LEMate: An Early Prototype of an Artificial Intelligence-Powered Learner Engagement Detection System for Low-Resource Classrooms

In contemporary educational settings, learner engagement is a critical factor influencing the effectiveness of teaching and learning processes. Traditional methods of assessing engagement often rely on subjective observations or self-reporting, presenting challenges in obtaining accurate and timely insights. This work introduces an Artificial Intelligence-Powered Learner Engagement Detection System, called LEMate, designed to objectively assess and analyse learner engagement in real-time focussing on low-resource classrooms and learners from neuroatypical populations. The LEMate system leverages state-of-the-art machine learning algorithms to analyse diverse data sources, including facial expressions, eye gaze, heart rate and interaction patterns with computers, to determine the learner’s learning state and returns one of the three possible outcomes (engaged, bored, frustrated) using a traffic light system. The multi-modal approach ensures a comprehensive understanding of engagement dynamics. The model is deployed in educational contexts in Bangladesh and preliminary results from piloters highlight the system’s potential to accurately capture and interpret learner engagement, demonstrating its value as an innovative tool for educators seeking to create dynamic and effective learning environments where resources are scarce. The ongoing development and refinement of the system contribute to the growing field of educational technology, paving the way for more personalised and engaging learning experiences for learners with diverse needs in developing countries.