II. THREE CONTEMPORARY ACCOUNTS OF PERMISSIBLE SELF-DEFENSIVE KILLING

Judith Thomson has influentially argued that if A would kill V unless she is killed, and by doing so A would violate V's right not to be killed, A loses her right not to be killed. As a result, V may permissibly kill her in self-defence.Footnote ³ Under her approach, because CA, RA, IT, NRA and NRT would all violate V's right not to be killed, they forfeit their own right not to be killed. Thomson argues that although the fault of the aggressor may be relevant for certain purposes, e.g. what you may do to defend yourself, it is irrelevant to decide whether you can defend yourself ‘when what is at question is his life against yours’.Footnote ⁴ Furthermore, she argues that the permissibility of V killing A in self-defence ‘goes hand in hand with the impermissibility of’ A fighting back, and that those two phenomena have a common source, namely, A's violation of V's rights. According to Thomson, then, the fact that A is causally threatening V's life, or very important goods, is a sufficient condition for A to be liable to being killed in self-defence by V.

Reacting against Thomson's framework, Jeff McMahan has defended an account of defensive killing which relies on the idea of moral responsibility. To be liable to be killed in self-defence it does not suffice that A causally poses an (objectively) unjustified threat to V; she must also be morally responsible for posing that threat.Footnote ⁵ Thus, under his approach both moral responsibility and the existence of an objectively unjustified threat are necessary and sufficient conditions for liability to defensive killing.Footnote ⁶ Liability again means that the attacker herself lacks the right not to be killed: V would be at liberty to kill A and A would be under a duty not to fight back. Accordingly, McMahan claims that NRA and NRT are not liable to being killed by V in self-defence, while IT is. This is hardly surprising. In so far as the liability of CA and RA is explained on the basis of the fact that they have both forfeited or lost at least some of their rights vis-à-vis V, and moral responsibility is a necessary condition for liability to defensive lethal force, NRA and NRT cannot be liable to defensive killing. This, in turn, means that V would be under a duty not to use lethal force against them, as this would entail violating their rights. If V defended herself she would become herself liable to being killed by NRA and NRT. In sum, the person falling on a victim through no fault of his own is not liable to being killed by his victim. Her victim is under a duty to let herself be killed. This, McMahan concedes, may be counterintuitive, but this is a bullet he is prepared to bite.Footnote ⁷

Some recent accounts defend an agent-relative justification for killing in self-defence which stands separate from the agent-neutral liability justifications, such as those of Thomson and McMahan. Jonathan Quong, for instance, suggests that each person has a ‘powerful agent-relative permission to avoid sacrificing or significantly risking their own life for the sake of others (absent any obligations voluntarily incurred)’.Footnote ⁸ As a result, he claims that neither NRT nor her victim is liable to being killed; neither of them has forfeited any of their rights. This, however, does not mean that it would be impermissible for any of them to kill the other in self-defence. The fact that they both retain the right to life does not mean that their action is necessarily impermissible. Permissibility in this type of case can be defended on the grounds that each of them holds an agent-relative prerogative.Footnote ⁹ Accordingly, this personal prerogative leads him to defend a set of more demanding conditions for liability to lethal killing.Footnote ¹⁰

I will assess the personal prerogative as a plausible justification for permissibly killing another individual in section V below. At this juncture I am more interested in the conceptual structure that Quong's argument presupposes. In particular, I want to focus on the distinction between Y being liable to being killed by X and X being merely at liberty to kill Y. As we have seen from the main arguments above, liability to defensive killing usually means that killing an individual would neither wrong her nor violate her rights.Footnote ¹¹ This proposition is often explained on the grounds that Y has lost or forfeited her claim-right not to be killed.Footnote ¹² But being liable to being killed certainly means more than this. For one, the fact that Y is liable to being killed by X standardly means that Y is under a duty not to fight back – she lacks the liberty to defend herself against X. Accordingly, it has been suggested that Y's liability means that X has a right to kill Y.Footnote ¹³ Furthermore, if X kills (liable) Y she would neither owe compensation to her estate, nor owe her any type of apology or acknowledgement of the sacrifice Y had to endure for X's sake. In sum, liability to defensive killing is standardly characterized as a situation of total moral asymmetry between individuals. This is important, because Thomson and McMahan's accounts, as well as much of the relevant literature, rely heavily on this binary alternative liable/non-liable. I find this ‘all-or-nothing’ solution implausibly rigid to adequately capture the nuances in NRT, NRA and IT.Footnote ¹⁴

