Philosophy › The 20th century › Applied ethics (bioethics, animals, global poverty)
The trolley problem
1967 (Foot); 1976 (Thomson)· Idea or argument· Ethics & how to live
Well documentedSolid records, texts or physical evidence.
A thought experiment: may you divert a runaway trolley so it kills one person instead of five, and is pushing someone into its path different?
See it on the Philosophy map →The idea
Idea: the case tests whether only outcomes matter. Philippa Foot (1967) asked whether a driver may steer a runaway tram from five workers onto one. Most say yes. Judith Jarvis Thomson (1976, 1985) named the problem and added the footbridge: pushing a heavy man off a bridge onto the track would also stop the trolley and save five. Most say no, though the numbers match. The puzzle is why. Main objection: critics say such artificial cases distort intuitions and teach little about real choices.
Why it matters
It became the standard test case for clashes between consequences and duties, and now informs moral psychology and debates about self-driving cars.
Where it comes from
Foot's 'The Problem of Abortion and the Doctrine of the Double Effect' (1967); Thomson's 'Killing, Letting Die, and the Trolley Problem' (1976) and 'The Trolley Problem' (1985). Joshua Greene's brain-imaging studies began in 2001.
Still debated
Whether the doctrine of double effect, the line between means and side effects, or something else explains the different verdicts is unresolved.
Connections (3)
▸The categorical imperative is tested by The trolley problemMostly agreed
The connection
Kant's formula of humanity forbids using a person merely as a means. That seems to explain the trolley verdicts: diverting the trolley kills one as a side effect, while pushing the man off the footbridge uses his body as a tool. But Kantians struggle to explain why diverting is permitted at all.
When and how
Foot (1967) and Thomson (1976, 1985) set up the cases; nonconsequentialists such as Frances Kamm have analyzed their variants in detail since the 1990s.
How sure are we?
The means versus side-effect reading is standard in the literature, and Joshua Greene's brain-imaging studies linked refusals to push with duty-based, 'deontological' intuitions.
Debates and open questions
Kant never discussed such cases; whether his ethics actually permits switching the trolley, and whether any single principle explains all the variants, is debated.
▸The trolley problem tests Utilitarianism: the greatest happiness principleWell documented
The connection
Utilitarianism counts outcomes: five lives beat one in both cases, so you should switch the trolley and also push the man. Most people accept the first but reject the second. That gap is a standard objection: either the footbridge intuition is wrong, as some utilitarians say, or the theory leaves something out.
When and how
Foot's 1967 paper and Thomson's footbridge (1976, 1985); Joshua Greene and Peter Singer later defended the utilitarian answer.
How we know
Surveys and lab studies consistently find most people switch but will not push; Greene's 2001 brain-imaging study in Science linked refusals to emotional responses.
Debates and open questions
Whether the footbridge intuition is moral insight or an emotional bias, as Greene and Singer argue, is a live debate.
▸The trolley problem resurfaced in debates about Philosophy of artificial intelligenceMostly agreed
The connection
Self-driving cars made the thought experiment concrete: if a crash is unavoidable, should the car protect its passengers or pedestrians, the many or the few? MIT's Moral Machine experiment gathered public verdicts worldwide. The case now frames broader questions about building ethical choices into machines.
When and how
MIT's Moral Machine website (2016), reported in Nature in 2018: about 40 million decisions from 233 countries and territories. Germany issued ethics rules for automated driving in 2017.
How sure are we?
The Nature paper and government guidelines document the link, though engineers often say real crashes rarely pose clean trolley-style choices.
Debates and open questions
Many argue trolley cases are the wrong model for AI ethics, which is more about bias, everyday risk and responsibility than dramatic dilemmas.