Attention Drift Detection in Online Learning Using Eye-Tracking Signals

Online learning environments require sustained attention by the learner to foster the acquisition of knowledge. Distractions and cognitive fatigue commonly cause attention drift to go unnoticed. To overcome this obstacle, a real-time detection system is introduced, which is based on eye-tracking cues that are detected using webcams. The system transforms incoming gaze data into a Cognitive Gaze Tensor and calculates an Attentional Energy Function that is the product of fixation stability and pupil dilation. The model creates these features with the help of a Bidirectional Temporal Graph network that allows for identifying subtle disengagement patterns. Largescale eye-tracking datasets show high performance with 93.1 % test-accuracy, 92.8% F1-score, 93.70% recall, 0.96 AUC, and a mean processing latency of 40 ms. The importance of the energy model and adaptive feedback loop is proved by ablation studies. The proposed framework can offer a cost-effective, scalable approach to the integration of attention monitoring into elearning environments and deliver timely interventions to maintain engagement with learners and enhance education results.