Testing the Basis for an Automated Design-by-Analogy Tool Through Comparison to Expert Thinking

Abstract

Our prior work presented a method for computationally structuring patent databases as a basis for an automated design-by-analogy tool. In order to examine the validity and sensibility of the prior work as the basis for a design tool, its output is compared in detail to expert designers’ mental models of the domain being structured, i.e., a set of 45 patents and their inter-relationships. The comparison sought first to gauge the intuitiveness and sensibility of the computational method of structuring to human minds, and further to ascertain whether any differences between the method’s and the experts’ structures indicate potentially novel or surprising ways of approaching the space of patents, or indicate that the output was nonsensical, invalid or needing modification in order to be useable. The results indicate that, when compared to expert thinking, the computationally generated structure is sensible in its clustering of patents and in its organization of these clusters into a structure or space. The results also suggest that the computationally-generated structure represents a version of the patent space upon which experts can find common ground and consensus — making it likely to be intuitive and accessible to a broad cohort of designers. Thus, the prior work which presented a computational method for structuring design databases has been found to offer a resource-efficient way of usefully representing the space that is sensible to expert designers, while still preserving an element of surprise and unexpectedness, making it promising as the basis for a computational design-by-analogy inspiration tool.