‘The moral irrelevance of moral coercion’

An agent A morally coerces another agent, B, when A manipulates non-epistemological facts in order that B’s moral commitments enjoin B to do what A wants B to do, and B is motivated by these commitments. It is widely argued that forced choices arising from moral coercion are morally distinct from forced choices arising from moral duress or happenstance. On these accounts, the fact of being coerced bears on what an agent may do, the voluntariness of her actions, and/or her accountability for any harms that result from her actions (where accountability includes liability to defensive harm, punishment, blame and compensation). This paper does not provide an account of the wrongness of moral coercion. Rather, I argue that, whatever the correct account of its wrongness, the mere fact of being coerced has no bearing on what the agent may do, on the voluntariness of her action, or her accountability for any resultant harm, compared to otherwise identical cases arising from duress and happenstance.


Introduction
An agent A morally coerces another agent, B, when A manipulates nonepistemological facts in order that B's moral commitments enjoin B to do what A wants B to do, and B is motivated by these commitments. 1 The use of human I'll use happenstance to describe cases in which there is a forced choice, but no wrongdoing (which includes natural accidents and innocent misadventure), as in Trolley: Trolley: A trolley is blown by the wind to where it will kill five innocent people. Bystander can save the five only by diverting the trolley down a sidetrack, where it will kill Victim. Bystander diverts the trolley. Victim is killed; the five are saved.
Again, Bystander is motivated by her moral commitments, rather than by any threat to her person.
Many philosophers argue that forced choices arising from moral coercion (henceforth, coercion) are morally distinct from forced choices arising from moral duress (henceforth, duress) or happenstance. According to these philosophers, the fact of coercion is morally significant in at least one of the following respects: it bears on what an agent may do (Kamm 2007;Bazargan 2014;Zohar 2014), the voluntariness of her actions (Mason 2012), and/or her accountability for any harms that result from her actions (where accountability covers a broad range of liabilities, including liability to defensive harm, punishment, blame and compensation (Mason 2012;Kamm 2007). The view that coercion is morally significant was also reflected in the 2015 US Department of Defence Law of War Manual, which held that, unlike other civilians caught in the crossfire, civilians being used as involuntary human shields by the enemy (as in Cache) may be disregarded in combatants' proportionality calculations (US Department of Defense 2015: 5.12.3). 3 The UK Joint Service Manual of the Law of Armed Conflict holds that harm to involuntary shields may be discounted relative to harm to other civilians (UK Ministry of Defence 2004: 2.7.2), such that harms inflicted on shields can be proportionate even if inflicting identical losses on non-shields would be disproportionate. Whether coercion is indeed morally distinct from duress and happenstance is thus of considerable importance. This paper defends Moral Irrelevance: Moral Irrelevance: When all else is equal, whether a person faces a choice between harms as a result of moral coercion, moral duress, or happenstance has no bearing on what she may do, the voluntariness of her actions, or her accountability for those actions.
The paper does not provide an account of the wrongness of moral coercion. 4 Rather, I argue that, whatever the correct account of its wrongness, the mere fact of being coerced has no bearing on what the agent may do, on the voluntariness of her action, or her accountability for any resultant harm, compared to otherwise identical cases arising from duress and happenstance. I assume throughout that the coercion is unjustified.
I begin by defending the claim that coercion makes no difference to what an agent may do. In Section Two, I consider two objections that might seem to easily defeat Moral Irrelevance. The first is that wrongdoing matters for what we may do, thus marking a distinction between harms arising from coercion and duress, and harms arising from happenstance. The second is that intentional rights violations are worse than unintentional rights violations, thus making it worse to allow the killing of the five in Duress Trolley than to allow the killing of the five in Coerced Trolley. I show that neither undermines Moral Irrelevance. In Section Three, I reject Saba Bazargan-Forward's claim that coerced harms ought to be treated as intentional by the coercee, thus making them harder to justify than non-coerced harms by altering their weight in the coercee's proportionality calculation (Bazargan 2014: 14). In Section Four, I argue that agents lack rights to resist being used as means by coercers that make a difference to what they may do. In Section Five, I reject the claim that one may resist moral coercion on the grounds that acceding allows the coercer to exploit morality. In Section Six, I argue that, other things being equal, one is as responsible, in the voluntariness sense of responsible, for harms that one inflicts or allows as a result of coercion as one is for harms that one inflicts or allows as a result of duress or happenstance. I reject Elinor Mason's claim that coerced actions are less voluntary than non-coerced actions, and offer an alternative explanation of why coercees are not accountable for harms resulting from coerced actions (Mason 2012). Section Seven concludes.
2 Two objections to moral irrelevance 2.1 The relevance of wrongdoing Moral Irrelevance is compatible with the view that wrongdoing bears on what an agent may do. Imagine that Bystander can save either Victim from being murdered, or Other Victim from being killed by a falling tree. If one ought to attach greater disvalue to harms that arise from wrongdoing compared to those that do not, 5 then, other things being equal, Bystander ought to save Victim rather than Other Victim.
