Monday, April 14, 2014

The Fat Man in the Mirror

By Michael Dorf

The latest issue of the New York Review of Books includes a review by Cass Sunstein of the new book, Would You Kill the Fat Man? The Trolley Probelm and What Your Answer Tells Us About Right and Wrong, by David Edmonds. (Only the very beginning of the review is available for free online. The rest is behind a subscription paywall, so if you're not a NYR subscriber, you'll have to trust me for my description.) The title of the book refers to a famous hypothetical scenario in moral philosophy, a variant of the so-called "trolley problem." In this post, I'll very briefly explain the problem, and then suggest that Sunstein's otherwise fair-minded review omits (express) discussion of an important feature of moral reasoning.

A) In the classic trolley problem, you see a trolley hurtling out of control towards five people tied to the track. You stand by a switch. If you throw the switch, you can divert the trolley to a second track, but there is one person tied to the second track. Do you throw the switch, resulting in the death of the one but the sparing of the five?

B) In the fat-man variant of the trolley problem, you are standing on a footbridge overlooking a track, where you see a trolley hurtling out of control towards five people tied to the track. Next to you is a (very) fat man. If you push him off the footbridge, his mass will block the trolley before it can run over the five. Do you push the fat man, resulting in his death but the sparing of the five?

People have various moral intuitions in the two cases. Act-utilitarians think it morally permissible or even obligatory to throw the switch and to push the fat man, because in both cases, the action leads to a better result: the death of only one, rather than five. Some act/omission deontologists think it morally impermissible to throw the switch or to push the fat man because in each case, they would be deliberately acting in a way that causes a death, rather than merely failing to act to save five lives. Most interestingly, many people (including me) think that it is morally permissible to throw the switch but morally impermissible to push the fat man. The book, the review, and the substantial literature on "trolleyology" explore the reasons for these intuitions (and others regarding still more variants), in order to discern either what morality really entails or, as in Sunstein's case, to question whether this whole method of inquiry is likely to lead to moral truth.

I'll come to what I think Sunstein omits shortly, but first, I should give the standard explanation for what I called the most interesting pair of intuitions--that it's permissible to throw the switch but not to push the fat man. The explanation goes under the heading of a doctrine--originally traceable to Catholic theology but now part of the general philosophical discussion--called "double effect." The doctrine allows that it is sometimes morally permissible to intentionally take an action that has as its aim averting one evil, even though it will also cause some other evil.

Real-world applications of the double-effect doctrine include pain medication that has the side effect of killing the patient (which is distinguished from taking an overdose for the express purpose of suicide), causing "collateral" deaths of civilians in wartime (so long as the use of force targets military combatants and collateral deaths are not disproportionate), and more controversially, abortion.

In trolleyology, throwing the switch is an instance of double effect, because the person who throws the switch intends thereby to divert the train in order to save the five, even though doing so will have the regrettable side-effect of killing the one. Pushing the fat man is not an instance of double effect, because the fat man's death is not a side-effect of saving the five; it is the means of doing so.

As I said, Sunstein's review is generally fair, although he himself appears to be a rule-utilitarian. He thinks that if there is a moral distinction between throwing the switch and pushing the fat man, it is one that has to do with following general rules that, in the aggregate lead to the greatest good for the greatest number, even if, in particular cases, following such rules leads to bad results. Thus, he acknowledges that there is overall utility in following a rule that one should not push people off of bridges into the path of hurtling trolleys, and that perhaps the benefits of following that rule, as a rule, outweigh the costs of occasionally forgoing the greater good of saving trolley passengers.

But Sunstein ends on a note of skepticism. The moral intuitions we have, he says, are heuristics that developed in ordinary circumstances. They may be "misfiring" in odd cases like the fat-man problem. Put differently, perhaps the rule that a rule-utilitarian ought really to be following is not "don't push people off of footbridges in front of trolleys" but "don't push people off of footbridges in front of trolleys unless doing so will avert a greater harm." The fat-man problem, in this view, shows that our intuitions do not perfectly map onto the right rule-utilitarian rules.

Yet Sunstein's skeptical account of trolleyology omits any express discussion of what is usually a key feature of deontological reasoning and indeed, of moral reasoning more generally. Deontologists do not typically say that each of our moral intuitions generalizes into a true moral principle. After all, our intuitions sometimes tug in different directions. Thinking about Jim Crow may lead one to think that legal distinctions based on race are wrong, but then thinking about race-based affirmative action in public universities may lead one to modify that principle to say something like race-based legal distinctions that subordinate traditionally disadvantaged minorities are generally wrong, but that ameliorative race-based distinctions are more often acceptable. Or not. One might conclude that color-blindness is the right principle after all.

