A cognitively based approach to computer integration for design systems

The kind of relationship a designer will have with a design assisting tool is dependent upon the kind of dialogue possible between the two, and any dialogue will be conditioned by the degree of mutual understanding existing between designer and assistant. In the first part of this paper we examine the methodology, currently prevalent in the computer-aided building design domain, that is supposed to address this problem: product modelling. We argue that product modelling is inherently incapable of fulfilling the goals that it sets itself primarily because it is an attempt to move a methodology from, and appropriate to, a closed domain to an unbounded domain. In doing so, it fails to address the basic problem: it fails to provide a medium for dialogue. Dialogue entails mutual understanding; to support a dialogue between tools and designer, it is not sufficient for the designer to learn about the tool, the tool's designers must develop a strategy for enabling the tool to understand the designer. The second part of this paper comprises the theoretical background for a pragmatic strategy for classifying representations held by design agents by mapping the internal structures of differing representations of objects onto one another. Research in cognitive science has shown that such structure mappings can form the cognitive basis for certain kinds of classification decisions. We briefly describe ongoing work aimed at applying these insights in order to enable machines to classify, and thereby interpret, nonspecific design representations. Copyright © 1996 Published by Elsevier Science Ltd.