Advancing self-directed learning in STEM education: integrating GPT-based learning aid with multimodal learning analytics
Chia‐Ju Lin, Wei‐Sheng Wang, Hsin‐Yu Lee, Pin‐Hui Li, Yueh‐Min Huang, Ting‐Ting Wu
Abstract
Precision education employs technology to diagnose learning processes and provide adaptive feedback tailored to individual needs. This study explores the impact of generative AI as a learning aid to enhance self-directed learning (SDL) in STEM education. In a four-week randomized controlled trial with 72 university students, a GPT-based learning aid was compared to a traditional FAQ tool. The GPT-based aid, powered by large language models (LLMs), offered personalized feedback, while the FAQ tool provided static responses. Results showed the experimental group outperformed the control group in learning performance and SDL abilities. Multimodal learning analytics (MMLA) revealed richer self-assessment, reflective thinking, and interactive behaviors among GPT-aid users, highlighting the transformative potential of generative AI in advancing personalized learning and SDL in STEM.