II. THE LEGALIST VIEW AND THE NEO-CLASSICAL VIEW

As was indicated above, there are in contemporary political philosophy two dominant approaches to the moral standing of combatants. According to the first, defended by Michael Walzer, combatants are liable to intentional attack, regardless of the justness of their cause. In order to be liable to attack, of course, an agent must lose, or ‘forfeit’, his (negative) right not to be attacked. On Walzer's view, soldiers lose their right not to be attacked because, amongst other things, they pose material threats to each other. Since all combatants usually do so, they are equally permitted, or at liberty (under no duty not to), to use force against their adversaries.Footnote ² The resulting ‘moral equality of soldiers’ leads to the normative separation of the jus ad bellum (justice in the declaration of war) from the jus in bello (justice in the conduct of war).Footnote ³ According to Walzer, compliance with JIB, rather than JAB, is a necessary and sufficient condition to be recognized as a moral equal.

Jeff McMahan, in a number of incisive critiques, challenges Walzer's position.Footnote ⁴ For McMahan, unjust combatants, i.e. combatants who pursue an unjust cause, are liable to attack, but lack a liberty to intentionally kill their (just) adversaries. Contrary to Walzer's view, McMahan assumes that just combatants, i.e. those combatants who pursue a just cause, are not liable to attack. They have not, in other words, forfeited their right not to be attacked. Consequently, unjust combatants violate negative duties not to cause harm when they target their just adversaries. On McMahan's view, then, the presence or absence of a just cause for war at the level of JAB largely determines the permissibility of the use of force once hostilities are under way.Footnote ⁵ Those who lack a just cause are not moral equals. If McMahan is correct, JIB and JAB are not independent of each other.

Let us refer to Walzer's and McMahan's positions as the Legalist View and the Neo-Classical View, respectively. Although, as we shall see, the Legalist View amounts to a quasi-Hobbesian position on self-defence, it largely models the ethics of war on the contemporary laws of war.Footnote ⁶ By contrast, the Neo-Classical View returns to an older, pre-legalistic understanding of just war theory, which operates with a less pronounced distinction between JAB and JIB. As McMahan acknowledges, his position can be traced back to the writings of Francisco de Vitoria, Francisco Suarez and, more problematically, Elizabeth Anscombe.Footnote ⁷

In order to get a better grasp of these two views, it is useful to place them within the following fourfold normative taxonomy. First, A's use of force against B, it is assumed here, is morally excused if A can be exculpated from wrongdoing. Second, A's use of force against B is morally permissible if A is under no moral duty not to use force against B, subject to necessity and proportionality. Third, A's use of force against B is morally justified if there is a strong normative reason in favour of A's actions, subject to necessity and proportionality. Fourth, A has a right to use force against B if (and only if) she holds a negative claim against (a) third parties not to interfere with her use of force and (b) B not to repel her attack, subject to necessity and proportionality. Obviously, it must be specified whether the resulting right to self-defence is a moral and a legal right. But because it is primarily interested in the ethics of war, the article does not seek to tackle this issue. It focuses on moral rights only.

Viewed in terms of this taxonomy, the Legalist View rests on the assumption that (unjust) combatants act permissibly when killing their adversaries. As we just saw, if all combatants are liable to attack, no combatant violates any negative duties not to cause harm. However, without offering a detailed account of McMahan's critique, the problem with the Legalist View's argumentative structure is that it resembles, in Walzer's work at least, a plea for excuses. For instance, combatants are excused for taking part in an unjust war because they (a) pose a threat to each other, (b) have a sense of patriotism, or (c) have been conscripted into the army against their will.Footnote ⁸ Yet, while these points might show that unjust combatants are not blameworthy for killing just combatants, it does not prove that the latter lose their right not to be attacked. In brief, the Legalist View tends to exculpate (unjust) combatants from wrongdoing instead of showing that they act permissibly.

There are a number of strategies available to advocates of the Legalist View to confront this problem. It is beyond the scope of this article to discuss all of these here, but it is useful to point out two prominent ones. First, as was indicated above, it is possible to appeal to a quasi-Hobbesian understanding of self-defence. Roughly, the Hobbesian position rests on the claim that it is permissible to kill material threats to one's self-preservation. But any appeal to Hobbes is problematic for two reasons. First, as McMahan shows, in the initial stage of a war of unjust aggression, combatantsvictim_state do not pose material threats to combatantsaggressor_state.Footnote ⁹ Second and more significantly, the Hobbesian strategy makes it difficult, if not impossible, to draw a normatively sound distinction, central to just war theorizing, between combatants and non-combatants; for it is easy to imagine situations where the presence of non-combatants poses an obstacle to the self-preservation of combatants.Footnote ¹⁰

The second strategy invokes the notion of consent. Yet, as Walzer himself concedes, one of the characteristics of war is that it severely restricts the possibility of rendering consent.Footnote ¹¹ Moreover, as McMahan points out, it is by no means clear that it is possible to render morally valid consent to being attacked.Footnote ¹² Taken together, these brief points show that it is doubtful whether the Legalist View can successfully bridge the justificatory gap between excuse and permission.

