Artificial Intelligence for the Analysis of Biometric Data from Wearables in Education: A Systematic Review.

Wearable devices provide reliable biometric measurements in different contexts, and AI algorithms are increasingly being used to analyze this data. The objective of this review is to examine the use of wearable devices to collect biometric data combined with AI algorithms in an educational setting. A systematic review was conducted through the PRISMA methodology, by searching the Scopus database for works that included wearables, biometrics, and AI algorithms. A total of 43 studies were included and examined. The objectives, the type of collected data, and the methodologies of the included studies were investigated. Most articles utilized machine learning and deep learning algorithms for classification tasks, such as detecting stress or attention. Other applications included human activity recognition (HAR) for classroom orchestration and emotional or cognitive state detection. Many of the studies applied knowledge from previous works to the educational context, resembling exploratory research. Conversely, some authors developed tasks and methodologies tailored to the educational context. The strengths and weaknesses of the presented studies were discussed to propose future research directions. The main findings of this review highlight the advantages of the combination of multimodal sensing and predictive modeling in education with the eventual prospect of personalization. The absence of standardized acquisition and reporting remains the main barrier to replication, benchmarking, and synthesis across studies.