Philosophy and the Science of Human Nature

PHIL 181 - Lecture 14 - The Trolley Problem

Chapter 1. Introducing the Categorical Imperative [00:00:00]

Professor Tamar Gendler: So there's two things that we need to do in today's lecture. The first is to finish up our discussion of deontology, which was necessarily quite rushed. We're trying to do Kant in roughly a lecture. But I do want to get to the end of that discussion. And the second, which will allow you to use your clickers and express your opinions, is to talk through the structure of Judy Thomson's trolley problem paper.

So you’ll recall from last lecture that our goal in understanding the very brief selection from Kant's Groundwork to the Metaphysics of Morals that we read was to try to make sense of the three claims that he makes in the first book, first chapter, of that volume.

And those claims, to which I've now added some underlining, are the following. The first is the claim that in order to have moral worth, an action needs to be done from duty. And the distinction that Kant is making there is the distinction between doing something in keeping with duty, that is, something that conforms to what morality demands of you, and doing something not merely in keeping with, but also from duty. And Kant's picture is that the moral worth of an action is determined not merely by it being in keeping with duty. That's a necessary but not a sufficient condition. What determines the moral worth of an action is that it be done in keeping with duty for the sake of being in keeping with duty, that is, it be done from duty.

The second thing that Kant says, the second proposition which he seeks to defend in the Groundwork, is the claim that an action done from duty--that's the thing we were talking about in the first claim--an action done from duty has its moral worth not in the purpose that is to be attained by it, not in what the Greeks would call its telos, its goal, its aim, but rather in the maxim according to which the action is determined. That is, what determines the morality of the action on the Kantian picture is the description under which the action is performed.

Now, a number of you came to office hours yesterday, and we had a rather lively discussion of how it is that one goes about determining what things count as maxims. And I encourage those of you who are interested in that question to take an ethics course, where you can work through Kant's writings on this question in more detail.

For the purpose of our class, all we need to hold on to is the simple idea that what Kant is interested in here are acts under a description, and that that description is going to have to satisfy a certain sort of test, as we'll find out in a moment. So that's Kant's second claim.

Kant's third claim is that duty, which is the central notion to deontology--deon, duty is at the core of deontology--duty is “the necessity of an action done out of respect for the law.” And the idea here is that not only do you need to act with the goal of conforming to the law in mind, not only do you need to do so in a way that you articulate your actions as falling under that norm, but you do so because you take the moral law to be morally binding upon you, because you recognize that it is what rationality demands of you.

The moral law turns out be the law of governing your behavior that you set for yourself as a rational being. It is the only aspect of your behavior on the Kantian picture that isn't determined by the contingent forces of the world around you. It's determined by your recognition of your role as somebody capable of binding themselves to a law that they themselves set.

So as I said, the reason we were interested in these three principles was to get to Kant's famous formulation of the categorical imperative. And we closed lecture last time by meeting one member of the categorical imperative family in response to Kant's question: “What sort of law can it be, the thought of which must determine the will without reference to any expected effect so that the will can be called absolutely good without qualification?” Kant sets himself, as I said, this cliffhanger of a question, and answers it with the first member of the categorical imperative family. It's the will's “universal conformity of its actions to law as such.” only thereby, says Kant, can one act autonomously and not heteronomously.

What does it mean to act autonomously as opposed to heteronomously? Let's look at the words: autonomous, heteronomous. You'll notice that they both have here the word nomos, that is, law. And they distinguish the law to which you're subjected by saying that in one case, it's an auto-nomos, and in the other case, it's a hetero-nomos.

Auto-nomos. What could that be? Let's think of other words where we have this prefix auto-. How about automobile? What's an automobile? An automobile is something that is self-propelled. It is propelled by its own strength. To act autonomously, on Kant's picture, is to act on the basis of a law that you yourself have imposed. You are auto-nomos, subject to a law that comes from within.

By contrast, what is it to be hetero-nomos? Well, what is it to be heterosexual? What it is to be heterosexual is to be attracted sexually to individuals that have a different gender than you do. So what is it to be hetero-nomos? It's to have law given unto you that comes from outside of you, that comes from something different from you.

So Kant's picture is that autonomy, self-lawgiving, is possible only when the law to which you conform your behavior comes not from the contingencies of the world, but comes from within.

