The dangers of ethical thought experiments

by Carl Pierer

159999773“Yes, I would let the five people die.”

To philosophers, and I mean to include all people interested in philosophical questions, this is a pretty standard response to a pretty well-known thought experiment: The Trolley Problem. But it is not only in philosophy that you get very uncanny scenarios when trying to clarify an idea by applying it theoretically. These thought experiments play an important role in fields as diverse as physics and arts, mathematics and literature, but the most infamous ones are probably to be found in philosophy, and in ethics particularly. Not only are they notorious, but in fact they face two challenges, which easily turn into dangers should we ignore them and base our argument on them.

First of all, thought experiments have to be distinguished from metaphors, since they serve different purposes. At first sight it might seem that they are poles apart. However, Dennett writes: “If you look at the history of philosophy, you see that all the great and influential stuff has been technically full of holes but utterly memorable and vivid. They are (…) lovely thought experiments. Like Plato's cave, and Descartes's evil demon, and Hobbes' vision of the state of nature and the social contract, and even Kant's idea of the categorical imperative.” Dennett here conflates a variety of famous philosophical scenarios under the heading “thought experiments”. Yet the structure of Plato's cave is completely different from Descartes's evil demon. In Plato's case there is no new knowledge gained. It is not a hypothetical scenario of how the world might be, but rather a more literary expression of how it actually is. The philosopher's ascent from the cave is figurative and an it does not serve the purpose of drawing some conclusion from this view, but rather to embrace the general idea that this is the philosophers' condition. It is a picture, an illustration of his idea rather than a method to develop a new belief. Descartes, on the other hand, imagines an evil demon who brings about a very sophisticated illusion of reality, making us think that all our experiences are real while they are merely his creations. It is an application of radical scepticism. Once we hypothetically accept this scenario Descartes asks whether any of our pre-demonic knowledge still stands. The difference between Plato's cave and Descartes's demon is that the former is a mere illustration of an idea. The latter, in contrast, serves to provide some new insight. Therefore, I propose to distinguish between thought experiments and metaphors. The purpose of the former has to be a more rigid one than that of the latter. We use thought experiments to test what happens if we apply our theoretical ideas. Its similarity to actual experiments should not be ignored. We peruse those hypothetical results, and only if we can accept them are we ready to accept a theory.

However, more often than not, thought experiments are used the other way round. Hypothetical scenarios are invented in such a way that our theories fail to deliver what is expected of them.

Take, for instance, Gettier cases. Gettier successfully conjures a scenario where the justified true belief theory (JTB) yields a result that contradicts our intuitive response. On the classic analysis, knowledge consist in having a belief that something is a fact. This belief has to be justified (by evidence of some sort) and true. Only then are we allowed to talk of knowledge. But we don't consider people in Gettier cases to know even though their reasoning meets all the conditions of JTB.

A Gettier scenario would be the following: You just woke up and look at your (analogue) clock standing by your bedside. You see that the clock displays 9am. Unfortunately, you don't know in this case that the clock stopped working yesterday evening (you were very tired that night and got to bed at 8pm). And so you form the justified belief that it is 9am. As it turns out, in this very moment it is indeed 9am. According to JTB you know that it is 9am. However, this does not fit our concept of knowledge, it is a mere accident that you got it right. So we find JTB insufficient on the grounds that it would attribute knowledge falsely. This shows that Gettier cases are proper epistemological thought experiments. By applying our JBT we see that it fails utterly. Therefore we have to go and look for a new account for what knowledge is.

Suppose we didn't let go of the theory that has been refuted by our hypothetical test. We could just embrace the unlikely consequence and state that it doesn't falsify the theory. This strategy has become known as “biting-the-bullet”. Such a response to Gettier cases would yield: ‘Well, it seems counterintuitive, but in these cases we do know. We know that it is 9am, simply our original ideas about knowledge are misleading.'


Consider, in contrast to epistemology, an ethical thought experiment: The trolley case in its extended version by Judith Jarvis Thomson (1976):

Frank is a passenger on a trolley whose driver has just shouted that the trolley's brakes have failed, and who then died of the shock. On the track ahead are five people; the banks are so steep that they wiII not be able to get off the track in time. The track has a spur leading off to the right, and Frank can turn the trolley onto it. Unfortunately there is one person on the righthand track. Frank can turn the trolley, kiIIing the one; or he can refrain from turning the trolley, letting the five die.

Deontologists, who think that killing is always worse than letting die, are compelled to answer Frank must not turn the trolley. The thought experiment is conceived in such a way that this is counterintuitive. Surely, it is better to save the lives of five people and only have one person dead? Yet, since this is an ethical thought experiment it is easier to bite the bullet here; Deontologists could simply claim ‘Well, it might feel wrong, but letting the five die is the right thing to do.'

However, the two instances of biting the bullet are fundamentally different. In the Gettier case, this is not an actual option. The point is: the justified true belief theory has been put forward to rule out coincidentally getting something right. Knowledge, we think, is more than luckily getting it right. Gettier cases are problematic since they show us a scenario where all conditions are met and yet it is somewhat “beyond the influence” of the agent that they are correct. The JTB theory tries to capture what knowledge means, not to change our concept of it. Yet, biting the Gettier bullet would entail such a change. Contrary to this, moral theories try to find what is morally right. In this, they can be prescriptive. They can tell us what would be right, even though we don't feel that it is. Epistemological theories, on the other hand, such as JTB are descriptive. They do not aim to alter the world as such, but to explain it.

