DOING ONE’S BEST, ALTERNATIVE POSSIBILITIES, AND BLAMEWORTHINESS

SUMMARY : My main aim in this paper is to improve and give further support to a defense of the Principle of Alternative Possibilities (PAP) against Frankfurt cases which I put forward in some previous work. In the present paper I concentrate on a recent and famous Frankfurt case, Derk Pereboom’s “Tax Evasion”. After presenting the essentials of my defense of PAP and applying it to this case, I go on to consider several objections that have been (or might be) raised against it and argue that they don’t succeed. I conclude by pointing out that my criticism of Pereboom’s example suggests a general strategy against other actual or possible Frankfurt cases.


. Introduction
The view that moral responsibility for something one has done requires freedom to choose other available options has found a characteristic expression in the so-called "Principle of Alternative Possibilities" (PAP). Harry Frankfurt's formulation of PAP is the following: "A person is morally responsible for what he has done only if he could have done otherwise" (Frankfurt 1969, p. 829). Frankfurt rejected PAP; his main argument against it rests on so-called "Frankfurt cases" or "Frankfurt-style cases".
The problem that Frankfurt cases raise for PAP defenders may be stated thus. Imagine an agent who, fully on her own and for her own reasons, decides and acts in a morally objectionable way, so that we find her clearly blameworthy. Like most of us on many occasions, she assumes that she had alternative possibilities (APs). However, fully unbeknownst to the agent, the situation includes a certain factor (say, a sort of brain-reading device) that would have intervened and prevented her from making a different decision if it had detected some sign that she was going to make it. The agent shows no such sign and the device remains totally inactive along the whole process. The agent deliberates, decides, and acts fully on her own, assuming, falsely, that she has APs to her decision. If, under these conditions, the agent is morally responsible for her decision and action, then APs are irrelevant to moral responsibility, and PAP is false.
In some previous work, and especially in Moya (2011a), I have defended PAP against Frankfurt-cases. My main aim in the present paper is to improve this defense and to reinforce it by meeting some objections that have been (or might be) raised against it. My understanding of PAP in this and other recent papers is shaped by the following assumptions: first, I take doings to cover not only actions, but also refrainings and omissions; so, I'll consider avoiding or omitting to do something, or refraining from doing it, as a way of "doing otherwise"; second, mental acts, and particularly decisions, are included under the scope of doings; third, I'll restrict PAP to moral blameworthiness, leaving aside praiseworthiness; 1 finally, I'll take PAP to refer only to morally significant ("robust", in John Fischer's terms) APs, roughly, APs whose availability to an agent is explanatorily relevant to the moral responsibility she bears, if any. More precisely, an AP that is available to an agent is robust just in case either aµ the fact that the agent does not choose it partly explains the blameworthiness she bears for what she does, if any; or bµ the fact that the agent chooses it partly explains why she is less blameworthy than she would have been otherwise, or even not at all. 2 The structure of this paper is the following. Sections 2 and 3 contain the essential aspects of my defense of PAP against Frankfurt cases. Section 4 focuses on a recent Frankfurt case designed by Derk Pereboom and shows how the previously exposed defense can be 1 My reason for restricting PAP to cases of negative moral responsibility (blameworthiness) is that I find it dubious that praiseworthiness requires APs. Some authors, such as Wolf (1990), Nelkin (2008) or Moya (2011b) have defended an asymmetrical view of the APs requirement, according to which APs are necessary for being blameworthy, but not for being praiseworthy. At any rate, the practice of dealing with PAP in terms of blameworthiness is rather common in recent literature on free will and moral responsibility.
2 As we will see, the explanation should be rational or justificatory, not merely causal. used to resist its intended conclusion. Sections 5, 6 and 7 are devoted to reinforcing my defense of PAP, and especially my criticism of Pereboom's Frankfurt case, by meeting a number of objections. The paper ends with some concluding remarks (section 8), which suggest how my defense of PAP, which is mainly applied to Pereboom's example, could be extended to cover other actual or possible Frankfurt cases.

. An Example and Some Variations
My defense of PAP rests on a principle that seems to be implicit in our attributions of moral responsibility. As a preliminary step to a statement of this principle, let me present for consideration an example and some variations of it: 3

Drowning Boy 1
David is walking along a relatively quiet beach when suddenly he sees a young boy in the water moving spasmodically, in real danger of drowning. David is a good swimmer and thinks he could save the boy's life by jumping into the sea and swimming towards the boy. However, he is quite a selfish person and, moreover, he is feeling very depressed, for he has lost his job a few days ago. So, instead of jumping himself in the sea and saving the boy, he calls out for help. Unfortunately, nobody is close enough and the boy drowns. Let us assume, plausibly enough, that, had David jumped in the water, he would, in all probability, have succeeded in saving the boy. I hope we would agree that David is culpable for not saving the boy's life. He did something good, namely to call out for help, but this is not enough for precluding him from blame, for there was something much better that he should and could have done, namely to swim towards the boy and save his life. The availability of this AP contributes to explaining why David is blameworthy for not saving the boy: he is so partly because he should, and could, have saved the boy himself but decided not to. This AP is, then, robust, or morally significant, according to our prior characterization of robustness. Additionally, choosing it (and acting accordingly) would have exempted David from the blame he now bears. I'll call APs of this last sort "exempting". So, the AP of swimming and saving the boy's life is for David both robust and exempting. 4 Let us now introduce a variation in the example and call it "Drowning Boy 2". Things are as in Drowning Boy 1 except for the fact that, unbeknownst to David, there is a very powerful current of water flowing between him and the boy. Even if, contrary to fact, David had jumped in the water and tried to save the boy, he would have failed, for the current would have drawn him far away and prevented him from reaching the boy. Again, David calls out for help but the boy dies anyway.