Contrary to the Legalist View, the Neo-Classical View maintains that unjust combatants are not permitted to target just combatants, subject to qualifications. This claim is, in McMahan's work, the result of his meta-argument that there is no difference between the ethics of war and (a non-Hobbesian understanding of) the ethics of domestic self-defence.Footnote ¹³ In regard to the latter, McMahan argues that (domestic) attackers who are morally responsible for a forced choice between lives, i.e. a situation where an individual must be harmed, (a) are liable to attack because they forfeit their right not to be attacked and (b) are not permitted to repel their victim's (V) counterattack. The last consideration is the outcome of our intuitive understanding of justice. Those who are morally responsible for creating a state of affairs in which an individual must be harmed are liable to bear the resulting costs. Consequently, the attacker should suffer the harm he would have otherwise inflicted on the victim.

In terms of our normative taxonomy, McMahan's theory of self-defence assumes that V's use of force is justified, as there is a justice-based reason in favour of targeting the (responsible) attacker (RA). In addition, if RA should suffer the costs that result from the forced choice between lives he created, V holds a moral right to self-defence against RA. To wit, RA (a) lacks a right of self-defence against V and (b) third parties are not permitted to intervene on behalf of RA against V. Since this article focuses on the ethics of war, it assumes that McMahan's account of the ethics of domestic self-defence is, by and large, normatively sound. In particular, let us accept that moral responsibility for a forced choice between lives is a necessary and sufficient condition to (a) forfeit one's right not to be attacked and (b) be prohibited from using retaliatory force. The next question is whether this model is relevant for and applicable to the ethics of war.

Given that, for McMahan, the ethics of war is no different from the ethics of domestic self-defence, the Neo-Classical View answers this question affirmatively. In doing so, it ultimately contends that unjust combatants are morally on a par with domestic RA. As a result, unjust combatants are not allowed to repel their just adversaries, who are holders of a moral right to self-defence. The idea of a right to self-defence also explains why the Neo-Classical View denies the independence of JAB and JIB. The lack of a just cause leads to a duty not to resist just combatants. However, while the Legalist View must bridge the gap between permission and excuse, the Neo-Classical View must show that its extension of the principles governing domestic self-defence to combatants in war is defensible. As we shall see, this is a complex and complicated undertaking.

These are the rough contours of the Legalist and Neo-Classical Views, respectively. It is noteworthy, though, that the debate between their respective advocates and critics has almost exclusively been concerned with the status of combatants during self-defensive wars. The Legalist View argues that combatantsaggressor_state and combatantsvictim_state must be considered as moral equals, whereas the Neo-Classical View maintains that the former are morally obliged not to target the latter. Since self-defensive war against unjust aggression is usually considered as the paradigmatic case of a just war, the focus on self-defence is hardly surprising. But, as section III explains, it is premature to conclude that these two approaches are also readily applicable to non-self-defensive wars, most notably MHI.Footnote ¹⁴

III. COMBATANT STATUS AND THE PROBLEM OF INTERVENTION

Amongst contemporary political philosophers, there is a broad consensus that MHI is sometimes morally permissible, perhaps even morally obligatory. Overall, it is fair to say that contemporary liberal political philosophy is pro-interventionist.Footnote ¹⁵ Although it is impossible to do justice to the intricacies of the theoretical debate about MHI here, let us assume that it is at least morally permissible to halt militarily certain large-scale atrocities, including genocide, politically motivated mass murder (‘crimes against humanity’), enslavement and ethnic cleansing. For the sake of convenience, the article defines MHI as the deliberate and non-consensual military interference by a state or group of states (henceforth the intervening state) in the internal affairs of another state (henceforth the target state) in order to halt what we can loosely term Atrocity Crimes.Footnote ¹⁶ Assuming that MHI is permissible, we must specify the relationship between intervening combatants and their counterparts in the target state. This issue, as was indicated in section II, is problematic for both the Legalist and the Neo-Classical Views.

Beginning with the Legalist View, combatants, just or unjust, who violate the rules of JIB aren't exculpated from wrongdoing. Compliance with JIB, as we saw above, is a necessary (and sufficient) condition for equal moral standing. But, as Walzer puts it, the atrocities a soldier commits ‘are his’.Footnote ¹⁷ Intuitively, it is inappropriate, if not perverse, to grant equal moral status to those who massacre defenceless individuals, destroy their homes or exterminate whole groups. But it is questionable whether the Legalist View can consistently support this intuition. Its underlying Hobbesian model of self-defence, for instance, would allow combatants engaged in the perpetration of Atrocity Crimes to defend themselves against intervening combatants. Since intervening combatants pose a material threat, they are presumably liable to attack. In order to avoid such a counterintuitive position, advocates of the Legalist View must develop an alternative normative framework that regulates encounters between intervening combatants and perpetrators of Atrocity Crimes.

