Torture and the Balance of Evils
Torture and the Balance of Evils
Abstract and Keywords
If (as the last chapter argues) criminal legislation should prima facie aim at prohibiting all and only moral wrongdoing, then the actual reach of the criminal law will depend on the nature of moral wrongdoing. Is it deontological or consequentialist in nature, for example? The general shape of our moral obligations and thus of their breach, wrongdoing, is approached in this chapter through what is known in criminal law as the issue of general justification, sometimes called ‘necessity’ or ‘balance of evils.’ This justification exists whenever the evil prevented by the doing of some act is greater than the evil of doing that act, considered by itself. The contours of this most general form of justification is taken to reveal the general shape of the wrongs that can be thus justified. A complex deontological form for moral wrongdoing is elucidated in this way, a form allowing thresholds over which consequences govern, exceptions, and limited scope, to moral ‘absolutes’ such as, ‘never torture.’
A legislator adopting the legal moralist theory of proper legislative aim makes moral wrongdoing central to legal wrongdoing. Prima facie for such a legislator all and only what is morally wrongful should be criminally prohibited. This places great weight on the idea of moral wrongfulness.
It is possible to think that morality consists of hundreds or thousands of specific injunctions of the Ten Commandments kind, and that that is all that can be said about moral wrongfulness. Yet few of the many philosophers who have spent much time in ethics have ever thought this. Rather, the assumption has been that morality has more structure to it than that and that a theory of substantive ethics is a description of that general structure.
At the most general level of ethics, that structure has been thought to be of one of two kinds. On one view morality consists in only one injunction to all of us all the time: maximize good consequences, and minimize evil ones. Of course, such a monistic theory of right action can (but need not) become very pluralistic in what are intrinsically good or intrinsically bad states of affairs. Nonetheless, the injunction to each of us is to maximize the one and minimize the other. On the alternative view, morality consists of many injunctions: we are enjoined not to kill, rob, pillage, etc. What keeps this view from collapsing into the first view merits careful scrutiny. For why could one not say that the adherence to each of these injunctions is intrinsically good, their breach, intrinsically bad, and then think that one is enjoined by morality to maximize the one and minimize the other?
What makes the second view distinct from the first is its refusal to make this last step. Rather, on the second view morality is personally addressed to each of us. Its injunctions say categorically to (p.670) each individual: ‘don’t you violate this injunction now, even if your doing so now would minimize violations of this injunction by you or others in the future’.
The most general question of substantive ethics (as opposed to meta-ethics) is which of these two views about the shape of morality’s injunctions is right. This is thus a kind of watershed issue for any theory of criminal law that would mirror its criminal prohibitions in the injunctions of morality.
In Part II I said that the royal road to the theory of moral culpability that underlay the general part of the criminal law lay in the excuses. Here there is an analogous truth; to see what general shape moral norms possess, the best heuristic is to look to the issues of justification. For it is here, where there is much to be said for doing something that is normally quite wrongful, that we can work our intuition pumps most vigorously in seeing the general shape of morality. Hard cases may make bad law, as Holmes warned us, but in ethics they are indispensable for sorting out what we think.
I will thus approach this most general question about wrongfulness through the issue of justification. Moreover, in this chapter I will pursue the issue of justification with respect to two actual cases on which I was asked to comment in the past. Since such cases involve deliberate torture by state security forces but where the potential for good consequences was high, the cases work our intuition pumps most vigorously.
In the first case, the No. 300 bus, a civilian bus operating within Israel, was hijacked by four Palestinian terrorists in 1984. Once the terrorists began killing passengers, the Israeli security forces stormed the bus. All four terrorists were reported killed in the retaking of the bus. As it happened, news footage subsequently revealed that two of the terrorists had been quite alive when they were lead away by the Israeli security forces. They were apparently beaten to death during interrogation, as one of their interrogators confessed to the newspapers upon his retirement recently.
The second case involved a junior officer in the Israeli Defence Force, a Lieutenant Nafsu. He had been accused of smuggling and hiding arms for the PLO. The interrogation which produced his admission in this regard had involved what was later called ‘physical pressure’. Such ‘pressure’ consisted of slapping, pushing, hair pulling, sleep deprivation, cold showers, and misinformation and threats about his family.
(p.671) These two cases brought to light the existence of a little known counter terrorist unit of Israel’s General Security Service (GSS). This unit, the Investigation Unit, was set up in 1967 after acquisition of the occupied territories raised larger fears of terrorist activities within Israel. The unit was not a prosecutorial unit, and, indeed, during the early years of its existence was forbidden from testifying in criminal trials. The function of the unit was to prevent terrorism, not punish it.
The prevention methods of the unit included interrogative techniques involving physical abuse. Such techniques were not revealed even to prosecutorial agencies, and when agents of the unit were allowed to testify in criminal trials, as a matter of policy they would lie about the methods of interrogation.
The revelation of the existence of this unit and its practices in the two cases mentioned gave rise to the appointment of a Commission to study the situation. Appointed by the Israel Supreme Court, and headed by Moshe Landau, a retired Justice of that court, the Commission in 1987 issued its report in two parts, one of which was declassified and published.1
This chapter explores the questions raised by the issuance of the Landau Commission Report: what is the legal and moral status of torture of terrorist suspects and others, when that torture is engaged in by an agency of the State of Israel for the purpose of extracting information potentially saving many Israeli lives? More specifically, was the Commission right in its retrospective conclusion that ‘the methods of interrogation … employed [in the past by the GSS] … are largely to be defended, both morally and legally …’?2 Was the Commission right in its prospective conclusion that no new legislation is needed to deal with the methods of interrogation of the GSS because ‘the GSS can turn a new leaf … within the framework of the existing law …’?3
The Commission’s Report strongly condemns the long-standing practice of the GSS to have its investigators lie to prosecutors and courts about the interrogative techniques used to obtain confessions. The Commission ‘utterly rejected’ any notion that the GSS can hide its methods by perjury.4 With this finding I shall not be concerned. More troubling is the Commission’s approval of the (p.672) morality and legality of interrogations such as that of Lieutenant Nafsu. Such assaults during interrogation, the Commission concluded, are neither immoral nor illegal when they are, for example, conducted ‘in order to induce him [the terrorist] to talk and reveal a cache of explosive materials meant for use in carrying out an act of mass terror against a civilian population …’.5 The Commission found it ‘self-evident’,6 ‘according to the concepts of morality implanted in the heart of every decent and honest person’,7 that such interrogations are justified. From this moral conclusion two legal conclusions emerged: first, that there is no civil or criminal liability of GSS personnel for past interrogations under the defence of necessity;8 second, that there is no immediate need for remedial legislation to govern future interrogations of terrorist suspects.9 In what follows, I shall examine these two conclusions, and their moral premise as they relate to the general defence known in the criminal law as the ‘balance-of-evils’ defence, a defence to which the Commission attached ‘central importance’ in reaching its conclusions.10
II. The Retrospective Question: was Torture Ever Justified Under the Defence of Necessity?
A. Necessity and the Balance of Evils
1. The Nature and Rationale of the Defence
In pursuing the question of the liability of GSS interrogators for past acts, I shall confine myself to criminal liability. I do this because both civil and criminal liability turn largely on the interpretation given to the defence of necessity, and, although there is a defence of necessity in tort law, the literature about it is not nearly so well developed as it is about the criminal law defence of necessity.11
(p.673) Section 22 of Israel’s Penal Law contains Israel’s version of the criminal law defence of necessity. It provides:
A person may be exempted from criminal responsibility for an act or omission if he can show that it was done or made in order to avoid consequences which could not otherwise be avoided and which would have inflicted grievous harm or injury on his person, honour, or property or on the person or honour of others whom he was bound to protect or on property placed in his charge: provided that he did no more than was reasonably necessary for that purpose and that the harm caused by him was not disproportionate to the harm avoided.12
A number of preliminary remarks may help to locate this defence within the law of crimes. First, it is a defence. It is not part of the prosecution’s prima facie case. The prima facie case against a GSS investigator employing Nafsu-like methods of interrogation against some suspect might be for assault, battery, or the more specific crime of using force or violence against a person for the purpose of extracting from him a confession or information.13 The actus reus and mens rea requirements of such crimes should usually be satisfied in cases like Nafsu’s, so that it is only by a defence like necessity that a GSS investigator may escape liability.14
Second, necessity provisions should be regarded as justification defences, not excuse defences. The precise borders of what is a justification and what is an excuse are a matter of some debate in current moral and criminal law theory.15 As set forth in chapter 1, a (p.674) justification shows that prima facie wrongful and unlawful conduct is not wrongful or unlawful at all. For example, killing another is prima facie both wrong and unlawful, but killing another in self-defence is neither because it is a justified act.16 By contrast, an excuse does not take away our prima facie judgement that an act is wrongful and unlawful; rather, it shows that the actor was not culpable in his doing of an admittedly wrongful and unlawful act. For example, killing another is both wrongful and unlawful, and it remains so when done by an insane person; but such an actor is not liable for his admittedly wrongful act because he was not culpable in his doing of it.
In some penal codes it is clear that necessity is a justification defence. For example, the American Law Institute’s Model Penal Code provides:
Conduct which the actor believes to be necessary to avoid a harm or evil to himself or to another is justifiable, provided that the harm or evil sought to be avoided by such conduct is greater than that sought to be prevented by the law defining the offence charged …17
This is not the language of excuse, for no mention is made of the actor’s diminished capacities or impaired opportunities to make the right choices; rather, it is the language of justification, for the defence is only available when the actor did what was on balance the right thing to do (that is, he prevented more harm or evil than he caused by his action). The language of the Israeli Code is less clearly focused on justification exclusively, for it only requires that the harm caused by an actor not be ‘disproportionate’ to the harm he avoided by his action. While such language might allow for an excuse version of the defence, it certainly allows for some justification version.18
(p.675) Third, necessity provisions are usually catch-all justificatory provisions. That is, most penal codes make available specific defences for justifications such as self-defence, defence of others, defence of property, use of force in law enforcement, and the like. These more specific defences capture the most recurrent situations in which one person is justified in using force against the person or property of another. General necessity provisions stand behind these more specific provisions in the sense that they allow a defence in any of the infinitely various circumstances that may arise whereby violating the criminal code is the right thing to do (i.e., it is justified) because the alternatives were all worse. Violating the traffic laws to get an injured person to the hospital and burning several houses to create a fire-break needed to save a city from fire, are common examples.
The rationale behind such open-ended, justificatory defences as section 22 should by now be apparent. Any criminal code should allow punishment only where it can establish moral blameworthiness. Such moral blameworthiness, in turn, can only be established where an actor was neither justified nor excused in his actions. Yet the situations that may justify an actor in doing something otherwise criminal are so various that no set of specific justificatory defences can describe them all—’fact is richer than diction’, as one ordinary language philosopher once put the point. Hence, the need for the open-ended, justification defence of section 22: to capture all those situations of justified behaviour not recurrent enough to warrant a specific defence for them.
2. The Fusion of the Legal and Moral Issues: The Moral Knowledge Needed to Apply the Defence
In light of the necessarily broad language used in provisions like section 22, the Commission’s observation that ‘this statutory provision has to be filled with interpretive content …’19 is something of an understatement. When an actor such as a GSS investigator attempts to apply such a provision to his own contemplated behaviour, or when a court applies such a provision to behaviour already done, it should be clear that at least one kind of moral knowledge must be applied. The actor and the court that judges him must know what is good and bad, beneficial and harmful, and it must know comparatively what sorts of things are worse than others. Actors and courts need nothing less than the (p.676) ability to rank-order quite different states of affairs from worst to best in order to judge whether some action produced more harm than good (and so was not justified) or produced more good than harm (and so was justified).
This kind of moral knowledge is required not only of courts seeking to apply the defence of necessity to situations like the Nafsu case. Citizens also must make the balance of evils required by this defence, and they must get the balance right if they are to avail themselves of the defence. One sees this last point in light of the differential treatment accorded mistakes of fact versus mistakes of value under the defence.20 Citizens who believe they are justified because of certain erroneous factual beliefs may nonetheless use the defence; citizens who believe that they are justified because of certain erroneous value judgements may not use the defence.
The reliance on this kind of moral knowledge by the law is familiar in contexts other than the necessity defence.21 In judging whether an action is negligent or reckless in either criminal law and torts, it is not enough to show that that action posed a significant risk of harm to others or that the actor was consciously aware of such a risk. In addition, the taking of the risk must be unjustified, that is, the consequence of not doing the action (and thus, not taking the risk) must be less bad than the harm risked.22 One must, that is, ‘balance the evils’ to decide whether an act is negligent or reckless.
It should thus be obvious that the balance of evils or necessity defence calls for the application of moral knowledge of the familiar kind I have described. Less obvious is whether the defence does not also call for a second kind of moral knowledge in addition to the knowledge of what are better and worse states of affairs. Consider the following example, familiar in philosophy and legal (p.677) theory for the past twenty years.23 S is a surgeon who, among other things, performs organ transplant operations. He has five patients, all on the verge of death for want of a vital organ. One needs a heart, two need one lung each, and two need a kidney apiece. V is a patient of S’s who is perfectly healthy, visiting S only for an annual check-up. May S kill V in order to harvest V’s organs, claiming as defence that he was justified by effecting a net saving of four lives?
Most people have a firm moral intuition that the answer to this question must be ‘no’, that it is not morally permissible in such a case to justify a prima facie wrongful action by the good consequences it produces. If this is true, and if there are other cases where good consequences cannot justify a prima facie wrongful action, then a second kind of moral knowledge is needed in order to be morally justified in doing some prima facie wrongful action. Needed is some moral knowledge of when good consequences may be used in justification, and when they may not. Needed is what I shall call knowledge of the proper domain for consequential calculation. When is the common saying that the ‘end cannot justify the means’ applicable, and when is it not?
One might think that this is only a moral problem, not a problem for applying the legal defence of necessity. After all, there is nothing in the language of either section 22 of Israel’s Penal Law or of section 3.02 of the Model Penal Code that requires that this second kind of moral knowledge be utilized in applying this defence. Compare both provisions to that of Germany, whose penal code restricts the necessity defence by the requirement that the ‘act be an appropriate means’.24 As construed by George Fletcher, ‘the principle of “appropriate means” signals absolute restraints on pursuing utility maximization’.25 Put in my language, (p.678) under German law there are some acts so wrong that even very good consequences cannot justify them.
Yet I doubt that there is the contrast between German and American law on this point that Fletcher suggests. Even without the ‘appropriate means’ language of the German code, and despite the clearly utilitarian leanings of the drafters of Model Penal Code section 3.02,26 American law does not justify acts like that of my hypothetical surgeon. As Sanford Kadish has observed, there are some acts:
fairly within the net-saving-of-lives, lesser evil doctrines that it is very doubtful courts would sanction—for example, killing a person to obtain his organs to save the lives of several other people, or even removing them for that purpose against his will without killing him. The unreadiness of the law to justify such aggression against non-threatening bystanders reflects a moral uneasiness with reliance on a utilitarian calculus for assessing the justification of intended killings, even when a net savings of lives is achieved.27
Kadish too notices that our moral knowledge about what makes actions right or wrong to do is not exhausted by calculating the good versus bad consequences of those actions. It is this second, non-consequentialist kind of moral knowledge that gives rise to the ‘moral uneasiness’ Kadish notices, a moral uneasiness as much a part of American law as is the consequentialist calculation itself.
Israel’s section 22 should be read analogously. Although lacking any explicit language restricting the domain for consequential justification to actions that are ‘appropriate means’, section 22 cannot be read to justify the harvesting of organs from one to save two or more. This, because Israel too seeks a criminal law that reflects its fundamental moral beliefs. One of those beliefs is that the end, no matter how good, does not justify the use of any means to achieve it.
Another temptation to be resisted here is the following. It might be thought that there is no ‘second kind of moral knowledge’ (p.679) needed to apply the necessity defence. Rather, this view would continue, the only knowledge needed is consequentialist in nature; but the notion of what is a good or a bad consequence needs to be enriched. In my surgeon hypothetical, for example, the net savings of lives is a good consequence, but overlooked is a very bad consequence of killing V: V did not just lose his life (as he might, say, in an avalanche); he was murdered, and any full balance of the evils must take this very evil consequence into account. When that consequence is factored in, this line of thought concludes, it will show that the surgeon did not achieve ‘the lesser evil’ by his killing V, even though he did effect a net saving of four lives. Thus, the only moral knowledge needed here is knowledge of good and bad consequences—recognizing that what is good and bad is not simply a function of harms people suffer but also of the moral rights they have that are violated, the moral duties others have that are violated, etc.
