LLM-Assisted Learning Analytics
Validating LLM-based diagnosis of students emotion and engagement in science inquiry learning, and associating students learning trajectories with engagement survey and performance.
This project investigates how large language models can be applied to learning analytics in science education. One line of work focuses on validating LLM-based methods for diagnosing students' academic emotions and engagement states during science inquiry tasks, using facial expression analysis and multimodal data. A second line of work traces students' learning trajectories over time, examining how behavioral patterns associate with self-reported engagement and learning performance.
The goal is to develop scalable, automated approaches for understanding students' affective and cognitive states in real time, enabling more responsive and personalized learning environments without relying solely on traditional survey instruments.
Conference Proceeding:
- He, J., Jin, B., Xie, K., & Zhang, D. (2025). Diagnose Academic Emotions from Facial Expressions: Relationship with Science Learning Performance in Web-Based Self-directed Learning. ISLS 2025.