However, intervening combatants will also encounter ‘regular’ combatantstarget_state who are not engaged in Atrocity Crimes. Rather, they are merely ordered to defend the target state against the intervening state. According to the rationale of the Legalist View, regular combatantstarget_state can potentially be viewed as moral equals. This is so because the target state's campaign against the intervening state resembles a more traditional self-defensive war. While the perpetration of Atrocity Crimes necessarily involves grave violations of the rules of conflict, the target state's self-defensive campaign can in principle be fought within the rules of JIB. If this is sound, advocates of the Legalist View could propose a ‘hybrid’ position on the status of combatants during MHI. The moral equality of soldiers applies to encounters between regular combatants, while a different framework, perhaps closely resembling the Neo-Classical View, regulates the relationship between combatants engaged in Atrocity Crimes and those seeking to halt them.

Intuitively, though, the hybrid solution is unattractive. Its first part, i.e. the argument that there is a moral asymmetry between intervening combatants and perpetrators of Atrocity Crimes, is intuitively sound. But the second claim, i.e. that there is a moral symmetry between regular combatantstarget_state and intervening combatants, isn't. This is so because, if the target state resists the intervention, greater barriers will exist to halting successfully Atrocity Crimes. If we assume, however, that Atrocity Crimes represent some of the gravest moral evils known to mankind, then, for reasons of proportionality alone, it is wrong to resist the interveners.Footnote ¹⁸ Instead, everything should be done not to obstruct the rescue effort. In this light, it is reasonable to argue that (a) self-defence against the intervening state does not, at the level of JAB, constitute a just cause for war (subject to further conditions) and (b) regular combatantstarget_state should not be permitted to resist intervening combatants.Footnote ¹⁹

These observations suggest that the Neo-Classical View is closer to our intuitions about MHI than the Legalist View. Although, as section IV explains, this is correct when it comes to encounters between intervening combatants and perpetrators of Atrocity Crimes, the application of the Neo-Classical View to regular combatantstarget_state remains problematic. Even if we assume that a self-defensive war against the intervening state is unjust, regular combatantstarget_state, at least in the initial stages of MHI, do not pose an immediate threat to intervening combatants. Rather, regular combatantstarget_state are the victims of aggression, albeit permissible aggression. Thus, defenders of the Legalist View can reply that, although a self-defensive war against the intervening state is unjust and delays the halting of Atrocity Crimes, there is a normative case in favour of considering regular combatantstarget_state as moral equals. As one can see, the Neo-Classical View does not deliver an unambiguous verdict on the status of (regular) combatants during MHI either. As one can see, MHI remains, in many ways, the ‘blind spot’ of contemporary just war theory.

If the above analysis is sound, the phenomenon of MHI is not readily grasped by the two dominant approaches to the status of combatants. Nevertheless, the following discussion shows that the Neo-Classical View comes closest to generating a normatively sound position on the status of combatants during MHI. To do so, sections IV and V first look at the status of perpetrators of Atrocity Crimes, whereas section VI turns to the status of regular combatants.

IV. THE STATUS OF PERPETRATORS OF ATROCITIES

As we saw in section III, Walzer thinks that the atrocities a soldier commits are his. But it is not clear what he means by this. To be sure, since their occurrence signals large-scale non-compliance with negative duties not to harm, Atrocity Crimes usually involve some form of wrongdoing. But the question is how we should assess it morally.Footnote ²⁰ To tackle this issue, let us utilize the Nuremberg Defence, which replaced the older Superior Orders Defence during the Nuremberg Trials. Contrary to the Superior Orders Defence, which demands that combatants prove that the order they carried out was duly authorized, the Nuremberg Defence requires combatants to show that (a) they thought that the actions set out by a (duly authorized) order were morally and legally permissible (moral perception criterion) and (b) that following the order was the only morally reasonable course of action available at the time (moral choice criterion).Footnote ²¹ As we shall see, a more philosophical engagement with the moral perception and choice criteria reveals that perpetrators of Atrocity Crimes are not merely responsible for their deeds, but often also morally culpable for them.

Beginning with the moral perception criterion, combatants are required to demonstrate that the actions set out by an order were morally permissible. This requires a set of principles that enables them to make such an assessment. These principles must be (a) universal in scope and (b) responsive to the highly stressful conditions combatants face. According to Larry May, the bare moral minimum we can expect from combatants is not to attack intentionally individuals who are materially innocent, i.e. non-threatening, and defenceless. The intentional killing of defenceless individuals, May rightly observes, is usually deemed a great evil across cultures.Footnote ²² In order to assess the moral permissibility of an order, then, combatants must primarily determine whether it entails the intentional killing of the innocent. Although the moral perception criterion does not appear extraordinarily demanding, it is possible to raise the following objection to it.

