Excuses and Alternatives

Abstract

A version of the principle of alternate possibilities (PAP) claims that one is only blameworthy for actions which one was able to avoid. Much of the discussion about PAP concerns Frankfurt’s counterexamples to it. After fifty years of refined debates, progress might seem hopeless. Yet, we can make headway by asking: “what’s our reason for believing PAP?” The best answer is this: lacking eligible alternatives—alternatives whose cost is not too high to reasonably opt for—is a good excuse. Yet, this principle is subject to straightforward counterexamples, unless it is given an epistemicised reading. And in this latter case, it does not support PAP. So, PAP is unsupported, at least for now.

1. Introduction

Several philosophers still endorse a version of the principle of alternate possibilities (PAP), according to which one is only blameworthy for actions one was able to avoid—that is, actions to which one had alternatives. Why endorse this principle? Because, presumably, lacking alternatives is a good excuse; and it is not fitting to blame those who have good excuses (at least full excuses, to which I restrict my attention in this paper). More precisely, it is lacking an eligible alternative—one whose cost is not too high to reasonably opt for—that is a good excuse. This is reflected in our quotidian interactions: “Don’t be harsh, I had no choice but to stay late at work,” “Sorry I did not call you, but I really could not” are very common phrases. ¹ This common practice seems to presuppose PAP, which might give confidence to the PAP defender.

I want to argue that this perhaps banal explanation is important. It has a crucial bearing on whether philosophers have successfully defended PAP. In essence, I contend that an adequate defence of PAP should not only defeat Harry Frankfurt’s putative counterexamples to PAP (1969) but should also secure a justification for PAP (perhaps by the above considerations).

Instead, the debate about PAP has been more narrowly focussed on Frankfurt cases themselves. This raised intricate metaphysical controversies regarding the nature of alternatives, desires, omissions, and actions. However interesting, these controversies can be bypassed by directly asking an ethical question: Is lacking eligible alternatives a good excuse? My answer is negative, which is bad news for the PAP defender.

Here is the precise argument I shall defend:

(1) Our (good) reason for believing PAP is, if anything, that lacking eligible alternatives is a good excuse.

(2) Yet, on close examination, lacking eligible alternatives is not a good excuse, unless the eligibility of an alternative is relativised to the agent’s epistemic situation.

(3) And, in this latter case, the principle that lacking eligible alternatives is a good excuse provides no support for PAP and thus is not a good reason for believing it.

Therefore, we have no good reason for believing PAP.

This kind of strategy is indirect since it does not attempt to disprove PAP but rather to undermine its motivation. And it is bypassing since it avoids metaphysical controversies about putative counterexamples to PAP. It has but rare and unsuccessful cousins in the gargantuan literature on PAP. ²

Section 2 argues for the first premise. Section 3 and section 4 present the controversies that my argument bypasses and an unsuccessful indirect strategy. Section 5 argues for the second premise, by drawing on this unsuccesful attempt. Section 6 argues for the third premise. Section 7 explores some of the few options the PAP defender is left with, including the acceptance of a new form of moral luck.

One word of conceptual caution is in order regarding moral responsibility and blameworthiness. First, I want to limit my discussion to cases of wrongdoing. Several philosophers (see e.g., Wolf 1993) have given us good reason to think that responsibility for rightdoing has different requirements. ³ Second, I want to conduct most of my discussion in terms of blameworthiness, rather than of responsibility for wrongdoing. What’s the difference? Well, we do much more than blame those that we recognise as responsible wrongdoers. Amongst other things, we demand explanations, compensations, apologies, and amends from them; we also withdraw our trust or good will from them; sometimes, we resent, exclude, or punish them. So, claims about responsibility for wrongdoing are much more general, and so difficult to establish, than claims about blameworthiness itself. As I stick to “blameworthiness,” I invite the reader to prudently extend my claims to some of the other adjacent reactions. I also invite the reader to patiently tolerate my occasional use of the word “responsibility,” which is unavoidable given the state of the literature on PAP. When I use it, I strictly mean responsibility for wrongdoing in the sense relevant to blame (and probably to other, though not to all, connected reactions).

2. Excuses and eligibility

PAP is underdescribed. It leaves open the question whether lacking alternatives is an excuse or an exemption. ⁴ This distinction, largely owed to Peter Strawson (1974) and Gary Watson (2004), goes as follows. Excuses apply to specific conduct: for instance, we excuse actions and omissions which are the result of ignorance or coercion. Exemptions, on the other hand, apply to agents more generally: for example, we exempt young children and those suffering from severe dementia. We exempt them from the kind of interpersonal relationship we have with other adults which presupposes that we are in general responsible for our conduct. Or at least, we exempt them from the obligations which are attached to these relationships. The excused have failed to satisfy an obligation, but do not deserve blame; the exempted were not under said obligation. Both exemptions and excuses render casting blame or attributing responsibility unfitting, but for different reasons.

