Philosophy and the Science of Human Nature

PHIL 181 - Lecture 16 - Philosophical Puzzles

Chapter 1. Sunstein on the Trolley Problem Continued [00:00:29]

Professor Tamar Gendler: OK. So I want to start out by finishing up the discussion that we began last class about ways of thinking about the perplexity that the trolley case gives rise to. And you'll remember that the perplexity that the trolley case gives rise to is that there's an apparent asymmetry in our responses to the bystander case and the fat man case, even though both of them seem arguably to involve killing one in order to save five.

And we looked, last class, at Judy Thomson's response, which says, look, there's no asymmetry in the two cases, because when we reflect on the additional hypothetical case where there's a third track on which you, yourself, are standing, we come to recognize that it's not morally acceptable to turn in Bystander, just as it's not morally acceptable to push in Fat Man.

At the other extreme, we looked at Josh Greene's response, which was that just as it's morally acceptable to turn the trolley in Bystander, it's morally acceptable to push the man in Fat Man. And to the extent that we're getting differential responses in those cases, says Greene, it's due to the fact that the emotional part of our brain response mechanism is activated by the up close and personal nature of the fat man case, and as a result, we give an answer that he thinks remains morally unjustified.

And what we started to think about at the end of last lecture was a third possibility, which lies somewhere in between the Thomson and the Greene, though closer to the Greene. And that's Cass Sunstein's argument that though our responses differ, and perhaps differ in ways that will be impossible for us to change, the cases are the same, deep down. And he's inclined, though not as certain as Josh Greene is, to think that if we want the cases to come together, what we ought to do is to push the fat man.

And you'll recall that his argument there proceeded as follows. He suggested that in that in non-moral cases, it's uncontroversial that we make use of heuristics, and that those heuristics, though useful, frequently lead us to errors, and then went on to contend that just as this occurs in non-moral cases, so too does it occur in moral cases.

And we left at the end of last class thinking about what goes on in Sunstein’s argument that in moral cases people often use heuristics. And you’ll recall that he gave a couple of examples from Jonathan Haidt's work of cases where people were expressing moral disapprobation toward actions for which they could find no justification. So consensual incest between siblings, cleaning your bathroom floor with the American flag: People were inclined to find those morally problematic, and to find them morally problematic even when, if pushed, they were unable to articulate what moral rule those things violated.

And what Sunstein suggests in the paper is that in general, we can look at the heuristics and biases literature and see instance after instance where the framing of a case affects our response to the case in ways that they do in non-moral cases, in moral cases as well.

So you'll recall that in the third lecture, right when we were learning how to use our clickers, which I should tell you, we're going to use a bit in this lecture, so you should take out your clickers. When we were first starting to learn our clickers, we were presented with the famous Asian disease case, which is the case that runs as follows.

A terrible disease has struck 600 people in your town, right? So there's 600 people in your town who are destined to die. You are the mayor, and two courses of treatment are available, plan A or plan B.

And I asked half of you to look at the green side of the description, which says that plan A is the one where 200 people will live, whereas plan B is one where there's a one-third probability that 600 will live, and a two-thirds probability that no one will live. And the other half of you looked at the exact same plan, but described not in terms of who will live, but in terms of who will die.

So plan A says, 200 of the 600 people will live, which means 400 will die. And here plan A means 400 of the 600 people will die, which means 200 of the people will live.

Nonetheless, and this was our very first clicker response, the attitudes that you had towards the cases differed, whereas when it was presented as the number of people who will live, 66% of you went with plan A, and only 34% with plan B. When we inverted the framing, the numbers came out exactly the opposite. So 66 of you favored plan A in the green case, 64 of you favored plan A in the blue case. But plan A and plan B are mathematically identical.

So perhaps something like this is what's going on in the trolley cases. And this is not a real clicker example, but imagine you were presented with the following case.

A terrible trolley is hurtling down the tracks towards six people in your town. You are the mayor, and two courses for the trolley are available for you, plan A and plan B.

And then I present you plan A, one person will be spared, which means, of course, that five people will die. Or plan B, that one person will die, which means, of course, that five people will be spared.

And there's an inclination, I think, to go with plan B in the blue case and plan A in the A [correction: green] case. And this generalizes. Depending on who we're focusing on in these moral dilemmas, we have different responses to them.

