Eye-tracking metrics for estimating workload and characterizing errors in conflict detection and resolution during...

Growing traffic density and airspace complexity demand adaptive decision-support tools that anticipate when controllers are approaching overload or conflicts are mishandled. Ocular behavior offers a single, unobtrusive stream that simultaneously reflects global mental workload (MWL) and moment-to-moment attentional allocation. The present study examined whether eye-tracking metrics can estimate MWL and expose the mechanisms underlying errors in conflict detection and resolution during simulated en route control. Twenty-four novice participants worked six 16-min radar scenarios that varied traffic load and sector complexity. A remote eye-tracker recorded pupil diameter, blink dynamics, and fixations on static and aircraft-centered areas of interest, while subjective MWL was sampled with the Instantaneous Self-Assessment and NASA-TLX scales. Higher traffic density increased self-reported MWL, enlarged pupils, reduced blinks and blink durations, and concentrated fixations inside the active sector, whereas higher traffic complexity increased MWL, reduced blinks, and concentrated fixations inside the active sector. Blink rate and pupil size accounted for most of the variance in MWL (up to 94%). In addition, two scripted conflict events were examined in greater detail. In the simpler conflict, errors primarily stemmed from failures in detection. Successful resolutions were characterized by sustained gaze on both converging aircraft and a higher frequency of altitude-change clearances, while failures showed reduced fixation times and a lack of interventions. In contrast, errors in the more complex conflict resulted from planning breakdowns despite initial detection. Successful resolutions in this case typically involved at least two interventions, whereas failures were associated with prolonged fixation times but insufficient corrective action. Thus, global ocular indices provide precise estimates of MWL, and gaze-action couplings can help anticipate errors in conflict detection and resolution. Embedding both levels of inference in adaptive ATC support systems could enable real-time MWL management, and proactive mitigation of separation-loss events.