TARGETED KILLING AND THE STRATEGIC USE OF SELF-DEFENSE
TARGETED KILLING AND THE STRATEGIC USE OF SELF-DEFENSE
Abstract and Keywords
Although governments that engage in targeted killing will usually invoke self-defence as a primary justification, they often seem on thin ice in doing so. We frequently have the sense that there is something very strategic and dishonest in the way self-defence is utilized in such cases. The most obvious example is the case in which the government uses an agent provocateur to induce a potential terrorist to take actions for which it then proceeds to kill him — in ‘self-defence’. This chapter examines a series of such strategic situations and suggests that, appearances notwithstanding, there is in fact nothing illegitimate about using self-defence in such strategic ways. That is not to say that our persistent unease about what governments are doing in such cases is without significance. But its significance turns out to lie mostly in what it teaches us about certain counterintuitive features of legal and moral rules, which is another that is explore here.
Although governments that engage in targeted killing will usually invoke self-defense as a primary justification, they often seem on thin ice in doing so. We frequently have the sense that there is something very strategic and dishonest in the way self-defense is utilized in such cases. The most obvious example is the case in which the government uses an agent provocateur to induce a potential terrorist to take actions for which it then proceeds to kill him—in “self-defense.” In this chapter I will examine a series of such strategic situations and will suggest that, appearances notwithstanding, there is in fact nothing illegitimate about using self-defense in such strategic ways. That is not to say that our persistent unease about what governments are doing in such cases is without significance. But its significance turns out to lie mostly in what it teaches us about certain counterintuitive features of legal and moral rules, which is another issue I aim to explore here.1
I. Take no prisoners
Consider the following scenario. Our government has learned that a long-sought major terrorist is holed up in a secret compound on the territory of a friendly power. With that other government's assent, we send a special forces unit into the compound to “take him out.” We do not expect him to surrender. More importantly, we do not really want him to surrender, because bringing him back home and putting him on trial is going to mean major legal headaches, in addition to the danger of terrorist reprisals.2 To this end, the President actually vetoes an initial (p.468) proposal that would have had us land in the target's compound with such overwhelming force as to possibly intimidate him into surrender. Instead, the President sends in a much smaller elite unit, which our target is likely to try to resist, thus justifying our shooting him on the spot. Let us suppose that things turn out exactly as planned: the target resists and gets shot. The government invokes the self-defense rights of legitimate government enforcement agents to justify what it did. Others, however, have suggested that there is something improper when the need for such self-defense has been strategically contrived by sending in a smaller force than would have been required to bring him in alive. After all, if the President had simply told his men to take no prisoners, that would surely have been illegitimate; but isn't what he actually did the functional equivalent, and therefore not really permissible either? Or so the critics’ argument goes.
What should one make of this criticism? Standard self-defense doctrine holds that you forfeit your right of self-defense if you provoke the attack against which you are defending. Walking into a bar, goading your worst enemywith everything you can think of until he loses his self-control and lunges at you, and then finishing him off—that is the kind of case where we generally do not want to view the person acting in self-defense as justified. Isn't this what the government has done when it sends in a smaller-than-optimal special forces unit that then serves as something akin to an agent provocateur?
