2022 | HAITICS

For decades, airplanes have been one of the safest modes of transportation, thanks to high standards in aircraft design, testing, operation, maintenance and inspection, as well as comprehensive pilot training.

The operation of aircraft, especially those carrying passengers, involves highly complex, safety-critical situations and therefore places considerable psychological demands on pilots. For this reason, pilots are usually assisted by a co-pilot who can take control in potentially critical situations.

As technology advances, AI systems could enable single-pilot operations, with one pilot controlling the entire flight. The system could serve as a digital co-pilot - provided that a safe and reliable interaction between AI and humans can be established.

A crucial prerequisite for such cooperation is that the system provides information in a way that the human pilot can easily understand and interpret in order to act appropriately, for example by taking control. Similarly, the system must be able to correctly understand and interpret human actions in order to deduce when it needs to act and provide support in critical situations or take complete control. To meet this challenge, we provide insights into ongoing research on the requirements for safe operation when humans and AI collaborate in a cockpit. By conducting controlled behavioral experiments in a flight simulator, we evaluate the impact of information exchange on safety.

Our findings can reveal strategies for facilitating collaboration between AI and humans while providing critical food for thought on the emerging challenges.