If we think about Josh Greene's crying baby case, where you're locked in a basement with 19 others and your crying baby, surrounded by enemy soldiers who will kill you if you are found, the dilemma that Greene presents subjects with is, should you smother the baby, whose cries will call the soldiers with certainty to your hiding location and cause them to kill all 20 of you? So very much like the Jim and the Indians case, but with an even more painful premise.

If you focus your attention in that case to the experience of putting your hand over the mouth of your screaming child, it is virtually impossible to judge that as the thing that is morally required. But if you redirect your attention even a tiny bit towards the two year old next to you, and the four year old next to her, and the old man in the other corner of the room, all of whom will die if you don't take this action towards the baby, your response to the case shifts.

And the shiftiness in the direction of our attention is something that's going to be endemic to all of these kinds of cases. To some extent, we're able only to focus on part of the world at a time. And As a result of that, it's incredibly difficult to hold in focus in a way that makes them seem stable--these kinds of moral dilemmas.

So Sunstein's suggestion is that this phenomenon, whereby features that have got to be morally irrelevant--right? It can't be morally relevant to what's the right thing to do in the trolley case whether you frame it in terms of the number who will live or the number who will die. At least, prima facie, that doesn't seem like the kind of thing that could be relevant. You're making exactly the same decision framed in two different ways. How could that be what makes the difference?

Sunstein's suggestion is that the mechanism that underlies the phenomenon that I've just described happens over and over and over again, not just in hypothetical trolley problem-style cases, but all the time in the kind of moral reasoning that we engage in as citizens of a democracy, trying to make judgments about distributions of resources, trying to make judgments about what sorts of laws should be put in place to regulate or incentivize certain kinds of behavior.

So in each of the following four domains, says Sunstein, we very often focus on heuristics, that is, the surface features of the phenomenon, rather than the target attributes, that is, the thing that we ultimately care about. Remember I talked last class about putting a skin on your phone so that it's easily recognizable, that gives you heuristic access to which phone is yours. But of course that decoration on your phone is useful as a way of finding your phone only in so far as it tracks the target attribute that you care about, namely, finding the phone which has in it the phone numbers that you care about having. And when targets and heuristics come apart, we're in trouble.

So, says Sunstein, when we're thinking about risk regulation, that is, what do we do with the fact that as human beings, lots and lots of the stuff we do has the potential for causing harm, but we don't want to spend our lives wrapped in large amount of Styrofoam, moving very slowly through the world so as not to bump into things. Given that we are willing to take risks, how is it that our tendency to use heuristics interacts with our regulation of them?

In cases of punishment, and this is the first topic that we'll turn to after break, Sunstein thinks we use heuristics in ways that cause us to behave in counterproductive ways in punishing both individuals and aggregates. In our hesitation to make certain kinds of choices in the area of reproductive medicine, thinks Sunstein, we risk mistaking the heuristics for the targets. And in taking the act-omission distinction so seriously, we risk mistaking heuristics for targets.

So we'll turn to the issue of punishment right after break, and we'll turn to the issue of act-omission in the later part of the lecture. What I want to do right now is to run through three examples of risk regulation via Sunstein's analysis. And the third of these-- I'm actually really curious, and so I want to see how the clicker numbers come out.

Chapter 2. Risk Regulation and Heuristics [00:11:08]

So Sunstein points out, and it seems to me that he's exactly right, that people are more likely to condemn a company when their behavior is described in ways that involve certainty than in ways that involve risk. So take company A, which produces a product that 10 million people use, which will kill 10 people. Of the 10 million people who make use of this product, 10 of them will have a reaction to it of a kind that will cause them to die. And the cost of eliminating that risk entirely would be $100 million.

There is a feeling, an inclination, at least, on the part of many, to think that the company ought to spend its money getting rid of that risk; that it's unacceptable to produce a product where 10 people are going to die. By contrast, if you frame the case in terms of probabilities, that 10 million people use the product, that it produces a risk of death of one per million, and the risk elimination is exactly as costly, this is the sort of thing that we allow all the time. Without this sort of risk tolerance, there would be no technological innovation, and most of the goods and resources that all of us have come to take for granted would never have come to be.

So Sunstein's contention here is that though the target attributes are identical in the two cases, in this case, 10 people are going to die, and saving them would have cost $100 million. In this case, 10 people are going to die, and saving them would cost $100 million, the target attributes are identical. In both cases, 10 people die, and saving them would cost the amount specified.