In Drowning Boy 2, saving the boy is not an AP that is available to David, although he falsely believes it is. It seems plausible to hold that, unlike Drowning Boy 1, David is not properly culpable for not saving the boy's life, because the boy would still have died even if he had tried to save his life. However, he is blameworthy, at a minimum, for not jumping in the water and not trying to save the boy's life himself. And even in this scenario David had an exempting AP: in these circumstances, his honest attempt to save the boy by jumping in the water and swimming would have exempted him from the blame he now deserves; and the reason is that this was the best that he could have done in order to save the boy's life. Moreover, this AP was also robust, for its availability to David contributes to explaining the culpability he now bears.
Let us finally consider another variation of the example, which we may call Drowning Boy 3. This variation differs from Drowning Boy 2 in that there is no powerful current between David and the drowning boy. More importantly, it differs from Drowning Boy 1 and 2 in that, instead of being a good swimmer, David cannot swim. David's inability to swim is not due to his lack of interest in learning, but to an insurmountable phobia of deep water. As in the first two versions, David calls out for help and, again unfortunately, the boy drowns because nobody is close enough to save him.
Whereas in Drowning Boy 1 and Drowning Boy 2 calling out for help is not an exempting AP for David, this AP is exempting in Drowning Boy 3, for there is nothing better which David could 4 The robustness of an AP and its exempting character are different properties. They are neither extensionally nor intensionally equivalent. The robustness of an AP has to do with its explanatory virtues concerning the agent's blameworthiness. Its exempting character refers instead to its capacity to preclude the agent's blameworthiness were she to choose it. Now, an AP can be robust and non-exempting, as well as exempting and non-robust. Moreover, whereas the robustness of an AP is a gradual property, its exempting character is all-or-nothing instead.
reasonably have done in order to save the boy's life. Unlike Drowning Boy 2, David could not even have tried to save the boy's life by jumping in the water and swimming, for he was not able to swim. In Drowning Boy 3, David is blameless for the boy's death, though he may justifiably feel regret for the death, which he could not have avoided. The AP of calling out for help was both robust and exempting, and David was precluded from blame in choosing it.
One lesson from this example and its variations is that what constitutes a robust, or even an exempting, AP is a highly contextual matter, and depends not only on what an agent believes she can do in a certain situation, but also on what she can effectively do in that situation. So, calling out for help is an exempting AP in Drowning Boy 3, but not in the other two versions. In the latter, there are some better things that David can do in order to behave in a morally right way. In Drowning Boy 1, he is culpable for not saving the boy, something he could have done and refused to do. In Drowning Boy 2, David is culpable for not trying to save the boy by jumping in the water and swimming, something he could have done but decided not to do.

. A Principle and Two Corollaries
A plausible way of accounting for our judgments about Drowning Boy and its variations is to view them as applications of a general principle that, as I said, seems to underlie our attributions of moral responsibility. I'll call it "DBB" (from "Doing one's best and blameworthiness"). It may be stated as follows: DBB: If someone cannot reasonably 5 do more than she actually does in order to behave in a morally right way, she is not morally blameworthy for not doing more. 6 5 The qualification "reasonably" is intended to avoid exaggerated demands (including supererogatory acts) from the agent in order to comply with morality and be blameless. So, for example, in Drowning Boy 3, David, in spite of being unable to swim, could have jumped in the water and tried to save the boy, with the consequence that he would have died himself. This would not have been a reasonable thing for him to do. 6 In Moya 2011a I proposed a related but different principle, which I called "C" (from "can"). It was as follows: "(C) If someone cannot reasonably do more than she actually does in order to fulfil her moral duties, she is not morally obliged to do more, and so she is not morally blameworthy for not doing more". (C) is the conjunction of two conditionals. The first ("If someone cannot . . . she is not morally obliged. . . ") is the contraposition of the Kantian moral principle that "Ought" implies "Can". The second (if someone "is not morally obliged . . . she is not morally blameworthy. . . ") expresses the plausible view that being blameworthy for doing (or I find DBB eminently plausible. It certainly seems unfair, an offence against our sense of justice, to morally blame someone who has made all her best efforts to act in a morally right way. 7 Now, both in Drowning Boy 1 and in Drowning Boy 2, we immediately realize that, from a moral point of view, David could certainly have done better than he did; this "better", however, is different in each case; in the former, he could have saved the boy's life; in the latter, he could not, but he could at least, and at most, have tried to do so. Now, we see that, if David had done either (in the corresponding context), he would have satisfied the antecedent of DBB, and he would not have been blameworthy.
There are two important consequences to be drawn from a reflection on these examples.
The first has already been suggested: whether an AP is robust or exempting is a highly contextual matter and depends on the particular circumstances of the case. But we can now make this suggestion more precise, for there is a systematic relation between the two parameters. To see this, consider which APs are exempting (and also robust) for David in the three versions of the example: -Drowning Boy 1: Jumping in the water, swimming towards the boy and saving his life.
-Drowning Boy 2: Jumping in the water, swimming towards the boy and honestly trying to save his life.
-Drowning Boy 3: Calling out for help in a sincere attempt to find someone who could save the boy's life.
