Comment on “Associative Duties and the Ethics of Killing in War”
Jeff McMahan
University of Oxford
Seth Lazar’s “Associative Duties and the Ethics of Killing in War” (Journal of Practical Ethics, Volume 1, Number 1) is an original, rich, challenging, and intricately argued contribution to our understanding of the ethics of war. Its main aim is to explain how fighting in a war can be permissible when warfare inevitably involves the killing of people who are not liable to be killed—a problem that is more extensive than it may seem.
Civilians are almost inevitably killed in war and Lazar accepts that few if any civilians are liable to be killed in war. In principle, of course, a war could be fought without killing civilians, and certainly without killing them intentionally. Yet it is scarcely possible to fight a war, or at least a war in the familiar sense, without intending to kill enemy combatants. Lazar believes, however, that many combatants are not liable to be killed. As I and others have argued, those who fight for a just cause in a just war (“just combatants”), and by permissible means, do nothing to make themselves morally liable to be killed—that is, they do nothing to forfeit or lose their right not to be killed. And Lazar himself has argued, in previous work, that many combatants who fight in wars that lack a just cause (“unjust combatants”) are also not liable to be killed. This is because the harm that, as individuals, they threaten to cause is insufficiently great or because the degree to which they are responsible for the harm they threaten is too low. I think he is right about this, though I think the proportion of unjust combatants who are not liable to be killed is in most cases lower than he thinks it is.
The fact that many combatants are not liable to be killed poses a problem for just war theory because, Lazar claims, “contemporary philosophers of the ethics of war … until recently unanimously affirmed … that in justified wars those whom we kill and maim are liable to be killed.” Since no contemporary philosopher writing on the ethics of war claims that all the civilians killed as a side effect in justified wars are liable to be killed, Lazar must mean that philosophers have claimed that wars are justified only when those killed intentionally, as a means of achieving the war’s aims, are liable to be killed. But if this were correct and yet many combatants on both sides were not liable to be killed, there could in practice be no justified wars. Pacifism would be the correct account of the morality of war. To avoid being committed to pacifism, Lazar thinks we should accept that “warfare necessarily involves violating rights,” including the rights of those who are killed without being liable to be killed, but also accept that “weightier reasons can override those rights violations, rendering warfare all things considered justified, though unjust.” (4)
The main aim of Lazar’s argument is to show that associative duties to those to whom we are specially related can sometimes override the rights of people not to be killed, “thus rendering some acts of killing [in war] all things considered justified.” (5)
(A brief parenthetical comment. When Lazar writes that philosophers have claimed that a justified war is one in which those killed intentionally are liable to be killed, he also comments that “Jeff McMahan…accepted this commitment without question.” (4) Just for the record, I did not accept this uncritically because I did not accept it at all. In a paper published in 2005, for example, I wrote that “while all just wars are morally justified, it seems that not all morally justified wars are just wars. … [T]here seem to be wars that are morally justified despite their requiring the targeting of those who are innocent in the relevant sense… The form of justification in these latter cases is familiar: in rare circumstances, considerations of consequences override constraints on action that would otherwise be decisive.” [“Just Cause for War” Ethics and International Affairs, p. 16.])
Lazar develops a plausible non-teleological or non-instrumental account of the significance of special relations and the way in which they ground associative duties. A substantial portion of his essay then seeks to show that associative duties can ground a permission, in a restricted range of conditions, to defend a person to whom one is specially related even at the cost of killing another person who is not liable to be killed. He then argues that this permission can apply in war and can explain the permissibility, at least in certain conditions, of killing combatants who are not liable to be killed. This Associativist Account of the permissibility of certain killings in war provides, he suggests, the best way of avoiding being committed to pacifism. Its appeal to associative duties supports the claim that “combatants on both sides of a war can, in some cases, fight justifiably.” (42) In particular, it explains how unjust combatants “can sometimes be justified in fighting, and on much the same grounds” as those that justify the belligerent action of just combatants. (42)
Lazar contends that, “at least in 1:1 cases, the duty to protect” a person to whom one is specially related “can override the general negative duty not to foreseeably…kill a non-liable person.” (19) In other words, it can be permissible for a third party to save the life of an innocent person to whom he is specially related even when, in doing so, he will knowingly kill an innocent or nonliable bystander as a side effect. I will refer to this as Lazar’s central claim.
