Leveraging Network Analytics to Examine the Impact of Generative Artificial Intelligence-Assisted Feedback on Inquiry-Based Discussion

Abstract

In inquiry-based discussion (IBD), it is essential to provide participants with effective feedback to promote engagement in knowledge construction and enhance the development of higher-order thinking. However, university instructors often struggle with monitoring multiple groups and delivering prompt and personalized feedback. Generative artificial intelligence (GAI), which can analyze text data and generate humanlike responses, offers potential solutions to mitigate these challenges. This study investigates the influence of GAI-assisted feedback on the IBD processes of pre-service teachers. A quasi-experiment was conducted with two classes (experimental: n = 53; control: n = 55) at a Chinese university. Epistemic network analysis was employed to model group IBD processes and compare groups with different characteristics (e.g., with/without GAI-assisted feedback, high/low engagement, high/low performance). Results show that GAI-assisted feedback significantly altered IBD dynamics. Collaboration self-efficacy was crucial for distinguishing group interaction patterns with the GAI chatbot. Moreover, groups in the experimental condition with high or low learning performance, engagement, and cognitive load showed diverse IBD interaction patterns. For example, groups with higher performance relied heavily on the GAI chatbot for idea generation without significant improvements in higher-order thinking. This study contributes detailed, process-oriented insights and implications on the adoption of GAI tools in IBD contexts.