Yubei Chang, Qingtang Liu, Yingxue Lu, Enhui Miao
Abstract Peer feedback is crucial for promoting deeper thinking and interaction. However, various challenges exist in conducting high-quality peer feedback in collaborative argumentation learning. Generative AI (GenAI) has shown potential for supporting peer feedback, but how GenAI can be efficiently applied to support peer feedback needs to be further explored. In addition, the effect of GenAI-supported peer feedback on argumentation performance and feedback quality remains unclear. To address these gaps, this study compared three peer feedback conditions: peer feedback (PF), peer feedback with GenAI (GenAI-PF), and peer feedback with GenAI supported by prompt scaffolding (PS-GenAI-PF). Forty-five student teachers were assigned to 12 groups of 3–4 students. All groups were randomly assigned to three peer feedback conditions, each including four groups. Kruskal-Wallis H test results showed that the GenAI-supported peer feedback groups performed better in argumentation performance. In particular, the PS-GenAI PF group performed better on “Rebuttal data and warrant” and “Addressing the opposing view”. For feedback quality, the GenAI-supported peer feedback group showed more explanation, suggestions, as well as neutral and negative feedback. The epistemic network analysis revealed that the PS-GenAI PF group had strong associations with negative emotions and higher-order feedback content, such as identification-suggestion and explanation-suggestion. Implications for the instructional design and implementation of GenAI-supported peer feedback in higher education are discussed.