Scientific Discovery in the Layperson

In this chapter, we argue that the process of scientific discovery is in many ways similar to the process that the layperson uses to learn and evaluate predictive relationships, with the point that that these techniques should be incorporated into models of scientific discovery. We review some findings that illustrate constraints on human learning and show how these findings manifest themselves in scientific discoveries. We concentrate on two aspects of the discovery process: evaluation of evidence, and scientific rhetorics, illustrating these aspects by two programs that perform common-sense reasoning tasks. OCCAM uses a theory of causality to constrain the learning of causal relationships. We show that this system can use the same mechanism to learn about the process of catalysis. ABDUL/ILANA models the process of argumentation. We show how the system can represent competing scientific hypothesis and use argumentation strategies to direct the reasoning process. We conclude with some observations about the consequences of this view of discovery. 1