Sunday, December 16, 2012

The Trolley Problem

Over on Bloggingheads, some psychologists were discussing the neurology of moral judgments. They discussed the trolley problem, which is pretty famous among moral philosophers.  The basic conundrum is this:
A trolley has lost its brakes, and is about to crash into 5 workers at the end of the track. You find that you just happen to be standing next to a side track that veers into a sand pit, potentially providing safety for the trolley's five passengers. However, along this offshoot of track leading to the sandpit stands a man who is totally unaware of the trolley's problem and the action you're considering. There's no time to warn him. So by pulling the lever and guiding the trolley to safety, you'll save the five passengers but you'll kill the man.
Most people pull the switch, killing the one man to save five. That wouldn't be so interesting by itself, but then the problem is extended to a seemingly similar problem:
As before, a trolley is hurtling down a track towards five people. You are on a bridge under which it will pass, and you can stop it by dropping a heavy weight in front of it. As it happens, there is a very fat man next to you – your only way to stop the trolley is to push him over the bridge and onto the track, killing him to save five. Should you proceed?
Most people would not push the fat man. These grave dilemmas constitute the trolley problem, a moral paradox first posed by Phillipa Foot in her 1967 paper, "Abortion and the Doctrine of Double Effect." I don't really like any of the popular resolutions as to why people think it's OK to kill the first guy but not the second.

I think a good resolution is that in the second case there is a significant probability that one does not save the 5 men, and instead merely kills the fat man. I've never pushed a fat man in front of a trolley, but I suspect most would simply run right over him and keep going. In the first case, if you  killed the one man you definitely save the five men, it's not possible to kill both the one man and the five men by switching tracks. In the other case the probability is clearly less than 1, perhaps only 0.1. That's the difference. As a rule, acting on a theory and killing x people with certainty to perhaps save 5x people is morally wrong, mainly because these theories are often wrong, so all you do is kill x people (eg, a lot of evil is legitimized as breaking eggs to make an omelette, but then there's no omelette). The move from certainty to mere 'highly likely in my judgment' is huge.


Anonymous said...

You're supposed to just assume the fat man would stop the trolley (with 100% certainty). It's a thought experiment - don't get hung up on whether that's realistic or not.

If you don't like the "fat man" case, consider the "transplant" case, also described on the Wikipedia page.

Eric Falkenstein said...

I know the philosophers assume there's a 100% chance the fat man stops the trolley, but that's their error. As they are trying to explain the preference, I think they are forgetting that most people's intuition rightfully puts this chance below 100%. Transplants also don't have 100% success rates. The lever switch is a 100% scenario. Certainty is very different than highly likely, that's my point.

Anonymous said...

In my opinion the real problem is attempting to derive useful, general moral insight from bizarre or freak events. I have never in my life encountered such a problem and I doubt most people have. Regardless, in the real world when faced with highly uncertain outcomes where morality is not clear, most people freeze and do nothing and "let things happen" rather than become an active participant. And of course a fat man would not stop a trolly. The entire problem is a hoax.

Anonymous said...

Bingo Falken. You got this one right.

Anonymous said...

For the lever problem the answer is to direct the trolley to whichever track has the psychologist. For the fat man problem, push him if he is a psychologist.

jason g said...

There's one major difference between the switch situation and the fat man situation, and I believe it plays (consciously or subconsciously) into people's judgments.

In the switch situation, the one man on the track is already in a dangerous situation. He's on an active trolley track which has the capacity at any moment to be carrying a trolley that will kill him.

The fat man is merely sitting on a bridge, and by pushing him, YOU are the one actively putting him into the dangerous situation that he did not choose for himself.

So, I can see a logical reason for a person to choose to throw the switch to kill the one man, but to choose not to push the fat man.

tomslee said...

If we accept the peculiar preconditions, assumptions, and hypotheses the philosophers wrap around this problem... then you push the fat guy. I don't understand why people try to say "but our intuitions say it's wrong" as if that's an argument. Intuitions can be wrong.

Anonymous said...

jason has it right - it is nothing to do with a calculus as to whether you think the fat man will stop the train or the occupants still get injured.
It is simply a direct act of throwing the man to his death versus the indirect act of causing it via throwing a lever. In the direct case the man will know you did it, in the indirect he will not. People will not do the direct but can happily do the indirect.

Anonymous said...

Well, Jason solved the problem. Discussion over.

MP said...