This simple distinction will work for our purpose. But we can note the possible existence of cases which it does not easily classify, such as borderline and mixed cases. Mixed cases feature an agent exempted within a certain domain but whose conduct is excused within another. Having infringed her obligation not to lie, an autistic individual might be excused for lying but exempted from the obligation to show empathy, the lack of which was manifested in lying. Borderline cases are those where an agent could be described either as having limited capacities (she is exempted); or as having adequate capacities that were hard to deploy in the circumstances (she is excused). A child exhibiting poor self-control in the face of danger or disappointment might be exempted because she is a child, or excused because it was too difficult for her in the situation to control her reaction. ⁵

The answer suggested by our moral practice is that lacking an alternative is an excuse, not an exemption. (I shall consider a third answer soon.) We sometimes see that, on a specific occasion, a person cannot help hurting her partner’s feelings by hiding her deeply rooted sentiments nor by expressing them, and so we excuse her. And we do not blame our friend for failing to call us if there was no signal on the train: we excuse him. While the exempted might also lack alternatives, this is not the way we describe them. We might claim that a severely mentally handicapped person or a child lacked basic emotional or motor control or comprehension, but not alternatives. Thus, lacking alternatives is not itself an exemption. And it is far from obvious that lacking alternatives is always implied by a condition that qualifies as an exemption. ⁶

I believe that these remarks have important implications for the kind of alternatives which are relevant to PAP. If lacking alternatives simpliciter is a good excuse, it must be in virtue of the fact that lacking eligible alternatives is a good excuse. ⁷ As we soon shall see, the same would not be true if lacking alternatives were an exemption.

An eligible alternative ⁸ is an alternative whose cost, either moral or material, is not too high to reasonably opt for—and thus one that an informed third party could generally expect one to opt for. ⁹ Typically, lacking alternatives is used as an excuse when there are in fact alternatives, but only ineligible ones. Say I live in London and at midday all the trains to Oxford have been unexpectedly cancelled. I could hire a taxi for £200. But if I don’t and instead email them, take the bus, and arrive at 13:45 for my 13:00 class, my students should accept my excuse. I did not have an eligible alternative. I had an alternative (the taxi), but it was unreasonably costly. And things would have been morally similar if I had had no alternative at all, for instance, if there had been no taxi to hire. I would have lacked an eligible alternative all the same.

The fact that most cases of the “couldn’t do otherwise” excuse feature the lack of eligible alternatives, not of alternatives simpliciter, does not show that people use the excuse incorrectly. Quite the contrary, it would be strange if the excuse were only used strictly. To see this, consider a bus driver who loses the control of his vehicle and has literally no alternative but to crash it into a wall, thereby causing severe but not mortal damage amongst the passengers. Amend the case and give the driver an ineligible alternative: add a ravine on the side of the road into which he can direct the bus. It would be odd if the first, but not the second driver, had access to a special excuse. The addendum makes no difference to the driver’s culpability, to what he should justify to us or be praised for, etc. precisely because the alternative is ineligible. The presence or absence of ineligible alternatives is morally irrelevant, or at least irrelevant to matters of blameworthiness. ¹⁰

As the taxi and bus cases show, what counts as an eligible alternative is decided in part by making a value judgement—something John Hyman (2015, 98) made clear when he introduced the concept. To focus on the taxi case, imagine that I was instead supposed to perform an urgent surgery at 13:00 in Oxford. In that case, the taxi ride was an eligible alternative to arriving late because the cost was reasonable, given the stakes. ¹¹ Value permeates the eligibility, and thus the relevance, of alternatives for PAP.

By contrast, the kind of alternatives that would be required if lacking alternatives were an exemption rather than an excuse would presumably not depend on value. So, if lacking alternatives were an exemption, it would not be in virtue of the fact that lacking eligible alternatives is an exemption (since the eligibility of an alternative depends on value). Suppose a toddler has eaten half of the clafoutis that was intended for the guests. Asking whether refraining from eating the dessert was an eligible alternative and, in particular, whether the cost of waiting for the guests was too high, is an odd way of determining whether we should exempt him. Whether, for instance, the guests were the toddler’s grandparents or the Queen of England seems irrelevant to the question of exemption, but it would be relevant to the question whether to excuse an older child.

To sum up: when we try to be more specific about PAP by asking whether lacking alternatives is an excuse or an exemption, we come to realise that it is an excuse, at least in our current practice. And if lacking alternatives is an excuse rather than an exemption, the kind of alternatives which are relevant to this excuse are eligible ones. Why? For one thing, when we excuse a subject by citing their lack of alternatives, it is almost always the case that they did in fact have alternatives, but that those alternatives were ineligible. For another, the presence or absence of a noneligible alternative makes no difference to the quality of our excuses. When someone claims to have lacked alternatives, this raises the partly ethical question as to whether there were eligible alternatives rather than the purely metaphysical question as to whether there were alternatives simpliciter.