By contrast, we may also conceptually account for permissible defensive killing without recourse to liability. We may argue that X is merely at liberty to kill Y even if Y has not forfeited or lost her right not to be killed. To illustrate this conceptual possibility we may recall that under most accounts of just war theory, a pilot X is at liberty to kill a proportionate number of non-liable non-combatants as a side-effect of an attack against a military objective. Unlike the liability framework described above, this situation allows for and in fact presupposes (though it is not by any means committed to) some sort of pro tanto symmetry between the two agents. Accordingly, X being merely at liberty to kill Y is compatible with Y not having lost her right to life. Furthermore, this entails that there would be some form of moral residue in X killing Y. This insight is usually explained through the notion that she would be (permissibly) infringing Y's rights.Footnote ¹⁵ This moral residue may entail that some form of compensation or apology is owed to Y, or at least it should be acknowledged that her rights were infringed.Footnote ¹⁶ In sum, to say that X is liable to being killed is conceptually very different from saying that Y is merely at liberty to kill her.Footnote ¹⁷ With this analytical distinction in mind, let us re-examine some of the leading accounts of self-defensive killing available in the literature.

III. AGAINST THE CAUSAL RESPONSIBILITY ACCOUNT: THE CASE OF MULTIPLE NRTS

Thomson's causal responsibility approach has received several criticisms in the literature. Many people believe that given that NRT is not exercising any form of agency, she cannot be violating V's rights any more than a tiger or a hurricane could be said to do so.Footnote ¹⁸ Accordingly, it would be at least odd to say that she has forfeited some of her rights. I think this criticism is compelling. However, I want to press a different line of argument here. I suggest this approach cannot successfully deal with a case of multiple NRTs. Let us consider the following example.

Most people would say that these 50 CAs are liable to being killed by V. Furthermore, this conclusion would most likely not change even if we raise the number of attackers to 100,500, or even 1,000.Footnote ¹⁹ The fact that they are liable to being killed means that they simply lack a right not to be killed. Accordingly, V is at liberty to kill them (not under a duty not to do so) but also, because she has not lost her right to life, she cannot be prevented or interfered with their killing, either by them or by a third party in their defence. More importantly, the 50 CAs are under a duty not to fight back. Finally, by killing them V is not infringing their rights, so she does not owe them any type of compensation or apology.

Consider now the case of 50 individuals who are only causally responsible for the threat against V:

According to Thomson's account these 50 NRTs would be liable to being killed by V. This seems to follow from the claim that by killing him they would violate his right to life. As a result, V would not only be at liberty to kill them; they would also be under a duty not to kill V pre-emptively. This conclusion, however, seems highly unpersuasive.

There are at least three ways in which Thomson could resist this conclusion. First, she may suggest that it need not follow from the premises. Namely, although under the causal responsibility account these NRTs would be liable to being killed, it would be impermissible for V to kill them on, e.g., lesser evil grounds. However, even if we were to concede that this move would not necessarily entail that it would be also impermissible to kill the 50 killers, something which would require further argument, this response would have an important flaw. The claim that the 50 passengers are liable to being killed implies that they lack the right not to be killed by V. Thus even if it is true that V would act impermissibly all-things-considered, Thomson's account would seem committed to the conclusion that the rights of the 50 passengers would not be infringed in any way, that they would not be wronged. I find this necessary implication of Thomson's view implausible. And the reason for this is precisely that they do not seem to have lost or forfeited their right not to be killed.

Second, Thomson could argue that what is doing the work here is the issue of numbers. But numbers in this case are used to show what is more difficult to see in the case of a single threat. Put differently, the relevant comparison is between the 50 killers (liable) and the 50 passengers in Tram (non-liable). Furthermore, the fact that rights are essentially non-aggregative should indicate that numbers do not necessarily make a difference to individual liability if we are not arguing on consequentialist or lesser evil grounds. Defenders of the causal responsibility position characteristically do not argue on such grounds, nor, I suggest, can they coherently do so. Accordingly, Tram seems a plausible heuristic device to show that one NRT is not liable to being killed.

