2025 ASEE Annual Conference & Exposition

This paper describes how generative AI (i.e., ChatGPT) was used to rapidly develop fictitious, yet realistic, qualitative interview transcripts for industrial engineering undergraduates to use as part of a human-centered design (HCD) problem. The curriculum for undergraduate engineering students is heavily focused on developing quantitative skills. However, engineering professionals may want or need to expand their skill set to also include qualitative methods. To that end, this research project will introduce and study qualitative methods training included in an existing industrial engineering course. Students in this mixed methods group, along with a comparison group of students who received standard quantitative-only methods training, are then asked to work through an HCD problem that includes both quantitative and qualitative data.

For the given design problem, students will be provided with 10 qualitative interview summaries in addition to standard quantitative anthropometric data tables to support their work on a design problem focused on workstation design. Collecting real qualitative interviews to support the design problem for this project was time and resource prohibitive. Instead, we used generative AI to rapidly simulate the fictitious interviews, adjusting the prompts as needed to construct realistic looking interview transcripts. After editing the transcripts to introduce more variability and distinction across the 10 interview transcripts, intentional "design seeds" were planted within the interview texts for students to potentially discover during their qualitative analysis. Our goal was to have recurrent themes (e.g., comments about the desire for dual monitors to enhance productivity), appearing across multiple interview transcripts in a variety of conversational ways. Students could then potentially discover these during their analysis of the interviews and include them in their workstation designs.

The human-AI teaming aspect of this work is especially notable as the research team and ChatGPT uniquely contributed to the creation of the interview transcripts that resulted in a final product that could not have been achieved alone in the time frame needed for the project. ChatGPT was able to produce rich, detailed interview transcripts to support the design problem by adjusting the prompts in a way that generated the detail needed to appear authentic. An example prompt was, "Write 500 word interview notes with a person sitting at workstation in a private office who seems frustrated with ability to concentrate, with that person answering questions about what is good and bad about their workstation, including the chair, desk, computer, and other workstation components." We could then produce additional interview transcripts by changing the wording to "…a person…who seems intensely focused…", or by using the same prompt but asking ChatGPT to generate different answers. Final editing of the transcripts by the research team was needed to introduce more variability across the interviews in terms of wording, as well as plant specific, recurrent "design seeds" across multiple interview transcripts for students to potentially discover. This project may inform industrial engineering and other faculty who wish to supplement their course design work for students with supporting materials using generative AI.