciding). This is true of the passage by Pereboom quoted above, and Fischer has at times expressed a similar sentiment:

Like Fischer, McKenna and Pereboom maintain that the mere possibility of having something different occur cannot plausibly contribute to an explanation of an agent’s being morally responsible for what she does. To do this, an alternative must be within an agent’s control.

The point that alternatives can be robust only if they are within an agent’s control is well-taken. It is certainly plausible that whatever intuitive force accompanies PAP stems from the view that moral responsibility requires what Fischer describes as “the ability to make a difference in the sense of selecting one from various paths the world could take, where these various paths are all genuinely available to the agent.”^{Footnote 28} Unless the availability of various alternatives enhances an agent’s control over the way things go, it is unclear how this would help ground attributions of moral responsibility.

Importantly, however, agents can see to it that the world takes one path rather than another, not only by acting but also by omitting to act.^{Footnote 29} Consider a normal case of me sitting at the bus stop, waiting for my bus to arrive. Once it arrives there are various paths the world could take that are genuinely available to me. One path involves me getting on the bus and making it home in time for dinner, and I can see to it that the world takes this path by deciding to get on the bus, which it is within my direct voluntary control to do. There are also a number of other paths that do not involve me getting on the bus or making it home in time for dinner, and I can see to it that the world takes one of these paths rather than the first one. One way I can do this is by making a different decision—either a decision not to get on the bus or a decision to do something else that is incompatible with my getting on the bus then. Importantly, however, this is not the only way I can control which path the world takes. Another way I can see to it that the world takes a path that does not involve me getting on the bus is by simply omitting to (decide to) get on the bus. This too is within my direct voluntary control. Just as there is nothing else I need to do in order to decide (not) to get on the bus—I simply decide—there is nothing I need to do in order to omit to (decide to) get on the bus—I simply omit to (decide to) do so.

Given that omissions may equally lie within agents’ direct voluntary control, there appears to be no justification for insisting that only actions are capable of constituting robust alternatives. As long as their claims are understood in a way that is compatible with that, let us assume that robustness requires everything that Fischer, McKenna, and Pereboom say it does. Again, this will be helpful in demotivating PAP only if it can be shown that robust alternatives of this sort are unnecessary for moral responsibility.

V. Robust Flickers of Freedom

We are now in a position to see why indeterministic Frankfurt-style cases are theoretically incapable of falsifying PRAP and, hence, why indirect Frankfurt-style attack (like direct Frankfurt-style attack) cannot possibly succeed.^{Footnote 30} To begin, assume that, like PAP,

1. (1) PRAP is a thesis about basic moral responsibility.

Also like PAP, PRAP is a conditional in which the referent of the consequent is fixed by that of the antecedent (and not the other way around). That is, PRAP requires that a person had a robust alternative to doing that for which he is basically morally responsible. Consequently,

1. (2) A counterexample to PRAP must be a case in which a person had no robust alternative to doing that for which he is basically morally responsible.

The first major step in this argument is to show that

1. (3) Agents in Frankfurt-style cases are basically morally responsible for deciding on their own to act as they do (not for so deciding simpliciter).^{Footnote 31}

Earlier I argued that the same reasons that support the view that moral responsibility for overt actions (such as raising one’s arm) derives from responsibility for certain mental actions (such as deciding to raise one’s arm) equally support the view that, at least in Frankfurt-style cases, agents’ responsibility for deciding to perform the relevant actions derives from their responsibility for so deciding on their own (for which they are basically morally responsible). There is another reason for thinking that agents in Frankfurt-style cases are basically morally responsible for deciding on their own to act as they do, and not for so deciding simpliciter, which we can see by considering the following question. Assuming that Bob (from Mele and Robb’s case) is morally responsible both for deciding on his own to steal Ann’s car and for deciding to steal Ann’s car, what should we think about the relationship between his responsibility for the former and his responsibility for the latter? Three possibilities present themselves:

1. (i) Bob’s moral responsibility for deciding to steal Ann’s car derives from his responsibility for deciding on his own to steal Ann’s car.

2. (ii) Bob’s moral responsibility for deciding on his own to steal Ann’s car derives from his responsibility for deciding to steal Ann’s car.

3. (iii) Neither Bob’s moral responsibility for deciding on his own to steal Ann’s car nor his responsibility for deciding to steal Ann’s car derives from the other; each is something for which Bob is basically morally responsible.

