Sean Daily

Somewhere right now, an engineer is writing the rules for what a self-driving car should do in the split second before a crash it cannot avoid. Swerve toward one pedestrian to spare three? Protect the passenger at any cost? Strip away the code and the sensors, and you are left with a puzzle philosophers have argued about since the 1960s: the trolley problem. It looks like a simple quiz about numbers. It is really a question about whether how a harm happens matters as much as how many people it touches. Here is the trolley problem explained, from its odd origin to the versions that still split rooms, so you can decide where you stand.

〰

The Trolley Problem Explained: Where It Came From

The famous trolley started life as a side example. In 1967 the Oxford philosopher Philippa Foot published “The Problem of Abortion and the Doctrine of the Double Effect” in the Oxford Review. Her goal was to get clearer about abortion by comparing it with cases involving people who were already born, so that the question of whether a fetus has full rights could be set aside for a moment.

One of those comparison cases was a runaway tram. You are the driver. It can be steered but not stopped. Ahead are five people; on a side track there is one. Should you turn? Most readers say yes.

Foot then set it beside a darker case. A judge faces rioters who will take revenge on part of the community unless someone is punished for a crime. He could frame one innocent person and save several lives. The numbers are the same as the tram, one against more than one, yet almost nobody thinks the judge should do it. Why the difference? Foot’s answer, as Britannica summarizes it, was that negative duties (the duty not to harm) weigh much more than positive duties (the duty to help). The driver who turns is choosing between killing one and killing five. The judge is choosing between killing one and failing to save five.

Keep that original setting in mind. The trolley was born inside a debate about medicine and the limits of what we may do to one person for the sake of others, not as a party game.

〰

Thomson’s Twist: The Bystander and the Footbridge

Judith Jarvis Thomson (1929–2020), who taught at MIT for four decades, gave the puzzle its standard form in “Killing, Letting Die, and the Trolley Problem” (1976) and its sequel, “The Trolley Problem,” in the Yale Law Journal (1985). She made two changes that broke Foot’s tidy answer.

First, she took you out of the driver’s seat. Now you are a bystander standing by a switch. If you do nothing, the trolley kills five. If you pull the lever, it kills one. You are no longer choosing between killing one and killing five; you could simply walk away and let it happen. Foot’s duty-based answer predicts you should not pull. Yet most people still say pulling is allowed.

Second, she moved you onto a footbridge. There is no switch. Next to you stands a large man. If you push him off, his body will stop the trolley and the five will live. Same arithmetic: one dies, five are saved. Ask yourself honestly whether you would push. Most people say no, and many say it would be monstrous.

That gap is the trolley problem in the narrow sense. It is not “should you save five or one?” It is: what explains why switching seems acceptable and pushing does not, when the outcomes are identical? Thomson called it, according to a New Statesman retrospective, “a lovely, nasty difficulty.”

Case	What you do	Typical verdict
Driver (Foot, 1967)	Steer the tram from five onto one	Permissible
Bystander (Thomson)	Pull a switch to divert the trolley onto one	Permissible
Footbridge (Thomson)	Push a large man off a bridge to stop the trolley	Impermissible
Loop	Divert onto a side track that rejoins the main line, where the one person’s body stops the trolley	Disputed
Transplant	A surgeon kills one healthy patient to give organs to five	Impermissible

〰

Three Ways to Explain the Gap, and Where Each Breaks

1. The doctrine of double effect

The oldest candidate goes back to Thomas Aquinas, who wrote in his discussion of self-defense that “Nothing hinders one act from having two effects, only one of which is intended” (Summa Theologica II-II, q. 64, a. 7). The idea, developed by later writers into a set of conditions, is roughly this: you may sometimes bring about a bad effect you foresee but do not intend, if the act itself is not wrong, the good is not achieved by means of the bad, and there is a serious enough reason.

Applied to trolleys: when you switch, the one person’s death is a side effect. If he somehow escaped, you would be delighted. On the footbridge, his body is the very thing that stops the trolley. You need him to be hit. That is intending harm as a means.

The trouble, as the Stanford Encyclopedia of Philosophy notes, is the “closeness problem.” Did you intend the man’s death, or only that his body make contact with the trolley? If the line between “means” and “side effect” can be redrawn that easily, the doctrine may not be doing the work it claims. Experiments by Joshua Knobe add a further worry: people’s moral reactions seem to shape whether they describe an outcome as “intended” in the first place.