This does not undermine Moral Irrelevance, even though unjustified moral coercion and moral duress both involve wrongdoing, and happenstance does not. This is because in cases of either moral coercion or duress, we are weighing rights violations against each other, rather than a rights violation against an adventitious harm. If Bystander kills Victim in either Coerced Trolley or Duress Trolley, Villain thereby violates Victim's right not to be killed. And if Bystander refrains from diverting towards Victim in either case, such that the five are killed, Villain thereby violates the five's rights not to be killed. In cases of moral coercion or duress, anyone who is harmed, or allowed to suffer harm, suffers a rights violation This feature of coercion and duress cases is therefore neutralised. One ought to simply weigh the harm that one will inflict on Victim against the harm one will otherwise allow to befall the five. This is also what one ought to do in a happenstance case. Thus, the fact that coercion and duress involve wrongdoing does not undermine Moral Irrelevance. This is true even if the fact of wrongdoing makes coercion and duress cases morally worse than happenstance cases. Bystander is not choosing whether to bring about either a wrongful harm or an adventitious harm in our cases. When Villain has wrongly set a trolley in motion (as in Coerced Trolley and Duress Trolley), Bystander is choosing between only wrongful harms. There is no option available to her that avoids anyone's having their rights violated. Whatever she does, Villain will violate the right(s) of whoever is killed.
Moral Irrelevance is thus compatible with the claim that Bystander ought to prevent wrongdoing in Double Trolley:

Intentional rights violations
Everyone who suffers harm as a result of unjustified coercion or duress will have their rights violated. But only some will have their rights intentionally violated. For example, if Bystander fails to divert in Duress Trolley, Villain will intentionally violate the five's rights, since it is their deaths that Villain is trying to bring about. In contrast, if Bystander fails to divert in Coerced Trolley, Villain will unintentionally violate the five's rights, since his intention is that Victim be killed. Thus, we might think that it is morally worse if the five are killed in Duress Trolley compared to Coerced Trolley, and that this bears on what it is proportionate to do to save the five in each case.
It is true that non-consequentialists often describe intentional harms as harder to justify than unintentional harms. But this is imprecise, because one might intentionally harm a person either eliminatively or in a way that makes use of her. Harming a person in a way that makes use of her is plausibly harder to justify than harming her as a foreseen side-effect. It is less plausible that eliminatively harming someone is harder to justify than harming her as a side-effect. This is 'The moral irrelevance of moral coercion' 3469 evidenced by the fact that several non-consequentialists think that it is permissible to eliminatively kill a non-liable person to save one's own life, but impermissible to kill a non-liable person as an unintended side-effect of saving one's own life. 6 Moreover, the fact that it is harder to justify some modes of harmful agency compared to others does not entail that unjustified instances of those modes of harming are, other things being equal, morally worse than unjustified instances of other modes of harming. On the contrary: other things being equal, unjustifiably killing someone as a side-effect is plausibly as morally bad as unjustifiably killing her eliminatively, or in ways that make use of her (Frowe 2019: 630-631;Tadros 2018: 291). At the very least, and sufficient for my argument here, the mode of agency of an unjustified killing does not seem to bear on what may be done to prevent it. Consider Attack: Attack: Terrorist lethally and unjustifiably throws a grenade at Victim, intending to kill him as a means of intimidating others. The grenade blast also kills Bystander as a foreseen but unintended side-effect. Bystander's death will also have the effect of intimidating others.
The fact that Terrorist intends to kill Victim and merely foresees that he will kill Bystander does not mean that killing Bystander is less bad than killing Victim. That the killings are equally bad is reflected in most legal systems, which would count both as murder. Nor do Terrorist's intentions give third parties any reason to save Victim rather than Bystander.
This moral parity explains why the literature on defensive harm has comparatively little to say about the modes of agency of unjustified killings despite the considerable interest in how modes of agency might settle whether a killing is justified. 7 Consider, for example, that Coerced Trolley does not specify whether Victim's death is eliminative or useful. Perhaps Victim is Villain's romantic rival, and Villain just wants Victim gone. In this case, the killing seems eliminative. Or perhaps Villain will derive pleasure from the killing of Victim, in which case he makes harmful use of Victim. 8 That writers do not tend to specify the mode of agency involved in such cases is not, I think, best explained by the fact that Victim's death is intentionally brought about by Villain, irrespective of whether Victim is eliminatively or usefully killed. Rather, it is best explained by the fact that killing is unjustified in each case, rendering modes of agency morally irrelevant for the purposes of permissible defence. 9 For example, it seems unlikely that Victim may use less force to thwart 6 This is, roughly, the difference between killing a non-liable threatener, such as someone who will otherwise non-responsibly land on you and crush you to death, and killing a bystander as a side-effect of defending yourself against some other threat. See (Thomson 1991;Frowe 2014: Chs. 1-2). 7 Seth Lazar's Sparing Civilians (OUP, 2015) is an exception. Lazar thinks that the ways in which unjust combatants unjustifiably kill just combatants or civilians bears on the badness of those killings. For rebuttal, see (Tadors 2018). 8 Sexually-motivated killings are like this, for example. being eliminatively killed compared to being usefully killed, or that there is a difference in the proportionate harm that third parties may inflict (on either Villain or bystanders) in the course of saving Victim in each case. Rather, since the killing is unjustified in each case, there is no difference in what Victim or third parties may to do to prevent it.