Combatants, critics may point out, face considerable epistemological difficulties when assessing whether someone is innocent. In some conflicts, especially those that are asymmetrical, enemy fighters are not directly marked or, even worse, pose as non-threatening individuals.Footnote ²³ As a result, mistakes are likely. This is certainly true. But history also shows that perpetrators of Atrocity Crimes deliberately target their victims when they expect the least resistance. Thus the situational context of Atrocity Crimes is often not analogous to conditions on a battlefield, asymmetrical or not. To illustrate the point, the Serb forces who executed 8,000 Muslim men in Srebrenica did not face resistance from local groups or UN peacekeepers.Footnote ²⁴ The killing of Jews in Eastern Europe during the Second World War, at the hands of the various military and civilian organs of the German state, customarily involved putting victims deliberately into a defenceless position, e.g. by stripping them naked and shooting them in the back. Under these circumstances, those who participate in Atrocity Crimes cannot cite non-culpable ignorance as an excuse for killing defenceless individuals.

Interestingly, research in social psychology and sociology supports this claim. It shows that perpetrators of Atrocity Crimes often encounter a cognitive dissonance, rather than epistemological restrictions, when carrying out their orders.Footnote ²⁵ That is to say, perpetrators know that killing defenceless individuals is morally wrong, but also, for various reasons, intend to comply with their orders. Although there may be different psychological strategies available to overcome this cognitive dissonance, perpetrators correctly perceive the ‘wrongness’ of the situation.

That being said, while perpetrators might even acknowledge that they targeted innocent individuals, they could maintain that they had ‘no choice’ but to obey their orders. This takes us to the moral choice criterion, which supplements the moral perception criterion. As its name suggests, it requires that there are morally reasonable alternatives to following an order. If the agent has justified concerns that disobedience is going to lead to his death, his scope for action seems severely limited. Indeed, since societies in which Atrocity Crimes occur are often authoritarian, perpetrators will commonly claim that they acted under duress. But although it is important to realize that authoritarian societies limit our capacity for action, the appeal to duress is not as strong as it initially appears.

First, there are well-documented cases where perpetrators of Atrocity Crimes did not act because they were under duress. According to Christopher Browning in his famous historical study of the role of the (German) Reserve Police Battalion 101 during the Final Solution in Poland, officers were asked whether they wanted to take part in the killing of Jews.Footnote ²⁶ Most of the men complied, but, strikingly, not out of fear of death. There is not a single documented case that the refusal to kill Jews led to the death of potential perpetrators. Instead, police officers acted out of loyalty to the head of the Battalion and a sense of camaraderie. Building on Browning's analysis, the noted German social psychologist Harald Welzer argues that killing was nothing more than a ‘job’ that had to be done properly.Footnote ²⁷

Second, Atrocity Crimes, Claudia Card points out, are morally distinctive because they involve the infliction of ‘intolerable’ harms.Footnote ²⁸ Gang rape, (sexually motivated) torture and even cannibalism often occur during Atrocity Crimes.Footnote ²⁹ In an influential article on the civil war in Bosnia, Catherine McKinnon describes how ‘one woman was allowed to live so long as she kept her Serbian captor hard all night orally, night after night after night’, and how Muslim women who were taken away to be killed ‘were raped, had their breasts cut off and their wombs ripped out’. Further, in their treatment of the moral uniqueness of the Holocaust, Avishai Margalit and Gabriel Motzkin stress that perpetrators do not only seek to exterminate their victims, but also (sometimes) deliberately and cruelly humiliate them in the process of doing so.Footnote ³⁰ The pure gratuitousness of violence, humiliation and (sexual) sadism that accompanies, even characterizes, Atrocity Crimes suggests that a shocking number of individuals are not coerced into their role, but take advantage of the ‘killing opportunities’ offered by their state.

Third, it is notoriously difficult to determine what constitutes a justified belief that one's life is endangered. Someone held at gunpoint will experience a different form of duress than someone who has a mere suspicion that his life is at risk.Footnote ³¹ The greater the likelihood that the dreaded event is going to occur immediately, the more it seems appropriate to speak of duress. Conversely, if the likelihood is low (the threatening party is bluffing) or the consequences of non-compliance are vague, the less it seems appropriate to speak of duress. Generally, it is by no means the case that every potential perpetrator of Atrocity Crimes faces an immediate threat to his life in case of non-compliance. The appeal of duress is, therefore, limited.

Finally (4), even if a potential perpetrator faced an immediate threat to his life, duress, in most cases, does not excuse the (intentional) killing of an innocent person.Footnote ³² For criminal law, David Rodin notes, does not recognize duress as an excuse for intentionally killing the innocent.

These four points indicate that an appeal to a lack of moral choice is not a sound exculpatory strategy. Although highly oppressive societies remove options for individual action, combatants who are ordered to participate in Atrocity Crimes do not always face moral alternatives that are entirely unreasonable. Moreover, although duress, as the criminal law recognizes, has some exculpating function, it does not excuse the intentional killing of an innocent person.

