From Diagrams to Design: Overcoming Knowledge Acquisition Barriers for Case Based Design

One important and often overlooked implication of case based design systems is the acquisition of cases, which are typically hand crafted by system designers. Not only does this make case acquisition difficult and time consuming, but it may also introduce unanticipated biases and limitations into the system. In this paper we demonstrate a case based design system capable of automatically acquiring cases from diagrams, and of using those cases to solve new design challenges. Moreover, we demonstrate that our representation framework enables general teleological reasoning in design that transcends some domain borders, greatly increasing the versatility and usefulness of the cases thus acquired. Motivation and Goals Previous research establishes that case based design systems can synthesize new models from existing design cases to meet new functional requirements. One important and often overlooked implication of these systems, however, is the acquisition of cases. Typically, cases are hand crafted by system designers, and co-evolve with the system that will use them. Not only does this make case acquisition difficult and time consuming, but it may also introduce unanticipated biases and limitations into the system. Perhaps the most obvious limitation is that hand crafting takes valuable time and requires skilled knowledge engineers. While a prototype system may only require a handful of cases to prove its theoretic capability, when deployed in more complex and varied domains, resource demands for large numbers of hand crafted design cases may not scale. A second limitation is the flexibility with which design cases can be specified. Even with a syntactically well defined representationa