We can now spell out the relation alluded to. As we can see, the standards for some action to be an exempting AP rise or fall as a partial function of both the agent's subjective epistemic perspective and the APs effectively available to her. We require more of David omitting) something requires being morally obliged not to do (or to do) it. Now, though I think that both propositions are true, not everybody accepts them. And those who reject either will also be led to reject (C). This is the reason why I have replaced (C) with DBB, which connects directly the agent's inability to do morally better and his blamelessness, and nonetheless remains eminently plausible. 7 I have deliberately used the rather vague expression "to behave in a morally right way", which permits different readings, in order for DBB to be acceptable from different perspectives on what morality and moral behavior consist in. Moreover, DBB is not overly demanding: it does not logically follow from it that, if someone can reasonably do more than she does to behave in a morally right way, then she is ipso facto blameworthy, even if it may give rise to a presumption to that effect. to exempt him from blame in Drowning Boy 1 than in Drowning Boy 2. In both cases, he believes he can save the boy personally, but in 2 he cannot, which makes the standards for exemption descend from saving the boy personally to merely (honestly) trying to save him. And we require less of David in Drowning Boy 3 than in Drowning Boy 2, for in the former trying has no point, given that David cannot swim; the standards for exemption fall again from trying to save the boy himself to calling out for help in an attempt for someone else to save him. In more colloquial terms, the more (the less) an agent can do to behave decently in a certain situation, the more (the less) he is required to do to be blameless.
A second consequence of our reflection on the examples is the statement of two additional principles. They can be considered as corollaries or maybe plausible extensions of DBB. The former establishes the conditions for an AP to be exempting in cases in which an agent is ignorant of her inability to do something, as happens in Drowning Boy 2. We'll call it "NBA-ign" (from "next best action under ignorance of inability"): NBA-ign: If, unbeknownst to her, an agent cannot do something A such that, if she did it, she would behave in a morally right way and be precluded from blame (and she knows that she would), then, in order to be so precluded, she should perform the next best action that reasonably was in her power to perform in order to behave in a morally right way, where "the next best action" may be characterized as trying or attempting to A, or at least taking some steps directed to A. (cf. Moya 2011a, p. 15) An agent who behaves as NBA-ign counsels satisfies the antecedent of DBB and is then blameless. This would have been the case for David in Drowning Boy 2 had he honestly tried to help the boy by jumping in the water and swimming.
The second principle applies to cases where the agent knows, or justifiably believes, that she cannot do something, as happens in Drowning Boy 3. We'll call it then "NBA-kn" (from "next best action under knowledge of inability"): NBA-kn: If an agent knows (or justifiably believes) that she cannot do something A such that, if she did it, she would behave in a morally right way and be precluded from blame, then, in order to be so precluded, she should perform the next best action that reasonably is in her power to perform in order to behave in a morally right way. (Cf. Moya 2011a, p. 16) Here we cannot characterize the next best action as trying to A, for trying to A seems to require the belief that A-ing is not beyond one's reach, which is not the case under the knowledge of (or justified belief in) inability.

. Back to Frankfurt Cases
We can come back to Frankfurt cases and see whether, and if so how, the preceding considerations have a bearing on them.
The challenge that Frankfurt cases pose to PAP defenders may be stated as follows. In such cases, the agent is fully ignorant that, owing to the presence of a potentially intervening factor, she lacks certain APs. Furthermore, she deliberates, decides and acts exactly as she would have done in the absence of such a factor, which remains causally idle. Now, how could this factor, and the corresponding lack of APs, have any bearing on the moral responsibility the agent may bear for her decision and action?
In the light of the preceding considerations, we can now try to provide a response to this challenge. Even if the factor never intervenes in the process of the agent's deliberation, decision and action, its mere presence is not without consequences, for, according to the preceding arguments, in reducing drastically the APs available to the agent, it lowers correspondingly the standards for an act to be a robust and exempting AP in that situation. So, a tiny event, which in normal circumstances, without the factor in place, would not exempt the agent from moral responsibility, would actually do so, according to principles DBB and NBA-ign, if performing that act was the most the agent could reasonably have done, in the circumstances, to comply with morality. Tiny, subtle acts can be robust and exempting APs in the rarified contexts of Frankfurt cases, so that, even if the agent is blameworthy, PAP need not be refuted.