Nor is there any difference between what may be done to prevent intentional unjustified killings and merely foreseen unjustified killings. Imagine that Villain coerces Bystander to divert towards Victim not in order to kill Victim, but rather because Victim is on the longer piece of track, and Villain loves to see the trolley in motion. Villain does not intend that Victim be killed. He merely foresees that Victim will be killed as a result of coercing Bystander to divert onto the longer track. But this has no bearing on what may be done to prevent Victim's death, given that Villain's bringing about Victim's death is unjustified.
If modes of agency are morally irrelevant when it comes to what may be done to prevent unjustified harming, the fact that Villain will intentionally violate the five's rights if they are killed in Duress Trolley, but unintentionally violate their rights if they are killed in Coerced Trolley, does not matter for what Bystander may do. It is not proportionate for Bystander to inflict more harm to prevent intentional rights violations compared to unintentional rights violations.
3 Coerced harms as intentional harms 3.1 Bazargan-Forward's account Saba Bazargan-Forward argues that when a coercee is determining whether she may accede to a coercer's wishes, she ought to treat the harm that the coercer intends her to cause or allow as intentional, rather than merely foreseen. Bazargan-Forward defends this claim using two cases, Hostage and Alley: Hostage: Villain threatens to kill fifty innocent hostages unless Bystander kills Victim. Alley: Thirty children are stuck in an alley. Bystander and Victim are also in the alley. Villain is overlooking the alley from a nearby rooftop. Villain has a bomb, and wants to kill Victim. He lacks a clear shot at Victim, but knows that if he hurls it towards the children, Bystander will divert it in the only other direction-towards Victim-in order to save the children.
It is, he argues, ''not infelicitous'' to describe Victim's death in Alley as intentional, even though Bystander did not kill her intentionally (Bazargan 2014: 13). This is because (i) Villain aims at Victim's death, and (ii) Bystander furthers that aim by substantially contributing to that harm (Bazargan 2014: 13). Thus, in terms of Footnote 9 continued proportionality of defence. But it suffices here that those who usefully kill are not necessarily more culpable than those who eliminatively kill. Hence, there is no necessary connection between the degree of permissible defensive force and the mode of agency of the threat.
'The moral irrelevance of moral coercion' 3471 agency, throwing the bomb at Victim in Alley is on a moral par with shooting Victim in Hostage, even though in Alley Bystander ''kills merely foreseeably rather than intentionally.'' (Bazargan 2014: 13-14, emphasis in original). Bazargan-Forward holds that intentional harms are harder to justify than merely foreseen harms, and receive greater negative weight in proportionality calculations. Thus, the fact that one is coerced affects the proportionality calculation regarding what one may do, compared to an otherwise-identical case that does not involve coercion.
Bazargan-Forward also rejects the view that intending is purely agent-relative, such that one must augment harms in a proportionality calculation only if one intends them oneself. Rather, he argues, there is an agent-neutral reason to ensure that non-liable people are not intentionally killed (Bazargan 2014: 14). That harms are intended by someone matters from the agent-neutral perspective, and affects proportionality. Bystander's killing of Victim in Alley ''should be weighed as heavily as it would be if it were committed intentionally, even if [Bystander] is actually committing it collaterally.'' (Bazargan 2014: 13). 10 Note that Bazargan-Forward is ambiguous about the type of intentional killing involved in Alley and Hostage. This is unfortunate, because his account relies on the thought that whether Bystander is justified in, for example, throwing the grenade at Victim in Alley depends on the mode of agency involved-that is, on what Villain intends. Since intentional killing covers both eliminative and useful killing, and useful killing is plausibly harder to justify, it matters precisely what Villain intends: not because it makes Villain's unjustified actions any better or worse, but because the mode of agency will set the justificatory threshold at which Bystander may accede. Bystander must ask, for example, whether she is justified in usefully killing Victim in order to save the five in Coerced Trolley, or usefully killing Victim to save thirty in Alley.
Bazargan-Forward's argument is most plausible if we construe his cases as involving useful killing-for example, if we assume that Villain wants Victim dead because he will derive pleasure from the killing-because, as I suggested above, it's unclear whether intentional eliminative harming is harder to justify than merely foreseen harming. In what follows, then, I'll assume that Villain intends the lethal use of Victim in both Alley and Hostage. If Bazargan-Forward's claim about the dominance of Villain's intentions is sound, this will affect what Bystander may do in in a coercion case compared to non-coercion cases. However, as I'll argue below, Bazargan-Forward's claim is mistaken.
