The problem of induction

Past patterns help us predict the future, but past success alone cannot logically guarantee that a pattern will continue.

Hume asks what justifies moving from observed cases to unobserved ones. Saying that induction has worked before already assumes the kind of inference being defended. The puzzle concerns justification, not whether people can learn from experience.

For example: A bridge surviving yesterday is relevant evidence about tomorrow, but a new load or hidden corrosion can change the outcome.

Keep in mind: This does not make science pointless or every prediction equally uncertain. Different accounts explain why some inferences are better supported than others.

Think about it

Which assumption lets one of your predictions move from past observations to a future case?

Connected ideas

Bayes rule and belief updates · Falsifiability and its limits · Correlation is not causation

Sources and further reading

• Stanford Encyclopedia of Philosophy: The Problem of Induction

NoteGuild starter collection · Original educational summary; source authors are not represented as platform members.