Let us now apply these abstract considerations to a recent and sophisticated Frankfurt case: Derk Pereboom's now famous "Tax Evasion". I shall concentrate on version 2 of this example. It's as follows: Tax Evasion (2): Joe is considering claiming a tax deduction for the registration fee that he paid when he bought a house. He knows that claiming the deduction is illegal, but that he probably won't be caught, and that, if he were, he could convincingly plead ignorance. Suppose he has a strong but not always overriding desire to advance his selfinterest regardless of its cost to others and even if it involves illegal activity. In addition, the only way that in this situation he could fail to choose to evade taxes is for moral reasons, of which he is aware. He could not, for example, [fail to] choose to evade taxes for no reason or simply on a whim. Moreover, it is causally necessary for his failing to choose to evade taxes in this situation that he attain a certain level of attentiveness to moral reasons. Joe can secure this level of attentiveness voluntarily. However, his attaining this level of attentiveness is not causally sufficient for his failing to choose to evade taxes. If he were to attain this level of attentiveness, he could, exercising his libertarian free will, either choose to evade taxes or refrain from so choosing (without the intervener's device in place). However, to ensure that he will choose to evade taxes, a neuroscientist has, unbeknownst to Joe, implanted a device in his brain, which, were it to sense the requisite level of attentiveness, would electronically stimulate the right neural centers so as to inevitably result in his making this choice. As it happens, Joe does not attain this level of attentiveness to his moral reasons, and he chooses to evade taxes on his own, while the device remains idle. (Pereboom 2009, p. 113) In contrast with the simplicity of Frankfurt's initial example, Pereboom's Frankfurt case is quite complex and full of details and specifications. Behind this complexity there is the long and subtle discussion that has taken place since Frankfurt's (1969) seminal article came out, as well as the important objections that have been raised in defense of PAP against those cases. In particular, Pereboom is anxious to avoid a trait that most authors have considered as question-begging against incompatibilists, namely the introduction, in the actual sequence of these cases, of a prior sign or event, beyond the agent's control, that causally necessitates the agent's decision. To avoid this problem, Pereboom assumes that the process is indeterministic; it contains at least an AP that is under the agent's voluntary control, namely the agent's reaching a certain level of attentiveness to his moral reasons against evading taxes. This AP per se leaves it open whether Joe will end up deciding to evade taxes or failing to decide this. But the decision to evade taxes is ensured by a counterfactual factor that would eventually cause the decision were it to sense the requisite level of attentiveness; this factor, however, never causes Joe's decision, which he makes on his own and for his own (selfish) reasons. It seems, then, that Joe is blameworthy for his decision to evade taxes, even if he has no robust AP to making it, and PAP is refuted by the example At this point in the argument, I think that the response that I favor suggests itself rather naturally. Though it is true that the device never intervenes, its mere presence introduces an important difference be-tween this case and a normal case, with no lurking intervener. The difference concerns what the agent would have to do in order to be blameless in either situation. Unbeknownst to Joe, with the device in place, and given the structure of the example, he cannot do more, in order to behave in a morally right way, than paying careful attention to moral reasons against evading taxes. He cannot decide not to evade taxes, for, in order to do this, and given his predominantly selfish character, he should first attend to moral reasons for doing so, but then the device would fire and take over Joe's deliberation and decision making. However, if paying careful attention to moral reasons is all that Joe can reasonably do to behave in a morally decent way in the circumstances of Pereboom's example, then, by DBB and NBAign, if he had done it, he would not have been blameworthy for not doing more, e.g. for not avoiding the decision to evade taxes. Freely becoming attentive to moral reasons is then, against appearances, an exempting alternative in Pereboom's Frankfurt case. This alternative is also robust, for the fact that Joe did not choose it, though it was available, contributes to explaining his blameworthiness for his decision. Hence, even if we judge that Joe is blameworthy for his decision to evade taxes, we can see that he had a robust and exempting AP, and PAP has not been refuted after all. Of course, if the device had not been present, paying careful attention to moral reasons would not have exempted Joe from blame; in that context, only deciding against evading taxes and doing so would have had that effect. Again, since, in the context of Pereboom's example, Joe could not do so, the standards for an AP to be exempting fall, with the effect that the next best action he can perform, which in this case is only paying careful attention to the moral reasons he is aware of, becomes an exempting AP. Since this AP is also explanatorily relevant, and so robust, PAP remains untouched by Pereboom's example. 8 However, before we can take this result as established, we have to confront important objections to the preceding line of argument. 8 Concerning more specifically Pereboom's example, an important line of criticism was initiated by Ginet (2002) and has been developed by Palmer (2011) andFranklin (2011). It is usually called "the timing objection". I think the objection is sharp and may well be finally successful, though Pereboom (2012, pp. 307-313) has offered a response to it. A recent criticism of the objection can be found in Hunt and Shabo 2013. My own proposal does not intend to compete with it, but rather to provide an alternative strategy to keep PAP safe from Pereboom's and other Frankfurt cases. I think my strategy is both more intuitive and more general than the timing objection. But this is for the reader to judge.

. Meeting Objections
An important problem for my view arises if we ask what exactly Joe is blameworthy for. My own judgment (and of course Pereboom's) is that Joe is blameworthy, not only for his low attentiveness to moral reasons against evading taxes, but also for deciding to evade taxes (and for doing so). Now, in order for my proposal to succeed, paying careful attention to moral reasons has to be a robust and exempting AP not only to not paying sufficient attention to moral reasons, but also to deciding to evade taxes. But whereas the former seems clearly right, the latter is problematic. Pereboom has seen this problem. 9 He writes: The Frankfurt defender can agree that Joe is blameworthy for not becoming more attentive to the moral reasons, and that for this he does have a robust, exempting alternative possibility. But it's intuitive that Joe is also blameworthy for deciding to evade taxes, and for this, at least prima facie, he has no robust alternative. (Pereboom 2012, p. 304) The objection can now be completed as follows. If Joe is culpable for deciding to evade taxes but has no robust and exempting AP to that, it seems that Tax Evasion (2) is a counterexample to PAP after all and my defense of this principle fails. Let me discuss this objection.
