Targeted killings of the controversial sort that will be focused on here are exemplified by the following hypothetical case:
Al is a special-forces sniper whose team is assigned the mission of killing a certain terrorist. This particular terrorist is responsible for fabricating devices that are used by suicide bombers who detonate their explosives in places frequented by large numbers of civilians. Al’s team is provided with reliable information regarding the terrorist’s plan to travel by car to his bomb factory. Al and his team are flown by helicopter to a suitable location along the terrorist’s route, where they wait for him to arrive. When the terrorist does come within range, Al shoots and kills him.
Al’s action is seriously problematic from a legal standpoint. First of all, his killing the terrorist possesses none of the features that would render it permissible according to the criminal law. Al is certainly not acting self-defensively, for example. Moreover, the status of his action within the laws of war is at best unclear. These laws do permit members of opposing forces in wars to kill each other, even if doing so is not self-defensive, or is not otherwise necessary to avert imminent threats of serious harm. But the laws of war have traditionally been interpreted as applying only to conflicts between the military forces of political communities, or at least to groups whose members overtly distinguish themselves from civilian noncombatants.1 On this view, the laws of war are inapplicable to a political community’s conflict with terrorist organizations, and hence can provide no basis for concluding that Al’s action is legally permissible.
The moral status of targeted killings is problematic for analogous reasons. Ordinary moral principles permit homicide only when necessary to prevent the loss of life or comparably serious harms. This restriction does seem to be relaxed in various ways for killings within wars. In particular, combatants on opposing sides can be morally permitted to kill each other even when, in doing so, they are not responding to imminent threats of serious harm. However, this relaxation of ordinary morality’s prohibition against homicide would seem to require the existence of a special set of moral norms that—like the laws of war—apply only to wars strictly so-called; that is, these moral norms of war would apply only to conflicts between the military forces of political communities. Accordingly, a “morality of war” would not apply to the targeted killing of terrorists, and would provide no better basis than ordinary morality for establishing the permissibility of these killings.
I will argue here, however, that this explanation of the moral status of targeted killing is mistaken because of what it assumes about the implications of ordinary morality. I will argue more specifically that targeted killings can be morally permissible according to ordinary moral principles of self-defense.2 Central to my argument is the idea that individuals, like the terrorist in our example, perform actions that are connected in morally significant ways with the actions of others (of suicide bombers, for example). More specifically, the idea is that, while the terrorist is not himself acting aggressively, his actions can be components of an action that has multiple agents, and that is jointly aggressive. As will be explained in the course of the discussion that follows, ordinary moral principles of self-defense whose applicability is commonly restricted to individually aggressive actions, can also be applied to actions that are jointly aggressive.
Actions can be jointly defensive as well as jointly aggressive. This move in the direction of collectivizing defensive and aggressive actions might seem to imply that, in the final analysis, targeted killings must be viewed as acts of national self-defense that are responses to aggressive actions on the part of terrorist organizations. I will, however, provide reasons for regarding appeals to national self-defense as, at best, unhelpful in the present context.
As a useful first step towards explaining how principles of self-defense apply to targeted killings, let us examine these principles in a somewhat broader context.
Given how theories of self-defense are invariably formulated, their primary applications are to defensive actions performed by individuals in response to aggressive actions performed by other individuals. Judith Thomson depicts a hypothetical situation of this sort in her classic paper “Self-Defense and Rights”:
Suppose Aggressor has got hold of a tank. He has told Victim that if he gets a tank, he’s going to get in it and run Victim down. Victim sees Aggressor get in his tank and start towards Victim. It is open country, and Victim can see that there is no place to hide, and nothing he can put between himself and Aggressor which Aggressor cannot circle round. Fortunately, Victim happens to have an anti-tank gun with him, and it is in good working order, so he can use it to blow up the tank, thereby saving his life, but of course thereby also killing Aggressor.3
Assuming that, when Aggressor launches his attack, Victim’s actions are morally innocuous, and assuming too that no one else is involved in this situation, Victim’s action is a paradigm instance of morally permissible self-defensive homicide.
