T-11: From Peer to Policy: AI’s Role in Learning

This study examines the integration of a multi-agent Artificial Intelligence (AI) system in undergraduate Computer Science education to support caring pedagogical practices in large classes. The system incorporates five specialized AI agents supporting knowledge synthesis, task management, collaborative learning, instructor guidance, and personalized student support. Through mixed-methods research, we investigate how these agents facilitate individual learning, collaborative knowledge building, and instructor decision-making. This study contributes to the use of agentic AI in caring educational practices while maintaining personalized learning experiences in increasingly complex academic settings.

We examined the effects of AI-based feedback from iWrite and ChatGPT versus peer feedback on Chinese ESL college students’ writing skill, self-efficacy, and enjoyment. Over a 16-week semester, 246 non-English major students were assigned to peer, iWrite, or ChatGPT conditions. Pre- and post-intervention assessments revealed no significant differences among feedback methods. Students’ English proficiency did not moderate the outcomes, indicating AI-generated feedback is as effective as peer feedback. These findings suggest AI tools serve as reliable complements to traditional feedback methods, offering promising avenues for improving the feedback process in ESL writing instruction.

AI is the most talked-about innovation in education—but remains largely untapped. Policies and regulations create barriers, keeping AI at the fringes of academia, full of promise but absent in practice. While concerns over data sovereignty and intellectual property dominate discussions, practical applications remain elusive. Our presentation reframes AI as a missing team member in course development. How might it enhance instructional design and media production? What tools exist, and what obstacles remain? Our project aims to revisit, a year from now, how today’s theories on AI in education align with future real-world insights.