Well said, people perceive a very different emotion when they must "do the dirty work" first hand. I think the simple beauty of this thought experiment is it illustrates how framing problems differently will effect decision making. Don't get in the weeds on the specifics. To know this of others and YOURSELF is the beauty. Remember, just because you know about flaws in human decision making does not it now doesn't apply to the now "enlightened you" - that would be... overconfidence :)

Eric Falkenstein said...

Jason: I don't see a difference between pushing the fat man, and pulling the switch in regards to their deaths, because both imply death pretty much with the same confidence. To me the crux of the difference is that pushing the man does not necessarily save the 5, but in the switch case it does necessarily do so.

One might say, assume it did, but that would require a difference example, say, by portraying a case where a fighter jet fires a missile that obliterates the trolley with its one occupant so nothing is left to hit the five men, in which case I would not hesitate to do so and I think most would agree.

Mercury said...

In many states if you kill an armed intruder in your own home who is threatening your life the DA will bend over backwards to ruin you, regardless of what the law actually says. In these two trolley examples there are so many ways for our philosopher hero to lose that it’s hard to know where to begin. So, in the real world the best course of action is to turn you back on the oncoming disaster and feign obliviousness. I hate to say it but the “Fat Tonys” (as opposed to the academics) standing on the platform/bridge probably have this figured out by now.

Here’s a thought experiment which we are seeing real-world examples of with ever-increasing frequency:

In response to a major national disaster/crisis, would you create a new federal agency, knowing that it would tax and hassle most people for the rest of eternity if it were 'highly likely in your judgment' that such an action would be perceived *today* as effecting more good than harm?

Anonymous said...

Fat men are hard to push, and they might push back, I think it is hard for the human mind to simply avoid these facts (as it is hard to imagine a fat man would stop a trolley).

In either case, discussing philosophy problems is a waste of time.

cig said...

I think what the experiment captures is that people act differently depending on the physical proximity of the victim (or beneficiary) of their actions.

If I'm correct, the potential flaws in the experiment's design are totally irrelevant: most people still wouldn't kill a man (regardless of weight) by pushing him to his death, even if it could with total certainty save five people, who are much further away.

The key parameter is distance: if you put say N people within equal pushing distance, and pushing a specific one to their death would save the others and not conversely, then you would, I expect, get back "rational" behaviour, of minimising the number of deaths within the equidistant group.

I think that the claim it's an unrealistic experiment is also not founded. This phenomenon can be observed in real life in many ways: many more people would happily steal a sandwich from an unattended dispensing machine than would steal one from a sandwich seller they just had eye contact with (given equal sandwiches, prices and task difficulty). Or similarly, if you're walking on the pavement and someone within 1 metre of you gets severely injured by some flower pot falling from a nearby balcony, you'll get more agitated than if they were 50 metres away, and way more than if you read on the Internet that someone got hit by a flower pot in Australia, even if in all cases the victim was previously unknown to you.

Most armies are organised to exploit this effect in many ways (uniforms, team building, hierarchy, etc) which are all contributing to add an artificial distance between the soldier and his colleague on the other side, because the closer your target is, the harder it is for non-psychopaths to kill fellow human beings.

I don't see there's much of a paradox to speak of either: that physical closeness is a parameter that reinforces the don't-harm-your-own-species taboo seems a fairly obvious outcome of natural selection at the species level.

Anonymous said...

There is another problem with this question. The assumption of perfect knowledge of outcomes prior to the decision point here is just as bad as peaking into the future in a backtest. On the other hand, a reasonable real-time judgement could be made that the person down the track could hear or see the train and jump out of the way. The entire question is so flawed no useful generalizations can be made.

Dave Pinsen said...

"Jason: I don't see a difference between pushing the fat man, and pulling the switch in regards to their deaths, because both imply death pretty much with the same confidence."

Eric, you're eliding Jason's point that guy on the track is already in danger while the fat man on the bridge isn't. Since the train is going to go one one of the two tracks, you could say that the single man on the one track already has a 50% chance of getting killed. You flipping the switch makes it a 100% chance, but he was halfway there already.

That's not the case with the fat man on the bridge. His chance of falling to his death without you pushing him is close to 0%. So you would be close to 100% responsible for his death if you pushed him.

Cig's point about physical proximity make some sense too, but there's another, related aspect. Killing someone by pulling a lever is psychologically easier (even if they are close to you physically) than killing someone with your bare hands, as you would by pushing the fat man off the bridge.