Leo Katz in his fine first book on criminal law theory takes this line on the necessity defence.28 Finding that the hypothetical surgeon clearly may not kill one to harvest the organs needed to save five, Katz concludes:
It would be wrong to say that a person is free to kill whenever he brings about a net savings of lives … Fortunately, the principle of necessity is more flexible than the argument of numbers out of which it grew: Applying that principle, we would probably conclude that in this case [the case of the surgeon] ‘the harm or evil sought to be avoided’ is not ‘greater than that sought to be prevented by the law defining the offense charged’.29
Yet this interpretation of the defence of necessity is unacceptable. Consider the following variation of the surgeon hypothetical. Susan is a surgeon who performs transplant operations, harvesting organs from healthy victims whenever it is necessary to do so to effect a net saving of lives. Susan in the near future will harvest such organs from five healthy patients in order to save more than five dying patients. Sharon is another surgeon who knows this and knows that the only way to prevent this is to kill Susan’s husband while he is on the operating table before Sharon. (Susan’s husband has just been rushed in to Sharon for an emergency operation; Susan is so attached to her husband that she will not be able to carry on for (p.680) some time after his death.) Sharon also has several dying patients, and should Sharon kill Susan’s husband Sharon would use the latter’s organs to save as many patients as possible. If killing a patient to harvest his organs is such a great evil, may Sharon perpetrate that very evil in order to prevent even more of it from being done at the hands of someone else such as Susan?
Someone accepting Katz’s interpretation of the necessity defence is committed to allowing Sharon to kill Susan’s husband and to use that person’s organs to save others. For notice that the very same moral view that makes Sharon’s act so evil guarantees that the evil Sharon seeks to avoid—Susan’s very same type of action, except done five times over—is much greater.30 In other words, Sharon would be justified on Katz’s view by having balanced the evils correctly. Yet I doubt that Katz or anyone else wishes to say that Sharon’s act is justified. Sharon should be allowed the necessity defence no more than Susan, should Susan go ahead and harvest organs through murder. Both have violated a moral restriction on the means they may use, a restriction that is not reducible to some consequentialist calculation, no matter how sophisticated.
B. Various Moral Theories About When Good Consequences Justify Otherwise Evil Actions
My two surgeon hypotheticals are only designed to get one’s intuitions in play, intuitions pointing to the need for a second kind of moral knowledge in applying the necessity defence to actual cases. We now need to examine the large moral issue actually involved to sort in a more systematic way the possible answers to my question of whether there are two kinds of moral knowledge needed here. Distinguishing between four sorts of moral theories frames the large moral issue I wish to examine. I shall discuss three in this section, paving the way for a detailed exposition of a fourth in the following section.
1. Act-Consequentialist Moral Theories
Someone who is an act-consequentialist in his moral theory will believe that there is no second kind of moral knowledge needed to apply the necessity defence, for an act-consequentialist judges the rightness (justifiability) (p.681) of any action by one criterion only: does the act produce better consequences than any alternative action reasonably available to the agent?31 Any act, be it torture, murder, rape, or whatever, may be done, on this view, if a net of good over bad results from it.
One problem with assessing this view is that by itself, it does not say very much. It only says: whatever is good, maximize it by your actions. Quite different versions of act-consequentialism can be produced by varying what it is that is thought to be good. One historically important variant here is act-utilitarianism. A utilitarian theory of the good holds that the only good state of affairs (to be maximized by the consequentialist principle) is the happiness, pleasure, preference satisfaction—or, more generally, welfare—of persons, each counting for one but only one (in Bentham’s famous phrase). Specifically not included in the utilitarian theory of the good are states such as the state of moral rights not being violated, moral duties being done, moral virtues being realized. Only to the extent that people prefer these states does the utilitarian factor them into his calculus.
The problem with this utilitarian form of act-consequentialism has been stated many times. Consider Sam Scheffler’s version of the objection:
[S]uppose that your country is waging a just war, and that an enemy agent you have captured tells you that he has planted a bomb in an area crowded with civilians and that, unless defused, it will soon go off, killing many people. Suppose that there is not enough time to conduct a general search for the bomb, and that all of your attempts to get the agent to reveal its location are unsuccessful. Suppose, however, that you have captured him with his family, and that by torturing his small child in front of him you could eventually destroy his resolve and get him to give you the information. Utilitarianism seems to imply not only that you may but that you must torture the child. These implications and others like them strike many people as entirely unacceptable.32
Whether one finds such implications unacceptable is one of the watershed questions the answer to which says a great deal about (p.682) the kind of morality to which one subscribes. For if the answer is no, that torture in these circumstances is morally right (and thus, legally justified), then one is probably a utilitarian as well as a consequentialist. For you would probably be saying that neither the rights of the innocent child, nor the virtue of his torturers, are of any weight in calculating the consequences of torturing the child. What does count as a bad consequence is the pain inflicted on the child, and, secondly, on his relatives and any others empathizing with him. Also counting as a bad consequence may be the preferences of many of your countrymen that what they see as the child’s moral rights not be violated. But these are costs that can be outweighed by the good you are producing, the net welfare of saving many lives. A plausible application of the utilitarian version of act-consequentialism is thus to torture the innocent child.
I find this conclusion to be morally repugnant. No one should torture innocent children—even when done to produce a sizeable gain in aggregate welfare. Nor should a state allow such practices. As the Landau Commission observed:
the methods of police interrogation which are employed in any given regime are a faithful mirror of the character of the entire regime … All the more so with respect to the interrogation methods of a security service, which is always in danger of sliding towards methods practiced in regimes which we abhor.33
If I had been a Soviet citizen, to continue this example, I would have been deeply ashamed that the security force that acted in my name (the KGB) castrated and killed an innocent Arab who happened to be the brother of one of the terrorists who had kidnapped Soviet diplomats—even though that act brought about the release of the kidnapped diplomats and may well have prevented their deaths.
What this shows is that I am not a utilitarian in my theory of what are good and bad states of affairs. What it does not yet show is that I am not a consequentialist. For one might reject the utilitarian’s exclusively welfarist notions of the good without abandoning the consequentialist principle that the right action is one that maximizes the good. One might, like Leo Katz discussed earlier,34 (p.683) enrich one’s theory of the good to include non-welfarist notions like rights-violations as bad, duty-keepings as good. And one might use such a theory to say that the violation of an innocent child’s right not to be tortured is so bad that it is not outweighed by the saving of many innocent lives.
One might. But no one on reflection should hold such a moral theory, for at least two reasons. One is the lack of any plausible account of why non-welfarist bads like killings and torturing, on the one hand, are so much worse than the corresponding welfarist bads such as naturally occurring deaths or pains, on the other.35 Is it significantly a worse death if one is murdered than if one dies by accident (stipulating that in each case the death is painless and without forewarning)? So much worse that it would be impermissible for us to prevent an accidental death but not a murder, when we can only prevent one but not both? Answering both of these questions affirmatively is required by the non-utilitarian, consequentialist moral theory here considered; and such answers do not seem very plausible.
The second problem for this non-utilitarian consequentialism stems from the kind of example we considered before with the surgeon. Even if a murderous death is so much worse than an accidental death, murder on this theory is still the right thing to do if it is the only way to prevent two or more murders by someone else. Even if torturing an innocent person is so much worse than that person suffering identical pain by accident, torture of the innocent on this theory is still the right thing to do if it is the only way to prevent the torture of two or more innocents by someone else. On the non-utilitarian, consequentialist theory under consideration, in other words, the KGB might well have done the right thing in castrating and killing the brother of the Arab terrorist if they were right in their calculation that this act, bad as it was, was necessary to prevent more acts of the same kind being done to Soviet diplomats.
For both of these reasons one might well reject even this more moralistic form of act-consequentialism. For it is not moralistic (p.684) enough in its implications for behaviour such as torturing the innocent.36 If one agrees with this insight, the most natural move is to adopt some non-consequentialist moral theory. Before I consider such theories, however, a softened form of consequentialism needs brief mention.
2. Rule-Consequentialist Moral Theories
A traditional move37 within consequentialist moral theory to avoid the unwanted conclusion that it is not only permissible but obligatory sometimes to torture an innocent child, is to adopt some version of rule-consequentialism. An acceptable definition of rule-consequentialism for our purposes is given by Gerald Barnes:
An act is right if and only if it conforms with an ideal set of rules; an ideal set of rules is any set of rules such that if everyone always did, from among the things he could do, what conformed with that set of rules, then at least as much good would be produced as by everyone’s always conforming with any other set of rules.38
As Barnes’s definition makes clear, the rule-consequentialist envisions a two-step procedure for any moral decision: first, you decide what the right rule is that you should adopt for your own and others’ behaviour; second, when confronting a particular moral dilemma, you resolve it by simply applying the rule. Consequentialist calculation enters in only at the first of these two stages of decision. That is, you decide what is the right rule to adopt in light of the maximally good consequences that would attend from universal obedience to that rule; but you decide what is the right act to do simply by applying the right rule, not by re-engaging in consequentialist calculations.
As applied to Scheffler’s example of torturing the innocent child of a terrorist in order to extract vital information from the terrorist, (p.685) a rule-consequentialist might say that the right rule to adopt is, ‘never torture an innocent’, even though he recognizes that once in a great while sufficiently good consequences would follow from such torture that, if he were allowed to calculate the consequences, individual acts of torture could be justified on consequentialist grounds. He might say this for a variety of reasons: because individual consequential calculations are factually difficult, error possible, and systematic error in the direction of too much torture rather than too little probable; because individual consequential calculations are time-consuming, time is itself a cost, and the gains achievable by calculating on each occasion to find the exceptional cases when torture is desirable are outweighed by the costs of calculation; because torture harms the character of those who do it or know of it and do not prevent it, no matter that it is done only when necessary to extract vital information, and such damage to generally desirable dispositions is a cost that will always outweigh the good consequences achievable in the few cases where torture of the innocent could otherwise be justified. For any of these reasons, a rule-consequentialist would conclude, the right rule is, ‘never torture the innocent’, and the right behaviour of GSS interrogators is simply to apply this rule without themselves attempting to determine whether in the particular situation they would cause more good than harm by torturing the innocent.
In some such way the rule-consequentialist attempts to soften the apparently immoral implications of consequentialist moral theory. There are two problems, however, for this softened form of consequentialism. One is Kant’s worry that even if consequential calculations justify the right rule they do not do so for the right reason.39 Such calculations make contingent what we experience as categorical. ‘Never torture the innocent’ can be a rule justified on both consequentialist and non-consequentialist grounds; but only the latter mode of justification squares with how we feel about such absolutes, which is not that they are dependent on complicated calculations of consequences that easily could come out the other way (i.e., in favour of some torture of the innocent).
(p.686) The second problem with rule-consequentialism is that it seems irrational (in its own consequentialist terms) when applied as a decision procedure for individual acts of decision. If the Tightness of action is ultimately a function of achieving the maximally good consequences available to the agent in that situation—which is what any consequentialist believes—then sometimes an agent ought to violate what he himself admits is the right rule. Suppose, for example, a GSS interrogator was certain about the immediately relevant facts in Scheffler’s hypothetical—he knows there is a bomb, that it will kill innocents unless found and dismantled, that the only way to find it is to torture the child of the terrorist who planted it. Suppose further he is already in possession of this information, and the costs of calculating utility are thus already ‘sunk’; suppose further that he himself is about to die and that he can keep his action secret, so that the long-term bad consequences stemming from his own or others’ corruption of character will be minimal. In such a case, adhering to the best rule will not be best, on consequentialist grounds. Adhering to the rule in such circumstances would be mere ‘rule-worship’ as one well-known consequentialist has described it.40
For both of these reasons it is difficult to accept rule-consequentialist theories of morality. Try as they might, they cannot accommodate either the content or the categorical nature of moral norms like ‘never torture an innocent child’. Logically, one is of course free to reject such moral norms instead of rejecting consequentialism. That, however, is an easier option to contemplate academically than it is to live with as one’s morality.41
3. A Simple Absolutist View: Never Torture
If one gives up the attempt to sophisticate consequentialist theories so that they seem less immoral in what they would permit, then one’s moral theory (p.687) will be non-consequentialist in nature. The non-consequentialist about morality denies that an action is right just because, of the alternative actions available to an agent, it is productive of the best consequences. He severs, in other words, the connection between good consequences and right action that consequentialism asserts to exist. A non-consequentialist might well say about Scheffler’s example, that it is always wrong to torture an innocent child, even when doing so produces better consequences (such as less torture of the innocent) than not doing so.
It is this feature of non-consequentialist theories that gives them the sobriquet, ‘agent-relative views’. Such theories view morality as consisting of norms directed to each moral agent, norms whose content says, ‘don’t you torture an innocent child’. Such norms are ‘agent-relative’ in the sense that they do not direct each of us to minimize bad states of affairs, such as murders and tortures. For if that were the content of moral norms, we would be justified in torturing one person ourselves whenever it would prevent the torture of many persons by others. Rather, such norms direct us not to do the forbidden acts, leaving it to others to obey or disobey such norms, as they will.
It is often thought that this view of morality’s nature can only be justified within a religious tradition.42 For religion may give one an authoritative text that can be the source of absolute norms like, ‘don’t kill’, ‘don’t torture’, etc. Yet Kant is a prominent example who did not think this, divorcing his morality from any religious basis. In any case, I shall not deal with these large questions here, but I shall assume that agent-relative theories of morality are no worse off with respect to their justifiability and truth than consequentialist theories.
The simplest agent-relative view—and the one most popularly called to mind when one thinks of non-consequentialist theories—is what I shall call the absolutist view of morality. On this view, the moral norms that make up morality are: (1) short and exceptionless injunctions, such as ‘thou shalt not kill’; (2) ‘absolute’ in the sense that they cannot be violated, whatever the consequences may be of not violating them on some occasion; and (3) applicable to (p.688) what we indirectly cause as well as what we directly do through our actions, applicable to what we allow to happen as well as to what we make happen.
It is perhaps this view of morality that underlies Amnesty International’s advocacy of a flat ban on torture in any and all circumstances. Such an absolutist view of morality may also motivate Article 5 of the Universal Declaration of Human Rights and Article 3 of the European Convention on Human Rights, both of which provide flatly that ‘no one shall be subjected to torture …’. Such a view of morality may also have influenced the Landau Commission, which found that ‘the pressure must never reach the level of physical torture …’43 and approved secret guidelines for coercive interrogative techniques for the GSS that ‘will be far from the use of physical or mental torture ….44 If morality consists of such absolutes as ‘thou shalt not kill’, it plausibly contains the norm, ‘thou shalt not torture’ as well.
I do not think that morality includes such an absolute, however, because I do not think that morality contains any such absolutes (i.e., norms with these three absolutist characteristics). Why moral norms must be more complex in order to be plausible I explore in the succeeding section as I lay out the more complex agent-relative view of morality I shall defend.
C. A Complex Agent-Relative View of Morality
1. When Consequences May and May Not be Calculated on Agent-Relative Theories of Morality
No agent-relative view of morality can plausibly hold that we may never calculate the consequences of various courses of action and decide what to do based on such calculations. Where moral norms are not violated by any of the actions we are considering, we are not only free but often obligated to choose our course of action by considering those consequences. When a moral theory tells us what sorts of things its norms prohibit and permit, it thus necessarily also tells us when consequences may and must be calculated—namely, wherever the norms of morality are not applicable.
Finding this permissible/mandatory realm for consequential calculation is thus largely a matter of specifying in detail the content of all the moral norms there are. This of course I shall not undertake, (p.689) although in the subsequent subsection I shall undertake to describe the content of the moral norm most relevant here, the norm forbidding torture of another. Arguably, most agent-relative theories of morality also take some systematic view on ‘the kind of thing to which absolutist prohibitions can apply’.45 I shall first examine the considerable suggestions that have been made in this regard in recent moral philosophy, and then seek to reorder these suggestions in line with what I take to be their underlying insight about when we may (not) justify our acts by their good consequences. I undertake this philosophical task in order to display the general view of morality that I think to be correct in its full complexity so that some judgement can be made as to its overall rationality. Those impatient for concrete application of my complex agent-relative view of morality to torture may wish to skip this discussion, for the bottom line of the analysis for torture is that none of these complexities, no matter how construed, makes torture permissible.
a. Suggested Limits on Consequential Calculation
(i) The allowing/acting distinction
Thirty years ago Philippa Foot suggested that there was a morally important distinction ‘between what one does or causes and what one merely allows’.46 To use one of Foot’s examples: if we rescue a group of five people from torture by another, rather than rescuing one person from such torture, where we cannot rescue all, we have only allowed the one to be tortured by someone else; whereas if we were to torture one person in our custody to prevent another from torturing five in his custody, that would be to do or cause torture ourselves. Foot’s sense was that morality places negative duties on us not to do or cause things like torture, whereas morality places positive duties on us not to allow things like torture to be done where we can prevent it. Foot’s further sense was that negative duties are much stricter than merely positive duties, so that (p.690) consequential calculation could be used to justify violations of the latter but less easily be used to justify violations of the former. In the first variation of Foot’s torture example, where ‘we are bringing aid (rescuing people about to be tortured by the tyrant), we must obviously rescue the larger rather than the smaller group’. In other words, since we will only be allowing one group or the other to be tortured whatever we do, we should guide our action by a calculation of the best consequences we can bring about. By contrast, in the second variation of Foot’s torture example our only way to prevent the torture of five is to torture one ourselves, and here we may not be justified by the good consequences we seek: ‘To refrain from inflicting injury ourselves is a stricter duty than to prevent other people from inflicting injury …’.47
Consider a second example, not involving torture, that may help to bring out the moral force of Foot’s distinction. Suppose we are on a dock and there are two groups of persons drowning in the lake below us. One person is by himself, and the other group consists of five people. Since all are about to drown, we have time only to rescue one group with the one rope on the dock. May (must) we throw the rope to the group of five, even though that means we thereby allow one to drown? Even if the answer to this question is ‘yes’, as I believe it is, may (must) we also throw the rope to save the five when the one has already seized the rope to climb to safety, and we must dislodge him to free the rope to throw it to the five? If we shake the one off the rope to his death, we have caused his death, not merely allowed it. It is much less intuitive that we may justify this violation of a negative duty not to kill by the good consequences of saving five.
