Guilty until proven innocent: Experiencing errors in training promotes relationship equity and trust calibration

The purpose of this study was to examine longitudinal trust formation with an autonomous system, focused on the impact an autonomous teammate's performance had on user trust, consequent monitoring behavior, and performance. Participants were randomized into one of two conditions; they either worked with a perfect or imperfect autonomous teammate to prevent missiles from striking their home base. Participants completed two phases of the task: a training phase, where their autonomous teammate either occasionally committed errors or not, and a testing phase, where their teammate always committed a singular error halfway through the task. Subjective trust ratings and participant performance data were collected and analyzed. Results demonstrated that working with an imperfect autonomous teammate in the training phase led to lower trust and increased monitoring behavior. Importantly, participants in the imperfect teammate condition prevented errors at a higher rate in the testing phase in comparison to participants who worked with a perfect teammate in the training phase. Results are discussed in terms of a new theory of relationship equity and the implications for training.