In this way, Kant is concerned with the same sorts of questions that Epictetus and Boethius are. Both of them are profoundly concerned with how it is that human freedom is possible, and Kant's picture is that human freedom becomes possible when you govern your actions on the basis of what you yourself decide to be, norms that you want to conform to.

In particular, when you conform your actions to the categorical imperative, which says, in the formulation that we see at the beginning of this section, that you should “never act in such a way that you cannot also will that your maxim”--there's your maxim again--“should become a universal law.” When you act in such a way that you don't take the contingencies of your situation into consideration--but rather that you think of yourself as one among any number of beings who in your situation would do exactly what you do--only then do you become free of the contingencies of circumstance.

So the picture is that in some ways, by stepping beyond the bounds of the contingent features of your experience, by stepping beyond the bounds of yourself, you thereby gain freedom from the contingencies of the world around you.

And Kant suggests that if you take this on as a picture, you will come to see that it conforms with the rules of rationality. So, he says, suppose you're confronted with a very particular case of an act that you want to perform under a particular maxim. The maxim you set for yourself is, “when I make a promise and it's going to be a pain in the neck for me to keep that promise, I'll break that promise.” That is, the maxim is, “it's OK for me to make lying promises.”

And Kant asks, suppose that you made, under that description, a promise that you didn't intend to keep. Can that maxim be universalized? Well, he says, suppose that it were. Suppose everybody, when they made promises, did so only with the thought that they would keep them when convenient and not when they were inconvenient. Were that to happen, says Kant, there would be no such thing as reliable promising. Why?

Well, because promises are like balconies. We don't step out on a balcony if there's a good chance that the balcony will break when we step out on it. And we don't step out on promises, on commitments that others make to us, unless we are close to certain that that promise will be preserved.

Chapter 2. Applying and Characterizing the Categorical Imperative [00:11:30]

So, says Kant, since the practice of promising would break down if everybody who found it convenient made lying promises, it is not in keeping with what the moral law demands of us that we make a lying promise. And, Kant suggests, this framework can be extended to all the kinds of duties that there are. There are, suggests Kant, two categories of obligation that we have towards ourselves and to others. We have duties to ourself and duties to other people. And, in addition, says Kant, we have perfect duties, things that we need always to do, and imperfect duties, things that we need sometimes to do.

In all four of these cases, says Kant in the reading that we did, we can see that the categorical imperative gives us guidance as to whether an action under a maxim is permitted. The action under the maxim is going to be permitted if it can be universalized, and it's going to be prohibited if it can't.

So if we ask the question, “is it alright to make lying promises,” and we say to ourselves, “suppose everybody made lying promises,” we discover that the act of lying promises is prohibited by the categorical imperative, because it can't be universalized.

Suppose we ask the question, “is it OK to commit suicide when feeling frustrated with the world?” And Kant says: suppose everybody did that. The practice, the convoluted argument would break down, because there would be nobody left to kill themselves. So goes the argument.

Suppose we ask ourselves whether we have a duty to cultivate our talents? Kant says: suppose nobody cultivated their talents. The world in which we lived would be one in which nobody would want to live. And consequently, we have a moral obligation to do so. Finally, he asks, “do we have an obligation to give money to those in need?” And asks again: what would happen if it were a universal law that nobody gave money to those in need? And again we discover a breakdown of an ordered world in which we want to survive.

Now, there's room for questioning--in fact, there's room for questioning all four of these derivations, though it's generally accepted that the lying promise derivation is the most effective of them. But let's look instead at what Kant says about them, if it were to turn out that these derivations worked.

Kant says, “these are some of the many actual duties whose derivation from the single principle above is clear.” "Is clear" is a bit of a stretch, but we can see how that derivation would go.

What does this tell us about morality? Kant says it tells us that when we take an act and try to determine whether it's moral, we need to check and see whether we're making an exception for ourselves. When we act, we need to be able to will that a maxim of our actions become a universal law. When we transgress, says Kant, we don't will that our maxim should become a universal law, but rather that the opposite of this maxim should remain a law universally.

So suppose you like to sit in the last two rows of this classroom, even though you arrive not late to class. Can you will that this become a universal law, or is this something that works for you only if others are willing to sit further in so that there's room for people on the stairs? Kant would say that the moral law demands of you that you move inward, because your sitting in those last two rows, despite your not late arrival, depends on other people doing something different.

