The Effects of Automation Expertise and System Confidence on Trust Behaviors

Abstract

Trust in automation is more likely to be appropriate when information about the automation's capability is available. The goal of this study was to determine how automation expertise and system confidence affected automation trust behaviors. Forty-one participants completed a target detection task while receiving advice from an imperfect diagnostic aid that varied in expertise (expert vs. novice) and confidence (75% vs. 50% vs. 25%, no aid). Results showed that participants were more willing to comply with the highly confident expert aid than the highly confident novice aid. Furthermore, participants were more apt to generate false alarms as system confidence increased. These results suggest that, similar to interpersonal relationships, humans appraise automation features such as confidence and expertise when deciding to comply with automation. Implications and direction for future research are discussed.