My point here is not to argue for any particular principle, but to note that trolleyology is not just about exploring our intuitions; it's about trying to find principles that reconcile our intuitions, and then, if that cannot be done, abandoning some of our intuitions in favor of more deeply held ones. It aims at reflective equilibrium, not naked intuitionism. Thus, the title of the post: We want to reflect on the fat man, not simply react to his case.

Sunstein titles his review "How Do We Know What's Moral?", and he implies in it that the method of interrogating our intuitions is not a reliable means of finding out. But he comes up short in two ways: First, his contention that our pre-reflective moral intuitions can be wrong is not an indictment of deontology, because deontology is not committed to the correctness of pre-reflective moral intuitions.

Second, to the extent that Sunstein's argument is more radical--i.e., to the extent that he thinks that our moral intuitions have nothing to do with moral truth but are simply heuristics that conferred survival value on groups of humans--Sunstein undermines not simply deontology but all of morality, including all forms of utilitarianism. After all, the claim that morality consists of maximizing utility is itself an appeal to moral intuition--a kind of meta-heuristic. If the intuition that it is wrong to push the fat man is simply a misfiring of a hard-wired rule that conferred evolutionary advantage, then so is the intuition that it is wrong to abstain from simple actions that could save more lives than they sacrifice.

The reduction of morality to heuristics is a kind of global moral skepticism that swallows all morality, not just deontology. Sunstein's (implicit) argument against double effect and deontology thus throws out the baby with the bathwater. If he really is a moral skeptic, then he would have no moral problem with throwing out a baby, but I don't read him to embrace moral skepticism, except perhaps inadvertently.


  1. Anonymous1:26 PM

    I am not skeptical of moral claims generally but I'm skeptical (perhaps even hostile) to abstract moral reasoning. My own intuition towards the trolly case is that I do not have enough information to make a decision and that given the fact I do not have enough information to make a decision I should do nothing. Another way of saying this is that I find the problem as posed too abstract and thus too artificial to allow for any insight into moral action in real life. In reality, a human always knows more than what is in the trolly hypothetical. He may know, for example, that the five men are all innocent and the one man is a convicted murderer. He may know that the one man is white and the five men are black. There are a significant number of possible real world permutations.

    It is quite common in the academic world to take the confusing morass of experience, induce general principles from it, and then deductively apply those principles to circumstances unrelated to the original induction. I think that whole approach to morality is a trick and a farce.

    I suppose another way of saying the same thing is--if I understand the OP correctly--is that what Professor Dorf calls "naked intuitionism" is the only responsible approach to morality.

  2. I don't think it follows that a moral skeptic would have "no problem" tossing a baby. Unless it is the case that this particular skeptic is a sociopath, he'd probably feel bad about throwing the baby. It's just that he wouldn't use moral language to describe a baby-tosser.

  3. Hi Mike,

    I'm sympathetic to your main point, which is that proper moral methodology requires reflective equilibrium, rather than unreflective acceptance of intuitions. But I'm not sure I agree with the more radical claim that radical moral skepticism follows from the claim that our moral intuitions are mere heuristics with survival value. Your argument that Sunstein is forced into radical moral skepticism relies on the assumption that moral theories themselves are intuitions (in the guise of meta-heuristics). This assumption is something that I predict Sunstein, and many utilitarians, will deny. Utilitarianism, they will say, is not an intuition. Intuitions are considered judgments about the moral permissibility/impermissibility of acting (or omitting to act) in particular actual or hypothetical cases. True moral theories, such as utilitarianism, they will say, are not even *supported by* intuitions: they are supported by reason, grounded in self-evident principles, such as that harm to fewer is a better state of affairs than harm to more. It's not that utilitarianism is some sort of meta-heuristic with survival value. The idea is that utilitarianism, like the Pythagorean theorem, is a general proposition that can be established by a rational proof.