Provoking an attack to use one's right of self-defense is a particular instance of a phenomenon sometimes described as contriving to create the condition of your own defense. German criminal law calls this the problem of the actio libera in causa—an “action free in its origins.” What is being referred to are cases in which someone deliberately, or maybe even just recklessly, creates situations in which he then finds a need to commit an act that would ordinarily be considered justified or excused by reason of necessity, or duress, or insanity, or lack of a voluntary act, or innumerable other potential defenses, but inasmuch as he deliberately contrived to create the situation, or just impermissibly risked its coming about, he no longer seems entitled to invoke those defenses. His actions being “free in their origin,” though no longer entirely “free in their execution,” deserve to be punished just as though he had no defense to begin with. That at least is the standard view.3 (p.469)
But why exactly is that? Why should contriving to create the conditions of one's own defense deprive one of the right to make use of that defense? One line of argument that has been offered to explain why that should be so goes something like this. Consider the case in which a storeowner contrives to create a situation of necessity so as to be able to do damage to his closest competitor's storefront without having to fear criminal liability. To this end, he starts driving at breakneck speed down the narrow alleyway where he knows the rival's store to be located. He anticipates that when he gets close to the store, there will probably be a cluster of people congregating in front of it, and that he will then face the choice of either running into them or veering to the side and damaging his rival's storefront instead. Naturally he would do the latter, thus getting to inflict the desired damage on his rival with impunity by invoking the law of necessity. Clearly he should not by this subterfuge be able to avoid liability. But how exactly should this kind of situation be analyzed? On what grounds could we actually deny him the necessity defense? What the most thoughtful commentators of this situation have said is that we should think about the matter in the following way. There is no doubt that the defendant acted rightly when, at the moment of encountering the cluster of people, he decided to veer into the storefront instead, and he should not on that account be held liable. He really did have a valid necessity defense at that moment. However, he did not have a valid claim to a necessity defense when he inaugurated the chain of events that led to this denouement—in other words, at the moment at which he first began hurtling down the alleyway. All that liability requires is that the prosecution be able to identify a moment in time at which he committed an act that constituted the forbidden actus reus of damaging another person's property, and that he did not at that time have a valid defense for that act. Therefore, when causing his car to hurtle down the alleyway, what the defendant did could probably be described as the first step of damaging another person's property, and for that step he did not have a justification. Necessity certainly did not require him to race down the alleyway. It only required him to swerve once he was already doing so. Ergo, he is liable.4 (p.470)
Possibly an even more compelling way to make this point is the following example. A person jumps out of a window, so as to land on top of his intended victim, whom in this fashion he kills. He then claims the involuntary act defense, because at the moment at which his body made contact with his victim's he was not committing a voluntary act. We would deny him the defense because we would view the involuntariness of his landing on his victim's head as irrelevant: he engaged in a voluntary act when he threw himself out of the window and that voluntary act qualifies as the actus reus required by a homicide—“causing another person's death.”
Interestingly enough, however, this argument against the legitimacy of contrived defenses does not work all that well when applied to contrived self-defense. The problem has to do with proximate causation. We are going to feel uneasy saying about the defendant who goaded his victim into an ill-considered attack, to which the defendant then responds with lethal force, that he (the defendant) inaugurated a chain of events at the end of which he may have had a legitimate right of self-defense but at the beginning of which he did not. The reason we feel uneasy is that the chain of events he inaugurated did not proximately trigger the victim's death. It did not do so because the victim's response seems like the kind of intervening act ordinarily considered to break the chain of proximate causation, thereby negating liability. To be sure, the matter is far from clear-cut. Whether the provoked victim's actions qualify as an intervening act is an uncertain matter under the usual understanding of the intervening cause doctrine. The arch-typical intervening act is one in which the intervening actor intends to bring about the consequence he brings about. Since the victim here quite obviously was not trying to bring about his own death, he is not an arch-typical intervening actor. On the other hand, if the provoked attacker had succeeded in killing the provoker, we would clearly not judge the provoker to have proximately caused his own death (inasmuch as in that case the intervening actor would in fact have brought about the intended consequences of his actions). But inasmuch as the provoked attacker's death is the result of an action by him that would be considered a non-proximate consequence of the provoker's actions—the attempt to kill the provoker—we seem to have pretty strong grounds for considering the provoker's action to be a break in the chain of proximate causation, even if the end result is not the provoker's but the provoked attacker's death.
One might, however, try an entirely different tack altogether to argue for the impermissibility of the provoking defendant's making a legitimate self-defense claim. Self-defense is legitimate if it is truly necessary for the defendant to save his life. But if he provoked the victim, was it truly necessary? Could he not simply have saved his life by not provoking him, rather than by provoking and killing him?
The difficulty with this argument becomes apparent if we try to make the time period involved in the process of goading a very extended one. Suppose a woman embarks on a relationship with a man who has some track record in starting to (p.471) behave abusively toward his partners many months, or even years, into the relationship. She is well aware of this. Do we want to say that she loses her right of self-defense on the ground that she could have avoided the whole problem by not entering into the relationship with him in the first place? What this brings out is an important ambiguity in the notion of necessity as it enters into the definition of self-defense. We do not really mean that you are only entitled to self-defense if that is logically necessary in some absolute sense. Necessity is relative to background circumstances. And we are surely willing to restrict the background circumstances that we allow as being relevant, most especially by only focusing on recently developed background circumstances. So for instance if our elite special forces units are kept generally small, in part with the thought that that might invite victims to engage in ill-considered resistance which might then justify killing them, it is hard to believe that anyone would want to seriously argue that they were not entitled to kill in self-defense because it was not strictly speaking necessary to do so, if one only took enough of a long-range view of the matter.
