Definition
Induction is the mode of inference that moves from a finite set of particular observations to a general claim that extends beyond them. From "every swan observed so far has been white" the inductive reasoner concludes "all swans are white" — a step that takes the conclusion past anything the evidence strictly entails. Inductive arguments are evaluated as stronger or weaker, not as valid or invalid: a good inductive argument makes its conclusion likely, but never necessary.
This is the engine of empirical science, statistical generalisation, and everyday pattern recognition. It is also the source of one of philosophy's deepest puzzles — the problem of induction, raised by Hume — which asks what licenses the inferential leap from past regularities to future ones.
Why it matters
How it works
The basic inductive schema enumerates instances and generalises: a is F, b is F, c is F; therefore all members of the relevant class are F. The argument can be strengthened by varying the sampled instances — different times, places, and conditions — so that the generalisation does not silently piggyback on a confounded common cause. It can also be strengthened by widening the sample size, since a regularity observed across many cases is less plausibly a coincidence. None of these moves makes the conclusion certain; they raise its probability.
Induction includes more than naive enumeration. Analogical inference, inference to the best explanation, and statistical reasoning are all inductive in the technical sense — each draws conclusions whose content exceeds the premises. The pivotal feature is that the truth of the premises does not guarantee the truth of the conclusion; it merely supports it. A single black swan invalidates an inductive generalisation that millions of white sightings had appeared to confirm. Induction is therefore the form of reasoning under which scientific theories are always provisional, and any future observation can in principle force revision.