Imperceptible harms

Why does an act that harms nobody detectably still seem wrong when everyone does it?

Here is a principle almost everyone holds without noticing: if what I did made no difference anyone could detect, then I harmed nobody, and if I harmed nobody, I did nothing wrong. It sounds close to a definition. It is also the principle that quietly excuses one more flight, one more trawler's haul, and staying home on election day.

Derek Parfit's contribution was to build a case where that principle produces a result nobody will accept, which is the most useful thing a philosopher can do to an assumption you did not know you had.

Take his example. A thousand torturers each have a victim, and each turns a switch that causes his own victim severe pain. Now rearrange it: the thousand torturers each turn a dial by one thousandth of a degree on every one of the thousand victims. Each victim still ends in exactly the same severe pain, but no single turn of a dial now makes any difference the victim could ever notice.

So ask the awkward question: has anyone done anything wrong in the second version? Each torturer can truthfully say that their contribution was imperceptible, that removing it would have changed nothing anybody could feel. If the principle we started with is right, a thousand people have each done nothing wrong, and yet a thousand people are in agony that these thousand people caused. Something in the reasoning has to give.

What exactly went wrong? Look at where the word "detectable" entered. It smuggled in a threshold — the idea that harm begins only at the point where a difference becomes noticeable. But nothing in the physics of pain, or of temperature, or of a fish population, honours that threshold. The dial turned by a thousandth of a degree did something; our nervous systems are simply not fine-grained enough to report it. Perceptibility is a fact about detection, not about causation, and we have been using a limit of our instruments as a moral boundary.

That reframing helps a great deal, and I want to be honest that it does not finish the job. One might try to repair the principle by rejecting imperceptibility altogether: perhaps every act really does make a tiny difference, and each torturer causes a tiny harm which we simply cannot measure. Sometimes that is exactly right. But some cases seem genuinely discrete — there are versions where the outcome is unchanged whether or not you act — and there the tiny-harm move has nothing to work with.

A second repair is probabilistic, and it is the strongest one available. Even where an effect looks all-or-nothing, there is usually a threshold somewhere: fish stocks collapse at some level of take, a supplier reorders in batches, an election is decided by some particular margin. You do not know where the threshold sits, so your act carries a small probability of being the one that crosses it — multiplied by an enormous harm. The expected harm of your act is then not zero, and can be roughly the aggregate harm divided by the number of contributors, which is exactly the answer intuition wanted.

Is that a resolution? Partly. It works well for voting and for genuinely lumpy supply chains. It works less well for the continuous cases — ocean temperature does not have a step in it — and it changes the subject slightly, since it makes your act wrong because of a lottery it entered rather than because of the harm it plainly helped constitute. No treatment of these cases is fully satisfying, and the philosophical literature on collective impact is still very much open.

The misconception being corrected is specific, and worth stating plainly: it is not that small harms do not count, which nobody believes when it is put that way. It is the threshold rule — that an effect too small to perceive is an effect of zero size. Parfit's cases work by making the aggregate undeniable while every individual contribution stays under that threshold, so the rule has to be abandoned rather than adjusted.

What replaces it is a shift in the test. Instead of asking did my act make a detectable difference, ask am I a contributor to a set of acts that together cause serious harm, and what does that make me responsible for. That question has answers, even though they are messier: your share, your reasons for participating, and what alternatives you had.

It is also worth noticing what the puzzle does not license. It does not follow that any individual act of driving carries the moral weight of the whole climate problem — aggregation cuts both ways, and your share is a share. Nor does it follow that individual restraint is the right response rather than collective arrangements; where a harm is constituted by thousands of participants, the effective lever is often a rule that binds all of them, which is a practical conclusion the puzzle supports far more directly than a personal one.

• Whether voting is best defended by expected value or on grounds that have nothing to do with outcomes.
• How the same structure appears in the tragedy of the commons, and why the remedies there are institutional.

Every term the collection defines is gathered in the glossary.