Of course, our example of Al and the terrorist is very different from Thomson’s. The terrorist is not acting aggressively, and Al’s action is not self-defensive, so the question of whether Al is permitted to kill the terrorist in self-defense does not even arise. Theories of self-defense almost invariably apply as well to other-defense, however. For example, a key component of Thomson’s account of self-defense is her claim that a person has no right to life if killing him is the only way in which to prevent him from performing an action that would violate another’s right to life (and other things are equal).4 Nothing in this account implies that it is restricted to self-defensive homicide. A person who does lack a right to life in virtue of satisfying the condition stipulated by Thomson can be permissibly killed by anyone in a position to do so.5
We must therefore consider whether targeted killings like the one carried out by Al could be morally permissible in virtue of being appropriately other-defensive. Taken in context, the following remarks by Andrew Altman and Christopher Heath Wellman suggest that they might answer this last question affirmatively:
Surely, it would have been permissible for someone to have assassinated Stalin in the 1930s. It seems, then, that political assassination is, in principle at least, a morally permissible means of stopping or halting human-rights abuses.6
These remarks raise an important question regarding the conditions that govern morally permissible defensive homicide. This question concerns what might be called the “inevitability condition,” according to which killing x in self- or other-defense is morally permissible only if x’s killing an innocent person would otherwise be inevitable.7
It might be thought that “highly probable” should replace “inevitable” in the aforementioned condition. But suppose that x is attacking y at time t, and that—at t—x’s killing y is highly probable if y does not kill x first. Suppose too that, as a matter of fact, x’s attack will fail regardless of what y does.8 Is y nevertheless permitted to kill x at t? I would argue that she is not. Probability considerations can, of course, play an epistemic role in the contexts we are examining. However, whether a defensive homicide is permissible depends on whether it is in fact the only way in which to avoid an innocent’s death—not on whether anyone reasonably believes that it is. Such epistemic considerations are (indirectly) relevant to moral culpability or blameworthiness after the fact, but they are irrelevant to moral permissibility before the fact.9
The underlying distinction here is between negative moral appraisals of actions and negative moral appraisals of their agents. My primary focus is on the former appraisals—not because they are more important than the latter—but because they are more basic in this respect: a person is morally blameworthy or culpable for performing a certain action only if the action is morally impermissible. Blameworthiness and culpability for acting also depend on what the agent could reasonably have been expected to believe about the nature of the action and its consequences; and here is where probability considerations enter into the picture.
The preceding remarks lead to a second point that needs to be made before proceeding—namely, that inevitability is distinct from imminence. That inevitability does not imply imminence becomes clear on considering the following variant of Thomson’s example:
Some Third Party (rather than Victim) has the anti-tank gun. Aggressor isn’t yet aware of Victim’s presence, but, when he does see Victim, he will launch his attack. Third Party can prevent Aggressor from killing Victim only if he fires his weapon before Aggressor begins his attack.
Third Party’s killing Aggressor in these circumstances would be preemptive other-defense, and would satisfy the inevitability condition referred to above, even though the threat posed by Aggressor is not imminent.10 Killing Aggressor would also be morally permissible according to some familiar theories of self-defense (including Thomson’s own theory).11
As it stands, however, our example of Al and the terrorist is not about preemptive other-defense, since nothing in the example suggests that the terrorist will inevitably attack and kill innocent people if he is not killed.12 But the example does seem relevantly similar to this variation on Thomson’s theme:
Prior to launching his attack on Victim, Aggressor must refuel his tank, and Accessory’s truck is the only available source of additional fuel. Accessory is happy to help, since she also wants Victim killed. The only way in which Third Party can prevent Aggressor from eventually attacking Victim and running him down with his tank is by destroying Accessory’s truck and Accessory along with it.
Third Party can save Victim’s life by—and only by—killing Accessory. Doing so would not be individually other-defensive, however (at least not obviously so), since Accessory is not attacking Victim. It is therefore difficult to see how theories that are aimed at determining the moral dimensions of individually defensive homicide can have anything to say about this case—or our example of Al and the terrorist.
