The trolley problem

What it tests Why does killing one to save five seem permissible in one case and not in the other, and what does the difference show about right and wrong?

The case

A runaway trolley (a tram, in British usage) is heading for five workmen who cannot escape. You are next to a lever. If you pull it, the trolley goes onto a side track where one workman is standing, and he will be killed instead. Most people say you may pull it.

Now change the story. You are on a footbridge above the track, and the only way to stop the trolley is to push a heavy stranger off the bridge into its path. He will die and the five will live. Most people say you must not. The numbers are the same, and the verdicts differ. The trolley problem is the problem of saying what explains the difference.

Philippa Foot started it in 1967, in a paper on abortion and the doctrine of double effect (the view that harm you foresee as a side effect can be permissible when harm you intend is not). She paired a judge, who could stop a riot only by framing and executing an innocent man, with the driver of a runaway tram who can steer from a track with five workmen onto one with a single man. Most people say the judge must not and the driver may. Her explanation: we owe more in not harming than in helping, and the judge would be killing one rather than letting five die, while the driver must choose between killing five and killing one.

Judith Jarvis Thomson named the problem in 1976 (she later explained that trams are called trolleys in America), and in 1985 replaced the driver with a bystander at a switch. The bystander is choosing between killing one and letting five die, yet most people think he may turn the trolley. Then Foot’s explanation fails, and something else must separate him from the man on the footbridge.

What it is meant to show

The cases test moral theories. “Save the greater number” gives one verdict for both; “never kill” gives the other; a theory that explains the pair has found a structure in our moral thinking. The method treats verdicts as evidence. In the 2009 PhilPapers survey, 68 per cent of professional philosophers would switch and 8 per cent would not.

Replies

Count the lives. Consequentialists deny that there is anything to explain: the right act in both cases saves five. Joshua Greene and colleagues found in 2001 that personal dilemmas such as the footbridge engage emotion-related brain regions more than impersonal ones such as the switch; Greene and Peter Singer use such findings to argue against deontological intuitions. The argument presupposes that emotional reactions are irrational or arational here, and those who reject that view of emotion need not follow it.

Intending and foreseeing. On the doctrine of double effect, which goes back to Aquinas on self-defence, the death of the one is a side effect when you pull the lever and a means when you push. Thomson’s Loop case tests this. The side track loops back to the main track, and one heavy man stands on it. His body will stop the trolley, and without him it would come round and kill the five. He is a means, yet most people still say you may divert it. Some take the paired verdicts as evidence that the means/side-effect distinction guides judgement; others deny that it explains the switch case.

Doing and allowing. Foot’s principle explains the judge and the driver but not the bystander. Frances Kamm adds five workmen on each track: the bystander must not turn the trolley, so harming has some priority over not aiding, and the five-to-one permission holds despite that priority, not because it is absent.

Deny the bystander’s permission. Elizabeth Anscombe, who defended an absolutist version of double effect, did not think it permits swerving the trolley. In 2008 Thomson changed her own mind: a bystander who could turn the trolley onto himself, and would not, is making a stranger pay a cost he will not pay, like stealing from someone to give to charity. William Fitzpatrick replied that morality cannot demand self-sacrifice.

Doubt the method. Roger Scruton complained, in On Human Nature (2017), that such dilemmas strip away every morally relevant relationship and leave only arithmetic, and Bauman and colleagues (2014) thought them too extreme to bear on real situations. Against that, a study in 45 countries found a death caused as a means judged less acceptable than one caused as a side effect.

Variations

If the man you could push is the villain who tied the five to the track, a majority are willing to push him. A death is also judged less acceptable when it results from the agent’s muscular force, as in pushing. In a transplant version, five patients could be saved with the organs of one healthy person.

Why it still matters

Designers of autonomous vehicles must decide how a car behaves when a collision is unavoidable. The Moral Machine experiment collected 40 million decisions from people in 233 countries and territories and found three major clusters of countries. A German ethics commission reported in 2017 that genuine dilemmas cannot be standardised or programmed in an ethically unquestionable way, and whether abstract verdicts should guide design is disputed.

The problem is unsolved, and each reply states a principle exactly enough to be tested against a case.