Lazar defends his central claim by presenting a transitivity argument. He offers both a stronger and a weaker version of the argument and says that he favors the stronger version. But the stronger version assumes that it is morally required to save 5 when doing so would kill an innocent bystander as a side effect of the redirection of a preexisting threat. This is too controversial to be a reliable foundation for his central claim. I will therefore briefly summarize and comment on the weaker version. But my main objection applies equally to both versions.
Lazar claims:
- It is permissible to redirect a meteor (or trolley) as a means of saving five nonliable people even though this has the foreseen side effect of killing one nonliable bystander.
- It is permissible to save one to whom one is specially related rather than save five nonliable people.
- From these two claims he infers that:
- It is permissible to redirect a meteor (or trolley) as a means of saving one to whom one is specially related even though this has the foreseen side effect of killing one nonliable bystander.
The logic of the argument seems to be this. The moral weight of saving five people is the same in cases 1 and 2. If not killing a nonliable bystander as a side effect has less moral weight than saving five, while saving one to whom one is specially related has greater moral weight than saving five, it follows that saving one to whom one is specially related has greater moral weight than not killing a nonliable bystander as a side effect.
There are various, though related, reasons for doubting the validity of this argument. There are, for example, counterexamples to the claim that “permissible when the alternative is” is a transitive relation. For whether an act is permissible can depend on what the alternatives are. Lazar cites one such counterexample from the work of Frances Kamm. There are others. (See, for example, Derek Parfit, “Future Generations: Further Problems,” Philosophy and Public Affairs 11 (1982), p. 131.) He argues that the reason why the “transitivity of permissions” fails in Kamm’s case does not apply to his argument. He may well be right about that and it may also be true that his argument differs in relevant respects from other similar arguments in which transitivity fails. There are, however, two other concerns that I will merely mention but not pursue.
One is that there may be what Kamm calls “contextual interaction” among the factors in the different cases. It may be, for example, that certain relevant considerations arise in choices in which one option involves killing that do not arise in choices among options that involve only saving and allowing to die.
The second concern is that, despite the assignments of numerical values that Lazar makes for heuristic purposes, the comparisons among killings and lettings die on which the argument depends cannot be precise. This is not because of epistemic limitation but because the relevant values or reasons may in reality be only imprecisely comparable. And when different options are only imprecisely comparable, transitivity may be undermined. (On evaluative imprecision and its significance, see Derek Parfit, “Toward Theory X: Part One” and “Toward Theory X: Part Two,” unpublished manuscripts.)
One reason I will not pursue these concerns here is that even if the argument is valid, it requires a further and doubtful assumption to have any serious relevance to the morality of killing in war. The killings in cases 1 and 3 that he claims are permissible are not only merely foreseeable rather than intended but also brought about by the redirection of a preexisting threat—the meteor. This latter fact has often been thought to be part of the explanation of why it is permissible in the standard trolley case to divert the runaway trolley that will otherwise kill five people onto a branch track where it will kill only one person. When Philippa Foot first introduced the trolley case, she contrasted it with a similar case in which to save five patients doctors must release a gas that will kill one other patient as a side effect, observing that while it seems permissible to kill one as a side effect of saving five in the trolley case, this does not seem permissible in the gas case. (“The Problem of Abortion and the Doctrine of Double Effect,” in her Virtues and Vices, p. 29.) Commenting on these cases in a later paper, Judith Thomson argued that the reason it is permissible to kill the one in the trolley case but not in the gas case is that in the trolley case one is merely “arranging that something that will do harm anyway shall be better distributed than it would otherwise be.” (“The Trolley Case,” in her Rights, Restitution, and Risk, p. 108.) In the gas case, by contrast, one is creating an entirely new threat to the one.
Since killing in war is almost always done via the creation of a new threat rather than through the redirection of an existing threat, the further assumption that Lazar’s argument requires to have significant implications for war is that there is no moral difference between the creation of a threat and the redirection of an existing threat. But, as Thomson contended, this is a dubious assumption.
