Attempting to automate compliance to aircraft collision avoidance advisories

As automation increasingly takes over functions requiring humans' skill- and rule-based behaviors, effective human-automation interaction increasingly depends upon human reasoning about when the automation should be disengaged. This paper uses the aircraft traffic alert and collision avoidance system (TCAS) as a case study where the pilot needs to decide how to respond to decision automation that generates collision avoidance advisories commanding vertical maneuvers. Of note, recent installations of TCAS can now automatically execute the advisories, potentially to improve compliance with TCAS advisories. This paper describes a Human-in-the-loop flight simulator study examining pilots' interaction with the automatic execution of advisories, including situations in which pilots chose to disengage it. While overall automatic execution appears to improve compliance to TCAS advisories in the aggregate, it was also disengaged in some of the same situations in which pilots historically had not manually flown compliant avoidance maneuvers. Further, the autopilot's subsequent reversionary mode settings after following the TCAS advisory may not be appropriate for returning to air traffic clearances. These results highlight the importance of designing automation to support effective pilot reasoning about traffic events, a task that remains in place while the pilot remains ultimately responsible for safety.