The heuristic attributes differ. This one is framed in terms of certainty; this one is framed in terms of risk. And we have a very good heuristic that goes like this. If 10 people are going to die from what you're doing, don't do it.

And Sunstein's contention is that the asymmetry in our response to these cases is irrational. Indeed, if we lifted this [clarification: second case] one to a risk of two deaths per million, and had this [clarification: first case] one with a certainty of 10, people would still be inclined to go with the first choice--sorry--to condemn the first choice, even though in that case, the second choice is clearly the worse one.

So as a result of mistaking the heuristic attributes for the target one, we make mistakes in what sorts of behaviors we permit. Sunstein thinks that this is what's going on in the case of emissions trading--cap and trade--of which he was an early advocate. In the model of emissions trading, polluters get given a license to pollute n units of pollution into the air, and those licenses then get to be traded on the market in such a way that, arguably, there's less pollution at lower cost.

Let's grant Sunstein the economics there. Even so, there is resistance to cap and trade. Because even if we're willing to concede that the target attribute--namely, that we've reduced the amount of pollution--is present, the heuristic attribute--“People are paying to pollute? You shouldn't be able to pay your way out of serious wrongdoing!”--strikes us as problematic.

Now, it's an interesting phenomenon that resistance to this sort of reasoning happens depending on the content [correction: context] from both the right and the left. So there is resistance to commoditization of things from the left, and there is resistance from the right to certain other sorts of framings that suggests that their responses in cases, for example, of reproductive technologies like cloning, are due, says Sunstein, to the heuristic, “don't play God.” And when confronted with the suggestion, “you're just using a heuristic there,” both sides respond with hostility to the smarty-pants academic analysis.

In the 1970s, it was common for advocates of the buildup of nuclear arsenals to make appeal to a notion called "mutually assured destruction" that we'll talk about when we talk about the prisoners' dilemma. The basic idea is that if both sides have enough weapons to knock the other side out, then neither will make use of them, because the deterrence function is too great.

There was resistance to that analysis from the left, because it felt too clever. There is resistance to the sort of analysis that Sunstein's posing here from both sides, because it cuts against the idea that we are introspectively transparent in such a way that our judgments are indicative of the things that we care about.

So the last example that I want to give you from Sunstein is our poll. Sunstein hypothesizes-- and are your clickers working? Sunstein hypothesizes that we are more uncomfortable being harmed by things which are meant to protect us than being harmed by things which aren't meant to protect us.

And he suggests that there is data showing that if people are given a choice between two cars-- the first car is one where there's a 2% chance if you're in an accident that you'll be killed by the steering wheel. And the second is a car where there's a 1% chance if you're in an accident, you'll be killed by the steering wheel, but in addition, there's a 1/10 of 1% chance that the airbag will kill you. And the question is, which car do you choose? The one where there's a 2% chance that you'll be killed by the steering wheel, or the one where there's a 1% chance that you'll be killed by the steering wheel, but a 10th of a percent chance that you'll be killed by the airbag, which was meant to protect you. And let's see how those numbers come out.

I have to say, I'm doing this poll because my intuitions didn't line up with Sunstein's, and I'm curious whether yours do. OK. Let's see how the numbers came out.

So 15% of you want to buy a car A, and 85% of you want to buy car B. So 85% of you are doing what is the statistically rational choice. But a good proportion of you are willing to risk greater harm so as to avoid this feeling of betrayal by that which is meant to protect.

So Sunstein's suggestion, just to sum up, is that in moral reasoning, frequently, we substitute heuristic attributes for target ones. And to do so is a mistake.

So what do the three responses to the trolley problem that we've considered suggest? Well, what Thomson says is this. She says, reconsidering our intuitions in light of alternative cases, like the alternative bystander case where you imagine yourself to be one of the people on the track, reconsidering our intuitions in light of alternative cases can lead to shifts in our assessment of those cases. And those shifts in our responses, she thinks, reveal something morally significant. We can learn from the contemplation of those specific cases what it is that morality demands of us.

Greene and Sunstein, by contrast, contend that our intuitive responses to cases frequently track features that are morally irrelevant, and that as a consequence, those features fail to reveal something morally significant.

The question is this. Is any of this a problem for Mill and Kant? Let's look back to the very opening pages of Mill's treatise on utilitarianism.

He writes there, and I didn't have you read this passage so there's no reason you should know that he says it. "Though in science, the particular truths precede the general theory, the contrary might be expected to be the case with a practical art, such as morals or legislation."