Stealing depends on other people respecting the laws of properly. Your not paying the toll on the subway depends on other people paying the toll so that there's enough money to keep up the subway.

When you make an exception for yourself, says Kant, you violate the moral law. And we'll return to this at the opening of our discussion of political philosophy, when we talk about prisoner's dilemmas.

Now, I mentioned in passing that you had met one member of the categorical imperative family, that you had met what's sometimes called the formula of the universal law: that one should “act only in accordance with that maxim through which you can, at the same time, will that it become a universal law.”

Kant puts the categorical imperative four different ways for a number of reasons. One of which, he says, it that in certain cases, it's easier to see how to apply the categorical imperative if we frame it in a slightly different way. And I want to introduce you to the other three, largely because the second of these is going to play a central role in the second half of today's lecture.

So Kant claims that it is equivalent to saying that “you should act only in such a way that you can will your maxim to be universal” to say that “you should act so as to treat humanity, whether in your own person or that of any other, in every case as an end, and never merely as a means only.” Do not use yourself as a means to an end, and do not use others in your interactions with them merely as means. Treat humanity and all others as ends in themselves.

Equivalent to that, says Kant, is the formula of autonomy, which we've already talked about briefly. “Act so that through your maxims, you could be a legislator of universal laws.” Act in such a way that you are self-lawgiver with respect to rules that reason endorses. And finally, a rather complicated notion, sometimes called the kingdom of ends formulation, “act in accordance with the maxims of a member giving universal laws for a possible kingdom of ends”--a harmonious society in which everybody exists according to the laws that you give.

As I said, we have only about an hour's worth of Kant, so we won't focus on the third and the fourth. But I do want to call your attention to the formula of humanity, because as I said, it's going to play a central role in Judy Thomson's ultimate diagnosis of what may be going on in our intuitions about trolley cases.

Chapter 3. The Aim of a Moral Theory [00:20:16]

So let me close the discussion of Kant by trying to connect it back to the mini-unit of which it is a part. You'll recall that we began this section of the course on the seventeenth, that is, last Thursday, with thinking about consequentialism as a moral theory. And the question that I want to ask is: what is there that is common to the two concrete moral theories that we've taken a look at the beginning of this unit?

We've looked so far at some of the differences between consequentialism on the one hand and deontology on the other. But I think it's important, in moving on to some of their practical applications, to think about what they have in common. And what they have in common is that both teleology, consequentialism--utilitarianism in the particular form that we found it--and deontology prohibit first-person exceptionalism.

Kant says: my desire may serve at bases for willed actions only if I can, at the same, coherently will that others in similar circumstances would act in a way that I am choosing to act. I'm only allowed to do things that I'm going to assume other people are also allowed to do.

And Bentham, quoted in Mill, Bentham, the great grandfather of utilitarianism, says, “everyone is to count for one, no one for more than one.” Mill, in his greatest happiness principle, speaks of the happiness of all, not the happiness from the subjective perspective.

So the challenge of morality is that of viewing the world not from the stance of your own needs as the most central set of needs in the world, but rather from the perspective of your own needs as one set of needs among those of six billion equally sentient beings.

Now the problem for morality is that the tendency towards first person exceptionalism, the tendency to take one's own needs as more important than the needs of anybody else is perhaps the most widespread and pervasive psychological bias. And when we get to the unit on political philosophy after March break, we'll talk about ways in which social structures are put into place to help deal with this sort of tension.

Even in the passages from Mill that we read for last Thursday, Mill talks about what sorts of attitudes it's important to cultivate in individuals so that they begin to view the world from a moral perspective. In the selections that we read from book ten of Aristotle's Nicomachean Ethics, Aristotle asked: in what way should society be structured to make it easy for people to act morally? And in some ways, the question with which we're going to close the course--how does rational versus nonrational persuasion work, what are the roles of literary as opposed to argumentative representations of the good life--is a version of this dilemma, operationalized. How is it, given an inevitable human tendency to take one's needs as more important than others, how is it possible to structure society in such a way that the needs of all are met?

So that's what I want to say by way of the Kant. And in the remainder of lecture, I want to ask you to take out your clickers and enjoy the ride.