Having realised this, the friend of PAP should seriously consider one of its variants, ¹² the “principle of eligible alternate possibilities”:

PEAP: One should be excused for one’s conduct if one lacked eligible alternatives (at some relevant time), viz alternatives whose cost is not too high to reasonably opt for.

To endorse PEAP is not to claim that PAP is false. On the contrary, it is to explain the grounds we have for thinking that it is true. But it follows that if we discover that PEAP is false, then we should no longer accept PAP—that is, unless and until we can find some other reason to believe it. In a nutshell, our reason for believing PAP is PEAP. Setting acronyms aside, our reason for believing that we are not blameworthy for unavoidable actions is that lacking eligible alternatives is a good excuse.

This is crucial for my purpose, so let me clarify how PAP and PEAP relate. First, PEAP logically implies PAP, given that good excuses undercut blameworthiness. For, if eligible alternatives are necessary for blameworthiness, a fortiori mere alternatives are. Second, PAP does not logically imply PEAP, since it is logically possible that lacking alternatives simpliciter, but not lacking eligible alternatives, is incompatible with blameworthiness. Third, PEAP putatively grounds PAP—it is our reason to believe PAP. This third claim is not a logical claim. In fact, we discover it by asking, as we have, whether lacking an alternative is an excuse or an exemption; by accepting that it is an excuse, and by seeing that if it is an excuse, then it is implausible that the excuse be lacking an alternative simpliciter as opposed to lacking an eligible alternative.

In this section, I have vindicated the first premise of my main argument:

(1) Our (good) reason for believing PAP is, if anything, that lacking eligible alternatives is a good excuse.

Differently put, our reason for believing PAP is undermined, at least temporarily, if we discover that PEAP is false. ¹³

In the next section, I turn to the second premise, which claims that, in fact, lacking eligible alternatives is not a good excuse, unless we understand eligibility differently than we have so far. Remember that I claimed an eligible alternative is an alternative whose cost is not too high to reasonably opt for. Things would be different if an eligible alternative were an alternative whose cost was not reasonably believed or known to be too high to reasonably opt for. Which conception of eligibility is preferable? I will come back to this question in section 6. For now, I will use the nonepistemicised conception of eligibility.

Before I move to my second premise, let me briefly consider an important objection to (1) that we can reconstruct from Helen Steward’s (2009; 2012) and Maria Alvarez’s (2009; 2013) work. Their key claim is roughly that action or agency requires alternatives (since it is the exercise of a two-way power or the result thereof). One does not φ, where φing is an action type, unless one can refrain from φing. ¹⁴ Thus, lacking alternatives is not an excuse, and is something different from the exemptions we considered. Suppose that my partner blames me for having spilt some wine on the carpet, but the dog was responsible. I am neither responsible nor exempted; yet I do not have an excuse. This was simply not my doing. According to Steward’s objection, the same goes when I lack alternatives.

Steward’s objection can be reconstructed as follows. She can claim that PAP is true ¹⁵ because:

1. 1. If I did not have alternate possibilities, I did not act.

2. 2. If I did not act, I was not morally responsible.

3. 3. Therefore, if I did not have alternate possibilities, I was not morally responsible (and so not blameworthy).

So, the truth of PAP has nothing to do with eligible alternatives; alternatives are necessary for responsibility because (mere) alternatives are necessary for true actions, not because lacking alternatives is a good excuse which it would be unfair to ignore (Steward 2009, 81–86).

I only have the space to voice reservations briefly. Regarding the first premise, the idea that action requires alternatives is a highly controversial conception of action. Philosophers have defended other appealing views such as that an action is the causing of a change (see Hyman 2015, chap. 1) or, more traditionally, that action is an intentional event. Steward’s view is more troubling than others because it is uncontroversial that we act and yet it is controversial that we have alternatives. This speaks, admittedly softly, against her view. ¹⁶ Regarding the second, it is implausible if it is read strictly: we can also be responsible for attitudes, omissions, if not sentiments, and states of affairs. This idea is featured in the common law, where criminal liability is thought to include omissions and perhaps possessions. ¹⁷ So, “actions” in the two premises should instead be read ecumenically: it should include actions, omissions, attitudes, and everything else for which we think someone might be directly responsible. Yet, the more we include, the more pressure we put on the first premise.

That being said, I suspect that there are other reconstructions of Alvarez’s and Steward’s argument in the vicinity. ¹⁸ I shall leave this aside for the rest of this paper, but it strikes me as a promising avenue for the PAP defender, albeit one which comes with significant theoretical commitments.

3. The controversy over Frankfurt cases

Is lacking an eligible alternative in fact a good excuse? I shall argue for a negative answer and, so, for the second premise of my main argument, in section 5 (to which the hasty reader can jump). But first, I want, in this section, to briefly discuss the literature on PAP and Frankfurt cases, and in the next, to consider a similar strategy to mine.