So in science, we look at particular instances. We discover we drop this object and it falls with acceleration A, we drop this object and we discover it falls with acceleration A, we drop this object and we discover it falls with acceleration A. And from that, we conclude that the law governing fall of bodies is that they fall with acceleration A.

So “though in science, particular truths precede the general theory, the contrary might be expected to be the case with a practical art, such as morals… A test of right and wrong must be a means of ascertaining what is right or wrong…and not a consequence of already having ascertained it." “The difficulty” (of building a theory out of judgments) “the difficulty is not avoided by recourse to” what is sometimes now called the moral sense--"a natural faculty," says Mill, "that discerns what's right or wrong in a particular case in hand, as our other senses discern the sight or sound actually present." So as if you can see whether a case is morally wrong. "Rather," he says, “moral reasoning, moral understanding, is a branch of our reason, not of our sensitive faculty. The morality of an individual action…is a question of the application of the law to an individual case…As a result, whatever steadfastness or constancy our moral belief has attained is due to the tacit influence of this reflectively available standard."

So Mill is building theory out of theory, not theory out of cases.

Kant. "Worse service cannot be rendered to morality than that an attempt to be made to derive it from examples. For every example of morality must itself first be charged according to principles of morality in order to see whether it is fit to serve as a model."

We have here, in some ways, embodied the dialogue of this course. To what extent is our capacity for rational reflection the best way to get at answers to questions that we care about? To what extent is our capacity for emotional response, for sensation, for instinctive judgment on the basis of presentation of particular cases, indicative of answers to the questions we care about?

So that closes the discussion of the trolley cases. And what I want to do in the second half of lecture is run through two kinds of puzzles which persist regardless of which of those attitudes that we take.

Chapter 3. Ducking vs. Shielding [00:23:26]

So the first is something that I presented you with as promissory note in the very first lecture, because this is one of the most fun papers that we're reading all term. And this is Roy Sorensen's paper with [Christopher] Boorse on ducking and sacrificing.

And you'll remember, that was the weekend that the senator from Arizona had been shot, so I couldn't do it with bullets. So you'll remember that the case I gave you is, you're standing in a line. You're the yellow guy. And a bear is rushing towards you. And you jump out of the line, and the bear eats the person behind you. Contrast that with the case where you're standing in a line. You're still the yellow guy. A bear is rushing towards you, and you reach behind you, pick up the guy, and put him in front of you, and the bear eats him.

The first of these is the classic what's called ducking case. That is, you're in a situation where there's a harm moving in your direction. You move out of the harm's way, and the harm hits someone else instead. The second is a classic sacrificing case. There's a harm moving towards you, and you make you use of another person as a shield.

So to duck is to avoid harm, thereby allowing it to fall on someone else. To sacrifice is to avoid harm by bringing about that the harm comes to someone else if you use that person as a shield.

And this is analogous to the act-omission distinction, one that we've already looked at, but it's wholly within the realm of acts.

Now what Sorensen and Boorse bring out in their article is how resilient this phenomenon is, regardless of how you mess around with the framing of the cases. So they give you the example of the mall gunman. There's a bullet coming towards you, and your choice is to leap aside or to pull somebody in front of you as a way of avoiding the bullet.

There's the speeding truck case. You're in a row of cars. There's a truck coming up behind you in such a way that it's going to crash into you. And you have one of two things that you do. In the first, you switch lanes, and the truck hits the car that was in front of you. In the second, you signal to a car that's behind you to switch into your lane, and the truck hits him.

There's the terrorists case. You're on an airplane. Libyan terrorists--quite timely to be speaking about Libya--Libyan terrorists, in this example, come onto your airplane and threaten to kill all Americans. You're a U.S. State Department representative, and on your briefcase is a U.S. State Department sticker, and the terrorists are coming down the aisle, and they're about to shoot you. Two possibilities. One, you cover your sticker with a Libyan airlines sticker, so they skip you and go and shoot the woman sitting next to you, the next one in line. The other, you switch briefcases with the person next to you, and so they shoot her instead of shooting you.

Or the sinking boats case. You're in an ocean. You're trying to get your signal, your boat is sinking, the guy-next-to-you's boat is sinking. You're trying to signal to an airplane above you to come and pick you up. And you can do one of two things. You can strengthen your signal, right? Make your light really strong, and then the airplane will come and rescue you. Or you can jam the signal of the other guy, making your signal relatively stronger so that the airline comes and picks you up.

