A Systematic Look at Releasing Terrorists (Column 608)

In the previous column I discussed the hostage deal carried out last week. In the course of that discussion I mentioned the consideration that it is right to release terrorists now even if there is a future risk that they will take lives, and perhaps even if the expectation is that more lives will be taken than those we are saving now. There I cited the Jerusalem Talmud as quoted by the Kesef Mishneh, which permits (or even obligates) one to endanger his own life to save the life of another, on the claim that “he is certain and I am doubtful.” Its argument is that the future risk is uncertain while the rescue now is certain; therefore it is preferable to save life now. I did note, however, that the accepted halakhic view is otherwise: a person is forbidden (or at least not obligated) to endanger his life to save another’s.

I also added there that regarding the Gilad Shalit deal I thought it was not right to carry it out because of the future risk. But in that case we were dealing with the release of more than a thousand terrorists in exchange for one captive. In our deal we were talking about a more balanced number of released terrorists with a risk that is not too high, particularly because the number of hostages was very large, and therefore (together with other reasons) I thought it right to do so.

How does this square with the claim that one may not take a risk upon oneself for the sake of a certain rescue? Can this depend on the numbers involved? What is the significance of the fact that we are dealing with a doubtful future risk versus a certain present rescue? Does the timeline play a role here? In this column I wish to present additional arguments for and against such hostage deals. Do not expect a definitive bottom line. My aim is to contribute to a more systematic discussion so that each person can form a reasonable position of his own.

Risk to an Individual and to the Public

The above Jerusalem Talmud and the halakhic decisors discuss the consideration of exposing one person to risk in order to save another individual. By contrast, in our case the risk looms over the public, and seemingly that makes the situation more stringent. I have written more than once (see, for example, in columns 284, 529 and more) that when there is risk to the public we must also factor in very small risks. Thus, for example, we saw there that one may extinguish a metal ember in the public domain on Shabbat even though the risk is very small, since many people pass by. In such a case, a small chance of harm is multiplied by the number of passersby, yielding a high expected loss. For example, if there is a one-in-a-thousand risk to an individual and ten thousand people pass by in a day, we expect about ten people to be harmed. That certainly justifies desecrating Shabbat.

In those columns I brought a question posed to Rabbi Mordechai Eliyahu. An army officer asked him whether he may turn on a flashlight on Shabbat in order to clear a soldier’s weapon after a shift. The assumption is that the risk of an accidental discharge is minimal (because one can check with a finger without a flashlight; the prior likelihood that a round remains in the chamber is very small; and even if it remains, it will not necessarily be discharged; and even if discharged, it will not necessarily hit another person). Therefore, for a private individual the answer should have been that this does not justify desecrating Shabbat. But since there are many officers and many soldiers, imposing a prohibition would lead to almost certain harm. Rabbi Eliyahu ruled that there is a consideration of probability multiplied by the number of cases that yields a significant expected loss. He permitted (and in fact obligated) desecrating Shabbat.

Note that in the case of the officer we are dealing with a situation different from the metal ember. In the ember case, one person desecrates Shabbat to save many who are at risk. There the justification is clear, since it saves lives with a very high probability. So it seems fairly obvious that the act is justified. But the officer’s not turning on a flashlight, in itself, does not bring about any future risk. The probability that a round remains in the chamber and will be discharged is tiny. The consideration there concerns this case as one instance within many similar cases. If all officers refrain from turning on flashlights on Shabbat, a risk will be created. The permission granted to that officer is based on a consideration akin to the categorical imperative (act in accordance with a maxim you would want to be a universal law; see column 122 and many more). Note that in that case, even after he was permitted to turn on the flashlight, he could tell himself that others will do as they will; his own act does not endanger life. The logical conclusion from his perspective would be to disregard the halakhic instruction. One could argue against him that if everyone disregards it, people will be harmed; but he will reply that others’ decisions are independent of his, and his choice affects only his own case. Therefore, at the end of the day, his desecration of Shabbat prevents only a very small risk and thus is not justified. Rabbi Eliyahu correctly rules that it is permitted (for him it is like the metal ember), but from the perspective of this particular officer only the categorical imperative would permit him to desecrate Shabbat (in that column I explained the complex relationship of the categorical imperative to consequential considerations). All of this of course cannot be said about the ember case, where a single act prevents a significant risk to many. There, even for the one who extinguishes (not only for the halakhic authority) there is a simple justification for desecrating Shabbat that is unrelated to the categorical imperative.