Bazargan-Forward's two conditions generate an implausibly broad account of when one must treat foreseen harms as if they are intentional. Consider Redundancy: Redundancy: Amos wants Brenda's life to go badly because he dislikes her. He falsely tells Brenda's manager, Clare, that Brenda has been spreading unpleasant rumours about Clare in the hope that Clare will fire Brenda. Clare's company's finances are such that someone must be made redundant. Nobody deserves to lose their job, so Clare puts all the employee's names in a hat and pulls out Brenda's name. Losing her job will make Brenda's life go badly.
Bazargan-Forward's criteria require Clare to treat as intentional the harm she foresees she will cause Brenda, since Amos aims at the harm of Brenda's life going badly, and Clare's making her redundant substantially contributes to that end. And yet it seems clear that Clare need not treat the harm she causes Brenda as intentional, nor treat making Brenda redundant as worse than making a different employee redundant. It is not enough, then, that one's actions coincide with what a wrongdoer wants. To be plausible, Bazargan-Forward's view must at least be restricted to cases in which the victim's suffering harm runs through the wrongdoer's planned causal route. If so, harms that the coercer intends get extra weight in the proportionality calculation only when the coercee does (more or less) precisely what the coercer intends her to do-that is, the coercee must conform to the plan, not merely bring about the desired end.

Agent-relative and agent-neutral constraints
Bazargan-Forward contrasts what he calls the agent-relative view of the constraint on intentionally killing with the agent-neutral view. On the agent-relative view, intentional harms matter in only the intender's proportionality calculation (Bazargan 2014: 14). There is no more reason to prevent someone else's unjustified intentional killing of A than to prevent someone else's unjustified collateral killing of B. On the agent-neutral view, in contrast, ''the moral relevance of the intention/foresight distinction is grounded in agent-neutral reasons to prefer the latter to the former.'' Bazargan-Forward argues that ''all agent-relative constraints generate a corresponding agent-neutral reason to promote compliance with the agent-relative constraint. That is, the agent-relative view entails the agent-neutral view.'' (Bazargan 2014: 14).
Consider a variation of Attack from Sect. 2: Rescue Attack: Terrorist lethally and unjustifiably throws a grenade at Victim, intending to kill him as a means of intimidating others. The grenade blast will also kill Bystander as a foreseen but unintended side-effect. Bystander's death will also have the effect of intimidating others. Rescuer can pull either Victim or Bystander out of the blast radius.
Bazargan-Forward argues that preventing intentional killings gives us a better world ''because agent-relative constraints are being met.'' (Bazargan 2014: 14) But if we understand the 'constraint' on useful killing as the claim that useful killing is harder to justify than collateral killing, it's unclear what it means to say that this principle is met, or satisfied, when all the prospective killing is unjustified. By preventing Terrorist from killing Victim, but allowing him to kill Bystander, Rescuer does not satisfy, or promote compliance with, the principle that it is harder to justify usefully killing compared to collaterally killing. Hence, satisfying or promoting compliance 'The moral irrelevance of moral coercion' 3473 with this principle cannot give Rescuer a reason to pull Victim, rather than Bystander, out of the way. Bazargan-Forward supports his claim that agent-relative constraints give rise to agent-neutral reasons by hypothesising two worlds, identical in all respects except that in the first, agent-relative constraints are consistently violated, and in the second they are consistently respected. Bazargan-Forward claims that if adhering to agentrelative constraints has only agent-relative value, we have no reason to prefer the second world (Bazargan 2014: 14-15). And yet the second world is clearly morally preferable.
Imagine that in Bazargan-Forward's first world, people violate agent-relative constraints by inflicting both disproportionate intentional harms and disproportionate collateral harms. This is a world in which people are causing unjustified harm. The fact of unjustified harm is of agent-neutral concern. This gives us agent-neutral reason to prefer the second world, in which people adhere to deontic principles, to the first. But it does not give us agent-neutral reason to care more about the intentionally inflicted unjustified harms than the merely foreseen unjustified harms. Thus, Bazargan-Forward has not shown that the intending/foreseeing distinction has agent-neutral significance in the way he claims, and thus has not shown that coercees ought to treat coerced harms as if they are intentional, in virtue of the fact that the coercer intends them.
We have further grounds to reject Bazargan-Forward's view once we flesh out a claim I made above-namely, that Clare in Redundancy should not prefer making another employee redundant rather than Brenda, in light of the fact that Amos intends that Brenda be fired. Consider a variation of Alley in which Bystander can divert the bomb towards either Victim, who is Villain's target, or Other Victim, in whom Villain has no interest. On the assumption that it is better to kill as a sideeffect, Bazargan-Forward's view holds that Bystander should simply divert the grenade towards Other Victim, since she has a choice between intentionally (usefully) killing one and merely foreseeably killing one. This seems mistaken. Bystander should, ideally, give Victim and Other Victim a fair chance of survival by tossing a coin. The fact that Villain aims at Victim's death does not bear on Other Victim's right to a fair chance. The same is plausibly true in Rescue Attack. Even if we grant that Terrorist has weightier reason to refrain from intentionally kill Victim than collaterally killing Bystander, this does not seem to entail that Rescuer ought to save Victim rather than Bystander when Terrorist unjustifiably threatens both. Again, Rescuer ought to give each an equal chance of rescue.
