Jump to ContentJump to Main Navigation
Targeted KillingsLaw and Morality in an Asymmetrical World$

Claire Finkelstein, Jens David Ohlin, and Andrew Altman

Print publication date: 2012

Print ISBN-13: 9780199646470

Published to Oxford Scholarship Online: September 2012

DOI: 10.1093/acprof:oso/9780199646470.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 22 June 2018



Targeted Killings

Michael S. Moore

Oxford University Press

Abstract and Keywords

This chapter approaches the question of the morality of targeted killing by developing a general theoretical framework for answering questions about what morality permits, forbids, and requires. The framework consists of three levels. At the first level is the consequentialist principle that one ought to choose that action from among the alternatives that will bring about the greatest net good in the world. This principle is decisive in determining the moral status of an action across a wide range of cases, but it can be preempted by principles that belong to the second level. At that level are the deontological principles that specify ‘strong permissions’, which allow an agent to choose an action that does not have the optimal consequences, and ‘strict obligations’, which require the agent to perform an action that does not have the optimal consequences. The right of self-defence is a strong permission, and the prohibition on deliberately killing one innocent person in order to save, say, three other innocents is a strict obligation. Normally, such permissions and obligations are decisive when they apply to a case, but at the third level of the framework is a principle that overrides those permissions and obligations in cases in which catastrophic consequences would otherwise ensue. Using this framework, a decision tree is constructed that shows at what point in the deliberation process, and how consequentialist and deontological considerations are to figure in a decision about the moral permissibility of an action.

Keywords:   targeted killing, morality, consequentialist principle, deontological principles, self-defence, permissibility

I. The moral topic to be examined here

Reasonably reliable sources indicate that the military attaché to the British Embassy in Berlin reported to the British government in 1938 that the SS guard around Hitler had gotten sloppy in one aspect of Hitler's security: Hitler apparently was reviewing troops within easy gunshot range for any marksman who located himself in the British mission.1 The British officer, Colonel Noel Mason-MacFarlane, volunteered to assassinate Hitler personally, urging that doing so would forestall the calamity about to descend on the United Kingdom and the rest of Europe.Neville Chamberlain's government disapproved the assassination, reportedly with the comment that “it wouldn't be sporting.”2 My topic, roughly, is whether Chamberlain was right in not ordering an assassination of the leader of a country with whom the United Kingdom was not (yet) at war.

The echoes of Berlin in 1938 are, of course, all around us today. Israeli and American security forces regularly engage in “targeted killings” of individuals who intend to harm Israeli or American interests.3 Such assassinations and killings do not easily fit under the rubric ofwartime killings of enemy combatants, no more than would have the proposed assassination of Hitler in 1938. They thus raise the moral question that is the topic of my chapter. (p.435)

The topic that interests me here is purely a moral topic, as distinguished from related topics in law or in political philosophy. The topic is whether such targeted killings can ever be the morally right (permissible and perhaps even obligatory) thing to do. Such a moral question is distinct from: (a) the legal question of whether such killings violate present international or domestic law; (b) the question of political philosophy, whether such killings ought to be made illegal under international or domestic law, even if they are not already so; (c) another question of political philosophy, whether such killings ought to be publically condemned or even privately believed to be immoral/illegal by those who run our government or by those who advise (formally or informally from the academy) those who run our government. These other questions about the law we either have or ought to have, or about the conventional beliefs it would be good for our leaders to express or to hold, are important questions too, but they are different from the moral question that interests me.What is morally right can diverge from what is (or ought to be) legally right because of a number of features of law as an institution.4 For one thing, the law must necessarily speak in generalities, whereas morality can be as fine-grained in its resolution as the situation demands. For another thing, the law has certain formal requirements (such as clarity, prospectivity, and the like) needed for its notice-giving and administerability functions that have no analogues in morality. Also, the law is rightly concerned with the incentive effects of its rules, so its content may vary from what would be morally ideal in order to optimize desirable incentives for the actors to whom it is addressed.

What is morally right can also diverge considerably from the content of political philosophies when they are concerned with: appealing to an “overlapping consensus;” appealing to all “reasonable” points of view or to points of view that no one could “reasonably” reject; justifying beliefs even to those most disadvantaged by principles making up the contents of such beliefs; achieving neutrality or impartiality between competing conceptions of the good; etc.5 These contemporary approaches to political philosophy also no doubt have their interest for some audiences, but they explicitly prescind away from the moral question that interests me. They each require a suspension of one's own considered moral judgments (in the name of some political ideal such as peace or agreement) in ways that rule out appeal to substantive moral theory. My own view is that such “political but not metaphysical” approaches overvalue peace and undervalue moral truth, but that is not something needing resolution here. It is enough to distinguish such political questions from the moral question to be examined here. Doing so, like (p.436) distinguishing the legal questions too, brings into tighter focus the question of moral correctness about targeted killings that interests me.

The next clarification needed as a preliminary here has to do with the deontic nature of the question asked. The question is deontic in the sense that it asks what we should do in these situations. That is distinct from the aretaic question of how one should feel about what it is we should do. It has often and correctly been pointed out that difficult moral choices are often “no-win” situations for those who make them, in the sense that no matter which way such actors choose, they feel regretful, ashamed, and guilty.6 Moreover, on some plausible theories of the virtues, people ought to feel these ways, even when they make the right choice.7 But notice that such aretaic judgments need not affect the deontic truth about actions: it can be right to assassinate the next Hitler, no matter how guilty that makes one feel and no matter how virtuous it might be to feel such guilt.

Likewise, one must put aside considerations about what it might be politic to say about what one should do in such situations. For there are surely contexts in which it is impolitic, harmful, or even immoral to say what is morally true. Perhaps the recent torture debate is illustrative. I have long believed that torture is sometimes permissible to prevent significant harm in certain circumstances.8 However, saying this to the wrong audiences can lead to impermissible abuses, Abu Ghraib being a case in point. So reasons to believe such moral truths as those about torture are not necessarily reasons to say publicly what one believes. Still, such (perhaps obligatory) restraint in expression does not touch on the truth of what is or is not expressed. And it is only the latter with which I am here concerned.

II. The generality of any plausible answer about the morality of targeted killings

What is needed to answer my moral query about targeted killings is something quite general. We need an ethical framework employable for many moral dilemmas besides that of targeted killings. For targeted killings scenarios are but an instance of what in (p.437) general are usually categorized as moral dilemmas.9 Such moral dilemmas are found whenever two or more quite plausible moral principles recommend diverging courses of action on particular occasions. Surely it is plausible, for example, to think that we are obligated not to murder people even when they say things that we do not like, as did Hitler. But it is also plausible that we should prevent the murders of millions of people, if we can do so. Such hard choices between conflicting moral principles are a familiar feature of our moral life. Well-known examples are to be found not only in contexts of assassinations and targeted killings, but also in questions of: triage in medical contexts, where medical resources are insufficient to save all;10 jettison on the high seas, where planks or boats are insufficient to save all who need them;11 rescue situations, where not all can be rescued; combat situations, where some must be sacrificed if others are to get through; starvation on the high seas or in remote regions, where the only available food is the flesh of others about to die anyway;12 mountaineering situations, where ropes supporting some must be cut if others are to survive;13 flame-outs and other crashing aircraft scenarios, where pilots must choose whether to redirect their aircraft from hospitals and schools to other, less populated locations; wartime choices of whether to warn civilian populations of impending attack where to do so would reveal one's penetration of the enemy's intelligence or communication systems;14 wartime decisions about deliberate redirection of enemy rockets from more to less crowded civilian centers;15 torturing terrorist suspects to reveal life-saving information;16 transporting such suspects to other countries for torture impermissible in one's own, when such tortures will reveal life-saving information;17 (p.438) choosing whether to separate Siamese twins who otherwise will both die, when such separation will surely kill one of such twins;18 choosing which child to save when one cannot save both;19 shooting down civilian airliners when they are targeted at large civilian populations or crucial governmental or economic functions;20 etc. The sad fact is that the world is full of such moral dilemmas, the general form of which is the same. To resolve any of them adequately demands recourse to something equally general, viz., something no less general than our overall deontic framework for all obligatory/permissible actions.

III. The possibility of finding objective answers to such questions of pure morality

There are a number of skepticisms about this enterprise that need to be defused for those otherwise inclined towards them. One stems from a general skepticism about morality itself. Common forms of such skepticism are:21 relativism, where there are moral truths but these are relative to the conventional moral beliefs prevalent in a given society; subjectivism, where moral truth is relativized to each individual's subjective moral beliefs; non-cognitivism, where moral expressions are interpreted as serving exclusively non-descriptive functions and thus lacking any truth value, being neither true nor false; and cognitivist error theories, according to which moral statements do not lack truth values but such statements are always false because what such statements say exists (such as justice, fairness, etc.) never in fact exists.

I have elsewhere urged that general moral skepticisms such as these are the disease of our age because they demotivate all moral enquiries, this one included. I have accordingly devoted considerable effort elsewhere to showing such skepticisms to be false.22 My general take on all such skepticisms is that they do not correspond to anyone's actual, lived moral experience. None of us experience moral choices as arbitrary and not mattering; none of us experience ourselves as incapable of moral error, either in our individual or in our collective judgments as a society. Recapitulating any detailed refutation of such skepticisms is a task beyond the present chapter, however. What can be done here is to look at more particular skepticisms about there being solutions to moral dilemmas that are not based in some more general skepticism about morality. (p.439)

One of these more particular skepticisms is to think that moral reasons are incommensurable, so that when they conflict (as they do in moral dilemmas) there is no weighing up to be done. Alternatively, one might think that conflicting moral reasons are often equally weighty and thus in equipoise in their recommended action. In either case the results for rational decision-making would be unhappy. There would be two possibilities: either we are not obligated either way, meaning we are free to do either action, or each obligation still applies to us with undiminished force, so that whatever we do, we act immorally.

The first of these possibilities is one interpretation of the old saying (of Cicero, Aquinas, Kant, Bacon, Holmes, and others) that “necessity knows no law.”23 On this view, such irresolvable conflict of moral reasons takes us “beyond morality” to a realm where we must do what we must do, for whatever mysterious reason as may move us.24 Morality, on this view, is only for simpler situations; when the going gets tough, it has nothing to say and we must navigate as best we can by our own lights.

