Experience, causation, and induction › The step from observed to unobserved

The argument usually called the problem of induction begins when Hume presses beyond the answer "experience." Suppose bread has nourished you in the past. That describes an observed pattern. Expecting apparently similar bread to nourish you again extends the pattern to another occasion. The new expectation contains something the record of earlier meals alone does not contain. Hume separates these propositions explicitly in E 4.16. Section IV, Part II.

Why take the further step? A demonstrative argument cannot establish that nature must continue in the same way, Hume argues, because a change in its course involves no contradiction. We can conceive apparently familiar objects producing unfamiliar effects. His claim is about conceivability and demonstration. He is not announcing evidence that nature is about to change or recommending that ordinary expectations receive equal weight with every imaginable alternative. E 4.18.

An argument from past experience seems more promising. We have relied on experience before, and it has often served us well. Why not trust it again? The trouble is that this answer itself moves from past cases to a new one. It assumes the very transition it was supposed to establish. To defend projection by saying projection has worked so far is to use projection in its own defense. Hume identifies that circularity in E 4.19. Section IV.

Return to the train. A detailed explanation involving timetables, staff, and maintenance can improve your prediction. It may reveal why this morning's punctuality differs from next week's circumstances. But the explanation still carries expectations about how those arrangements will operate outside the observed cases. Adding mechanisms can make an inference better informed without turning its factual premises into logical necessities. The example illustrates Hume's question; it is not an additional proof that all predictions are equally uncertain.

This last distinction prevents an overstatement. The challenge is stronger than the observation that predictions can fail: it asks for the foundation of the move that makes prediction possible. Yet Hume continues to distinguish careful reasoning from careless reasoning. Failure to provide a noncircular foundation for experiential inference does not by itself show that choosing between two evidence-based expectations is pointless. The remaining sections explain why inquiry and action continue.