Pilot Dependence on Imperfect Diagnostic Automation in Simulated UAV Flights: An Attentional Visual Scanning Analysis.

Abstract

An unmanned air vehicle (UAV) simulation was designed to reveal the effects of imperfectly reliable diagnostic automation a monitor of system health parameters on pilot attention, as the latter was assessed via visual scanning. Four groups of participants flew a series of legs under different automation conditions: a baseline (no automation) control, and automation which was either 100% reliable, 60% reliable with a low-threshold bias to produce false alerts, and 60% reliable with a high threshold to produce misses. A high workload mission completion task and ground surveillance task were simultaneously imposed. Consistent with the reliance-compliance model of imperfect automation developed by Meyer (2001), miss-prone automation removed visual attention from the surveillance task, while FA-prone automation delayed the alert-driven attention shift to the system monitoring task.