Thomson’s theory is a case in point. Its central thesis (components of which were stated above) can be put as follows: killing x is morally permissible if x lacks a right to life; and x lacks a right to life if killing him is necessary to prevent him from violating another’s right to life (and other things are equal). Since it does not appear that killing Accessory is necessary to prevent him from violating Victim’s right to life, Thomson’s theory does not seem to imply that Third Party is permitted to kill Accessory. And, given the relevant similarity of this case to that of Al and the terrorist, Thomson’s theory does not appear to imply that Al’s targeted killing is permissible.13
Of course, self- and other-defense situations are not the only ones in which people are permitted to prevent the deaths of innocents by killing non-innocents. There are also what might be called “self-preservation” and “other-preservation” situations, the latter of which might be thought to include Third Party’s killing Accessory and Al’s killing the terrorist. By itself, however, this change of direction would be incapable of solving the problem at hand. This is because the permissibility of killing individuals in circumstances like those surrounding Accessory and the terrorist depends on connections between their actions and the actions of others.
Hence, an explanation of the permissibility of Third Party’s killing Accessory would need to incorporate an account of how Accessory’s and Aggressor’s actions combine to create a lethal threat for Victim. In a parallel fashion, the permissibility of Al’s killing the terrorist could not be explained without locating the terrorist’s actions within a nexus that includes the actions of others associated with his terrorist group—suicide bombers in particular. It is worth bearing in mind, by the way, that Al’s action was part of a joint effort that included contributions by many individuals serving in various capacities, some of whom (planners, for example) were not directly involved in the shooting.
These remarks shift the focus of this inquiry from actions that are individually defensive or aggressive to actions that are in some sense collectively defensive or aggressive. This shift in focus leads quite naturally to the idea that targeted killings are matters of national self-defense—that their permissibility follows from the role they play in defensive actions by political communities in response to aggression by terrorist organizations.
Certain targeted killings are morally problematic because neither ordinary morality nor (assuming there is such a thing) the morality of war seems to provide a basis for establishing their permissibility. Ordinary morality seemingly will not do because the targeted killings in question do not avert imminent threats of death or comparably serious harm to anyone. And since a morality of war would apply only to conflicts between political communities, it would be inapplicable to the killing of terrorists. However, if targeted killings were matters of national self-defense, then—contrary to much of what has been said to this point—principles of self-defense, understood as components of ordinary morality, might indeed be relevant to the morality of targeted killings.
It is important to recognize that the appeal to national self-defense that we are considering is meant to establish that individual targeted killings (for example, Al’s killing the terrorist) are morally permissible.14 For this to work, there must be valid arguments whose central premises are exemplified by “The United States has a right of self-defense,” and whose conclusions are exemplified by “Al is morally permitted to kill the terrorist.”15 But how do we proceed logically from the central premises of such arguments to their conclusions?
A natural first step would be to infer that, because political communities have a right of self-defense, they have a right to establish policies that are aimed at defending themselves against aggression. Then, assuming that targeted-killing policies have this aim, the next step would be to infer that political communities have a right to establish such policies. However, even if the argument were valid to this point, there would be no logical way in which to derive conclusions about the moral permissibility of particular targeted killings. For example, there would be no valid way to infer that, because the United States has a right of self-defense, Al is permitted to kill the terrorist. Hence, even if a political community is fighting a defensive war against terrorists, this has no logical bearing on whether any specific targeted killings are morally permissible.
The logical problems associated with appeals to national self-defense become especially clear on attempting actually to employ theories of self-defense to show that particular targeted killings can count as morally permissible defensive homicides.
Consider Thomson’s theory, for example, according to which x is permitted to kill y in self-defense if and only if y would otherwise violate x’s right to life (and other things are equal). Applying this theory to our example of Al and the terrorist, we have “The United States is permitted to kill the terrorist if and only if the terrorist would otherwise violate the United States’ right to life (and other things are equal).” Even if sense could be made of the idea that the United States has a right to life, this proposition is obviously incapable of being used as a basis for inferring that Al’s killing the terrorist is morally permissible. Moreover, a similar result would be equally obvious if other theories of self-defense (Uniacke’s or McMahan’s, for example) were appealed to.
While our example of Al and the terrorist might be locatable within the broad context of a conflict between the United States and the organization to which the terrorist belongs, explaining its defensive character requires a much narrower context. As was suggested in the preceding section, this narrower context cannot be restricted to Al and the terrorist, but must also include members of broader groups with whom Al and the terrorist are respectively connected in certain ways. For convenience, Al’s group will be referred to as an assassination team (consisting of members of the U.S. military and perhaps the CIA), and the opposing group will be referred to as a terrorist cell.16
It seems reasonable to say that the assassination team is responding to aggressive actions on the part of the terrorist cell. The fact remains that the assassination team does not kill the terrorist cell; rather, Al kills the terrorist. And it is the moral permissibility of Al’s action that is in question. Answering this question requires examining certain of the ways in which the actions of members of a group are related to actions of the group as a whole.
