Discovery of Mental Metadata Used for Analogy Formation in Function-Based Design

Abstract

Applying previous solutions to solve new problems is a core aspect of design. In this context, analogies provide a mechanism to reapply previous solutions in new ways, but analogy formation is limited by a designer’s knowledge. One approach toward improving a designer’s analogy-forming capabilities is to provide an easy-to-use computational means of retrieving a wide breadth of relevant analogies. This work aims to answer what types of similarity are commonly used to draw design analogies, and whether some types of similarity are used more frequently in compound analogy versus single analogy. In this study, an experiment was performed to observe and document the types of information that designers found useful when forming analogies during conceptual design. A categorization of this information is sought in order to inform (1) the types of similarity data to store in an intuitive design-by-analogy database and (2) the form that a search query should take. The experiment consists of a design task and a follow up interview. Ten mechanical engineering graduate students specializing in design participated. These participants were interviewed, and their internal knowledge queries were encoded to reflect their objectives, thought process detail, direction of reasoning, and subject behavior type. Each conceptual design is cataloged according to whether it represents a compound analogy, a single analogy, or no analogy. The results show little difference between the types of information used in compound versus single analogy. Function, flow, and form information were all observed during analogy formation, indicating that all three types of information should play a role in a design-by-analogy database, regardless of generative goal. Notably, flow behavior was a commonly observed type of abstract similarity across domains. This points to the value of capturing flow behavior abstraction in engineering analogy databases.