Chapter 4. The Trolley Problem [00:25:02]

So as you know, the paper which we read for today is a great and intricate paper by the philosopher Judith Jarvis Thomson written in the mid-1980s in response to an earlier paper by another mid-century woman philosopher, Philippa Foot. And what Philippa Foot and Judy Thomson are interested in in these papers is a systematic exploration of a number of cases which seem to evoke, in most subjects, pretty powerful intuitions about what the right thing to do is, but which seem to adduce intuitions the explanation for which is hard to systematize.

So as you know well, the first case, which we'll call trolley driver--there's the driver--is this. There's a trolley, hurtling down a track, in such a way that it’s going to kill five people. But it turns out that there is a second track onto which the trolley could be diverted, where there is only one person.

And the question is this. The trolley driver is driving the trolley. It's heading towards the five people in such a way that he is going to kill the five. Should he, is he morally required to, morally prohibited from, or perhaps neither prohibited nor required but rather just permitted, to turn the trolley onto the track where there is only the one?

Now notice that though this is framed as an idealized problem, the diversion of threats is something that decision-makers face all the time. Suppose there is an airplane that has become incapacitated, that's falling in such a way that it's going to land on a large city, and it would be possible to divert the plane so that it falls on a less populated area instead. Suppose that there is an illness which is taking the lives of many people, but if one quarantines those who are ill, causing that number to die, the rest of the population will be spared.

In case after case, we face dilemmas with roughly the structure of trolley. So though these cases are idealized, in the sense that we're granting ourselves that we know with certainty what the outcomes will be, it is, I think, not a useless exercise, even if our concern is real-world morality, to think through what the right thing to do is in these cases.

So let's start with the case of the trolley driver, driver of the trolley. The trolley is heading down the track in such a way that the driver will, with the trolley, kill five people if he doesn't turn it. He faces the choice of turning the trolley onto the track where there is the one. Question: Is it morally mandatory for him to turn a trolley from the five to the one? Is it morally permitted, but not morally mandatory, for him to turn the trolley from the five to the one? Or is it morally prohibited for him to turn the trolley from the five to the one?

OK. So let's see how the numbers come out. I'm going to write these down, because we're going to need them for later.

So 7% of you, very, very small number, think that it's morally prohibited for him to turn the trolley. The vast majority of you, close to two thirds, think that it's morally permitted, but not morally mandatory, for him to turn the trolley. And about 30% of you, roughly a third, think that he is morally required to make that turn.

Case number two: Transplant. You're running a hospital. Five people show up at the hospital, all of them destined to die, because one needs a lung, and one needs a heart, and one needs a kidney, and one needs a liver, and one needs a brain.

So they're all going to die, and you are the doctor. And into the emergency room walks a perfectly healthy young man who has a heart and a lung and a liver and a kidney and a really good, active brain. And if you were to chop up that man and give his parts to the five suffering individuals, you could save the five at the cost of the one.

Question: For the doctor, is it A, morally mandatory to chop up the healthy man to save the five, B, morally permitted but not morally mandatory to chop up the healthy man, or C, morally prohibited to chop up the healthy man?

So let's see how numbers come out. So. 85% of you think it is morally prohibited to cut up the one to save the five. 9% of you think it is morally permitted, but not morally mandatory. And 6% of you--off to med school--think that it is morally mandatory to chop up the healthy man.

So what's going on here? So Philippa Foot, who was the person who initially presented this juxtaposition, has a hypothesis. And her hypothesis is this. That in the trolley driver case, the choice that the driver faces is between killing one and killing five. Whereas in the transplant case, the choice that the doctor faces is between killing one and letting five die. And if we were to graph these on what I'll call the bad-o-meter, which tells us how bad things are, we would discover that letting five die is bad, but killing one is worse, and killing five is even worse.

And so this seems to give us the answer that since killing five is worse than killing one, then in the trolley driver case, it's OK for him to turn the trolley, but since killing one is worse than letting five die, then in the doctor case, it's not OK to chop up the one man. Because in the doctor case, you have to kill once to save five, whereas in the trolley case, the driver has to kill one in order not to kill five.

And that seems to accord pretty well with your intuitions. 93% of you think it's permitted, in the trolley case, to turn the trolley, whereas only 14% of you think it's permitted in the doctor's case to kill the one.

