Does Accountability and an Automation Decision Aid’s Reliability Affect Human Performance in a Visual Search Task?

Abstract

Some automation aids are not 100% reliable and can therefore provide incorrect information to users. If humans trust and comply with faulty automation aids, performance errors may result. In combat identification, performance errors may include fratricide or failure to identify enemies. In operator-automation teams, human accountability may reduce errors resulting from over-compliance and automation bias. The goal of the proposed study was to evaluate if reliability of an automation aid (60%, 80%) and accountability level (accountability, non-accountability) would influence human performance on a visual search task. Fifty-two undergraduate students performed a simulated combat identification task, aided by automation. Those in the 60% reliability condition committed more identification errors. However, this difference was eliminated for accountable participants. The findings suggest increasing operator accountability may improve identification performance and meliorate the adverse effects of unreliable automation.