Taken together, our discussion of the moral perception and choice criteria shows that participants in Atrocity Crimes are often not only morally responsible for their actions. Rather, since they often lack a sound exculpating reason for killing the innocent, we can assume that they are morally blameworthy for what they do. In this sense, perpetrators of Atrocity Crimes exceed the basic moral requirement that, for the Neo-Classical View, is necessary and sufficient to establish liability to attack, namely, moral responsibility. If the above analysis is sound, perpetrators of Atrocity Crimes are culpably responsible for the threat they pose. If this is true, we can interpret Walzer's claim that the atrocities a soldier commits ‘are his’ as an indication of moral culpability. Hence it is not too far-fetched to argue that perpetrators of Atrocity Crimes can be classified as what the ethics of (domestic) self-defence refers to as culpable attackers (CA), rather than RA.

If perpetrators of Atrocity Crimes and CA are morally equivalent, our analysis has interesting implications for the ethics of war in general and the debate about the status of combatants in particular. First, while the Neo-Classical View argues that moral responsibility is necessary and sufficient for combatants to incur liabilities to attack, the presence of moral culpability indicates that the considerations of necessity and proportionality that govern the use of interventionist force are not as strict as in cases involving RA.Footnote ³³ It would, for instance, be permissible to kill a higher number of CA than RA. Faced with the choice of killing perpetrators of Atrocity Crimes with RA and CA status, respectively, intervening combatants should target the latter. Second, since the use of force against CA is usually the least controversial case for the ethics of self-defence, the equivalence of CA and perpetrators of Atrocity Crimes strengthens, though it does not fully establish, the normative case for a (moral) right to intervene, which is correlated to a negative duty, falling upon the target state and some of its combatants, i.e. those engaged in the perpetration of Atrocity Crimes, not to resist the interveners.

V. THE STATUS OF NON-RESPONSIBLE PERPETRATORS OF ATROCITIES

Nonetheless, there is at least one group of perpetrators, responsible for some of the most horrific atrocities, which falls into the category of neither CA nor RA. Child soldiers are often used to perpetrate Atrocity Crimes. Children have little deliberative agency due to their lack of moral development.Footnote ³⁴ Moreover, apart from brutalization and brainwashing, they are often drugged by their captors in order to lower their threshold for aggression and ensure compliance. It is hard to see how child combatants could possibly be considered as morally culpable for their actions. They do not even seem morally responsible for them, or at least their moral responsibility is strongly diminished. Intuitively, then, child combatants seem to be excused for participating in large-scale wrongdoing. For the sake of convenience, let us refer to those types of attackers who lack moral agency and are, as a result, neither culpable nor responsible for a wrongdoing as Non-Responsible Attackers (NRAs).

For the Legalist View, of course, the category of NRAs does not pose a normative problem. As was observed in section II, because of its Hobbesian roots, it does not differentiate between various categories of attackers. Considerations of moral responsibility or moral agency more generally are not relevant to determine liability to attack. To use a slightly hyperbolic example, for the Neo-Classical View, a person who becomes an involuntary physical threat owing to a sudden gust of wind is as liable to attack as an assassin who is about to carry out a ‘hit’ he has been planning for weeks. For the Neo-Classical View, by contrast, the case of NRAs poses a problem, since NRAs do not forfeit their right not to be attacked. This is so because NRAs lack the precondition for the incurrence of liabilities to attack, namely, moral agency. In what follows, let us briefly discuss whether the theoretical problem of NRAs, as well as its ‘real-world’ manifestation in the case of child soldiers, undermines the Neo-Classical View.

To begin, defenders of the Neo-Classical View can argue that self-defence against NRAs, though impermissible, is excusable.Footnote ³⁵ This claim raises a number of problems that we do not need to discuss here. But even if we assume that it is sound, it is problematic for the ethics of MHI because the victims of Atrocity Crimes do not defend themselves against NRAs. Rather, a third party acts on their behalf. However, the third party, unlike the victim of NRAs, cannot cite its own interests as an exculpating reason for killing NRAs.Footnote ³⁶ This is so because the rescuer is not himself threatened by NRAs. In fact, because neither NRAs nor their victims have lost their (respective) rights not to be attacked, it does not seem that, ceteris paribus, an intervening party could (and should) give preference to the rights of the victims over those of non-responsible perpetrators of Atrocity Crimes. Non-intervention, in other words, seems to be the only morally legitimate strategy. The question, though, is whether other things are equal. There are two proposals that suggest that they aren't.

The first proposal contends that a third party can be excused for killing NRAs on behalf of a victim if (and only if) they share a special relationship. Using an example from domestic society, a father may be excused for defending his child against NRAs. But unfortunately, the case of MHI is not comparable to the father–child relationship. Although considerations of self-defence may inform the decisions to intervene, the immediate victims of Atrocity Crimes are, in most cases of MHI, non-citizens of the intervening state. An appeal to special relationships is thus likely to fail.

