#445: Knowledge-Building Discourse in Generative AI-Mediated Collaborative Problem Solving: A Student–Expert Comparison

This study examines how students and experts differ in their engagement with generative AI (GenAI) during knowledge-creating dialogue. Twenty undergraduate students (10 pairs) and one pair of expert educators collaborated with ChatGPT-4o to address a socio-scientific issue. Using the Knowledge Building Discourse Explorer (KBDeX) and “Good Moves” coding, we analyzed temporal degree centrality and discourse networks to capture epistemic and dialogic processes. Results showed that students’ discourse was largely anchored to AI-generated ideas, with limited transformation through dialogue, whereas experts engaged in self-sustained cycles of problem definition, evaluation, and reframing, using GenAI selectively to extend inquiry. These findings suggest that differences lie not in the availability of ideas but in how discourse is organized to support knowledge creation, highlighting the importance of scaffolding epistemic and dialogic practices for effective GenAI use.