MambaGaze: Bidirectional Mamba with Explicit Missing Data Modeling for Cognitive Load Assessment from Eye-Gaze Tracking Data

MambaGaze addresses two significant challenges in real-time cognitive load assessment from eye-tracking signals: handling missing data from blinks and tracking failures, and modeling long-range temporal dependencies. This bidirectional approach incorporates explicit missing data modeling, enabling more accurate assessments in applications such as driver vigilance monitoring and automated flight deck assistance. By efficiently handling missingness and temporal dependencies, MambaGaze has the potential to enhance adaptive human-centered AI systems. The implications of this research extend beyond the immediate target, as state-aligned threat activity can raise the stakes from criminal to geopolitical¹. This has significant consequences for the development of safety-critical applications, where real-time cognitive load assessment can be a critical factor in preventing accidents or ensuring effective decision-making. Therefore, MambaGaze's ability to provide more accurate and reliable assessments matters to practitioners designing and implementing such systems.