Presentation
Querying humans about their understanding of the task environment does not support supervision of a highly automated work system
DescriptionHighly automated systems improve workplace productivity and safety but can introduce risks, especially when automation failures are rare. Interventions are needed to facilitate appropriate human supervisory oversight. Three experiments were conducted to investigate whether periodically presenting queries about the current and future state of a task display (environment) could improve participants’ detection of a single failure of automation in simulated air traffic control (ATC). University student participants completed two 30-min ATC scenarios. In the manual scenario, they performed all tasks manually. In the other scenario, automation accepted/handed off aircraft and detected and resolved aircraft conflicts (violations of aircraft separation). The conflict detection automation failed once late in this scenario. Participant speed and accuracy to intervene to this automation failure was recorded. It was predicted that periodically querying humans about the task environment would facilitate automation failure detection, however the empirical outcomes provided no support for this. Results instead highlighted potential risks associated with this approach, which degraded participants’ ability to detect the conflict missed by automation. Even queries that specifically directed participants attention to the common cruising altitude and converging trajectory of the aircraft pair involved in the impending automation failure provided no benefit to detecting that conflict.
Event Type
Lecture
TimeThursday, October 22nd11:40am - 12pm PDT
Location
Similar Presentations