Supplementary file 2_Eye-tracking metrics for estimating workload and characterizing errors in conflict detection and resolution during simulated en route air-traffic control.docx

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.

不断增长的交通密度与空域复杂度，亟需能够预判管制员即将面临过载或冲突处置失当情况的自适应决策支持工具。眼部行为可提供一条单一且无干扰的数据流，同时反映全局心理负荷（Mental Workload, MWL）与逐时刻的注意力分配情况。本研究旨在探究眼动追踪指标能否估算心理负荷，并揭示模拟航路管制场景下冲突探测与处置失误的内在机制。24名新手管制员参与了6组时长16分钟的雷达管制场景实验，场景的交通负载与扇区复杂度各不相同。研究采用远程眼动仪记录受试者的瞳孔直径、眨眼动态，以及对静态区域与以航空器为中心的感兴趣区域的注视情况；同时通过即时自我评估量表与NASA任务负荷指数量表（NASA-TLX）采集主观心理负荷数据。更高的交通负载会提升受试者自我报告的心理负荷，使瞳孔扩大、眨眼频率与眨眼时长降低，并将注视点集中于活动扇区内；而更高的空域复杂度同样会提升心理负荷、降低眨眼频率，并将注视点集中于活动扇区内。眨眼频率与瞳孔大小可解释心理负荷的绝大多数变异（最高可达94%）。此外，本研究还对两组预设的冲突事件进行了更细致的分析。在较简单的冲突场景中，失误主要源于探测失败。成功处置该冲突的表现为：持续注视两架交汇航空器，且高度调整指令的发布频率更高；而处置失败的表现则为注视时长缩短且未采取任何干预措施。与之相对，在更复杂的冲突场景中，即便初始探测成功，失误仍源于规划环节失效。该场景下的成功处置通常需要至少两次干预措施，而处置失败则表现为注视时长过长，但未采取足够的纠正措施。综上，全局眼部指标可精准估算心理负荷，而注视-动作耦合机制则有助于预判冲突探测与处置过程中的失误。将这两级推理机制嵌入自适应空中交通管制（Air Traffic Control, ATC）支持系统，可实现实时心理负荷管理，并主动规避航空器失隔事件。