Exploring the impact of AIGC on learning performance: a moderated mediation analysis between system-paced and self-paced video learning
Bin Jing, Hailiang Yang, Zhongling Pi, Yu Zhou, Hongliang Ma · Interactive Learning Environments · 2025
Artificial intelligence-generated content (AIGC) has potential in video learning settings, but its effectiveness depends on how it is used. This study integrated a GPT-3.5-Turbo-based question-answering tool into a video learning system to examine how AIGC-supported questioning affects learning under system-paced and self-paced conditions. We conducted a 2 (pacing: system vs. self) × 2 (AIGC: without vs. with) experiment involving 160 undergraduate students, analyzing their cognitive load and learning performance. A moderated mediation model was built to explore how AIGC influenced learning performance, with cognitive load as the mediating variable and pacing as the moderating variable. Findings revealed that AIGC reduced extraneous cognitive load, increased germane cognitive load, and improved learning performance in the self-paced condition. However, AIGC did not yield significant positive effects in the system-paced condition. The model analysis indicated that the AIGC’s impact on learning performance was mediated by extraneous cognitive load and depended on the self-paced environment. These findings support using AIGC to enhance video learning, with recommendations for instructional designers and teachers to utilize AIGC to support learners in self-paced environments.