Note first of all that statements of the truth-conditions for propositions that do attribute actions to groups always contain references to appropriate actions on the part of members of those groups. For example, if it is true that Green Bay played in the 2011 Super Bowl, then this is because individuals who were members of the Green Bay Packers on February 6, 2011 performed actions of certain sorts on that date. And if it is true that the Royal Shakespeare Company performed Romeo and Juliet during March of 2010, then this is because of certain of the things done by members of the Company during that period.
Football games and dramatic performances are not simply collections of individual actions, however. More specifically, the truth-conditions for the proposition that the Royal Shakespeare Company performed Romeo and Juliet would not be equivalent to anything like this: x1 performed a1 at t1 and x2 performed a2 at t2 and .... In addition to such references to actions performed individually, a statement of the truth-conditions for the proposition in question would also include references to actions that are performed in concert or jointly with others—references that presuppose the concept of joint agency. Note too that, although the proposition on which we are focusing concerns a dramatic performance, not all of the actions to which it refers are individual dramatic performances. Some of these references are to actions on the part of individuals who manipulate scenery or control lighting, for example.
In a parallel fashion, if the assassination team is responding to aggressive actions on the part of the terrorist cell, then this is because members of the respective groups act jointly in certain ways. Moreover, just as the individual actions that compose a dramatic performance need not themselves be dramatic performances, so the members of the terrorist cell can be engaged in joint aggression even though not all of the actions composing their joint action are individually aggressive. Similarly, actions of the assassination team can be jointly defensive without all of their actions being individually defensive.
In order to elucidate the implications of these considerations for the permissibility of targeted killings, an account of joint action—and its companion concept of joint agency—is required. With this account in hand, it will be possible to explain how familiar theories of self-defense imply that targeted killings like the one in our example can count as morally permissible defensive homicide.
Joint actions differ from individual actions in that, while the latter are performed by single agents, the former are performed by multiple agents and have individual actions as components. Here is a homely example that illustrates this distinction:
Dale’s car has a dead battery. Roy offers to help with her problem by connecting her battery to his by means of jumper cables, and Dale accepts his offer. When Roy completes the connection, he signals Dale who is at the controls. She engages the starter and the car starts.
The car is started by Dale and Roy. However, while the proposition that Dale and Roy started the car is true, the proposition that Dale started the car and Roy started the car is false. We might interpret this proposition as implying that a pair of people started Dale’s car; that is, a group containing Dale and Roy as members. However, a more perspicuous interpretation would refer to the truth-conditions for the initial proposition.
A statement of these truth-conditions would refer to certain individual actions performed by Dale and Roy respectively, and to ways in which these actions are connected with each other. It is in virtue of the nature of the individual actions that are respectively performed by Dale and Roy, and of how these actions are connected with each other, that it is true that they started the car. That is, over and above the individual actions respectively performed by Dale and Roy, there is a joint action that consists in the starting of Dale’s car, and whose agents are both Dale and Roy.
In order for Dale and Roy to be acting jointly in the relevant sense, they must be exercising joint agency. And this latter concept can be explained in light of the following, more detailed version of the example:
Dale wants her car to be started. She believes that, if she engages the starter and Roy does his part, then the car will start. Roy wants Dale’s car to be started. He believes that, if he connects her battery to his and Dale does her part, then the car will start. Dale’s desires and beliefs lead her to engage the starter (call this action A). Roy’s desires and beliefs lead him to connect the two batteries (call this action B). Assuming that the car is otherwise in working order and the cables are properly connected, Dale’s performing A initiates a sequence of events that merges with the sequence of events initiated by Roy’s performing B, forming a sequence of events that results in the starting of Dale’s car.
While it is not true that Dale started her car, and it is not true that Roy started her car, each of them performs an action (A and B, respectively) that is a component of the joint action that consists in starting Dale’s car. They exercise joint agency in doing so, in virtue of the common contents of the beliefs and desires that lead them to perform the individual components of their joint action. Although Dale and Roy’s joint action and joint agency are distinct from their individual actions and exercises of agency, the former are explicable in terms of the latter. As a result, the only agents involved in the example are concrete individuals.
