Simplicity and Complexity in Human Concept Learning

The topic I’d like to talk to you about today is, I think, one of the oldest and most basic in cognition: how we learn from examples. As most famously pointed out by Hume, when we make a finite number of observations of an enduring phenomenon, there is no strictly logical (i.e., deductive) basis for forming any firm generalizations about it. Instead we must “induce,” that is, make educated guesses about what its general properties might be. The need for this is especially clear in the case of category learning, or as it is sometimes called, concept formation. We see a few examples of a category—say, a straight-backed chair, a plush armchair, and threelegged stool—and must guess the true form of the category (chairs).