The following discussion serves two important goals. First, it shows how intricate the surrounding metaphysical disputes are, and so how beneficial a bypassing strategy like mine can be. Second, it introduces the notion of a robust alternative, which, even if it raises further problems, is an important precursor of the notion of an eligible alternative. In a word, by examining this literature, we see what controversies my strategy avoids, but also how indebted I am to its participants.

Consider Frankfurt’s counterexample to PAP. Jones is about to commit a crime for his own reasons and without any interference. But Black wishes to make sure that Jones will in fact commit the crime, so he secretly implants a device in Jones’s brain. The device will be activated if Jones chooses not to commit the crime. If activated, the device will reliably make Jones change his mind and commit the crime. As it happens, Jones decides to commit the crime on his own. So, he is responsible for committing a crime even though he lacked alternatives, contra PAP.

The broader lesson, according to Frankfurt, is that what matters for responsibility for φing is why the agent φed and nothing else. ¹⁹ Since lacking alternatives might be irrelevant to why an agent φed, it is not always relevant to moral responsibility. In fact, there is a difference between making an agent φ and depriving her of alternatives regarding φ. When these two things come apart, we see that alternatives are not required for responsibility. Frankfurt notes that this speaks in favour of compatibilism, since the truth of determinism does not explain why a specific agent φs (determinism is at best explanatory relevant). These are appealing claims, but they have been subject to ingenious objections. ²⁰

Yet, the literature has generally focussed on counterexamples to PAP themselves. Defenders of PAP spent considerable energy arguing that Frankfurt cases failed in one of two main ways. Each tried to exploit the notion that Jones retains alternatives in the cases, and so that these cannot be true counterexamples to PAP.

The pure defence (Steward 2009; Alvarez 2009) argues that Jones did in fact have alternatives to φing: he could have refrained from φing. Here is why. If I move your fingers on a piano keyboard, you are not playing the piano. Similarly, if Black moves Jones in the counterfactual scenario, Jones is not acting—his limbs are moving. ²¹ Black might instead be merely influencing Jones but, in that case, Jones retains the ability to resist. So, however we describe the case, Jones had an alternative, namely refraining from φing.

The fine-grained defence (Ginet 1996; van Inwagen 1997; Capes and Swenson 2017) concedes that Jones is not responsible for φing simpliciter. Rather, he is responsible for φing in a certain way. True, Jones could not have refrained from committing a crime, but he could have refrained from committing a crime on his own or from committing a crime at that precise instant.

Some others take a mixed approach. For instance, Kadri Vihvelin (2013, chap. 4) follows the fine-grained defence for “bodyguard” Frankfurt cases, i.e., cases where Black would intervene if Jones actually decided not to do the crime. In such cases, it is clear that Jones could decide to do otherwise, which is enough to make him responsible. And Vihvelin comes closer to the pure defence with “preemptor” Frankfurt cases, i.e., cases where Black would have intervened preemptively, long before Jones’s change of heart, if he had predicted it. In such cases, Jones’s relevant abilities (including his ability to refrain from φing) are left unaltered. ²²

These different approaches have been questioned. John Fischer (1994), one of the most ardent critics of PAP, recognised early on that Jones did in fact have some alternatives in Frankfurt cases. But he thought that these “flickers of freedom” failed to save PAP. Why? After all, strictly speaking, for PAP to be true, all that needs to be the case is that there be no situation where someone is both responsible and lacks alternatives. So, flickers of freedom do disqualify putative counterexamples to PAP. However, this is shortsighted for reasons already revealed in section 2. The point can be put metaphysically and epistemologically. Metaphysically: if PAP is true, it is in virtue of the fact that one cannot be responsible (blameworthy) in the absence of certain relevant alternatives, not just any. To see this, consider someone who lacks alternatives to φ. Now, give her one “flickery” alternative: she could refrain from φing by suffering a heart attack, or perhaps by accidentally knocking herself out. This addition is irrelevant to moral responsibility for φing. Similarly, imagine a person who only has these flickery alternatives and remove them: this subtraction is similarly irrelevant. Thus, the existence of mere alternatives is not relevant to moral responsibility. Epistemologically: our reason for believing PAP is that lacking certain alternatives is incompatible with moral responsibility (blameworthiness), not just any. We have no reason for believing that the addition or the subtraction of just any alternative like having a heart attack or knocking oneself out is relevant to moral responsibility.

