A runaway trolley is about to kill five people tied to the track. You're standing at a lever that can divert it onto a side track, where it will kill one person instead. Most people, asked this, say pull the lever — five lives outweigh one, and doing the arithmetic feels almost obligatory. Now change one detail: there's no lever. The only way to save the five is to push a large stranger off a footbridge into the trolley's path, where his body will stop it. Same trade, five for one. Most people now say no. The philosopher's question isn't which answer is right. It's why the same number of lives, traded the same way, gets a different verdict depending on how your hands are involved.
What the two versions actually change
Consequentialism, most famously utilitarianism, judges an action purely by its outcomes: if pulling the lever and pushing the man both end with one dead instead of five, a strict utilitarian calculus says they're equivalent, and the flinch most people feel at the footbridge is just squeamishness getting in the way of the correct sum. But the flinch is consistent and widely shared enough that it's worth taking seriously as data rather than dismissing as noise, and it points at a real distinction utilitarianism doesn't have room for: using a person's body directly as the means of saving the others, versus a side effect of redirecting a threat that was already in motion.
That distinction is closer to home in deontological ethics, and specifically in Kant's categorical imperative — the idea that treating a person merely as a means to someone else's end, rather than as an end in themselves, is wrong regardless of the resulting head count. The footbridge does that explicitly: the stranger's body is the tool. The lever doesn't, quite — the trolley kills the one person as a side effect of being diverted, not because his body was recruited to physically stop it. Two ethical frameworks, pointed at the same numbers, produce two different maps of where the moral line actually sits.
Why it's not really about trolleys
Nobody who studies this thinks runaway trolleys are a policy problem. The setup is deliberately stripped of everything except the moral variable being tested, which is exactly what makes it useful — real dilemmas are never this clean, and the trolley's whole job is to isolate one variable (how directly you're involved in the harm) instead of drowning it in the mess a real emergency would come wrapped in.
What we're still unsure about
The clean numbers are also the thought experiment's weak point. Real ethical decisions almost never come with guaranteed outcomes — you don't actually know the trolley will only kill one person on the side track, or that pushing the stranger will definitely stop it — and critics argue that stripping away that uncertainty to make the arithmetic tidy also strips away most of what makes real moral decisions hard. A thought experiment built to isolate one variable can end up teaching a lesson that doesn't transfer cleanly to situations where nothing is certain.
This sits inside Consequentialism & Utilitarianism, one of seven topics in Ethics, one of five domains in Philosophy, one of seventeen subjects the app can quiz you on.