The second proposal – which I want to defend here – involves what I call the Numbers-Harm Thesis. As its name suggests, it comprises two elements. The first appeals to the normative weight of numbers, especially the claim, defended by some non-consequentialist moral theorists, that one should rescue the greater number.Footnote ³⁷ If this is correct, the number of potential victims, which is likely to be high in Atrocity Crimes, does count for our moral assessment of the situation. In addition, as was mentioned above, the harm inflicted during Atrocity Crimes is morally distinctive because of its severity, if not gratuitousness. If we combine this observation with the fact that Atrocity Crimes usually involve large numbers of victims, the killing of NRAs, according to the Numbers-Harm Thesis, at the hands of a third party is at least excusable, notwithstanding the absence of a special relationship between the rescuing party and the victims of Atrocity Crimes.

But it is worthwhile stressing that even though NRAs threaten large numbers of victims with gratuitous levels harm, they, in the absence of moral agency, do not lose their right not to be attacked. The Numbers-Harm Thesis should not be used to define negative duties not to cause harm out of existence. They are too important in our ethical thinking to be cast aside like that. Instead, the Numbers-Harm Thesis provides intervening combatants with a normative reason that at least excuses non-compliance with negative duties not to cause harm when targeting child soldiers or other types of NRAs. If this is sound, intervening combatants excusably infringe the rights of NRAs.

In the present context, the concept of a rights infringement raises two critical questions. First, it must be clarified whether child soldiers, unlike culpable participants in Atrocity Crimes, are permitted to defend themselves against intervening combatants. The answer is negative because the intervening combatants do not pose an unjust threat and, therefore, have not lost their right not to be attacked. Nevertheless, since their rights (no less) are infringed, NRAs have an agent-relative excuse for attempting to repel intervening combatants. On similar grounds, intervening combatants have an excuse for repelling child soldiers or other NRAs.

The second question is whether it is potentially excusable for a third party to intervene in order to defend one side against the other. Let us first consider whether a third party is justified in assisting intervening combatants who have come under attack from child soldiers. Since the institution of the military establishes strong special relationships between its members, intervening combatants may potentially be excused for defending their comrades against child soldiers. But since the Numbers-Harm Thesis already permits intervening combatants to target child soldiers, an appeal to special relationships appears superfluous. It is possible to reply, however, that one must distinguish between force aimed at rescuing fellow intervening combatants and force aimed at rescuing the victims of Atrocity Crimes. Strictly speaking, the Numbers-Harm Thesis only permits the latter. But in practice the two acts are so closely intertwined that it would be difficult to separate them.

Let us now consider whether third parties are at least excused in assisting child soldiers who have come under attack from intervening combatants. On the one hand, it would be strange to argue that child soldiers stand in a special relationship to their superiors. In general, special relationships should have moral weight if (and only if) they fulfil certain preconditions. For instance, they should be non-exploitative, respectful of the rights held by the members of an association qua individuals, valued by the members of an association and not detrimental to the rights of non-members of an association. None of these conditions are met in the case of an army that uses child soldiers to perpetrate Atrocity Crimes. On the other hand, as was argued above, the relationship between children and parents or the close-knit relations between members of a village may provide parents or fellow villagers with an excuse to defend child soldiers against intervening combatants. That being said, one should mention that parents or villagers should assist intervening combatants by, if necessary, killing those who irresponsibly (and culpably) put children into the firing line.

Finally, it must be pointed out that principles of ‘due care’ and criteria of necessity and proportionality are extremely stringent in cases involving NRAs. In regard to the former, intervening combatants should accept greater risks to themselves when confronting child soldiers. They are usually not obliged to accept these risks in cases involving RA or CA. In regard to the proportionality and necessity criteria which form part of JIB, intervening combatants are allowed to target an almost unlimited number of RA and CA if doing so means that they will not have to kill child soldiers.

Admittedly, the case of NRAs challenges the Neo-Classical View. But unless Atrocity Crimes are exclusively perpetrated by NRAs, the existence of an exception to the rule does not undermine the Neo-Classical View as such. Having dealt with the moral status of the perpetrators of Atrocity Crimes, we must now turn to the status of regular combatants who are summoned to the defence of the target state. In particular, we must find out whether, like combatants who participate in Atrocity Crimes, they are also bearers of the duty of non-resistance which is correlated to the right to intervene.

VI. THE STATUS OF COMBATANTS WHO RESIST JUST INTERVENERS

As was pointed out in section III, advocates of the Legalist View, the main competitor to the Neo-Classical View, can potentially endorse a hybrid position on the status of combatants during MHI. They can argue that because regular combatantstarget_state are not involved in the perpetration of Atrocity Crimes, they should be recognized as moral equals. Unlike perpetrators of Atrocity Crimes, they are the victims of (just) aggression without having posed an unjust threat. In order to challenge this kind of reasoning, the Argument from Derivative Liability, to be developed below, contends that regular combatantstarget_state who defend their state against combatantsintervening_state are morally complicit in Atrocity Crimes. In the language of the criminal law, they are derivatively liable for Atrocity Crimes. If this is sound, the Argument from Derivative Liability would be a clear indicator that the Neo-Classical View, rather than the Legalist View, governs the relationship between combatantsintervening_state and regular combatantstarget_state, giving the former a right intentionally to target the latter.

