How AI Processing Delays Foster Creativity: Exploring Research Question Co-Creation with an LLM-based Agent

Yiren Liu,Si Chen,Haocong Cheng,Mengxia Yu,Xiao Ran,Andrew Mo,Yiliu Tang,Yun Huang,Haocong Chen
DOI: https://doi.org/10.48550/arXiv.2310.06155
IF: 6.4588
2023-10-09
Human-Computer Interaction
Abstract:Developing novel research questions (RQs) often requires extensive literature reviews, especially for interdisciplinary fields. Leveraging Large Language Models (LLMs), we built an LLM-based agent system, called CoQuest, supporting RQ development through human-AI co-creation. We conducted an experimental design with 20 participants to examine the effect of two interaction designs: breadth-first and depth-first RQ generation. The results showed that participants found the breadth-first approach more creative and trustworthy upon task completion. However, during the task, they rated the RQs generated through the depth-first approach as more creative. We also discovered that AI processing delays allowed users to contemplate multiple RQs simultaneously, resulting in more generated RQs and an increased sense of perceived control. Our work makes both theoretical and practical contributions by proposing and assessing a mental model for human-AI co-creation RQs.
What problem does this paper attempt to address?