User-Centered Design of an Intelligent Pilot Advisory System for Non-Emergency Scenarios: A Case Study

Kurzfassung

Artificial Intelligence (AI) is being used in an increasing number of fields, so including it aviation as well only comes natural. Therefore, the Intelligent Pilot Advisory System (IPAS) is being developed as a research platform. and is supposed to assist pilots in their decision making. Previously, the research was focused on the IPAS’s application in emergency scenarios. In this thesis, the IPAS is extended for use in non-emergency scenarios to support pilots in normal operation. Hereby, this research explores what functions the system could offer to pilots that are useful to them, pilots’ willingness to use such a system in non-emergency situations, and how to design this system for optimal usability, focusing on explainability and trustworthiness. To design a suitable system, a workshop was held with experts to brainstorm possible functionalities. Hereby, the idea for a Mission Monitoring and Advisory Function (MMAF) was developed. The MMAF encompasses the IPAS to continuously monitor the conditions along the flight route, as well as at the destination airport and to provide updates, strategic insights, and recommendations to the pilots. Additionally, the IPAS is supposed to identify suitable alternate airports in the area. To develop a User-Interface (UI) for such a system, current guideline from research and industry were considered and an iterative user-centered design process was utilized. Lastly, the system was tested in a mixed-methods, between-subjects study with 18 pilots in an Airbus 320 cockpit simulator to evaluate its usability and the impact of explanatory details on user’s trust. The results showed that the pilots were willing to use such a system and that the developed interface has a good usability with a System Usability Scale (SUS) score of 75.49 and a Post-Study System Usability Questionnaire (PSSUQ) score of 2.52. Regarding the perceived trustworthiness of the system, there was no significant difference between both groups, except in the perceived sufficiency of information provided to establish adequate trust. Here, in the second session, the group with more explanatory details rated that they perceived their information to be more sufficient to trust the system adequately. Furthermore, the results highlighted areas for improvement in the interface and showed other possible functionalities to consider in the future.