Finally, defenders of the causal responsibility approach may suggest that considering a single NRT non-liable would also be unpersuasive. This would be equating her to the innocent bystander, and this too is deeply counterintuitive. I agree. Nevertheless, this proposition is based on the ‘all-or-nothing’ approach, which I suggest we need to abandon. Put differently, the fact that NRT is not liable to being killed need not entail that V lacks a liberty to kill her, that is, that V is under a duty to let herself be killed by a single NRT. Admittedly, more argument is needed to fully account for this claim. I will provide a framework to make sense of this proposition in section V below. For now it may suffice to acknowledge that this move is not readily available against my criticism in so far it is theoretically consistent to argue that V is at liberty to kill NRT even if we admit that NRT is not herself liable to being killed.

IV. AGAINST THE MORAL RESPONSIBILITY ACCOUNT: THE CASE OF MULTIPLE ITS

I am certainly not alone in arguing that the causal responsibility approach of liability to defensive killing is over-inclusive.Footnote ²¹ A reasonable move within the ‘all-or-nothing’ approach is to make the conditions for liability more demanding, and a plausible way to do this is to require A to be morally responsible for the threat she poses. Under this view, although NRTs are not liable to being killed by Vs, innocent threats (ITs) are. Take the often-cited case of conscientious driver (IT) described at the outset. McMahan has influentially argued that this IT is liable to being killed by V. He suggests that IT is ‘liable because [s]he voluntarily engaged in a risk-imposing activity and is responsible for the consequences when the risks [s]he imposed eventuate in harms’.Footnote ²² He recognizes that what he considers as the relevant moral difference is ‘of comparatively slight moral significance’, but that given the costs of the driver's actions ‘cannot be divided, it is fair that he should suffer them all rather than impose them all on the pedestrian’.Footnote ²³

It is easy to see that if one takes the latter proposition as a given, our intuitions back McMahan's conclusion. But this choice is based on the ‘all-or-nothing approach’, i.e. the centrality of liability in his account of defensive killing. By contrast, if one accepts that considering IT non-liable does not necessarily mean that V would have to endure the full cost of the situation, McMahan's point seems to lose much of its purchase power. In fact, many people would intuitively agree that putting IT under a duty to let herself be killed simply because she decided to take a ride in her well-maintained car is a slightly draconian understanding of what morality requires from us. I therefore join Ferzan, Rodin, Bazargan and Tadros in considering it implausible to think that IT has forfeited or lost her right to life after having done what can reasonably be asked from her.Footnote ²⁴ Nevertheless, the fact that different people have different intuitions on this particular case should alert us against being too confident on what intuitions tell us.Footnote ²⁵ Accordingly, we need to resort to a compelling theory of defensive rights in order to reach firmer conclusions vis-à-vis the moral situation of IT. Again, I think a fruitful way forward is to look at situations with multiple threats.

Suppose it is not a car as in IT but a bus that needs to make an emergency stop. Take, for instance,

Just as in our initial IT case above, the 50 passengers have knowingly created a minor risk for V and due to a freak event they are now directly threatening V. If V is to save his life he would have to kill the 50 passengers in the bus. Would we still be prepared to accept that the 50 ITs are liable to being killed by V, i.e. that he would not wrong them if he killed them? I think not and, again, I suggest the better way of accounting for this conclusion is to argue that IT is not liable to being killed by V either.

An advocate of the moral responsibility account could complain that 50 ITs being non-liable to be killed by V is entirely compatible with both 50 CAs and one IT being so liable. Nevertheless, I am not sure this move is available to McMahan. Admittedly, an important aspect of his approach is that the degree of an agent's responsibility has a direct impact on the degree of her liability. McMahan distinguishes between culpable threats, partially excused threats, excused threats and innocent threats.Footnote ²⁷ All of them are, in turn, distinguished from NRTs, that is, a person who without justification causally threatens to harm someone (V) in a way to which she is not liable, but who is in no way morally responsible for so doing. This would explain why a single CA may be liable to being killed if she threatens V with chopping off her hand – a non-deadly threat – but an IT would most likely not.Footnote ²⁸ Similarly, their different degrees of responsibility may also account for 50 CAs being liable to be killed, while 50 ITs are not so liable.

