Abstract
The focus group methods remain a cornerstone of co-design practice but continue to face persistent challenges, such as participant recruitment, groupthink, and the efficiency of extracting design insights. This paper explores how Large Language Models (LLMs) could offer an alternative approach while preserving the benefits of common focus group methods. Therefore, we propose FocusGPT, a prototype system where customizable LLMs-based multi-agents simulate stakeholders to facilitate discussions and generate design insights. Through a formative study with 16 design professionals, we identified key requirements for developing and refining such a system. Our informal pilot trials suggest that FocusGPT could potentially serve as an additional option to help designers gather different perspectives. Despite promising applications, significant limitations exist in the accuracy of multi-agents representation, designer over reliance on LLMs outputs, and the black-box nature of LLMs. This research contributes to the understanding how LLMs might responsibly augment, rather than replace, focus groups in co-design practice.
Keywords
AI-enhanced Design Tool; Co-design Practice; Large Language Models
DOI
https://doi.org/10.21606/iasdr.2025.327
Citation
Xu, S., Lai, Y., Dong, L., Duh, H.,and Zhou, Z.(2025) FocusGPT: The Design of LLMs-based Focus Group for Empowering Co-design Practice, in Chang, C.-Y., and Hsu, Y. (eds.), IASDR 2025: Design Next, 02-05 December, Taiwan. https://doi.org/10.21606/iasdr.2025.327
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Conference Track
Track 4 - Human-Centered AI
FocusGPT: The Design of LLMs-based Focus Group for Empowering Co-design Practice
The focus group methods remain a cornerstone of co-design practice but continue to face persistent challenges, such as participant recruitment, groupthink, and the efficiency of extracting design insights. This paper explores how Large Language Models (LLMs) could offer an alternative approach while preserving the benefits of common focus group methods. Therefore, we propose FocusGPT, a prototype system where customizable LLMs-based multi-agents simulate stakeholders to facilitate discussions and generate design insights. Through a formative study with 16 design professionals, we identified key requirements for developing and refining such a system. Our informal pilot trials suggest that FocusGPT could potentially serve as an additional option to help designers gather different perspectives. Despite promising applications, significant limitations exist in the accuracy of multi-agents representation, designer over reliance on LLMs outputs, and the black-box nature of LLMs. This research contributes to the understanding how LLMs might responsibly augment, rather than replace, focus groups in co-design practice.