Computer-aided mind map generation via crowdsourcing and machine learning

Abstract

Early-stage ideation is a critical step in the design process. Mind maps are a popular tool for generating design concepts and in general for hierarchically organizing design insights. We explore an application for high-level concept synthesis in early stage design, which is typically difficult due to the broad space of options in early stages (e.g., as compared to parametric automation tools which are typically applicable in concept refinement stages or detail design). However, developing a useful mind map often demands a considerable time investment from a diverse design team. To facilitate the process of creating mind maps, we present an approach to crowdsourcing both concepts and binning of said concepts, using a mix of human evaluators and machine learning. The resulting computer-aided mind map has a significantly higher average concept novelty, and no significant difference in average feasibility (quantity can be set independently) as manually generated mind maps, includes distinct concepts, and reduces cost in terms of the designers’ time. This approach has the potential to make early-stage ideation faster, scalable and parallelizable, while creating alternative approaches to searching for a breadth and diversity of ideas. Emerging research explores the use of machine learning and other advanced computational techniques to amplify the mind mapping process. This work demonstrates the use of the both the EM-SVD, and HDBSCAN algorithms in an inferential clustering approach to reduce the number of one-to-one comparisons required in forming clusters of concepts. Crowdsourced human effort assists the process for both concept generation and clustering in the mind map. This process provides a viable approach to augment ideation methods, reduces the workload on a design team, and thus provides an efficient and useful machine learning based clustering approach.