The Problem of Induction

Inductive arguments project observed regularities to unobserved cases. It is, however, notoriously difficult to justify this method of reasoning. The infamous “problem of induction” is, it has seemed to many, intractable.

Inductive arguments appear to rest on an assumption that nature is uniform. Such arguments take the past to be a good guide to the future, or observed cases to be a good guide to unobserved cases. This will only work if there are patterns in nature, if things are generally the same as each other.

How can this assumption be justified? How can we be sure that nature is uniform, that things will carry on as they have always carried on? The only obvious evidence that we have for this principle is past experience: thus far, past observations have been a good guide to the future. Based on the success of induction in the past, we infer that it will continue to be successful in the future.

This, though, is an inductive argument; it takes the past to be a good guide to the future. The argument in support of the uniformity of nature therefore assumes precisely what it is trying to prove: that nature is uniform. Such circular arguments as this commit a logical fallacy: begging the question; they prove nothing.

To see what is wrong with this method of reasoning, we may imagine a community of counter-inductivists, a community who reason that the past is a bad guide to the future, the unobserved cases will be unlike observed cases. Counter-inductivists

We, it seems, have no more evidence for the efficacy of induction than counter-inductivists have for the efficacy of counter-induction. Our preference for induction appears to be entirely arbitrary. This is the problem of induction.