The Trolley Problem

In 1967 Philippa Foot published a paper in the Oxford Review called 'The Problem of Abortion and the Doctrine of the Double Effect'. On its first page, in the course of testing whether a principle from Catholic moral theology could do the work asked of it, she introduced a runaway tram and a driver who can steer it onto a track where one man will die rather than a track where five will. She used it and moved on. It has since generated a research literature, a field of experimental moral psychology, and an entire genre of internet joke.

It is worth recovering what it was for, because the artificiality that makes it a joke is the same feature that makes it work. Foot is running a controlled experiment. The outcomes are held fixed (one death instead of five) so that if judgements differ between versions, the difference cannot be about the numbers. Everything else has been stripped out precisely so that one variable can be varied.

Judith Jarvis Thomson sharpened it into the pair everybody now knows. In the first case a trolley is heading for five people and you can throw a switch to divert it onto a side track where one person stands. In the second you are on a footbridge above the track beside a large man, and the only way to stop the trolley is to push him off in front of it. The arithmetic is identical. Around ninety per cent of people will throw the switch. Around ten per cent will push.

That gap is the whole subject. A strict consequentialist has to give the same answer in both cases, because five deaths are worse than one and nothing else in the description is morally relevant. So either the intuition in the footbridge case is tracking something real that the theory omits, or it is a psychological artefact that ought to be overridden. Both conclusions are live and neither is comfortable.

The traditional explanation is the doctrine of double effect, which descends from Aquinas. It holds that it can be permissible to bring about a harm you foresee but do not intend, and impermissible to intend the same harm as your means. Divert the trolley and the man on the side track dies because he happens to be standing there; if he noticed and stepped clear, your plan is unaffected and you would be delighted. Push the large man and his body is the mechanism; if he survives, you have failed. In the first case his death is a side effect, in the second it is the method.

The principle is not idle. It does real work in medicine: administering morphine in doses that will shorten a patient's life in order to relieve pain is treated quite differently from administering morphine in order to shorten life, though the pharmacology is identical. Whether the distinction is genuinely moral or merely a way of describing what you are doing so that you can live with it is the central dispute, and Foot herself was not uniformly confident about it.

Her own preferred explanation was different and, to my mind, sturdier: the distinction between doing harm and allowing it. Negative duties not to harm are stringent. Positive duties to aid are real but weaker and more easily outweighed. This is why you must not push the man and need not bankrupt yourself for a stranger, and it is why diverting an existing threat is not the same act as initiating one.

Thomson then supplied the case that shows how much is at stake. A surgeon has five patients who will each die without a different organ, and a healthy visitor in the waiting room whose organs would match. The arithmetic is the same as the trolley. Essentially nobody accepts it. Sophisticated consequentialists answer at the level of rules (a hospital that harvested visitors would empty within a week, so the practice maximises nothing) but this rescues the verdict by abandoning the direct calculation the theory was built on, which is a fair hit.

The standard objection to all of this is that reasoning from such cases is worthless, because real moral situations involve uncertainty, relationships, institutions, and consequences for trust, and the trolley deletes every one of them. That is entirely correct as a description and wrong as an objection. The cases were never decision procedures. Their job is to show you that your principles are not what you said they were, and on that job, sixty years on, they still work.