(ii) The intending/foreseeing distinction
A much more ancient distinction, going back at least to Aquinas, focuses on the intentions with which we act.48 There is a distinction (p.691) between what we intend and what we foresee. Suppose, to use Foot’s torture example, we foresee to a certainty that the one we have to leave when we rescue five will be tortured by another. There is no sense in which we intend that he be tortured. His being tortured by another is neither an end of ours, nor is it a means to that which is an end, the saving of five from torture. It is only a side-effect we know will come about by our rescuing the five and leaving the one. On the other hand, were we to yield to the threat of the tyrant—that we torture one else he will torture five—we would intend to torture the one. We would not have the torture of the one as an ultimate end—our ultimate end is to save five from torture—but we would have his torture as an intended means.
The ‘doctrine of double effect’ of Catholic theology has long used this distinction to draw the limits of consequential calculation. Where we intend the violation of a moral absolute like ‘don’t torture’, whether as an end or as a means, we may not justify our action by the good consequences it will produce; where we only foresee (even to a certainty) that some act will take place that will violate a moral absolute, we may use good consequences to justify our action. In Foot’s torture examples the doctrine of double effect thus yields the same result as Foot’s doctrine that negative duties prevail over positive duties: we may justify our leaving one knowing he will be tortured, not justify our torturing one, intending to do so, even though in each case our actions will produce the good consequence of saving five from torture.
(iii) The foreseeing/risking distinction
As Sanford Kadish has observed with regard to moral absolutes like ‘don’t kill’, the ‘consequentialist standard is most firmly in evidence when unintended killing is involved. It is revealing that the judgements in this area [of risk creation] … permit life to be yielded when the costs of saving it, in terms of the comforts, conveniences and satisfactions of many, seem too high’.49
There are actually two distinctions at work here. One is between knowing to a practical certainty that someone will die or be tortured, and knowing only that there is a substantial risk that this will be the case. This is a distinction between perceived degrees of (p.692) likelihood that some wrong action will take place. The second distinction is between knowing that somebody will be killed or tortured—somebody in the sense of no one in particular but someone—and knowing that some identified person will be killed or tortured.
The force of these distinctions may be appreciated with a familiar example.50 Statistically we know that activities such as high speed freeway transportation, tall building construction, underground coal mining, etc. cost a certain number of lives each year. Yet we do not know on any given occasion that any particular person will be killed; we only know that there is a risk to someone on each occasion of driving, construction, mining, etc. In such circumstances we have no trouble at all in justifying the risk taken in terms of the undesirable consequences attendant upon lowering it—consequences such as higher transportation, construction, or mining costs. Consequential calculation, as Kadish observes, reigns supreme in such circumstances.
Our attitude changes considerably, however, when we deal with a trapped truck driver, construction worker, or miner. Now we know who is at risk, and we know the risk of death is a near certainty unless we act. Now consequential calculation seems much less acceptable. ‘Save him whatever the cost’ is much more descriptive of our attitude. One might thus think that one of the markers of the domain for consequential calculation is the distinction between risk to someone versus certain harm to some particular one: we may calculate the consequences to justify our action (or inaction) in the former case but not in the latter.
(iv) The victim-in-peril/victim-not-in-peril distinction
It is commonly thought that it matters (for purposes of permissible calculation of the consequences) whether the victim of our action was already in peril anyway. Abortion, for example, is often thought to be an easier decision when doing nothing will result in the death of both the foetus and the mother than when doing nothing will only result in the death of the mother. For in the former case but not the latter the victim of the abortive procedure—the foetus—will die anyway. It is here that the Catholic doctrine of (p.693) double effect operates least intuitively for it prohibits killing the foetus equally in both cases.51
Some would thus substitute for the doctrine of double effect the following principle: when the victim of your proposed action is already in peril of suffering the harm you are contemplating inflicting, you may be justified in going ahead if the consequences (such as saving the mother) are sufficiently good; but where the victim is not in such peril, you are absolutely forbidden to inflict such injury upon him, no matter what the consequences.
Moral theory and criminal law commentary are both rich in examples that implicitly employ this principle. The official commentary to the Model Penal Code section 3.02 on the balance-of-evils defence explicitly approves application of that defence to the situation where a mountaineer must cut the rope holding him to his fallen companion else both, rather than one, will fall to their death.52 Similarly, in situations of shipwreck two men who simultaneously come upon a plank sufficient to float only one of them may each with justification fend off the other; for whoever is the victim is already in peril of drowning.53 Whereas where one has already seized the plank (and thus removed himself from the peril of drowning), others may not thrust him off the plank to save themselves.54
The most famous application of the defence of necessity in the criminal law, the nineteenth-century lifeboat cases, also belong in this category.55 In Holmes, the lifeboat was too heavily laden to (p.694) stay afloat so the defendant lightened it by throwing some of the passengers overboard. In Dudley and Stephens the lifeboat was inadequately provisioned, so two of the crew killed another so that they could survive by eating his flesh and blood. Those critical of the courts’ convictions in those cases often ground their criticism on the principle mentioned above. As Glanville Williams characterizes these cases, ‘there is no choice as to who is to die; the only choice is how many are to die’.56 The victim already being in peril makes these cases much easier to decide for the defendants than they would have been in the absence of this feature:
[T]he act of killing B may be of either of two kinds. B’s death may be slightly accelerated, in the sense that, had A done nothing, B would still have died shortly afterwards. Or B may be deprived by the killing of what would probably have been a number of years of life; in other words, but for the killing he would have been safe. A killing of the former type can obviously be regarded much more leniently than the latter …57
(v) The distinction between redirecting old threats and creating new threats
Due to two influential articles by Judith Thomson many have come to accept what she calls a ‘distribution exemption’ from the absolutist prohibitions of morality.58 As she herself describes this ‘exemption’: we can ‘make be a threat to fewer what is already a threat to more’.59 In other words, we can justify killing or torturing a few with the good consequences of preventing thereby the deaths or tortures of many when the manner in which we kill or torture is only to redirect a pre-existing threat not of our own making. We cannot justify such actions, even by the good consequences of saving many, when the manner in which we kill or torture is to create a new threat to our victims.
Examples speak with greater clarity than generalizations here. Consider three examples. The first is Thomson’s own example where we stand at the switch of a railroad junction.60 We see a run-away (p.695) trolley car which we cannot stop; we can only direct it out to one of two tracks by pulling the switch beside us. If we do not pull the switch, the trolley will continue on the main track killing five trapped workmen; if we pull the switch, we will direct the trolley onto a spur track on which there is only one trapped workman. Thomson urges that we may (and perhaps even must) pull the switch, because in doing so we are only redirecting the pre-existing threat presented by the runaway trolley.
A second example comes from World War II, where British counter-intelligence had penetrated the German intelligence network in England that was responsible for calling in the accuracy of the German rockets aimed for London.61 The British discussed whether they might not feed into the German intelligence system slightly inaccurate reports on rocket hits, thereby redirecting the German rockets from more populous central London to less populous outlying areas north of London. Had the British done this, Thomson’s principle would have approved it, for their act would only have been one of redirecting a pre-existing threat, and thus allowed the consequential calculation about a net savings of British lives.
A third example is provided by the official commentary to Model Penal Code section 3.02.62 Suppose, the Code drafters tell us, we were at the side of a dyke during the time of quickly rising flood waters. If we do nothing, the dike will hold, diverting the waters downstream where a city will be flooded with great loss of life. Alternatively, if we blow up the dyke, the flood waters will be averted onto a farm below us, killing all of its inhabitants. The drafters of the Code make clear that if we blow the dyke and kill the inhabitants of the farm, we will be justified (by the net savings of lives) because we have only averted a pre-existing threat that was going to kill someone.
In each case what seems to matter is that our actions only redirect a pre-existing threat. To confirm this, contrast these three examples with the following three variations: (1) we start a trolley to run over one, because unless we do so another will kill five; (2) we (like Truman in World War II) bomb two towns killing 150,000 inhabitants, in order to induce the surrender of Japan and to (p.696) prevent the loss of the millions of Japanese and American lives that would result from an invasion of the home islands;63 (3) we start a flood by blowing up a dam in order to kill the inhabitants of a farm, because unless we do so a terrorist will blow up an entire city with a hidden nuclear device. In each of these last three examples, the same net savings of lives will be effected by our actions as in the three previous examples. Yet we may think this consequentialist consideration to be of less relevance or significance in the second set of examples than in the first. Such a difference, Thomson urges, is explained by her principle about creating new, versus redistributing old, threats: good consequences justify only when we redistribute a threat, not when we create a new one to some victim(s).
(vi) The distinction between one’s own and others’ projects
Bernard Williams has urged that there is an important ‘distinction between my killing someone, and its coming about because of what I do that someone else kills them: a distinction based, not so much on the distinction between action and inaction, as on the distinction between my projects and someone else’s projects’.64 Williams’s thought is that we are each peculiarly responsible for what we do through our actions; such actions form part of the project we each have that makes each of us the person that we are. We are not responsible for harms that someone else causes, for those are part of their projects’, this, Williams thinks to be true even when we make possible the others’ actions that produce harm.
Consider one of Kadish’s examples: in ‘the often-discussed hostage cases, in which a band threatens to kill two persons in their power in order to obtain the death of one person in the custody of another group … the group may desist from protecting the wanted person and permit the band to enter and kill him …’.65 Although Kadish presents this as an example of an omission where we may use good consequences to justify it, Williams would treat it as an example of an action we do (sending out the one, or stepping out of the way) that allows another to kill him. When the terrorist band kills him, although our actions made it possible, it was not our doing—it was the terrorist band’s doing. Therefore, we can justify our action by the good consequences we achieve, a net saving of (p.697) lives. By contrast, we may not ourselves kill the one the terrorist band wants killed, even though if we do not do so they will kill two hostages in their custody.66 Williams would say that this is because that would force an action alien to us (killing an innocent) into the set of actions that define who we are and what projects speak for us. In other words, here we should not calculate consequences to decide what it is right to do.
Foot’s earlier discussed torture example is also explainable on this ground. To alter it slightly (to excise its omission aspect), we may send one to be tortured by another if by so doing we will distract the torturer enough to allow five to escape torture; we may not torture one in our custody to prevent a terrorist from torturing five in his custody (he has threatened to do just that unless we torture one). The former act does not make torture our project so we may justify it by the good consequences we achieve; the latter act would make torture part of our project and, by Williams’s distinction, cannot be justified by the equivalently good consequences of saving five from torture.
In a related context the criminal law of Great Britain, like that of many American states, declines to allow the defence of duress when the charge is murder. The thought behind this limitation is straight out of absolutist morality: the intentional killing of an innocent is so awful that no threatened consequences can justify or excuse anyone in doing it.67 Yet for a time the English courts held that, despite the assimilation of accomplice liability to the liability of a principal, accomplices to murder may avail themselves of the duress defence.68 For an accomplice only makes possible by his (p.698) action the killing by another; he does not himself kill. He may only drive the killer to the site of their crime, for example. In such a case, his act may be justified or excused by threats of death to himself and his family, even if an actual killing by him could not be.
(vii) The causing/doing distinction
Tom Nagel once argued that ‘what absolutism forbids is doing certain things to people, rather than bringing about certain results’.69 As Nagel elsewhere puts his distinction, it is ‘between what one does to people and what merely happens to them as a result of what one does’.70 Thus, to elaborate on one of Nagel’s examples, if we ourselves bomb a village we do certain things to the villages which morality absolutely forbids, namely, we slaughter and maim innocent civilians;71 such doings cannot be justified by the good consequences they may bring, such as shortening the war. On the other hand, those who manufacture the bomb casings eventually used by those who bomb civilians only bring about the result that they are slaughtered and maimed; presumably, because the death of the villagers is ‘what merely happens to them as a result’ of what the bomb casing manufacturer does, the manufacturer of such bomb casings can be justified by good consequences such as a shortening of the war and a lesser loss of life.72
b. Reassessing These Limits on Consequential Calculation
These seven distinctions, as thus far presented, may seem a heterogeneous lot. What, if anything, do they have in common? Why should they have anything to do with drawing the line between permissible and impermissible consequentialist justification? We may answer both questions by penetrating the surface features of these seven distinctions to get at their common core.
As I argued in chapter 9, there are three conditions for the full expression of our moral agency in a way that renders us morally responsible for some bad state of affairs. First, we must act. Only (p.699) by our voluntary bodily motions do we even begin to put our stamp upon the world. Second, we must desire or intend (or perhaps even believe with certainty) that our actions will have those causal or other properties that will make them wrongful. And third, the bodily movements that we will must themselves cause the bad state of affairs necessary for there to be wrongdoing.73
If we re-examine the seven suggested limits on consequential calculation, we can see that they each instantiate one or another of these three conditions of responsibility. Take Foot’s distinction of allowing (where good consequences count) versus doing (where they do not). Despite Foot’s claim that ‘there is no … general correlation between omission and allowing, commission and bringing about or doing’,74 the heart of Foot’s distinction is the omission/commission distinction. Foot gives several reasons to think otherwise. The first is the Ordinary language’ observation that our use of the word ‘allowing’ is not always idiomatic if we align it with ‘omitting’. This is true but irrelevant. It is true because it is a fact about the English language that verbs may be used in the active (rather than the passive) voice and yet describe omissions. It is irrelevant because it is the act/omission distinction that marks an important moral difference, not the active/passive voice distinction.
This gets us to Foot’s second reason, which is moral and not conceptual. Foot must think that the most stringent duties of morality enjoin us not only not to harm others but, sometimes, positively to help others. Thus, when we omit to give food to a starving beggar because we want his organs for medical research, ‘we are inclined to see this as a violation of a negative rather than positive duty’.75 That surely is not right.76 Although we do wrong in refusing the food to Foot’s beggar when we are motivated by the hope of harvesting his organs, surely we do much worse in poisoning him for his organs. Our more stringent negative duty is not to kill him; it does not include the positive injunction to save him.
(p.700) Foot’s more recent reason for refusing to collapse the doing/allowing distinction into the action/inaction distinction is the ease with which the latter distinction can be manipulated by us to give us only the less stringent duty. Foot’s example: ‘it would be possible to change the moral character of certain trains of events by such simple expedients as building respirators which needed to be turned on each day’.77 Yet the building of such Omission machines’ is itself an action done with culpable intent, so that if we now fail to turn on the respirator we will have earlier violated a negative duty not to kill (including not to kill indirectly by creating a machine that will kill for us unless we stop it). So Foot’s allowing/doing distinction, properly construed, is nothing more than the familiar omission/action distinction, and thus is directed towards the first condition, that of voluntary action.78
It takes no similar reconstruction to see that the distinctions between purposeful and knowing action, and between knowing and risk-taking action, instantiate the second condition of fully expressing our moral agency. As I argued in chapter 9, we are more culpable when we seek to achieve some bad state of affairs than when we do not aim at it but nonetheless know it will come about as a result of our actions; and we are more culpable when we know our acts will produce such immoral results than when we only do an act aware that there is a risk of such results coming about because of what we do. We should see these two distinctions as competing attempts to locate the most significant break points in culpability, which competition need not be resolved to make our general point here: it is the culpability of the actor that here makes the difference whether consequential justification is permissible or not.