The second of these possibilities is what Kant described as “inconceivable;”25 namely, we are damned if we do and damned if we don't, because all possible courses of action (or inaction) are forbidden to us. Kant was surely offering up a bit of hyperbole when he said that such a morality was inconceivable; yet conceivable or not, such a morality would surely be unfair. Morality, on such a view, would become a kind of joke on the human race, the sort of thing some not-so-benevolent Greek God would foist off on us.

These more particular skepticisms about moral dilemmas make morality either ridiculously non-constraining (the first possibility) or ridiculously unfair (the second) in its condemnations. It is implausible that conflict frees us of all constraints (the first possibility), and equally implausible that conflict guarantees us moral failure (the second).The morality most of us experience looks nothing like either of these possibilities. (p.440)

Two remaining skepticisms are different. They assume that morality may deliver up answers to the most difficult of moral dilemmas, answers that are both objectively true and epistemically accessible to us. Yet for one reason or another, morality's answers are not the ones we should use when we act to resolve such dilemmas.Something else, it is thought, “trumps” morality when it comes to deciding what to do.

The first skepticism here is that the law is what makes morality irrelevant. The idea is that law obligates obedience by its subjects, so that what the law condemns, morality cannot condone, and what the law condones, morality cannot condemn. This view surfaces, for example, in the arguments of those who regard treaty obligations as decisive of what is morally permissible in the international arena; it also surfaces in the arguments of those who excuse any who obey lawful orders in the military, no matter how outrageous the actions such orders require of their subjects.

Such a conclusive role for law in our practical reasonings would certainly render irrelevant enquires such as this, divorced as they are from the law. Yet law surely has no such conclusive normative force, no matter what may be its pretensions to the contrary. Immoral but lawful military orders are not conclusive of what those subject to them should do; treaty-based and other international legal obligations are not conclusive of what our political leaders should do in targeting terrorists for killing, torture, or anything else. It is debatable whether the law even prima facie obligates citizens or official obedience;26 however that debate is resolved, it surely is not plausible to cede law conclusive moral authority. The question of what it is morally right to do is open, no matter what the law may say on the subject.

A second thought (about what may “trump” morality) is more diffuse and difficult to pin down. Yet it surfaces in some form in all discussions of moral dilemmas between academics and those “in the field.” Put simply, it is that morality is too impractical, too idealistic, too unrealistic, to be used in actual practice. It is academic, in a pejorative sense of that term. Morality on this view is said to be “trumped” by common sense, or practical realities, or some such thing.

Yet what exactly is this thought? Can it really be the claim that there are reasons for action more important than moral reasons? That morality, like the law, is just one more consideration among others, and not always the most important one to boot? One might well so regard conventional morality; that is, the practices and beliefs of a people with whom one only partly agrees. But can any of us so regard his or her own best thoughts about what morality really requires? (p.441)

Again, we all experience some weakness of will. We all know that, like St Paul in Romans vii, we can know the good but do the bad. We may even excuse such akrasia in ourselves. But that is worlds apart from thinking it right to do the morally wrong thing because common sense, practical realities, or whatever, so dictate.

I conclude that the only sense to be made of this common enough skepticism is the epistemic sense. Surely the idea only amounts to this: abstract moral theory often glosses over important nuances of particular moral choices, and insofar as it does so, such abstract theory is to be corrected or even ignored. Yet what makes nuances important is nothing other than another moral theory—a better, because more nuanced, moral theory—but moral theory nonetheless. There is no higher standard for determining what we ought to do than a moral standard—recognizing that that has to be as nuanced, as context-sensitive, as responsive to real interests, as moral correctness demands.

IV. The desirability of finding objective answers to such questions of pure morality

To say that a thing is possible to do is not, of course, to say that such a thing is desirable to do. I have yet to answer why anyone should care about my moral question as I have defined and clarified it. Motivational questions are usefully divided up by those who are asking them. Consider first those actors within the American and Israeli intelligence community who are now making, will make, or have made, such decisions about targeted killings. For those presently making such decisions, or those who will do so in the future, what could be more relevant than assessing the morality of their prospective actions? Once one puts aside the skepticisms just mentioned, such actors have to ask and answer my moral question in order to decide rationally what to do.

Even for actors who made such decisions in the past but are no longer making them (such as John Rizzo of the CIA),27 there is reason to ask and answer my question of moral assessment. Like all of us, such actors should want to be “in session with themselves”28 to assess whether they acted in a morally proper way or not. Even when no corrective actions are possible, each of us should want to know how it stands with our moral ledger. We want to ask Harry Truman's kind of retrospective question (about his decision to bomb Hiroshima and Nagasaki),29 Earl Warren's (p.442) kind of question (about his decisions regarding the Japanese interments during the Second World War),30 the Shin Bet's kind of question about whether they did the right thing in beating two terrorists to death after they had taken the 300 Bus in the Sanai in the 1980s.31 We all want to assess how we have done in this life, and our resolutions of moral dilemmas (where there was something to be said for the choice we did not make) can be a critical part of this assessment.

Most of those reading this chapter will not, of course, be persons who have ordered or themselves performed targeted killings, nor will they do so in the future. The morality of targeted killings thus does not affect their moral ledgers directly one way or the other. Yet we all should be concerned about the morality of our government's actions when that government acts in our name. One of Abraham Lincoln's favorite quotations was of John Winthrop, early governor of the Massachusetts Bay Colony. Winthrop famously told those early Americans that they should create a “new City upon a Hill for the eyes of the world were upon their efforts.”32 Few governmental actions speak for us as loudly as those of our military and intelligence services. The stakes are high, the temptations severe, the measure of who we are clearly visible for all to see. If our military or intelligence communities behave badly, they do so in our name and we are ashamed; whereas if they behave well in such circumstances, they make us proud. We all thus have an interest and a stake in the morality of governmental actions such as targeted killings.

For those in positions of power—legal policymakers—the interest in this question is even more direct. I earlier distinguished my moral question from the legal question of what current law provides. I also separated the moral question from the political philosophy question of what the law ought to provide. Yet even while reaffirming these distinctions, we should see that the moral question is not irrelevant to the question of what the law does or should provide in regards to targeted killings. Criminal law, for example, largely punishes only when it can blame, and on any plausible theory of punishment, criminal law should only punish those who are morally blameworthy.33 Being morally blameworthy, in turn, requires one to have done something morally wrong, either in fact or at least in one's own mind. Whether an act (such as targeted killings) is morally wrong thus enters into both the criminal law we have and the criminal law we ought to have, even though an act's moral status is not by itself determinative of that act's legal status. (p.443)

V. Two straightforward but unsatisfactory views of when any action is morally right

There are two attractively simple views about what morality demands of us and what it permits us to do. One is usually termed a consequentialist (or “agent-neutral”) view,34 of which the most familiar variant is utilitarianism. A consequentialist believes two things: first, that the basic kinds of things that are valuable are states of affairs, second, that right action consists in maximizing good states of affairs, and minimizing bad states of affairs.

Consequentialism thus consists of two theories. The first is a theory of the (intrinsic) good. Such a theory marks out some state(s) of affairs as intrinsically good, others (perhaps but not necessarily derivatively) as intrinsically bad. All other states of affairs can be judged instrumentally good or bad, depending on how they contribute causally to what is intrinsically good or bad.

Utilitarianism usefully illustrates this first aspect of consequentialism.35 Utilitarians believe that only one thing is intrinsically good: human welfare. Exactly what such welfare consists of varies within utilitarian theory: Bentham equated human welfare with pleasure, John Stuart Mill with the broader notion of happiness, the post-Paretian economic utilitarians with preference-satisfaction (which may or may not give pleasure or make one happy). Normative (or “objective”) utilitarianism equates human welfare with identifiable human needs, rather than desires. But however fleshed out, the general idea is that human welfare is intrinsically good and all other states of affairs are instrumentally good insofar as they contribute to such human welfare, bad insofar as they detract from it.

The second theory needed by consequentialists is a theory of right action. On the plausible supposition that if something is good, more of that thing is better, the consequentialist theory of right action is a maximizing theory: right action maximizes intrinsically good states of affairs and minimizes intrinsically bad states of affairs. For a utilitarian-consequentialist, thus, actions are morally right when but only when they maximize human welfare. Each person's welfare counts equally in doing such utilitarian calculations; that is, your happiness gets no extra weight when it is you doing the calculation, just because that happiness is yours. Utilitarianism is not a kind of rational egoism, in other words.

The radically different but equally simple view of the deontic part of ethics is usually termed the deontological (or “agent-relative,” or “non-consequentialist”) (p.444) view.36 Being rational and moral on this view does not consist in doing those actions that maximize good states of affairs and minimize bad ones; states of affairs are not the locus of value at all, on deontological views. Rather, it is actions themselves that are either intrinsically right or intrinsically wrong to do. Right action consists in conforming to moral norms that either permit or require that certain kinds of actions either be done or not be done. Such norms do not direct agents to maximize the satisfaction of such norms; rather, they are personally directed (“agent-relative”) norms categorically telling each agent either to do, or not to do, certain actions. They say to each moral agent, “Don't you do some action A” (such as torture), not even when A-ing on this occasion would minimize A-ing overall in the long run.

Two examples may help make plain the distinction between consequentialist and deontological views of morally right action. The first comes from the waning days of the Soviet Union. Middle Eastern terrorists had captured four Soviet diplomats. They then made certain demands on the Soviet Union accompanied by threats to kill the diplomats if the demands were not met.37 The terrorists then killed one of the diplomats to show they were serious in their demands. The KGB knew who the terrorists were but were powerless to attack them directly. So the KGB killed an innocent relative of one of the terrorists, a relative who had had nothing to do with the kidnapping or killing of the Soviet diplomats, and let it be known that they (the KGB) would kill every relative of the terrorists in the Middle East if the remaining three Soviet diplomats were not released unharmed. Believing the KGB, the terrorists released the three Soviet diplomats, and no further kidnappings of Soviet officials again took place in the Middle East.