This is what Fischer had in mind when he maintained that the alternatives identified to defend PAP should be robust ones, alternatives which, in his terms, have enough “oomph.” Less metaphorically, a robust alternative is an alternative which grounds the responsibility of an agent (Fischer 1994, 140) or explains her blameworthiness (Pereboom 2014, 10). This is the concept of robustness, which is accepted in the literature. But there are different possible conceptions of robustness. For Fischer and Pereboom, a robust alternative must involve free or voluntary behaviour salient to the agent. ²³

With this apparatus in mind, how should we assess the pure and fine-grained defences of PAP? We enter the land of metaphysical controversies. Take first the pure defence. According to it, Jones has the alternative to refrain from committing a crime by being moved (his limbs move). A problem immediately arises: if, in the alternative sequence, Black plays with Jones’s brain, can’t he make sure that Jones will act? Black could instill an irresistible desire in Jones to commit a crime. The pure defender then insists, controversially, that irresistible desires are problematic, if not impossible. ²⁴ Fischer might then question whether, if a desire is strong enough, the alternative consisting of ignoring it counts as robust. (Consider torture: a strong enough desire to stop suffering might explain why resisting longer is not a robust alternative.) To determine what alternatives count as robust, we must then tackle difficult questions about desires.

Turning to the fine-grained defence. In the most refined Frankfurt cases (see e.g., Hunt 2007; Fischer 2012; Pereboom 2014), the counterfactual agent would intervene if Jones showed that he was about to decide to change his mind. The fine-grained defenders insist that, although Jones could not decide to do otherwise, he could simply omit to decide on his own (by doing whatever it takes to trigger the chip). Crucially, they insist that this alternative is robust in Fischer’s sense of containing a free or voluntary behaviour (e.g., Capes and Swenson 2017, 969–72). It is robust because an omission can be free even if it is not preceded by a decision. But that’s controversial. To determine which alternatives count as robust, we must then tackle difficult questions about free omissions. ²⁵

This is only a rough summary of a fifty-year-old debate. But it shows that the fate of Frankfurt cases depends on a disagreement about the nature of robust alternatives, which turns on further metaphysical controversies (about free omissions and desires amongst other things). Can we bypass these metaphysical quandaries?

4. An unsuccessful indirect strategy

Michael McKenna (2003) argued that we can and, like me, he thought that the way to go was to reconceptualise robust alternatives. According to Fischer and Pereboom’s conception, it is a metaphysical and epistemic property: a robust alternative includes free or voluntary behaviour salient to the agent. McKenna thought this misses a key subjective ingredient: deliberative relevance. McKenna (208) confesses he finds it hard to define it, but he thinks that examples can give us a good enough sense of what it is. To illustrate, here is a (simpler) case of my own:

dessert: Mike likes receiving people at home, but he is hopeless at cooking desserts. He always goes for one of two things: either he gets baklava from the local Turkish bakery, or he makes crepes at home. Making crepes is more taxing, but it is a nicer touch for the guests. Now, Mike knows of other options such as going to the grocery store to get lesser quality and more expensive baklava. Tonight, Mike is wondering what he will do. Unbeknownst to him, his guests hate crepes. So, successwise, it would be better for Mike to buy baklavas, whether from the Turkish bakery or from the grocery store. But getting baklavas from the grocery story is not a significant alternative for him in his deliberation. This remains true if the Turkish bakery is closed, unbeknownst to him.

The idea is that even if, successwise, it would be better for Mike to get the worse baklava than to make crepes, this option is deliberatively irrelevant (and so not robust). This alternative is not in the cards (at least at this point) and should not be. As for the option to go to the Turkish bakery if it is closed unbeknownst to Mike, it is deliberatively relevant but not robust, since it is not a real option.

On the grounds that only deliberatively relevant alternatives should be considered when assessing PAP, McKenna (2003, 209) suggests permuting PAP with:

PSA: An agent S is morally blameworthy for performing action A at t only if she had within her control at t performing an alternative action B such that (1) performing B at t was morally less bad than performing A at t, and (2) it would have been reasonable for S to have considered performing B at t as an alternative to performing A at t given S’s agent-relative deliberative circumstances.

To simplify, an agent is only blameworthy if she had a robust alternative, where robustness implies deliberative relevance. (For McKenna, PSA plays the role that PEAP plays for me: it logically implies PAP but is not logically implied by PAP, and it grounds the truth of PAP.)

McKenna then argues that PSA can be shown to be false, using an elaborate counterexample. Here is a summary.

Casper has ten seconds to press on one of two buttons, “Good” or “Bad.” He knows that if he presses “Bad,” he will receive one million dollars drawn from the bank accounts of a million college professors. If he presses “Good,” he will cure a village from a dangerous disease at no cost. However, unbeknownst to Casper, a brain lesion makes it impossible for him to press the “Good” button (this mimics the bakery being closed in dessert).

According to McKenna, Casper lacks robust alternatives because the option to press “Good,” while deliberatively relevant, is blocked. And the option to press nothing, while real, is deliberatively irrelevant. But Casper is blameworthy if he presses “Bad” straightaway. So, PSA is false: someone can be responsible in the absence of deliberatively relevant alternatives.