    Here's an analogy for you. Before Dedekind and Weierstrass, the idea that there could be infinitely many points in a finite line segment seemed absurd. Many mathematicians and philosophers treated this as a kind of rock-bottom intuition. Then a bunch of 19th century mathematicians, following the lead of Newton and Leibniz, put the theory of infinity on a sound mathematical footing. The theory was so powerful that the mathematical intuition it contradicts was brushed aside, and rightly so. But the theory of Dedekind cuts and the epsilon-delta definition of mathematical continuity were not themselves intuitions (or even supported by intuitions). They were theories, useful theories, fruitful theories, based on coherent definitions and axioms. This is how many utilitarians think of their own theory.

    At this point, the debate shifts to whether moral theories are provable a priori. Proofs have been offered, but I don't find any of them convincing. The ball, now, is in Sunstein's court, to offer a proof that is better than the ones that have failed in the past. Good luck to him.

  4. A couple of additional comments.

    1. The initial trolley case was invented by Philippa Foot, and the bystander version of it (which is now standard) and the fat man case were developed by Judith Jarvis Thomson. These two great philosophers should get credit for these wonderful test cases.

    The trolley case was actually initially a challenge to the doctrine of doing and allowing (DDA), and only later was it used to support the doctrine of double effect (DDE). The DDA says that it is more difficult to justify doing harm than it is to justify allowing harm. But, in that case, shouldn't it be morally impermissible to turn the trolley onto the one in order to save the five? After all, turning the trolley is an action that leads to harm. In Foot's version of the case, you are the trolley *driver*, and when you drive the trolley onto the five or onto the one, you are committing an act (of killing) either way. And when two doings are compared, one of which leads to less harm than the other, it is permissible to choose the least harmful option (in this case, diverting the trolley). But Thomson cleverly changed the case by turning the trolley driver into a bystander at the switch (who can divert an empty runaway trolley by pulling a lever). Foot's reaction to this kind of case was to treat diverting an already existing causal sequence as a special kind of behavior, morally akin to allowing rather than to doing (in just the way that pulling the plug on a respirator is a kind of allowing to die rather than a form of killing).

    The fact that Foot's theory can accommodate the trolley case without giving up the DDA means that the DDE is not *needed* to account for our intuitions in the case. The most likely explanation for the moral permissibility of diverting the trolley, then, has to do with its consistency with both the DDA and the DDE.

    2. Edmonds' question "would you kill the fat man?" is exactly the *wrong* question. Who knows that you would do or what I would do? Maybe you would push the fat man and I would freeze. Maybe I would try to reason with the fat man as the five are crushed. But who cares? The question is not what *would* I do in that case. The question is what *should* I do in the case. Is it *morally permissible* to push the fat man onto the tracks below/to divert the trolley onto the one? THAT is the question.

    3. JimmyD needs to read the relevant literature. In all of these cases the potential victims are strangers, all of them innocent, none of them members of your family, all the same age, skin color, etc. The point is to use these cases to test such theories as the DDA, the DDE, and also consequentialism (one version of which is utilitarianism).

  5. I think the best contemporary philosophical defense of the role of intuitions in keeping with Professor Dorf's mention of the role of reflective equilibrium is Robert Audi's The Good in the Right: A Theory of Intuition and Intrinsic Value (2004).

    One glaring problem from my end is the attempt to draw conclusions about the nature of moral life or metaethical principles from what are clearly extreme moral situations, one that attempts to reconcile our existing moral sentiments and principles with situations of emergency and dire necessity, cases in which the DDE is relevant. The "exceptional cases" are just THAT, and we can hardly draw too many conclusions outside the scope of their specific moral specification. David Wiggins asks us to consider, for instance, a case in which a tyrant has kidnapped my wife and children and is threatening to torture and kill them unless I serve as his assassin and eliminate a group of his enemies. In this case, argues Wiggins, the DDE does not license me saving my loved ones by killing others (strangers) on behalf of the tyrant. More vividly and perhaps controversially, "Double Effect does not license us to buy off the terrorist who threatens to kill three thousand innocent people by giving in to his demand that we execute an innocent person. Nor, though, is it Double Effect that forbids this. In the case of these demands, it is other teachings that will encourage the agent to reflect that terrorism only breeds terrorism; and other teachings that encourage him to think that, if the terrorist's threat is carried out, it will not be the agent, who is responsible for the doing of the thing threatened, but the terrorist. If the agent has a duty in the matter, it may be said it is the same as everyone else's, namely to add his own voice to the clamour that demands that, as soon as possible, the underlying conflict be imaginatively and magnanimously resolved. The counsel does not derive from Double Effect, however.