Nevertheless, distinguishing between different degrees of responsibility does not translate to the issue of numbers too readily, at least not in the way McMahan's argument would need. Suppose CA is hiding behind five innocent bystanders. Although we would normally say that it would be impermissible for V to kill CA if he can only use a bazooka that will kill all six of them, we would maintain that CA is liable to being killed by V. Rights are typically individualistic creatures – at least the right to life is; they capture the relationship between two concrete individuals. Thus, the fact that there are innocent people next to a culpable attacker who are not liable to being killed does not mean that the culpable attacker is or becomes herself non-liable, but only that it is impermissible to shoot at her (despite her being liable). If V were to shoot and due to sheer good luck only killed the culpable attacker he would not wrong her or violate her rights. This holds, even if we admit that she should probably be prosecuted for impermissibly endangering the five bystanders.

Recently, McMahan has introduced the notion of ‘proportionality in the aggregate’ to sort out this type of difficulty. Proportionality in the aggregate, he claims, is sensitive both to the level of responsibility of the attacker or threat, and to the numbers involved. When

Now given that McMahan claims that proportionality is internal to liability (and proportionality in the aggregate would be internal too) this accounts for the fact that whereas one IT would be liable to being killed by V, 50 ITs would not.

I find his division of labour unconvincing. Let us go back to the case of a culpable assassin (CA) who uses five innocent bystanders as shields. If V has no other way to kill CA than by also killing the five shields, I agree with McMahan that it would be impermissible for V to kill CA. However, contra McMahan it would be odd to suggest that CA would herself become non-liable to being killed by V. Such a position would entail, implausibly in my view, that she (CA) would be wronged by V if V were to shoot; that V would violate her rights. McMahan could object that by including bystanders I am ‘aggregating’ apples with oranges, in so far as bystanders are clearly not liable to defensive killing.Footnote ³⁰ But I see no reason why we can ‘aggregate’ ITs but not ITs, CAs, Bs, and so on.

Let us take a situation in which each individual may be considered prima facie liable to defensive killing, or at least satisfies the agential conditions for liability. Namely, suppose CA is now the driver of a bus carrying 50 ITs, and that she culpably seeks to kill V by running over her. In this case, proportionality in the aggregate would again be committed to the view that if the only way V can defend herself is by blowing up the bus, CA would not be herself liable to being killed by V.Footnote ³¹ Again, the fact that it is implausible to say that CA has herself a right against V that she not shoot at her, that CA will herself be wronged by V if she were killed, makes this suggestion unconvincing.

Finally, McMahan may retort that the latter two implications need not follow from his initial premise. He has argued that even if CA is not liable in this type of case, it may still be the case that she is not wronged by V if she is killed.Footnote ³² The reason for this is that killing CA may achieve some good effects independently of the physical defence of the victim. For instance, it may allow V to assert her dignity or moral status in the face of the wrong done to her and deter others. Similarly, he claims, many people would accept that CA deserves some harm, and that she may be liable to some harm even if she is not liable to being killed.

This suggestion seems ad hoc. But more importantly, I think it is unconvincing for three reasons. First, this solution would make liability largely irrelevant in a significant number of situations, something which is at odds with McMahan's overall approach to defensive killing. To argue that asserting one's dignity or moral status, coupled with the fact that an individual deserves some harm, or is liable to some harm short of death, suffice for her not being wronged if she is killed would seriously undermine the central role that liability plays in accounting for the permissibility of defensive killing. Second, to claim that CA is not liable in this situation means that she has a prima facie right not to be killed by V. I very much doubt that these further goods would suffice to override the value of such a right under most circumstances. This suggestion would ultimately be incompatible with the strength we typically ascribe to rights in moral and legal discourse. Finally, and perhaps decisively, I suspect most people will think that CA has forfeited her right and that the reason it is impermissible for V to shoot at the bus has nothing to do with her rights being violated.

In sum, I suggest that McMahan cannot have it both ways – i.e. one IT liable to being killed but 50 ITs not. Given that he acknowledges that the plight for IT's liability is quite weak, and that we may assume that 50 ITs being so liable is a rather implausible position to sustain, we may have to consider IT not liable after all.

