Diagnosing Cognitive Control with Eye-Tracking Metrics in a Multitasking Environment

In multitasking environments, such as military flight missions, effective task prioritization is crucial to ensure overall flight safety. Cognitive control plays a vital role in this process, balancing stability for goal pursuit and flexibility for reacting to unexpected events. This can be challenging, since cognitive stability is also associated with more difficult task switching and cognitive flexibility is linked to distractedness. This study explores the use of eye-tracking metrics to diagnose the cognitive control state of operators within an adaptive assistance system in a multitasking environment. Three studies, involving 144 participants, manipulated control modes using a task prioritization strategy in a low-fidelity flight simulator. Eye movement data was recorded at 1000 Hz. The study employed eight supervised machine learning algorithms for binary classification, namely random forest, k-nearest neighbors, support vector machines, naïve bayes, decision trees, logistic regression, linear discriminant analysis, and XGBoost. The average accuracy of 0.89 demonstrates that the recognition of the cognitive control mode via eye-tracking is feasible. The findings suggest the potential integration of eye-tracking metrics into real-time user state diagnosis for adaptive assistance systems, especially in safety-critical human-machine systems. However, further research is needed to validate the classifier in more realistic flight environments with an expert sample.