Discussions of complicity usually involve two elements.Footnote ³⁸ The conduct element, the actus reus, shows how an accessory's (S) actions are causally related to a wrongdoing perpetrated by a principal agent (principal/P). S typically enables P to do x by aiding, encouraging or counselling P (or via a combination thereof). The mental element, the mens rea, sets out a number of ‘mental conditions’ that must be met for S to be morally responsible for wrongfully aiding P. In order to explore the relevance of complicity for MHI, it is easier to start with actus reus. We can then turn to mens rea.

To begin, if regular combatantstarget_state are complicit in the perpetration of Atrocity Crimes, they are also (at least) morally responsible for atrocious mass killing. From a moral perspective, they are on a par with those individuals who directly carry out the killing. But this claim seems striking for three reasons.

First, regular combatantstarget_state are not involved in Atrocity Crimes themselves and might not even be close to the site where they take place. Second, regular combatantstarget_state do not cause perpetrators to engage in Atrocity Crimes. Regular combatantstarget_state are not, say, threatening to kill or otherwise punish potential perpetrators should they fail or refuse to comply with their orders. Third, it is questionable whether regular combatantstarget_state aid the perpetrators of Atrocity Crimes at all. Regular combatantstarget_state do not support perpetrators by directly acting towards them, e.g. by transporting weapons to the killing site or providing other forms of assistance such as food, water and medical supplies.Footnote ³⁹ Regular combatantstarget_state are merely ordered to defend their state against combatantsintervening_state. Contrast this with an example that I refer to as Gun Case: Sgun gives P a gun that P uses to kill V. Here we encounter a positive act, the handing over of the weapon, for which Sgun is clearly responsible and perhaps even culpable. In light of these arguments, it sounds reasonable to maintain that regular combatantstarget_state are only responsible for actions restricted to their encounters with combatantsintervening_state.

The Argument from Derivative Liability negates this type of reasoning. It emphasizes that, by defending the target state, regular combatantstarget_state are morally responsible for sustaining a state of affairs in which very serious wrongs are committed. Hence they are (indirectly) morally responsible for Atrocity Crimes. From the perspective of the actus reus criterion, there are three points in support of this position.

First, the criminal law does not require that, in order to be derivatively liable for x, S causes P to engage in x. Consequently, in order to have accomplice status, regular combatantstarget_state do not need to cause the perpetrators of Atrocity Crimes to act. On the other hand, if it can be shown that P would have been, say, killed by S if P did not do x, it is likely that S's derivative responsibility increases up to a point where S himself could become the principal. If the perpetrators of Atrocity Crimes only engage in wrongdoing because their comrades would otherwise kill them, combatantsintervening_state should try, for reasons of necessity and proportionality alone, to minimize the infliction of harm on the former, while increasing their attack on the latter.

Second, an agent does not need to act positively towards a principal in order to be the latter's accessory. It is sufficient that the accessory's actions place P in a position where P can carry out the wrongdoing. If S helps P rob a bank by distracting the police, S has the same level of derivate liability as P's second accomplice who hands P the dynamite to break open the bank's safe. By analogy, if regular combatantstarget_state successfully repel intervening combatants, perpetrators of Atrocity Crimes are in a position to continue their attack on their victims. Regular combatants are thus morally on a par with, say, those soldiers charged with transporting weapons to the killing site.

Third, and directly related to the preceding point, it is also not required that the accessory is present when the wrongdoing is carried out. If, in the bank robbery case, S distracts the police three blocks away from the bank, his moral responsibility does not diminish. If regular combatantstarget_state render assistance that shields the perpetrators of Atrocity Crimes, it does not make a difference whether they are present at the killing site or not. The point is that the perpetrators of Atrocity Crimes will not be able to continue the assault on their victims if their comrades fail to halt the advance of combatantsintervening_state. The possibility that the accessory is absent from the actual execution of the wrongdoing does not stand in the way of considering him complicit in it.

Although we have overcome these initial obstacles, we must note a related problem for the actus reus criterion. S's contribution to P's design is not always (morally) impermissible. To use Gun Case as an example, Sgun's act of giving P the gun may not be in itself morally wrong. Critics can therefore argue that it is absurd to maintain that Sgun should be derivatively liable for V's murder. But we can reply that, under these specific circumstances, Sgun is under a duty not to give P the gun. Sgun's liberty to dispose of his property (‘the gun’) as he pleases must be balanced against the rights of the prospective victim.Footnote ⁴⁰ If the latter's rights are in danger of being violated, an otherwise morally permissible act can be wrong. In the context of Atrocity Crimes, it is clear that we must balance the Legalist View's moral permission to participate in an unjust war against the rights of the victims of Atrocity Crimes. Even if one generally supports the Legalist View, one could maintain that it does not apply to MHI for exactly this reason. This point lends further support to the Neo-Classical View.