So it looks like this bad-o-meter is pretty well capturing the intuition of those of you in this classroom. And in fact, empirical studies that have been done on thousands and thousands of people throughout the world suggest that your intuitions are pretty much in line with the intuitions of most.

But there's a problem. Case number three: trolley bystander. Here's Jim. Poor Jim. Really bad luck. First he shows up in this Latin American town, and he's supposed to shoot some Indians. Now here he is, next to a trolley which is hurtling down a track, about to kill five people. But here there is, in the middle of the track, a switch that if Jim turns will cause the trolley to kill the one.

Question: For Jim, the bystander, is it morally mandatory for him to turn the trolley so that instead of the trolley hitting the five, it hits the one? Is it morally permitted, but not morally mandatory for him to turn the trolley? That's answer two. Or is it morally prohibited for him to turn the trolley? Let’s see how those numbers come out.

OK. And here are your numbers. 15, 70, 15. These are very, very similar to the distribution of answers that you gave in the driver case. In the driver case, 63% of you thought it was morally permitted, whereas here 70% of you think it's morally permitted. In the driver case, 30% of you thought it was morally mandatory. Here slightly fewer of you think it's morally mandatory, 15%. And in the driver case, 7% of you thought it was prohibited. Here 15% of you think it's morally prohibited. So there's a little change, but not a lot of change.

Here's the problem. Remember that in Foot's analysis of the case, we knew that letting five die was a little bit bad, that killing one was worse, and that killing five was worse than that. Trolley driver faced the choice of killing one versus killing five. In transplant, you face the choice of killing ones versus letting five die. But what's going on in the bystander case?

Well, in the bystander case, Jim, Jim of the bad luck, faces a choice between killing one--diverting the trolley onto the track in such a way that Jim kills that guy--or letting five die. Letting the trolley hit the five that it's going to hit inevitably.

But in contrast to the doctor case, where 85% of you thought it was prohibited to kill the one in order to save the five who would otherwise die, in this case, 85% of you think it's at least permitted to kill the one in order to let the five die. Let me do that again for you. 85% of you--watch the bad-o-meter--think that it goes the other way.

Now what's going on? We thought we had a solution to the problem. The solution to the problem that differentiated transplant from trolley driver was the distinction between killing and letting die. And all of a sudden, there's a whole bunch of you who seem to be saying about bystander that letting five die is worse than killing one. You must think that, or you wouldn't think that it's morally, at least, permitted for him to turn the trolley.

Moreover, stuff gets even worse. Suppose that the hospital case comes about as follows. Five healthy individuals show up at the hospitals, and a doctor--either because he's tired or because he wants to get the insurance benefits of which he is a beneficiary if there are a lot of sick patients in his hospital--poisons the five who show up, in such a way that one of them needs a liver, one of them needs a kidney, one of them needs lungs, one of them needs a heart, and one of them needs a brain. And so as a result of what that man has done, these five individuals will die.

And it's a few hours later, and he thinks, “Oh! I forgot about the categorical imperative! Shoot! What am I going to do?” And up shows a healthy individual, and he thinks, “Oh, Good! I've got a solution, here, I can chop him up. Heart, lung, kidneys, liver, brain, and I can save the five. OK.”

Question: For the doctor who has poisoned the five individuals who earlier showed up at the hospital, who now faces the option of saving their lives by killing the one, is it A, morally mandatory to chop up the healthy man, B, morally permitted, but not morally mandatory, to chop up the healthy man, or C, morally prohibited to chop up the healthy man? And let's see how the numbers come out.

All right. So. 82%, 11%, 7%. So your numbers here are almost identical to what they were in the original doctor case. There it was 6, 9, 85, here it's 7, 11, 82. Almost no difference.

But let's go back to our bad-o-meter. We are going to set aside the killing and letting die question, and just look at the kill one versus kill five.

So we know from trolley driver that killing one is pretty bad. But according to most of you, according to 93% of you, killing five is worse than killing one.

OK. Poison doctor. So here's what choice the doctor faces. He can kill five--right? He's poisoned them, and now they're going to die. Or he can kill just one--that one healthy guy who just showed up. And then the other five won't die. 82% of you told me that it was better for him to kill five than to kill one. Let me show you this again, on the bad-o-meter. 85% of you thought that it was better for him to kill five than to kill one.