Sorensen gives case after case about this. If beetles are eating your roses, it's OK to put beetle repellent on your roses, which will cause them to go over to your neighbor's house, but it's not OK to put beetle attractant on his roses.

We have this strange tendency, over and over and over, to think that ducking is OK and that shielding is not.

Now the perplexity that Sorensen and Boorse consider is that it seems like there's no systematic way to account for these kinds of discrepancies in intuition. So you might think: Look, the problem with these cases is that when you tie up your opponent's feet, when you're trying to outrun the bear, or when you push him in front of you in the shooting cases, you interfere with fair competition. And that fair competition is what matters in these sorts of circumstances.

But of course, there are plenty of these circumstances where the competition was unfair to begin with. And nonetheless, it seems problematic. Even if the guy whom you're trying outrun the bear in front of is a much slower runner than you, so that you were certain to win, it still doesn't seem OK to tie his shoes together.

The fairness of the competition doesn't seem to be what's driving the intuition. So perhaps, they say, it's that in each of the shoving cases, what you do is somehow an included wrong. It's wrong to pick somebody up and carry them in front of you. Whereas, it's OK just to duck down so that something hits them. It's wrong to steal somebody's briefcase. It's wrong to jam somebody's signal.

But, they point out, it seems just as bad to put the person in front of you in a friendly way by saying, “wouldn't you like to see the beautiful view?” as it does to pick him up and put him in front of you. It's just as problematic, they suggest, to scare somebody into jumping off a cliff by yelling, "E equals mc squared!" to surprise them, as it is to cause them to jump off the cliff by yelling a racial epithet.

The included wrong doesn't seem to be what's explaining our response. So, too, and I'll leave you to read these responses on your own if you haven't had the chance already, so, too does the act-omission distinction or the doing-allowing distinction not seem sufficient to do the work. So, too, does the idea that what matters is if you were the locus of a causal chain, the originator of some sequence of causality. So, too, does the doctrine of double effect not seem to account for all of these cases. So, too, does appeal to Kant's notion of rights in contrast to utilities not seem to explain all of these cases.

So Sorensen and Boorse somewhat reluctantly consider a conclusion of skepticism. Which is roughly, this is a perplexing feature of our psychology.

But we, having listened to the first half of this lecture, have one more alternative explanation. And I don't promise that it will work in every case, though it seems pretty promising. Which is that what's going on in the ducking and shielding cases is the overapplication of a heuristic. In general, it does seem like moving out of the way of a harm is not a bad thing to do, whereas putting somebody into the track of a harm is a bad thing to do. So perhaps these first set of puzzles can be explained by means of heuristics.

Chapter 4. Moral Luck [00:31:08]

In the last fifteen minutes of lecture, I want to focus on a set of puzzles which, I think, can't. And for these, you'll need your clickers.

So let's start with four drivers. The first of them, Lucky Albert, or Lucky Alert, does the following. He gets into his car. He has his car in perfect condition. He pays attention at every light. He drives in an extremely safe way. And at the end of the day, gets home from work. That's it. That's lucky alert.

Question: When Lucky Alert drives home, setting aside whether he has his mistress in his car with him, setting aside whether he's bought a car that has a high rate of emissions as opposed to buying a Prius. In driving home, setting aside all the other things that Alert might have done morally wrong, did he do something morally blameworthy, driving home from work, having fully fixed his car, and doing no harm to anyone along the way?

So this is not a trick question. So if you think Lucky Alert did something morally blameworthy, setting aside all the things that are morally blameworthy about driving a car, push one. Whereas if you think he didn't do anything morally blameworthy, push two.

So what you're judging is, is driving home from work, all things considered, if nothing bad happens, a morally problematic thing to do? And let's hope-- OK.

So there's always that 5%. Those anti-car crowds. You're the ones going to med school and chopping up our poor healthy guy in the waiting room. 95% of you think Lucky Alert did nothing morally blameworthy.

Let's meet Lucky Alert's twin brother, Unlucky Alert. Here's what Unlucky Alert did. Exactly what Lucky Alert did. Except as he neared his house, a child ran out in front of his car and he hit the child. OK? Unlucky Alert did exactly what Lucky Alert did. Left work, checked his tires, stayed alert the entire time, drove at safe and proper speeds. But due to bad luck, on his way home killed a child.

