Triggering Cockpit Alerts Using an Eye-Tracking-Based Measure of Monitoring Performance

Abstract

This study explored the potential for enhancing pilot performance via an alerting system that adapts according to an eye-tracking-based measure of monitoring. The novel measure combines gaze analysis with system state assessment to estimate the pilot’s understanding of the current system state. On this basis, an alerting system was developed to direct pilot attention to unnoticed system state changes, thereby improving system state monitoring. In a flight simulator study involving 10 participants in a generic jet cockpit, we compared the adaptive alerting system with a no-assistance condition. Although alerting improved the participants’ performance in two tracking tasks, it adversely impacted performance in a third task. Nonetheless, alerting resulted in a decrease in both variance and mean detection time of critical changes. Participants’ subjective ratings were generally positive, yet they criticized the lack of transparency of the alerting mechanism. Alerts triggered based on eye-tracking and system state show potential for improving operator task performance. Nonetheless, for the system to reach its full performance potential, it is critical that the operator understand the principles underlying the alert triggers. False positives and alert design were identified as key areas for improvement to maintain user trust and task flow.