Reflective equilibrium

When a cherished principle and a firm intuition collide, which one ought to give way?

You hold a principle you have thought hard about and would defend in public. Then a case arrives where the principle delivers a verdict you find, on reflection and without any wish to be difficult, plainly monstrous. One of them has to go.

The question as posed already assumes there is a boss — that either principles govern intuitions or intuitions govern principles, and the only work is to remember which. Both answers have distinguished defenders. Before choosing between them, it is worth asking whether the assumption survives inspection.

Take the tidy answer first: the principle wins, because principles are considered and intuitions are just feelings. Now ask where the principle got its authority. Almost certainly you accepted it because it captured, in a compact form, a range of cases you were already confident about. Strip those cases away and what is left? Not much — a form of words. So the principle's standing is borrowed from intuitive verdicts, which makes it awkward for it to overrule one on the grounds of superior rank.

Try the other tidy answer: the intuition wins, because intuitions are the data and theory answers to data. But ask the same question of it. Why do you trust this reaction? Partly because it fits with things you believe generally — it is not free-standing either. And we know some intuitions are worth very little: they shift when the same case is described differently, and they run reliably in the direction of the person's own advantage.

So neither item is a foundation, because each was already leaning on the other. Nelson Goodman made exactly this move in Fact, Fiction, and Forecast (1955), and about deduction rather than ethics, which is what makes it striking: a rule of inference is justified by conforming to inferences we actually accept, and an inference is justified by conforming to the rules. He called it a delicate process of mutual adjustment, and denied that the circle was vicious. Rawls took the method into moral philosophy in A Theory of Justice (1971) and gave the resulting state its name.

Once you see the two as mutually supporting, the question changes shape. It is no longer "which type outranks which" but "which of these two particular items is holding up less of the structure". That is answerable, and the answer varies case by case. What you are comparing is independent support: how much each item is backed by things other than the one it is now fighting with.

That gives us something to test. For intuitions, Rawls proposed a filter — he counted only considered judgements, made where we are calm rather than upset, informed about the facts, not standing to gain, and confident rather than hesitant. An intuition that fails those tests is downgraded before the contest begins. For principles, the analogous test is breadth: a principle earning its keep across many independent cases has more claim than one that was reverse-engineered from a favourite example. Neither test is mechanical, but both are real work, and they are the reason this is not merely a shrug.

Now what is the stopping point? Not agreement, and not comfort. It is a state where every revision still available to you would cost more coherence than it buys — move the principle and you lose more cases than you gain, discount the intuition and you undermine judgements you rely on elsewhere. That is what makes it an equilibrium rather than a verdict: it is a balance of forces, it holds only until something new pushes on it, and it is explicitly revisable.

Two honest complications. Norman Daniels distinguished narrow equilibrium, which balances only principles against judgements, from wide equilibrium, which brings in background theories — of the person, of society, of what morality is for — to give the system some purchase from outside the pair. And the strongest objection has never been answered to everyone's satisfaction: coherence among unreliable inputs is achievable without getting anywhere near the truth. Peter Singer pressed this hard, arguing that intuitions are the residue of evolution and upbringing and deserve no evidential standing at all. There is no agreed theory of when an intuition is signal and when it is noise, and anyone presenting this method as settled is overselling it.

Three clarifications. First, this is not "split the difference". Very often one side loses outright — if the intuition is a considered judgement backed by a hundred parallel cases and the principle was formulated last week, the principle should go, and saying so is the method working, not failing. Second, it is not relativism by another name. It does constrain: an equilibrium must be internally consistent, and the filter on considered judgements excludes a great deal. But it also does not guarantee a unique destination, and the possibility that two careful people settle at different equilibria is a live objection rather than a feature. Third, the equilibrium is a resting point, not a conclusion. A new case, or a better theory of why we react as we do, reopens it.

• Wide reflective equilibrium and whether background theories genuinely break the circle or merely widen it.
• Singer's debunking argument, and the experimental work on framing effects that gives it teeth.
• Whether two reasoners can reach incompatible equilibria and what, if anything, follows.
• Goodman's original version for inductive and deductive rules, where the same structure appears without any moral content.

Every term the collection defines is gathered in the glossary.