The problem of induction,

sometimes known as Hume’s problem, has to do

with justifying a very basic sort of nondeductive

inference. We often seem to infer from observation

that some sample of a population has a certain

attribute to the conclusion that the next

members of the population we encounter will

also have that attribute. When you eat a piece of

bread, for example, you are concluding from the

many times in the past that bread has nourished

you, that it will also do so this time. But it is conceivable

that bread should have nourished in the

past, but not this time. It isn’t a necessary, analytic,

or a priori truth that the next piece of bread you

eat will be like the ones you have eaten before.

How does your inference bridge the gap? It is

natural to appeal to various general principles

that one has discovered to hold. But, as Hume

points out, the future application of principles

found reliable in the past presents exactly the

same problem. For example, consider the most

general principle of all, that the future will be

like the past. All one has really observed was

that, in the past, the future was like the past.

How does one know that in the future it will be?

The problem of induction is stated in Hume’s An

Inquiry Concerning Human Understanding, Section

IV, and discussed by Salmon, “The Problem

of Induction.”

inductive argument See deductive argument.