The Fundamental Difficulty

In any case, the conclusion is that a very small risk that threatens the public (or many people, not necessarily “the public”) justifies an otherwise forbidden act, since the expected harm in such cases is large. Seemingly that is the situation in a deal releasing terrorists. There too the risk is to the entire public, and such risk must be reckoned by the statistical expectation of the likely outcome. That is, suppose that in the Shalit deal the threat posed by the released terrorists was one in a million, but there are ten million citizens under that threat; from our perspective it is as if the lives of ten people are in real danger. Is it justified to prefer Shalit’s life over the ten future victims? Note that this is not a risk to ten people. The risk is to ten million people. The ten are those who will certainly die as a result of the deal. That is not the deal’s risk but its price. There would seem to be no justification to save one life at the cost of sacrificing the lives of ten others. Taking risks can be permitted, but here the risk faces a large group, and therefore we should consider the expected outcome as if it were certain. This conclusion is straightforwardly consequentialist and does not depend on the categorical imperative.

At first glance this resembles the famous trolley dilemma (see columns 555, 570, and others). In one version, a person is at a railroad switch with a lever that can divert the train from its current track to another. He sees a train approaching, and on its track five workers are sleeping and will die if it continues. He can divert the train to another track, but there lies one person who will die. Should he divert the train (the consequentialist consideration) or leave it on the current track (better to refrain from action)?

There it seems the dilemma is between an active deed that will cause the death of one person and passive inaction that will allow five to die. Does it matter that in the diversion case the one who dies does so due to my act, or should we compare only the outcomes, i.e., the numbers who die? By contrast, in a hostage deal the dilemma is somewhat different. In both options I am not performing an act that kills anyone. At most I release terrorists who may kill others by their own choice. That is also not certain to happen, and it may be that we will be able to protect against them. Alternatively, if I do not release them, even then I did not kill the hostage by my action; at most I did not save him.[1] It is hard, then, to say that there is an option of direct killing. Here the consideration should seemingly be purely consequentialist. On the consequentialist plane it is fairly clear that one may not release the terrorists, since that will lead to more deaths. Our deal can perhaps be analogized to a parallel dilemma regarding rescue: Suppose that in two places in a river there are drowning people. In one place one person is drowning and in the other three are drowning, and I can save either the one or the three. It seems clear that one should save the three, since again there is no “action vs. inaction” consideration here but only a comparison of outcomes.

The conclusion is that the discussion of deals releasing terrorists is unrelated to the above Jerusalem Talmud. There we were dealing with endangering my life to save another, whereas here we are dealing with saving one versus saving many. In such a case, it would seem that only the consequentialist consideration governs. The obvious conclusion is that we should assess the expected average harm from the deal, and if the expected number of future victims is greater than the number of people we are saving now, we should not make the deal. This yields a clear distinction between the Shalit deal, where one person was saved, and the current case where we had 240 hostages. If we look at what has happened so far, over two hundred female and juvenile prisoners were released in exchange for about a hundred hostages, and it is very doubtful that this release itself will lead to harm to more than a hundred Israelis in the future. This is unlike the Shalit deal, which proved to be very destructive.

Additional Considerations

However, the picture is not that simple. There are further intuitions on which people rely when discussing such deals. The threat in question is future, uncertain, and to undefined individuals (we do not know who will be harmed, if anyone).

The fact that it is future should ostensibly make no difference. The timeline does not play a role here. The fact that it concerns undefined individuals also ostensibly does not matter, since in the end we are talking about human life (as in the Talmudic discussion of one who throws a stone into a crowd, Ketubot 15a and parallels). But the fact that the harm is uncertain can indeed play a role, since while the expected loss is ten people, it is in principle possible that no one will be harmed at all. Moreover, the result that will occur depends at least partially on us (if we act properly—unlike on Simchat Torah—perhaps we can prevent it). It is true that on average this is what is expected to happen, but it is only an average. It is hard to liken this to a certain outcome of harm to ten people. True, we saw above that halakhically one treats expectation as a certain result, but that is not necessarily correct for our discussion.

It is important to understand that with regard to desecrating Shabbat, as in the case of the metal ember, it is reasonable to permit even when the danger is future, uncertain, and to undefined individuals, since there we are not comparing the loss of one life to the loss of another life, but rather a transgression versus loss of life. As is well known, any outcome involving loss of (significant) life justifies desecrating Shabbat; thus these distinctions do not necessarily change the ruling, and one should permit desecrating Shabbat in all such cases. But in our case we are comparing numbers of casualties, i.e., the risk to the hostages versus the future risks, and in comparisons between such alternatives the parameters I listed may carry weight.

