Exploring Expert and Peer Tutoring Dynamics in Online Learning Discussions: A Temporal Learning Analytics Approach

• In Event: Supporting Student Learning in Online Settings

Abstract

Online discussion forums offer rich data for understanding tutoring interactions in STEM education. This study uses learning analytics to examine temporal dynamics between expert and peer tutors in 100,000 math posts from Math Nation, a K–12 platform. We applied transformer-based models to identify tutoring behaviors and used multilevel vector autoregression (VAR) to analyze sequential patterns. Findings show peer tutors posted nearly twice as often, offering more acknowledgments and feedback, while expert tutors were more directive. In mixed threads, expert tutors shaped peer responses by discouraging directive help. This study advances scalable behavioral modeling and sheds light on how tutoring roles unfold over time in online learning.

Authors

• Wanli Xing, University of Florida

• Zifeng Liu, University of Florida