The remaining four distinctions each has to do with the third condition of responsible agency, causation. Take the victim in-peril/not-in-peril distinction. This I take to be proxy for a crude cause-in-fact judgement: when the victim is in peril, in the sense that he was going to die anyway, the temptation is strong to say we have not really (in fact) caused his death. His death in such cases (p.701) was overdetermined in the sense that the same kind of death would have occurred anyway, only at a somewhat later time. I say this is a crude cause-in-fact judgement because literally one does cause another’s death even when the victim would have suffered the same kind of death had the actor not acted. Yet our intuition remains that one has not very much altered the world from the way it would have been in any event when the harm one causes was about to be visited on the victim anyway, without one’s action. In such cases we may have the sense that our responsibility is diminished because our causal responsibility is more in question.79
The fifth distinction that Judith Thomson has emphasized, between redirecting a pre-existing threat and creating a new threat, also gains its plausibility from some causal intuitions. Thomson herself finds her proposed ‘distribution exemption’ puzzling: ‘I do not find it clear why there should be an exemption for, and only for, making a burden which is descending upon five descend, instead, onto one. That there is seems to me very plausible, however.’80 My suggestion is that the distribution exemption has as its rationale the idea that we only are a ‘small cause’ when we redirect a pre-existing threat not of our making; that, as such, our causal responsibility (and thus moral responsibility) is less than when we create a new threat and thus ‘largely cause’ the harm ourselves.81
The sixth distinction introduced by Bernard Williams, between our own and others’ projects, is quite clearly built on causal notions. In their justly celebrated book, Causation in the Law,82 Hart and Honoré go to some length to describe one dominant notion of proximate causation in both law and ordinary thought. (p.702) In the law we call this the ‘direct cause’ conception of proximate causation. A direct cause of some harm is any cause in fact where there is no intervening cause. An intervening cause either is that abnormal conjunction of natural events we think of as a coincidence, or it is a free, informed, deliberate, voluntary action of a third party that intervenes between the defendant’s action and the harm.
The moral force of Williams’s ‘my project/someone else’s project’ distinction stems entirely from the causal intuition Hart and Honoré described. Specifically, Hart and Honoré tell us, our causal responsibility ends where some third party’s voluntary action intervenes between our act and the harm our act in fact causes. This is why I may do an act that results in another being tortured by someone else, while I may not torture someone else myself to achieve the same good consequences. This is why I may step aside and thereby make possible the shooting of another by a terrorist band, while I may not shoot the one myself to achieve the same good consequences. Where I am causally responsible for the moral wrong—the second of each of these pairs of examples—I may not justify myself with good consequences, whereas where I am not causally responsible I may, for I am not nearly so responsible.
The seventh distinction Tom Nagel once emphasized, that between causing and doing, also derives whatever moral force it possesses from a distinction in our causal responsibility.83 To begin with, there is a certain moral arbitrariness in the distinction as Nagel stated it, for it is an etymological accident whether the English language happens to have a ‘causally-complex’84 action verb to cover some situation we are judging. If it does (‘killing’, ‘torturing’, ‘maiming’, etc.), then we at least can classify the event as a doing; if it does not (‘avalanching’, ‘house-collapsing’, etc.), then what we did can only be classified as a causing because there is no doing to have been done. That causing death is a frequent enough occurrence that our language has a verb to cover it (‘kill’), but that causing damage by avalanche is not, is a morally arbitrary difference.
(p.703) Nagel’s linguistic distinction therefore has to stand proxy for some less morally arbitrary distinction. Although Nagel himself preferred to keep his distinction ‘unanalyzed’,85 Jonathan Bennett86 once put his finger on the relevant feature. We call a causing-death a killing only when there is a certain immediacy between what we do and another’s death. We see this feature of our English usage best in the present-perfect tense, as in ‘I am now killing him’. If we are shooting someone, this is an appropriate description; if we are loading the gun before we find our intended victim, this is not the appropriate description. In the latter situation we are only preparing to kill him. If later on we do find and shoot him, only the act of shooting but not the act of loading was an act of killing, even though both were acts causally necessary for the victim to die.
Immediacy of causation matters morally for the reasons lawyers have explored with their concept of proximate causation. Even without the existence of an intervening cause (in Hart and Honoré’s sense), spatial/temporal remoteness between our act and a harm matters to our causal responsibility. Caesar’s crossing the Rubicon may well be a necessary condition of some future nuclear accident, but Caesar’s causal responsibility does not extend so far. Such diminished causal responsibility also betokens a lessened moral responsibility for the remote harms we in fact cause.
c. Ruminations on the Rationality of Agent-Relative Views of Morality
If these reductions are right, then our view of the domain of things to which moral norms apply is altered somewhat. Such norms categorically forbid intentionally done (or perhaps intended) complex actions like killing or torturing. Put in other words such norms prohibit our doing those voluntary acts that directly and proximately cause states of another’s being killed or being tortured, when such acts are accompanied by an intent (or perhaps foresight) that such states will result.87
(p.704) By reducing a seemingly heterogeneous range of intuitive distinctions to the three distinctions that centrally matter to an agent’s prima facie moral responsibility, what besides clarity have we accomplished? At a minimum, we have shown a kind of coherence to these intuitions, together with a consistency between them and the general idea of agent-relative theories of morality. The coherence point is probably obvious enough: if I am correct, the intuitions other philosophers have captured with their examples and distinctions are not ad hoc, isolated instances where consequential calculations are permissible even on agent-relative views. Rather, the general principle behind all of them is the principle that blame-worthiness matters in determining when we may justify our acts by their good consequences: when a contemplated act would make us fully culpable were we to do it, we are forbidden to do it, irrespective of the good consequences we could achieve by doing it; when that act would not make us as fully blameworthy—because of the absence or diminishment of a voluntary act, culpable mental state, or causal responsibility—then we may do it, using those same good consequences as our justification.88
I sometimes think this to be a profound insight, sometimes a trivial one. In the former moments I am struck by the very systematic coherence of the distinctions above ordered around the three conditions of prima facie responsibility. In the latter moments I am struck by the thought, ‘What else would you expect of agent-relative views of morality?’ For such views interpret morality to say to each of us, ‘Don’t you intentionally do a morally prohibited action.’ Such views are by their nature focused on the moral agency of each of us, so how could one be surprised that when the conditions of that agency are fully realized, one’s intuitive sense that good consequences are irrelevant is strongest? (This latter insight is what I mean when I say that these intuitions are consistent with agent-relative views.)
(p.705) Seeing both the coherence and the consistency of our principles for ruling out consequential calculation raises a troubling question about the rationality of agent-relative theories of morality. Why should each of us act so as to preserve his or her moral innocence? Why is our own moral blameworthiness so important that it is morally forbidden to us to intentionally kill an innocent who was not in danger, if by so doing we can prevent the intentional killing of five others by someone else? I do not know a very good answer to this question, although there are quite a few stabs at it in moral philosophy.89 I do know that I share Sam Scheffler’s response to his own inability to find an intelligible answer to this question: ‘I find myself in certain moods more impressed by the particular intuitions to which the [agent-relative] restrictions respond than by the general difficulties encountered in attempting to motivate them …’.90 I thus shall leave for another day this very general problem in moral philosophy and assume that good consequences cannot justify our voluntarily acting in such a way that we intentionally cause certain states of affairs.
d. The Limits on Consequential Calculation Applied to Torture
Assuming the correctness of the complex agent-relative view just described, how do we judge the torture of innocents by GSS interrogators when such torture is conducted, for example, in order to locate a hidden bomb? The answer is surprisingly straightforward: there is nothing in any of the seven distinctions we have examined leaving any room for consequentialist justification in such a case. Torture is a voluntary act, not an omission to prevent pain; torture in the circumstances imagined is fully intentional—not just foreseen, but intended as a means to some further end; the victim of GSS torture is known, he is not a member of some larger class at (p.706) risk for torturous treatment; such a victim is not already in peril of harms like torture, so that it is only the GSS that creates such a peril; there is usually no pre-existing threat of someone being tortured, with the only choice of the GSS being as to who will suffer it; there is no voluntary act by someone else that intervenes between the acts of the GSS interrogator and the pain of his victim, nor is there any spatial/temporal remoteness. In short, all of the conditions of prima facie blameworthiness are as fully met by the GSS interrogator as by the hypothetical surgeon about to harvest the organs of one to save five. The moral answer to both must be the same: you cannot torture or kill the innocent, even to achieve what are admittedly good consequences like saving five lives.
There is thus nothing in this first aspect of complex agent-relative moral views that licenses torture by the GSS. I now turn to the other two complexities that characterize this moral view to see if the same conclusion holds.
2. Complex Norms with Exceptions versus Exceptionless Absolutes
a. The General Need for Exceptions to Moral Norms
There are very few (if any) plausible candidates for the kind of short and snappy injunctions that can be fitted on one stone tablet. Thou shalt not kill’ is a poor candidate for such a norm, because to be plausible it has to have exceptions built into it. Not telling lies, keeping our promises, or not committing adultery, are even less plausible candidates for exceptionless absolutes. What would we think of a person who could save his family from wrongful killing by another but refused to do so because the only means available involved: killing the aggressor, telling a lie to the aggressor, or not keeping a promise to the aggressor? Despite Kant’s famous views to the contrary, I would think that such a person not only lacks virtue, but that he is derelict in his moral duties.
The point is that moral norms must have exceptions implicit in them to have any moral plausibility. To be morally plausible, ‘thou shalt not kill’ must be taken to be an elliptical reference to a much more complicated norm: ‘Don’t kill, unless in self-defence, to protect your family, to aid in a just war lawfully declared, to execute those deserving of the death penalty such as Adolf Eichmann etc. Outside of a text-based view found only in religious or conventionalist views of morality, there is no reason to suppose that (p.707) moral truths are simple. Indeed, as science shows us about the world of non-moral fact, there is every reason to think that moral laws are quite complex—as complex as the scientific laws that govern the behaviour of physical objects.91 No one expects all of science to be statable in ten rules of very short compass.
So the second complexity that must be introduced into any plausible agent-relative view is the possibility that norms like ‘never torture’ have implicit exceptions clauses to them. It may seem that such ‘possibility of an indefinitely large number of exceptions’ renders this into a consequentialist view of morality,92 but this would be a mistake. We must distinguish exceptions to norms—exceptions which are simply a further filling out of the content of norms—from overriding norms on occasion even when we admit that they are by their terms applicable.93 The distinction is as familiar as the common law lawyers distinction between distinguishing a precedent case’s holding (by showing how an unstated exception to the holding exempts the present case from its true reach) and overruling a precedent case’s holding (where we admit its applicability but refuse to follow it in the case at hand).
With this distinction in mind, we can see that consequentialism is the view that we should override moral norms like, ‘don’t torture’, whenever the balance of good over bad consequences say we should. A non-consequentialist view of the complex kind here considered does not allow such overridings of moral norms; it does allow moral norms to be quite complex in their content because of its allowance of exceptions to norms like, ‘don’t torture’. The only way the complex agent-relative view would collapse into a consequentialist view of morality would be if one counted as ‘exceptions’ items like, ‘unless good consequences dictate otherwise’. Such ‘exceptions’ would of course be nothing more than an invitation for a general consequentialist override of the norm in question.
(p.708) b. Possible Exceptions to the Norm, ‘Never Torture’
(i) Consent of the victim
As an illustration of how exceptions to norms are not consequentialist overridings of norms, consider the situation in which the victim consents to what otherwise would be a violation of his rights. Often such consent operates to transform a prima facie wrong action into a rightful action.94 For example, rape, where consent radically alters the nature of the action; or trespass, where consent creates a property interest (a ‘licence’) where otherwise there was a wrong. In such cases we do not reach the question of consequential justification overriding some moral norm because the norm itself—when fully spelled out so as to include an exception for consent—was not violated.
Consent is often not a defence to serious crimes such as homicide. This may, however, be for reasons unconnected to the moral rightness or wrongness of the defendant’s action. Euthanasia, for example, is legally defined as murder just about everywhere, but not because the consent of the victim makes no moral difference; rather, there are utilitarian worries that if there were a defence of consent, many unconsented-to homicides might go unpunished because the only surviving witness—the defendant—can lie about the victim’s consent with impunity.95
Even those who do not share my view that in certain circumstances euthanasia is not wrong at all may think that it is not as wrong as an unconsented-to homicide. Consider again the lifeboat cases. In Holmes, where some of the passengers had to be jettisoned if any were to live, would it be wrong to release someone (who is in some way restrained) in order that he might voluntarily jump overboard and save the boat? To push him in, if he is paralysed and requests such an action because he cannot do it himself? If either of these actions would be wrong at all, surely they are not as wrong as drowning one who does not volunteer.
What such situations show is the possibility that exceptions to moral norms need not make an all-or-nothing difference in the wrongness of an action to which they apply. Consent of the victim, for example, might lessen but not eliminate the wrongness of doing (p.709) the action to which consent was given. If this is so, then notice how plausible it is to say that consequential justifications may be given in those exceptional instances of killing, torturing, etc., that are not very wrong. Put in terms of the crowded lifeboat situation: even if it would be wrong to aid another to kill himself when in full possession of his faculties and for understandable reasons he wishes to do so (the general prohibition against euthanasia), it is not so wrong that it cannot be balanced against, and outweighed by, the saving of the lives of those remaining in the lifeboat.
Since terrorist suspects never consent to be tortured, the exception we are here examining needs to be broadened considerably if it is to be of any use to us here beyond a general illustration of how exceptions to moral norms operate in this complex agent-relative view of morality. One suggestion arising from the lifeboat cases is that if you consent to a fair procedure for determining who must suffer a harm, you cannot complain when that procedure determines that it will be you who suffer the harm.96 This is based on an assumption of the risk intuition: one who voluntarily assumes a risk of which he is fully aware has less cause to complain when the risk materializes than someone who did not so assume it.
Such an intuition has no limitation to assumption of the risk of selection procedures in situations of common peril. Consider in this regard Judy Thomson’s example:
Suppose an avalanche is descending toward a large city. It is possible to deflect it onto a small one. May we? Not if the following is the case. Large City is in avalanche country—the risk of an avalanche is very high there. The founders of Large City were warned of this risk when they built there, and all settlers in it were warned of it before settling there. But lots and lots of people did accept the risk and settle there, because of the beauty of the countryside and the money to be made there. Small City, however, is not in avalanche country—it’s flat for miles around; and settlers in Small City settled for a less lovely city, and less money, precisely because they did not wish to run the risk of being overrun by an avalanche. Here it seems plain we may not deflect the avalanche onto Small City to save Large (p.710) City: the Small Cityers have more claim against it than the Large Cityers do.97
What is doing the moral work in cases like Thomson’s, as in the lifeboat-lottery cases, is not the bare fact of consent to run a certain type of risk.98 Consent, unlike a promise, creates only very weak content-independent reasons for action in morality.99 What makes it not wrong to prefer Small City to Large City is the unjust enrichment that would take place if such preference were not made. Where there is a benefit attached to running a certain risk, and where some opt for the benefit and the risk while others eschew both, it would unjustly enrich those who have benefited if the cost of their benefit (the risk) were transferred to someone else.
If we apply this discussion to torture, there is perhaps some weak argument that terrorists assume the risk of harsh treatment if caught. For if there is an established policy of so treating terrorists, and if it is publicized so that it is known, then those who seek (what they regard as) the benefits of terrorist activities cannot complain if the risks attendant upon obtaining those benefits materialize.
There are two things wrong with this argument. One stems from my doubts whether even explicit victim-consent could make torture permissible. The reason that victim-consent eliminates or lessens the wrongness of various actions harming the victim is because such consent shows there to have been no harm to this person (even if an identical state of affairs would be harmful to others). As Tom Nagel describes the moral force of victim-consent:
[T]here can be complications [to a deontological restraint]. One is the possibility of someone volunteering to be subjected to some kind of pain or damage, either for his own good or for some other end which is important to him. In that case the particular evil that you aim at is swallowed up in the larger aim for deontological purposes. So the evil at which we are constrained not to aim is our victim’s evil, rather than just a particular bad (p.711) thing, and each individual has considerable authority in defining what will count as harming him for the purpose of this restriction.100
My doubt is whether torture would not be wrong even when consented to (presumably by a masochistic victim), for torture harms both torturer and victim no matter what the latter thinks about it.
The other thing wrong with the assumption of the risk argument is that it is far too sweeping in its implications to be acceptable. Such an argument could be made in support of any government policy, no matter how horrible, so long as it was well-publicized and regularly applied.101 The reason this argument may seem persuasive is due, not to any assumption of the risk by terrorists, but to their blameworthiness as terrorists. It is thus to the possible exceptions relating to this aspect of the situation to which I now turn.
(ii) Self-defence and defence of others analogies
One thing the assumption of the risk discussion highlights is the difference it seems to make whether it is a terrorist or an innocent who is tortured. Other possible exceptions suggest themselves to account for this intuitive difference in terms of the differential responsibility of terrorists versus non-terrorists. The most obvious place to start is with the well-established exceptions to norms like ‘don’t kill’ that exist for self-defence and defence of others situations. This right to resist aggression has often formed the basis for drawing distinctions between combatants and innocents in discussions of the morality of war,102 and I shall use it for the same purpose.
