Effects of Virtual Pedagogical Agents and Instructional Content Type on Learners' Experience in Online Learning: Evidence from Behavioral, Eye-Tracking, and EEG Data

In plain English

Forty students learned from either a human-like or a non-human pedagogical agent, teaching either traditional knowledge or popular-science content, while the researchers recorded questionnaires, eye tracking and EEG. The match mattered more than the realism: human-like agents worked better for traditional knowledge, non-human agents worked better for popular science, and recall improved for popular science only with the non-human agent. The practical read is that "more photorealistic" is not a default improvement — the right agent appearance depends on the content it is delivering. The sample is small, single-institution and mostly female, and voice differences were confounded with appearance.