Doing versus allowing

Why does causing a harm feel worse than allowing the very same harm?

A runaway trolley will kill five people. You can divert it onto a side track where it will kill one. Most people say: divert. Now a surgeon has five patients dying of organ failure and one healthy visitor whose organs would save them all. Almost nobody says: cut. The arithmetic is identical — one dies, five live — so something other than the arithmetic is doing the work.

The obvious candidate is that in the second case you do the killing, while in the first you redirect a harm already in motion. But naming a feeling is not justifying it. What exactly is the difference, and does it deserve the weight we put on it?

Notice first what the paired cases are for. Any single case has too many confounds — fear, disgust, professional role, the surgeon's promise to patients. Pairs that match on everything except the suspect variable are the philosopher's controlled experiment, which is why the literature is full of them.

So run the tightest pair. Smith stands to inherit if his young cousin dies, and drowns the boy in the bath. Jones stands to inherit likewise, walks in intending to drown him, and finds the boy has slipped under by himself — so he watches, ready to push the head down if it surfaces. James Rachels put this case in 1975 to make one point: if doing versus allowing carried intrinsic weight, Smith should be worse than Jones. Most readers judge them equally monstrous. Motive, knowledge, ease of rescue and outcome were held fixed, and the asymmetry vanished with them.

That should make us suspicious rather than dismissive. It suggests the asymmetry we feel in ordinary cases may be tracking things that usually travel with doing — worse motives, more certainty of harm, less cost to acting otherwise — rather than doing itself.

Still, four candidate distinctions are seriously defended, and each one earns a counterexample.

Agency. Perhaps what matters is whether the harm flows from you or from a threat you merely fail to stop. But in the trolley case the harm does flow from your hand on the lever, and most still permit it — so causal agency alone cannot be decisive.

Intention, via the doctrine of double effect. Perhaps harm you intend is worse than harm you merely foresee: the surgeon needs the visitor dead, the diverter would be delighted if the one man escaped. This handles the pair elegantly — and then meets the loop variant, where the trolley is diverted onto a side track and stopped only by the body on it. Here the one man's death is needed, yet many still permit diverting, which is exactly the reverse of what double effect predicts.

Using a person as a means. A close cousin: the visitor is a resource, the man on the side track merely unlucky. The loop case bites here too, and so does the objection that "using" is hard to define without already knowing the verdict you want.

Rights against interference. Perhaps our duty not to harm is simply stronger than our duty to aid, so a claim not to be cut is more stringent than a claim to be saved. This fits a great deal of ordinary morality — and it is what makes the Smith and Jones pair uncomfortable, since Jones violates no such right and is nonetheless a monster.

One empirical footnote is worth having. The asymmetry is not a local quirk: large studies of moral judgement across many countries find the doing/allowing and intention effects showing up broadly, though their size varies with culture and framing. But robustness is a fact about us, not a licence. That an intuition is near-universal is compatible with its being an artefact of how minds evolved — the debunking argument several philosophers press — and others answer that a widely shared, stable judgement is the only kind of evidence moral theory ever has.

Three refinements hold the pieces together.

First, the asymmetry is real as a description of judgement and unsettled as a claim about morality. Almost everyone agrees people judge this way; nothing yet follows about whether they should.

Second, the candidates are not rivals to be scored and ranked. Real cases activate several at once, which is why isolating one requires such contrived scenarios — and why the contrivance is a legitimate complaint, since intuitions about surgeons and trolleys may not transfer to famine relief or emissions.

Third, even the canonical readings shift. Judith Jarvis Thomson, who did more than anyone to make the trolley case famous, later argued you may not turn the trolley. When the field's own authorities revise their verdicts on the founding example, treating any of this as settled would be a mistake.

• Whether the distinction can bear the institutional weight placed on it in medical ethics, where withdrawing treatment and administering a lethal dose are treated very differently.
• The debunking debate: what, if anything, follows for morality from discovering the psychological origins of a moral intuition.

Every term the collection defines is gathered in the glossary.