We should deal at the outset with the possibility that resisting aggression is not an exception to norms against killing and torture, but is only an excuse for violating such norms. Such a possibility may occur to us because self-defence situations often happen quickly, bring our self-preservative instincts into action (instincts that we know are difficult to resist), and such situations are without criminal liability even if there are many culpable aggressors attacking us (so there is no net savings of lives to justify our actions) and/or the aggressors are themselves morally innocent (p.712) actors.103 Yet, self-defence and defence of others are not excuses. We do no wrong when we kill or injure an aggressor, and we are not culpable when we intend to kill or injure an aggressor, and thus we need not excuse ourselves from anything.
Intuitively, we see this justificatory (rather than excusing) nature of self-defence with the following sort of example. Imagine someone (‘D’) with bad character who enjoys inflicting suffering on others. Through no fault of his own he is attacked in his home by a mugger (‘V’), who demands his wallet. D refuses to give up his wallet, whereupon V tries to kill him. D defends himself by killing V. It was not a hard choice for D. He neither agonized nor was he at all fearful, being confident of his abilities to defend himself. Indeed, he welcomed the opportunity (as he put it later) to ‘get in a free killing’ of someone, enjoying every thrust of his knife into V’s body. D is a pretty dreadful person and does not possess any of the diminished capacities or opportunities for choice we think of as prerequisite for excuse. Yet D is entitled to the defence of self-defence because he did no wrong in killing V.
How one fleshes out this intuition that self-defence is a justification has been the subject of some debate in criminal law theory and philosophy.104 There are two sorts of accounts of the moral basis of self-defence as a justification, victim-based accounts and agent-based accounts.105 The victim-based accounts focus on the culpability of the victim of the homicide (who is culpable because he begins as the aggressor even though he ends up as the victim). Such culpability is variously said to forfeit, override, or specify out of existence the aggressor’s right to life106 or to discount the value of (p.713) his life, either entirely or in part. One problem with such victim-based accounts is that they do not seem to square very well with the ‘innocent aggressor’ cases,107 for it is difficult to think that those who innocently pose a threat to us forfeit their right to life or have the value of their life discounted. Another problem with such accounts is that they leave unexplained why, when the aggression is ended, the aggressor’s life regains its value or why his right to life is re-established. Such problems motivate the alternative, agent-based account, which speaks of a defendant’s right to resist aggression.108 Yet these latter accounts also are problematic in that they do not go very far—for an explanation of why it is not wrongful to kill an aggressor in terms of a right to kill an aggressor comes perilously close to tautology. Wanted is some more general account of why we each have such a right.
If we attend to our sense of fittingness when an aggressor trying to kill someone is instead killed himself, we may gain some insight into the moral basis for self-defence. Notice that the aggressor creates a situation in which someone must be killed, either he or his intended victim. He has wrongfully created a threat of harm that his intended victim can now only redirect but not eliminate. Since he is the one creating such a threat, he in all fairness is the one to be selected when someone has to bear the harm threatened.109
This account helps to explain the innocent aggressor cases. Aggressors may be innocent for a variety of reasons.110 They may be trying to kill you or others, but are morally and legally excused (insane, very young, under duress of others, etc.); or they may not even be trying to kill you but are nonetheless physically part of a threat that is trying to kill you. Consider Sandy Kadish’s illustration of the first of these two classes of cases:
Suppose a terrorist and her insane husband and 8-year-old son are operating a machine gun emplacement from a flat in an apartment building. (p.714) They are about to shoot down a member of the diplomatic corps, whose headquarters the terrorist band is attacking. His only chance is to throw a hand grenade (which he earlier picked up from a fallen terrorist) through his assailant’s window. Probably under Anglo-American law he will be legally justified in doing so. His right to resist the aggressor’s threat is determinative.111
The relatively innocent participants in this scenario—the insane husband and the 8-year-old child—are permissibly killed because they too are causers—non-culpable causers, but causers nonetheless—of the situation that demands someone die. They too are thus most fairly selected as the recipients of the harm whose threat they have set in motion.
One can perhaps extend this analysis to the second category of innocent aggressor cases. Consider the situation in which a terrorist (T’) takes on innocent bystander (‘V’) as a shield to use as T seeks to kill another (‘D’). If D’s only way to defend himself against T is to shoot through V, it is common to suppose, morally and legally, that he may do so.112 The reason for this, I would think, lies in the Ownership’ of the risk (of T’s violence) that V, through no fault of his own, has acquired. T’s taking V as a shield was V’s misfortune; V did nothing to deserve it but it’s his nonetheless, like disease or a natural catastrophe. D is entitled to resist the shifting of that risk from V to himself, not because V caused the risk, but because it’s his risk nonetheless.113
The application of this account of the moral basis of self-defence to GSS torture of terrorist suspects should by now be obvious enough. Terrorists who are captured do not now present a threat of using deadly force against their captors or others. Thus, the literal law of self-defence is not available to justify their torture. But the principle uncovered as the moral basis for the defence may be applicable. For if the terrorist knows the location of hidden bombs, or of buried hostages, or of caches of arms, or if he knows of future terrorist acts to be executed by others that he has aided, he has culpably caused the situation where someone must get hurt. If hurting (p.715) him is the only means to prevent the death or injury of others put at risk by his actions, such torture should be permissible, and on the same basis that self-defence is permissible.
Torturing the terrorist in order to get him to reveal information that will prevent the very harm he has set in motion may be likened to the following scenario. Suppose another wishes to kill you. At an isolated location where no medical attention is available he puts poison in your food, which you eat; seeing you swallow the last bite, he then tells you what he has done so that you will know that you are dying; he also swallows the only supply of the antidote to the poison you have taken. If you can catch him, may you cut him open to extract the antidote in order to save yourself, even though you know such cutting will certainly kill him?
My own answer to this question is a rather clear ‘yes’. True enough, his attack on you is over, if we construe ‘attack’ as an action and not as its continuing effects; so the literal law of self-defence may well not be available. Yet the poisoner has culpably caused the situation where someone must die. He is, to my mind, by far the best candidate to suffer the harm that someone must suffer.
Such an exception to the norm against torturing has of course no application to the torturing of the innocent, even when, as in Scheffler’s earlier example,114 such torture (of the terrorist’s child) is the only way to induce the terrorist to reveal the location of a deadly bomb. For the child of the terrorist did not cause, nor is he himself part of, the threat created by the terrorist activity. As Robert Nozick observes, ‘some uses of force on people to get at an aggressor do not act upon innocent shields of threats; for example an aggressor’s innocent child who is tortured in order to get the aggressor to stop wasn’t shielding the parent’.115
Less clearly within or without this exception are instances of torture of someone who had no hand in the terrorist activity that gave rise to the threat but who nonetheless possesses information that, if disclosed, would prevent the threatened harm from occurring. For example, the spouse of the terrorist who has planted a bomb had no hand in the activity that led to the bomb’s planting, but she has been told by her husband where the bomb is located, and she refuses to disclose its location. May she be tortured for the information?
(p.716) One avenue that suggests that she might would be to consider her a part of the threat posed by her husband’s activities because she fed, clothed, and supported him during the time of those activities. Yet, I would reject such a suggestion. It is like treating spouses of soldiers as combatants for purpose of drawing the line between morally permissible versus morally impermissible modes of warfare.116 However that difficult line is drawn, surely there are some civilians (non-combatants) who may not be killed wherever and whenever they can be found, and for that to be the case spouses of combatants cannot themselves be considered combatants.
Nonetheless, my hypothetical spouse is unlike other civilians in that she could prevent the harm by no greater an act than speaking up. If we repair to our self-defence analogy, perhaps again this will focus our intuitions on this aspect of the hypothesized situation. Consider the following example of Jeffrie Murphy, what he calls ‘the case of the homicidal diabetic’:
He is chasing you through the woods of an enclosed game preserve, trying to kill you for sport with a pistol. However, because of his medical condition, he must return to a cabin in the middle of the preserve every hour in order that his aged mother can give him an insulin shot. Without it, he will take ill and die and will thus be forced to abandon his attempt to kill you. Even if blocking that insulin shot seems your only hope, killing the mother in order to do it would be a very doubtful case of self-defense.117
Yet is it so clear that we may not torture the mother in order to get her to reveal the insulin supply, or kill her, if she is told what her son is trying to do, told how she can prevent it, and she refuses? Does she not, by that refusal, become a part of her son’s threat against which we are entitled to defend ourselves?
Consider a variation on Lon Fuller’s famous trapped spelunkean explorers.118 The explorers, it will be recalled, are trapped in a blocked cave without hope of rescue before they starve to death. They debate whether they may kill one of their number in order to save the rest by feeding on his body. The variation is: in the midst of this debate appears a cave dwelling hermit who knows another route of egress from the cavern; because he despises intruding (p.717) spelunkers, and to deter further intrusions, he refuses to reveal the route to safety. If someone has to be eaten to save the rest, would not such a Bad Samaritan be the best candidate? Short of eating him, would not torturing him to extract information he ought to reveal anyway be preferable? Troubling as they are, I think the answers to these questions to be ‘yes’. The Bad Samaritan who could prevent anyone dying and who refuses to do so for no good reason119 becomes part of the threat to be defended against, and should be treated accordingly.
(iii) Prepaid punishment of the generally culpable
I now wish to return to the victim-based accounts of self-defence. I do this not because they are good accounts of self-defence, for they are not. Rather, such accounts allow us to glimpse a third possible exception to norms like ‘don’t kill’ in addition to the exception we have just examined that encompasses the right of self-defence.
There is a common moral intuition that culpable wrongdoing deserves punishment. This is the retributive urge that we examined in Part I. Retributivism asserts that it is not wrong to punish the guilty and, indeed, it is the state’s duty to punish those who deserve it. Deserved punishment, in other words, is for a retributivist one of the exceptions to norms like ‘don’t restrain’ or even, perhaps, ‘don’t kill’.
There are two obvious problems with extending this exception to torturing culpable terrorists on some theory that it is only a ‘prepaid punishment plan’: one, torture is not a morally permissible form of punishment; two, GSS interrogators are not courts and are not in a position to make the findings of guilt that only courts can make. Both these considerations combine to rule out any blanket exception in favour of torturing the quite blameworthy. (Otherwise, it would be morally permissible to torture terrorist suspects even when there is no life-saving information to be extracted from them.)
Yet remembering the possibility that not all exceptions to moral norms operate in an all-or-nothing fashion, perhaps it is less wrong (p.718) to torture the blameworthy than the innocent. Put crudely, their lives and rights are worth less precisely because they are morally odious individuals deserving of punishment, even if we are not their rightful punishers and even if death or torture is not their rightful punishment.
Consider the provocation defence to murder in Anglo-American criminal law to illustrate this possibility.120 There is what I would call an unofficial version of the defence, unofficial in the sense that it is not the stated reason for the defence but it is, often enough, the reason that guides jurors in their application of the defence to the cases they must decide. This unofficial rationale comes out in the following kind of case: D is the husband of W. W was forcibly raped by her gynaecologist (‘V’) in V’s office on one of W’s regular visits to him. With no witnesses, W was unable to get the authorities to prosecute V for rape. Her deep distress over the incident leads to her suicide. D, in full possession of his faculties, and after giving the matter considerable thought over a period of several months after W’s death, purchases a gun, walks into V’s office, and shoots him dead with every bullet in the gun. D then waits for the police in V’s office, as he intended to do all along.
Despite the lack of any ‘passion overbearing reason’—the official, excuse rationale for the defence—the jury found for the defendant on his provocation defence, resulting only in a manslaughter, not a murder, conviction. It is not hard to reconstruct their reasoning: V was a very blameworthy human being who pretty much deserved what he got. True, there is no death penalty for rape—D’s punishment was excessive. True, D was not the person to impose punishment of any kind. But D’s action was much less wrong than it would have been had V not been the moral leper that he was.
If one agrees with the jury’s reasoning in this actual case, as I do, then one glimpses the following possibility: while the killing or torture of another is wrong, where it is the killing or torture of someone deserving of serious punishment, it is significantly less wrong, (p.719) so much so that good consequences may justify the doing of it. Thus, in situations inviting a balance of evils, one may justifiably select those deserving of punishment anyway to bear the evil someone must bear. If you have only enough dialysis machines for ten, but twenty need them, you may exclude the morally odious first;121 if the lifeboat is sinking, and some must go to save the rest, the known murderers amongst the passengers are good candidates for being the first to be tossed overboard.
Saying this is not to justify executing or torturing to death those who have culpably caused the death of innocents, as the GSS (apparently) did to those terrorists who took the No. 300 bus in 1984. But had those terrorists possessed information about the location of another bus full of civilians about to be destroyed by another terrorist band, and if the only way to prevent that loss of innocent life was to torture the captured, culpable terrorists, then such torture or death is morally permissible. The act is much less wrong because done to culpable terrorists, and the good consequences achieved may justify doing that lesser wrong.
3. Non-Absolute Moral Norms: Threshold Deontology
Apart from the exceptions that the content of moral norms must have for them to be plausible, a third modification of absolutism is the softening of the ‘whatever the consequences’ aspect mentioned earlier. This aspect of absolutism is often attributed to Kant, who held that though the heavens may fall, justice must be done. Despite my non-consequentialist views on morality, I cannot accept the Kantian line. It just is not true that one should allow a nuclear war rather than killing or torturing an innocent person. It is not even true that one should allow the destruction of a sizable city by a terrorist nuclear device rather than kill or torture an innocent person. To prevent such extraordinary harms extreme actions seem to me to be justified.
There is a story in the Talmudic sources that may appear to appeal to a contrary intuition.122 It is said that where the city is surrounded and threatened with destruction if it does not send out one of its inhabitants to be killed, it is better that the whole city should perish rather than become an accomplice to the killing of one of its (p.720) inhabitants. Benjamin Cardozo expressed the same intuition in rejecting the idea that those in a lifeboat about to sink and drown may jettison enough of their number to allow the remainder to stay afloat. As Cardozo put it:
Where two or more are overtaken by a common disaster, there is no right on the part of one to save the lives of some by the killing of another. There is no rule of human jettison. Men there will often be who, when told that their going will be the salvation of the remnant, will choose the nobler part and make the plunge into the waters. In that supreme moment the darkness for them will be illumined by the thought that those behind will ride to safety. If none of such mold are found aboard the boat, or too few to save the others, the human freight must be left to meet the chances of the waters.123
There is admittedly a nobility when those who are threatened with destruction choose on their own to suffer that destruction rather than participate in a prima facie immoral act. But what happens when we eliminate the choice of all concerned to sacrifice themselves? Alter the Talmudic example slightly by making it the ruler of the city who alone must decide whether to send one out in order to prevent destruction of the city. Or take the actual facts of the lifeboat case124 to which Cardozo was adverting, where it was a seaman who took charge of the sinking lifeboat and jettisoned enough of its passengers to save the rest. Or consider Bernard Williams’s example, where you come across a large group of villagers about to be shot by the army as an example to others, and you can save most of them if you will but shoot one; far from choosing to ‘sink or swim’ together, the villagers beg you to shoot one of their number so that the rest may be saved.125 In all such cases it no longer seems virtuous to refuse to do an act that you abhor. On the contrary, it seems a narcissistic preoccupation with your own Virtue’—that is, the ‘virtue’ you could have if the world were ideal and did not present you with such awful choices—if you choose to allow the greater number to perish. In such cases, I prefer Sartre’s version of the Orestes legend to the Talmud: the ruler should take the guilt upon himself rather than allow his people to perish.126 One should (p.721) feel guilty in such cases, but it is nobler to undertake such guilt than to shut one’s eyes to the horrendous consequences of not acting.
I thus have some sympathy for the Landau Commission’s conclusion that ‘actual torture … would perhaps be justified in order to uncover a bomb about to explode in a building full of people’.127 If one does not know which building is going to explode, one does not have the consent of all concerned to ‘sink or swim’ together. On the contrary, one suspects that like Williams’s villagers, the occupants of the building, if they knew of their danger, would choose that one of their number (to say nothing of one of the terrorist group) be tortured or die to prevent the loss of all. In any case, the GSS interrogator must choose for others who will pay the costs for his decision if he decides not to act, a cost he does not have to bear; this situation is thus more like my variation of the Talmudic example than the original.