Why does Pereboom think that becoming attentive to moral reasons is not a robust AP with respect to Joe's decision? After the preceding text, Pereboom goes on: What is the motivation for thinking that becoming more attentive to the moral reasons now becomes a robust alternative to deciding to evade taxes? Agreed: it is the next best action available to him. But it is not per se exempting for deciding to evade taxes, since, without the device in place, even if he did become more attentive he could decide to evade taxes, and, we might suppose, he even would be likely so to decide. (Pereboom 2012, p. 304) So, one reason Pereboom has for denying the robustness of becoming attentive to moral reasons with respect to Joe's decision to evade taxes is that, without the device in place, Joe could end up making this decision after all. I will deal with this consideration below. There is, however, a previous problem with this way of arguing. I fully agree that this AP would not exempt the agent without the device in place. For suppose that Joe becomes attentive enough to moral reasons, but that he finally decides to evade taxes nonetheless. He could not dream of getting off the moral hook for this decision by claiming that he was attentive to moral reasons against making it. (Let me add that, far from precluding him from moral responsibility, this claim might probably aggravate his blameworthiness, for it would strongly suggest that Joe made that decision in full awareness that it was immoral). In that situation, without the device in place, only deciding not to evade taxes would exempt the agent from blame for deciding to evade taxes. But in arguing for the non-robust character of becoming attentive in this way, Pereboom is subtly denying, without justification, what I have been at pains to establish through Drowning Boy and its variations, namely that the robust and exempting character of an AP is a highly contextual question, strongly dependent on what an agent can effectively do in a particular situation. Hence, a certain way of acting can be robust and exempting in one context and not in another if what the agent can do in one of them differs from what he can do in the other, as happens, e.g., with calling out for help in the different versions of Drowning Boy. If this is correct, as it seems to be, Pereboom is not legitimated to infer that becoming attentive to moral reasons is not a robust and exempting AP to deciding to evade taxes in Tax Evasion (2) from the fact that it is not exempting in another context, with no device in place, which differs crucially from the first in what the agent can effectively do.
As for the substance of Pereboom's consideration against the robust character of becoming attentive to moral reasons with respect to deciding to evade taxes, namely that, after becoming so attentive, and without the device in place, Joe could decide to evade taxes, I think it does not succeed. I hope we would agree that, in Drowning Boy 1, David's honest (and, by hypothesis, successful) personal attempt to save the boy would have been a robust AP to his free refusal to attempt it. However, on Pereboom's argument, this AP would not have been robust, for, after swimming for a while toward the boy, it was open to David to decide to stop trying and let the boy die. Worse still, if the argument were correct, Joe's deciding not to evade taxes would not even be a robust AP to deciding to evade taxes, for, after deciding not to evade taxes, Joe could change his mind and decide to evade taxes after all.
Central to my defense of PAP, as we have seen, is a contextual conception of the robust and exempting character of an AP. But Pereboom contends that this conception is fueled by an allegiance to PAP itself and hence begs the question against the Frankfurt theorist: The feature of the context that Moya claims to alter the standard from the usual one is the unavailability of what would under normal circumstances be an exempting alternative. However, this claim is itself fueled by the robust alternative possibilities requirement, which is what's at issue. If Moya's proposal for altering the usual standard could be justified by considerations independent of this requirement, then perhaps it should sway the Frankfurt defender. But the justification Moya sets out, while not implausible, is not independent of the requirement. (Pereboom 2012, pp. 304-305) In response, I can say the following. First, I do not think there is a "usual" standard for exempting alternatives which can be established in "normal" circumstances. If we consider Drowning boy, none of its three variations is more "normal" than the others. The three situations are equally possible ordinary situations. They differ in what Joe can effectively do in each of them and this determines different kinds of actions as robust and exempting APs. And none of these APs is more usual than the rest. I agree that Frankfurt cases are not normal, ordinary situations; normally, I hope, there is no lurking device prone to intervene if we, e.g., consider moral reasons for acting in a certain way. But a central effect of the device is to reduce what the agent can effectively do, and there is no reason why this reduction should have different effects on the exempting APs in a Frankfurt case than it has in normal, ordinary situations represented, e.g., by the different versions of Drowning Boy. Second, and more importantly, I think that the variations in the standards for robust and exempting APs (and principles such as DBB, NBAign and NBA-kn, which regiment such variations) are not fueled by an allegiance to the APs requirement. They are inspired instead by intuitions about fairness and unfairness in ascriptions of blame raised by reflection on ordinary situations. 10 The underlying idea is that blaming someone who has made his best efforts to behave in a morally decent way, though not uncommon, is unjustified and unfair.
Such an agent does not deserve blame and is not blameworthy. So, I think that my proposal supports PAP, but does not rest on an implicit acceptance of it.

. Robust Alternative Pathways
The preceding considerations are mainly negative. They are intended to show that Pereboom's arguments against my defense of PAP are not successful. However, it would be important to argue in more positive terms for the claim that becoming attentive enough to moral reasons is for Joe a robust AP, not only to not becoming so attentive, but also to his decision to evade taxes.
The problem, to be precise, concerns the robust character of the AP, for there is no doubt that the AP is exempting, not only concerning the lack of attentiveness to moral reasons, but also the decision to evade taxes. For if Joe had become attentive enough to moral reasons, the device would have been activated and caused Joe's decision to evade taxes; hence this decision would not have been free and Joe would not have been culpable for it. 11 Becoming sufficiently attentive to moral reasons in Tax Evasion (2) is robust with respect to Joe's final decision just in case either aµ Joe's becoming so attentive would have rationally explained (for causally it certainly would) his blamelessness for that decision, or bµ Joe's not becoming so attentive (as is the case in the example) rationally explains his blameworthiness for that decision. Pereboom denies a and, on this basis, presumably b. His denial of a rests on his view that, in order for an AP to be robust, the agent has to justifiably believe (or at least have some cognitive sensitivity to the fact) that, by choosing that AP, she would be (or at least she would likely be) exempted from blame. 12 And Joe, according to Pereboom, does not satisfy this condition, for he "is not aware that due to 11 Some thinkers would throw serious doubts about whether what happens in the alternative sequence deserves to be called a decision or even an action (e.g. Alvarez 2009, Steward 2009); I think there are good reasons for such doubts, but I'll leave this question open. 12 This epistemic condition is intended to exclude cases where the agent is exempted from blame by sheer luck or coincidence (cf. Pereboom 2009, p. 112;2001, p. 26;2003, p. 194). In addition, according to Pereboom, a robust AP has also to be exempting. His reason for such a strong conception of robustness is that, according to him, the intuition behind PAP is "the 'off the hook' intuition: to be blameworthy for an action, the agent must have been able to do something that would have precluded this blameworthiness" (Pereboom 2009, p. 114). However, this position is by no means uncontroversial. PAP is likely to express rather an intuition of control over one's behavior.  (2012, p. 304). Given this, he will hold that Joe's blamelessness for his decision, as a result of his attentiveness to moral reasons, would be a matter of sheer luck and would not have a rational explanation, but only a purely causal one. Instead, I accept b and, on this basis, within certain limits, a as well. In my view, the fact that Joe did not become attentive enough to moral reasons is part of a rational and justificatory explanation of his blameworthiness for his decision. And his blamelessness, had he become attentive, could be rationally explained too, even if an element of luck was also involved. Let me then argue for this view.
