Individual differences in the calibration of trust in automation.

Abstract

The objective was to determine whether operators with an expectancy that automation is trustworthy are better at calibrating their trust to changes in the capabilities of automation, and if so, why. Studies suggest that individual differences in automation expectancy may be able to account for why changes in the capabilities of automation lead to a substantial change in trust for some, yet only a small change for others. In a baggage screening task, 225 participants searched for weapons in 200 X-ray images of luggage. Participants were assisted by an automated decision aid exhibiting different levels of reliability. Measures of expectancy that automation is trustworthy were used in conjunction with subjective measures of trust and perceived reliability to identify individual differences in trust calibration. Operators with high expectancy that automation is trustworthy were more sensitive to changes (both increases and decreases) in automation reliability. This difference was eliminated by manipulating the causal attribution of automation errors. Attributing the cause of automation errors to factors external to the automation fosters an understanding of tasks and situations in which automation differs in reliability and may lead to more appropriate trust. The development of interventions can lead to calibrated trust in automation.