Moreover, it is, to borrow Bazargan-Forward's phrase, not infelicitous to say that Victim is unjustifiably killed in Alley, Hostage, Coerced Trolley and Duress Trolley, since in each case Villain unjustifiably manipulates the facts such that Bystander ought to kill Victim. We need not infer from this that Bystander unjustifiably kills Victim-indeed, we cannot infer this, since Bazargan-Forward and I agree (at least in Duress Trolley) that Bystander ought to kill Victim. 11 Likewise, the reason it is not infelicitous to say that Victim is intentionally killed in Alley is that Villain intends that Bystander kill Victim. It doesn't follow that Bystander intends Victim's death, nor that she should weight it as an intentional harm when thinking about what she may do. Both intention and justification are indexed to agents and their actions, rather than to outcomes.

Helping evil to succeed
Say we agree that Bystander should not treat coerced harms as intentional, and that coercion is not distinctive in making Bystander party to wrongdoing, since this also occurs in duress cases such as Duress Trolley. Nevertheless, although Duress Trolley results in wrongdoing, Bystander does not help evil to succeed in this case. She rather frustrates Villain's end, and what she does is not part of Villain's plan. In Coerced Trolley, in contrast, Bystander both brings about the end that Villain wants, and plays the role in the plan that Villain has designed for her. We might think that letting oneself be used as means, especially in the pursuit of evil, is morally significant, and makes cases of coercion morally worse than cases of duress or happenstance. If this worseness bears on what Bystander may do, this will falsify Moral Irrelevance.
Before we explore this thought, we should clarify how the moral worseness of coercion might bear on what Bystander may do. For coercion to make such a difference, we must adopt one of two views. 12 The first holds that diverting in noncoercion cases, such as Duress Trolley and Trolley, is supererogatory (that is, permissible but not required), whilst diverting in coercion cases as is impermissible, reflecting coercion's moral worseness. The second holds that diverting in noncoercion cases is required, whilst diverting in coercion cases is supererogatory (again, reflecting coercion's moral worseness). If diverting is either supererogatory in both coercion and non-coercion cases, or impermissible in both, coercion makes no difference to what Bystander may do.

Being treated as a means
Moral Irrelevance is compatible with the claim that Bystander has a stronger complaint against Villain in cases in which she is used compared to cases in which she is not. Villain treats Bystander as a means in Coerced Trolley, manipulating her for his own ends. If we require Bystander to divert, we require her to let herself be treated as a means by Villain for the sake of the five. However, Moral Irrelevance denies that this makes a difference to what Bystander may do.
To see why, consider Duress Trolley and Trolley. In these cases, saving the five requires Bystander to treat herself as a means for their sake: she must make herself useful to them. Some writers argue that this renders diverting the trolley supererogatory, rather than required (Walen and Wasserman 2012: 554). I've argued elsewhere that this is a mistake. A duty to rescue just is a duty to make oneself (or one's resources) usefully available to others (Frowe forthcoming). Thus, the mere fact that Bystander must treat herself as a means for the sake of others cannot defeat a duty to rescue: if it did, it would defeat all duties to rescue.
Moreover, and importantly for our purposes here, the only plausible version of the claim that coercion makes a difference to what Bystander may do depends on the claim that diverting is required in non-coercion cases, even though Bystander thereby treats herself as a means. Recall the two possible views of how the alleged moral worseness of coercion makes a difference to what Bystander may do. The first holds that diverting in non-coercion cases is supererogatory, whilst diverting in coercion cases is impermissible. But it is implausible that Bystander in Coerced Trolley may not allow herself to be used as a mean by Villain for the sake of the five, such that diverting is impermissible. Villain will wrong her Bystander by treating her in this way, but Bystander may decide to suffer this wrong for the sake of the five. If my earlier arguments for rejecting the dominance of Villain's intentions are sound, then we have no other reason to think that it is impermissible for Bystander to divert. Given this, proponents of the view that coercion makes a difference ought to adopt the second view that letting oneself be used by a moral coercer is supererogatory, and thus diverting in Coerced Trolley is supererogatory. To get a difference in permissions between coercion cases and non-coercion cases, then, we must hold that diverting in non-coercion cases, such as Duress Trolley and Trolley, is required.