Many think that the agent-relative view just sketched, allowing as it does consequences to override moral absolutes when those consequences are horrendous enough, collapses into a consequentialist morality after all. Glanville Williams, for example, in his discussion of the legal defence of necessity, recognizes the agent-relative view that ‘certain actions are right or wrong irrespective of their consequences’ and that ‘a good end never justifies bad means’.128 Williams nonetheless concludes that ‘in the last resort moral decisions must be made with reference to results’.129 Williams reaches this conclusion because, as Williams sees it, the agent-relative slogans just quoted reduce to the claim ‘that we ought to do what is right regardless of the consequences, as long as the consequences are not serious’.130
Contrary to Williams, there is no collapse of agent-relative views into consequentialism just because morality’s norms can be overridden by horrendous consequences.131 A consequentialist is committed by her moral theory to saying that torture of one person is justified whenever it is necessary to prevent the torture of two or more. The agent-relative view, even as here modified, is not committed to this proposition. To justify torturing one innocent person requires that there be horrendous consequences attached to not torturing that person—the destruction of an entire city, or, perhaps, (p.722) of a lifeboat or building full of people. On this view, in other words, there is a very high threshold of bad consequences that must be threatened before something as awful as torturing an innocent person can be justified. Almost all real-life decisions a GSS interrogator will face—and perhaps all decisions—will not reach that threshold of horrendous consequences justifying torture of the innocent. Short of such a threshold, the agent-relative view just sketched will operate as absolutely as absolutism in its ban on torturing the innocent.132
Despite this logical difference between agent-relative views of morality with a threshold, and consequentialist views of morality, there is of course a real danger that as a matter of psychological fact people will tend to confuse the two. As Bernard Williams has noted, it is very easy to slide from ‘the idea that there was nothing that was right whatever the consequences’—the non-absolutist but agent-relative idea I have defended here—to ‘the idea that everything depends on consequences’—the consequentialist view I reject.133 For the threshold of exceptionally awful consequences is bound to appear as a somewhat arbitrary line wherever it is drawn. By the familiar process of adding just one more straw to the camel’s back, people may thus push back the line further and further until the exceptional becomes the usual, and the balance of consequences always determines what it is right to do.134
(p.723) That this is a psychological danger is a reason not to tell people that there are thresholds of awfulness that justify prima facie immoral behaviour. It is not a reason to doubt that there are such moral thresholds for consequential calculation. It is the truth of this last proposition that governs the retrospective question with which I am here concerned. How one handles the psychological danger I pick up when I discuss the prospective question of new rules for GSS interrogations.
Even if threshold deontology does not collapse into consequentialism, many think such a view of morality to be arbitrary and irrational. Why should goodness of consequences not count at all and then, at some point, count enormously in the sense that it fully determines the rightness of action? What is overlooked by those convinced by this suggestion is that the threshold deontologist is not committed to denying that consequentialism is one moral principle among others. As Tom Nagel concedes, ‘there is one important component of ethics that is consequentialist and impersonal’.135 If our behaviour does not violate a moral norm like ‘don’t torture’ (and is not excepted by a moral permission like ‘you need not give away all of your money’), goodness of consequences does determine the rightness of actions.
Thus, for a threshold deontologist consequences always ‘count’. For behaviour violating deontological restraints, however, until the threshold is reached the principle that makes such consequences count—the consequentialist principle itself—is outweighed by other moral principles. As the consequences get more and more severe, the consequentialist principle becomes of greater weight as applied to this situation, until at some point (the threshold) the consequentialist principle outweighs competing principles of morality. Even before such a threshold was reached, consequences counted but were of insufficient weight to determine the rightness of actions. An analogy may help here: imagine water slowly rising behind a dam until eventually it spills over. There is nothing arbitrary about thinking both that there is no spillover until the threshold of the dam’s height is reached, and that each bit of water always counts in determining whether water will spill over the dam or not.136
If the rationality of a threshold-deontological view of morality is to be questioned, it will have to be on grounds other than these. (p.724) Perhaps the worry is not the apparent suddenness with which consequences seem to count for everything where before they counted for nothing; rather, the worry may be that any point we pick for a threshold beyond which consequences determine the rightness of action may seem arbitrary. Do we need there to be 500 people in a building about to be blown up by a bomb before we may torture an innocent to find the bomb’s location, or will 450 do?
Although this worry can surely give rise to quite genuine perplexity and anxiety when we make practical decisions, it is not the basis of any very powerful objection to threshold deontology as a moral theory. For this is no more than the medieval worry of how many stones make a heap. Our uncertainty whether it takes three, or four, or five, etc., does not justify us in thinking that there are no such things as heaps. Similarly, preventing the torture of two innocents does not justify my torturing one, but destruction of an entire city does. And if an actual case arises where I have to say whether preventing the death of 450 people in a building is sufficiently good to justify the torture or killing of one innocent, I would want to know more details as to what was done to the innocent to extract the information—for where the threshold lies depends in part on the degree of wrong done.
4. Torture and the Balance of Evils
This has been a long and complicated discussion of a complex agent-relative view of morality. If one accepts the morality defended here, a brief statement of its implications for torture is as follows. First, one of morality’s firmest norms is that torture is impermissible. We each have a moral duty not to torture another human being even when our doing so would prevent the torture (or other harm) to other human beings. Second, in the circumstances in which GSS interrogations took place, there is no construal of the limits of permissible consequential calculation that weakens or alters this absolutist conclusion. Whenever a state agency such as the GSS tortures, it does so through the intentional actions of its agents, to victims who are not in peril of torture anyway, and without the intervention of any third party or any complicated chain of natural causation. Third, the moral ban against torture applies less firmly to those who culpably cause the need for torture by planting the bomb that needs removal, etc.; such persons may be tortured when absolutely necessary to remove the threat that they have caused. Fourth, those who could remove the threatened (p.725) harm at little cost to themselves are also less wronged if they are tortured to induce them to do what morally they ought to do anyway. When sufficiently good consequences can only be obtained by the torture of such persons, morality does not forbid it. Fifth, nothing in the preceding points justifies torturing the innocent, no matter how necessary such torture may be to prevent bad consequences. Only the most horrendous of consequences could justify the torture of those who neither caused a threat of serious harm nor become a part of such threat by refusing to remove it. The likelihood of such horrendous consequences ever actually following from not torturing an innocent are so remote that no GSS interrogator is likely to have faced such a situation.
If I am correct about these moral conclusions, then they should also form the basis for legal conclusions about the criminal liability of GSS interrogators for their past acts of torture. For Israel’s balance-of-evils defence, like that of the Model Penal Code, is best construed to require just the kind of non-consequentialist moral reasoning that I have here attempted to display.
To actually apply these general moral conclusions to particular cases, much more would have to be said. First, there is the matter of ‘justificatory intent’.137 To avail himself of a balance-of-evils defence, a GSS interrogator must have tortured in order to prevent the harm whose threat justifies the torture. Use of torture as an investigative tool to aid in obtaining confessions for use in prosecuting terrorists does not meet this requirement, even if on some occasions such torture turns up life-saving information. Second, the GSS interrogator must have believed that the torture would in fact produce the kind of good consequences that alone can justify it. If life-saving information was being sought by torture, then there must have been some evidence on which the GSS interrogator based his beliefs: (1) that the terrorist had such information; (2) that the techniques adopted were likely to get him to divulge it; and (3) that the information would not have been rendered useless by the terrorist’s organization having taken preventive measures on learning of the capture of one of its members.
(p.726) Third, the GSS interrogator must have adopted the ‘least restrictive alternative’ in his interrogative methods. I have deliberately refrained from defining ‘torture’ throughout my discussion so that euphemisms like ‘pressure’ or ‘harsh treatment’ do not make the moral dilemmas seem easier than they are. Still, it should be plain that there is a world of difference morally between the slight tortures of sleep deprivation and the severe tortures of physical mutilation. In order to satisfy the balance-of-evils requirement that the torture has been necessary to save lives, the GSS interrogator must have believed that no less draconian interrogative technique would have been as effective as the one he employed.
Lastly, and in an abundance of caution, let me reiterate the obvious. Nothing said herein implies that torturing terrorists in retaliation for their terrorist acts is justified. Torture is an evil, and is only eligible to be justified when a greater evil is prevented. Torture for its own sake or for retaliation is not a balance of evils. It is only evil.
III. The Prospective Question: is New Legislation Needed?
A. What Behaviour of the General Security Service in Interrogating Terrorist Suspects Is Desirable?
Before deciding whether new legislation will better achieve some goal, one first must ask what the good is that such legislation should be trying to achieve. Specifically, what behaviour by its GSS interrogators should Israel desire for the future? To a large extent we have answered this prospective, legislative question by our answer to the retrospective, adjudicatory question. For open-ended provisions like Israel’s Penal Law section 22 invite the kind of all-things-considered value judgements needed by a legislator no less than a judge. For the future Israel should want its GSS interrogators to get the moral balance described in this chapter right as much of the time as is humanly possible. Israelis should want this because they seek a state whose behaviour and whose morality they can openly affirm. As the Landau Commission observed, ‘the methods of police interrogation which are employed in any given regime are a faithful mirror of the character of the entire regime …138 It is (p.727) difficult to be proud of a state whose police forces engage in ‘interrogative methods like those practiced in regimes which we abhor’.139 Thus, however one resolves the moral questions I explored in the preceding part of this chapter, one should employ that resolution here as well.
B. What Laws Will Produce That Behaviour?
The interesting question is how by law one may best achieve the morally right behaviour described before. The Landau Commission helpfully distinguished three general possibilities:140 (1) to regard the interrogation of terrorist suspects by the GSS as in a ‘twilight zone’ that is ‘outside the realm of law’, so that what levels of ‘pressure’ the GSS uses, and when, would be up to it alone; (2) to promulgate laws to guide GSS behaviour while not intending to enforce those laws by their terms when they are violated by the GSS; (3) to promulgate rules that are designed both to guide GSS behaviour and to be enforced according to their terms. I shall explore each option briefly.
1. GSS Interrogations as ‘Beyond the Law’
There are a number of reasons that might incline one to the view that the activities of a country’s security forces are beyond the law. One stems from the ancient idea that situations of necessity are so extreme that the law can hope to have no influence on the motives that move citizens to act in such situations. As Francis Bacon put this point, in utilizing Cicero’s well-known two men on one plank hypothetical: ‘if either there lie an impossibility for a man to do otherwise, or so great a perturbation of judgement and reason as in presumption of law man’s nature cannot overcome, such necessity carrieth a privilege in itself.141
There are two ways to interpret this idea from Bacon. One is to say that acting in situations of necessity is done literally outside the reach of a nation’s positive laws, so that normal proscriptions against murder and torture do not apply. Building perhaps on his well-known views that law has to serve an essential action-guiding function in order to be law at all,142 Lon Fuller has his fictional (p.728) ‘Justice Foster’ take this interpretation of Bacon.143 Alternatively and preferably, Bacon should be taken to mean that a nation’s positive laws do reach behaviour in situations of necessity, but that one of the things those laws provide is that necessity is a defence.144 The latter interpretation keeps necessitous behaviour governed by law and does not lead to the odd conceptual puzzles of how certain behaviour within a state’s borders cannot be subject to that state’s laws (even to the extent of having those laws grant a liberty to engage in that behaviour).145
In any case, however Bacon is construed, his ‘psychological impossibility’ of conforming to the law has no application to GSS interrogation of terrorist suspects. GSS interrogators, and those who make policy for them within the GSS, are not clinging desperately to a plank. It is not impossible for the law to have its educative and deterrent effect on GSS behaviour, so this cannot be a good reason for regarding that behaviour as ‘beyond the law’.
A second reason sometimes suggested here is that it is unfair and unfitting for lawyers and judges after the fact to judge the behaviour of those acting in extremis. Leo Katz nicely catches this intuition:
Somehow what goes on in the spacious, soft-carpeted, high-ceilinged courtrooms of civilization seems to have little bearing on what men do to each other under circumstances as extraordinary as those in which the sailors and the spelunkers found themselves.146
What cannot fairly be judged in a courtroom, one might think, should be seen as ‘beyond the law’.
Yet this too is an indefensible view to adopt for GSS behaviour, for both the reasons I advanced against Fuller’s interpretation of (p.729) Bacon: (1) one better alleviates any unfairness of after-the-fact judgement by having law that takes into account the psychological difficulties under which the behaviour judged was done, rather than saying such behaviour is beyond the reach of law at all; and (2) a state organization adopting deliberate policies does not act in extremis, but in fact engages in what looks like traditional law enforcement functions as a matter of deliberate state policy.
A third and deeper reason might be suggested for why behaviour like that of the GSS is beyond the law. That is the suggestion that the hard choices made in situations of necessity are beyond morality itself, and thus are beyond that reflection of morality we call the criminal law. Bernard Williams suggested such a view of morality as it applies to situations like that of torturing a terrorist’s child in order to save a building full of innocent lives. A moral agent, Williams tells us, may in such situations:
find it unacceptable to consider what to do … Logically, or indeed empirically conceivable they [the situations] may be, but they are not to him morally conceivable, meaning by that that their occurrence as situations presenting him with a choice would represent not a special problem in his moral world, but something that lay beyond its limits. For him, 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.147
Such a view does not depend on a psychological premise, as did the two earlier reasons we considered. Williams is not suggesting that reason is psychologically unhinged when we face the kinds of choice discussed in this article. Rather, it is morality itself that is left behind in such choice situations. Thus, Williams concludes, he would decide his hypothetical case of ‘Jim’148 in favour of Jim’s killing one innocent native in order that nineteen other innocent natives not be killed by someone else.149 But such a decision would not be dictated by morality:
Instead of thinking in a rational and systematic way either about utilities or about the value of human life, the relevance of the people at risk being (p.730) present, and so forth, the presence of the people at risk may just have its effect. The significance of the immediate should not be underestimated … we are not primarily janitors of any system of values, even our own: very often, we just act, as a possibly confused result of the situation in which we are engaged. That, I suspect, is very often an exceedingly good thing.150
I find this view of morality, and like views put forward by Elizabeth Anscombe151 and Tom Nagel,152 to be totally unacceptable. Such views treat morality and even practical rationality as ‘running out’ when the going gets tough. Morality and practical reason supposedly have nothing to say on what we ought to do; Sartre-like, we are told, we must just choose. Such views do not square with our experience with acute moral dilemmas. We never feel ‘that from a moral point of view it cannot matter any more what happens’. What makes moral dilemmas such as those we have been exploring so acute is that it does matter what we choose to do. That is why we think as hard as we can about them and try to discover the best of an admittedly not very happy set of answers.
Williams’s view makes morality only fit for a world that is better off than the world we actually inhabit. Morality is only applicable when things are ‘nice’; moral theories are ideal theories in the pejorative sense of the word. Even Williams does not really believe this, for he can resolve hard moral dilemmas like the one he constructs about ‘Jim’. GSS interrogative techniques in the face of terrorism might well be beyond the law if those techniques were beyond morality, but since the latter is not the case, neither is the former.
A fourth reason for thinking the GSS’s war on terrorism is beyond the law lies in the unsavoury nature of torture. Intentionally inflicting pain on another human being for any reason is degrading and disgusting. One might think that even though such behaviour can be justified on occasion, its odious character prevents the law from giving it any recognition. Law, one might think—unless it is going to ban flatly such unsavoury behaviour—cannot otherwise regulate it.
An example of what detailed legal regulation of torture would look like is provided by the T’ang (AD 618–907) and Ch’ing (AD 1655–1911) Codes of China, which allowed torture of witnesses as well as suspects to obtain information and confessions.153 Both (p.731) Codes define with precision permissible instruments of torture, which instruments had to be governmentally inspected and stamped; bamboo for beating, for example, is divided between light and heavy, with precise measurements of each. The amount of torture is specified: for beatings, for example, the T’ang Code allowed a suspect to be beaten a maximum of three times during a sixty-day period, with a total limit of 200 blows during this period; the Ch’ing Code allowed thirty blows per day, with the heavy bamboo. The circumstances of use was specified: for the very painful ankle-squeeze, only homicide or robbery justified its use, and even then, it was never to be used on a woman.
There is surely something grotesque and ugly about detailed regulations of this kind. (Surely part of the reason why Israel has kept secret its classified regulations on GSS ‘pressure’ on terrorists is a similar distaste.) Yet part of what makes such regulations so distasteful is that they only tell half the story, which is the evil that may be done by torture; they do not tell of the evil being prevented whenever torture is justified. If I am right in the moral conclusions reached earlier, those are conclusions of which no one should be ashamed. The law, like the morality that underlies it, can speak openly about them. One can say what the Parker Commission in England said of British interrogative techniques when used against Irish Republican Army suspects in order to save life:
[T]he answer to the moral question is dependent on the intensity with which these techniques are applied and on the provision of effective safeguards against excessive use … Subject to these safeguards, we have come to the conclusion that there is no reason to rule out these techniques on moral grounds and that it is permissible to operate them in a manner consistent with the highest standards of our society.154
I conclude that there is no good reason to think that the subject of permissible torture must remain outside the reach of law. On the other hand, there are quite good reasons why such behaviour should remain subject to court-enforced law. The main reason is obvious: without such legal restraint, the potential for abuse is (p.732) enormous. The American experience with a like claim made by the Nixon Administration—that national security placed government agents outside the laws against burglary and the Fourth Amendment—underlines the obvious dangers of telling any governmental agency they are ‘beyond the law’. It took Sir Thomas More, Lord Coke, and their common law brethren several centuries to convince the English kings that they too were subject to the law. It took a unanimous US Supreme Court and much American travail to convince Richard Nixon that the same applied to him. It is not a lesson that should need repeating in Israel.
