Automation and Interpersonal Trust: The Sociotechnical Implications of Unreliable Automation

This study examined the potential interactions between human trust and human trust of automation on outcomes such as attribution of credit and blame, trust in partner, trust in automation, team performance, and perception of reliability.Using a micro-world computer program called Networked Fire Chief, 186 participants in 93 teams were assessed on a variety of outcomes, including attribution of blame for errors and credit for success, trust in their human partner, trust in the automation, perceptions of the reliability of the automation, and performance of the team.These teams were comprised of either teams of friends or teams of strangers to compare levels of human trust, and experienced either 100% reliable automation or 70% reliable automation.Results showed that individuals blamed themselves for mistakes less when they were in the 100% reliable automation condition, and that there was a three-way interaction of human trust, reliability of the automation, and perfect automation schema (PAS) on the perception of the reliability of the automation.The relationship of the variables with PAS changed depending on whether the participants were in Matched Trust or Mixed Trust environments, suggesting context incorporating human trust characteristics is important for automation research going forward.Implications of these findings are explored, as well as possible future directions.