Trust in Autonomous Cars Does Not Largely Differ from Trust in Human Drivers when They Make Minor Errors

Abstract

Studies have explored the factor of trust in autonomous cars (ACs), and it has been shown that their ability and performance are crucial for determining trust. However, little is known about the effects of minor errors without involving negative consequences such as property damage and fatalities. People are likely to expect automation technologies to perform better than humans. It was, therefore, hypothesized that minor errors would destroy expectations and significantly decrease trust in ACs. This study aimed to investigate whether minor errors have a more negative effect on trust in ACs than in human drivers. Two experiments were conducted ( N = 821) in Japan. Two independent variables were manipulated: agent (AC and human) and error (error and no-error). Some participants were shown videos depicting ACs and human drivers making minor errors, such as taking a longer time to park (Experiment 1) and delaying to take off when traffic lights turned green (Experiment 2). These minor errors did not violate Japanese traffic laws. Others watched videos in which no errors occurred. The results of the two-way analysis of variance did not show evidence that the agent type modulated the negative effects of these minor errors on trust. Minor errors did not lead to a significant difference in trust levels between ACs and human drivers. This study also indicated that people expected ACs to not make more errors than humans did. However, these expectations did not increase trust in ACs. The findings also suggest that minor errors are unlikely to cause an underestimation of ACs’ capabilities.