2. Don’t use people merely as a means

A Kant-inspired answer says pushing treats the man as a tool, while switching does not. This fits the footbridge and the transplant surgeon nicely. Then comes the Loop. In that variant the side track curves back to the main line, so diverting saves the five only because the one person’s body blocks the trolley. Now switching also uses him as a means. Many people still feel switching is acceptable here, and the Kantian rule has a hard time saying why.

3. Redirecting a threat vs. creating a new one

Thomson’s own proposal was that you may redirect a threat that already menaces the five onto the one, provided doing so does not violate the one’s serious rights by other means. Switching deflects the existing trolley. Pushing starts a new attack on a new victim. Later philosophers, including Frances Kamm, built cases designed to strain even this rule.

And here is a twist that surprises many readers. In a 2008 paper, “Turning the Trolley,” Thomson argued that the bystander should not pull the switch after all, reversing the position she had defended in 1976 and 1985. The philosopher who made the case famous ended up rejecting the intuition most people bring to it.

〰

From Armchair to Lab: Brains, Surveys and Cars

In 2001 Joshua Greene and colleagues published an fMRI study in Science using dilemmas like these. Their central claim was that dilemmas differ in how strongly they engage emotional processing, and that those differences shape the judgments people make. Many readers took this as a hint that the footbridge “no” is partly an emotional reaction to up-close, personal force rather than the output of a hidden principle. Others replied that an emotional response can still be tracking something morally real. The data alone do not settle which.

Then the trolley went online. MIT’s Moral Machine experiment, reported in Nature in 2018, gathered 40 million decisions in ten languages from people in 233 countries and territories, all about what an autonomous vehicle should do in unavoidable crashes. The researchers found broad global preferences, but also three major clusters of countries whose preferences differed, and those differences lined up with institutions and deep cultural traits.

That finding cuts two ways. If you think intuitions reveal moral truth, which culture’s intuitions count? If you think they are just habits, why build them into cars at all?

There is also a backlash. Critics complain that “trolleyology” has become a mini-industry of ever more convoluted scenarios whose intuitions are hard to trust. Real emergencies come with uncertainty, panic and incomplete information; the textbook trolley offers perfect knowledge and exactly two options. You might reasonably ask whether a case that strips out everything messy is still teaching you about real life.

〰

The Choice the Trolley Leaves You With

Go back to the engineer and the car. If outcomes are all that matter, the code is easy to state: minimize deaths. But that same rule, applied consistently, also tells the surgeon to harvest organs, and almost no one accepts that. If how harm happens matters, then the car’s designers need a principle that distinguishes swerving from “using” someone, and you have just seen how slippery those principles are.

So you are left with a trade-off rather than an answer. You can trust the footbridge intuition and accept a moral theory with awkward exceptions. Or you can trust the arithmetic and accept conclusions that feel wrong in your stomach. Thomson herself moved between camps over thirty years. Which would you rather defend in front of the one person on the side track?

〰

Questions to Think About

• Would you pull the switch but refuse to push? If so, try to state the rule that explains both answers. Does your rule survive the Loop?
• If the people on the tracks were your family, would your answer change? Should it?
• When a gut reaction and a principle disagree, which one should give way, and how would you know?

〰

FAQ

Q: Who invented the trolley problem?

Philippa Foot introduced the runaway tram case in 1967. Judith Jarvis Thomson developed the bystander and footbridge versions in 1976 and 1985, and those versions are what most people mean today.

Q: Is there a correct answer to the trolley problem?

No answer has won general agreement. The main proposals (double effect, harming vs. not helping, not using people as means, redirecting threats) each fit some cases and stumble on others.

Q: What does the trolley problem have to do with utilitarianism?

A view that counts only outcomes says to save the five in every version, including the footbridge. The puzzle is pressing precisely because most people’s judgments do not follow that pattern, which suggests something beyond outcomes is shaping them.

Q: Do self-driving cars really face trolley problems?

Designers do have to decide how vehicles behave in unavoidable crashes, which is why projects like the Moral Machine exist. Critics point out that real crashes involve uncertainty the clean thought experiment leaves out.

This article is an introduction, not a substitute for reading the original texts.