Skip to main content

A chicken on a farm is fed every morning by the same man. Day after day the pattern holds, and if the bird could reason it would grow more confident with each sunrise that breakfast is coming. Then one morning he arrives with a knife. Bertrand Russell told this story in 1912, and his verdict was dry: "The man who has fed the chicken every day throughout its life at last wrings its neck instead, showing that more refined views as to the uniformity of nature would have been useful to the chicken." Much of our own reasoning about nature, it turns out, is built the same way.

What the field asks

Scientists ask what the world is like. Philosophers of science ask a question one step back: what makes a method a good way of learning about the world, and what scientific knowledge actually amounts to. Can evidence ever prove a general law? What separates science from its imitations? Courts, health agencies and voters face these questions constantly.

Hume's problem

The oldest of these questions was posed by the Scottish philosopher David Hume. In his Enquiry Concerning Human Understanding of 1748 he noticed that every step from observed cases to unobserved ones rests on an assumption. In his words, "all inferences from experience suppose, as their foundation, that the future will resemble the past." We have seen bread nourish us, so we expect the next loaf to do the same. But what justifies the assumption that nature is uniform?

It cannot be proved by logic, because a world in which the pattern suddenly breaks involves no contradiction. Nor can it be proved by experience without going in a circle, since the argument that induction has worked before and so will keep working is itself an inductive argument. This is the problem of induction: reasoning from observed cases to unobserved ones seems to have no justification that does not assume what it sets out to prove. Hume's point was about justification, not practice: we cannot help expecting regularity, but we cannot show the expectation is rational.

Answers that have been tried

One tempting escape is to say that induction delivers probability rather than certainty. The difficulty returns at once, because saying the next loaf will probably nourish us still projects a past pattern forward. A second answer, developed by the German philosopher Hans Reichenbach in the 1930s, is a pragmatic vindication: if any method can find the patterns in nature, induction will, so it is our best bet even though it cannot be proved to succeed.

Many philosophers today prefer Bayesian approaches, named after the eighteenth century minister Thomas Bayes. They treat belief as a matter of degree and use the rules of probability to say exactly how much a piece of evidence should shift confidence in a hypothesis. Critics reply that the starting degrees of belief still need a source, so Hume's question simply moves there.

Two newer puzzles

In the twentieth century the problem grew new heads. The German American philosopher Carl Hempel noticed in the 1940s that confirmation behaves strangely. A black raven plainly supports the law that all ravens are black. But that law is logically equivalent to the claim that every thing that is not black is not a raven. A green apple fits the second claim. So, by apparently innocent steps, a green apple or a white shoe seems to confirm that all ravens are black. This is the raven paradox. A common reply accepts the conclusion but shrinks it: since non-black things vastly outnumber ravens, a white shoe confirms the law, but only by a tiny amount.

The American philosopher Nelson Goodman then posed a sharper puzzle in 1955. Call an object grue if it is examined before some future date and found to be green, or not examined before then and is blue. Every emerald ever examined has been green, and so also grue. The evidence fits "all emeralds are green" and "all emeralds are grue" equally well, yet the two predict different colors for emeralds first examined after the date. Goodman's lesson was that the hard question is not only whether to project patterns but which categories are fit to project. His own answer appealed to terms with a long record of use, which he called entrenched.

Worked example: a new drug

A trial gives a blood pressure drug to 4,000 adults and finds that it works without serious side effects. Regulators approve it for millions of people. Every part of that step is inductive: the trial patients stand in for future patients and the studied group for the wider population. Hume's problem does not make the approval irrational, but it explains how the reasoning can fail. If the trial left out pregnant women, older adults or people on other medicines, the pattern may not carry over, much as the chicken's pattern did not. Goodman's puzzle appears here as a practical choice of category: does the evidence concern "adults," or "adults like those in the trial"? The data alone cannot settle which description to project. That is why monitoring continues after approval. Every generalization is a bet on the uniformity of nature, and the bet is safest when the sample resembles the people it is meant to protect.

Quiz

5 questions · Practice: scored, not saved

1. What lesson did Goodman draw from the grue puzzle?
2. What is the problem of induction?
3. What is a common reply to the raven paradox?
4. What makes an object grue, in Goodman's definition?
5. According to Hume, what do all inferences from experience assume?
6. What is Reichenbach's pragmatic vindication of induction?
7. What does the raven paradox suggest a white shoe can do?
8. Why, on Hume's view, can logic not prove that nature is uniform?

CivilSimian.com created by AxiomaticPanic, CivilSimian, Kalokagathia