McKenna’s strategy is ingenious (even if, regrettably, its inclusion of a special brain lesion brings us back to contentious thought experiments), but it fails. As Robinson (2014) has argued, the strategy fails because it does not consider omissions properly. ²⁶ For, omitting to do that which one knows to be wrong is always a deliberatively relevant alternative. And once we include omissions in PSA, the counterexample falls short. Casper had the robust alternative of omitting to press the “Bad” button. This option was robust because it was a real one and a deliberatively relevant one. Conceded, there is a mismatch between, on the one hand, the way to omit which Casper thinks is best, and, on the other hand, the way to omit which is actually available to Casper. But PSA remains unrefuted.

To rule out all robust alternatives, we might have to tweak Casper’s brain to remove the possibility of omission. But this seems like a one-way ticket to the metaphysical funfair: whether Casper lacks robust alternatives will remain controversial. This does not hold the promise of an indirect strategy.

5. A new indirect strategy

McKenna was dead right in opting for an indirect strategy, a strategy which does not try to directly refute PAP. He was also right to include a value-infused component in his conception of robustness. But he was wrong, I think, to suggest that the missing ingredient was deliberative relevance.

The missing ingredient of robustness is, as we anticipated, eligibility. In section 2, we saw that our reason for accepting PAP is that lacking eligible alternatives is a good excuse. Eligibility is the missing ingredient of robustness. An ineligible alternative does not adequately ground the agent’s responsibility or blameworthiness. An alternative that is too costly, morally or materially, should be irrelevant to responsibility or blameworthiness, just like an alternative which is nonsalient to the agent. Thus, PAP defenders cannot respond to Frankfurt cases by invoking the presence of ineligible alternatives in them.

This opens up a new strategy. If we can identify scenarios—whether Frankfurt cases or others—where an agent is blameworthy yet lacks eligible alternatives, this will be a problem for PAP defenders. In particular, it will undermine their reason for believing PAP: that lacking eligible alternatives is a good excuse (PEAP). There we will have a new indirect strategy against PAP.

These scenarios do not require much in the way of tricky brain manipulation. We can start with:

fischer’s case: In this kind of variant on the Frankfurt-type scenario, if it becomes clear to the counterfactual intervener—he is an excellent judge of such things—that the agent is going to decide to do something else [than shooting Smith or than shooting Smith with a certain motive], he will use his machine to destroy the agent’s brain and thus kill him instantly! […] [T]he agent in some sense has access to another path—the path on which he is instantaneously killed (not a very promising path, by the way!) But this path does not contain any voluntary behaviour by the agent or anything with sufficient “oomph” to ground moral responsibility.

(Fischer 2003, 242)

As we have seen in section 4, Fischer’s reasoning is resistible. Both the fine-grained defender (Capes and Swenson 2017, 969–72) and the pure defender (Alvarez 2009, 69–70) can argue that Jones has a robust alternative, in Fischer’s sense, to shooting Smith.

But the case is a good starting point, which we must amend in aim and in detail. Regarding the aim, the case must be taken to show that lacking an eligible alternative is not a good excuse. ²⁷ This will show that PEAP is false. As we have argued in section 2, if PEAP falls, then so does our justification for PAP. PAP might still be true, but we lose our reason for believing that it is.

Regarding the details, Jones might have an eligible alternative in fischer’s case because it seems reasonable to demand that he sacrifice himself instead of shooting Smith. After all, self-defence does not warrant killing innocent bystanders and the Common Law does not recognise duress as a defence for murder. ²⁸ So, we need to amend the case:

wallet-or-death: Just like fischer’s case, but Jones’s crime is to steal Smith’s wallet. (Unbeknownst to him, he would be killed if he refrained to steal, or if he showed the slightest sign of hesitancy.)

Now, the option of dying in order to refrain from stealing Smith’s wallet is too costly. Jones lacks eligible alternatives in spite of being blameworthy for stealing the wallet. Now we have refuted PEAP and so undermined PAP.

Let us take stock. With Fischer and Pereboom, I focussed on robust alternatives, i.e., alternatives which, in principle, could ground or explain responsibility or blameworthiness. But, against them, I denied that the best way to conceive of robustness is metaphysical. With McKenna, I accepted that robustness is a partly moral concept but, against him, I argued that the missing moral ingredient of robustness is eligibility. This allowed me to work with cases much simpler than Frankfurt’s or McKenna’s because removing eligible alternatives requires no brain manipulation or lesion. This paved the way to showing that PEAP, our reason for believing PAP, is false. My counterexample does not have the (now recurring) problem that omissions posed to the others. In fact, in wallet-or-life, while to omit is plausibly robust in the sense of Pereboom and Fischer, it just is not an eligible alternative. Jones’s life is no reasonable price for a wallet.

At this point, the reader might (and should!) question my conception of eligibility. Can we not simply maintain that whether an agent has eligible alternatives depends on her epistemic situation? Accordingly, Jones in fact had an eligible alternative, since he had no reason to think his life was on the line. This might be the best conception of an eligible alternative. But remember that the premise I intended to defend in this section was:

(2) Lacking eligible alternatives is not a good excuse, unless the eligibility of an alternative is relativised to the agent’s epistemic situation.