Question: Did Unlucky Alert do something morally blameworthy? If yes, push one. If no, push two. And I'm going to write down the numbers on the first case, which were 5 and 95. OK. So let's see how the numbers come out on this.

Here, 81% of you think he didn't do something morally blameworthy, but we're up from 5% to 19% on people who think he did do something morally blameworthy.

Let's turn to our third case. Here's Mr. Lucky Cell Phone. Here's what Mr. Lucky Cell Phone does. He gets into his car and starts driving home from work. And on his way home from work, he talks on his cell phone, but you know what, nothing else happens. And he gets home from work having harmed no one.

Question: Did Lucky Cell Phone do something morally blameworthy in driving home from work talking on his cell phone? And let's see how these numbers come out.

OK. So your verdict here. 78% of you think Lucky Cell Phone did something-- you guys, I don't believe you. I mean, you're anticipating the next case! All of you talk on your cell phones when you drive all of the time, and you don't think of yourself as doing something morally blameworthy!

OK. These are not valid data. This has to do with where they are embedded in this experiment.

All right. So since you've already answered question four, let me just ask it of you. Unlucky Cell Phone drives home from work while talking on his cell phone. Child runs out in front of his car, and--OK. Question. During his drive home, did Unlucky Cell Phone do something morally blameworthy?

And let's see how the numbers come out. All right. Let's see where Unlucky Cell Phone's big red line comes out. OK. So now we've got a complete shift from the original one, and in fact different from our previous case.

OK. So what these examples demonstrate is a phenomenon known as moral luck. We have two people here, Lucky Alert and Unlucky Alert, who do exactly the same thing, but Unlucky Alert's actions caused the death of an innocent victim. And whereas only 5% of you think Lucky Alert did something wrong, 19% of you think Unlucky Alert did something wrong.

Here we have, in similar fashion, somebody who in a very slight way has taken a risk, which in this case had no bad consequences and in this case had very severe bad consequences. And 92% of you condemn Unlucky Cell Phone.

The phenomenon that this illustrates is a phenomenon known as moral luck. Cases where an agent is assigned moral blame for an action or its consequences, even though the agent didn't have full control over that action or its consequences.

Right? It's not the case that Unlucky Alert or Unlucky Cell Phone wanted the child to run out in front of his car. It's not the case that Unlucky Alert or Unlucky Cell Phone could have done it anything different at that moment. The child was in front of the car, and the car hit the child.

Moral luck is perplexing because we seem to have two competing commitments when we think about moral responsibility. On the one hand, we seem to accept something which we might call the control principle: that moral praise and blame shouldn't be assigned in cases where the action or the consequences lie beyond the agent's control.

And I can see that many of you subscribe to the control principle, because 81% of you thought that the unlucky alert driver did nothing morally wrong, even though he killed a child with his car. And the reason you're inclined to think that he did nothing wrong in that case, I suspect, is because your judgment in that case, as Mill said, is regulated by a principle to which you tacitly subscribe. Namely, something like the control principle.

It's intuitively plausible, says Nagel, that people can't be morally assessed for what's not their fault, or for what's due to factors beyond their control. If you bump into me, and I trip, and I accidentally fall on the red button that causes the nuclear war to start all over the planet, it's not my fault. I mean, it's really a terribly bad thing that the planet is destroyed, but I just tripped.

By contrast, and directly in competition with the control principle, it seems, as the moral luck principle states, that in some cases, moral praise and blame should be assigned even where the action or consequences lie beyond the agent's control.

The difference in your responses between the lucky and the unlucky cases indicate the degree to which you tacitly subscribe to that. So you went from 95% of no blameworthiness to 81%. So you shifted--15% of you shifted your view as a result of something beyond his control. In the cell phone case, again, roughly 15% of you shifted your view.

The problem is that both of these principles are incredibly difficult to let go of. The control principle relies on the following kind of reasoning. In general, we have a pretty good sense of what kind of factors increase the blame- or praiseworthiness of an action. In general, if an act is voluntary, that is, if you've done it not out of coercion and not out of mistake--right? If you specifically chose to perform the action that you performed--then you get more praise for doing it if it was a good action, and more blame for doing it if it was a bad action.

Likewise, if you had full information, if you were aware of its likely consequences, you knew that that was water, or you knew that that was cyanide that you were giving the person to drink, it increases the degree of praise- or blameworthiness.

And these are pretty robust responses that fall out not merely of our analysis of cases, but also out of our understanding of the principles that seem to underlie moral responsibility.

