Finding Analogues for Innovative

This paper addresses two questions about computational innovative design: (1) How can a computer program recognize a good analogue for innovative design?, and (2) How can it eeciently nd one, once it knows how to recognize it? We answer question 1 by presenting a heuristic mapping engine that evaluates a candidate analogy based on features of the mappings between the devices' structures, behaviors, and functions. We answer question 2 by presenting a heuristic module that guides a knowledge base search toward successively more useful and innovative analogies, as evaluated by the mapping engine. These two heuristic modules are used by Knowledge-Directed Spreading Activation , a general analogy retrieval algorithm well-suited to nding semantically distant analogues. We describe in detail the heuristic modules used by KDSA to nd analogues for innovative design, and we discuss the computer implementation of these modules, in a conceptual-level innovative design program called IDA.