    One last, metaethical if not epistemic point. Without in any way explaining how this may in fact be the case, I suggest we take an attitude to the various ethical theories proffered: Kantian or intuitionist ethics, utilitarian or consequentialist ethics, moral particularism (or 'ethics without principles'), and care and virtue ethics, to cite the most prominent or popular theories, such that we assume each view captures some measure of moral insight or truth (the specific measure or degree of which may differ, even strongly so). On this account, we would endeavor to appreciate when and where a particular ethical tradition or theory may relevant or applicable, illuminate something essential or incidental to ethical life in the main or in particular circumstances or situations. I suspect if we take this route, "trolleology" as it were, will prove to illuminate only one very small part of morality and the ethical life in general.

  6. Anonymous6:34 PM

    This comment has been removed by the author.

  7. Anonymous6:50 PM

    I deleted my post above because I did not see the post by Patrick O'Donnell before I started typing and his final paragraph expresses what I would express if I were a more articulate and less argumentative person. I endorse his perspective in its entirety.

  8. The trolley riders accepted the risk of riding a trolley. Why does the fat man have to pay for their risks?

  9. @Patrick: The vast majority of moral theories agree on the ordinary, non extreme cases. Otherwise they would lose much of their appeal. Both consequentialists and non-consequentialists can give you reasons not to lie, cheat, steal, coerce, harm, maim, kill, and so on. For example, the act-utilitarian will say that a particular lie under ordinary circumstances will produce more bad consequences than good consequences, the rule-utilitarian will say that the consequences of everyone telling lies under ordinary circumstances will be worse than the consequences of the implementation of a general anti-deception policy, and the Kantian will say that imagining a world in which a maxim of lying has been universalized and in which a person lies involves a contradiction in conception. So, in order to decide which moral theory is better than the others (or which is correct), we *need* to consider extreme cases. That is precisely why they are there.

    The fact that different moral theories agree on many cases does not mean that they all capture some measure of moral *insight*, even if it means that they capture many moral truths. Consequentialist reasons for action are completely different from non-consequentialist reasons for action. Insight in morality comes from understanding *why* (i.e., the reasons why) actions are permissible/impermissible. It's only if you've captured the reasons why that you count as having moral insight.

    No major moral theory of which I am aware would condone killing a group of innocents as a way of preventing a terrorist from killing one's spouse and children. Of course, if one *did* kill the group of innocents, one might be *excused* (because one is deciding under extreme duress). But excuses do not justify.

  10. I think the likely more common actual explanation for the "interesting" set of intuitions - throw the switch don't push the man - is that the Fat Man variation on the problem is irredeemably flawed as an example. There is just a far greater degree of reliability that throwing the switch will successfully save someone's life.

    It's just very easy to imagine oneself botching the push so that he falls and dies without saving anyone, or to question whether even a perfect push really would divert the trolley, compared to imagining throwing the switch failing to divert the trolley. Or even failing to get the guy over the bridge railing, in which case you're standing there with a guy you've just tried to kill looking at you is horror and rage.

    You can ask for a stipulation that pushing the man is as certain to divert the trolley as is pull throwing the switch, but you can't forcibly insert that assessment into respondents' heads and make them use to assess how they'd act.

    It's not clear to me that there's a way to clean up these problems with the hypothetical while preserving the clarity of the intuition it exposes.

  11. Anonymous3:33 PM

    Sam writes, "Insight in morality comes from understanding *why* (i.e., the reasons why) actions are permissible/impermissible."

    Some would say that but others would disagree. One of the reasons I approve of Patrick's comment is that my (non-moral) intuition told me that he grasped the fact that my first post in this thread was implicitly adopting the stance of moral particularism. Indeed, I would go so far as to say at a meta level what Patrick wrote is at least consonant with moral particularism. (I am not claiming that Patrick subscribes to moral particularism only saying a person who adopts the stance of moral particularism can agree with his perspective.)

    Moral particularism disagrees with Sam's statement. Moral particularism holds there are no reasons why a person adopts a moral stance because moral particularism rejects abstract moral reasoning. A moral person can--as a descriptive matter--explain the data that lead them to a specific moral decision in a specific circumstance but such a description doesn't explain *why* it only explains *how*. It therefore follows that what Sam calls "insight" and Michael calls "reflection" are both illusory narratives. To be blunt, they are fooling themselves when they claim that reflection leads to a "deeper intuition" or that reasoning leads to "insight". Neither /process/ reflects how people make decisions in the real world.