But, as above, if Bystander is required to divert in non-coercion cases, such as Duress Trolley and Trolley, she is thereby forced to treat herself as a means. And, in Duress Trolley she is forced to treat herself as a means because of Villain's wrongdoing. If she is nonetheless required to divert in Duress Trolley, despite being wrongly forced to treat herself as a means for the sake of the five, I doubt that the fact that she will be wrongly used as a means by Villain in Coerced Trolley can make a difference to what Bystander may do, rendering diverting supererogatory rather than required. This is because it is plausibly harder to justify making people treat themselves in harmful ways than it is to justify requiring them to allow others to treat them in harmful ways. This is reflected in the fact that Bystander need not lethally divert the trolley towards herself to save five, but (as I have argued elsewhere) Victim must allow her to divert towards him to save five (Frowe 2018). 13 If so, the additional wrong that Bystander endures in Coerced Trolley-of letting herself be treated as a means by someone else-is less serious than the wrong of being forced to treat herself as a means that she also suffers in Duress Trolley. 14 13 People disagree about whether Victim may harm Bystander to save himself once the trolley has been diverted. It is less controversial that he may not prevent the saving of the five. It is certainly less controversial if one grants that Bystander is required to divert as, as I have argued, proponents of the view that coercion makes a difference must do. 14 I take the arguments in Sects. 2 and 3.1 to defeat the objection that in Coerced Trolley Bystander is used in the service of an unjustified goal (whereas in Duress Trolley she is uses herself to secure a justified goal). Killing Victim is not unjustified from Bystander's perspective, and the fact that Villain intends Victim's death does not give Bystander further reason not to kill Victim in addition to the independent reasons she has not to kill him.
Wrongs of this sort are not additive in the way that, say, suffering physical harms might be. If saving five lives suffices to justify requiring Bystander to both kill Victim and treat herself as a means in Duress Trolley, adding the lesser wrong of letting oneself be used cannot tip the balance such that diverting becomes supererogatory in Coerced Trolley. And if, as I have argued elsewhere, Victim is required to let himself be killed by Bystander for the sake of the five in Duress Trolley, it seems especially implausible that Bystander may refrain from saving the five in order to avoid letting herself be used in Coerced Trolley. What is proposed for Bystander is a much lesser wrong than is proposed for Victim in each case.
In sum, the fact that acceding to coercion involves letting oneself be used as a means in an unjust plan does not generate different permissions in coercion cases compared to non-coercion cases, and does not undermine Moral Irrelevance.

Exploiting moral constraints
A common objection to acceding to moral coercion is that, by acceding, one allows Villain to manipulate or exploit morality (e.g. Zohar 2014: 168. For similar legal arguments, see Canestaro 2004: 461;Dinstein 2016: 185). Villain exploits Bystander's commitment to doing the right thing in order to do the wrong thing himself. I've set aside here the question of whether acceding to coercion encourages further coercive acts, and thus whether there are consequentialist reasons to resist moral coercion. But there might be non-consequentialist objections to allowing wrongdoers to profit from others' moral commitments in this way. If one may resist bringing about coerced harms, but not equivalent non-coercive harms, this would undermine Moral Irrelevance. Call this the Exploitation Objection.
Although the Exploitation Objection is appealing, I do not think it is well grounded. If we generally refuse to allow others to exploit morality, we commit ourselves to implausibly permissive results. Consider self-defence. Unjust attackers can sometimes orchestrate things so that an innocent victim's using defensive force is impermissible. One might, for example, deliberately attack someone in circumstances in which she has only disproportionate means of defence at her disposal, thereby hoping to exploit her moral commitments. The most plausible version of the Exploitation Objection holds that one may use somewhat greater force against such attackers compared to those who do not orchestrate things in this way, given our stronger reason to prevent this kind of exploitation. Thus, using otherwise disproportionate means is permissible against an exploitative attacker, but not against a spontaneous attacker, or merely lucky attacker. This seems implausible. Whether an unjust attacker has given any thought to their victim's available means of defence seems irrelevant to whether their victim may defend herself. Other things being equal, it is no less bad to attack a person when one is merely lucky that they have only disproportionate defensive means than to attack a person where one has ensured that they have only disproportionate defensive means. Refraining from defence in either case allows the attacker to benefit from one's moral commitments, and whether the attacker intended to so benefit does not bear on the seriousness of the wrong the victim suffers. But if we grant that 'The moral irrelevance of moral coercion' 3477 victims may also use disproportionate means against merely lucky attackers, in order to treat these cases alike, we are effectively rejecting the proportionality constraint on self-defence. Given the counterintuitiveness of that result, we have good reason to deny that concerns about exploitation generate special rights to resist moral coercion that we lack with respect to duress or happenstance.