On the plausible view that the killing of innocents is bad, a consequentialist should approve of the KGB's action of killing the one innocent relative of the terrorists. The KGB caused one bad state of affairs, to be sure, but that same act prevented three other equally bad states of affairs from taking place, namely, there were no killings of the three remaining Soviet diplomats. On the consequentialist view, it does not matter that it was the KGB that did some killing, or that the killings prevented would not have been by the KGB—for consequentialism is “agent-neutral” in the sense that the identity of the agent does not matter. What matters is that three bad states of affairs were prevented by bringing into existence one bad state of affairs; the world is better off by a net gain of two bad states of affairs that do not exist.

A deontologist sees the case much differently. The relevant moral norm says, “Don't you kill an innocent.” That other killings (by you or by others in the future) will be prevented (p.445) by the killing of an innocent here and now, is neither here nor there for a deontologist. The norm is personal and absolute: no matter what, “Thou shalt not kill.”

I customarily raise this KGB example in my criminal law classes to illustrate the difference between consequentialist and deontological ethics. One year at the University of Virginia a burly male student at the back of my class interrupted my presentation by proclaiming that he was a former KGB Colonel who regularly taught this case in his teaching of KGB officers in Moscow. I asked him how he taught the case; his reply was pure consequentialism: “big success—one dead Arab, innocent to be sure, but three not-dead Soviet diplomats who were equally innocent.” To test his deontological intuitions I asked, “But what about human rights?” His reply was again pure consequentialism, but of a non-utilitarian kind: “Rights? You want to talk about rights? One right violated in order to prevent three equally important rights from being violated. Net gain, two rights not violated.” I concluded with a reminder of what many find compelling about deontology: “Amnesty International must have given you guys their annual award for human rights protection. Or did you miss something essential about rights (and their correlative duties): that they categorically are not to be violated, even in the name of minimizing rights-violations?”

The second example is self-defense. Except for a few Quakers and others, everyone concedes that each of us does no wrong in killing an attacker when necessary to save our own life. Most people also agree that it is permissible to kill an attacker who threatens some evil less than our own death, an evil such as rape, other forms of sexual assault, or kidnapping; most also think that we may kill as many attackers as it takes to save our own (one) life.38 These moral conclusions should be problematic for a consequentialist. For in these cases we seemingly cause a worse state of affairs to exist in order to prevent a less bad state of affairs from existing. This remains true even if we discount the lives of aggressors; at some point the harm threatened by them should become small enough, or the harm done to a large enough number of them great enough, that any consequentialist-based permission to kill them evaporates. The deontological intuition is that our right of self-defense never evaporates in these ways, that for example, we can kill without limit all who are trying to kill us.

There are deep problems for each of these two simple views of morality. Among other things, consequentialism seems strikingly out-of-tune with moral intuitions that are very difficult to give up. Paradoxically, consequentialism's illness of fit has two opposed dimensions: consequentialism is both too demanding of us, and it does not demand enough of us. It is too demanding of us when it requires us to lay down our own life, for example, when our only way to save ourselves is to kill many culpable agressors who are out to kill us. It is not demanding enough, for example, when it permits us to kill one innocent in order to deter the killings of three other innocents by others. (p.446)

Likewise, the simple deontological view I have sketched has deep problems, of which I shall mention three. (Most of these three have to do with such a simple deontological view being too demanding of us.) The first has to do with the content of the absolute (or categorical) norms of deontological morality. “Thou shalt not kill,” for example, is only a start at specifying when morality forbids us to kill. If one has only one stone tablet with which to work, fair enough, this is an important start. But in a modern criminal code with between 7,000 and 15,000 prohibitions, surely we can (and must) more completely specify the content of the norm against killing. “Thou shalt not kill, unless in self-defense, in defense of others, in effecting a lawful arrest of a fleeing and dangerous felon, in fighting a just war . . . etc.,” comes closer to what we intuitively think can rightly be demanded of us.

The second problem has to do with the force—the “absoluteness” or “categorical” nature—of deontological moral norms. Kant, a well-known proponent of deontological ethics, once proclaimed that it was better that the Heavens fall and the human race should perish than that an injustice be done.39 This rightly strikes many people as absurd, a bit of hyperbole that cannot literally be true. If the only way to find and defuse a nuclear bomb planted by a terrorist somewhere in New York City is by torturing to death his child in front of him I, like many others, think you torture the child to death if necessary. Allowing the catastrophe of millions of deaths to occur because one innocent's rights must be violated gives too strong a force to those rights (and correlative duties) to be plausible.

The third problem for a simple deontological view of ethics lies in the potential for conflict between our absolute obligations. Kant perceived the problem, and proclaimed that a conflict of such obligations was “inconceivable.”40 Like the protagonist in the film, The Princess Bride, I do not think Kant means what we mean by “inconceivable.” For surely such conflict of obligations is not only conceivable; it is seemingly quite possible. If I am categorically obligated to save you, and equally obligated to save another, and both you and she are in peril and I cannot save both, I seem to be in a no-win situation: whatever I do, I do wrong. This is not inconceivable; but it is surely unfair. If a view of morality gives us no chance of moral success because of such conflict of obligations, that is a reason to reject such a view.

These problems of under-and-over demandingness keep me from accepting either of these simple views of ethics, a pure consequentialism or a pure deontology. Yet I raise these simple theories not just to reject them, but to learn from them. They each have their intuitive tugs, do they not? The trick is to combine them into an overall view of ethics that accommodates our intuitions better than can either simple view. (p.447)

VI. The three-level analysis of ethics

My own long-held view of the deontic part of ethics is best unpacked in terms of three levels of analysis.41 The basic picture is of an omnipresent consequentialism, one that is nonetheless subject to two kinds of deontological restrictions, those of strong permissions and of stringent obligations. These restrictions themselves are subject to an override in situations of moral catastrophe. Let me examine this picture, one level at a time.

Most of what we do is governed by consequentialist reasons. I sometimes put this as a slogan: “Man cannot live by deontology alone.” Rational decision-making for most of our daily decisions is rational and moral, not because the actions decided upon instantiate categorical norms of permission or obligation, but rather because such actions produce good states of affairs or prevent bad states of affairs. Take the decision as to when a law school should begin its classes for the fall semester. Plausible reasons for choosing one date rather than another are considerations of faculty convenience, maximizing student summer employment opportunities, coordinating with the start dates of other units on campuses, etc. Implausible would be an argument of deontology, such as one I heard many years ago: “Whatever the balance of consequences in favor of moving the start date, it cannot be moved because faculty coming to the school have the right to the start date in place when they came.”

In general, it is extremely plausible, is it not, that the right thing to do is the action that produces the best consequences, on the whole? Whatever one's theory of the good, surely it is usually rational and moral to choose actions productive of more of that good rather than less. We thus operate in an omnipresent, background sea of consequentialist reasons. Such reasons always bear on our actions, even in cases where they lose out because of deontological considerations. The omnipresence of such consequentialist reasons is what makes the kinds of life-and-death decisions earlier mentioned so hard and so problematic. It is this omnipresence of consequentialism that gives rise to moral dilemmas of the kind I want here to examine.

The force of consequentialist reasons can become so compelling that such reasons can constitute moral obligations. These are non-deontological obligations, to be sure, but genuine obligations nonetheless. We each have reason, for example, to rescue strangers from perils when we can do so at little risk or inconvenience to (p.448) ourselves. Unlike the libertarians, I take this to be obligatory: we do wrong when we omit to help others in such circumstances. Yet the wrong we do is a consequentialist-based wrong, not one based in deontology. We can see this by seeing the ease with which we can justify apparent violations of this duty by good consequentialist reasons: I can easily justify omitting to save one if I instead save two, in situations where I cannot save all. Being based on good consequences, our duty to rescue evaporates when those consequences are not, on balance, as good as they could be. Whereas if our duty of rescue is deontological (as I think it is for our children, for example), such good consequences (as saving two rather than one) would not justify the omitting to save one.

This gets us to my second level of analysis: deontological suspensions of consequentialist reasons. There are two kinds of such suspensions: deontological permissions and deontological obligations. Sometimes we are permitted to do actions (such as self-defense) even when they are not productive of the best consequences on the whole; sometimes we are obligated not to do actions (such as the KGB's action of killing an innocent Arab) even though such actions would be productive of the best consequences on the whole. The right way to see these restrictions is as a kind of overlay, or “side-constraint,” on consequentialist reasons. In such cases our deontological reasons do not outweigh our consequentialist reasons; rather, our deontological permissions and obligations preempt the consequentialist reasons from having their normal sway in determining what is rational and moral to do.

It is a long-standing puzzle in ethics just why or how consequentialist rationality can be preempted in these two ways.42 I have no new solution to that puzzle to lay before you here. Yet it is very intuitive, is it not, that we are permitted to lead our lives in ways not always productive of the best consequences? How could one justify, for example, a good dinner out on the town if one was required to maximize good consequences (as by giving the money for the dinner to those more in need of basic sustenance)? How could one justify, as another example, refusing to have bone marrow extraction if that was the only way to save another (whose DNA uniquely matches your own) from death? A pure consequentialism never preempted by deontological permissions is a saintly mode of being that is not required (and, to my way of thinking, is not always even virtuous). Likewise, the KGB's killing of one innocent to save three would be justified by true consequentialists; yet it (and thousands of examples like it) flies in the face of what most of us think is moral to do. I cannot take your property, your spouse, your children, your bone marrow, or your life, just because doing so will make them and others better off more than doing so makes you worse off. Our rights-based duties have a stickiness to them that rejects as irrelevant consequentialist justifications such as these. (p.449)

We have then, deontological permissions (such as the right to defend ourselves against wrongful aggressors), and deontological obligations (such as the duty not to torture or kill) that exist even though doing or omitting to do these actions is not productive of net good consequences. Such permissions and obligations operate as overlays on omnipresent consequentialist reasons, overlays that preempt such reasons from having their normal determinative effect on rational choice.

Merely combining consequentialism and deontology in this way so far only alleviates the problems besetting a pure consequentialism. (It does this by making such a side-constrained consequentialism neither too lenient nor too demanding.) It does not alleviate the three problems of deontology I earlier alluded to, problems largely having to do with the seeming over-demandingness of deontology. So I need to say more about how this three-level analysis of ethics is not subject to these three problems.

