Investigating the role of an AI-based assistance system for the decision-making process of pilots– an interview study

The introduction of cockpit assistance systems based on Artificial Intelligence (AI) hold the potential to significantly increase flight safety and efficiency, especially in complex and time-critical situations. However, there is limited understanding how an AI-based system could effectively assist pilots in the decision-making process. This interview study investigates the decision-making process of pilots in a complex decision-making scenario with and without the support of a Level 1B AI-based assistance system (EASA Roadmap 2.0, 2023). In a complex in-flight diversion scenario, the AI-based assistance system (named Virtual Crew Assistant) provides an assessment of the risks and benefits of suitable alternates in a user-friendly interface, enabling a direct comparison of the given alternates. Semi-standardized expert interviews were conducted with 12 commercial aviation pilots. The pilots were asked to perform a decision-making process for a complex scenario - using the conventional FOR-DEC decision-making model - with and without the support of a static prototype of the Virtual Crew Assistant (VCA). During the decision-making process, the experts were instructed to think aloud, and were also requested to rate the workload for each condition. Working with the AI-based system workload was rated significantly lower than working without the AI-based system. In addition, conditions for Trust, Explainable AI, Usability and Limitations of AI-based assistance systems in the cockpit were investigated. Other possible functions of the AI-based system suggested by the experts are given and extending the human-machine interaction to higher collaborative AI-Levels is discussed.