Consider, first, that an important feature of Pereboom's Tax Evasion (2) is that Joe is primarily blameworthy for his decision to evade taxes, and only secondarily for his overt act of evading them. 13 Instead, in specifying what David is responsible for in the Drowning Boy series, I have pointed mainly to actions that he performed or refused to perform (such as calling out for help, jumping in the sea, saving the boy or attempting to save him). So, in order to come closer to Pereboom's example, let me design a new Drowning Boy example, where David will be primarily blameworthy for his decision not to help the boy. Let me call this example Drowning Boy D (where "D" stands for "decision"). It would be as follows.
Drowning Boy D: Take as our starting point Drowning Boy 1. In the new example David also decides not to jump in and help the boy himself, and he is blameworthy for this decision. Now, however, he could not have decided to jump in and help the boy himself for the following reason. Owing to some traumatic experience in his early childhood, which he has forgotten, if David had taken the possibility of jumping in the sea and saving the boy seriously, by attending to the important moral reasons in favor of this option, he would have suffered (because of some special features of the particular situation) a serious attack of anxiety and deep fear, so that he could not even have decided to jump in and help the boy.
I would agree with Pereboom that, in a case like this, which has important similarities with Tax Evasion (2), David is blameworthy for deciding not to jump in and help the boy himself, even if he 13 At least since the formulation of the so-called "dilemma defense" of PAP, it is common in the design of Frankfurt cases, and the debate about them, to focus on decisions, rather than overt actions, as the primary objects of moral responsibility. There are good reasons for doing this, though I will not discuss them here.
could not have decided to jump. 14 But I disagree for what concerns the robustness of the APs open to David. I think it is highly plausible to hold that David's AP of attending to the strong moral reasons in favor of jumping in and taking the possibility of doing so seriously was robust concerning his decision not to help the boy. As happens with Tax Evasion (2), there is no doubt that the AP is exempting, for this alternative pathway, had David taken it, would have ended in a serious anxiety attack, so that he would not have made the decision that he actually did and would not have been culpable for making it. It is true that, as happens in Tax Evasion (2) too, there was an element of luck involved in this exemption, for David was not aware that he would suffer an anxiety attack if he were to start this alternative pathway. It would seem, then, that David's blamelessness would not have had a rational explanation at all. I shall come back to this issue below. Anyway, this in itself does not show that the AP is not robust, for the question of its robustness, we may recall, is also the question whether the AP is relevant to a rational, justificatory explanation of David's blameworthiness for his actual decision. And I think it is. Let me argue for that.
Why is David blameworthy for his decision? I gladly agree with the Frankfurt theorist that an important reason for this blameworthiness is that David deliberated and decided on his own and for his own reasons, without any coercion, and that he made that decision in full awareness that it was morally wrong, for it was in clear conflict with the moral reasons for helping the boy that David was aware of. However, against the Frankfurt theorist, I contend that, in addition to those factors, the fact that the indicated AP was available to David is also part of a true rational explanation of his blameworthiness. Our judgment about David's culpability for his actual decision receives also a strong support from the intuition that, from a moral point of view, he fell very short of doing the best he could have done. In the situation at hand, David was able to take steps toward the morally right decision to help the boy himself, by paying serious attention to moral reasons, and the fact that he freely avoided taking any such steps partly explains, in a rational way, why he is blameworthy for his decision. David's AP, then, is clearly robust. Now, there seems to be no justification not to apply a similar reasoning and conclusion to Joe's decision to evade taxes in Tax 14 In the same way as, in Drowning Boy 2, David is not properly culpable for not saving the boy because he could not have saved him, I agree that David is not culpable for not deciding to help the boy himself, either, for he could not have made that decision. But, as stated, he is culpable for his actual decision not to help.
Evasion (2). 15 In the same way as, in Drowning Boy D, David was blameworthy for deciding not to help the boy partly because he did not try or take any steps toward deciding to help him, in Tax Evasion (2) Joe is blameworthy for deciding to evade taxes partly because he did not even try or take any steps towards deciding not to evade them. At no moment in the process that ends with his decision to evade taxes does he show any sign of taking the possibility of not evading taxes seriously, although he had evidence (namely the moral reasons he was aware of) in favor of that possibility. In freely disregarding this evidence, by not becoming attentive enough to those reasons, he freely prevented a necessary condition for deciding not to evade taxes from being satisfied. Hence he is blameworthy for deciding to evade taxes partly because he did not pay enough attention to his moral reasons against such an evasion.
