Repository logo

Optimizing designer cognition relative to generative design methods


Generative design is a powerful tool for design creation, particularly for complex engineering problems where a plethora potential design solutions exist. Generative design systems explore the entire solution envelope and present the designer with multiple design alternatives that satisfy specified requirements. Although generative design systems present design solutions to an engineering problem, these systems lack consideration for the human element of the design system. Human cognition, particularly cognitive workload, can be hindered when presented with unparsed generative design system output, thereby reducing the efficiency of the systems engineering life cycle. Therefore, the objective of this dissertation was to develop a structured approach to produce an optimized parsing of spatially different generative design solutions, derived from generative design systems, such that human cognitive performance during the design process is improved. Generative design usability foundation work was conducted to further elaborate on gaps found in the literature in the context of the human component of generative design systems. A generative design application was then created for the purpose of evaluating the research objective. A novel generative design solution space parsing method that leverages the Gower distance matrix and partitioning around medoids (PAM) clustering method was developed and implemented in the generative design application to structurally parse the generative design solution space for the study. The application and associated parsing method were then used by 49 study participants to evaluate performance, workload, and experience during a generative design selection process, given manipulation of both the quantity of designs in the generative design solution space and filtering of parsed subsets of design alternatives. Study data suggests that cognitive workload is lowest when 10 to 100 generative design alternatives are presented for evaluation in the subset of the overall design solution space. However, subjective data indicates a caution when limiting the subset of designs presented, since design selection confidence and satisfaction may be decreased the more limited the design alternative selection becomes. Given these subjective considerations, it is recommended that a generative design solution space consists of 50 to 100 design alternatives, with the proposed clustering parsing method that considers all design alternative variables.


Rights Access


systems engineering
generative design


Associated Publications