Use of Generative AI for Boundary Crossing in Interdisciplinary Collaborative Research

Abstract

Ph.D. students in interdisciplinary education programs face challenges in achieving effective collaboration. To support them in overcoming their challenges, we designed a workshop using generative artificial intelligence (AI) as a broker device to create partially shared objects for boundary crossing. In the workshop, groups of students collaborated by coordinating research themes proposed by ChatGPT with the input of their research information. After their collaboration, experts evaluated improvements in the proposed themes. We then analyzed four high-outcome and three low-outcome groups to identify key discourse moves for successful knowledge-building discourse. Epistemic Network Analysis (ENA) and discourse analysis revealed that compared to low-outcome groups, high-outcome groups actively enhanced explanations, connected/synthesized their discourse, and shared relevant disciplinary domain and research-design knowledge to improve their themes without persisting with those that they initially judged as non-promising. We utilized these results to propose a conjecture map toward future systematic design-based research on interdisciplinary education.