The first problem was that of content. “Thou shalt not kill,” while short and snappy and nicely fitting onto one stone tablet, misses all the situations where killing seems permissible to almost all of us. The existence of deontological permissions can help here, for they can supply the content to the exceptions deontological norms of obligation need in order to be plausible. If our deontological permission to defend ourselves, for example, supplies an exception within the content of the deontological obligation not to kill, the full statement of the obligations is not, “Don't kill,” but rather, “Don't kill except in self-defense, except in . . . ,” etc.

Not all exceptions to deontological norms of obligation need be deontological permissions. The exceptions for killings necessary to protect other people who are strangers, and killings necessary to stop fleeing felons, for example, might well not be based on deontological permissions. If so, such exceptions only remove such killings from the content of a deontological obligation; they do not by themselves make such killings right. Such killings are still subject to our consequentialist reasons, and become right only when productive of net good consequences.

The distinction between exceptions based on deontological permissions and those that are not, illustrates an important fact about permissions. There are two kinds of permissions: those that make an action right irrespective of that action producing net good consequences (deontological permissions); and those that make an action right only if that action produces net good consequences. I call the first of these, the deontological ones, strong permissions, and the second weak permissions.43 Weak permissions are simply the absence of deontological obligations; such absence returns one to the “background, omnipresent sea” of the unrestricted consequentialist calculus determinative of right action.

These two kinds of exceptions, based on either strong or weak permissions, are needed to make plausible a not-too-demanding deontology. The simple absolutes (p.450) we teach our children can only be heuristics for the more complicated norms of deontological obligation, completed as these norms must be by numerous exception clauses. As long as such exception clauses do not include things like, “unless productive of net good consequences on the whole,” I see nothing problematic for a more sophisticated deontology that regards such exceptions as an integral part of the content of its norms of obligation.

The distinction between weak and strong permissions just introduced in my discussion of the first problem of deontology will also help with the third problem.The third problem was that conflict between deontological obligations, while not inconceivable, was unfair and therefore problematic for the truth of any moral theory generating much of such conflict. Once one sees the existence of weak permissions by virtue of some exceptions in the content of norms of obligation, one should also see the possibility of there being weak permissions of a more general sort. And the generality of these latter weak permissions might help alleviate the potential for conflict between the deontological norms of obligation.

To explain, what makes conflict between deontological obligations widespread and inevitable would be the broad scope of such obligations. If they apply generally to omissions to save as well as actions that kill, known riskings of death as well as intending of deaths, causings of death as well as allowings of death, etc., then conflicts of obligation will be endemic. If the KGB, for example, was deontologically obligated not to omit to save the three Soviet diplomats as much as it was so obligated not to kill the one innocent Arab, then its choice was inevitably a conflict-ridden one. Whereas if its obligation to save its diplomats was only a consequential one, whereas if its obligation not to kill the Arab was deontological, there is no conflict of deontological obligations.

What might these conflict-reducing scope distinctions look like, and from whence do they spring? I have elsewhere charted in some detail six main distinctions here.44 They are: (1) the distinction between intending and foreseeing. I may consequentially justify causing a harm that is merely foreseen when I could not consequentially justify causing that harm when I intend to do so. The most famous example of this medieval distinction is the killing of noncombatants in a just war: I can justify killing such noncombatants by the justness of the war only if I foresee that my bombing of strategic targets will surely kill them; I cannot justify the killing of noncombatants by the justness of the war if I intend their deaths as a means to winning the war (say, by dispiriting the enemy population so that they sue for peace). (2) The distinction between action and omission. I can consequentially justify failing to prevent some evil, whereas I cannot consequentially justify acting so as to cause such evil to come about. Thus, I can omit to save one drowning person when (p.451) I use the only rope I have to save two; but I may not save the two by drowning one who is saving himself with the only available rope. A famous historical example is Churchill's decision not to alert the citizens of Coventry about the bombing of their city intended by the Germans during the Second World War.45 Saving them by alerting them would have tipped off the Germans that MI-6 had penetrated the German code system, thus lengthening the war and even putting its outcome in jeopardy. Imagine if Churchill had tried to keep secret the MI-6 success in breaching German communications in a more active way: he had all the civilian population of Coventry killed, because an unidentified one of their number was about to tell the Germans of MI-6’s success. Surely Churchill could justify not saving the citizens of Coventry by good consequences in a way that he could not justify acting so as to cause the deaths of those same citizens. (3) The distinction between doing and allowing. I can consequentially justify allowing someone to die, whereas I cannot consequentially justify killing that same person. This medieval distinction is often illustrated by the difference between active and passive euthanasia: if medical equipment is needed to save others, I can remove it from the patient for whom it is doing less good, even though I know that he will die without the equipment; whereas I cannot actively kill that same patient in order to free up the equipment that can then be used to save others. In the first case I allow nature to take its course by removing a defense to that course that I earlier provided, and thus prevent something from preventing death (which is why these cases are commonly called cases of “double prevention”); whereas in the second case I cause death (by an injection of poison, say). Morally the first case is easier to justify by good consequences than is the second. (4) The distinction between causing and aiding. I can consequentially justify aiding another to cause some harm even when I cannot consequentially justify causing that harm myself. The British duress cases illustrate this distinction.Under British law, one cannot justify killing an innocent because of the duress of threats by another. But (for a time, at least) one could justify aiding another to kill an innocent, if such aid were the only way to avert threatened harm to oneself or one's family. Thus, in one well-known case, the defendant could justify driving the IRA killers to where they could kill a British policeman by the threats of the IRA otherwise to kill the defendant's family; but averting such threats could not justify the defendant in actually killing the policeman himself.46 A more contemporary use of this distinction was by the CIA in its program of “extraordinary renditions”: one could justify flying terrorist suspects to other countries such as Egypt for torture there by those countries’ police, by the good consequences of extracting life-saving information,47 whereas one could not justify torturing those suspects oneself by those same good consequences. (5) The distinction between initiating a (p.452) threat and redirecting a threat initiated by others. One can consequentially justify redirecting a threat initiated by others when one cannot consequentially justify initiating such a threat to others oneself. Philosophers have long illustrated this distinction by the example of a switchman who must choose whether to turn a runaway trolley onto one track or another, when each track has some trapped workman on it who will be killed by the trolley.48 A more real-life example is provided by another of Churchill's wartime decisions, this one having to do with the V-1, V-2 rocket barrage of London.49 The German targeting of their rockets depended on spies calling in the hits; MI-6 had penetrated the German spy network, so by deliberate mis-information Churchill's agents were able to redirect the German rockets from more populated areas of London to less populated suburbs. The rockets were going to kill some; Churchill's redirection decision caused them to kill less rather than more. The American Model Penal Code provides yet another example of this distinction in its licensing of the blowing of a dike and thereby diverting a flood about to inundate an entire village, even when the diverted flood will kill a farmer and his family. (6) The distinction between causing a harm not about to occur anyway, and merely accelerating a harm about to happen without one's contribution. One can consequentially justify accelerating an inevitable harm even when one could not justify causing such a harm when it was not inevitable that it would occur. Thus, when all mountaineers on a rope are about to die, some may cut the rope saving themselves, even though this means the others will die a little sooner than they otherwise would have.50 When two Siamese twins are attached in such a way that both will die within the space of a year or two, doctors may sever the two, saving one, even at the cost of accelerating the death of the other.51 When a Nazi officer cruelly threatens a mother with the death of both of her children unless she chooses one over the other, she may choose to save her son (as in the film) or the daughter, even though that condemns the other child to death.52 And in cases of shipwrecks, when all will die unless one is pushed off the plank that's insufficient (p.453) to support all,53 or some are thrown from the sinking lifeboat,54 or some are eaten to sustain the rest55—in such cases, such actions can be justified by the net good consequences of saving some.

Intuitive as these distinctions are to many people, they cry out for some explanation as to why they mark the line of permissible consequentialist justification. Why are we weakly permitted to allow or omit to prevent harms we cannot cause, foresee harms we cannot intend, aid others to cause harms we cannot ourselves cause, redirect or accelerate harms we cannot ourselves initiate? Why is the scope of our deontological obligations limited in these ways, rather than other, equally conflict-reducing ways?

The answer lies in the general shape of our deontological obligations. Not only do particular obligations have exceptions to them specific to such obligations—see above—but there is a general shape to the content of our deontological obligations that limits their scope and thus their potential for conflict. That shape is dictated by the four major determiners of desert that exist in morality: act, cause, intent, and counterfactual dependence. We are most responsible, most blameworthy, most deserving of punishment, for our acts 56 (rather than omissions) that cause 57 (rather than fail to prevent, allow, redirect or aid) harms to others, when those harms are intended 58 (rather than foreseen or risked) and when those harms counterfactually depended on our acts59 (rather than being such as were going to happen anyway). It is these four desert-determiners that dictate the general shape of our most stringent duties, which are our deontological norms of obligation.

To say that we are weakly permitted to fail to prevent or allow harms to others, to aid others or nature to cause harms, to foresee or risk harms, or to cause harms about to occur anyway is not, of course, to say that we are (bottom line) right or justified to do any of these things. Such weak permissions betoken no more than an absence of deontological obligation (because we are outside the scope of such obligations). That means that we are back in the land where consequentialism holds full sway. Such (p.454) weakly permitted actions or omissions will be right and justified only if such actions or omissions produce the best consequences overall. Weak permissions, in other words, do not permit actions as such; rather, they permit consequentialist reasons to do their justifying work, if any such justifying consequentialist reasons exist.

The existence of these general kinds of weak permissions, which are part and parcel of the general shape of the deontological norms of obligation, reduces but does not eliminate the potential for conflict between such obligations.60 It thus reduces the force of the third objection to deontology considered above. Yet we so far have said nothing about the second objection to deontology's seeming over-and under-demandingness. This was the moral catastrophe objection. It seems very counterintuitive that the Heavens should fall or that the human race should perish, rather than a deontological obligation be violated.

Enter the third layer of the analysis. I have urged that consequentialist reasons are omnipresent in the determination of rational and moral action (the first layer); further, that the force of those consequentialist reasons can be suspended by either strong permissions or deontological obligations (the second layer). Now see that those deontological overlays can themselves be overridden by consequentialist reasons whose force exceeds some threshold of moral weight. This so-called “threshold-deontology” is the subject of the third layer of analysis.