A third way to argue for depriving the special forces unit of their right of self-defense is to say that what we have here is a form of entrapment. In a literal sense, we arguably do, but in a legal and moral sense, the situation is far from clear. What we are talking about here certainly is not the entrapment defense in the strictly legal sense—where it is simply an argument someone is allowed to raise against being convicted of a crime into which the government entrapped him. This is not a case of a government seeking to punish someone for something they entrapped him into doing, for the simple reason that they killed him before the issue of punishment even arose. Moreover, if they had tried to arrest and prosecute him, he would have had trouble making a doctrinally sound entrapment defense because it is usually restricted to non-violent crimes, and putting even that technical issue aside, he might have had trouble proving the lack of predisposition that the entrapment defense requires. But more important than any of these slightly formalistic points is the fact that the entrapment defense is hardly an uncontroversial moral principle. In fact, its moral footing continues to be much debated and is if anything more widely questioned than the restriction of the right of self-defense with regard to provoking agents. In a similar vein, one might try to make some sort of estoppel argument against the special unit's claim of self-defense. But estoppel arguments are on an even less sound footing than entrapment, from which in this case they seem barely distinguishable.
But rather than continuing with what I have been doing so far—which is to answer individually every separate argument that has been raised against allowing a provoker to raise the claim of self-defense—it is perhaps more effective to respond in a more general way to the skepticism regarding the provoker's right of self-defense. But that will require some stage-setting.
Elsewhere I have argued—in a joint paper with the economist Alvaro Sandroni—that all rule-based decision-making produces certain kinds of cycles and that these (p.472) cycles necessarily invite maneuvers of the kind that the provoking agent is engaged in. Let me here give a very brief sketch of that argument.5
First, a bit of background. Sandroni, who constructed the model on which our argument builds, had been interested in the kind of cyclical choices that psychologists often hold up as the hallmark of human irrationality. Here are two telling examples.
(1) When someone is given the choice between watching movie X and watching movie Y, he chooses to watch movie Y. When given the choice between watching movie Y by himself or with a handicapped person, he chooses to watch it with a handicapped person. But when given the choice between watching movie Y with a handicapped person or movie X, without a handicapped person, he chooses movie X. In other words, his choices are cyclical. Why is that? What he is doing seems bizarre and irrational, at least from a purely formal point of view. Informally, it is not so hard to make out what is going on here. The person would be embarrassed to admit to not wanting to watch a movie in the presence of a handicapped person, which is why he chooses to watch it with him rather than without him. But given the opportunity to escape doing so by watching an entirely different movie, which he has the option of watching all by himself, he has no trouble doing so, because he can say that he happened to prefer that movie, though of course we know the opposite to be true, since we know that given the opportunity he would choose Y over X.
(2) When someone has to choose between staying in the office and working, or going home to relax instead, he would opt for going home. When given the choice between going home to relax or visiting a sick friend in the hospital, he would opt for visiting his friend. Alas, when given the choice between staying in the office and working, or visiting a sick friend, he would choose staying in the office. Once again we have a cycle, which at least on a formal level seems strange and perplexing. Of course, informally we find nothing strange and perplexing about it. The person is too embarrassed to admit that he would rather stay away from the hospital, and since he can only do so if he has work as an excuse, he chooses to stay at the office.
Sandroni pointed out that a natural way to describe, or model, what is going on here is to think of the chooser as maximizing his preferences subject to the requirement that he respect, first, the usual feasibility constraint (also known as the budget constraint)—that is, only choose something that is actually available to him—and second, and this is where Sandroni's interesting innovation comes in, the constraint that one's choice be “rationalizable,” his term for a choice that can be justified in terms of applicable legal or moral rules. That extra constraint somewhat unexpectedly gives rise to cycling. The cycling in turn gives rise to the possibility of (p.473) certain strategic actions that allow one to attain a seemingly forbidden objective by being manipulative. One is supposed to visit his friend, but one can avoid doing so by making sure not to be home but in the office. One is supposed to not discriminate against a handicapped person, but one can nonetheless do what comes to the same thing, by choosing a movie different from the one the handicapped person is going to be watching, and so on.