V. A QUASI-SYMMETRICAL FRAMEWORK

In the previous two sections I have provided a set of examples meant to show that it is theoretically problematic to claim that NRT or IT are liable to being killed by V. I have thereby argued that the causal and the moral responsibility accounts of liability to defensive killing fit well neither with the individualistic character of moral rights, nor with the strength and function we standardly ascribe to them. I will not pursue the issue of liability to lethal force any further here. The purpose of this article is not to identify a set of necessary and sufficient conditions for liability to defensive killing, but rather to suggest that liability plays no relevant part in sorting out this type of situation.

In section II I rejected that considering NRT or IT non-liable necessarily entails that V is under a duty to let herself be killed. This type of reasoning is based on an ‘all-or-nothing’ approach which I have argued seems implausibly rigid. I will now argue that NRT, NRA, IT and V belong in a more granulated framework; that they are in a pro tanto symmetrical position. This suggestion is not particularly new. Agent-relative accounts characteristically argue, at least vis-à-vis NRT, that each individual (NRT and V) has an agent-relative permission to avoid sacrificing or significantly risking her own life for the sake of others (absent any obligations voluntarily incurred).Footnote ³³ My main claim of innovation is to suggest that the area in which we should talk of pro tanto symmetry is wider than this position admits; that it is not solely governed by this prerogative, but rather by other sorts of considerations which warrant far more attention than they have received. Put differently, it seeks to relocate an important part of the discussion on permissible killing by identifying as the point of departure a situation of pro tanto moral symmetry between the relevant agents, at least with regard to none of them having lost the right to life. But before I present this account, let me succinctly clarify where the problem with agent-relative accounts lies.

As indicated above, Quong has argued that if neither NRT nor V is liable to being killed, it would be permissible for each to kill the other in self-defence on the basis of an individual’s personal prerogative.Footnote ³⁴ This prerogative would override the other's right not to be killed. However, as McMahan has persuasively argued, conceding this type of role to the personal prerogative proves too much. The reason for this is that it would end up conferring on V a liberty to kill an innocent bystander (B) as a means of self-preservation.Footnote ³⁵ If the personal prerogative suffices to provide a permission to kill someone who is not liable to defensive killing, advocates of such view would need a further argument about why this does not include bystanders whose death is necessary for V to save herself. Quong has sought to resist this objection by reference to the moral difference between ‘eliminative’ and ‘opportunistic’ killing.Footnote ³⁶ Unlike eliminative killing, opportunistic killing entails a situation in which someone's death is not itself necessary for V's survival. Given that we may standardly accept that killing someone opportunistically is morally worse than killing her eliminatively, and that V would kill NRT eliminatively whereas she would kill B opportunistically, this move may allow the personal prerogative to account for the permissibility of killing the former and the impermissibility of killing the latter.Footnote ³⁷

Nevertheless, this distinction does not fully dispel McMahan's initial concern. Put differently, his objection need not entail that it is permissible to kill any innocent bystander, but only some of them. McMahan illustrates this point by reference to the following example:

If V were to cross the bridge he would be killing B eliminatively, and not just opportunistically. In McMahan's words, ‘shaking [B] off seems clearly the elimination of an obstacle rather than the strategic use of her resource for one's own benefit’.Footnote ³⁸ Yet if it is permissible for V to give her personal interests priority over those of others, it is not clear on what grounds Quong can consistently argue that it would be impermissible for her to cross the bridge. Accordingly, this example shows that Quong's resort to the personal prerogative commits him to the permissibility of V killing (at least) such an innocent bystander.Footnote ³⁹ Given how strong our intuition that it is impermissible to kill innocent bystanders in self-defence is, this would suffice to reject Quong's agent-relative proposal.Footnote ⁴⁰

I believe there is a more appealing way to sort out cases of permissible defensive killing amongst non-liable individuals than merely resorting to an agent-relative prerogative. A standard way to sort out this type of situation in law and morality is by recourse to the idea of conflicts of prima facie rights.Footnote ⁴¹ Namely, it is often admitted that prima facie rights exist outside of particular circumstances and they are bound to conflict with other prima facie rights. This obtains in many different contexts, such as the possible conflict between the right to freedom of speech and the right to someone's integrity, or between the right of property over a particular good and the right of someone else to destroy that property to protect a particularly important good of hers.Footnote ⁴² Accordingly, once we examine the concrete situation we may assign one of the parties an all-things-considered right, liberty, and so on, but we would standardly claim that the other person's prima facie right is neither lost nor violated, but (permissibly) infringed.