  12. @ Michael: when I lecture on this material, and I have done so regularly for years, the responses I get along the lines you suggest target the switching mechanism far more than they target problems with the pushing mechanism. Students at all levels (especially high school students, but also college students at all levels), even before they are willing to answer the question of moral permissibility, ask whether there is a way to fiddle with the switching mechanism so that the trolley jumps the tracks, gets stuck at the switching location, and so on. Of course, similar questions can be asked about the fat man, but almost everyone to whom I have put these cases do not suppose that there will be any trouble pushing the fat man onto the tracks below in sucks way that his body will stop the train.

    But in any event, after a little discussion, most everyone internalizes the stipulation that we know with certainty what will happen in each case. This is important to the intuition extraction process. And when this happens, reaction to the trolley case and to the fat man case is remarkably consistent: it is permissible to turn the trolley but impermissible to push the fat man. Really, it's time that we stop dismissing these cases, and consider very seriously what they have to teach us about morality and justification.

    @jimmyd: I am not a moral particularist, and i dont think that patrick is either, but this doesn't matter much here. There is no relevant *moral* difference between the how and why of morality. The how question can be given a psychological reading, in which case it is different from the why question. But that is irrelevant to the point at issue here.

    Even moral particularists accept that there are moral reasons. They just don't think that moral reasons can be generalized into universal exception less moral truths. You're barking up the wrong tree. Besides, moral particularism just has a terrible time explaining why intuitions about a wide range of doing/allowing cases and intending/foreseeing cases are remarkably stable and consistent. Only substantive moral generalizations can explain this kind of pattern, assuming that our intuitions reflect insight into the moral facts.

  13. Sam, I guess I don't see it quite that way and am not (as yet?!) persuaded by your argument, in part because it appears to trade on a model derived from science (e.g., conducting tests at the margins so as to converge on the best theory; I'm not opposed to such testing, but rather to the attempt to draw grand implications or conclusions with regard to what might be considered the 'better' or 'best' ethical theory generally). While there's no doubt some agreement as to what constitutes "common morality," the arguments, say, for and against the use of torture (not an 'extreme case,' given its political salience and urgency), or how to treat non-human animals, can come out very differently depending on ethical theories one draws upon (and, yes, they can arrive, on occasion, at the same conclusion). Perhaps for the (comparatively small class of) professional philosophers, extreme cases, as "thought experiments" that are highly improbable or almost surreal, are in fact used in the manner you cite, but I don't think they have that relevance or necessity for the rest of us: in short, we'd be foolish to prefer one ethical theory over another in our lives based on the results of such experiments.

    Wiggins' recent book on ethics well illustrates how the different ethical theories do tend to demonstrate unique insight or applicability to certain situations or cases. It reminds me of the manner in which Robert Goodin demonstrates the necessity of utilitarianism in public reasoning with government officials or democratic representatives, while in my private or intimate life, such as it is, I may find the perspective of virtue ethics more compelling and then use something along the lines of reflective equilibrium to assess the former arguments when voting. In the Hindu tradition, for instance, we see a complex mix of both something like Kantian ethics (dharma arguments), as well as consequentialist reasoning (in political treatises and some religious epics), and the tradition (for better or worse; rightly or wrongly, although I suspect the former) appears perfectly content to see value in both perspectives. And I'm inclined to see the Mohist tradition in Chinese philosophy as an attempt to combine consequentialist ethics with what, from the vantage point of Western ethics, would appear to be values also derived from non-consequentialist ethics.

    Well, just some initial and perhaps ill-considered thoughts. I suspect my prior commitment to at least the spirit of Jain epistemology precludes me from seeing matters otherwise: I simply do not believe we need arrive at a decision as to which ethical theory is, once and for all, "the best" (perhaps this is why, of late, I've been drawn to 'narrative ethics' as sketched, for instance, in the work of Grant R. Gillett; a home for ethics with many rooms). Perhaps naively or mistakenly, I'm constitutionally sceptical about both the need for and possibility of such a project.

  14. They may be "misfiring" in odd cases like the fat-man problem. Put differently, perhaps the rule that a rule-utilitarian ought really to be following is not "don't push people off of footbridges in front of trolleys" but "don't push people off of footbridges in front of trolleys unless doing so will avert a greater harm." The fat-man problem, in this view, shows that our intuitions do not perfectly map onto the right rule-utilitarian rules.fut 14 coins  elo boosting  fifa ultimate team coins  lol boost