Threshold deontology answers the second objection to deontology directly; such a threshold deontology denies that we are obligated not to torture an innocent in situations where such torture will save the planet, save a city, preserve a culture, etc. It also denies that we may exercise any of our permissions (such as our strong permission to defend ourselves) if such action will have these catastrophic consequences. Threshold deontology restricts the operation of strong permissions and deontological obligations to the normal world most of us face throughout our daily lives, a world where no such huge catastrophes loom before us.

Threshold deontology, popular though it is amongst philosophers who are deontologists in their ethics, is largely supported by intuition alone. Deontology becomes too counterintuitive without interposing some such threshold for overriding deontological permissions and obligations. Some find this kind of intuitive support disquieting (although the older one gets, the shorter the argument chains one finds acceptable). Others find the lumpiness of threshold deontology queer: consequentialists reasons do not count, and then all of a sudden they do if they are weighty enough. Others find the vagueness of where any such threshold might be located troubling, accompanied by a disquieting sense of arbitrariness if a precise threshold is specified.61 (p.455)

I do not find any of these worries sufficiently worrisome as to abandon threshold deontology. Indeed, if anything were to be abandoned in the face of moral catastrophe kinds of examples, it would be deontology itself. Deontology without thresholds, á la Kant, would be too implausible to be accepted.

This third level of analysis provided by threshold deontology can be seen as a kind of skewed consequentialism. This characterization will be accurate if one adopts what I call “sliding scale threshold deontology.” What slides on this view is the threshold: more serious moral wrongs require a higher threshold of bad consequences averted than do less serious moral wrongs. This is not the normal consequential balance, but a skewed one whereby the doing of a single moral wrong by the agent doing the calculation is equated (at the threshold) to a large measure of bad states of affairs not due to one's agency that would be averted by the doing of that wrong.

The three-level analysis just sketched invites the following sorts of questions about practices such as that of targeted killings. One can think of these seven questions, in the order presented below, as a kind of ordered decision procedure, one amongst others possible that yield a rational way to proceed. First question: is there a strong permission to engage in the action (say, a particular targeted killing)? If the answer is yes, that requires that one ask a second question: is this strong permission nonetheless overridden by the catastrophic consequences of its being exercised? If this answer is no, then one is justified in doing the action without asking further questions; if this answer is yes, then one is not justified in doing the action without asking further questions. If the answer to the first question is no—that there is no strong permission to engage in the action—then ask a third question, the question of general consequentialist balancing: will the action produce more net good consequences than its alternative(s)? If the answer to this third question is no, then one is not justified in doing the action without asking any further questions. But if the answer to this general consequentialist question is yes, then one must ask a fourth question: is the action prohibited by a deontological obligation, at least prima facie? If the answer is no—there is no deontological prohibition—then one is justified (given that the action is productive of net good consequences) in doing the action without asking further questions. But if the answer is yes, that the action prima facie instantiates a deontologically prohibited act-type, then ask a fifth question: is there nonetheless a weak permission to do such action embedded as an exception within the content of the deontological norm of obligation? If there is such a weak permission, and since the action also is productive of net good consequences, then one is justified in taking the action without asking further questions. But if the answer is no, then ask a sixth question: is there nonetheless a weak permission to do the action because the general nature of the action (or omission) falls outside the scope of all deontological prohibitions, this one included? Is it a merely foreseen or risked action rather than an intended act? Is it not an act at all, but an omission? Is it a mere allowing or an aiding or redirecting or accelerating? If the answer to this (p.456) sixth question is yes, then one is justified (given that the action is productive of net good consequences) in doing the action without asking further questions. But if the answer here is no—that there is no weak permission to do an act deontologically prohibited to one—then ask a seventh and final question: is the act nonetheless necessary to prevent some moral catastrophe that is over the threshold of one's threshold deontology? If the answer to this last question is no, then one may not do the action; if the answer is yes, then one may (and indeed, must) do the action.

Since this is a complicated decision procedure, I have schematized it on the following decision tree:

Admittedly this decision procedure has a bit of a mechanical feel to it, something akin to painting-by-numbers. But this is because it is algorithmic and ploddingly methodical in making plain the connections between the different questions, and because it is complete. Much of what we know here is intuitive, so that (like much of our inference-drawing in many areas) in practice we can be elliptical.In cases of horrendous consequences looming on one side or the other (nodes (2) and (7) above), for example, these will in practice be so salient that we may well focus on them first or even exclusively, determinative as they are of our decisions if they exist. The same may be said for the existence and application of a deontological norm of prohibition (nodes (4), (5) and (6)): the application of such a norm may be sufficiently clear, and the absence of strong permissions or threshold overrides so obvious, that we focus our decision on this alone.

I thus do not pretend that the decision procedure schematized above will always or even often be the actual psychological process a rational actor should go through explicitly and consciously. Still, the schema organizes all of the questions potentially (p.457) relevant in assessing the morality of a given action or practice, and is useful for that reason. That is how I intend to use it in the discussion of targeted killing that follows.

VII. The three-level analysis applied to targeted killings

I shall here sketch the relevant variables to assess the morality of targeted killings. I make no attempt to be definitive; my purpose is to illustrate how one should decide this question more than it is to decide it. I shall proceed using the seven-question decision procedure outlined in the last section.

First question: is there a strong permission to kill terrorists about to engage in violence against us? One obvious strong permission to invoke here as that of self-defense, as a number of papers prepared for the Penn conference anticipated.62 I am not concerned with the technical legal notions of self-defense of either domestic criminal law or the international law of war. Sticking to the moral question relieves us of any need to abide by these legal doctrines (such as the imminence requirement) except insofar as they capture or reflect some underlying moral truth.

In morality the privilege to defend oneself is but an instance of a more general principle.63 When someone attacks us meaning serious harm, and when the only way to prevent such harm is by our own use of harm-causing force, the attacker has made it inevitable that someone be harmed, either himself or his intended victim. The general moral principle behind self-defense (and defense of others) is that he who culpably causes the necessity of some harm being visited on somebody, is himself the right person to be made to suffer such harm. And this is true, without regard to whether the harm inflicted on the attacker is greater or lesser than the harm he would inflict on his intended victims.

Such a principle has support in rather robust intuitions of fairness and appropriateness. When someone has tried to harm another, and yet the harm he has put in motion comes back to him rather than his intended victims, we often recognize such outcomes as peculiarly fitting as a kind of “divine justice.” Think of the would-be murderer's bullet that kills him instead of his intended victim. Such sense of fittingness carries over to situations where we must initiate a new force, rather than (p.458) redirect the wrongdoer's old one, as in Philippa Foot's well-known example:64 a fat man has wedged himself, face outward, in the only opening to a cave filled with people and rising water. He had done this to drown all inside.Surely we may use dynamite to blow him out of the cave's entrance, since it was his act that made it necessary that either he or the group die. (Notice this same act of blowing open the cave seems permissible if there's only one who is trapped, and a large group blocking the entrance in their attempt to drown the one.)

As applied to targeted killings, the strong permission (to turn the hard end of a hard choice on he who culpably caused the necessity of such choice) would seem to apply to killings of those terrorists who otherwise would visit harm on us. The only interesting questions here are factual, not moral: how certain is it that the terrorist targeted will in fact do serious harm if not assassinated? Bomb-makers, suicide bomb-carriers, those who plan such things, seem easy examples of justified targeting, even if there will also be harder cases of less certainty.

More difficult moral questions are faced when the threat emanates from an innocent source. In domestic criminal law we call these “innocent aggressor” cases.65 The child who does not know what a gun is but who will shoot you with it unless you shoot her first, the insane (and thus not blameworthy) attacker,66 the coerced or duped assassin, the man whose body will fall on you unless you disintegrate it,67 etc., are innocent aggressors. In the international arena, think of bin Laden's amused comment on video that some of the 9/11 hijackers were ignorant of what was in store for them; or think of those couriers ignorant of the deadly peril in what they carry. Can these too be killed, if necessary to prevent the harms they threaten, however innocently? I think the standard answer—and the right answer—is “yes” domestically, and it should be “yes” internationally. Whether these are justified killings, or only excused ones, is a close question, but here too I think morality makes such killings right and not merely wrong but excusable.

A second deontological permission possible here has to do with retributive punishment. According to retributivists (in which I include myself),68 we are obligated to punish those culpable wrongdoers who deserve it. Such categorical obligation to punish carries with it an equally categorical permission: we are permitted to punish the guilty, even when doing so is not conducive to achieving the best consequences on the whole. Such obligation-cum-permission will justify targeted killings (p.459) whenever those killings are of terrorists whose past acts merit a sentence of death.The justification here is backward-looking and punitive, not forward-looking and preventative, as in self-defense.69

Examples here that come to mind include the 2011 U.S. Navy SEAL killing of Osama bin Laden, as well as the decision of Golda Meir to kill the terrorists who had killed Israeli athletes at the 1972 Olympics in Munich. To my mind, these were unproblematically justifiable killings on this retributive ground alone. True, the process for determining guilt was a bit abbreviated and informal in that no court ever passed on the guilt of those executed. Yet when capture and trial is not feasible for one reason or another, better these executions of deserving wrongdoers than no punishment at all.

Second question: if there is such strong permission, is it nonetheless not available because the consequences of preemptively killing terrorists are so horrendous that they override the permission? The consequences that come to mind here are those brought forward by liberal political theorists: law and order on the international scene will disappear, no one will trust or respect any country that engages in such practices, terrorism against us will only be heightened and prolonged by such practices, etc. Conceivably such kinds of consequences could reach a magnitude that could override a nation's right of self-protection.70 Witness the sacrifice the French and British governments asked of the Czechs in 1938, giving up the only defensive line the Czechs had with which to exercise their right of self-defense against German aggression, in the name of avoiding the horrendous consequences of world war. Yet as in 1938, the question is an empirical one: will these bad consequences really be averted if targeted killings do not take place? Equivalently, will the international order generally, trust or respect of America specifically, or the self-restraint of terrorists up to now, disappear or even be much diminished if America and Israel continue their practices of targeted killings? I doubt this very much. The only thing that makes a horrendous level of these consequences seemingly plausible is the confusion of empirical likelihood with a kind of formal postulate of political philosophy. The postulate is that we should only do what we would allow others to do, or what we could give reason to them to accept what we do, or what some neutral observer stripped of any beliefs in the justness of our cause versus theirs, would accept. These postulates (that I put aside earlier) are no substitute for the needed empirics. The empirical questions are not, for example, whether everyone should (p.460) be allowed to do what we do;71 rather, it is whether they will do what we do. And the answer to this empirical question seems to be rather clearly no.