If in Lazar’s first case the only way to prevent the meteor from landing on the five were to blow it up by detonating a bomb that would itself kill a nonliable bystander as a side effect, it seems that this would not be permissible. And it seems the same would be true if in his third case the only way to prevent the meteor from killing one’s child were to create an explosion that would kill a nonliable bystander. This should not be surprising given the difference between people’s intuitive reaction to the trolley case and their reaction to the gas case. But if it is right that there is a moral difference between killing via the redirection of a preexisting threat and killing via the creation of a threat, it seems that the plausibility of Lazar’s transitivity argument is restricted to cases in which the killing is done via redirection. This means that the scope of the argument is highly limited and that it is virtually irrelevant to the justification of killing in war, which is seldom done by the redirection of a preexisting threat.
Some people, of course, believe that it is permissible for an individual to act in self-defense or self-preservation in a way that will create a threat that will kill a nonliable bystander as a side effect. But this is a minority view. Even those who claim that there is an agent-relative permission to kill a wholly innocent or even nonresponsible person who threatens one’s life are usually reluctant to accept that this permission extends to an act of self-preservation that creates a threat that will kill a bystander as a side effect.
But even it if is not permissible to kill a nonliable bystander as a side effect of an act of self-preservation, it might be permissible to kill such a person as a side effect of an act of saving a person to whom one is specially related. For an agent-relative permission and an associative duty are distinct sources of reasons, and it may well be that at least some of one’s special relations to others are sources of stronger reasons than is the relation of identity one bears to oneself. It might, for example, be permissible for a parent to kill a nonliable bystander as a side effect of saving her child even when it would not be permissible for the child to save himself if doing so would unavoidably kill the same bystander as a side effect.
But even a view of this sort has little relevance to the permissibility of killing in war. One reason is that such a view is plausible, if at all, only in the case of the most significant special relations, such as the relation between a parent and child. It would not be permissible to kill a nonliable bystander as a side effect of saving one’s neighbor, even though being neighbors is a special relation that has a certain degree of moral significance. But in war a soldier almost never knows that an act that would kill an enemy combatant is necessary to save the life of someone as closely related to him as his child. In general, the most he can know is that the act may slightly reduce the risk to someone closely related to him of being killed by the enemy.
There is, however, one exception to this, which is that a combatant can sometimes know that his killing an enemy combatant is necessary to save the life of one of his comrades-in-arms. Lazar cites the claim of J. Glenn Gray and others that the most important factor that motivates combatants to continue to fight rather than to flee or surrender is the compulsion they feel to protect their close comrades. A specialist on the psychology of war reports that “in military writings on unit cohesion, one consistently finds the assertion that the bonds that combat soldiers form with one another are stronger than the bonds most men have with their wives.” (Quoted in Lt. Col. Dave Grossman, On Killing, 1995, p. 149.) Perhaps this special relation strengthens the justification that combatants have for killing enemy combatants even when the latter may not be liable to be killed.
But I doubt that this is true. Suppose the combatants under attack are just combatants and the enemy combatant that one of them must kill to save the other is an unjust combatant. In that case, the unjust combatant threatens a just combatant with death and hence is morally liable to be killed, even if the degree of his responsibility for the threat he poses is low. There is already a liability justification for killing him; hence the appeal to an associative duty is otiose.
If instead the combatants under attack are unjust combatants and the enemy combatant who threatens one of them is a just combatant, it is then very unlikely that the special relation between the two unjust combatants could justify the killing of the just combatant. There are at least four reasons why this is so, some of which are recognized by Lazar himself.
First, the bonding that the unjust combatants have experienced is not so much a reason for fighting as it is a result of fighting. The bond has developed because of their having been in combat together. But because their war is unjust, the bond has arisen because of their shared participation in an activity that is objectively wrong. Indeed, they did wrong to get themselves into the situation in which the bond developed and now motivates them to kill people who are not liable to be killed. As Lazar acknowledges, following Thomas Hurka, the contaminated nature of the relation between them diminishes or even vitiates altogether its moral significance.
Second, as Lazar also concedes, the associative duty to protect someone to whom one is specially related has at most only a weak application when the that person is liable to be harmed. And when an unjust combatant will otherwise kill a just combatant, that unjust combatant is liable to be killed. Thus, even if the special relation that the one unjust combatant bears to the other were highly morally significant, it would not justify a combatant in killing a nonliable person as a means of defending his comrade-in-arms against an attack to which he was liable.