Of these options, (ii) appears to be the least plausible. Indeed, it is difficult to imagine what reason there could be for thinking that Bob’s moral responsibility for deciding on his own to steal Ann’s car derives from his responsibility for so deciding simpliciter—particularly since it seems so natural to explain the latter in terms of the former. This is evinced by the fact that so many proponents of Frankfurt-style argument (including Mele and Robb, and even Frankfurt himself) have sought to support the judgment that agents are genuinely morally responsible for their decisions in these cases precisely by pointing out that these are decisions the agents made on their own. For this reason, (iii) too seems problematic insofar as it suggests that there is no connection between Bob’s responsibility for deciding to steal Ann’s car and his responsibility for so deciding on his own. Simply on the face of it, then, (i) appears to provide the most (if not the only) plausible way of thinking about the relationship between Bob’s responsibility for deciding on his own to steal Ann’s car and his responsibility for so deciding simpliciter.

From (2) and (3) it follows that

1. (4) A counterexample to PRAP must be a case in which a person had no robust alternative to deciding on his own to act as he did.

As noted earlier, here I want to grant that having robust alternatives involves everything Fischer, McKenna, and Pereboom say it does (again, so long as their claims are not interpreted as requiring robust alternatives to be actions). Thus, let us assume that

1. (5) An alternative is robust if and only if it was (i) morally significant, (ii) deliberatively significant, (iii) an alternative that the agent would understand to be a way of precluding himself from being (basically) morally responsible for what he did, and (iv) within the agent’s voluntary control.

For the sake of simplicity, most (if not all) of the debate surrounding PAP, Frankfurt-style cases, and robustness has limited the focus to cases involving blameworthiness. Following that custom, let us assume that the relevant actions in these cases are morally bad actions for which the agents are meant to be not only morally responsible but blameworthy.

The next major step is to show that the alternative of voluntarily omitting to decide on one’s own to do that which one knows to be wrong satisfies all the conditions for robustness identified in (5). To begin, I take it to be noncontroversial that

1. (6) Omitting to decide on one’s own to do that which one knows to be wrong is a morally significant alternative to so deciding on one’s own (in all Frankfurt-style cases).

Arguably, omitting to decide on one’s own to do that which one knows to be wrong (all things considered) is always a morally significant alternative to (because it is morally better than) deciding on one’s own to do that which one knows to be wrong. Possible exceptions to this might be cases in which a person knows that her deciding to A is unavoidable, regardless of whether she decides on her own to A (say, because she knows that she will be forced to decide to A unless she so decides on her own). Of course, in such cases it is no longer obvious that performing the action would be morally wrong, as opposed to just morally bad. Regardless, since it is an essential feature of Frankfurt-style cases that the agents neither believe nor ought to believe that there are counterfactual interveners about, ready to force them to decide to A unless they do so on their own, these considerations do not apply in Frankfurt-style cases.

I also take it to be evident that

1. (7) Omitting to decide on one’s own to do that which one knows to be wrong is a deliberatively significant alternative to so deciding on one’s own (in all indeterministic Frankfurt-style cases).

Assuming that agents in these cases are morally responsible for what they do (as we are to suppose), omitting to decide on their own to do that which they know to be wrong is a morally significant alternative that it is reasonable for us to expect them to consider as an alternative to deciding on their own to do that which they know to be wrong.^{Footnote 32}

Similarly, any agent satisfying the epistemic conditions for morally responsible agency will understand that she can avoid being (basically) morally responsible for doing something by simply omitting to do so on her own. For any morally competent agent will understand that one is not (basically) morally responsible either for actions one is forced to perform nor for actions one does not perform at all, and these are the only possibilities when one omits to perform an action on one’s own.^{Footnote 33} Thus,

1. (8) Omitting to decide on one’s own to perform a given action is an alternative that one would understand to be a way of precluding one from being (basically) morally responsible for so deciding (in all Frankfurt-style cases).

Next, it is trivially true that

1. (9) Voluntarily omitting to decide on one’s own to do that which one knows to be wrong is within one’s voluntary control whenever one voluntarily omits to do so.

From (5) through (9), it follows that

1. (10) Voluntarily omitting to decide on one’s own to do that which one knows to be wrong is a robust alternative to so deciding on one’s own (in all indeterministic Frankfurt-style cases).

That is to say, voluntarily omitting to decide on one’s own to do that which one knows to be wrong is an alternative that, when it remains open to agents in indeterministic Frankfurt-style cases, is a distinctly robust alternative possibility. Premise (10) should not be understood as saying that this alternative in fact remains open to agents in all indeterministic Frankfurt-style cases. (That claim comes next.)