Risk to Undefined Individuals

An aspect few consider relates precisely to the fact that we are dealing with a threat to undefined individuals. Suppose that in the Shalit deal there was a one-in-a-million risk to all residents of Israel. As noted, the expected loss is ten people, and ostensibly that is not justified for one hostage. But now consider the following: If we ask every person in Israel whether he is willing to take a one-in-a-million risk in order to save Gilad Shalit’s life with certainty, I think a reasonable person would answer in the affirmative. We take upon ourselves far greater risks for far less important goals (for example, we cross a street or drive a car), and saving a life certainly justifies it. If so, all citizens are willing, each one, to take a risk in order to save the hostage’s life, and that alone suffices to justify the deal, even if the expected loss is greater than the number of hostages saved.

Of course this is conditioned on the risk to each individual being small (which it surely is), and on the fact that we are dealing with undefined individuals. A person who is expected to face significant risk will not agree to that in order to save another’s life. The consent here rests on the fact that my own chance as an average Israeli citizen of being harmed is minuscule.

This would seem to assume the halakhic position of the Jerusalem Talmud that a person is permitted (even if not obligated) to endanger himself to save his fellow. Without that, the consent of each citizen is irrelevant, since he would be forbidden to do so. But I think that is not correct. Beyond the fact that the Jerusalem Talmud speaks of an immediate (and perhaps certain) risk—the risk is certain, not its outcome—more importantly, it speaks of significant risk. The level of risk at issue here (one in a million) does not enter the halakhic dispute, since it is not considered risk at all. Again, consider crossing a street or driving a car, which we would have had to forbid for any purpose other than saving life. Since these are small risks, this is not a level of risk that is forbidden to take. Note that the risk to each citizen is very small, even though the overall expected loss is the death of quite a few people.

True, we saw that in the case of the metal ember or clearing weapons in the army, forbidding the act places a very small risk upon many, and there we actually permit. Seemingly, in light of what I have said here, we should forbid in those cases as well, since the risk to each individual is minuscule. But as noted, in those cases we are not comparing numbers of casualties but a transgression for the sake of saving life. Permission to transgress is not conditioned on the risk level but on the expected result. Even when we formulate the permission in terms of risk level (for I wrote above that for small risks one does not receive permission to transgress), the reference is to the expected outcome, not to the probability. What permits the transgression is saving life (the outcome), and the probability/risk is only a measure of that.

In short, a consideration like the categorical imperative is not relevant when discussing risk versus rescue, only when discussing moral or halakhic “cost.” Questions of risk and rescue are inherently comparative and therefore consequentialist; thus the discussion revolves around an actual comparison, not a hypothetical test like the categorical imperative. And yet, my suggestion above is that when discussing such a deal and comparing the outcomes, it is appropriate to take into account the individual’s probability rather than the expected aggregate outcome. The reason is that we are dealing with uncertainty regarding undefined individuals. We ask each person, whose chance of being harmed is very small, whether he agrees or not.

It follows that the common argument—what makes the current hostages preferable to the lives of future victims (whether more or fewer)—does not necessarily hold water. At least according to the approach I present here, those comparisons are incorrect.

The Root of the Dispute: Metaphysics and Morality

We can now perhaps understand the nature of the debate over such deals. Proponents think that if there is consent by each individual citizen to take upon himself risk, one may carry out such a deal even when the expected aggregate harm is greater. Especially when we add the fact that there is a chance that the future harm will in any case be smaller (we are speaking only about an expected outcome). Opponents, by contrast, argue that we must compare the numerical expectations of the alternatives (the expected values of the options). If you re-read the debates, you will find this accurately describes many of the arguments now taking place.

This brings us again to the correlation I noted at the end of the previous column: for some reason this debate maps onto Right versus Left. I explained there that one of the fundamental differences between Right and Left lies in metaphysics (ontology): right-wing thinking tends to think in collectivist terms; in war it sees two collectives facing each other. From that point of view one weighs the expected harm to the whole collective, not the perspective of the individual citizen and whether he consents. Left-wing thinking, by contrast, places the individual at the center (the collective is a fiction), and thus for it the collective is merely the sum of citizens. Hence, if each individual citizen consents to such a deal, there is no barrier to carrying it out.

