Shamya Karumbaiah
Incoming Assistant Professor, Educational Psychology, UW-Madison

Abstract: K-12 classrooms are increasingly using AI-based educational software, which in practice, is jointly facilitated by both teachers and AI tutors. In recent years, AI tutors have been designed with the awareness of such human-AI hybrid nature of teaching in ecological settings. Recent research has also developed teacher-facing analytics dashboards to augment teachers’ real-time awareness of students’ learning and experience with the system. Yet, the strategies teachers use to jointly enact the adaptive learning experiences with the AI tutor and how these impact students’ learning and engagement remain understudied. In this talk, I first propose a multimodal conceptualization that recenters teachers’ roles and practices in such human-AI partnerships with an intention to augment human abilities. Then, I present a multimodal methodology designed to support research on: 1) quantifying the active roles teachers play in AI-enabled classrooms, 2) unveiling the processes involved in teachers’ mediation of students’ learning with AI tutors, and 3) generating results for teacher reflection on their teaching practices with AI tutors. To illustrate the proposed methodology, I present exploratory case studies involving data on teacher position, gaze, dialogue, and dashboard use to understand the relationship between teacher practices (e.g., attending, visiting, help-giving) in the physical classroom and student learning and engagement with an AI tutor.