Problem Solving in Open Worlds

This book presents a computational model of design problem solving in which problems may be under-specified and domain knowledge may be incomplete. The approach combines plausible inference methods for generating, adapting, and evaluating designs, together with a flexible control strategy that accommodates constraints emerging late in the design process.

Novel features of this model include a problem-solving architecture that integrates case-based reasoning and constraint posting to generate plausible designs, domain-independent algorithms and heuristics for adapting proposed designs, and new mechanisms and metrics for evaluating solutions in terms of their integrity and completeness. This model of design is implemented in a computer program called JULIA that interactively designs the presentation and menu of a meal.