Cognitive Mechanisms of Learning

4Learning a Concept from Examples of Concepts: Induction

Induction “is an ‘inferential’ approach, which, from premises of a specific scope, taken as truths, aims to obtain a more general conclusion; however, we cannot usually be absolutely certain that this conclusion is true” (Guy Politzer, 1990). Inductive reasoning plays a part in a large number of everyday situations. For example, when I receive a package in the mail, it is often delivered before 9am. Based on these observations and on my knowledge of mail delivery circuit times, I induce the knowledge (potentially false) that “I live in a part of town which is close to the mail distribution center”.

The “learning a concept from examples of concepts” situation, introduced in Bruner, Goodnow and Austin’s pioneering A Study of Thinking (1956), was the first learning “paradigm” put forward in cognitive psychology and AI. Notwithstanding increasingly sophisticated adaptations¹, this paradigm still lies at the heart of most models of human and machine learning.

The authors considered that human beings have a fundamental capacity to acquire and use categories. Firstly, by categorizing different events as equivalent, the organism reduces the complexity of its environment. For example, we learn to consider the leaves of all of the trees in a forest in summertime as “green”, despite the vast range of different hues. Categorization is also a means of identifying objects, events, etc. – for example, the ability to identify a noise ...