When AI Joins the Team: How Personas Shape Learning and Collaboration
As generative AI systems increasingly participate in human collaboration, they are no longer mere tools but social actors with distinct personas that shape how people think, argue, and learn together. This seminar explores how AI personas, supportive or contrarian, affect both epistemic and socio-emotional processes in collaborative learning. Building on experimental studies of human-AI teamwork, I discuss evidence of a social blindspot: humans often fail to detect AI teammates, yet these undetected agents subtly restructure group discourse and reasoning patterns. Drawing on the framework of Argumentative Knowledge Construction (AKC), I show how AI personas influence the balance between critical elaboration, consensus building, and cognitive conflict in group reasoning tasks. The findings reveal that supportive personas enhance psychological safety and participation, while contrarian personas trigger productive friction that deepens conceptual exploration, but only under certain conditions of trust and transparency. By integrating insights from learning analytics and interaction analysis, this talk highlights new directions for designing generative AI collaborators that foster both intellectual engagement and social cohesion in educational and professional contexts.
Lixiang Yan is an Assistant Professor at the Institute of Artificial Intelligence in Education, School of Education, Tsinghua University. His research lies at the intersection of artificial intelligence, educational technology, and learning analytics, focusing on how cutting-edge technologies, particularly multimodal data analytics and generative AI, can be leveraged to support and enhance human learning. He has led multiple international and national research projects, including the OpenAI Research Grant, the Monash University FIT Early Career Researcher Seed Fund, and a UNESCO project, and has contributed as a core member to an Australian Research Council (ARC) Discovery project. His recent work has been published in top-tier journals such as Nature Human Behaviour, Nature Reviews Psychology, Computers & Education, and the British Journal of Educational Technology (BJET), and has received several Best Paper Awards or nominations at leading international conferences, including LAK and AIED. He currently serves on the Early Career Editorial Board of the British Journal of Educational Technology and actively advances research frontiers in generative AI literacy, educational AI agents, and multimodal learning analytics, with a strong emphasis on their responsible and effective integration into educational practice.