My amended case establishes (2); the objection is irrelevant because of its ‘unless’ clause. This takes us to the third and last premise of my main argument to undermine PAP.

6. Relative eligibility

Is eligibility relative to the epistemic situation of the agent? If so, it would appear that we can save PEAP after all (even if this leaves [2] untouched). We could say that lacking an eligible alternative is an excuse when an eligible alternative is one whose cost is such that it is reasonable for the subject to opt for it, given what she knows (or has reason to believe) in the circumstances. In wallet-or-death, whether Jones is blameworthy depends on whether he knew (or had reason to believe) that he would be killed if he did not steal. If he knew, he did not have an eligible alternative and should not be blamed. Whereas, if he did not know, he did have an eligible alternative and should be blamed. The principle that he should be blamed only if he had an eligible alternative is therefore saved. PEAP stands. ²⁹

We can make this important point slightly differently, by keeping a nonrelativised reading of eligibility, but by amending PEAP:

PEAP*: One should be excused for one’s conduct if one knew, or reasonably believed, that one lacked eligible alternatives (during the relevant period of time)—viz alternatives whose cost is not too high to reasonably opt for.

I have a slight preference for this second way of epistemicising PEAP, ³⁰ but it does not matter much for our purpose. Everything I say about one applies to the other. Note also that PEAP*, like PEAP, can be understood as specifying a requirement on blameworthy conduct: a blameworthy agent must not know, or have reason to believe, that she lacks alternatives. ³¹

Yet my main argument already makes it clear that this amendment will not save PAP. Its third premise reads:

(3) If eligibility is relativised to the agent’s epistemic situation, the principle that lacking eligible alternatives is a good excuse provides no support for PAP and, thus, is not a good reason for believing it.

Differently put, PEAP* does not constitute a good reason for believing PAP. This is not because it is false but because it provides no support for PAP.

Why? PEAP* states an epistemic excusing condition, and thus an epistemic requirement on blameworthines. But PAP does not state an epistemic requirement on blameworthiness; it states a metaphysical requirement on blameworthiness. Yet, a principle about an epistemic requirement is not a good reason (on its own) to believe in a principle about a metaphysical requirement. So PEAP* is not a good reason to believe PAP, hence (3) is true.

This reasoning calls for nuances. We should be prudent in segregating epistemic conditions and metaphysical conditions such as control conditions on blameworthiness. They are connected. The acquisition of knowledge normally supposes control, if only the control required to maintain focus. ³² And, as Alfred Mele (2010) has rightly suggested, control requires knowledge, perhaps because control requires acting for reasons which require knowledge (as Hyman [2015, chap. 7] argues). In this case, we might want to conclude that if lacking control is an excuse, so is complete ignorance (since it necessarily rules out control).

However, for an epistemic condition to support a metaphysical condition, we need to fill a gap. This might be easy if the metaphysical condition involves the recognition of, and reaction to, reasons. But this is harder in our case. A complete ignorant might have dozens of alternatives, while a perfect knower might have none.

To illustrate further the disconnect between PAP and PEAP*, consider how differently they stand to the free will debate. If PAP is true and if we all lack alternatives at all times, as some free will sceptics maintain, then we are never blameworthy. This is true regardless of what we believe or should believe about the matter. By contrast, if PEAP* is true, no such thing follows. It only follows that, in worlds where free will exists, some, but not all, subjects are excused (unless they all have the same relevant reasons or knowledge); and the same goes for worlds where free will does not exist. So, unlike PAP, PEAP* could not be used by sceptics to argue convincingly that no one could ever possibly be blameworthy.

PEAP* is also very different from PEAP regarding its relation to PAP. Like PEAP, PEAP* exculpates ³³ some people that PAP does not. But unlike PEAP, PEAP* fails to exculpate some people whom PAP exculpates. Put another way, PEAP extends the domain of the exculpated PAP sets, whereas PEAP* defines quite a different set of exculpated agents. Thus, PEAP* is not a reason, at least certainly not a sufficient reason, for believing PAP. The two principles seem related in name alone. ³⁴

In passing, PEAP* does not support PAP, but it might support PAP*, Frankfurt’s (1969) and Strawson’s (1974) principle according to which one is excused for actions that she performed because she lacked alternatives. For, if Jones acted because he lacked alternatives, he must at least believe and perhaps know that this is true. This raises a further question: Should we simply accept PAP* or should we also grant PEAP*? If both principles are true, then an agent is excused if she acted with the knowledge that she lacked eligible alternatives, even if she did not act on that knowledge. According to PEAP*, a driver is excused for bumping into someone else’s car ³⁵ if he was aware that there were no eligible alternatives, such as braking, even if his reason for bumping is malevolent. If PAP* but not PEAP* is true, this is not necessarily the case. ³⁶ To clarify, PAP*, no more than PEAP* and for the same reason, does not support PAP.