Third question: are the consequences of targeted killing on the whole good? This is a large question. We need to introduce some distinctions to make it more manageable. The first is a distinction I have intentionally suppressed until now. This is the distinction between justifying an individual decision to target a particular terrorist for assassination, on the one hand, and justifying the general practice of targeted killings, on the other.

To use this distinction here is not to return to the much-rejected idea that there is a kind of indirect, or two-step consequentialism, often called, “rule-consequentialism.” Such a form of consequentialism has long been known to be incoherent, collapsing as it does either into act-consequentialism or irrational rule-worship.72 Rather, the consequentialism urged here is always an act-consequentialism; that is, a calculus which one uses to decide how to act on particular occasions. It is just that some actors, because of their institutional roles, have a choice set that is limited to the enactment of general rules. As Rawls pointed out years ago,73 legislators, for example, can only decide on the enactment of statutory rules; they cannot make individual decisions under those rules because that is not their institutional role.Similarly here, policymakers within the CIA and elsewhere in government are not themselves in the business of deciding whether and when to target particular terrorists. Their choice set consists of general policies defining the practice of targeted killing that they think will produce the best consequences on the whole.

It might be thought that the consequentialist calculation (about the general practice of targeted killings) has at least the potential to preempt any calculation of consequences by CIA officers doing the targeting. As long as the general rule is not, “use your discretion,” in other words, it might be thought that individual CIA decisions are wholly governed by the general policies decided upon and not by any further consequentialist calculation. Yet this is not so, because (among other things) the normative viewpoint of the officer in the field is not the same as the viewpoint of those making general policies. There is always a “gap” between the viewpoints of those who make rules, and those who apply them.74 As much as the former like to pretend that their general decisions fully govern the particular decisions of officers in the field, in fact those officers have to make their own consequentialist (p.461) calculations. Thus we have two consequentialist balances to assess: one, does the practice of targeted killings produce net good consequences, and two, does some particular targeted killing produce net good consequences?

Separate as these two questions are, there is an impact that the answer to each has on the answer to the other. The field officer, when calculating the consequences of a particular targeting, should take into account those long-term consequences (such as precedential effect, reputational injury to his country, disbelief in international law, etc.) to which his individual action may contribute.75 Similarly, the policymaker when calculating the consequences of some general policy on targeted killings, should take into account that his policy will be imperfectly applied by those in the field. The latter's calculation, in other words, should not be, “What would be the best rule to enact if universally followed?” Rather, it should be, “What is the best rule to enact in light of how I predict it will actually be applied in the field?” Despite these interactions, we still have two separate consequential calculations to consider.

Consider first the general practice of targeted killings, as say it is currently being carried out by the intelligence communities in the United States and Israel. Are these practices producing better consequences than their conceivable alternatives (such as a flat ban on such killings, or an expansion of them)? Now the considerations of political philosophers mentioned above in discussion of Question Two can come into their own, pitched perhaps at a less hysterical level. It does cost America's reputation something, does it not, to kill people around the world that it alone judges to be a threat to its interests, when those people are not combatants in any declared war? The damage done to America's reputation by the Bush Administration's well-publicized practices of torture illustrates how serious a harm this can be.

On the other side of the balance, of course, is the intended benefit of targeted killings: They are intended to prevent great harms to America, Americans, and the citizens of our allies. Whether much of such benefit is achieved by the practice of course depends on the accuracy of the targeting, a matter difficult to assess from public information. In principle it is certainly possible that these gains in security and safety outweigh the costs to reputation (and respect for law) that are also consequences of the practice.

The consequentialist calculations of officers in the field are even more difficult to assess from the armchair. For everything depends on the details of individual cases.We are told that the CIA has levels of review on proposals to assassinate given terrorists, and that numerous candidates proposed for assassination are declined by this higher order review.76 Both Israeli and American targeting also make efforts (p.462) to avoid “collateral damage,” the euphemism for killing or injuring those near those targeted but who have no culpable connection with them.77 If this works as it should, it is plausible that net good consequences are produced in some (many, most?) cases of targeted killings as currently practiced.

Even in cases where mistakes are made—the one targeted turns out not to have been a threat to U.S. or Israeli interests—those decision-makers assessing their own responsibility need to distinguish the question of whether they did the right thing from the question of whether they were culpable in doing the wrong thing. As the criminal law extending from Aristotle to the current Model Penal Code reflects, these are two distinct moral questions.78 Acts of mistaken targeting in fact do the wrong thing, but if such targetings are reasonably believed to be necessary to prevent serious injury by those who do or order them, culpability is lacking. For those like John Rizzo “in session with himself,” regret at wrong choices would be appropriate (and aretaically, demanded), but guilt would not.

There is an interesting skewing question about doing the consequentialist calculation at either level, that of general practice or of particular targeting. Before, I noted that a sliding scale threshold deontology could be seen as a skewed consequentialism, skewed by the great but not infinite weight each actor should give to the fact that it is his or her agency that is involved in causing some harm or evil to exist. Now consider skewing the consequentialist calculation by the identities of the victims, not the agents. Standard consequentialisms follow Bentham's dictum that “each is to count for one, but only one.” A consequentialism skewed by the identities of victims and beneficiaries, by contrast, might weight the interests of the “near and dear” greater than those of strangers.

As applied to the present situation, such a skewed consequentialism might weight the lives of U.S. citizens more heavily than the lives of non-U.S. citizens. Consider as an example of such skewing the decision about the altitude at which bombers should fly prior to the introduction of “smart bombs.” Suppose the facts are these: the numbers of innocent noncombatants killed varies directly with the altitude of the bombers (viz., the higher the planes, the greater civilian casualties because of the lesser accuracy of the bombs). Suppose further that the number of airmen killed varies inversely with the altitude, because the higher the plane the less vulnerable it is to anti-aircraft fire. Assume (contrary to fact, but to simplify) that the accuracy in hitting the intended targets does not change with altitude.

A standard consequentialism, where each noncombatant life is equal to one airman's life despite their different citizenship, would recommend bombing at that altitude where one airman will be lost for each noncombatant killed, whereas a victim-skewed (p.463) consequentialism would recommend a somewhat higher altitude, valuing a U.S. airman's life as the equivalent of some large number of noncombatant, non-U.S. citizen lives. A radically skewed consequentialism would not count noncombatant, non-U.S. citizen lives at all in assessing the proper level of risk to U.S. airmen.

One of the features we like about the use of unmanned drones to do our targeted killings, of course, is that we eliminate the risks to U.S. airmen that would exist if we used manned aircraft. If this substitution were at the cost of increased civilian casualties, we would be facing the same question of differential weighting as in my older bombing scenario.

My own sense is that some skewing of the consequentialist calculus is appropriate in military contexts such as these. Putting our own airmen at risk up to the level equal to the risk imposed on innocent noncombatants seems too idealistic; ignoring the interest of the latter entirely seems too heartless. We have a vague-by-degree privilege to prefer our own in making such consequentialist balances, difficult as that is to quantify.

Fourth question: are we prima facie debarred from achieving whatever balance of net good consequences as can be achieved by targeted killings, because of a prima facie deontological duty prohibiting such killings? As long as we recognize the work done by “prima facie” here (in putting aside momentarily the possibilities of strong or weak permissions or threshold overrides), surely the answer to this question is in the affirmative. Targeted killings are intentional killings of human beings.79 As the contemporary electronic commentary on the proposed 1938 assassination of Hitler notes, “Assassination is hard to justify when you're not at war—it is generally called ‘murder’.”80

We have seen two possible exceptions to this deontological norm (in the form of strong permissions to defend ourselves and to punish past culpable wrongdoing) that, if they exist, would render this prima facie prohibition idle. Yet at this stage of our decision procedure we should assume that these permissions do not exist.In which case, the prima facie deontological prohibition threatens to override the consequentialist balance in favor of such killings. More pertinent at this stage in our deliberations would be the existence of weak permissions, as those exist either in the form of further exceptions to the deontological norm against murder, or in the form of scope limitations on all such deontological norms. I shall examine these possibilities in the fifth and sixth questions below.

Fifth question: are there further exceptions to the norm prohibiting murder beyond self-defense and retributive punishment that are relevant here? Few come to mind. (p.464) Perhaps if terrorists forfeit their rights (to life and bodily integrity, among others) by engaging in acts of terrorism threatening those rights in others, that constitutes a (weak) permission to kill them, a permission exercisable only if net good consequences are produced by such killings. Such rights-forfeiture ideas are the basis for one of the extant theories about the moral basis for self-defense.81 Yet neither in that context, nor in this, are such forfeiture theories plausible.

Sixth question: are some ways of arranging for the deaths of terrorist suspects outside the scope of deontological obligations generally, and thus weakly permitted? Most targeted killings are central cases of acts (not omissions), strongly causing (not merely aiding, allowing, or redirecting) the deaths of terrorists who are not about to die anyway (versus mere accelerations of inevitable deaths) where those deaths are intended (not merely foreseen or risked). Such central cases of targeted killings cannot thus avail themselves of what I earlier called the scope limitations common to deontological obligations generally.

There are some less central cases, however, where these distinctions may have some bite. One of these has to do with the killing of noncombatants, typically spouses, children, and other relatives of a targeted terrorist. In cases where the killing of the terrorist himself is (strongly or weakly) permitted by an exception (such as extended self-defense or retributive punishment), the justifiability of the action that also kills non-involved others will depend on whether their killings were deontologically prohibited. Here the doctrine of double effect may do some work, for in such cases the intelligence services can truly say that they did not intend the death(s) of those relatives of the terrorist who were at the wrong place at the wrong time. Indeed, in one well-known case, that of the killing of the Hamas military commander Salah Shehada by Israeli forces in 2002, the Israelis not only did not intend, they did not even know that they would also kill Shehada's14-year-old daughter as well as several other relatives and neighbors.82 While they did know that his wife always accompanied him, and did knowingly accept her death as inevitable, they no more than risked the deaths of the others. In which case the Israeli forces were weakly permitted to kill Shehada where and when they did, if such killing at that time was necessary to prevent even greater evil than was the death of Shehada's relatives and neighbors. (Remember, a weak permission only permits the consequentialist balance to be made; the action itself is permitted only if in addition that consequentialist balance favors that action.)

