Plataforma EEG para la monitorización grupal de la atención en entornos de enseñanza

Abstract

There are several aspects in education that concern the educational community, such as objectively measuring the level of attention students pay during the course of a class, early detection of students with learning difficulties (LD) or low academic performance. Historically, the assessment of student attention has been carried out through mere observation by the teacher. However, this method, although accurate most of the time, is subjective and is very difficult to apply in online learning environments. Electroencephalography (EEG) is a technique that has been widely used to study cognitive and mental states of people, such as attention, stress, fatigue, or drowsiness, and its use could play a significant role in this context. The main objective of this interdisciplinary thesis is to develop an innovative teaching platform based on EEG brain activity recording capable of immediately evaluating students' attention in class, whether in-person or online, objectively. Additionally, this EEG system must be wearable, noninvasive in the educational context, easy to deploy for teachers and students, and costeffective. To validate the platform, a study was conducted in a real classroom with high school students. The study's first objective was to demonstrate that such a platform is feasible, reliable, and robust, and that it can be easily deployed in a realistic setting. For this purpose, EEG recordings of multiple students were made while they performed an evaluative task during a class. The results showed a simple correlation between beta brain waves (associated with attention) and academic performance. The second objective was to study the feasibility of using the platform for the proper detection of students with some form of LD. Using Machine Learning (ML) or Artificial Intelligence (AI) models, we found evidence suggesting that our platform can be effective and convenient for early detection of LD symptoms in high school students. Lastly, ML techniques such as Support Vector Machines (SVM), Random Forest (RF), Neural Networks (NN), or Logistic Regression (LR) were applied to build a prediction system for students' academic performance based on the information generated by our platform, achieving 100% accuracy in predicting students who pass or fail a task. These results demonstrate the utility of the platform as an educational tool, providing valuable and objective real-time feedback to teachers, enhancing the effectiveness of teaching-learning processes, both in face-to-face and online environments.