Third, Lazar’s central claim concerns the permissibility of killing a nonliable bystander. But a just combatant is not a bystander; he is a just threatener vis-à-vis the unjust combatant he threatens. Killing him is therefore wrong for two distinct reasons: it would not only wrong him but also prevent him from achieving his just aim.
Fourth, and finally, Lazar’s central claim concerns the justifiability of killing a nonliable person as a side effect—that is, an unintended effect. But the killing of a just combatant as a means of saving one’s fellow unjust combatant is an intended killing, and the constraint against killing a nonliable person as a means is, as Lazar recognizes, stronger than that against the killing of a nonliable person as a side effect.
This is perhaps the main reason why Lazar’s central claim has little relevance to the justification for killing in war. That claim is that it can be permissible to kill a nonliable person as a side effect of the redirection of a preexisting threat away from someone to whom one is significantly specially related. But the killing of combatants in war, which is what needs to be justified in principle if pacifism is to be avoided, is much more often intended than unintended and is almost always accomplished by the creation rather than the redirection of a threat.
It is, however, unclear what Lazar takes the scope of his argument to be. While his central claim is about the justifiability of killing as a side effect, he is clearly aware that the refutation of pacifism requires a justification for at least some intentional killing of combatants, and particularly unjust combatants, who are not liable to be killed. Thus, he writes that “the real challenge is to show that [associative duties] can license some intentional killing of combatants, without also permitting intentional killing of noncombatants.” (34)
His effort to meet this challenge begins with four reasons why it is more seriously wrong intentionally to kill nonliable noncombatants than it is to kill nonliable combatants. Of these four reasons, the one to which he devotes most space is that “it is more wrongful to kill nonliable people who are defenceless and vulnerable than to kill those who can fight back, or who are less vulnerable.” (36) Although common, this claim has always seemed implausible to me. It implies, for example, that it is more seriously wrong to kill just combatants using bombers or long-range artillery than to kill them in close combat—so that an unjust combatant could say in mitigation, “I killed him, which was wrong, but at least I didn’t kill him from a safe distance.” But I will not discuss this further here, as the more important question is whether there are reasons to think that combatants’ associative duties can make it permissible for them intentionally to kill other nonliable combatants when the killing would be impermissible in the absence of the associative duties. (For a powerful critique of the idea that it is more seriously wrong to harm nonliable people who are defenseless than to harm otherwise similar people who are not entirely defenseless, see Jonathan Parry, “Community, Liability, and Just Conduct in War,” forthcoming.)
Lazar’s discussion of this issue is confusing because of the puzzling way in which he uses the terms “eliminative” and “opportunistic” to apply to acts of harming or killing. On page 33, for example, he restates his central claim in this way: “our duties to protect those we share valuable relationships with can override the duty not to kill a nonliable person, at least in 1:1 cases where the victim is killed foreseeably and eliminatively, rather than intentionally and opportunistically.” In discussions of the ethics of harming and killing, the term “foreseeable” is often used as shorthand for “foreseeable but unintended” and it is reasonable to assume that that is what Lazar means here. But if a person is killed only foreseeably and not intentionally, he is not killed eliminatively, in the sense in which the latter term is used in the literature. For the distinction between eliminative and opportunistic killing has hitherto been understood as a distinction between two forms of intended killing.
Warren Quinn, who introduced the distinction, first defined direct harmful agency as “agency in which harm comes to some victims…from the agent’s deliberately involving them in something in order to further his purpose precisely by way of their being so involved (agency in which they figure as intentional objects)” and indirect harmful agency as “harmful agency in which either nothing is in that way intended for the victims or what is so intended does not contribute to their harm.” He then suggests that the revised doctrine of Double Effect that he proposes might “strongly discriminate… against direct agency that benefits from the presence of the victim (direct opportunistic agency) and more weakly discriminate…against direct agency that aims to remove an obstacle or difficulty that the victim presents (direct eliminative agency).” (“Actions, Intentions, and Consequences: the Doctrine of Double Effect,” Philosophy and Public Affairs 18 (1989) , p. 344.) Both eliminative and opportunistic agency are thus defined as forms of “direct” agency—agency that affects a victim intentionally. One reason for understanding the distinction this way is that harm that is merely a foreseen side effect is not instrumental in eliminating a threat. But to call harming or killing eliminative is to acknowledge that it is instrumental in eliminating a threat from the person harmed or killed rather being a means of avoiding a threat of which that person is not the cause.
