Learning design knowledge as generalization

Generalizing design knowledge is a process of extracting knowledge from design-provideddata. Although it is one of the most common waysof “learning” design knowledge, generalizationhas a fundamental weakness: except for special occasions, the results of generalization can never be validated. Inquiries into generalization have therefore dealt with questions ofwhat are the best criteria for guiding the generalization. This paper argues that generalizing design knowledge involves three essential tasks; knowledge representation, a description languagefor design examples, and generalization operators. An application of generalizing empirical networks from design examples is described to illustrate these three tasks in the development of a generalization system.