I suggest that this framework helps illuminate the underlying moral landscape in the cases of NRT, NRA and IT. Let us consider NRT first. You recall, in this case A is causally threatening B through no fault of his own and the only way for B to save herself is to kill A. Consider now the following situation:Footnote ⁴³

This is a tragic situation: whoever presses the green button first will kill the other. I believe that most people will agree that it is permissible for both A and B to press their green button.Footnote ⁴⁴ Neither of them can invoke a sufficiently strong reason for the other one to let herself be killed. Furthermore, their situation is not sorted by reference to lesser evil considerations. That is, the reason why A and B can each decide to save herself is best accounted for on grounds of the moral strength of their right not to be killed coupled with their agent-relative personal prerogative. For it is plausible to argue that in this type of situation it is permissible for each of them to give greater weight to her interests than those of the other person (much the same as most people will agree that, as the old joke goes, A and B can both permissibly seek to outrun the other when a bear is chasing them). The way to account for this situation would therefore be that each one's interest in not being killed together with their personal prerogative would suffice to give each one of them an all-things-considered liberty to press the green button (in the sense that neither of them is under a duty not to do so).

Interestingly, though, the fact that neither A nor B is liable to being killed need not lead to an all-things-considered moral symmetry between them in every situation. Consider now:

I do not know of any plausible account of right's forfeiture that can consistently argue that A forfeited her right to life in the former case, but the five people have not forfeited theirs in the latter. As per my argument in section III, I assume that none of the relevant individuals in these cases is liable to being killed. Notwithstanding this fact, most people will argue that whereas it is permissible for B to kill A in Age difference, it is impermissible for her to kill the five individuals in Numbers. By the same token, we would suggest that it is impermissible for A to ‘fight back’ in Age difference and for B to ‘fight back’ in Numbers. Arguably, the reason for this is that the harm that B and the five individuals would suffer in Age difference and Numbers, respectively – a harm they have the right not to suffer – makes their killing impermissible. Or better put, that A's interest in staying alive (even if coupled with her agent-relative prerogative) is not sufficiently important to allow her to override B's interest in not being killed in Age difference, whereas B's interest in staying alive (coupled with his agent-relative prerogative) is not sufficiently important to override the interest of the five individuals in Numbers.

These examples lead us to three preliminary observations. First, the proposed explanation simply acknowledges that rights are not the ultimate source of moral value, but rather work as middle-level reasons which help us tackle difficult philosophical issues. In Raz's terms, they ‘belong to the ground level of practical thought in which we use simple-to-apply rules’.Footnote ⁴⁵ When we face a situation of conflict between prima facie rights we need to focus on the underlying moral considerations. Second, these cases illustrate that it is possible that conflicts of rights lead to one of the parties being (merely) at liberty all-things-considered to kill the other(s), while others lead to her being under a duty all-things-considered to let herself be killed. Each of these possibilities is compatible with none of them being liable to being killed. Finally, Killings shows that conflicts of rights are not exhausted by lesser evil considerations. Or better put, lesser evil considerations need not play a relevant role in explaining why it is permissible to kill a non-liable individual. This normative work is to a significant extent performed by the strength of moral rights (much as liability, i.e. the lack of such moral rights, plays a critical role in the situation of CA and RA).