The final major step in the argument is to show that

1. (11) It is an essential feature of indeterministic Frankfurt-style cases that it is possible for agents in these cases to voluntarily omit to decide on their own to do that which they know to be wrong.

It is precisely because it is possible for these agents not to perform the relevant actions on their own that the ensuring mechanisms are there in the first place; otherwise, they would be superfluous, which they are not. Without a mechanism that stands ready to force the agents to make the relevant decisions unless they do so on their own, there is nothing to guarantee that the agents make the relevant decisions at all. Crucially, however, even if these ensuring conditions are able to guarantee that the agents make the relevant decisions, they are not able to guarantee that the agents make the relevant decisions on their own. There is no way to guarantee that an agent will make a certain decision on his own because it is simply impossible (indeed, incoherent) to force an agent to do something on his own (that is, without being forced). As a result, it is always possible for an agent not to do a thing on his own. In indeterministic cases, this option is available right up until the moment of choice, within the scope of agents’ voluntary control.^{Footnote 34} The only way to prevent an agent from voluntarily omitting to decide on her own to perform an action she is considering performing, without causally determining what she does, is to render her unconscious or otherwise not in control of her behavior, thereby precluding her responsibility. Thus, provided they are morally responsible for what they do, agents in these cases will be able to voluntarily omit to decide on their own to do so.

To be clear, I allow that it is possible to construct indeterministic Frankfurt-style cases in which agents are morally responsible for deciding to perform some action despite being unable to intentionally omit to decide on their own to perform that action.^{Footnote 35} But an omission need not be intentional in order to constitute a robust alternative possibility. It must be within an agent’s control, to be sure, but an omission can be within an agent’s control (even direct voluntary control) without being intentional. ^{Footnote 36}

The kind of omission I here have in mind corresponds to a class of action that some philosophers have dubbed “nonintentional action.”^{Footnote 37} To illustrate, consider a dentist who performs a necessary procedure on a patient and, in so doing, causes the patient to be in pain. Suppose that, as she is performing the procedure, the dentist is consciously aware of the fact that she is causing the patient pain, even though she has no intention of doing so and is doing everything she can to minimize the patient’s pain and discomfort.^{Footnote 38} Because the dentist is in control of her actions and is aware that she is causing the patient pain as she is doing it, it would seem to be a mistake to say that she does this unintentionally. At the same time, however, it would also be inaccurate to say that she intentionally causes the patient to be in pain since causing the patient pain in no way figures into the content of any intention she has with respect to performing the procedure. Fortunately, as some theorists point out, there is a middle ground: “insofar as an agent who is A-ing is neither aiming at A-ing nor trying to A, either as an end or as a means to (or constituent of) an end, she is not intentionally A-ing; insofar as an agent is A-ing knowingly and nonaccidentally, she is not unintentionally A-ing; and actions that are neither intentional nor unintentional are nonintentional.”^{Footnote 39} I suggest that, like actions, omissions too can be nonintentional in this way. Just as one can knowingly and voluntarily perform an action and yet not perform it intentionally, one can knowingly and voluntarily omit to perform an action without intentionally omitting to perform it.

To see this, think back to the ordinary case of my sitting at the bus stop from earlier. Once my bus pulls up to the stop, there are various paths the world can take. Some paths involve my getting on the bus and making it home in time for dinner, while many others do not. Of those that do not, some paths involve my getting up to do something else then (climb a tree, go to a bar, stroll through the park), while other paths do not involve my doing anything other than remaining seated until the bus pulls away and simply omitting to get on the bus. Consider this last path. If my remaining seated until the bus pulls away were caused by (or otherwise appropriately related to) a decision or an intention not to get on the bus, then we could reasonably say that I intentionally omitted to get on the bus. If, instead, my remaining seated until the bus pulls away were the result of my getting lost deep in thought and failing to recognize that the bus had arrived, then this would seem to be a clear case of my unintentionally omitting to get on the bus (and not something I did voluntarily). But what would nonintentionally omitting to get on the bus amount to? What would it look like to knowingly and voluntarily, yet not intentionally, omit to get on the bus? Suppose that, when the bus arrives, I am still uncertain whether to get on the bus. I recognize that the bus has arrived at the stop and will soon depart without me unless I (decide to) get on the bus. As the seconds tick by, I continue to remain seated at the bus stop. I have not decided (and do not intend) to get on the bus, nor have I decided (or otherwise come to intend) not to get on the bus. Perhaps I continue to deliberate about whether to get on the bus, perhaps I do not. Either way, as each moment passes, I am aware, of course, that I am not currently getting on the bus and have not as of yet decided to do so. Eventually, the bus pulls away without me.

