The Trolley Problem

A runaway trolley is speeding toward five people tied to the track, unable to move. You’re standing next to a lever. Pull it, and the trolley switches to a side track, where it will kill one person instead of five. Almost everyone asked this question says: pull the lever. One death is a smaller tragedy than five. So far, this looks like simple arithmetic. The trolley problem gets interesting once you change the mechanics while keeping the arithmetic exactly the same, and watch the verdict flip.

The footbridge: same math, opposite answer

The philosopher Philippa Foot introduced the original lever case in a 1967 paper on abortion and unintended consequences. Judith Jarvis Thomson, in a 1985 Yale Law Journal article titled simply “The Trolley Problem,” added the case that made it famous. Same runaway trolley, same five people ahead, but now you’re on a footbridge above the tracks next to a large stranger. The only way to stop the trolley before it reaches the five is to push him off the bridge; his body is heavy enough to stop it, killing him but saving the five. No lever, and your own body isn’t heavy enough to do the job. Asked this version, most people say no. One death, five lives saved, identical to the lever case on paper, and the answer reverses anyway.

The loop: testing what “using someone” really means

Thomson’s same 1985 paper adds a third version built to isolate exactly what’s driving that reversal. The trolley can be diverted onto a loop of track that curves back around and rejoins the main line, so it would still reach the five, unless a person standing on the loop is heavy enough to derail it first. Structurally, this is the footbridge case: the one person’s body is the actual mechanism that saves the five, not an unlucky side effect. But it’s operated with a lever, the same physical action as the original case. Answers to the loop version split much more evenly than either of the other two, which is itself evidence about what’s actually driving people’s judgments.

Rival explanations

The classical answer is the doctrine of double effect, traced to Thomas Aquinas’s Summa Theologica: harming someone as a foreseen but unintended side effect of pursuing a good outcome is treated differently from harming them as the actual means to that outcome. Diverting the trolley in the lever case kills the one man as a side effect of saving the five; pushing him off the bridge kills him as the means. The loop case is the hard test for this theory, because the one person there is also used as a means, yet many people’s answers drift back toward the lever case anyway.

Immanuel Kant, in the 1785 Groundwork of the Metaphysics of Morals, had already offered a more formal version of the same intuition, nearly two centuries before Foot or Thomson ever wrote about a trolley: act only in ways that treat a person as an end in themselves, never merely as a means to someone else’s end. Kant doesn’t need a separate doctrine about intention to explain the footbridge case. Diverting the trolley treats the one man’s death as an unwanted side effect of a rule aimed at everyone’s survival; pushing him off the bridge treats his body as equipment for stopping a trolley, which is exactly the kind of use Kant’s formula rules out no matter what anyone intended.

The psychologist Joshua Greene took a completely different approach, in a widely cited 2001 Science paper: he scanned people’s brains while they considered trolley-style dilemmas and found that “personal” harm, the kind that requires touching someone with your own hands, activates emotion-related brain regions far more than “impersonal” harm like flipping a switch, regardless of what any doctrine says about means and side effects. On this account, what actually drives the footbridge verdict isn’t a principled moral distinction at all. It’s a much older, more visceral aversion to using your own body to hurt someone up close.

John Rawls, in his 1971 A Theory of Justice, asked a different kind of question altogether: not why people react to a case the way they do, but what rule they’d actually agree to in advance, from behind a “veil of ignorance” that hides which position in the scenario they’ll end up occupying. Michael Sandel’s Justice: What’s the Right Thing to Do? (2009), built from his popular Harvard course, opens its own discussion of utilitarian reasoning with this same trolley case, before turning, later in the book, to Rawls’s veil of ignorance as a rival standard for what counts as fair. Run that test on the lever case yourself: would you agree in advance to a rule letting a bystander divert deadly danger onto whichever one person happens to be standing on a side track, not knowing whether that person might turn out to be you?

Why it isn’t just an armchair puzzle

These aren’t just competing theories argued over in seminar rooms. Military doctrine on collateral damage leans explicitly on the doctrine of double effect to distinguish permissible harm to civilians from war crimes. Self-driving cars occasionally face a version of the same choice in miniature, swerve into one bystander to avoid five, and in 2018 a Nature paper reporting the Moral Machine experiment collected judgments on exactly this question from millions of people across more than two hundred countries, finding real, systematic cultural variation in where the line gets drawn. None of this is hypothetical anymore. Somebody has to actually decide, in code or in policy, which version of the trolley problem the world runs on.

You can run the four classic variants on yourself, quickly, before reading any explanation of what they’re testing: Would You Pull the Lever?