As we point out in our joint paper, these situations have ready analogues in the law, most especially in the law of criminal law defenses. Consider the law of duress. If I am threatened with torture, unless I have committed a rather serious crime, including something as serious as killing several people, I might well qualify for the duress defense if I submit: the law cannot expect me to be a hero, the explanation would presumably run. By contrast, if I were threatened with the destruction of a manuscript I have devoted my life's work to, unless I commit the identical crime—killing several people—I naturally will not qualify for the duress defense if I submit. Now suppose that I have in the past endured significant torture just to save the manuscript. Then it seems we just landed in another cycle, this time a legal one. As between suffering the destruction of the manuscript or being tortured, I would rather be tortured. As between torture and committing several killings, I would rather carry out the killings. Yet as between committing several killings and suffering the destruction of my manuscript, I would rather allow my manuscript to be destroyed (because I will do what's necessary to avoid doing something illegal). This cycle too gives rise to possibilities of circumvention, like the psychological cycles considered earlier.
For instance, imagine that I am able to protect myself against the people who are threatening the destruction of my manuscript (unless I do what they want) by borrowing a lot of money from a loan shark which I then use to hire people to safeguard my manuscript. When it comes to paying off the loan shark, I do not have the funds, as a result of which he threatens to torture me unless I commit several killings in his behalf. Now I am likely to qualify for the duress defense. If one tried to create doctrines that made such exploitation of legal cycles impossible, they would turn out to be tantamount to the most extreme and unyielding kind of utilitarianism—the sort that would inflict punishment every time I choose to do anything at any moment in time that does not maximize overall social utility, more or less understood in traditional Benthamite ways. (If there are other ways to block the exploitation of legal cycles, they are going to be at least equally unappealing.)
There are entirely uncontroversial examples of this kind of cycle in the law of self-defense. Suppose a robber says to me “Your money or your life.” I stand fast and refuse to give him my money, whereupon he attacks me, whereupon I kill him, which would, of course, be perfectly legitimate on my part. Hidden within this simple interaction is the exploitation of a cycle much like the one involved in duress. Given the choice between killing a thief and losing my money, I am required by law (p.474) to lose my money. Given the choice between losing my money and losing my life, I am free to lose my life. Given the choice between losing my life and taking the attacker's life instead, I am free to kill my attacker. The result is that if I want to get around the prohibition against choosing my money over my attacker's life, I can do so by first choosing to put my life at risk for the sake of my money, and then saving my life by killing the attacker. Obviously, something analogous is really being done when I contrive to create the circumstances of self-defense by putting an agent provocateur—the special forces unit—in the victim's vicinity.
II. Redirected self-defense
Let us now turn to a different type of targeted killing case, raising a different set of questions about the strategic use of self-defense. Let us suppose that there lives on the territory of a generally unfriendly foreign power a peaceable holy man who has inspired a fanatical, violence-prone following. Members of his movement are in the process of preparing a violent attack upon us. It would solve our problems if we could eliminate the holy man with a targeted killing. But he is not actually doing anything to deliberately foment acts of aggression against us. In fact he is not only opposed to such actions, he strongly counsels against them. But it does no good. Those who like his spiritual message are led, in significant numbers, to regard our nation as the devil and to plan on violent actions against us.
It would be convenient for us to get rid of him, but it seems we cannot. Neither self-defense nor necessity, as conventionally understood, would seem to allow us to do anything to bring him down. There is no ground for self-defense because he is not actually attacking anyone. And to kill him on grounds of necessity would be a classic case of killing one innocent to save a larger number, which the defense of necessity, without more, would not usually be held to permit. The excuse of duress, I suppose, might fare a little better, but not much: first, because its scope might not reach that far, since we are not quite in the classic duress situation where someone tells us to do something lest he injure us (admittedly, this is a somewhat controversial restriction on the duress defense), and second, and more importantly, because of the interesting fact that excuses are not the kind of defense a government can readily invoke. (To be sure, there is something mysterious and perplexing in this which seems worth exploring, but I will not concern myself with it here.)
