September 2, 2025
New paper led by I3T Lab's PhD students appearing at IEEE ISMAR 2025: Will You Be Aware? Eye Tracking–Based Modeling of Situational Awareness in Augmented Reality.
In this paper, we developed the first deep learning approach to predict a user’s situational awareness in augmented reality settings using eye tracking data from AR devices.
In a study involving 30 volunteers performing CPR with AR assistance during simulated emergencies, we demonstrated that loss of situational awareness is linked to increased visual fixation on virtual content. To predict situational awareness from gaze data, we introduced a graph neural network, FixGraphPool, which structures gaze events into spatiotemporal graphs that effectively capture dynamic attention patterns. The model predicts situational awareness with 83% accuracy, outperforming both classical machine learning and state-of-the-art time-series models. These findings pave the way for real-time, automated monitoring and preservation of situational awareness in AR-assisted tasks.