Predicting Attention Allocation in Dual-Task Scenarios Using Multidimensional Eye-Tracking Metrics

Attention allocation is essential for multitasking, as it reflects the distribution of limited cognitive resources across tasks. Traditional eye-tracking metrics, such as fixation and glance ratios, provide limited insight into complex attentional processes. This study employed a comprehensive set of eye-tracking measures, including pupil diameter, eye movement amplitude, velocity, and duration, to assess attentional dynamics. We designed a dual-task scenario with varied task difficulty, elicited different levels of attention allocation. Analysis of eye position data revealed significant correlations between self-reported attention and multiple eye metrics. We proposed a CNN-based attention allocation classification network that outperformed conventional pattern recognition methods, demonstrating the approach’s effectiveness. These findings demonstrate that combining multidimensional eye-tracking metrics enables a more comprehensive assessment of attention allocation. This approach enhances understanding of cognitive resource distribution in dual-task scenarios and provides a promising method for monitoring attentional states in complex environments.