Now suppose that the holy man's followers are indeed about to launch their long-planned attack against us. They will do so from a place in which they are surrounded by many innocent civilians. We are considering a preemptive attack upon them that would take out the attackers and their weapons just as they are about to strike. It has the unfortunate drawback of costing the lives of many innocent (p.475) civilian bystanders, but given that we are essentially responding to an attack by another country upon us, we would be well within our rights. Suppose further, however, that all of this could be prevented, if instead of preempting the attack along such conventional lines, we simply arranged for the death of the innocent holy man. Would that be legitimate? Notice that in doing so we are not simply saving some innocents who are about to die by taking the life of some other innocent. Rather, we are choosing to substitute for the deaths of several innocents whom we are entitled to kill, despite their innocence, the death of another innocent whom we are not entitled to kill.
But will this in fact work? Such an argument is usually dismissed because it sounds at first indistinguishable from the standard utilitarian claim that it is all right to do whatever achieves a net saving of lives, which many would dismiss by drawing the analogy to organ-harvesting doctors and lynch-mob-appeasing prison wardens. But in fact our situation is different in some crucial respects, and the net saving of lives argument here is on much stronger, even if not entirely unassailable, ground.
To see why, let's begin by adapting the infamous trolley example to our kind of situation. Imagine that we could take preemptive action against would-be attackers by building a long ramp, an incline, down which we could chase a trolley which would run over, and kill, the innocents that happen to surround our would-be attackers and in due course would kill our would-be attackers as well. This is something we are obviously entitled to do.6
Now add the familiar second track to this incline, such that if we were turn the trolley, it would run into our holy man instead, killing him, sparing the innocents, and stopping in their tracks the attackers who depended on the sustenance they drew from his message. Would it not be perfectly acceptable to turn the trolley here, just as it is in the standard trolley case?
To be sure, we cannot ordinarily use this kind of a trolley arrangement to liquidate just anyone by setting things up in such a way that we take aim at one group of innocents and then avoid hitting them by turning the trolley onto a second track. But that is because we are ordinarily not entitled to take aim at a group of innocents in the first place. Here things are different, because we are entitled to take aim at them, in the course of exercising our right of self-defense. Turning the trolley now seems an entirely different matter. (p.476)
Most of the details of the trolley arrangement are surely inessential for this argument to work. If we simply directed a strike against the holy man, that seems just as permissible as engaging in the trolley maneuver—once one has registered that it is not essentially different from the trolley case. And so it would seem that at least in a context in which we are entitled to take innocent lives in the course of defending ourselves—and probably for other reasons as well—we will often be entitled to cash in that chit for other innocent lives, in this case for a targeted killing.
Some people are going to find my analysis here too coarse-grained. There are many situations in moral theory in which chits like this just do not work—you really cannot cash them in, however Pareto-optimal it would be if you could. A good illustration is what one might call the problem of delayed self-defense: take an ordinary case of self-defense, in which rather than killing my attacker outright I allow myself to be injured by him, hoping to survive his attack without having to kill anyone. Having been injured badly, however, it turns out I can only survive with a heart transplant, for which he is the only potential donor. Can I avail myself of his heart? Can I claim that this is a kind of extended self-defense? Some people might say yes (Michael Moore has so argued); most, however, would say no. Not killing him in the first place did not give me a chit I can cash in later on, however mutually beneficial it would be for him and me alike if I could.7
Some people would claim that my holy man case is just such a case where the logic of chits fails, where the right to commit one killing cannot be exchanged for the right to commit a different killing that happens to have fewer victims. They might invoke the famous fat man example posed originally by Judith Jarvis Thomson in one of her well-known articles on the trolley problem. She asks us to consider the following variation of the trolley case: “As before, a trolley is hurtling down a track towards five people. You are on a bridge under which it will pass, and you can stop it by dropping a heavy weight in front of it. As it happens, there is a very fat man next to you—your only way to stop the trolley is to push him over the bridge and onto the track, killing him to save five. Should you proceed?” The widely shared intuition about this case is that doing so is off limits. The natural question to then ask about my holy man case is whether killing him to spare those innocent civilians is not rather like killing the fat man to spare the five on the track?
