Overview
The sun has risen every day for billions of years. So it will rise tomorrow, right? David Hume said: “Prove it.”
Core Idea
Induction: Reasoning from specific observations to general rules. (I saw 100 white swans -> All swans are white). The Problem: You can’t justify induction using logic (it’s not a logical contradiction that the sun won’t rise). You can’t justify it using experience (that’s circular: “Induction worked in the past, so it will work in the future” is induction).
Formal Definition (if applicable)
Uniformity of Nature: The assumption that the future will resemble the past. This is the unprovable axiom of science.
Intuition
The turkey is fed every day by the farmer. Every day, its confidence grows that the farmer loves it. Then comes Thanksgiving. The turkey used induction.
Examples
- Black Swan: Europeans believed all swans were white until they went to Australia.
- Grue Paradox (Nelson Goodman): Define “grue” as green before 2030, and blue after. Emeralds observed today are green, so they are also grue. Induction supports “All emeralds are green” AND “All emeralds are grue.” But in 2030, they can’t be both.
Common Misconceptions
- “Science proves things.” (Science supports theories, but it can never strictly prove them because of the problem of induction. A new observation could always overturn the theory.)
Related Concepts
- Falsificationism (Popper): Science doesn’t use induction to prove true; it uses deduction to prove false. We can never prove all swans are white, but seeing one black swan proves they aren’t.
- Abduction: Inference to the best explanation.
Applications
- Machine Learning: AI is basically an induction machine. It assumes past data predicts future data. (Overfitting is the failure of induction).
- Risk Management: “It’s never happened before” doesn’t mean it won’t.
Criticism / Limitations
We have no choice but to use induction. It’s “animal faith.”
Further Reading
- Hume, An Enquiry Concerning Human Understanding
- Goodman, Fact, Fiction, and Forecast