Furthermore, I suggest that this account of conflicts of rights can readily sort out McMahan's criticism against agent-relative accounts of defensive killing. Namely, he plausibly complains that Quong's personal prerogative would entail allowing V to kill at least some innocent bystanders. But this criticism does not affect the conflict of rights framework I advocate here even if it makes room for the personal prerogative. The reason for this is precisely that the personal prerogative is but one consideration in this broader framework, and arguably not the one doing most of the normative work. To see this, let us go back to McMahan's case of Bridge. Unlike the situation in Killings, where the conflict of rights was between A and B's prima facie right not to be killed by the other, the conflict between the prima facie liberty of the person fleeing the threat (A) and the prima facie claim-right of the bystander (B) sitting on the bridge would be equivalent to a conflict between A's prima facie claim-right to be saved by B and B's prima facie claim-right not to be killed by A. Following the moral difference between killing and letting die it seems clear that all else equal the latter would override the former, i.e. that it would be morally impermissible all-things-considered for the fleeing victim to cross the bridge and kill the innocent bystander even if she had no other way to survive.Footnote ⁴⁶ Accordingly, unlike the case of NRT, B stands apart in a pro tanto asymmetrical position vis-à-vis V.Footnote ⁴⁷

In sum, this framework allows for both symmetrical and asymmetrical positions while resorting neither to the notion of liability nor exclusively to the personal prerogative. It neither falls back on mere lesser evil considerations. Instead, it takes the strength and conceptual features of moral rights as its central underlying consideration. Admittedly, sorting out conflicts of prima facie rights can get messy. But this is not a problem for my account in so far as my main purpose is to show that these cases are better captured by this type of analysis than by the ‘all-or-nothing’, liability-centred framework.

Before concluding, let me show how this framework can help us provide a more compelling account of IT.Footnote ⁴⁸ In the previous section, I argued that considering IT liable to being killed is theoretically problematic in so far as it is difficult to consistently make the case that one IT is liable to being killed but 50 are not. I will now take this point further. Namely, I suggest that even if we concede that moral responsibility affects the way in which we should sort out conflicts of prima facie rights, this hardly entails that IT is liable to be killed by V. For one, even if, for the sake of argument, we grant that IT and V are in asymmetrical moral positions all-things-considered, i.e. that IT is under a duty to let herself be killed by V, it still does not follow that she is liable to being killed. The better view is that if V were to kill IT she would be (permissibly) infringing her rights, much as in Killings, Age difference and Numbers above. Namely, even if we agree that V is at liberty to kill IT whereas IT is under a duty not to fight back, there will be some form of moral residue as a result of her death. IT would be owed some kind of recognition, or apology by V, which may be symbolized by some form of compensation to her estate or acknowledgement of the harm she endured. This implication shows precisely that, in killing her, V would be (permissibly) infringing IT's rights, i.e. that IT is not liable to being killed.

Furthermore, I suggest the framework I am proposing here gets us significantly closer to adequately identifying the point at issue and the terms of the debate. Namely, I have argued that there is broad agreement both on the fact that it is permissible for V to kill IT, and that were she to do so she would be liable to compensate IT's estate or at least would owe some kind of apology or recognition of the sacrifice IT had to endure. The only question that remains at issue is whether IT is under a duty to let herself be killed by V, or whether it would be permissible for her to kill V. I have argued that the only way to do justice to this question is to take as our starting point the consideration that IT has not lost her right not to be killed. This clearly alters the terms of the discussion as presented by McMahan. For it is far easier to suggest that IT is under a duty to let herself be killed if we assume that she has lost her right to life than if we must point to a consideration sufficiently important to override such right to the extent that she must be put under such a duty. Put differently, I readily admit that the fact that she is morally responsible (though innocent) for the threat she is now posing is among the relevant considerations that must be taken into account in solving the conflict of prima facie rights between her and V. But I find it implausible to suggest that this consideration alone suffices to override the weight of her right in staying alive. This conclusion seems all the more persuasive if we admit that it is permissible for individuals to give greater weight to their interests than to those of another person, as the personal prerogative allows. So I suggest that under the framework hereby advocated IT seems closer to Killings than to Age difference and this fits well both with the conceptual apparatus of rights and with their strength in moral reasoning.