More occasionally, the omitting, allowing, aiding, and redirecting distinctions may have some bearing on the morality of a terrorist's deaths. In the taking of Al Qaeda's one-time chief of security, Abu Zubeida, for example, he was thrice wounded before capture. One could imagine justifying omitting to provide life- (p.465) saving medical treatment (or discontinuing any already begun) in a situation where his death was necessary to prevent even worse consequences.83 Likewise, where U.S. intelligence knows of a planned Israeli assassination, or vice versa, and could stop it, the omission to prevent such targeted killings by others is surely open to consequentialist justification. Likewise, the sharing of information and other minor forms of aiding targeted killings by others is also open to consequentialist justification, as would be redirecting other nations’ targetings from less to more useful targets.

Suicide bombers and those who, willingly or unwillingly, accompany them to their deaths, present an easy example of justified targeting. Such individuals are about to die anyway, and even if they are coerced into doing what they are doing or are acting in ignorance, the fact of their inevitable death licenses a consequentially justified killing of them. Accelerating the inevitable is open to consequentialist justification in a way that killing someone not under pre-existing threat is not. This is even true of decisions that many of us would find difficult psychologically to make, such as shooting down a passenger airliner commandeered by terrorists in order to prevent destruction of national monuments and the deaths of those who occupy them. Psychologically difficult as these decisions would be to make, morally these are not hard cases.

In these kinds of cases, intelligence and military services may act or fail to act in ways that are not subject to deontology's prohibition on murder. Like all weak permissions, however, that only means that action or inaction in such cases may be justified by the good consequences produced or allowed; the actions/inactions are themselves permissible only if such net of good consequences are actually in the offing.

Seventh question: are some targeted killings justified simply by the catastrophic consequences thereby avoided, without regard to any other permission, exception, or scope limitation? Surely the answer to this last question is sometimes, “yes.” The 1938 proposed assassination of Hitler is to my mind of this character. If that assassination would have prevented the Second World War and the Holocaust, it would have been justified and permissible on this basis alone. The only room for quibbling here might be on epistemic grounds: did Chamberlain's government in 1938 know enough about Hitler and his plans to ground a sufficiently certain prediction as would warrant Hitler's murder? The British colonel in Berlin, Colonel MacFarlane, surely thought so, and of course Winston Churchill had seen the danger for some time. But however much one wants to quibble on these epistemic grounds, isn't the substantive moral principle clear? Hitler would have been justifiably assassinated in light of the horrendous consequences of his survival. (p.466)

In our own times there are equally clear cases “beyond the threshold.” Should a targeted killing be necessary to prevent the detonation or even acquisition of nuclear weapons or of biological weapons of mass destruction, that killing too would be justified on this ground. For most deontologists with whom I am familiar (to a person, threshold deontologists), the hard questions do not lie in these clear cases.They lie in the borderlands. Israel's preemptive strike on the Iraqi nuclear facilities in 1981 come to mind in this regard, as does any similar action the United States might yet take against North Korea or Iranian nuclear capabilities. Or consider the Israeli assassination of Gerald Bull, an expert in barrel ballistics who was assisting the Iraqis under Saddam Hussein in developing a supergun capable of hitting Israel from Iraq.84 If such guns (three were planned) were only capable of firing conventional warheads in Iraqi hands, are such bad consequences enough to cross the threshold of moral permission?

VIII. Conclusion

My topic has been the morality of targeted killings. It has not been about their legality, either under present law or under better law that ought to exist. Even if one agrees with each of my moral conclusions, it would be a mistake to think that one should then simply write them into law. By way of example of the need for caution here, consider the seventh and last point in the preceding section, that of threshold deontology. I meant what I said there: over the threshold, one's duty is to do what it takes—kill innocents, torture, etc.—to prevent true moral catastrophes. Yet quite good reasons might yet stay the hand of the judge or legislator from writing a threshold override into law, for its potential for misapplication may be both large as well as asymmetrical: intelligence agents may well overpredict catastrophe and undervalue non-citizen lives if they were directed by law to apply such a standard.Plus, one suspects, often those acting in such extreme circumstances do not need the law's encouragement to do what they need and ought to do. In which case, one may well not want the law to say explicitly what is morally true.

Still, for the reasons given earlier, it matters how morality stands with respect to practices like targeted killings. It matters to those who do such things, and it matters to those of us in whose name those things are done. My aim in this chapter has not been to be definitive on this question; only to get those who order or execute such decisions to see the possibility of ordered, rational analysis of such issues, and to take seriously the implications of that analysis for the morality of what they do. Such thoughtfulness about morality is not pie-in-the-sky, ivory tower academics; it is where decent people live.


(1) Roger Moorhouse, Killing Hitler: The Plots, the Assassins, and the Dictator Who Cheated Death (Random House, 2007) 190.

(2) Evan Thomas, “Assassination Is a Two-Edged Sword,” Al Arabiya News, July 4, 2011.

(3) For a description of contemporary American practices, see Tara McKelvey, “Inside the Killing Machine,” Newsweek (February 21, 2011) 34–7; for a description of recent Israeli practices, see Alan Dershowitz, Pre-Emption: A Knife that Cuts Both Ways (W.W. Norton, 2006) 121–40.

(4) For a discussion of some differences between legal and moral permissibility in this context, see Jeff McMahan, “Targeted Killing: Murder, Combat or Law Enforcement?” in this Volume, ch. 5.

(5) As an example of such questions of political philosophy, see Jeremy Waldron, “Justifying Targeted Killing With a Neutral Principle?,” in this Volume, ch. 4.

(6) Herbert Morris explores a variety of “non-moral guilt” feelings, of which this is a species, in his “Non-Moral Guilt” in F. Schoeman(ed.), Responsibility, Character, and the Emotions (Cambridge University Press, 1987).

(7) On the “emotional tail” people are thought virtuous to feel in these circumstances, see Bernard Williams, “Ethical Consistency,” Proceedings of the Aristotelian Society, Supp. Vol., Vol. 39 (1965) 103, 107–8.

(8) Michael S. Moore, “Torture and the Balance of Evils,” Israel Law Review, Vol. 23 (1989) 280–344 reprinted in Moore, Placing Blame: A General Theory of the Criminal Law (Oxford University Press, 1997) ch. 17; Moore, “Patrolling the Borders of Consequentialist Justification,” Law and Philosophy, Vol. 27 (2008) 35–96, reprinted in Moore,Causation and Responsibility: An Essay in Law, Morals, and Metaphysics (Oxford University Press, 2009) ch. 3.

(9) On the nature of moral dilemmas generally, see Michael Moore, “Moral Reality Revisited,” Michigan Law Review, Vol. 90 (1992) 2423–533, 2463, reprinted in Moore, Objectivity in Ethics and Law (Ashgate, 2007).

(10) A famous example was described by Life magazine in 1958, where Seattle formed a “God Committee” to decide who should receive the then scarce respirators needed to live. (The Committee was so named because it attempted to rank potential recipients by the worthiness of their lives.) On triage generally, see Jonathan Glover, Causing Death and Saving Lives (Penguin Books, 1977) 203–27.

(11) As in United States v. Holmes, 26 Fed. Cas. 360 (Cir. Ct. F.D. Pa 1842), where some passengers were jettisoned from a sinking and overcrowded lifeboat.

(12) Regina v. Dudley and Stephens, 14 QBD 273 (1884) is the most famous murder/cannibalism case. For how common these were in the nineteenth century, see A.W.B. Simpson, Cannibalism and the Common Law (Oxford University Press, 1984).

(13) An example behind the Model Penal Code's Section 3.02 balance of evils provision. See Model Penal Code and Commentaries (American Law Institute, 1985). For a real-life example, see the well-told tale (recently made into a film) by Joe Simpson, Torching the Void (Harper Collins, 1988).

(14) William Stephenson, in his A Man Called Intrepid: The Secret War (The Lyon Press, 1976), describes a decision of this kind by Churchill during the Second World War with respect to the German bombing of Coventry.

(15) Another Churchill decision during the Second World War. Ibid. at 414. See also Jonathan Glover, Causing Death and Saving Lives (Penguin, 1977) 102.

(16) I discuss some Israeli cases of this in Moore, “Torture and the Balance of Evils,” supra n. 8, and some American cases of this in Moore, “Patrolling the Borders” supra n. 8.

(17) The euphemistically labeled “extraordinary renditions” of the American Central Intelligence Agency (CIA) were an exempt of this. The CIA's practice in this regard is discussed in some detail in Jane Mayer, “Outsourcing Torture,” The New Yorker (February 14, 2005) 106–23.

(18) See In re A [2001] Fam. 147, when the court faced the issue of whether to separate Mary from Jodie, when to do so would kill Mary but when not to do so would allow both to die.

(19) The choice in the film, Sophie's Choice, Dir. Alan J. Pakula, Incorporated Television Company, 1982.

(20) President Bush's choice regarding United Flight No. 93 on September 11, 2001, had the passengers on that flight not taken matters into their own hands.

(21) I survey a variety of general moral skepticisms in Moore, “Moral Reality,” Wisconsin Law Review, Vol. [1982] 1061–156, reprinted in Moore, Objectivity in Ethics and Law, supra n. 9.

(22) Moore, “Moral Reality” supra n. 21; Moore, “Moral Reality Revisited”supra n. 9.

(23) O.W. Holmes, The Common Law (Little, Brown, 1881) 47; I. Kant, The Metaphysical Elements of Justice, J. Ladd, trans. (Bobbs-Merrill, 1965) 41–2; Sir Francis Bacon, Maxims, reg. 25; Bacon is excerpted in Shedding, Ellis and Heath, The Works of Francis Bacon (Parry and McMillan, 1859) 343.