Lazar, by contrast, defines eliminative killing as killing in which “the killer derives no benefit from the victim’s death that he would not have enjoyed in the victim’s absence.” (19) He then observes that in his case 3 when the agent diverts the meteor away from the person to whom he is specially related but toward a nonliable bystander, the killing of the bystander is eliminative. This is not the way the term has been used by others, but this would not matter if it were not that the deviant use makes it unclear what Lazar means to say about the role that associative duties might have in justifying the most common form of killing of combatants in war—namely, killing to eliminate the threat that a combatant poses. Thus, in restating his central claim in the essay’s penultimate paragraph, he writes that “the argument advanced above was that in 1:1 cases, A’s duty to protect B from lethal harm can justify the foreseeable infliction of eliminative lethal harm on C, but that it cannot justify opportunistically harming C.” (43) The problem with this statement is that the lethal harm that Lazar describes as both foreseeable and eliminative could be either unintended or intended. The natural interpretation would be to read “foreseeable” as implying unintended. Yet the context is a discussion that purports to explain how the appeal to associative duties can help to justify the intended killing of combatants in war. And it is the justification of intended killing that is needed to support Lazar’s claim that unjust combatants “can sometimes be justified in fighting, and on much the same grounds as” just combatants. Yet what he concedes, in the same sentence, is that his argument cannot justify is opportunistic harming, not that it cannot justify intentional harming. Here he is repeating his earlier concession that associative duties “cannot justify the opportunistic killing that war inevitably involves.” (34) In these passages I find that I simply do not know what he means to be asserting. The suggestion that war inevitably involves opportunistic killing suggests that he may understand all killing as a means to be opportunistic. In that case, the killing of enemy combatants as a means of averting the threat they pose would be opportunistic. But this understanding of the term is incompatible with the definition he gives on page 19, which is consistent with Quinn’s. Yet according to this definition of opportunistic killing, it is not true that war inevitably involves opportunistic killing. The aims that are served by the killing of enemy combatants in war rarely require the presence of those combatants for use as a means. If the combatants were not there, there would be no need to fight. What war inevitably involves—indeed almost necessarily involves—is killing of enemy combatants that is eliminative, in Quinn’s sense, rather than opportunistic (though sometimes killing that is primarily eliminative can have an opportunistic dimension as well, as when killing is intended not only to eliminate the threat the immediate victims pose but also to intimidate the victims’ fellow combatants).
In the end, I cannot find an argument in the essay for the extension of Lazar’s central claim so that it also applies to the intentional, eliminative killing of nonliable unjust combatants. This is not to say that I think such an argument cannot be made. I suspect that it can. That is, I do not find it implausible to suppose that a just combatant’s associative duties do in some cases strengthen the justification for the eliminative killing of unjust combatants. But Lazar’s confusing and, I think, unnecessary invocation of the distinction between eliminative and opportunistic killing has been an impediment to his ability to produce such an argument.
There are, moreover, other good reasons why it can sometimes be permissible to kill nonliable unjust combatants; so we need not fear being compelled to embrace pacifism. One such reason is that unjust combatants are themselves responsible for making it reasonable for just combatants to believe that they are liable to be killed. Another is that when they are killed intentionally because their adversaries cannot know that they are not liable, killing them is morally like killing as a side effect. It does not come within the scope of the constraint against the intentional killing of people who are not liable to be killed. (For defenses of these claims, see Jeff McMahan, “Who is Morally Liable to be Killed in War,” Analysis 71 (2011), pp. 555-59.) A third and perhaps more important reason is that even when unjust combatants are not liable to be killed, they are normally liable to some degree of harm less than that involved in being killed. It might therefore be permissible to kill them if part of the harm they would suffer in being killed could be justified on the ground that they are liable to it, while the remainder could be justified on ground that it is the lesser evil when the alternative is to allow the achievement of their unjust cause.
Finally, I also cannot find an argument in the essay that explains how an appeal to associative duties can help to vindicate Lazar’s conclusion that “combatants on both sides of a war can, in some cases, fight justifiably.” That is, I cannot find an argument that shows that the associative duties of unjust combatants can justify their intentional killing of just combatants. And here I do think that no such argument, or at least a plausible one, can be made.