On the use of ECG and EMG Signals for Question Difficulty Level Prediction in the Context of Intelligent Tutoring Systems

Abstract

A fundamental drawback of traditional Intelligent Tutoring Systems (ITS) is that, unlike human tutors, they are not able to understand the emotional state of their users and adapt the learning process accordingly. This work explores the potential use of affective computing techniques for providing an affect detection mechanism for ITS. Electrocardiography (ECG) and electromyography (EMG) signals were recorded from 45 individuals that undertook a computerised English language test and provided feedback on the difficulty of the test's questions. Features extracted from the ECG and EMG signals were then used in order to train machine learning models for the task of predicting the self-perceived difficulty level of the questions. The conducted supervised classification experiments provided promising results for the suitability of this approach for enhancing ITS with information relating to the affective state of the learners, reaching an average classification F1-score of 75.49% for the personalised single-participant models and a classification F1-score of 64.10% for the global models.