What has been said here about the Dale/Roy example can be generalized as follows:
x and y act jointly in bringing about state of affairs S if and only if (a) x and y desire that S obtains, and each believes that there is another person who also desires that S obtains; (b) x and y each believes that, if she acts on these beliefs and desires, and if the other person does so as well, then S will obtain; (c) x’s beliefs and desires lead her to perform action v, and y’s beliefs and desires lead him to perform w; (d) x’s performing v and y’s performing w initiate causal sequences that merge to form a sequence that results in S’s obtaining.
The joint agency exercised by x and y consists in the common contents of the beliefs and desires that initiate the merging causal sequences that produce S.17
This account of joint action and agency can straightforwardly be extended to situations containing more than just two agents.18 Moreover, joint actions can be composed not only of individual actions, but also of other joint actions. If, in our example, some friend of Roy’s helps him connect the cables and does so with appropriate beliefs and desires, then their joint action is a component of a larger joint action performed by Dale, Roy, and the friend.
Because joint actions admit of this sort of structuring, they can be quite complex, and involve large numbers of agents. As was pointed out above, plays and games provide contexts within which complex joint actions are commonly performed. So do construction projects, sessions of legislative bodies, and battles. Regardless of the composition of a joint action, however, its agents are always concrete individuals exercising joint agency.
Now recall our “Accessory” example, in which Aggressor and Accessory both want Aggressor to kill Victim, and in which he will do so if and only if Accessory is not prevented from refueling Aggressor’s tank. If Third Party were to refrain from acting, then the proposed account of joint action and agency would imply that Victim’s death would result from a joint action performed by Accessory and Aggressor. Theirs would be a joint action in virtue of the merging causal sequences resulting from their individual actions, and in virtue of the common contents of the beliefs and desires with which these actions would respectively be performed.
If Third Party were to destroy Accessory’s truck before he could refuel Aggressor’s tank, then Third Party’s action would be preemptively other-defensive. While killing Accessory would not preempt an aggressive action on his part, it would preempt a jointly aggressive action on the part of Accessory and Aggressor.
We can now return to our original example of Al and the terrorist. Let us assume that, if nothing is done to stop him, the terrorist will provide explosive devices to suicide bombers who cannot be prevented from detonating their devices and thereby killing many innocent people. This case is similar in obvious and significant respects to our “Accessory” example. That is, the terrorist bomb-maker and the suicide bombers (and perhaps others as well) are performing a jointly aggressive action that will result in the deaths of innocents unless something is done to prevent this joint action from being performed.
And there is only one way in which to prevent its performance: Al must kill the terrorist. If Al does so, then his targeted killing is an act of preemptive other-defense against joint aggression. And in virtue of the ways in which Al’s action is connected with individual actions performed by other members of the assassination team, together they perform an action that is jointly defensive.
Having explained how targeted killings can count as defensive homicides, we can now consider whether there are conditions under which they are morally permissible. Doing so will require examining the moral properties of joint actions.
Like individual actions, joint actions can be morally permissible, impermissible, or required. They can also be actions that their agents have a right to perform, or actions that violate the rights of others. Moreover, joint actions possess moral properties in virtue of possessing the same non-moral properties that determine the possession of moral properties by individual actions. If, for example, a joint action would result in the deaths of innocent people, then it violates the rights to life of those people, and is therefore morally impermissible (other things being equal).19
Additionally, the agents of joint actions can possess moral properties in virtue of the nature of their agency. In particular, they can be morally blameworthy or praiseworthy for contributing to the performance of joint actions which themselves have relevant moral properties. If, say, a joint action is morally impermissible, and if its agents perform their individual actions with appropriately bad intentions and lack excuses for what they do, then they are blameworthy for their contributions to the joint action.
Now, in the realm of individual actions, theories of self-defense typically imply that defensive homicides are permissible only as responses to actions that possess certain sorts of moral defects. According to Thomson, for example, an action possesses the relevant sort of moral defect if it will violate someone’s right to life if its agent is not killed. And according to Susan Uniacke and Jeff McMahan, the relevant defect consists in an action’s posing a certain kind of threat to others.20 The moral defects in individual actions to which these theories refer can also be present in joint actions. Specifically, killing one or more agents of a joint action might be necessary and sufficient to prevent the action from violating someone’s right to life. Or a joint action might pose the sorts of threats to which Uniacke’s and McMahan’s theories refer.
