Machine Learning-Based Cognitive Load Prediction from Multimodal Neurophysiological Data
Dr. V. Maniraj1*, N. Mahalaksmi2
Abstract
A better understanding of students’ cognitive load is essential to optimize instruction. However, it could be challenging in the classroom, as it is difficult for teachers to observe each student continuously when dozens of students co-exist. The rise of multimodal learning analytics now allows for predictions of individuals’ cognitive load using neurophysiological features. Therefore, this research aimed to equip teachers with insights into students’ cognitive load in high school math classes by leveraging neurophysiological data. Results highlighted that the teacher’s cognitive load assessments for the overall class do not generalize well to individual students, especially on mental effort. In contrast, multimodal neurophysiological models demonstrated a notable correlation, accounting for 15.3% of perceived task difficulty and 25.9% of mental effort variance (R2), providing additional information for teacher assessments. Our study demonstrates that multimodal neural correlates may help teachers gain a deeper understanding of students’ cognitive states, enhancing instructional design accordingly.
Keywords:
Cognitive load; multimodal learning analytics; neurophysiological data; mental effort assessment; instructional optimization
