Year: 2026
Type: conference-paper
Abstract: Generative AI (GenAI) is increasingly used by university students as a conversational feedback partner in academic writing and revision. This raises questions about how different AI feedback chatbot designs influence students’ feedback interactions in authentic higher education contexts. This mixed-methods study compares the effect of two purpose-built AI feedback chatbots (instructive vs. reflective) on students’ feedback interactions, feedback perceptions, and reflection during an abstract writing exercise. Conducted in a master-level course (N = 23), the between-subjects study combined pre- and post-surveys, draft and revised abstracts, AI chat logs, and observation notes from mixed-group discussions. Quantitative results showed that students in the instructive condition had significantly higher feedback perceptions and greater self-perceived improvement in abstract quality, while chat log analysis revealed significantly higher levels of feedback interaction in the reflective condition. Qualitative analyses indicated that participants in the instructive condition often uncritically adopted AI suggestions, whereas students in the reflective condition showed greater deliberation, monitoring of writing decisions, and project-level reflection, but also increased frustration when the chatbot did not meet expectations of direct solutions. Observational field notes from mixed-group discussions revealed students’ critical reflection on their use of the two AI designs, where students articulated tensions between efficiency, autonomy, and responsibility for self-regulated writing. Overall, the findings show that chatbot design significantly shaped students’ feedback interactions, feedback perceptions, and self-assessed improvement. We found that explicitly comparing instructive vs. reflective AI feedback chatbots in the classroom can stimulate productive peer discussions about critical and self-regulated uses of GenAI in academic writing.
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7204741491