Generative Design Solution Space Parsing: An Evaluation of User Experience, Workload, and Performance

Generative design systems produce myriad design alternatives that comply with stated requirements. Since generative design systems yield the greatest benefits during conceptual design, requirements are often ambiguous and are comprised of mixed variables (e.g., categorical, continuous, etc.), which leads to a generative design solution space with a plethora of options for the user to review. With a plethora of design alternatives to choose from, many of which are similar, increased user workload leads to inefficient design selection processes. Subsequently, inefficient design selection processes could result in a negative user experience and improper design alternative selection. Therefore, it is imperative that generative design systems leverage parsing methods that methodologically reduce the quantity of design options that are presented to the user, while retaining novel designs from distinct solution space regions. Although parsing a solution space can yield smaller subsets of design alternatives, it is also imperative to consider how the subsets are presented to the user. A user study (N=49) was performed to evaluate user performance, workload, and experience during a generative design selection process, given manipulation of both the quantity and filtering of parsed subsets of alternatives. Subsets were filtered using cluster analysis using one of seven parameters, where participants experienced two filters across seven iterations each. Results show that cognitive workload is reduced when a design solution space consists of 50 to 100 design alternatives, with a clustering parsing method that considers all design alternative variables. Study findings can further be applied to other domains where a user is presented with a plethora of alternative options, requiring a method for improving the decision-making process.