My own sense is that the holy man scenario differs in some crucial ways from that of the fat man. To see this, consider a variation of the fat man case that is rather closer to the holy man case. Suppose that, as in the original trolley problem, we can spare the five by diverting the trolley onto another track. But let us also suppose that on that other track there is not just one potential victim, but two. (p.477) Finally, suppose that we only push the fat man onto that second track once we have diverted the trolley toward those two victims. My own intuitions about this case would permit us to do this. By diverting the trolley, we are choosing to target two innocents to save the life of five other innocents. We did so permissibly, of course, but that does not change the fact that we are actually killing those two, rather than merely letting them die. If we can spare those two innocents by substituting some smaller number of innocents, then that seems to me acceptable because we are no longer simply saving some innocents, in whose death we had no active part; rather, we are choosing to avoid actively killing a certain number of innocents by killing some other smaller number of innocents, a small but significant difference.
III. A back-up argument: minimizing rights violations
A standard argument for a targeted killing is the following. If we do not eliminate the potential attacker, or for that matter the holy man, on the grounds that we would be violating their rights if we did, we are thereby facilitating a much greater number of rights-violations down the road. To begin with, there are the rights violations carried out by the would-be attacker and by the followers of the holy man, but those are not the only ones to worry about. There are also the rights-violations that are likely to be carried out by our own side in the course of any conflict: unjustified detentions of suspicious aliens, abrogation of civil rights in the interest of national security and so on. Not that we think we are justified in engaging in such conduct—if we were, they would not be rights-violations—it is just that we can statistically predict that there will be many of those, whether we like it or not. That is just how it is. If one cares about rights, surely one should prefer a single rights-violation now in lieu of many more later on.
This kind of argument is what Robert Nozick called a utilitarianism of rights, and he viewed it as one which an advocate of rights, or at least a deontologist, would generally reject.8 If one is a deontologist, one ought to respect rights, not seek to minimize rights-violations. That means, somewhat paradoxically, that one ought to avoid violating a right here and now, such as the right not to be the victim of an unjustified targeted killing, even if the person or governmental entity that is respecting that right would be causing more rights-violations in the long run, such as more individuals being the victim of unjustified killing.
I am willing to go along with that. What I am not willing to go along with is a conclusion usually drawn from it: namely that it would be better if one did not violate a right than if one did. “Would be better” turns out to have several possible senses, and while it might well be better not to carry out the targeted killing in the (p.478) sense that it is what one ought to do, it turns out, on reflection, not to be better in some other, equally important sense—namely in the sense that one might well be less blameworthy if one were willing to engage in such a violation here and now. In other words, if someone were to review the moral ledger of someone who decided to respect a certain right here and now at the expense of violating many more such rights later on, and compared that to the moral ledger of someone who ignored that self-same right here and now, so as to forestall having to commit many more rights-violations later on, this last person might well end up with the better moral ledger.
To see this more clearly, let's consider for a moment the venerable Regina v. Dudley and Stephens, the lifeboat case in which Thomas Dudley decides to kill the youngest of the sailors, Richard Parker, so that he and the other two sailors, Stephens and Brooks, would have a shot at survival by feeding on his remains.9 Let us suppose, not implausibly, that Dudley had acted with criminal recklessness in taking a comparatively slight boat like the Mignonette on an ocean voyage to Australia—so criminal in fact that if Parker, Stephens and Brooks had all died in the resulting shipwreck, he would be guilty of their murder. Finally, let us suppose that we agree with the court's judgment that Dudley's killing of Parker was wrong, that it was too much like the classic utilitarian horror stories in which killings that achieve net savings of lives are nevertheless impermissible, like those I alluded to earlier: the organ-harvesting surgeon who saves five people one of whom needs a heart, two of whom need kidneys, and two of whom need lungs, by killing a fifth; or the prison warden who surrenders a prisoner to a lynch mob so as to avert a riot in which many more innocent lives would be lost.
Let us now compare what would have happened if Dudley had done the right thing, and abstained from killing Parker, and what would have happened if he had done what he actually did, killed him. In the first case, he would have been guilty of three murders—that of Stephens, Brooks and Parker. In the second case, he would only be guilty of what he actually was found guilty of: the murder of Parker. In short, Dudley's legal position is improved by doing the wrong thing.
