Lightening the load: Effects of AI agent generated humor and question-asking type on cognitive load and learning in online instruction

Abstract

This study investigates the effects of humor and question-asking strategies used by generative AI agent on teaching in online learning environments. A 2 (Humor: humor vs. no-humor) × 2 (Question-asking Type: open-ended vs. close-ended) between-subjects factorial design was employed with 116 Chinese university students. Participants interacted with AI agent that delivered instructional content and question-asking consistent with the experimental conditions. Measured outcomes included perceived humor, perceived agent value, intrinsic motivation, positive emotion, cognitive load, and knowledge retention. Results indicated that humor significantly enhanced learners’ intrinsic motivation, positive emotions, perceived agent value, and knowledge retention while also reducing intrinsic cognitive load (ICL). The findings underscore the effectiveness of incorporating humor in AI agent to positively influence both affective and cognitive dimensions of learning in online contexts.