Induction

Induction is the traditional proposal that we create knowledge by finding patterns in accumulated observations and generalizing from them — that evidence can support or make probable a conclusion. It is meant to solve the central epistemological problem of how knowledge is created, supplementing deduction so that we can reach empirical truths rather than only abstract ones.

CF, following Popper, holds that induction is a myth that cannot work even in principle. The decisive logical objection: infinitely many patterns fit any finite data set, and they imply contradictory conclusions about future cases, so induction gives no usable guidance about which pattern to extract. Slogans like “the future will resemble the past” smuggle in selective attention — the future resembles the past in infinitely many ways and differs in infinitely many ways. Worse, the supporting relation itself is illusory: any piece of evidence is logically compatible with infinitely many ideas, so evidence can only contradict or fail to contradict an idea, never single one out as more supported. This makes induction a species of justificationism.

Induction also presupposes the intelligence it claims to explain — pattern-finding requires the very judgment in question — and treats minds as passive vessels that absorb regularities, the bucket theory of mind CF opposes. CF’s replacement is conjecture and refutation: an evolutionary process of guessing bold explanations and eliminating errors by criticism. Crucially, CF treats induction’s failure as logically prior to comparisons of usefulness — a method that works poorly still beats one that cannot work at all. See also the problem of induction and CF’s wider critique of empiricism.