Let’s recap. (1) Our (good) reason for believing PAP is, if anything, that lacking eligible alternatives is a good excuse. (2) Yet, on close examination, lacking eligible alternatives is not a good excuse, unless the eligibility of an alternative is relativised to the agent’s epistemic situation. (3) And, in this latter case, the principle that lacking eligible alternatives is a good excuse provides no support for PAP and thus is not a good reason for believing it. Therefore, we have no good reason for believing PAP. This argument allowed us to bypass fifty years of metaphysical debates (summarised in section 3). Differently put, it allowed us to show that robustness, which was conceptualised as part of these debates, had been underexploited and incomplete—it should include eligibility and be used in an indirect challenge against PAP.

I do not expect PAP defenders to surrender. They might for instance claim that PAP is irreducibly intuitive and that I was wrong in the first place to ask why one should believe in it. In the next section, I explore a different concession that they might try to negotiate.

7. The price of PAP

Faced with my argument, the PAP defender has an option, which is to make an ethical concession. She can accept almost everything I have said so far except for my judgement about Jones in wallet-or-death. This forces her to conclude that, in this case, he is morally lucky. For Jones has a good excuse for his crime: he had no eligible alternative. The PAP defender can then argue that Jones is blameworthy for something else, like harbouring some reprehensible desires. But then she should not blame Jones with the same force as if he were blameworthy for stealing. And she should expect to be asked for a verdict on a case where Black presides over Jones’s desires—he would shoot him if Jones had good will towards Smith. In this last version of the case, the PAP defender must excuse Jones entirely.

This seems like an exorbitant and unusual concession. It is exorbitant because we can imagine a plethora of counterintuitive cases. Two employees who could not afford to lose their job are paid to work in separate electronic stores. Each has received from her boss a stock of defective phones. Both sell the items, knowingly swindling customers, for no reason except their lack of concern. Yet, unbeknownst to them, whereas the first employee would have been fired if she had not proceeded, the second would have suffered no consequence. If we grant PEAP in its nonepistemicised version (and PAP), we have to exculpate the first, who lacked eligible alternatives to swindling, but presumably not the second, who had such alternatives.

Some might say that this is just good old moral luck: If PEAP is true, some agents are off the hook simply in virtue of circumstances that they were unaware of and which they have no control over (e.g., Black’s existence, the different boss). Therefore, two identical agents with the same flawed motives and conduct are not necessarily equally blameworthy. This works like typical cases of moral luck, where, for instance we morally distinguish a driver who is guilty of accidentally dooring a cyclist from a driver who was just as negligent but hurt no one. (see Nagel 2012, chap. 3) Consequences seem to directly matter to blameworthiness.

However, PEAP introduces an unusual form of moral luck (different from the moral luck of the two drivers) because here (e.g., in the case of the two employees), not only are the conduct and motives identical, but the consequences of the conduct are identical. What pulls the agents apart is their modal circumstances: one had eligible alternatives, the other one did not.

This concession is also unusual given the stance that PAP defenders have taken over the years. They have normally ³⁷ assumed that they could respond to Frankfurt cases at no moral cost. In particular, they agreed with Frankfurt’s verdict about the responsibility of individuals like Jones ³⁸ and yet challenged Frankfurt’s explanation of that verdict. But this middle ground is now closed. PAP defenders cannot have their cake and eat it too, namely good grounds for believing PAP plus the most intuitive moral judgements.

8. Conclusion

I have revisited the debate about Frankfurt’s counterexamples to the principle of alternate possibilities by asking why the principle is true—if, in fact, it is. My provisional answer was that lacking eligible alternatives is a good excuse. But, I argued, this can only be plausible if eligibility depends on the epistemic situation of the agent. Yet, if this is the line we take, we need to concede that we have lost our reason for believing PAP. For PAP does not specify epistemic, but rather metaphysical, requirements for blameworthiness. The two traditional responses to Frankfurt cases, the pure defence and the fine-grained defence, cannot help us avoid these counterintuitive consequences. In fact, my guiding motive is precisely the desire to avoid the metaphysically murky grounds on which these admirable defences stand. Even if they successfully show that counterexamples to PAP fail, PAP will remain unmotivated.

If PAP must be saved, its defenders should look for a different reason to accept it—perhaps Steward’s and Alvarez’s conception of agency briefly discussed in section 2. Failing that, they will have to do something almost unheard of in the past fifty years. They will have to accept the following new form of moral luck. Sometimes, two identical agents in identical worlds with identical actions and motives resulting in identical consequences are not identically blameworthy. For one of them, but not the other, has unbeknownst to her a good excuse since she lacks eligible alternatives because of a discrete counterfactual agent. Whether exorbitant or not, this is the current price of PAP. For myself, I prefer not to pay it.