Nicod's 1930 criterion holds that observing a confirming instance increases belief in a universal generalization

confirmation-theoryphilosophy-of-sciencejean-nicodraven-paradoxepistemicsbayesian-reasoning

French philosopher and logician Jean Nicod (1893–1924) proposed a rule now known as Nicod's criterion for how evidence confirms a universal hypothesis; it reached English-language readers through Foundations of Geometry and Induction (1930), the posthumous translation of his French work — the reference Leike & Hutter themselves cite is the 1961 Presses Universitaires de France edition of Le Problème Logique de L'Induction. As Jan Leike and Marcus Hutter gloss it in their 2015 paper testing the criterion against Solomonoff induction: "observing an F that is a G increases our belief in the hypothesis that all F s are Gs." In its classic illustration — "all ravens are black" — seeing a black raven should raise confidence in the hypothesis, while an irrelevant observation (a red herring, a white shoe) should not.

The criterion is a foundational move in twentieth-century confirmation theory and underlies Hempel's raven paradox: the logical equivalence of "all ravens are black" and "all non-black things are non-ravens" implies, by Nicod's own rule, that observing a white shoe should also confirm "all ravens are black" — a conclusion most readers find intuitively absurd. Nicod's criterion and the paradox it seeds have a century-long afterlife in philosophy of science, and — via Leike & Hutter's 2015 proof that Solomonoff induction violates it — a second life inside the mathematics of universal artificial intelligence.