(24) The late Bernard Williams’ well-known views. Williams “Utilitarianism: Against” in J.J.C. Smart and Bernard Williams, Utilitarianism, For and Against (University of California Press, 1973) 92:

there are certain situations so monstrous that the idea that the processes of moral rationality could yield an answer in them is insane: They are situations which so transcend in enormity the human business of moral deliberation that from a moral point of view it cannot matter any more what happens.

I discuss these views in Moore, “Torture and the Balance of Evils” supra n. 8, in Placing Blame,supra n. 8, 729–30.

(25) Kant, supra n. 23, 25.

(26) Some of the now classic discussions of the “new anarchists” are M.B.E. Smith, “Is There a Prima Facie Reason to Obey the Law?,” Yale Law Journal, Vol. 82 (1973) 950–76; Joseph Raz, The Morality of Freedom (Oxford University Press, 1986) chs 2–4; Heidi Hurd, “Sovereignty in Silence,” Yale Law Journal, Vol. 99 (1990) 945–1028, 1007; Hurd, “Challenging Authority,” Yale Law Journal, Vol. 100 (1991) 1611–77.

(27) Rizzo's role in targeted killings is described in McKelvey, supra n. 3.

(28) Edmund Cahn's phrase, in describing Learned Hand reflecting on questions of good moral character and the like. Edmund Cahn, The Moral Decision (Indiana University Press, 1955).

(29) Truman's thoughts in this regard are described in David McCullough's Truman (Simon and Schuster, 1992).

(30) Earl Warren's later reflections on his Japanese internment decision are described in G. Edward White, Earl Warren: A Public Life (Oxford University Press, 1982).

(31) Three Shin Bet lawyers participated in these decisions, one of whom, at his retirement recently, said he had no regrets about those decisions.

(32) John Winthrop, “A Modell of Christian Charity” (1630), reprinted in Hanover Historical Texts Project (Massachusetts Historical Society, 3rd Series 1838).

(33) See Moore, Placing Blame, supra n. 8.

(34) For an introduction to these terms, see the classic essays on this topic collected in Samuel Scheffler, Consequentialism and Its Critics (Oxford University Press, 1988).

(35) An overview of utilitarianism is supplied in Anthony Quinton, Utilitarian Ethics, 2nd edn (Duckworth, 2007).

(36) For an introductory overview of deontological ethics, see Larry Alexander and Michael Moore, “Deontological Ethics” in Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Fall 2008 edition), available at 〈http://plato.stanford.edu/archives/fall2008/entries/ethics-deontological/〉 accessed November 3, 2011.

(37) See Ihsan A. Hijazi, “Beirut Captors Free 3 Russians After a Month,” New York Times, October 31, 1985, A-1; Charles Chi Halevi, “A Hard United Line on the Mideast,” Chicago Tribune, August 22, 1989, 17.

(38) The Model Penal Code's conclusions, e.g., reflecting what I take to be the common moral view. Model Penal Code and Commentaries, Section 3.04.

(39) Kant, supra n. 23, 100.

(41) Explored by me in Moore, Causation and Responsibility, supra n. 8, 36–41. The three-level analysis is far from universally accepted, although it is getting some adoptions by other ethicists.See, e.g., Jeffrey Brand-Ballard, who regards the analysis as being an “illuminating framework for thinking about the relationship between deontological and consequentialist reasons.” Brand-Ballard further “recommends [this framework] to deontologists. It seems as internally consistent and plausible as any deontological position with which I am familiar.” “Moral Emotions and Culpability for Resultant Harm,” Rutgers Law Journal 2012, forthcoming.

(42) See the survey of possibilities in Scheffler, supra n. 34.

(43) Moore, Causation and Responsibility, supra n. 8, 39–40.

(44) The six is a simplification of the larger number of such distinctions: ibid. at 42–76; Moore, Placing Blame, supra n. 8, 689–703.

(45) As described in Stephenson, supra n. 14.

(46) Director of Public Prosecutions for Northern Ireland v. Lynch [1975] AC 653.

(47) Mayer, supra n. 17.

(48) The “trolley discussion” in philosophy begins with Philippa Foot, “The Problem of Abortion and the Doctrine of Double Effect,” Oxford Review, Vol. 5 (1967) 5–15. There is now a considerable literature.

(49) Stephenson, supra n. 14, 414.

(50) As the Model Penal Code too concludes. See Model Penal Code and Commentaries Section 3.02. Such practice also seems accepted within the contemporary mountaineering community (Joe Simpson, supra n. 13), although it was not the norm in the nineteenth-century British mountaineering community. Thus, the Whymper party disaster on the descent of the first climb of the Matterhorn in 1865 led to an extensive investigation as to whether the rope holding four fallen climbers to three remaining climbers had been cut or had simply broken, the assumption being that if it was cut that would have been improper (despite the then inevitable death of all seven climbers).

(51) As the court concluded in In re A [2001] Fam 147.

(52) As in the film, Sophie's Choice, supra n. 19. That the mother is (perhaps virtuously) racked with guilt, does not mean that she was not justified in saving one of her children when otherwise both would die.

(53) That either person on a plank sufficient to float only one, can throw off the other to his death, has been concluded by Cicero, Aquinas, Bacon, Kant, Holmes, and Glanville Williams. See the citations in Moore, Placing Blame, supra n. 8, 693 fn. 53.

(54) The conclusion of the court (in dictum) in United States v. Holmes, 226 Fed Cas. 360 (3rd Cir. 1842)

(55) Although the court came out the other way in Regina v. Dudley and Stephens, 14 QBD 273 (1884), as Brian Simpson documents in his Cannibalism and the Common Law, the custom of nineteenth-century mariners was to sanction the killing and eating of some when the alternative was for all to die.

(56) See Michael Moore, Act and Crime: The Implications of the Philosophy of Action for the Criminal Law, 2nd edn (Clarendon Press, 1993, 2010).

(57) Moore, Causation and Responsibility, supra n. 8, ch. 2; Moore, Placing Blame, supra n. 8, ch. 5.

(58) Moore, Placing Blame, supra n. 8, chs 9, 11; Michael Moore, “Intention as a Marker of Moral Responsibility and Legal Punishability” in Antony Duff and Stuart Green (eds), The Philosophical Foundations of Criminal Law (Oxford University Press, 2011).

(59) Moore, Causation and Responsibility, supra n. 8, ch. 18.

(60) See G.E.M. Anscombe, “War and Murder” in Walter Stein (ed.), Nuclear Weapons: A Catholic Response (Sheed and Ward, 1962).

(61) See, e.g., Larry Alexander, “Deontology at the Threshold,” San Diego Law Review, Vol. 37 (2000) 893–912.

(62) See Phillip Montague, “Defending Defensive Targeted Killings;” Russell Christopher, “Imminence in Justified Targeted Killing,” in this Volume, chs 10 and 9, respectively; see also Peter Vallentyne, “Enforcement Rights Against Non-Culpable Non-Just Intrusion,” forthcoming, Ratio.

(63) I have long relied on Phillip Montague's notion of self-defense, as first laid out in Montague, “Self-Defense and Choosing Among Lives,Philosophical Studies, Vol. 40 (1981) 207–19. See also Moore, “Torture and the Balance of Evils,” supra n. 8, in Placing Blame, 712–17. That the Bush Administration also adopted this theory of self-defense (in the Yoo-Bybee memoranda on torture)does not dissuade me from its correctness.

(64) See Foot, supra n. 48.

(65) The modern discussion of the innocent aggressor issue begins with George Fletcher,” Proportionality and the Psychotic Aggressor: A Vignette in Comparative Criminal Law Theory,” Israel Law Review, Vol. 8 (1973) 367–90.

(66) The example of Fletcher, “Psychotic Aggressor,” and of the Model Penal Code and Commentaries §3.04.

(67) The example of Robert Nozick, Anarchy, State, and Utopia (Oxford University Press, 1974).

(68) My main defense of there being a kind of justice that is retributive in nature (as opposed to corrective or distributive), is in Moore, Placing Blame, supra, n. 8, chs 2–4.

(69) I raised the punishment of the deserving as a possible exception to the norm against torture, but only as a partial exception because of torture's general impermissibility as a mode of punishment. Moore, “Torture and the Balance of Evils”, supra n. 8 in Placing Blame, 717–19. Targeted killings, by contrast, utilize a permissible mode of punishment, viz., death.

(70) Kevin Govern examines such questions in “Operation Neptune Spear: Was Killing Bin Landen a Legitimate Military Objective?,” in this Volume, ch. 13.

(71) If this universalizability principle were in issue, notice that what it is we do is open to a detailed description allowing only those who, like us, are in the right, to do targeted killings.

(72) The locus classicus of this argument is in David Lyons, The Forms and Limits of Utilitarianism (Clarendon Press, 1975).

(73) John Rawls, “Two Concepts of Rules,” Philosophical Review, Vol. 64 (1955) 3–32. Rawls himself did not always carefully distinguish the role-relativity of different institutional actors’ choice sets, from the incoherent rule-utilitarianism that is not in any way role-dependant.

(74) See Larry Alexander, “The Gap,” Harvard Journal of Law and Public Policy, Vol. 14 (1991) 695–701.

(75) Nicely explored in Heidi Hurd, Moral Combat (Cambridge University Press, 1999).

(76) McKelvey, supra n. 3.

(77) Ibid.; see also Dershowitz, supra n. 3, 121–40.

(78) Moore, Placing Blame, supra n. 8, 45, 191–3, 403–4.

(79) As further defended in Fernando Teson, “Targeted Killing in War and Peace: A Philosophical Analysis,” in this Volume, ch. 15.

(80) Available at 〈http://www.imdb.com/title/tt0360715/〉 accessed November 3, 2011, for Killing Hitler (TV 2003).

(81) On self-defense as an instance of rights forfeiture, see Judith Jarvis Thomson, “Self-Defense,” Philosophy and Public Affairs, Vol. 20 (1991) 283–310.

(82) Recounted in Dershowitz, supra n. 3, 132.

(83) The facts are only imagined. The actual facts were that medical treatment was delayed in order to get Zubeida to talk, not to allow him to die. LA Times, March 6, 2003.

(84) Dershowitz, supra n. 3, 122–3.