2025 Theses Doctoral

Exploring the Effects of AI-Generated Pedagogical Agents in Instructional Videos on Learning

Lim, Suh Young (Jullia)

This dissertation examined the effects of AI-generated versus human pedagogical agents on learning outcomes, cognitive load, and visual attention in multimedia learning. Using a 2 x 2 mixed factorial design, 58 adult participants viewed two instructional videos that varied by agent type (AI or human) and video order. Learning outcomes were measured through retention and transfer tasks, cognitive load was assessed using a subjective scale, and visual attention was recorded through eye-tracking data focused on the agents’ face, eyes, and hands.

Findings indicated no significant differences in retention or transfer between AI and human agents, suggesting that well-designed AI agents can be as instructionally effective as humans. However, video order significantly influenced retention, with declarative-first sequences leading to higher scores. While cognitive load scores did not differ significantly, descriptive trends suggested that procedural content presented first may increase perceived effort.

Eye-tracking results showed that human agents consistently drew more attention to facial and eye regions, while AI agents attracted greater attention to gesture-related areas. These results highlight the importance of content sequencing and the alignment of visual and auditory cues. The study provides practical and theoretical insights for designing effective AI-generated instructional agents that support engagement and learning in multimedia environments.