I wrote there that, on the metaphysical plane, my personal view is in the middle: every person wears two hats—he is an individual and also a member of a collective—and neither hat cancels out the other. In our case I am, in principle, inclined to accept such a deal that risks the future for the sake of the present, but that is because there is an individual element in the picture that we should not ignore. If the individual is willing to sacrifice himself (i.e., to take a risk) for the sake of another individual, the collective should not force him not to do so. This is, of course, only when the risk is at the level of the individuals themselves. Where the leadership foresees damage to the collective (not only that many will die, but that we will lose the campaign, forfeit deterrence, and the like), there is certainly room to coerce individuals and refuse such a deal even if individuals are willing. This is highly relevant to our current situation, where there is a heated debate over continuing to exchange hostages, and it seems the families of the hostages are in direct confrontation with the government (see the demonstration this past Saturday night).[2]

An Example: Identifying Marks in Lost-and-Found

There is a Talmudic topic in the laws of returning lost property that combines some of the elements raised here, and it may even add another dimension.

The Gemara (Bava Metzia 27b) discusses why it is permitted to return a lost item based on identifying marks (simanim):

Rava said: If you say that identifying marks are not biblically valid, how can we return a lost item based on identifying marks?

If returning based on marks is from the Torah, then the Torah told us to return with marks—apparently because providing the marks is clear evidence that this is the loser. But if biblically one may return only with witnesses, how can the Sages enact returning with marks? According to the view that biblically it does not work, the assumption must be that there is concern that the one giving the marks is not the true owner (he knows the marks some other way), and if so, returning the item to him based on marks would be theft from the true owner.

The Gemara answers:

Because it is agreeable to the finder that we return based on marks, so that when he loses something, it will also be returned to him based on marks.

That is, there is an assumption that the finder consents to return by marks, so that if he himself loses something and has no witnesses, he can get it back with marks and without witnesses.

Now the Gemara challenges:

Rav Safra said to Rava: Does a person do himself a favor with money that is not his?

What do we care about the finder’s consent? The loss is borne by the owner, not by him. The Gemara therefore revises the rationale:

Rather, it is agreeable to the owner of a lost item to accept that it be returned based on marks, since he knows he has no witnesses, and he says: “Not everyone will know its definitive marks, but I will provide its definitive marks and retrieve it.”

Now the logic is as follows: a reasonable loser agrees that we return based on marks, for if we were to return only with witnesses, every loser would risk having no witnesses and would not receive his lost item even though he can provide marks.

Essentially, there is a comparison between two policies: requiring witnesses only, or returning based on marks. If we return based on marks, there is concern that we may give the item to one who is not its owner (a false positive), but if we return only with witnesses, a situation can arise in which the true owner has no witnesses and will not get his item even though he can provide marks (a false negative). So which is preferable? The Gemara assumes that it is more likely for a true owner not to have witnesses than for a fraudster to provide correct marks.

It is, of course, possible that we do not weigh the two possibilities equally. Perhaps we prefer to give the item to a non-owner rather than fail to give it to the fraudster (“let not the sinner profit”), and then even if the probabilities are equal, or even the reverse of what I described above, it would still be preferable to return based on marks. Something akin to this was written by Maimonides (Sefer HaMitzvot, Negative Commandment 290) about the halakhic requirement for two witnesses and the rejection of circumstantial evidence in court:[3]

“To acquit a thousand guilty persons is better and more desirable than to execute one innocent person for a single day.”

So too in criminal law it is accepted to require proof beyond a reasonable doubt, because if there is reasonable doubt there is a (small) chance that we will convict an innocent person, and it is preferable not to convict a thousand guilty than to convict one innocent.

To grasp the meaning of this, consider a case in which the judges think there is an 80% chance that the defendant is guilty. If we acquit him (because 20% is reasonable doubt), there is a high chance (80%) that we acquitted someone guilty; if we convict him, there is a very small chance (20%) that we convicted someone innocent. Even so, in such a case we acquit him, and effectively accept an 80% chance of error, because we prefer a false negative to a false positive. If the same ratio applies in returning lost items, there is no necessity that the probability ratio be as I presented earlier.

For our purposes, we see in this sugya that society operates on the basis of the considerations of its individual members. If a reasonable person consents to the risk of losing his property, then that will be the general policy for the whole society—even though overall many may lose from it, and as we saw, it is not necessarily the case that those who lose are fewer than those who gain.