Pereboom would insist that Joe was "not aware that due to the device becoming more attentive [ . . . ] would [ . . . ] render him blameless for this decision [to evade taxes]" (2012, p. 304). My response, and my reason for holding that David's blamelessness, had he chosen this AP, would not have been a merely lucky result, but also a rationally explicable outcome (the presence of luck notwithstanding), rests on taking a slightly different perspective on the case: Joe could reasonably foresee (or at least he had some cognitive sensitivity to the fact) that if he did not become attentive enough to the moral reasons he was aware of, he would end up (or would likely end up) deciding to evade taxes. Nevertheless, he freely omitted to pay due attention to those moral reasons, and this is partly why he is blameworthy for his decision. Even if an element of luck was also present, being blameless for something for which he is actually blameworthy, namely deciding to evade taxes, was in Joe's hands in a foreseeable and non-accidental, non-chancy way, namely by his honestly taking steps toward the decision not to evade taxes, even if, unbeknownst to him, the decision itself was not within his reach.
My strategy for defending PAP against Pereboom's example has some similarity to Widerkers's (2006, p. 173) response to it in terms of the concept of derivative responsibility. According to Widerker, Joe's blameworthiness for his decision to evade taxes, which he could not avoid, was derivative from his prior blameworthiness for freely failing to attend strongly enough to moral reasons, a failure that he could have avoided. But, as Widerker points out, PAP applies only to cases of direct responsibility and is immune to cases of derivative responsibility. However, I think that my strategy is preferable in that it does not essentially use the notion of derivative responsibility, whose applicability to Pereboom's example is not fully clear, as Pereboom himself (2009, p. 116;2012, pp. 305-307) has argued in his response to Widerker.
The strategy I have developed bears also some similarity to Margery Naylor's response to Frankfurt's initial paper. According to Naylor (1984), 16 in Frankfurt's original example Jones had APs, namely to do A on his own or not on his own (as a result of the intervention). Naylor's response was criticized by several authors, such as Kane (1996, p. 41) andFischer (1994, pp. 140 ff.). Fischer raised doubts about the robustness of the AP Naylor points to in Frankfurt's example, namely Jones's doing A not on his own (non-freely, in our terms). However, I think that my preceding considerations have plausibly met this objection in showing how in particular contexts, which include Pereboom's example, 17 doing or omitting something non-freely, or against one's will, can be a robust AP to doing or omitting it (freely).
As an additional and important remark, which is sometimes forgotten in the discussion on these topics, let me insist that a defense of PAP doesn't require that having robust APs available be the only or even the main explanation of the agent's culpability. According to PAP, APs are necessary for moral responsibility; this is fully consistent with the existence of other necessary conditions for moral responsibility, without which an agent would not be blameworthy. What is required in order to defend PAP is only that the availability of morally relevant (robust) alternatives be a non-dispensable part of a true rational explanation of the agent's moral responsibility. So, as I did in dealing with Drowning Boy D, I also suggest, concerning Tax Evasion (2), that what explains our judgment that Joe is blameworthy for deciding to evade taxes is both the (libertarianism-friendly) consideration that, from a moral point of view, he could have done reasonably better and the (compatibilism-friendly) consideration that he deliberated and decided on his own, with no coercion, etcetera.

. Further Objections: Modifying the Example
Someone could agree with the preceding considerations and accept that becoming attentive enough to moral reasons is a robust, morally relevant and exempting AP for Joe in Tax Evasion (2), but retort that the example can be redesigned in such a way that a necessary condition for the agent to have access to this or any other morally relevant alternative, and also the sign that would activate the device, is a mental act that is clearly morally irrelevant and non-robust. 18 Suppose, for example, that Joe is so constituted that, in the situation at hand, he could become attentive enough to moral reasons only if he were to imagine, say, a beautiful sunset on a beach. We may plausibly assume that Joe is fully unaware of the causal necessity relationship between imagining the sunset and attending to moral reasons. 19 On this version of the example, Joe is unable to attend seriously to moral reasons against that decision unless he previously imagines the sunset. And, as this imagining is the sign for the device's activation, he does not have access to the (robust) AP of attending seriously to moral reasons. But here it could not be said that imagining the sunset is something morally better (a "next best action") that Joe should have done but did not. Even if this AP is exempting, it is clearly morally irrelevant and non-robust. It would be wrong to say that Joe is blameworthy for deciding to evade taxes, even partly, because he did not imagine the sunset. Nevertheless, the objector could go on, Joe is blameworthy for this decision because he made it on his own and for his own selfish reasons, while having no robust AP available. PAP would then be false.
I think the objection can be met in more than one way. One way is to doubt that, on this version of Pereboom's example, Joe's capacity for practical reasoning, at least in those circumstances, is sound enough for him to be considered as a morally responsible agent. We expect a normal agent, with a normal degree of competence in practical and moral reasoning, to be directly sensitive and responsive to moral reasons against certain morally objectionable ways of acting, such as evading taxes, with no need of having a chancy, fortuitous thought, fully unconnected, from a semantic and rational point of view, with those moral reasons. On the version of the example we are considering, then, the agent's ability to respond to moral reasons is dubious, and with it his moral responsibility. PAP remains safe again.
A related response is to hold that, on this version of the example, Joe does not have enough control over his process of practical deliberation, which casts doubts over his moral responsibility for the result of this process, i.e. his decision to evade taxes. Given that Joe is unaware of the necessity relationship between imagining the sunset and his due attention to moral reasons, it could not reasonably be said that he should imagine the sunset; the imagining, then, can be assimilated to an involuntary mental event, a thought that simply might cross his mind. Without this chancy, lucky event Joe will be unable to attend seriously to moral reasons and to decide in accordance with them. But then Joe's control over his moral reasoning and decisionmaking is seriously diminished and the judgment about his moral responsibility becomes compromised, which leaves PAP untouched.
