EDUCATIONAL WELL-BEING IN THE DIGITAL ERA: CHALLENGES, OPPORTUNITIES AND IMPLICATIONS FROM BIOMETRIC ENGAGEMENT ANALYSIS

Abstrak

The increasing application of digitalization has created unprecedented opportunities to enhance access while simultaneously introducing complex challenges to student well-being. This study presents a comprehensive empirical analysis of the Student Engagement Biometrics Dataset (SEBD-2024), comprising multimodal learning sessions recorded from 300 students across three content types (Text, Video, Interactive) and three difficulty levels (Easy, Medium, Hard). Using electroencephalography (EEG) frequency-band power spectral density measurements across five neural bands (Delta, Theta, Alpha, Beta, Gamma) alongside eye-tracking metrics (pupil dilation, blink rate, fixation duration, saccade velocity), we extract, analyse, and contextualize biometric correlates of student engagement. Our Random Forest classifier achieved 35.3% cross-validation accuracy across three engagement classes, only marginally above random baseline (33.3%), revealing a critical finding: raw biometric signals from these modalities, analysed in isolation, exhibit insufficient discriminative power to reliably stratify engagement states. This null-hypothesis insight is itself a major contribution, highlighting the multi-modal complexity of educational well-being and the need for longitudinal, contextualized, and subjective-report-integrated frameworks. We draw direct connections between these findings and the broader discourse on digital well-being, cognitive fatigue, and the neurological substrate of engaged learning, proposing a five-pillar Well-Being Estimation Framework (WEF) for digital education systems.