2. Acoustic Separation: One Law for the GSS, Another for the Executive?
Sometimes the best way to hit a target is not to aim at it. When shooting an arrow in a steady crosswind, one does best by aiming to the side of the target by that amount one estimates the wind will carry the arrow during its flight. Similarly in law, sometimes a legislator best achieves her targeted behaviour by not describing that behaviour in the content of her laws. This will be the case whenever those who must obey such laws have a ‘steady crosswind’ in their motivation, viz., whenever they are systematically prone to erring one way rather than another.
It is often said that this condition is met in necessity cases where the life being saved includes the actor’s own. Self-preservation being the instinct that it is, the argument goes, actors systematically err in their own favour as they calculate when life must be taken to save even more life. Dudley and Stephens 155 was rightly decided, on this view, because it gave the right incentive structure to those in lifeboats: since the court held that any homicide will be murder, those in extremis will wait until death is almost upon them before they kill another. That is optimal because it results in life being taken only when it is truly necessary to preserve more life. Whereas, it might be thought, a rule that did not flatly ban homicide but allowed it when necessary to save more life would encourage too much killing, given everyone’s self-preservative instincts.
There is of course the nagging problem of fairness to actors like Dudley and Stephens who did not overly favour themselves and who did kill only when it was necessary to save more life. This the court took care of by inviting executive clemency, which was eventually and quietly granted to Dudley and Stephens.
(p.733) This divorce between the stated law that citizens (including GSS interrogators) are to apply to their behaviour, and an unstated targeted behaviour that the lawmaker wishes to achieve with the stated law, Meir Dan-Cohen has called ‘acoustic separation’.156 The acoustic separation is between what the GSS interrogator hears (the stated law, for example, ‘never torture’) and what some other official hears (the unstated, targeted behaviour, for example, ‘torture only when justified in Moore’s long and complicated chapter’). If the other official is the Executive, then a pardon or other form of executive clemency would be appropriate for those GSS interrogators whose behaviour was ‘on target’ even though violative of the stated law—only, of course, when such pardon or clemency could be given without public attention.
The psychological premise that would motivate this view has some accuracy when applied to GSS interrogators. Such interrogators, if allowed to calculate at all, will probably err in favour of too much torture too often. This may be because: they have just witnessed the death or injury of innocents at the hands of the terrorists they are about to interrogate, so that their fellow-feeling for such suspects is at low ebb; their patriotism inclines them to favour Israeli lives being saved over non-Israeli lives being lost; the accepted criteria for success in their job is in terms of arms discovered, lives saved, etc., not in terms of rights and dignities preserved; and, last but not least, brutal treatment brutalizes the inflicter of it.
For this reason, were I writing Israel’s laws, I would make no provision for the ‘threshold deontology’ I described in my complex agent-relative view of morality. I assume that a patriotic GSS interrogator will torture an innocent person when the consequences of not doing so are so horrendous that the consequential override of the threshold deontologist is morally required. I further assume that he will do so no matter that the stated law forbids it. Whereas were the stated law to include a provision saying that a GSS interrogator may do anything to anyone if the consequences of not doing so are horrendous enough, probably more torture would take place than could be morally justified. I thus think Israeli law should contain a flat ban on torture of innocents.
(p.734) Whether such a flat ban should encompass those who possess information whose transfer would save lives and injury, but who refuse without justification to divulge such information, is a difficult question. Certainly the same danger of GSS interrogators systematically erring in favour of too much torture is present. On the other hand, the moral error is less serious, and the likelihood of error is less (because the area of morally permissible torture is greater), if the person interrogated does possess life-saving information. I conclude, with some misgiving, that torture of those possessed of life-saving information, even though not themselves terrorists, should not flatly be banned. Nor, of course, should torture of terrorist suspects themselves.
Even the flat ban on torture of the innocents has some unfairness built into it. Those GSS interrogators who torture the innocent on those very rare occasions when it is justified—the horrendous consequences threshold—have cause for complaint that they will be punished although they did the right thing. Their behaviour was the targeted behaviour even though violative of the stated law. Of course, like Dudley and Stephens and the GSS personnel pardoned in the No. 300 bus affair, such interrogators may hope for executive clemency, the granting of which alleviates the unfairness. Still, a legislator sharing the complex agent-relative moral views I described earlier has to be willing to impose some unfairness as a cost of flatly banning even torture of the innocent.
3. A Law of Torture That Means What It Says
Other than the flat ban on torture of the innocents above described, Israel should have a law on torture that means exactly what it says. Open to question, however, is exactly how much the law should say on this subject. How much detail should be written into law, and how much should be left to individual decision under more general standards?
There are two kinds of details that could be written into law, corresponding to the two kinds of moral knowledge required to apply the balance-of-evils defence. First, one could specify both sides of the consequentialist balance. One could spell out the evils being done, like the Chinese torture schedule adverted to earlier; and one could spell out the evils being averted, such as numbers of lives, etc. Presumably, one would create a kind of sliding scale, more evil tortures requiring greater evils averted. Second, one could specify the (p.735) limits on consequential calculation I described in my complex, agent-relative view of morality.
I think there are good reasons not to do either of these things. I do not think such reasons have to do with it being dishonourable to spell out the limits of legal torture. What is morally right cannot be dishonourable (although it certainly can be painful and guilt-producing). Rather, such reasons have to do with J. L. Austin’s earlier quoted aphorism that Tact is richer than diction’.157 The circumstances making torture of terrorists and others justifiable are too various to be captured by ‘per se’ rules of justified torture. This is the reason that section 22 of the Israeli Penal Law, and like provisions of American and German law, are formulated as broadly as they are. Any greater precision will be at the cost of accuracy.
Consider the American experience in trying to define the like concept of negligence. Oliver Wendell Holmes thought that negligence could eventually be reduced to much more determinate rules, such as his rule that a motor car must ‘stop, look, and listen’ when coming to a railroad crossing. Anyone knows that such a rule is both over- and under-inclusive of cases with respect to negligence. It is over-inclusive because it includes cases where not stopping is not negligent because one can see clearly in both directions; it is under-inclusive because stopping, looking, and listening may not be enough precautions to be reasonable, as where there is a lot of noise and vision is blocked so that the motorist who only stops, looks, and listens cannot tell whether a train is coming. The lesson is that fact is richer than diction for concepts like negligence or necessity. Sir William Scott thus had it right when he wrote in 1801:
The law of cases of necessity is not likely to be well furnished with precise rules … whatever is reasonable and just in such cases, is likewise legal. It is not to be considered a matter of surprise, therefore, if much instituted rule is not to be found on such subjects.158
Other than the flat ban on torturing the innocent, section 22 is all the law on torture Israel should want or need.
What is needed is something that by law is very hard to achieve, namely, a shared set of attitudes within the GSS that abhors torture as inherently evil while at the same time allowing that, on (p.736) occasion, some evils are greater. The procedural reforms within the GSS suggested by the Landau Commission are helpful in fostering such attitudes. Having higher level officials sign off on extraordinary interrogative techniques, and steering those techniques away from the kinds of physical abuse that breeds sadism in those who use them, are steps in the right direction.
(1) Landau Commission, Landau Commission Report (Jerusalem: Government Printing Office, 1987).
(5) Landau Commission Report at 61.
(11) On the defence of necessity generally, see Leo Katz, Bad Acts and Guilty Minds (Chicago: University of Chicago Press, 1987), at 8–81; George Fletcher, Rethinking Criminal Law (Boston: Little, Brown, 1978), 774–98; Glanville Williams, Criminal Law—The General Part (London, 2nd edn., 1961), at 722–46; Joshua Dressier, Understanding Criminal Law (Mathew Bender, 1987), at 249–57; Arnolds and Garland, ‘The Defence of Necessity in Criminal Law: The Right to Choose the Lesser Evil’, Journal of Criminal Law and Criminology Vol. 65 (1974), 289–301; Glazebrook, The Necessity Plea in English Criminal Law’, Cambridge Law Journal, Vol. 30 (1972), 87–119; Tiffany and Anderson, ‘Legislating the Necessity Defence in Criminal Law’, Denver Law Journal, Vol. 52 (1975), 839–79; Comment, ‘Necessity Defined: A New Role in the Criminal Defense System’, UCLA Law Review, Vol. 29 (1981), 409–46.
(12) Penal Law, 1977 (LSI Special Volume), s. 22.
(14) That arguments of necessity enter a criminal trial only as matters of defence is not as obvious as it may seem. Both the actus reus and mens rea elements of the prosecution's prima facie case may have hidden necessity-like requirements in them. In construing what acts are prohibited by a criminal statute, courts will often go against the plain meaning of statutory language in order to exempt desirable violations of the statute. See, e.g., Kirby v. United States, 74 US (7 Wall.) 482 (1868) (literal obstructing of the federal mails held not to be an ‘Obstructing’ within the meaning of the statute when done to effect the arrest of a federal mail carrier wanted for murder); see generally Moore, ‘The Semantics of Judging’, Southern California Law Review, Vol. 54 (1981), 151–294; Glanville Williams, Criminal Law, supra n. 11, at 724–8. Likewise, in considering what mental states suffice for the mens rea requirements of recklessness or negligence, courts consider only risk-takings that are unjustified. See, e.g., Model Penal Code, s. 2.02(2)(c) and (d).
(15) See Fletcher, Rethinking, supra n. 1, at 759, 762, 799–800–11; Dressier, Understanding, supra n. 11, at 179; Kent Greenawalt, ‘The Perplexing Borders of Justification and Excuse’, Columbia Law Review, Vol. 84 (1984), 1897–927.
(17) Model Penal Code, s. 3.02 (Proposed Official Draft, 1962).
(18) The failure of the Israeli Code to distinguish justification from excuse in its necessity provision could be due to the consensus of influential criminal law theorists in Israel that no legal consequences should attach to the distinction (that is, that a defendant and those who aid or resist him are equally liable or exempt from punishment no matter whether the defendant claims necessity as a justification or necessity as an excuse). See Gur-Arye, ‘Should the Criminal Law Distinguish Between Necessity as a Justification and Necessity as an Excuse?’, Law Quarterly Review, Vol. 102 (1986), 71–89; Kremnitzer, ‘Proportionality and the Psychotic Aggressor: Another View’, Israel Law Review, Vol. 18 (1983), 178–214, at 196–9. One can believe this and yet believe that it is crucial to recognize the distinction when one analyses provisions such as s. 22, for the criminal law mirrors morality here and in morality the distinction is basic.
(20) See J. Dressier, Understanding, supra n. 11, at 251–3. It is this feature of the law of necessity that rules out any argument by terrorists that prevention of their future acts of terror is not a good consequence that may justify GSS torture. Terrorists may sincerely believe that their killing of innocent civilians is not wrong but right because it produces more good in the long run. They are just as incorrect in that belief as would be a GSS interrogator who believes that torturing an innocent child can be justified by the good consequences in the long run. The defence of necessity protects in neither case such moral error.
(21) See Michael Moore, Law and Psychiatry: Rethinking the Relationship (Cambridge: Cambridge University Press, 1984), at 83–4.
(23) Philippa Foot, ‘The Problem of Abortion and the Doctrine of Double Effect’, Oxford Review, Vol. 5 (1967), 5–15, reprinted in Foot, Virtues and Vices (Oxford: Oxford University Press, 1981); Sanford Kadish, ‘Respect for Life and Regard for Rights in the Criminal Law’, California Law Renew, Vol. 64 (1976), 871–901, reprinted in Kadish, Blame and Punishment (New York: Macmillan, 1987); Harris, ‘The Survival Lottery’, Philosophy, Vol. 50 (1975), 81–7; Thomson, ‘Killing, Letting Die, and the Trolley Problem’, The Monist, Vol. 59 (1976), 204–17, and Thomson, ‘The Trolley Problem’, Yale Law Journal, Vol. 94 (1985), 1395–415 (both trolley articles are reprinted in J. Thomson, Rights, Restitution, and Risk (Cambridge, Mass.: Harvard University Press, 1986); L. Katz, Bad Acts, supra n. 11, at 35 (subsequent references to these articles are to the reprinted versions).
(24) St GG, s. 34.
(26) The utilitarian and consequentialist flavour of the reasoning behind s. 3.02 of the Model Penal Code is accurately reflected in a student Note, ‘Justification: The Impact of the Model Penal Code on Statutory Reform’, Columbia Law Review, Vol. 75 (1975), 914–62, at 921–8. See also Kent Greenawalt, ‘Violence—Legal Justification and Moral Appraisal’, Emory Law Journal, Vol. 32 (1983), 437–97, at 465.
(27) S. Kadish, ‘Respect for Life’, supra n. 23, at 123. Compare the seemingly contrary assumption of G. Fletcher, Rethinking, supra n. 11, at 787–8, G. Williams, Criminal Law, supra n. 11, at 729, and Greenawalt, ‘Violence’, supra n. 26.
(30) The general form of the argument in the text is pursued with great clarity by Samuel Scheffler, The Rejection of Consequentialism (Berkeley: University of California Press, 1982), 87–93, 98–101.
(31) For this characterization of act-consequentialism, see Bernard Williams's contribution to J. J. C. Smart and Bernard Williams, Utilitarianism: For and Against (Berkeley: University of California Press, 1973), 86.
(32) Scheffler, ‘Introduction’, in Samuel Scheffler, ed., Consequentialism and Its Critics (Oxford: Oxford University Press, 1988), at 3.
(34) See the discussion about Katz earlier, text at nn. 28–9. For more general defences of non-utilitarian consequentialisms, see Tim Scanlon, ‘Rights, Goals, and Fairness’, in S. Hampshire, ed., Public and Private Morality (Cambridge: Cambridge University Press, 1978), reprinted in Consequentialism and Its Critics, supra n. 32, at 74–92, and Sen, ‘Rights and Agency’, Philosophy and Public Affairs, Vol. 11 (1982), 3–39, reprinted in Consequentialism and Its Critics, supra n. 32, at 187–223.
(36) One might also reject act-consequentialism on the strictly consequentialist ground that if followed as a guide to moral decisions by individual agents it would not in fact produce the best consequences. See Donald Regan, Utilitarianism and Cooperation (Oxford: Oxford University Press, 1980), 12–53.
(38) Barnes, ‘Utilitarianisms’, Ethics, Vol. 82 (1971), 57–64. On the different varieties of rule-consequentialism, and problems with each, see Richard Brandt, ‘Toward a Credible Form of Utilitarianism’, in Baruch Brody, ed., Moral Rules and Particular Circumstances (Englewood Cliffs, N.J.: Prentice-Hall, 1970); David Lyons, ‘Utility and Rights’, in Jeremy Waldron, ed., Theories of Rights (Oxford: Oxford University Press, 1984); Scanlon, ‘Rights, Goals, and Fairness’, supra n. 34, at 74.
(39) I. Kant, Groundwork of the Metaphysic of Morals, J. H. Paton, trans. (New York: Harper, 1964). As Tom Nagel captures this Kantian insight: ‘No doubt it is a good thing for people to have a deep inhibition against torturing children even for very strong reasons, and the same might be said of the other deontological constraints. But that does not explain why we find it almost impossible to regard it as a merely useful inhibition.’ Nagel, View, supra n. 35, at 179.
(40) Smart, in J. Smart and B. Williams, Utilitarianism, supra n. 31, at 44. For the same criticism, see Peter Railton, ‘Alienation, Consequentialism, and the Demands of Morality’, Philosophy and Public Affairs, Vol. 13 (1984), 134–71, reprinted in Consequentialism and Its Critics, supra n. 32.
(41) Jack Smart's willingness to embrace consequentialism even at the cost of giving up common sense morality's constraints, like ‘never hang an innocent man’, has earned Smart the following entry in The Philosophical Lexicon (D. Dennett and K. Lambert, eds., 7th edn., 1978), at 8: ‘Outsmart, verb; to embrace the conclusion of one's opponent's reductio ad absurdum argument. (As in) “They thought they had me, but I outsmarted them. I agreed that it was sometimes just to hang an innocent man.” ‘
(42) Williams, in J. Smart and B. Williams, Utilitarianism, supra n. 31, at 90. I -examine this question in Moore, ‘Good Without God’, in Robert George and Christopher Wolfe, eds., Liberalism, Modernity, and Natural Law (Oxford: Oxford University Press, 1996), concluding that God does not help here.
(46) Foot, ‘Abortion’, supra n. 23, at 26. George Fletcher has recently urged that the Talmudic example of keeping water for oneself when there is only enough for one presupposes adherence to an acting/allowing distinction, for self-love could hardly justify taking the water from another (which would be a killing, not an allowing to die). George Fletcher, ‘Defensive Force as an Act of Rescue’, Social Philosophy and Policy, Vol. 7 (1990), 170–9.