What I am imagining here is not as unusual as it might sound. After all, all we really need to construct this kind of scenario is to take any one of the classical utilitarian dilemmas—like the organ-harvesting case or the lynch mob case—and to imagine someone who has wrongfully inflicted some kind of serious harm that he could mitigate if he engaged in one of these unpalatable utilitarian trade-offs. Imagine, for instance, a ruler who first whips up the frenzy of the crowd against an innocent prisoner, but then wants to backtrack. He will minimize his blameworthiness by agreeing to the execution of the innocent prisoner, so as to prevent the lynch mob from doing even worse things, for which he could otherwise be blamed, given his role in whipping them up.10 (p.479)
What this means for a government engaged in targeted killings is that even if we consider those killings illegitimate, either because the arguments I have put forward in earlier sections regarding their legitimacy are not persuasive, or because they do not fall within the range of situations to which my arguments would apply, they might well serve to improve the moral position of the government engaging in them. By engaging in such killings, even if they are illegitimate, the government might well be reducing the overall amount of illegitimate conduct it will end up being guilty of. That does not mean that it did the right thing, not by any stretch. Only that it will end up with a better moral ledger, as it were, than if it had done the right thing!
This is without a doubt a very paradoxical situation for the government to be in. It means that if they were to ask God what he would recommend they do, he would tell them not to commit the targeted killing. But if they then proceeded to ask him whether they would be more likely to enter the pearly gates if they engaged in the targeted killing than if they did not, he would tell them that targeted killings are the way to go.
IV. Root causes
I have laid out in this chapter three lines of argument for certain defensive uses of targeted killings. Each of them has something strange and paradoxical about it. But what exactly makes them so strange and paradoxical? What makes them that, in my opinion, is that they reveal features of legal rules and legal reasoning that we would regard as irrational in the course of ordinary decision-making.
Start with the strategy of provoking intense resistance to an arrest so that the special forces can then simply take out the intended target, rather than having to arrest him. The legitimacy of this strategy depends on the ubiquity of cyclical, or intransitive, choices in a legal system, something which many would regard as the hallmark of an irrational person, if an individual decision-maker were to exhibit it.
The strategy of killing the holy man as an alternative to killing other innocents in the course of self-defense violates what is sometimes called the principle of independence of irrelevant alternatives. The standard joke used to illustrate this particular principle of rationality is the man in a restaurant who, having chosen to order steak rather than chicken, then learns that he overlooked that there was also fish on the menu, and promptly changes his order to chicken. Absurd, of course, yet something analogous is going on in our evaluation of the holy man scenario. As between killing the holy man or letting a lot of innocents be killed as a result of the pernicious influence of his teachings, we are ordinarily inclined to opt for the latter and reject the option of killing the holy man. But if a third option—that of killing other innocents in the course of self-defense—is introduced, then suddenly (p.480) the killing of the holy man becomes acceptable. That third option plays the same role as the fish alternative in the restaurant and it is thus bound to seem strange that it should affect the outcome.
Finally, the strategy of minimizing rights-violations goes against a desirable feature of rational decision-making that decision theorists like to call non-monotonicity. We expect sound moral and legal choices to have the basic property that if we choose to do the right rather than the wrong thing, we end up in a legally and morally better position than someone who chooses to the wrong rather than the right thing. (We expect moral choices to be monotonically related to overall moral status.) It just happens not to be true.
There is in fact nothing accidental about the fact that legal reasoning violates these canons of rational decision-making. It has long been known that collective decision-making routinely violates these canons. Those are the familiar paradoxes of social choices—the impossibility theorems of Arrow, Sen and others. It has also come to be accepted that individual decision-making, if it seeks to reflect a multiplicity of judgmental criteria, closely resembles collective decision-making and produces analogous paradoxes. But that is exactly what legal decision-making is: individual decision-making based on a multiplicity of criteria. This fact is sometimes disguised by the fact that we are applying rules, but the rules themselves are a device for aggregating a multiplicity of criteria into a decision-making device. Legal decision-making thus cannot avoid those paradoxes either. Our paradox-fraught assessment of targeted killings is just a special case of all that.
What I have so sweepingly asserted in this last paragraph is really a mouthful. I certainly do not consider it to be obviously right—though I do consider it to be right, else I would not have asserted it, just not obviously right. Indeed, I have made the defense of this thesis the subject of an entire book called Why the Law Is So Perverse.11 This larger claim about the root causes of the dissonance between the actual solidity and the apparent feebleness of the arguments for targeted killings is thus one the reader has to either take on faith—or read the book.
