Diagrammatic Reasoning: An Artificial Intelligence Perspective

摘要

A common motivation for developing computationalframeworks for diagrammatic reasoning is the hope thatthey might serve as re-configurable tools for studyinghuman problem solving performance. Despite the ongoingdebate as to the precise mechanisms by which diagrams,or any other external representation, are used inhuman problem solving, there is little doubt thatdiagrammatic representations considerably help humanssolve certain classes of problems. In fact, there area host of applications of diagrams and diagrammaticrepresentations in computing, from data presentationto visual programming languages. In contrast to boththe use of diagrams in human problem solving and theubiquitous use of diagrams in the computing industry,the topic of this review is the use of diagrammaticrepresentations in automated problem solving. Wetherefore investigate the common, and often implicit,assumption that if diagrams are so useful for humanproblem solving and are so apparent in humanendeavour, then there must be analogous computationaldevices of similar utility.