Case-based Learning for Design

We present a general framework for an adaptive, user-interactive design support system. Adaptability requires the acquisition and usage of huge amounts of highly structured knowledge. Interactiveness, i.e., the ability to perform and react in tight co-operation with the user, is of special importance in complex domains like industrial building design. Case-based learning will be applied to provide the knowledge necessary to support subgoaling, pre-organization of the case-base, similarity assessment, and adaptation. Prior work (B orner 1994a, B orner 1994b) focused on structural similarity assessment and adaptation. Our main intention here is case-based subgoaling and its interaction and relation to the former via some appropriate knowledge representation. The paper gives the underlying ideas and the basic algorithms of the framework. For its practical evaluation the framework has been partially implemented in SynGraph, a module of the knowledge-based system Fabel{Idea.