Counting happiness

Mill made two changes in chapter 2 that matter for every motion here.³ The first is quality. Some pleasures are better than others, not just larger, and anyone who has known both prefers the higher kind: "better to be Socrates dissatisfied than a fool satisfied." The second is impartiality. The happiness that counts is not the agent's own but that of everyone concerned, and the agent must be "as strictly impartial as a disinterested and benevolent spectator." Mill also answers the obvious complaint that nobody has time to calculate. People have been learning the effects of actions for the whole of human history, he says, and ordinary moral rules are the record of what they learned.

The theory also leaves things out, and they matter as much. In the same chapter Mill says motive has nothing to do with whether an act is right, though it has a great deal to do with how we judge the person. Someone who saves a drowning stranger for a reward has still done the right thing. Nothing has value unless it touches someone's experience. This makes utilitarianism the school most at ease with AI systems on their own terms. A system that reliably reduces suffering is doing good, whether or not anyone can say how.

Bentham applies the same test to blame. In chapter 13 he says that all punishment is itself a harm, and is justified only to the extent that it prevents a greater one.⁴ Blame is a tool for changing what people do next. Where it cannot change anything, it is cost without return.

Each motion turns on one of these commitments. "Crash arithmetic" turns on impartiality: the passenger counts as one person, no more and no less. "The opaque benefit" turns on what explanation is for: it was always a means of making systems safer, and a system already shown to be safe has met the end without it. "The careful builder" turns on chapter 13, since punishing care that failed by bad luck prevents nothing. "Borrowed judgement" turns on Mill's view of motive, and asks what exactly is lost if your choices improve and you are happier.

The standard objection is that counting can license things nobody should accept, such as sacrificing one innocent person to save five. A virtue ethicist adds that an impartial spectator has no friends, and that a theory with no room for the special weight of your own child has misdescribed love. Mill's reply is that the rules people have learned, including the rule against killing innocents, are themselves the best available guide to the general happiness. That defence is strong. It is also an admission that the counting cannot always be done directly, and a debater on this side should know where the admission leads.