But naturally, while the fulfilment of the actus reus criterion is a necessary condition for derivative liability, S, as the mens rea criterion demands, can only be derivatively liable for x if S is also aware of the relevant circumstances that lead to x. In Gun Case, for instance, Sgun is only complicit in V's murder if he knows that P is plotting an attempt on V's life. More precisely, the mens rea criterion for complicity, on which much of the Argument from Derivative Liability hinges, requires that:

1. (1) S intends to do the acts through which he assists P and is aware that these acts aid P to engage in a wrongdoing.

2. (2) S must know what P is about to do, that is, he must know the relevant circumstances. This is not to say that S must be fully aware of all facets of P's plans. Sometimes culpable ignorance, i.e. ‘closing one's eyes to the obvious’, can also lead to complicity.Footnote ⁴¹

The first part of the mens rea criterion necessitates a brief engagement with the role of intentions in assessments of complicity. It is possible to approach the issue in two ways. On the one hand, one can argue that S's intention to aid P is necessary to establish S's derivative liability. In this respect, Christopher Kutz stresses the centrality of what he calls ‘participatory intentions’ for the theory of complicity.Footnote ⁴² On the other hand, it is debatable whether S's intention to contribute to P's wrongdoing is really necessary for S to be considered complicit in it. The second approach thus contends that we should hold S responsible for what he does to further P's wrongdoing, regardless of S's intentions. As a result, S is derivatively liable for P's wrongdoing, even if S does not intend to assist P. To establish S's derivative liability, it is (merely) necessary that S is morally responsible for aiding P.

The Argument from Derivative Liability tends towards the second approach. That is to say, it is interested in moral responsibility rather than intentions, though the latter can, of course, increase a person's responsibility. To illustrate the point, consider the case of the Patriotic Soldier who knows that Atrocity Crimes are taking place, but denies that he intends to aid their perpetrators by resisting combatantsintervening_state. Known to be sincere, the Patriotic Soldier intends to protect the target state and its culture from the intervening state. The Patriotic Soldier is opposed to Atrocity Crimes, but foresees that the actions necessary to defend his country contribute to the continuation of mass murder. The Argument from Derivative Liability assumes that the Patriotic Soldier is complicit in Atrocity Crimes, since he can be held morally responsible for actions that assist the perpetrators in carrying out their deeds. Similarly, if, in Gun Case, Sgun disagrees with P's plan to kill his brother but sells P the gun out of sheer greed, he is still derivatively liable for the murder.

Turning to the second component of the mens rea criterion, in order to be derivatively liable for x, S must have knowledge of P's plan to engage in x. In the present context, the question is whether regular combatantstarget_state have sufficient information about the occurrence of Atrocity Crimes so as to fulfil the mens rea criterion. This is problematic because they are, as we just saw, usually not present at the killing site. Moreover, while Sgun may know that P has a severe grudge against his brother, regular combatantstarget_state have usually no personal knowledge of the orders issued to potential perpetrators of Atrocity Crimes by the target state's government. Fortunately, both objections can be dealt with relatively quickly.

First, MHI is usually a response to an already existing state of affairs. There are reasons to be sceptical that even an entirely authoritarian society can completely cover up Atrocity Crimes. For example, in Cambodia, the epitome of an authoritarian state, many people had heard of the killing fields through eye-witness accounts, word-of-mouth and other more informal channels of communication.Footnote ⁴³ It is unlikely that there will be no information about Atrocity Crimes available to regular combatantstarget_state.

Second, while it is true that regular combatants have no knowledge of the orders the perpetrators received, they can reasonably be expected to consider that their government is capable of ordering Atrocity Crimes. After all, as was noted in section IV, regimes that order or tolerate Atrocity Crimes are often highly authoritarian and have a history of (1) committing violence against their own people and/or (2) engaging in acts of unjust aggression against other societies. If regular combatants do not consider this possibility, they really do close their eyes to the obvious.

If these points are sound, we can assume that regular combatantstarget_state also fulfil the second part of the mens rea criterion. They can, therefore, be held derivatively liable for Atrocity Crimes. More precisely, the rendering of support to those engaged in Atrocity Crimes entails the violation of negative duties not to worsen the condition of potential victims. For the Argument from Derivative Liability, regular combatantstarget_state, even though they may not intend to assist the perpetrators of Atrocity Crimes, are morally responsible for the violation of these duties, therefore incurring derivative liabilities to attack. Contrary to the Legalist View, then, regular combatantstarget_state should not be considered as moral equals. As a result, the Argument from Derivative Liability rejects potential proposals for a hybrid position on the status of combatants during MHI. Rather, the Neo-Classical View should apply to all groups of combatants in the target state, i.e. those engaged in Atrocity Crimes and those summoned to the defence of the target state. This reinforces the aforementioned idea of a right to intervene. If the above arguments are sound, intervening combatants do not only hold a claim against perpetrators of Atrocity Crimes not to resist the intervention, but also against regular combatantstarget_state.