So we have these two super-duper, excellent principles that seem to explain what was going on in the trolley case. On the one hand, that killing one was worse than letting five die, and then all of a sudden, bystander made us think, oh no, we don't have that intuition. And then we have the intuition that at least killing one was better than killing five, and the poison case made us rethink that as well.

Now, there's an obvious issue that may be making the moral difference here. There's a temporal difference between when the killing of the one and the killing of the five happened. And perhaps, says Thomson, perhaps that's what explains our intuition in the doctor case. Perhaps it is because the killing of the five has become a letting die, as the result of time, that it's misleading to describe this as a kill one versus kill five case.

But the temporal is not going to help us with the transplant versus bystander case. Those seem pretty clearly to be both cases where one faces the choice between killing one and letting five die. And whereas it seemed pretty clear to most of you in transplant that killing one was worse than letting five die, it seems pretty clear that for most of you in bystander, it's the other way around.

So the trolley problem is the problem raised by these dancing arrows. How is it that we systematize our intuitions about killing and letting die, given that they appear to come apart in these cases?

So Thomson suggests that whereas the transplant case gives us a choice between killing one and letting five die, as does bystander, there is a potentially relevant difference between them. And that is that in the case of transplant, you are using the one as a means. You're using the one as a way to achieve the outcome of saving the other five.

Whereas in bystander, when you're diverting the trolley onto the track where that one individual is, you're not using that individual as a means, and--oh my goodness, I told you we'd get back to Kant, and we have! What did Kant's formula of humanity say? Kant's formula of humanity said, “so act as to treat humanity, whether in your own person or that of any other, in every case as an end, and never merely as a means only.”

So maybe that's our solution. Maybe the problem in bystander is that since you're not treating him as a means, it's OK to kill the one. Whereas in transplant, since you are treating him as a means, it's not OK to kill the one, and consequently, you're morally obliged to let the five die.

Well, says Thomson, that can't be quite right. Suppose that you're Jim, standing next to the trolley. Trolley's on its usual path to kill the five. But here, instead of the straight track on which the one sits, there's a looped track, and the one is in the middle of the track in such a way that if you divert the trolley, it will hit him, thereby saving the five.

Question. In the case that Thomson calls “loop,” is it A, morally mandatory to turn the trolley, that's one. B, morally permitted, but not morally mandatory, or C, morally prohibited? So remember, trolley's heading down the track towards the five. You're Jim. The trolley's going to hit the five, or you can divert the trolley onto the track with the one, and because the one is there, you will cause the trolley to stop.

OK? So let's see how the numbers come out on this.

OK. So 68, 18, 14. Not a lot of difference. You've answered bystander in loop in almost the same way that you've answered all our other trolley cases. Generally the distribution has been 15, 70, 15. That was bystander. Here it's 14, 68, 18.

But notice that in this case, you were using the guy on the loop track as a means! You're using him as a means to your end. You're trying to stop the trolley by using his body. Kant didn't help us enough!

Restock. Let's take stock again, says Thomson. Perhaps some of the work is being done by some notion of rights. Perhaps what's going on in the transplant case, the one where you guys are not going to let the doctor chop up the healthy man, is that you would be violating that man's rights. And perhaps it's true that “rights trump utility.” That is, that when somebody has a right to bodily integrity, that takes precedence over the needs of the many. The utilities of the five that are going to saved.

And we'll close this lecture with the final example, one that's meant to test that hypothesis. And we'll begin next lecture by talking about some of the reasons that people tend to give this response.

So suppose now that instead of the looping track, there's a bridge. And suppose that on that bridge is our fairly large gentleman. And you are now faced with the following dilemma. The trolley's heading down the track. It's about to kill the five. And here's how Jim the bystander could stop it. He could push the fat man off the bridge, and thereby cause the trolley to be stopped in its tracks by his weight.

Question. For the bystander in fat man, is it morally mandatory to push the fat man, morally permitted but not morally mandatory, or morally prohibited to push the fat man?

And let's take responses, and I'll leave you with those numbers and a remark about them as our close. So let's see whether we get any shift in the fat man case.

My goodness! That looks awfully different. What is going on? So remember, our classic distribution is that we have roughly 70% here, and no more than 10% in the prohibited. All of a sudden, 78% of you think it's prohibited. What's going on? Cliffhanger! We'll talk about it on Tuesday.

[end of transcript]