Another version of the example that can raise problems for my proposal has been designed by Pereboom himself (2009, p. 117;cf. 2011, p. 414). This is Tax Evasion (3). It is similar to Tax Evasion (2) except that now it is causally necessary for Joe's failing to decide to evade taxes in this situation "that he imagine, with a certain degree of vividness, being at least fairly severely punished for doing so, a mental state he can produce voluntarily" (Pereboom 2009, p. 117;2011, p. 414). It is important, however, that, in a posterior modification, Pereboom adds that, if Joe's moral reasons were stronger, e.g., if the amount to be evaded were substantially higher, Joe could decide not to evade taxes for moral reasons alone, with no need to imagine the punishment. This addition is relevant in order to avoid the plausible objection that Joe is not responsive to moral reasons, which might threaten his status as a morally responsible agent. 20 Pereboom has initially designed this version as an attempt to meet the objection of Widerker's to Tax Evasion (2), which we have already alluded to, according to which Joe's responsibility for deciding to evade taxes is only derivative and hence does not threaten PAP. However, Tax Evasion (3) can also be seen as raising an objection to my own proposal, for it seems that in it, and unlike the attentiveness to moral reasons in Tax Evasion (2), imagining being severely punished is not a "next best action", something morally better that it was reasonable for Joe to do in the circumstances, at least if we assume that Joe was unaware of the causal necessity relation between this imagining and the decision against evading taxes. 21 So it is not reasonable to hold that Joe is blameworthy for his decision to evade taxes (at least partly) because he didn't imagine the punishment. Unlike paying attention to moral reasons, imagining the punishment does not seem to be a robust, in the sense of morally significant and explanatory AP for Joe's blameworthiness. Even if following it would exempt him from blame, this would have happened by luck. In other words, there doesn't seem to be a next best action that Joe could have performed, and which could be pointed at as a means to save PAP.
Here, too, there are several ways in which this objection can be met. 22 To start with, let us accept that, as Pereboom stipulates, imagining the punishment is in fact causally necessary for Joe's failing to decide to evade taxes. Still, it does not follow from this that Joe did not have a morally better AP available. Since he was aware of, and sensitive to, moral reasons that counselled against evading taxes, we may safely assume that Joe could have made a greater effort to pay more attention to those moral reasons and to act in accordance with them. Given the stipulation, this effort, no matter how strenuous, would have been powerless unless it was accompanied by the indicated imagining. However, had he made this effort, Joe could not have done reasonably better in order to behave in a morally right way, and so by DBB and NBA-ign, he would not have been blameworthy.
As a second and more important rejoinder, I think that the stipulation about causal necessity is very hard to reconcile with Pereboom's depiction of Joe as morally reasons-responsive. As I pointed out, this feature of Joe is important to dispel possible doubts about his status as a morally responsible agent. So, according to Pereboom, Joe has to be conceived as reasons-responsive: he could decide not to evade taxes without imagining being severely punished if he had stronger moral reasons for that decision than he now has. But if this is so, then the stipulation about causal necessity becomes problematic, for, if Joe can respond to stronger moral reasons and decide against evading taxes without imagining the punishment, it is very hard to accept that it is literally causally impossible for him to make that morally right decision on the basis of his actual moral reasons alone, which he is sensitive to. It can be very hard and effort-demanding to make moral reasons prevail over non-moral ones in one's actions; but being very hard is one thing and being strictly causally impossible is another. It is one thing to say that, given his selfish character, it would be very difficult for Joe to make his moral reasons override his self-interest, and it is quite another thing to say that it is causally impossible for him to do so. Given Joe's awareness, sensitivity, and responsiveness to moral reasons, simply stipulating that it is causally impossible for him to make the right decision for moral reasons alone is not enough to make the example look psychologically credible and convincing enough to refute PAP. Moreover, if Joe had taken his moral reasons seriously and made the effort to decide against evading taxes, he could have succeeded, for the device was only sensitive to his imagining being punished, but not either to a careful attention to moral reasons or to an effort to decide for those reasons.

. Concluding Remarks
Though in this paper I have dealt only with one recent Frankfurt case and a few variations of it, the defense of PAP which I have recommended here suggests a general strategy to defend this principle against other possible or actually proposed Frankfurt cases. In spite of differences, these cases must have two traits in common, namely that the agent has no robust, morally significant APs to what she decides or does and that she is nonetheless morally responsible for deciding or doing it. However, in the light of such principles as DBB and NBA-ign, designing cases with these two features can become a very difficult task indeed. If deterministic assumptions are avoided, as it is advisable to do in order to avoid the possibility of begging the question against incompatibilists, then there will be some APs left, which, for the case to be convincing, must be non-robust. But even if these, in ordinary situations, with no lurking device, might not be exempting, or even robust, some of them will easily become so in the context of Frankfurt cases, provided that there is nothing better which the agent can reasonably do in order to behave in a morally right way. In this case, even if we judge that the agent is blameworthy, we also see that she has robust and exempting APs after all, and PAP will remain unscathed. Moreover, as we have seen in a variation of Tax Evasion (2), if the morally right decision is depicted as depending on a morally and explanatorily irrelevant AP, rationally unconnected with that decision, then the agent's control over her moral reasoning appears as seriously diminished and the intuition of her blameworthiness is seriously threatened.
So, for all I know, PAP may well be true. 23