VI. A PROBLEMATIC IMPLICATION FOR THE MORALITY OF WAR?

Before concluding, let me take up a final difficulty. Saba Bazargan has plausibly suggested that tightening the conditions of liability to lethal force in the way hereby advocated may have problematic implications for the morality of killing in war.Footnote ⁴⁹ For one, it may directly impact the assessment of revisionist or neo-classical theorists regarding the moral inequality between just and unjust combatants. If minimally responsible threats are not liable to being killed in war, and most unjust combatants are only partially or minimally responsible, this would seem to put them on a par with just combatants. This would not only make it harder to justify resorting to force from an ad bellum perspective, as Bazargan indicates, but it would also make a majority of the unjust combatants non-liable to defensive killing. Just combatants would then only be at liberty to kill unjust combatants, but the great majority of the latter would also be at liberty to fight back. In this section I argue that this implication does not follow from the premises. Or better put, I argue that advocating more demanding conditions for liability to defensive killing on the basis of the argument hereby put forward does not lead per se to this particular implication. If the moral equality between just and unjust combatants is to be defended, it would have to be defended on other grounds.

To examine this issue let us consider the other standard case of an IT usually discussed in the literature: Resident. McMahan presents it thus:

McMahan considers Resident an innocent though morally responsible threat, and thereby liable to lethal force.Footnote ⁵¹ I find both these propositions spot on. What I will challenge is that the latter proposition follows from the former in the way he suggests.

Interestingly, not everyone agrees with McMahan's conclusion. Bazargan, for instance, suggests that Resident is not liable to being killed by Twin brother, and that she is in a position of pro tanto moral symmetry with him. This means that it would be permissible for Resident to defend herself if Twin brother were to pre-emptively defend himself against Resident's threat. It would be permissible for each of them to kill the other. Yet in order to defend this conclusion, Bazargan presents us with a slightly revised version of the case. In his version, Resident now opens the door to Twin brother and she has a reason to be armed due to the fact that she has been ‘harassed’ in the past.Footnote ⁵² Put differently, in his version of this case, it may be plausibly argued that resident makes a more reasonable call than McMahan's Resident, who kills Twin brother from a position in which she was not yet being threatened.

I am not persuaded that Bazargan's amendments to the case suffice to warrant his preferred conclusion. Yet, I believe they (a) illustrate well why McMahan's version is intuitively compelling; and (b) point in the right direction as to why Resident helps us explain that making the conditions for liability to lethal harm more demanding need not lead to a symmetrical position between the overwhelming majority of just and unjust combatants. Indeed, I find McMahan's version of the case largely uncontroversial because under a plausible account of self-defensive rights his Resident would lack the liberty to kill Twin brother. The reason for this is mainly that killing him was both unnecessary and disproportionate under the circumstances. In so far as Twin brother had not shown any sign of intending to kill Resident, or of being armed, and that Resident was ‘safely’ inside his home and armed, it seems impermissible for him to kill Twin brother pre-emptively even if all the evidence suggests he is Serial killer.Footnote ⁵³ Having a weapon imposes very stringent obligations on its use, and I do not believe they were sufficiently met in this situation.Footnote ⁵⁴

By contrast, consider the following version of this case:

I believe that Resident 2 shows that a case in which Resident does everything that is morally required from her before shooting at Twin brother does not support the intuition that she is liable to being killed (as McMahan claims), even if she is ultimately mistaken. Put differently, McMahan suggests that Resident takes a great moral risk,Footnote ⁵⁵ but in his version of the case he fails to highlight that Resident seems to impermissibly take too great a moral risk. I suggest that it is this consideration that lies at the core of the intuition that Resident is liable to being killed. Furthermore, Resident 2 shows that the argument advocated in this article does not directly translate into the moral equality of just and unjust combatants in war. At least not in a way that would be problematic for revisionist just war theorists. For one, some combatants would be in a position akin to that of CA or, at least, RA. Even those rank-and-file soldiers who are not culpable would generally be in a situation which is fundamentally closer to Resident than to IT. That is, we should hold combatants to stringent duties of care before they decide to go to war, i.e. intentionally impose lethal threats on other human beings.Footnote ⁵⁶ Finally, and perhaps decisively, Numbers and Age difference show that even if we grant the point that my framework puts just and unjust combatants in a pro tanto symmetrical position, i.e. that ad bellum unjust combatants would under certain circumstances not be liable to be killed by just ones, it can still insist on the claim that it would be all-things-considered permissible for just soldiers to kill unjust ones and impermissible for the latter to fight back.Footnote ⁵⁷ In that situation the rights of unjust combatants would be permissibly infringed.