There lies a further danger hidden behind ethical thought experiments. Thomson gives two instances of the trolley scenario: the one just mentioned and one where Edward faces much the same situation as Frank but with the slight difference that he is the driver of the trolley, rather than a mere passenger. The alternatives open to him, since he is the driver, are killing one or killing five. Thomson thinks that the cases of Frank and Edward are different even though their choices aren't. Thomson follows Foot in making us think that Edward – as a driver – has a responsibility for the trolley: “Mrs. Foot thinks that if Edward does nothing, he kills his five, and I agree with this: if a driver of a trolley drives it full speed into five people, he kills them, even if he only drives it into them because his brakes have failed. But it seems to me that if Frank does nothing, he kills no one. He at worst lets the trolley kill the five; he does not himself kill them, but only lets them die.” One would assume that this entails different moral responsibilities for the two, but Thomson continues: “Yet I take it that anyone who thinks Edward may turn his trolley will also think that Frank may turn his. Certainly the fact that Edward is driver, and Frank only passenger could not explain so large a difference.” In her article this seems a consistent position, for she argues against a distinction between killing and letting die. This consistency stems from a circularity: The driver faces a decision between killing and killing, whereas the passenger between killing and letting die. This is explained by the fact that the driver is the driver. Yet, both are – at least – permitted to turn the trolley, since a distinction here cannot be explained by their different status. Therefore, killing and letting die are the same.

The problem with this argument lies with the assumption that their different positions cannot explain why not both of them should be allowed to turn the trolley. If being a driver can account for facing a decision between killing and killing, then why does this not entail different responsibilities? Saying that being a driver explains the former but not the latter is arbitrary, unless we presuppose that killing and letting die are the same. But presupposing this conclusion renders the argument circular.

This confusion is due to what we read between the lines. Naturally, we think that a professional has some responsibility, since it is his job – in Edward's case – to drive the trolley. So even while doing nothing, he would still be driving the trolley, whereas Frank, the passenger, would only start driving the trolley if he chose to. Yet, there is a further sense in which Edward can be called the driver. To make this point clear, we have to modify Thomson's thought experiment slightly.

Imagine that, longing for revenge after his last encounter with a philosopher, Descartes' evil demon interfered with poor Edward's trolley. Edward doesn't have any responsibility for the state of affairs, and there is no way he could have avoided the situation he finds himself in. Let us assume furthermore that Edward has just in this very moment been assigned to be the driver. In this altered thought experiment the situations of Frank and Edward are truly alike, except for the fact that Edward is nominally the driver. But then, if Edward is entirely innocent and only driver of the trolley because he has been named by the evil demon, why does he face a decision between killing and killing? The first sense of being a driver that Thomson invokes is that of being a professional. A professional faces different choices than a layman, simply because it is his job. Therefore, being a driver explains why Edward can only choose killing. However, the second sense of being a driver is merely nominal. We can think of it as something like a title: not Sir Edward, but Driver Edward. Then it becomes apparent this nominal difference cannot account for a substantial difference in choice. Thomson's argument seems sound since she tacitly moves from the first sense (with its explanatory power) to the second sense (which lacks this power). If we used the first sense throughout the argument, then the premise “Yet, both are – at least – permitted to turn the trolley, since a distinction here cannot be explained by their different status” would be wrong by definition. In contrast, if we used the second sense consistently, then – as the modified thought experiment shows – it is not clear why Edward faces a decision between killing and killing, and Frank does not.

Here, another risk of ethical thought experiments is revealed. They strongly depend on our background knowledge and on what we understand from what has been left out. Thomson's argument seems convincing because we associate “driver” with “responsibility”, which then yields the desired link that the driver can only kill. However, if we empty the word “driver” of that unexpressed association then we find that there is no difference between the two cases, because Edward doesn't have to choose between killing and killing any more so than Frank does. The problem is that thought experiments can be used very effectively even if (or because) they attempt to prove what they presuppose. This danger is lurking within any thought experiment; its set up has to be rigorous or else it is merely a rhetorical trick. It might then nicely illustrate a certain theory, but definitively be incapable of proving anything new. We might go as far as to call those sloppy rhetorical devices metaphors rather than thought experiments.

Thus are the two dangers of ethical thought experiments:

First, they cannot be used as conclusive disproofs. In cases where they appeal to our emotions and our initial thoughts about a particular scenario they run into difficulties. For we can always claim that our moral theory is normative, that it determines what is right and wrong, and hence we shouldn't bother too much about what we intuitively feel about the situation. There is no way a thought experiment could make a normativist change her mind, for she doesn't have to pay attention to the actual world at all. Second, they might seem convincing and clear even though they are not. They might rely on connotations or associations of words and find contrasts where they do not exist. As with every other bit of philosophy, we have to pay attention to use non-ambiguous language in thought experiments as well, in particular with ethical thought experiments. And we have to be critical of the purpose of our experiments. Do they prove anything new or are they simply illustrations of our dearly held theories? Are they rigorous tests of our theory or mere metaphors?

So, in the light of these two difficulties and even though it feels a bit odd, I still hold:

“Yes, I would let the five people die.”

Works cited:

Dennett, D. C. (1995). Intuition Pumps. In J. Brockman, The Third Culture: Beyond the Scientific Revolution. Simon & Schuster. Retrieved from

Thomson, J. J. (1976). Killing, Letting Die and the Trolley Problem. The Monist, 204-217.