6 Coercion, voluntariness, and accountability Elinor Mason argues that coercion, being the result of another agent's interference, undermines the voluntariness of coercees' actions in a way that does not arise in cases of duress or happenstance (Mason 2012: 196). Voluntariness is a necessary condition of accountability, which includes, amongst other things, one's liability to punishment, criticism, or compensation. 15 Thus, Mason concludes that coerced agents are less accountable for their actions, and their bad effects, than agents who face otherwise identical choices as a result of duress or happenstance (Mason 2012: 184). Mason aims to explain two intuitively plausible claims. First, a coercer is typically accountable for harms that result from their coercion. Second, a coercee is typically not so accountable (Mason 2012: 202). The first claim is relatively easily explained: the coercer is accountable because she has responsibly brought about the situation in which the harmful actions are performed. It is harder to explain why coercees are not also accountable, given their deliberate performance of harmful actions that contribute to coercers' unjust plans.
Mason frames her account of coercion as an analysis of Bernard Williams' famous Jim and the Indians case: Indians: Jim happens across Captain, who is about to unjustly kill twenty captured Indians. Captain tells Jim that if Jim kills one Indian, Pedro, Captain will let the other nineteen go.
Although Mason does not describe it as such, Indians is a case of moral coercion. Captain manipulates the facts in order that, to avoid the morally worse outcome, Jim must facilitate Captain's end of unjustly killing Pedro. As Mason puts it, Jim ''does what he ought'' when he accepts Captain's proposal (Mason 2012: 186). 16 For ease of exposition, I'll assume that Mason would also grant that if Coerced Trolley is indeed a case of moral coercion, this entails that Bystander ought to divert, since moral coercion just is for the coercer to set up the coercee's reasons such that they ought to accede.
Mason argues that Jim is dominated by Captain's coercion, and thereby alienated from his lethal action. Alienation, Mason argues, is ''the feeling that one is not in control of one's actions-it may or may not be related to a fact of the matter about whether one is.'' (Mason 2012: 195) When one is dominated, one's actions are not voluntary, and thus one is not accountable for them. One's action are not involuntary-one is still is an agent-but rather non-voluntary. Mason argues that the mere limiting of one's options does not render an action non-voluntary. Rather, what matters is the feeling that someone else is choosing for us. As she argues, ''the fact that another agent set our reasons up (rather than blind forces) is crucial […] Compatibilists agree on the crucial point illustrated by Frankfurt's examples, that whereas determination by the causal history of the universe does not interfere with moral responsibility, being determined by another agent does.'' (Mason 2012: 196) Note, then, that the deliberate setting up of the coercee's reasons is key on Mason's account. Alienation is created by the fact that the coercer ''is treating the coercee like an object, in that she has treated the coercee's reasons responsiveness as a feature to be manipulated'' (Mason 2012: 196). This prevents cases such as Duress Trolley from counting as coercion. Even though Bystander's reasons for action in Duress Trolley are generated by Villain's wrongdoing, Villain is not treating Bystander in any way at all, much less trying to get Bystander to do something by manipulating her reasons responsiveness. Rather, Bystander acts contrary to Villain's plans, frustrating his end.
By way of support for moral significance of this distinction, Mason points out that it is appropriate to resent a coercer, but not blind forces, just as one can appropriately resent an intentional, but not an accidental, kick. Mason argues that it is ''thus part of our concept of moral responsibility that responsibility is undermined by the interventions of other agents.'' (Mason 2012: 196) Other agents, having wills of their own, can ''take the place of 'self' in self-determination'' (Mason 2012: 196). But blind forces cannot, and so cannot dominate an agent in the voluntarinessundermining way (Mason 2012: 196).
If Mason is correct, this undermines Moral Irrelevance, which holds that whether a person faces a choice between harms as a result of coercion, duress or happenstance has no bearing on the degree of an agent's agential responsibility for the harms she inflicts or allows, nor on her accountability. In what follows, I argue that we should reject Mason's account of voluntariness. I then offer an alternative explanation of why coercees are not accountable for the harms they inflict.

Resentment and voluntariness
It is appropriate for a coercee to feel resentment that the range of her permissible options has been unjustifiably limited by a coercer, such that she is obliged to save the five, or obliged to kill Victim. But it is similarly appropriate for Bystander to feel resentment in Duress Trolley. Here too, Bystander is wronged by being unjustifiably forced to make herself useful to the five, and to kill Victim. And yet, this case does not satisfy Mason's criteria for undermining voluntariness. Villain does not deliberately manipulate Bystander's reasons, treating her reasonsresponsiveness as a feature to be exploited for his own ends. Villain makes no 'The moral irrelevance of moral coercion' 3479 coercive proposal at all, and killing Victim does not further Villain's end. On the contrary, diverting frustrates Villain's end of killing the five. Given that resentment is appropriate in each case, the appropriateness of resentment in coercion cases does not support Mason's view that one's capacity for self-determination, and hence the voluntariness of one's action, is specially undermined by coercion-by the deliberate setting up of one's reasons. In neglecting the category of non-coercive wrongdoing (i.e. duress), Mason overlooks a much more straightforward explanation of when resentment is appropriate, and why. Appropriate resentment tracks whether one is wronged, rather than whether one's agency is diminished. This is why resentment is inappropriate in a blind forces case, such as Trolley. Resentment does not, then, pick out a special relationship of domination or agential interference, but merely the fact of wronging. Thus, the fact that one can appropriately feel resentment for being deliberately put in one's position does not entail that one's action is less voluntary compared to when that position arises through happenstance or duress.