Hence, the theories to which I have alluded could naturally and plausibly be extended from individual defense to joint defense. Rather than attempting to develop any of these possibilities for theories proposed by others, however, I will do so for one that I have defended on a number of occasions.21
This theory focuses on situations in which individuals face “closed choices” in the distribution of harm.22 In a closed-choice situation, an individual x cannot prevent harm from befalling some members of a group G (that might include x), although x can determine which members of G are harmed. The theory implies that, if some member of G culpably created the closed choice situation, then, other things being equal, x is morally permitted (as a matter of justice) to distribute the harm to that individual.23
Closed-choice situations can clearly be created by multiple agents acting jointly, and these agents can be culpable for doing so. Justice permits the harm to be distributed to as many of these agents as is necessary to prevent it from being inflicted on innocent potential victims. Suppose, for example, that three individuals jointly and culpably create a situation in which x can prevent the death of an innocent person by—and only by—causing the death of one or more of the individuals who created the situation. Then x is permitted to cause the deaths of as many of these individuals as is necessary to defend their intended victim. This same line of reasoning is applicable to our “Accessory” example. It also applies to our original case of Al and the terrorist. It implies that Al is morally permitted to kill the terrorist, and that the killing is defensive in nature.
This result can be generalized, and applied to any targeted killing that is relevantly similar to Al’s killing of the terrorist in the circumstances that we are currently envisioning. These are killings that satisfy the following condition: a number of individuals are culpable for jointly creating a closed-choice situation in which killing some number of them is necessary and sufficient to prevent the loss of innocent lives. This condition is satisfied, as many of the individuals who created the closed-choice situation can be permissibly targeted and killed as is necessary to prevent the loss of innocent lives. Such killings would count as morally permissible defensive homicide even if those who are killed are not themselves performing actions that are individually aggressive.
Although my specific concern here is with the question of whether targeted killings can be morally permissible, the approach to answering this question that I have proposed has broader implications for the morality of killings that occur within wars.
As was pointed out much earlier in the discussion, belief in the need for a special morality of war is based largely on the proposition that certain homicides in wars that seem clearly to be permissible, turn out to be impermissible according to the principles of ordinary morality. For example, shelling trucks that are carrying reinforcements to the front lines can be morally permissible. Yet, assuming that those reinforcements pose no imminent threats to anyone, shelling them would appear to be prohibited by ordinary morality.
In order for the proposition that there exists a special morality of war to be even minimally plausible, however, the kinds of conflicts to which it applies must be narrowly restricted. Traditionally, the morality of war has been restricted to conflicts between political communities. But tradition is no substitute for an argument and, in fact, the idea of a special morality of war cannot withstand close scrutiny. Its weakness has become especially clear in recent years, with increases in the occurrence of “asymmetrical” wars.
The problem here has a number of sources, but one that seems to me to be especially important, has received almost no attention. I refer to the assumption that wars must be understood as waged either by political communities per se, or by members of those communities acting individually. This assumption is a barrier not only to providing an adequate basis for morally appraising acts of war, but even for accurately describing the events that occur within actual wars.
In addition to references to actions on the part of political communities and on the part of combatants acting individually, references to joint actions are necessary for both moral appraisals and descriptions of acts of war. Indeed, wars should be thought of as being composed of joint actions whose agents are members of the opposing sides. Some of these joint actions are extremely complex, while others are not. Compare, for example, the D-Day invasion with an attack by an American patrol on a German pillbox that occurred during that invasion.
Now, neither the invasion nor the patrol’s attack appears to be defensive. But both occur within the context of a response by the forces of a number of political communities to aggression by the forces of other political communities. As was pointed out earlier, a complex joint action can be defensive (aggressive) even if not all of its components are defensive (aggressive). And the aggressive component of a defensive joint action can be morally permissible according to ordinary morality because the defensive joint action is morally permissible. In this way, principles of ordinary morality can be extended to actions that are commonly regarded as open to moral appraisal only within a special morality of war. These ordinary moral principles apply not only to actions that are responses to culpable aggression, but also to those (alluded to in n. 23) that are responses to nonculpable but impermissible aggression.