Air traffic controllers communication analysis as a proxy of task demand and mental workload: Using voice recording markers for safety critical task

• The present work was a feasibility study prompted by CRIDA A.I.E., the Spanish Reference Center for Research, Development and Innovation in ATM. • The objective is to develop a simple model able to use behavioural markers to detect high task demand conditions as a proxy indicator of MW. • Such a method would offer the advantage of making unobtrusive and immediate predictions. • Our results indicate that we can infer ATCOs’ mental workload by observing shifts in their behaviours and strategies in response to task demands. • For long-term application, a classification model would help determine when and for how long an ATCo’s task load or MW exceeds a critical safety threshold. The anticipated increase in air traffic over the coming decades will lead to a more congested airspace, thereby heightening the mental workload of air traffic controllers, who are responsible for maintaining the safety, reliability, and efficiency of air travel. Sustained overload is recognized in the literature as a significant factor influencing performance. Developing computational models for real-time monitoring of air traffic controllers’ mental workload is critical for devising effective task support strategies to prevent the detrimental effects of declines in human performance. This research stems from consideration related to simple mental workload model for air traffic controllers based on behavioural markers that can be unobtrusively collected in real time from their voice recordings. These markers may reflect the task load and complexity experienced by a human operator, which in turn can be used in this study as a proxy indicator of mental workload. In this study over 80 h of voice recordings from real air traffic controllers were analysed to examine changes in behaviours and strategies that they might employ in response to variations in task demands while performing their tasks. Results indicate that traffic density levels can be predicted through behavioural measures derived from their voice communications. This suggests that variations in task demands lead to changes in their behaviours and strategies to effectively manage the situation and, thereby, reflect their mental workload. This early-stage model will be cross-validated and refined in a subsequent research stage using well-established psychophysiological and subjective measurements.