Data-Driven Heuristic Induction From Human Design Behavior
Abstract
Through experience, designers develop guiding principles, or heuristics, to aid decision-making in familiar design domains. Generalized versions of common design heuristics have been identified across multiple domains and applied by novices to design problems. Previous work leveraged a sample of these common heuristics to assist in an agent-based design process, which typically lacks heuristics. These predefined heuristics were translated into sequences of specifically applied design changes that followed the theme of the heuristic. To overcome the upfront burden, need for human interpretation, and lack of generality of this manual process, this paper presents a methodology that induces frequent heuristic sequences from an existing timeseries design change dataset. Individual induced sequences are then algorithmically grouped based on similarity to form groups that each represent a shared general heuristic. The heuristic induction methodology is applied to data from two human design studies in different design domains. The first dataset, collected from a truss design task, finds a highly similar set of general heuristics used by human designers to that which was hand-selected for the previous computational agent study. The second dataset, collected from a cooling system design problem, demonstrates further applicability and generality of the heuristic induction process. Through this heuristic induction technique, designers working in a specified domain can learn from others’ prior problem-solving strategies and use these strategies in their own future design problems.
Document Details
- Document Type
- Pub Defense Publication
- Publication Date
- Oct 14, 2020
- Source ID
- 10.1115/1.4048425
Entities
People
- Christopher McComb
- Jonathan Cagan
- Lucas Puentes
Organizations
- Carnegie Mellon University
- Defense Advanced Research Projects Agency
- Pennsylvania State University