Using Multiple Imperfect Diagnostic Automation

Due to the difficulty of performing multiple tasks, operators in complex environments are often aided by automation. Because automation is not always perfect operators must decide how much to trust in and depend on the automation aids. Theoretically operators can adjust their level of trust using either a component-specific or a system-wide trust strategy. This study tests these two theories. 36 participants monitored two gauges, each with an automated aid at different reliability levels, while engaged in a pursuit tracking task that simulated an unmanned aerial vehicle (UAV) mission flight. The data suggest that participants do not evaluate the reliability of each gauge independently (i.e. component-specific trust), but instead combine their experience from each automation aid and derive one overall perceived reliability value for the entire system consistent with the system-wide trust hypothesis.