Of course, as per the discussion in Sect. 4.1, Bystander is plausibly entitled to feel greater resentment in Coerced Trolley than in Duress Trolley. In Coerced Trolley, Bystander suffers the further wrong of being required to allow herself to be treated as a means by Villain. However, given that, in both cases, Villain's wrongful actions make it impermissible for Bystander to do anything other than kill Victim, I doubt that Bystander's action is less voluntary in one case rather than the other, if we hold everything else equal. The voluntariness of one's action seems to be a function of one's commitment to the end, combined with (one's belief in) the range of (permissible) alternatives. 17 Insofar as Bystander is equally committed to saving the five in all three cases, and believes that she lacks permissible alternatives, each diversion seems equally attributable to her agency.
Mason's account implies that the more one's action reflects someone else's will, the less it must reflect one's own, and thus the more alienated one is from the action and the less voluntary the action becomes. This seems like a mistake. Reflecting will is not zero-sum. I can fully and voluntarily perform an action that someone else wants me to perform. I can fully and voluntarily perform an action even if I perform it merely because someone else wants me to do so. I might go to the doctor solely to alleviate my partner's unfounded concerns, treating their request that I go as giving me content-independent reason to go. But I would not be acting non-voluntarily, even though my acting is substantially reflective of my partner's will, and I would be fully accountable for my actions. Likewise, the fact that Bystander's diverting the trolley is not part of Villain's plan in Duress Trolley doesn't mean that diverting more robustly reflects Bystander's will than does diverting in Coerced Trolley.
I suggest, then, that domination-being forced to act in service of another's goal-does not bear on voluntariness in the way Mason supposes. Insofar as, in each case, Bystander has no morally permissible alternative to diverting the trolley, Bystander's killing of Victim need be no more voluntary in Duress Trolley or Trolley than in Coerced Trolley.

Accountability
If we reject the view that coercees' actions are necessarily non-voluntary, how can we explain coercees' lack of accountability-that is, their immunity from punishment, blame, compensation claims and so on? I suggest that in cases of moral coercion, coercees' lack of liability to these costs is explained by the fact that they act with objective moral justification. After all, Mason and I agree that moral coercion works by manipulating the facts such that the coercee ought to accedethat is, by adjusting the balance of moral reasons such that Jim ought to shoot Pedro, and the trolley ought to be lethally diverted towards Victim. That they act with objective justification explains why coercees are not liable to either ex post or ex ante cost with respect to the harms they impose (McMahan 2008;Frowe 2018).
This lack of accountability-understood as the absence of liability to bear costs, and to criticism or blame-is not a distinctive feature of coercion cases. Bystander is not liable to bear costs in any of the Trolley cases, since she acts with objective moral justification in each. Other things being equal, she is equally agentially responsible for the harms she imposes whether she acts as a result of coercion, duress or happenstance, and may be fully agentially responsible for her actions. It is her lack of morally permissible alternatives that undermines the usual connection between responsible agency and accountability. Responsible agency is a necessary condition of accountability, but it is clearly not sufficient. Accountability is precluded if the agent is either morally required to act as she does, or the alternatives are so costly to her as to be supererogatory.

Conclusion
I have argued that the widely-held view that moral coercion bears on what an agent ought to do, the voluntariness of her actions, and/or her accountability for those actions, is mistaken. Whether one faces a choice between harms as a result of coercion, duress, or happenstance is irrelevant to each of these considerations. Coercees should not treat coerced harms as intentional. Nor do coercees have rights to resist being used as a means that permit them to refraining from imposing harms that they would be required to impose in comparable cases of duress or happenstance. If the duty to rescue is not defeated by the harm one will impose, nor by being forced to treat oneself as a means, it cannot be defeated by the lesser wrong of allowing oneself to be used as a means. The claim that acceding to coercion allows coercers to exploit morality is also unpersuasive. So too is the view that coerced actions are, necessarily, less voluntary than non-coerced actions, and that this best explains coercees' lack of accountability. This lack of accountability is best explained by the fact that those who act with objective justification are not accountable for the harm they inflict. This explanation holds across cases of coercion, duress, and happenstance.
War and Peace, Stockholm University and the Yeoh Tiong Lay Centre for Politics, Philosophy and Law, King's College, London), and the Annual Conference of the Society for Applied Philosophy in Cardiff. I am grateful to those audiences for helpful discussion. Thanks also to Tom Douglas, Saba Bazargan-Forward, Kieran Oberman, Victor Tadros, Jonathan Parry, Massimo Renzo and two anonymous referees for extremely helpful written comments, and to Elinor Mason for helpful correspondence.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.
Funding Open Access funding provided by Stockholm University. Knut and Alice Wallenberg Foundation, Grant Number 1521101.