Evaluating Evidence for Motivated Discovery

Inductive discovery can be said to proceed through the accumulation of evidence in attempts to refute the current theory. Typically, this has involved a straight choice between the falsi cation of a theory and its continued use, with little or no consideration given to the possibility of error. This paper addresses the problems of error and evidence evaluation. It considers a variety of di erent kinds of error and uncertainty that arise through interaction with an external world, in both scienti c discovery and other inductive reasoning. In response, it proposes a generally applicable model for the evaluation of evidence under di ering motivations, and describes its implementation in the MID system for motivated discovery.