What it was designed to show

Philippa Foot introduced what became the trolley problem in a 1967 paper on the doctrine of double effect — a principle from Catholic moral theology holding that it can sometimes be permissible to bring about a harmful consequence as a side effect of achieving a good end, even when it would be impermissible to bring about that same harm as a means to the same end. Foot’s original scenario was a trolley (tram) with failed brakes hurtling toward five workers on the track. You are the driver. You can steer onto a side track, where there is one worker. Do you steer? Most people say yes. Five lives against one — the calculus seems clear enough.

What the scenario was designed to do was create a case where the utilitarian calculation (minimize deaths: steer) and the doctrine of double effect might come apart, and to probe whether the doctrine itself was coherent. The one person on the side track dies as a consequence of your steering, but not, strictly speaking, as your means of saving the five; their death is a foreseen but unintended side effect. This is the kind of distinction the doctrine of double effect is designed to track. Foot was asking whether this distinction does any real moral work, or whether it is a piece of post-hoc rationalization for an intuition that is better explained in other terms.

Judith Jarvis Thomson added the footbridge variant in a 1985 paper. The scenario is now different: you are on a footbridge above the tracks, watching the runaway trolley approach five workers. Beside you is a large man. If you push him off the bridge, his body will stop the trolley, saving the five. He will die. Physically, the outcome is identical to the original: one death, five saved. But the overwhelming majority of people who said “yes, steer” in the first case say “no, do not push” in the second. Thomson used this divergence to argue against straightforward utilitarianism: if the only thing that mattered morally were the numerical outcome, the two cases should feel equivalent.

What it actually shows

The trolley problem does not definitively prove anything — it is not an argument — but it does something philosophically more useful: it reveals a structure of moral intuition that is consistent across cultures and resistant to easy debunking, while being resistant to easy justification as well.

People who would steer in the original case overwhelmingly refuse to push in the footbridge variant, and this divergence is robust: it holds across gender, culture, educational background, and philosophical training. The difference that seems to matter — or that people feel matters — is something like the distinction between redirecting a threat and using a person as a tool. In the footbridge case, the large man is not part of the threat; he is a bystander whose body you intend to weaponize for your purposes. In the original case, you are redirecting a harm, and the one worker’s death, while foreseen, is not the mechanism by which the five are saved.

Whether this felt distinction is a morally defensible principle or a psychological artifact has been argued for nearly sixty years. The neuroscientist Joshua Greene used trolley problems in fMRI studies around 2001 and found that the footbridge variant activated emotional processing areas of the brain more intensely than the original case, suggesting that our aversion to pushing is driven by an emotional rather than a rational response to personal physical contact. This led Greene to argue that our “deontological” intuitions — the sense that some things are just wrong regardless of consequences — were evolutionary holdovers, unreliable guides to actual moral truth, and that utilitarian reasoning, being more impartial and systematic, should take precedence. This conclusion was vigorously contested. The fact that an intuition has an evolutionary or emotional origin does not tell us whether it is tracking something morally real.

How it has been used and misused

The trolley problem has become the most discussed thought experiment in moral philosophy, which means it has also become the most abused. In academic philosophy, trolley variants have proliferated — loop tracks, transplant surgeons, runaway trolleys on carousels — until the apparatus sometimes obscures more than it illuminates. The point of a thought experiment is to isolate a specific variable; when the scenarios multiply without a clear target, they generate intuition-pumping noise rather than insight.

In popular discourse, the trolley problem has been applied to autonomous vehicle programming, military drone targeting, pandemic resource allocation, and dozens of other real-world contexts. These applications are sometimes genuine and illuminating: the question of how a self-driving car should be programmed to respond in unavoidable crash scenarios really does have trolley-problem structure. But they are often misleading, because the trolley problem assumes certainty (you know exactly what will happen if you steer or do not steer) while real ethical decisions involve radical uncertainty, incomplete information, institutional constraints, and systemic effects that single-decision scenarios cannot capture. Using the trolley problem as a direct template for policy is a category error.

The most serious misuse is the claim that the problem “proves” utilitarianism or “proves” deontological ethics, depending on which horn of the dilemma the arguer wants to defend. It proves neither. What it shows is that we have consistent moral intuitions that cannot all be satisfied simultaneously, and that any moral theory will have to either accommodate or override some of them — a challenge for every theory, not a victory for any.

What remains genuinely unresolved

After nearly sixty years of philosophical attention, several things about the trolley problem remain genuinely open.

First: is the personal-force distinction morally real, or is it a psychological artifact? We feel that pushing is worse than steering, and this feeling is consistent and cross-cultural. But we have not established whether what we feel is tracking something morally relevant about direct versus indirect causation, or whether we are simply more repulsed by physical contact. These are different questions, and the empirical literature does not settle the philosophical one.

Second: even if we accept that the footbridge case is worse than the original, we have not agreed on why. Is it because using a person as a means violates their dignity? Is it because it treats the large man as a mere instrument? Is it because our obligations not to harm are stronger than our obligations to aid? Different moral theories give different answers, and the scenario does not adjudicate between them.

Third: does our intuitive response in the footbridge case reflect the actual morality of the situation, or does it reflect a moral psychology shaped by evolutionary pressures that no longer apply? If our sense that pushing is wrong is an emotional response to physical proximity rather than a track of anything morally real, should we override it? And if we conclude we should override it, do we have any remaining grounds for trusting moral intuitions at all?

No one really knows. The trolley problem has been enormously productive not because it solved anything, but because it made the difficulty visible in a form that could not be shrugged away.