The Socratic Challenger: a structured GenAI-assisted workflow for undergraduate research inquiry

Abstract

Generative artificial intelligence (GenAI) tools are increasingly used during early stages of academic inquiry, yet their role in supporting research-question development remains unclear. This mixed-methods study evaluated an eight-step GenAI-assisted workflow implemented in an undergraduate ecology course. The workflow alternated AI-supported exploration with literature verification, revision, and human feedback. The study population comprised students enrolled in the course, and all 45 students who consented to research use of their coursework were included; therefore, no sample-size formula was applied. Quantitative data consisted of ordinal ratings of the perceived contribution of each workflow step, while qualitative data consisted of students’ written reflections. The rating items corresponded directly to the eight implemented workflow stages, and qualitative themes were independently reviewed and refined by two researchers. Quantitative results showed that perceived usefulness differed modestly across steps (Friedman χ 2  = 14.53, p  = 0.0426; Kendall's W = 0.046), although the median rating for every step was “Helped a lot,” and no pairwise comparison remained significant after correction. The ratings showed acceptable internal consistency (Cronbach's α  = 0.745). Qualitative analysis indicated that students viewed AI primarily as a thought partner that helped them narrow and refine ideas, while the broader scaffolded workflow supported progress through literature engagement, structured pacing, and feedback. Together, the findings suggest that GenAI can support undergraduate research inquiry when embedded within a structured process that requires students to verify evidence, evaluate suggestions, and retain responsibility for the final research question.