Classifying Cognitive Load in Digital Textbook Reading Using Electroencephalogram Signals

In the new era of machine learning and digitalization, we have observed more textbook offered in multi-modal format in various education scenarios. However, it remains unclear how learners adapt to this transition and how their cognitive load changes during this process. In this paper, we leverage electroencephalography (EEG) as an indicator and explore the potential of machine learning solutions for understanding textbook readers' cognitive load. We conduct a within-subject study, collecting participants' EEG data as they read multiple textbook articles. Then, we experiment with different inputs, algorithms, and evaluation approaches to classify three levels of cognitive load. The results show that our approach achieves an F1 score of 60% for three-class cognitive load classification using classical machine learning models trained via a text-independent validation approach, which is significantly higher than baseline classifiers. These results open up new opportunities for cognitive load detection using EEG signals.