This is a prime example of nonintentional omission, which is when one knowingly and voluntarily omits to do something but without intentionally omitting to do it. Because I did not intend not to get on the bus in this example (nor to do anything else incompatible with my getting on the bus then), it is implausible to maintain that my omitting to do so was intentional. And because I was fully aware that I was omitting to (decide to) get on the bus every moment I remained seated, it would seem to be a mistake to say that my omitting to get on the bus was unintentional or accidental. What remains is nonintentional omission. Although I was aware that I was not getting on the bus, I had not settled on not getting on the bus—I did not intend not to do so. And even though I did not decide or intend not to get on the bus, I was nevertheless fully in control of my behavior. I was not subject to threat or coercion of any kind, and my getting on or not getting on the bus remained entirely up to me. Thus, there appear to be no grounds for denying that my omitting to get on the bus in this case was a voluntary omission on my part.^{Footnote 40}

From (10) and (11), it follows that

1. (12) It is an essential feature of indeterministic Frankfurt-style cases that agents in these cases have robust alternatives to deciding on their own to act as they do.

And from (4) and (12), it follows that

1. (13) Indeterministic Frankfurt-style cases are theoretically incapable of serving as genuine counterexamples to PRAP.

So long as it remains open to agents in these cases to knowingly and voluntarily omit to decide on their own to do that which they know to be wrong, they will continue to possess robust alternative possibilities. And since there is no way to eliminate the possibility of agents knowingly and voluntarily omitting to decide on their own to act as they do without undermining the judgment that they are morally responsible for their actions, agents in these cases will continue to possess robust alternative possibilities so long as they are morally responsible for what they do. Thus, either agents are not morally responsible for what they do in these cases or they continue to possess robust alternative possibilities. Either way, these cases will fail to falsify PRAP and so will not be useful in mounting a successful indirect Frankfurt-style attack on PAP.

VI. Conclusion

Toward the end of the 1968 film, The Lion in Winter, there is a scene where Richard, Geoffrey, and John have been locked in the dungeon by their father, King Henry II, and are waiting to be executed. Eventually, upon hearing footsteps coming down the stairs and believing they had finally seen sunlight for the last time, Richard stands to his feet, defiantly declaring, “He’ll get no satisfaction out of me. He isn’t going to see me beg.” “Why, you chivalric fool,” Geoffrey chides, “as if the way one fell down mattered.” “When the fall is all there is,” Richard replies, “it matters.”

Critics of the flicker strategy have argued that the alternatives that remain open to agents in Frankfurt-style examples, including the option of omitting to perform the relevant actions on their own, are unable to do the work set out for them, that they are not the kinds of alternative on which proponents of the ability to do otherwise have traditionally been focused, and that they are too diminutive to matter. Granted, generally believing that it remains up to us not only how we perform the actions we freely perform but whether we perform them at all, this is not the full-fledged freedom with which we are normally concerned and which we commonly take ourselves to possess. Still, when this flicker is all there is—when the ability to knowingly and voluntarily omit to do on one’s own that which one knows to be wrong is all the freedom that remains—it matters.

Indeed, that such flickers matter—and, moreover, that they matter in such a way that they are able to help ground ascriptions of moral responsibility—is, I suggest, the main lesson to be drawn from Frankfurt-style cases. Although many philosophers have been convinced by these cases that we ought to reject entirely the ability to do otherwise as a necessary condition for moral responsibility, this is not something that Frankfurt-style cases are (even in principle) capable of showing. Instead, what they show is that there are certain ways of understanding this condition that ought to be rejected. In particular, proponents of an alternative-possibilities condition for moral responsibility have commonly held that the relevant alternative must be an alternative intentional action (or at least an intentional omission), so that an agent will be morally responsible for deciding to A only if it was possible for her to have intentionally done something other than decide to A (for instance, decide not to A or decide to do something else that is incompatible with A-ing). But just as Locke’s locked room demonstrates that one can be morally responsible for performing an overt (intentional) action even if one could not have avoided performing that overt action, Frankfurt-style cases show us that one can be morally responsible for making a decision even if it was not possible for one to have avoided making that decision. These cases are instructive, not because they show that PAP is false or ought to be rejected, but because they serve to identify the alternative possibilities that can plausibly be thought to be intrinsically relevant to moral responsibility. Their value is in showing us how best to understand this principle, not in demonstrating why we should abandon it.