Avatars vs. Video Presence: Effects of Instructor Presence on Cognitive Load in Video-Based Learning

This study examines the impact of instructor presence on cognitive load and learning outcomes in video-based learning environments, addressing a significant gap in optimising instructional design for digital education. Utilising eye-tracking technology to measure pupil dilation, a reliable proxy for cognitive load, this research compares three experimental conditions: no instructor presence, physical instructor presence, and avatar-based instructor presence. Thirty-three undergraduate participants engaged with educational videos, and their cognitive load was assessed through pupil dilation while learning gains were evaluated using pre- and post-tests. Findings indicate that physical instructor presence induces the highest extraneous cognitive load due to non-verbal distractions, negatively affecting learning outcomes. Conversely, avatar-based instructor presence effectively balances cognitive demands by reducing extraneous load and fostering germane cognitive processing, enhancing learning outcomes. The absence of an instructor minimises distractions and moderates cognitive load but results in only moderate learning gains, highlighting the importance of instructor presence in video-based learning, particularly for complex materials requiring contextual support and guided instruction. This research underscores the potential of avatars as a scalable and efficient instructional tool, especially in remote and asynchronous learning contexts. By simplifying visual cues and employing purposeful gestures, avatars mitigate extraneous distractions while maintaining instructional presence. These findings suggest that avatars can bridge the gap between the absence of an instructor and the potential overload associated with physical instructors. The study also demonstrates the value of integrating physiological measures like eye tracking into educational research to refine instructional designs further. This approach offers real-time insights into cognitive processing and learner responses, reducing biases inherent in self-reported measures. This work contributes actionable insights into designing scalable, effective educational technologies that optimise cognitive load and improve learning outcomes, paving the way for innovative approaches in modern e-learning.