The role of AI avatars and human instructors in video-based learning: an eye-tracking study on cognitive load and social presence

Abstract

Grounded in the Cognitive Theory of Multimedia Learning (CTML) and Cognitive Load Theory (CLT), this study examines how different teacher representations (human instructors, AI avatars, and photo-based synchronized videos) affect students' learning outcomes and psychological factors, including cognitive load, social presence, source credibility, and mind wandering. Eye-tracking analysis was used to capture learners' visual attention patterns, focusing on fixation duration, fixation count, and transitions between areas of interest. The research employed a quasi-experimental post-test control group design with 51 middle school students in southeastern Türkiye. Although teacher-representation condition did not produce significant mean differences across the variables, gaze-based indicators revealed that longer fixation duration on the instructor area was associated with higher academic achievement and germane cognitive load, as well as stronger perceptions of social presence. In contrast, less attention to the instructor corresponded with increased mind wandering. These findings suggest that teacher visibility and sustained visual attention, rather than the representation type itself, are critical correlates of learning effectiveness in video-based instruction.