Javascript must be enabled to continue!
Applying Physics-Informed Neural Networks and Eye-Tracking Data to Gain Cognitive Insights in Human-Centred Applications

Eye movements are a rich, low-cost, non-invasive signal of human cognition, but modelling them accurately and interpretably remains difficult: conventional deep networks are data-hungry and opaque. This work applies Physics-Informed Neural Networks (PINNs), which embed the physical equations of eye motion directly into the learning objective, to eye-tracking data, pursuing two goals: predicting gaze more faithfully and using the resulting models to derive cognitive insights for human-centred applications. Across on-screen and wearable (eye-tracking glasses) settings, a pure PINN outperforms feed-forward, graph, and temporal-convolutional networks and their PINN-hybrids, reaching a mean absolute error of 0.735° and R² = 0.996 with statistically significant margins, while remaining interpretable and data-efficient. Building on this foundation, the same ocular pipeline (gaze, fixations, saccades, blinks, and pupil dynamics) is being extended to index cognitive states such as attention and fatigue, with early work on reading comprehension. The central argument is that physics-informed modelling offers a responsible, interpretable, and sustainable route to eye-based cognitive sensing, with applications spanning education, accessibility, healthcare, and human–computer interaction. The poster presents the method, published results, the emerging cognitive-insight direction, and open questions for discussion.

Attachment language: English File type: PDF document Presentation Poster
Updated: 20-10-2026 11:23 - Size: 1.39 MB
Attachment language: English File type: PowerPoint presentation Presentation File
Updated: 20-10-2026 11:23 - Size: 4.1 MB
Speakers
Konstantinos RoumpasMichalis Xenos

Back

ACM GreekCHI-CYCHI Symposium 2026


We are excited to invite you to the ACM GreekCHI-CYCHI Symposium 2026, held on November 6-7, 2026 jointly in Athens, Thessaloniki, Nicosia and Online.