Correspondingly, it seems like the absence of those features decreases moral blameworthiness or praiseworthiness. If you do something under coercion, if you do something accidentally, it's less your responsibility. And if you do something out of lack of information, if I, fully thinking that I'm giving you something totally healthy, end up giving you something that harms you, we tend to think that the degree of blameworthiness is mitigated.

The control principle simply says that without a difference in these factors, how could there be a difference in blame- or praiseworthiness? If we hold these factors constant, we must be in a situation where there's no difference in moral responsibility.

By contrast, the moral luck principle is also really forceful. It seems undeniable that there are cases where we assess moral praise and blame in the absence of control. The driver case was one. One of the cell phone users hits the child, the other one doesn't. The first is morally blameworthy.

I leave the stove on in my house, or your house. I go to visit you, and I leave the stove on in your house. I go out for the day. When I'm unlucky, it causes your house to burn down. When I'm lucky, it doesn't. It seems, even if you think both were bad things to do, a much worse thing to leave on the stove and burn down your house than to leave on the stove, simpliciter.

Nagel gives the example of leaving the bath running with the baby in it. An irresponsible thing to do, but immeasurably more problematic when the baby drowns as a result.

Or the case where you and I have similar characters. I stay in Germany; you don't. It's the 1930s. I become a Nazi; you live your life in a way that makes no moral demands of you.

So there's three kinds of responses that we can give to moral luck cases. We can give a rationalist response. We can say, luck simply can't play a role in moral evaluation. And we can either take the extreme that, one might think, a pure Kantian takes, that all the agent is responsible for is his will, and those things over which he has full control. Or you can take what might be an extreme Millian version. That the agent is responsible for all the consequences of his actions, and that the attitude makes no difference.

You can take an irrationalist attitude towards this. You can say that luck can play some role in moral evaluation. Or, though I think this is ultimately difficult to maintain, you can say that as a matter of fact, we never know how responsible somebody is for an action until we see what its consequences are. That when I hypothesized that these cases were identical, I was idealizing in an illegitimate way.

Now, it seems like that third response might work for the classic cases of moral luck which I've been describing, cases which we would call resultant luck, where there's luck in the outcome of the action. I perform an action, and it happens to go awry in a way that I didn't expect. That's one class of cases of moral luck.

But it's harder to see how we can use that explanation for some of the deep and profound instances. So take constitutive luck. Some of you were born with genes that make it easier for you to behave in altruistic ways, and some of you weren't. Some of you were raised in families that were supportive of certain kinds of moral outlooks, and some of you weren't. Is your character that resulted from those features something for which you are responsible, and if it's not, how is it something with respect to which moral praise and blame can be assessed?

Take circumstantial luck, which Jonathan Shay discussed in our Achilles in Vietnam--luck regarding the agent's surroundings. Sometimes the circumstances you're in either create or reveal otherwise hidden features of your character. Does that mean, since they are in part a matter of luck, that you are not thereby morally responsible for what you did?

Finally, if we start thinking about our actions from the perspective of free will, it becomes hard to carve out any space in which we're responsible for what we do. It's a general fact about the world that actions and consequences are in general determined partly by features outside, or at least outside the control of the agent.

So we start thinking about why it is that we respond that way in Trolley, and it turns out it's because the emotional part of our brain is lighting up. But why is that happening? Well, that's happening because of blood flow, happening in a certain way in our brain. And why is that happening? Well, the blood is flowing in a certain way because of what certain kinds of molecules are doing.

And as we think this through, the area of genuine agency, says Nagel, seems to shrink to an extensionless point.

So I leave you for March break with the following perplexing non-solution to a really profound moral problem. Nagel suggests that the problem of luck has no solution because something in the idea of conceiving ourselves as agents is incompatible with the undeniable fact that actions are events, and people are things.

“As the external determinants of what someone has done are gradually exposed in their effect on consequences, character, and choice itself, it becomes gradually clear to us that actions are indeed events, and that people are indeed things. As a result of this, nothing remains which can be ascribed to the responsible self, and we're left with nothing but a portion of the larger sequence of events which can be deplored or celebrated, but not praised or blamed.”

Nonetheless, giving up the language of praise and blame is to remove from our conceptual repertoire what is perhaps the most important tool that we have. And coming to a stable perspective on these matters seems enormously difficult.  So I'll see you all at the end of vacation.

[end of transcript]