J said...

If you have drop of common sense, you will away and never talk about the tragedy you witnessed. Throwing a man into an incoming train is crime and you will surely hang. The judge can see the five healthy men who MAY have been in danger, but one cannot be 100% sure of hypothetical happenings. As time passes, there is less and less certainty that the five were condemned to unevitable death, maybe the brakes could have worked, maybe they could have escaped, may be their prayer would have been answered. maybe you misjudged the situation and there was no danger at all. The death of five men is highly hypothetical, while the man who you threw into the rails has a name, there is a body, there are pictures of bloody pieces of flesh in the paper and his widow and pitiable orphans are claiming justice. Your intentions, Sir, are not material.

Eric Falkenstein said...

Dave:considering this is a famous philosophy question, reasonable people can disagree...nonetheless, I don't think the lone guy on the track is 'already in danger'. If I do nothing, he is in zero danger; he's not 'halfway' there.

chucho said...

If you will not kill the fat man, then you should not throw the switch either. In both cases, your actions have resulted in the murder of a human that would not have occurred otherwise. The utilitarian logic of throwing the switch ultimately leads us down a very dark road. This problem is notoriously difficult because it makes utilitarian calculation seem relatively easy and natural. But as the second case shows, this line of reasoning can lead to all kinds of bad actions, like murder or even genocide.

Eric Falkenstein said...

cucho: I don't think they are the same. As you note, pushing the man can lead to genocide. When has genocide actually save 5x as many people as it killed? Never? That's why genocide is wrong, the theory is wrong...the fat man doesn't stop the train. If every genocide preceded a great flowering, genocide wouldn't have such a bad rap, but it's so obviously untrue, it's indefensible.

chucho said...

Eric, you've stated a moral rule in your reply: "genocide is wrong." Another moral rule could be, "killing an innocent is wrong." The onus is on the switch throwers to explain why this rule is not valid or can be overridden in the first case. As others have tried to explain, maybe the track worker is not an "innocent", or in your own case you've tried to express it in probabilities, but these exceptions only tip-toe around the central issue:
if you reject a rule-based moral code, how do you define the ethical parameters for your actions? "It depends" is just relativism.

Eric Falkenstein said...

I think 'killing an innocent is wrong' is not a good moral rule because it clearly isn't true. There are just wars in which many innocent civilians and even combatants (think 17-year old Germans in 1944) are killed. One statistically kills people by driving a car. Many people think killing an 8-month fetus is OK (I don't, but I agree a 2-month fetus is fair game). It's always costs vs. benefits.

jason g said...

Eric, thanks for your response. I think the thing people are forgetting here is that we are not talking about the "best" or "most utilitarian" or "logical" course of action (at least i wasn't, anyway.) We are talking about how people would respond (primarily emotionally) in either of the difficult situations.

So, yes it may be that 2 different actions result in precisely the same outcome (and i believe that is supposed to be assumed in the experiment.) But, they FEEL so much different to the person making the decision.

The human mind is capable of rationalizing a great deal. Throw a switch and an unseen person dies to save 5 -- you did your duty.

Push someone to directly cause their death in front of you -- well, you can't rationalize that away.

I believe that is the primary mechanism accounting for the empirical results of this thought experiment.


Dave Pinsen said...


Absolutely, reasonable people can disagree. It's interesting to consider why we disagree though. I think the crux of it is this:

You see this problem as the philosopher probably intended: the trolley in question is the only one, in which case you're right that the lone guy on the track is not in any danger already.

But I (and I'm guessing Jason and others that agree with him), assume that that section of the track is used sometimes, and there are other trolleys that might use it and hit the lone man, regardless of our actions in this case. So a guy lingering on railroad tracks has already put his life in danger in a way that a guy standing on a bridge has not (though my initial comment that he was "halfway there" overstated the case).

Anonymous said...

Replace the Fat Man with an innocent (and cute) 18 year old girl, and the 5 men are as good as dead.

On serious note, in the most pure form this is about our minds separating reality between "us" and "the world around us". This separation means that inaction will never bear the same moral consequences as action.

But in practical, less pure experiments, this is about uncertainty, as Eric said.

Unknown said...

This is not about uncertainty. You could also believe the guy in case 1 could wiggle off the track to save himself. It depends on what you are willing to consider.

This is about the basic moral problem of treating a person as a means to an end. In case 1, you are acting on the trolley. In case 2, you are acting on a person. Seems pretty straight forward.