(47) Foot, ‘Abortion’, supra n. 23, at 28–9, for both variations. Sometimes Foot believes we may justify the violation of a negative duty by good consequences. She construes her famous trolley example to be of that kind, where a trolley driver is justified in driving over and thereby killing one trapped workman if the consequence of his not doing so would have been to drive over and kill five.
(48) On the doctrine of double effect, see Foot, ‘Abortion’, supra n. 23; Nagel, View, supra n. 35, at 179–80; H. L. A. Hart, ‘Intention and Punishment’, in Punishment and Responsibility (Oxford: Oxford University Press, 1968); Jonathan Bennett, ‘Whatever the Consequences’, Analysis, Vol. 26 (1966), 83–102; G. E. M. Anscombe, ‘Modern Moral Philosophy’, Philosophy, Vol. 33 (1958), 1–19; John Finnis, Natural Law and Natural Rights (Oxford: Oxford University Press, 1980), at 118–25; Charles Fried, Right and Wrong (Cambridge, Mass.: Harvard University Press, 1978), at 31–2.
(50) See generally Charles Fried, An Anatomy of Values, Part III (Cambridge, Mass.: Harvard University Press, 1970).
(52) Model Penal Code s. 3.02, Comments at 8 (Tent. Draft No. 8, 1958): ‘So too a mountaineer, roped to a companion who has fallen over a precipice, who holds on as long as possible but eventually cuts the rope, must certainly be granted the defense that he accelerated one death slightly but avoided the only alternative, the certain death of both.’ For a real life example of this situation, see Joe Simpson, Touching the Void (New York: Harper and Row, 1988).
(53) The example is an ancient one, to be found in both Bacon and Kant. Justice Holmes's version was: ‘If a man is on a plank in the deep sea which will only float one, and a stranger lays hold of it, he will thrust him off if he can.’ O. W. Holmes, The Common Law (Boston: Little, Brown, 1881), at 47. See the treatment of the plank hypothetical in I. Kant, Metaphysical Elements of Justice, J. Ladd, trans. (1965), at 41–2. See generally Williams, Criminal Law, supra n. 11, at 738.
(55) R v. Dudley and Stephens (1884) 14 QBD 273; United States v. Holmes, 226 Fed. Cas. 360 (3d Cir. 1842). For commentary on these famous cases, see A. W. B. Simpson, Cannibalism and the Common Law (Oxford: Oxford University Press, 1984); Fuller, ‘The Case of the Speluncean Explorers’, Harvard Law Review, Vol. 62 (1949), 616–45; Katz, Bad Acts, supra n. 11, at 8–62.
(62) Model Penal Code s. 3.02, Comments at 8 (Tent. Draft No. 8, 1958).
(67) Although the defence of duress is best thought of as an excuse, not a justification, the English doctrine here discussed does not attend to that distinction. I accordingly feel some justification for treating the law of duress as an apt analogy illustrating the principle discussed in the text.
(68) Lynch v. Director of Public Prosecutions  All ER 917. When an accomplice does more than make possible the principal's intervening act—where he actually participates in the killing—then duress is no longer available as a defence. (See Abbott v. R  All ER 140; see also Katz, Bad Acts, supra n. 11, at 62–9.) This limitation also illustrates the principle in the text, because where the accomplice actually holds the victim while she is being skewered with a sabre wielded by the principal (the facts of Abbott), the principal's killing is not an act intervening between that of the accomplice and the harm; rather, both the sword thrust of the principal and the holding by the accomplice jointly kill. Failing to attend to this difference, the English courts have recently abolished the doctrine that accomplices may ever avail themselves of the defence of duress when the charge is murder.
(72) Judy Thomson at one time came close to adopting Nagel's distinction. Needed, she thought, was some account of ‘what it is to bring something about by doing something’. J. Thomson, Rights, supra n. 23, at 92. With such an account, she thought, perhaps we could make sense of the intuition ‘that what matters in these cases in which a threat is to be distributed is whether the agent distributes it by doing something to it [which then causes harm to the victim], or whether he distributes it by doing something to a person’.
(73) The Model Penal Code nicely duplicates these three conditions of responsibility in ss. 2.01 (voluntary action), 2.02 (culpable mental states), and 2.03 (causal responsibility).
(76) I am unclear as to Foot's ultimate views on her earlier starving beggar example. It may be she would now agree. See Foot, ‘Immorality, Action and Outcome’, in Ted Honderich, ed., Morality and Objectivity (London: Routledge, 1985), 23–38, at 37, n. 6 where she appears to reclassify the example as a violation of a merely positive duty but with a direct intention.
(77) Foot, ‘Immorality’, ibid., at 24. Leo Katz, in his delightful Ill-Gotten Gains (Chicago: University of Chicago Press, 1996), celebrates this possibility for deontological ethics. For additional criticism of Foot on this point, see Warren Quinn, ‘Actions, Intentions, and Consequences: The Doctrine of Doing and Allowing’, Philosophy and Public Affairs, Vol. 18 (1989), 334–51.
(81) This is not an unproblematic rationale, for it requires us to make sense of the idea that causation can be a more-or-less affair, a relation capable of being more strongly or less strongly present by matters of degree. In ch. 131 questioned whether the causal relation is this kind of scalar phenomenon, more like the relation ‘bigger than’ than the relation ‘brother of. Nonetheless, the common moral intuition to which Thomson's distribution exemption points us does seem to be built on this scalar intuition about causal responsibility. The most obvious candidate for a suitably scalar conception of the singular causal relation would be a physicalistic conception where there can be varying magnitudes of force and the like. I discuss some of such conceptions in ch. 7, without resolving the question of their viability.
(82) H. L. A. Hart and A. M. Honoré, Causation in the Law (Oxford: Oxford University Press, 1959).
(83) Nagel himself appears to have construed his own former distinction (between doing and causing) quite differently; now he aligns it with the intending/foreseeing distinction. Nagel, View, supra n. 35, at 179.
(87) I think that such combination of intention and action as the object of morality's agent-relative prohibitions allows me to escape the difficulties for either alone so elegantly explored in Heidi Hurd, ‘What in the World Is Wrong?’, Journal of Contemporary Legal Issues, Vol. 5 (1994), 157–216. Nonetheless, serious difficulties no doubt remain in spelling out how moral norms can conjunctively prohibit both culpable intentions and wrongful actions.
(88) Notice that the other aspect of culpability, excuse, is absent in this articulation of the range of permissible consequential justification. Just because one may be somewhat coerced does not increase the permissibility of consequential justification. An excused, or partly excused wrongful action remains wrongful in the face of good consequences. This is true in the actual world (for wrongdoing) and in the possible world created by the actor's mental states (for culpability). One may still be excused or partly excused for such intentionally done wrongful actions; but the conditions of excuse do not insinuate themselves into the objects of agent-relative norms in the way that the act, intention, and causation conditions do. Put more simply, the excusability of an action or intention in no way affects the issue of their justification.
(89) See, e.g., Fried, Right and Wrong, supra n. 48; Nagel, View, supra n. 35, at 175–88. It does seem to me that any answer to this question will have to focus on the moral integrity or well-being of the agent to whom moral prohibitions apply and not on the unfairness or rights of the victim. (See Joseph Raz, Morality of Freedom (Oxford: Oxford University Press, 1987), at 284–7.) This suggests the need for a virtue-based account of deontological constraints, recognizing that we tend not to think of all duties as ultimately being duties we owe to ourselves (to be virtuous). For a sceptical discussion of this approach, see Samuel Scheffler, ‘Agent-Centred Restrictions, Rationality, and the Virtues’, Mind, Vol. 94 (1985), 409–19, reprinted in Consequentialism and Its Critics, supra n. 32, at 243, 254–5.
(90) Scheffler, Rejection, supra n. 30, at 114. Scheffler himself ultimately allows the difficulties to overcome the intuitions, at least for agent-relative prohibitions if not for agent-relative permissions.
(91) I argue for this in Moore, ‘Moral Reality’, Wisconsin Law Review, Vol. , 1061–156; Moore, ‘Moral Reality Revisited’, Michigan Law Review, Vol. 90 (1992), 2424–533; Moore, ‘Good Without God’, supra n. 42.
(92) Of course, for social rules (unlike the laws of morality), an indefinitely large number of unknown exceptions defeats the purpose served by the existence of social rules. For the difference, see Moore, ‘Three Concepts of Rules’, Harvard Journal of Law and Public Policy, Vol. 14 (1991), 771–95.
(93) See John Rawls, ‘Two Concepts of Rules’, Philosophical Review, Vol. 64 (1955), 3–32, where Rawls distinguishes exceptions to the norm against breaking a promise from overriding that norm. See also Raz, Morality of Freedom, supra n. 89, at 361–2.
(94) See Heidi Hurd, The Moral Magic of Consent’, Legal Theory, Vol. 2 (1996), 121–46.
(95) See H. L. A. Hart, Law, Liberty, and Morality (Stanford: Stanford University Press, 1963), at 31.
(96) In Holmes, supra n. 55, the Court stated that if a fair procedure had been used to select who was to be thrown overboard, then the act would have been justified. In Lon Fuller's variation, it is not only the fairness of the selection procedure that makes a difference, but the victim's having consented to the procedure ex ante (i.e., before he knew he would draw the short straw). See Fuller, ‘Speluncean Explorers’, supra n. 55; Katz, Bad Acts, supra n. 11, at 51–6.
(98) Although on occasion such ‘bare fact of consent’ does do some moral work. What (of a non-utilitarian sort) can be said in favour of killing any fleeing felon (the old common law rule for policemen) was that he assumed the risk of being shot when he ran after the warning, ‘stop or I’ll shoot’.
(99) A ‘content-independent reason’ for action is a reason created by an act like a promise, or a vow, or a request of a friend. Such acts create new reasons to do what is promised, avowed, or requested, reasons whose force as reasons is independent of the goodness of the actions promised, etc. See generally Raz, Morality of Freedom, supra n. 89, chs. 2–4.
(101) On the problems with assumption of the risk arguments like that dealt with in the text, see Leo Katz, ‘The Assumption of the Risk Argument’, Social Philosophy and Policy, Vol. 7 (1990), 138–69.
(104) Compare Kadish, ‘Respect for Life’, supra n. 23; Thomson, ‘Self-Defense and Rights’, in Thomson, Rights, supra n. 23; Dressier, Understanding, supra n. 11, at 199–201; Montague, ‘Self-Defense and Choosing Between Lives’, Philosophical Studies, Vol. 40 (1981), 207–19; Wasserman, ‘Justifying Self-Defense’, Philosophy and Public Affairs, Vol. 16 (1987), 356–78; Montague, ‘The Morality of Self-Defence: A Reply to Wasserman’, Philosophy and Public Affairs, Vol. 18 (1989), 81–9.
(106) These possibilities are discussed in Thomson, ‘Self-Defense’, supra n. 104. Rashi, a medieval commentator on the Talmud, appears to have adopted such a victim-based account of self-defence, for he treats the killing of a burglar as a trivial act: such a blameworthy person as a burglar ‘has no blood’ in the sense that he should be treated as one who is already dead. See Marilyn Finkelman, ‘Self-Defense and Defense of Others in Jewish Law: The Rodef Defense’, Wayne Law Review, Vol. 33 (1987), at 1257–87 at 1284–6; Fletcher, Rethinking, supra n. 11, at 256.
(110) For a variety of examples, see George Fletcher, ‘Proportionality and the Psychotic Aggressor: A Vignette in Comparative Criminal Theory’, Israel Law Review, Vol. 8 (1973), 367–90; Robert Nozick, Anarchy, State and Utopia, (Cambridge, Mass.: Harvard University Press, 1974), at 34–5.
(113) Cf. Scanlon, ‘Rights, Goals, and Fairness’, supra n. 34, at 90, who suggests that judgements of a non-consequentialist sort may generally reflect a ‘bias of the lucky against the unlucky’. The alternative account of the shield type of innocent aggressor cases is that the actor is excused, but not justified, in shooting through the shield to save himself.
(119) That the Bad Samaritan has no good reason to refuse to prevent the harm is crucial here. Otherwise, we might think the healthy individual who refuses to give up his life so that five others may live with his organs is a Bad Samaritan deserving of sacrifice. Yet such an individual has a good reason not to give up his life and his organs: it is a very costly sacrifice to him. Similarly, if the spouse of a terrorist could show that she had good reason not to speak—a death threat by the terrorist organization, for example—then she too may not be the Bad Samaritan subject to harsh treatment. (I owe both of these points to Heidi Hurd.)
(120) Compare McAuley, ‘Anticipating the Past: The Defense of Provocation in Irish Law’, Modern Law Review, Vol. 50 (1987), 133–57, with Dressier, ‘Provocation: Partial Justification or Partial Excuse?’, Modern Law Review, Vol. 51 (1988), 467–80. An alternative legal example of where the degree of wrong done is affected by the blameworthiness of the victim to whom it is done, is provided by the felony-murder rule. If the person killed during the commission of a felony is a co-felon, it is common to refuse application of the felony-murder rule to the killer; this, presumably because taking the life of a co-felon is less wrong than taking the life of an innocent.
(122) Tosefta, Trumot 7, 23; Jerusalem Talmud, Trumot 8, 4(46b).
(123) Cardozo, Law and Literature and Other Essays and Addresses (New York, 1931), at 113.
(126) Jean-Paul Sartre, ‘The Flies’, in No Exit and Three Other Plays (New York, trans. L. Abel, 1955).
(132) As Amartya Sen has observed, a threshold-deontology will in some sense be a consequentialist system: ‘Such a threshold-based constraint system must rest ultimately on consequential analysis, comparing one set of consequences (badness resulting from obeying the constraint) with another (badness of violating the constraint itself, given by the threshold), and its distinguishing feature will be the particular form of the consequence-evaluation function.’ Sen, ‘Rights and Agency’, supra n. 34, at 187–223, 190–1, n. 8. This will particularly be true of a threshold deontology that varies the threshold (where consequences begin to count) directly with the degree of wrongness of the act to be justified by its consequences. (That is, the greater the wrong to be done, the higher the threshold of bad consequences averted has to be to justify the action as right.)
Saying all of this still does not collapse threshold deontology into consequentialism. Even if the goodness of consequences is always relevant to the Tightness of actions for a threshold deontologist, the goodness of consequences does not determine the Tightness of actions as it does for a consequentialist. Contrary to Sen, the degree of (agent-relative) wrongness of an action does not translate into some degree of (agent-neutral) ‘badness of violating the constraint’. The threshold deontologist does not think that the wrongness of his doing a certain act is just another bad consequence that can be outweighed by the better consequences of preventing like acts by others.
(136) I owe the analogy to Joseph Raz.
(137) On justificatory intent, compare the differing views of Fletcher, Rethinking, supra n. 11, at 557, 559–60, with Moore, Act and Crime, supra n. 84, at ch. 7. Fletcher thinks (along with current law) that an actor whose act is justified objectively but who lacks justificatory intent is guilty of the completed crime; I think that such an actor does no wrong but because he did such a wrong in his own mind, his culpability merits punishment for attempt.
(141) Shedding, Ellis, and Heath, The Works of Francis Bacon (Philadelphia: Parry and McMillan, 1859), 343.
(142) Lon Fuller, The Morality of Law (2nd edn., Stanford: Stanford University Press, 1969).
(143) As Justice Foster puts it: ‘I take the view that the enacted or positive law of this Commonwealth, including all of its statutes and precedents, is inapplicable to this case, and that this case is governed instead by what ancient writers in Europe and America called “the law of nature”.’ Fuller, ‘Speluncean Explorers’, supra n. 55, at 620.
(144) The difference between these two interpretations of Bacon can be seen by attending to the distinction between what modal logicians call external versus internal negation: Foster's interpretation is that it is not the case that the law prohibits spelunkean homicide (external negation); the alternative interpretation is that it is the case that the law does not prohibit spelunkean homicide. The law covers the situation under the second interpretation; it does not under the first.
(153) See Ch’u T’ung-Tsu, Local Government in China (Cambridge Mass.: Harvard University Press, 1962), 125; D. Bodde and C. Morris, Law in Imperial China (Philadelphia: University of Pennsylvania Press, 1973), 97–8. (I am indebted to Hugh Scogin for this information.) For other examples of legalized torture, see E. Peters, Torture (New York: Basil Blackwell, 1985).
(156) Meir Dan-Cohen, ‘Decision Rules and Conduct Rules: On Acoustic Separation in Criminal Law’, Harvard Law Review, Vol. 97 (1984), 625–77’.
(157) J. L. Austin, ‘A Plea for Excuses’, Proceedings of the Aristotelian Society, Vol. 57(1956), 1–30, at 21.
(158) The Gratitudine, (1801) 165 ENG REP 450, at 459.