(1) Nils Melzer, Targeted Killings in International Law (Oxford University Press, 2008)
(3) The problem of the actio libera in causa has received surprisingly little attention in the English language literature. The exceptions include Miriam Gur-Arye, The Actio Libera in Causa (Harry Sacher Institute for Legislative Research and Comparative Law, 1984); Paul H. Robinson, “Causing the Conditions of One's Own Defense: A Study in the Limits of Theory in Criminal Law Doctrine,” 71 Va L Rev 1, 3 (1985); Larry Alexander, “Reconsidering the Relationship Among Voluntary Acts, Strict Liability, and Negligence in Criminal Law,” 7 Social Philosophy and Policy 84 (1990); Douglas Husak and Brian McLaughlin, “Time Frames, Voluntary Acts, and Strict Liability,” 12 Law and Philosophy, 95 (1992); Michael S. Moore, Act and Crime: The Philosophy of Action and its Implications for the Criminal Law, (Oxford University Press, 1993), 35–6; Leo Katz, “Proximate Cause in Michael Moore's Act and Crime,” 142 U. Penn. L.Rev. 1513 (1994); Leo Katz, Ill-Gotten Gains (University of Chicago Press, 1996); Claire Finkelstein, “Involuntary Crimes, Voluntarily Committed” in Criminal Law Theory: Doctrines of the General Part, Stephen Shute and A.P. Simister (eds) (Oxford University Press, 2002), 147; Claire Finkelstein and Leo Katz, “Contrived Defenses and Deterrent Threats: Two Facets of One Problem,” 5 Ohio Law Journal 479 (2008). The German language literature, however, abounds with in, depth studies of this problem. Here is a sampling—with a preference for overviews to guide the curious: Michael Hettinger, Die Actio Libera in Causa: Strafbarkeit wegen Begehungstat trotz Schuldunfaehigkeit? Eine historisch-dogmatische Untersuchung (Duncker & Humblot, 1988); Joachim Hruschka, Strafrecht nach logisch-analytischer Methode (de Gruyter, 1983); U. Neumann, Zurechnung und Vorverschulden (Duncker & Humblot, 1985); Dorothee Sydow, Die Actio Libera in Causa nach dem Rechssprechungwandel des Budgerichtshofs (Lang, 2002); Hubert Stuehler, Die Actio Libera in Causa de lege lata und de lege ferenda: Eine Analyse von Rechtsprechung und der Literatur verbunden mit einem Gesetzgebungsvorschlag (Ergon 1999); Rene Zenker, Actio Libera in Causa: Ein Paradoxon als oeffentlicher Strafanspruch in einem vom Schuldprinzip geprageten Rechststaat (LIT Verlag 2003).
(4) See, e.g., Robinson, supra n. 3, 4–8, and Moore, supra n. 3, 35–6.
(5) Alvaro Sandroni and Leo Katz, Why Law Breeds Cycles (2010), manuscript on file with authors; Vadim Cherepanov, Timothy Feddersen, and Alvaro Sandroni, Rationalization (2009) manuscript on file with authors.
(6) Philippa Foot, “The Problem of Abortion and the Doctrine of the Double Effect” in Virtues and Vices (University of California Press,1978); Judith Jarvis Thomson, “Killing, Letting Die, and the Trolley Problem,” 59 The Monist 204–17 (1976); Judith Jarvis Thomson, “The Trolley Problem,” 94 Yale Law Journal 1395–1415 (1985); Francis Myrna Kamm, “Harming Some to Save Others,” 57 Philosophical Studies 227–60 (1989); Morality, Mortality, Vol. 1: Death and Whom to Save From It (Oxford University Press, 1993); Morality, Mortality, Vol. 2: Rights, Duties, and Status (Oxford University Press, 1996).
(7) Michael Moore, “Torture and the Balance of Evils,” 23 Isr. L. Rev. 280, 323 (1989). For a more general exploration of this type of issue, see Claire Finkelstein, “A Contractarian Argument Against the Death Penalty,” 81 N.Y.U. L. Rev. 1283 (2006).
(8) Robert Nozick, Anarchy, State and Utopia (Perseus Book Group, 1974) 28–9
(9) Regina v. Dudley and Stephens  14 QBD 273 DC
(10) As in Nozick, supra n. 8, 28–9.
(11) Leo Katz, Why the Law is So Perverse (University of Chicago Press, 2011)