A Multiagent Approach to Evaluating Innovative Component Selection

Though there is a clear correlation between profitability and innovativeness, the steps that lead to designing a "creative" product are elusive. In order to learn from past design successes, the impact of design choices on the innovativeness of an entire product must be identified. This problem of quantifying the impact of design choices on a final product is analogous to the problem of 'credit assignment' in a multiagent system. We use recent advances in multiagent credit assignment to propagate a product's innovativeness back to its components. To validate our approach we analyze products from the Design Repository, which contains thousands of products that have been decomposed into functions and components. We demonstrate the benefits of our approach by assessing and propagating innovation evaluations of a set of products down to the component level. We then illustrate the usefulness of the gathered component-level innovation scores by illustrating a product redesign.