“Repeated Game”

Note an important difference between the situation described here and a hostage deal and the release of terrorists. In the case described here we are dealing with a game that repeats again and again. People constantly lose things, and the dilemma of whether to return based on marks must be viewed through the prism of a repeated game. I noted in the previous column that game theory distinguishes between a one-shot game and a repeated game. For example, in the prisoner’s dilemma (see it in column 122), people may act selfishly and perhaps profit from it, but they must take into account that in the next round their opponent will likely respond accordingly. Therefore the considerations change when I know we will play again. In lost-and-found, a person’s consent to the possibility that his item will be returned to a fraudster based on marks is nourished by the knowledge that in future cases he will be the claimant with marks and will want items returned to him.

What does this say about a hostage deal? One could argue that it is not expected to recur, and even if it does, it will recur once or twice and not on this scale. Not all of us expect to be kidnapped. On the other hand, I raised earlier the consideration that an individual agrees to take a small risk to save someone who is presently in certain danger. Now think, even if only hypothetically, about a situation in which this “game” repeats itself again and again. Next time it might be he who is the hostage, and then he will certainly want people to take a small risk to bring him back. This adds another consideration to what we saw above. Previously I argued that a person consents to take a small risk to save another—that was presented as an altruistic (though not dramatic) consideration. But now we can see that this consideration has a self-interested facet, like in lost-and-found. It is worthwhile for a person to take this risk as insurance for a case in which he himself is kidnapped. And even if this game will not repeat often and my chances of being kidnapped are small, there remains a kind of categorical-imperative-type consideration that directs me to consent.

The conclusion is that precisely the concern that more such cases will occur—which is usually presented as an argument against such deals—increases the legitimacy of making such a deal. This is, of course, not a contradiction. There is an aspect of repetitiveness that argues against the deal (it encourages additional kidnappings) and an aspect that argues in its favor (if anyone can be kidnapped, each person has greater motivation to consent to taking a small risk to save the current hostage).

Dependence on Numbers

How does all this affect the difference between various deals with different numbers of hostages and of those released (for example, Shalit versus the current deal)? The principled argument about risk to undefined individuals stands in both cases. We are dealing with a very small risk to each of us versus the certain rescue of one person or of two hundred people. Seemingly, if I accept the consent argument, there should be no difference. Those who oppose it compare expected outcomes, and there the differences may be significant. If you hold a left-leaning position that centers the individual, there should be no difference between the deals. Future risk is always preferable to present, concrete danger. Such left-leaning people would support the Shalit deal and also the current deal. Right-leaning people can (but need not) oppose both.

Even one who holds a middle position (two hats) can distinguish between the deals. In principle it is always permitted to take risk to save another person, but as I noted, in a case of many hostages or many released prisoners, the state enters the picture and does not allow the individual citizen to volunteer to take risk. There the individual consideration is subordinated to the collective one, because a deal involving large and/or disproportionate numbers has ramifications for the condition of the collective and not only for risks to individuals. Our national-strategic situation with 240 hostages is not the same as with one hostage. And the release of over a thousand dangerous prisoners for one hostage is not the same as a few hundred not-so-dangerous releasees for hundreds of hostages. From a collective, strategic perspective, quantity can turn into quality. It is no wonder that the war aims play a more central role on the right side of the political map, while on the left side the return of the hostages occupies the entire screen.

Summary

As noted, the considerations presented here do not necessarily lead to a clear conclusion in a given case. But they allow us to understand the various positions and considerations, to refute incorrect claims (or raise counterclaims), and, above all, not to act from the gut based on general slogans.

Now each reader can fill in each box with the weight that seems correct to him: what outcome is expected under each alternative; whether this affects our collective (political and security) situation; whether he is right-leaning, left-leaning, or subscribes to the “two hats” thesis; and thus formulate his own position systematically. In my eyes, the method and mode of thought are no less important than the conclusion you reach.

[1] This is not even comparable to Maimonides’ ruling (Laws of the Fundamentals of the Torah 5:5) about gentiles besieging a city and demanding “Hand over one of you.” Most commentators see handing over that person as an act of killing and think this is why it is forbidden. Here, the hostage is already in their hands and I am merely not rescuing him. Incidentally, even in that case I do not agree with that interpretive assumption, and I explained this in columns 254, 438, 567, and others.

[2] Incidentally, I strongly suspect that the hostages there are reciting agenda-laden messages about the hardship in captivity (presented in a highly tendentious manner, with full cooperation from the media, of course) in order to promote their preferred order of priorities: the return of the hostages above all (and thereby vilifying the government and its priorities).

[3] Incidentally, in terms of judicial majorities, we follow the majority in court. Here that approach is not applied, and we do not fear an error that will obligate the innocent. See Michael Vigoda’s article on proposals to change criminal procedure for precisely this reason. See also the distinction I made in column 327.