The first self-driving taxi was launched in 2016 in Singapore. Since then, more cities around the world have introduced automated taxis. Past studies have identified trust as an important component...Show moreThe first self-driving taxi was launched in 2016 in Singapore. Since then, more cities around the world have introduced automated taxis. Past studies have identified trust as an important component to ensure users' safety when interacting with automated vehicles and benefitting from the available technology. Nonetheless, automated systems are not failure-proof and can result in fatal accidents. The concept of trust in automation is related to interpersonal trust: A trustor puts themselves at risk by expecting a certain outcome. Therefore, we conducted a driving simulator study at Leiden University to investigate the development of trust in automation compared to interpersonal trust after experiencing a system malfunction. We tested 41 participants and divided them into two groups. Participants were either passengers in a conventional taxi or in an automated vehicle. During the simulated drive, participants rated their trust verbally on a 7-point Likert scale. Additionally, trust in automation or a taxi driver was assessed through a questionnaire before and after the experiment. Our results showed that the development of trust did not differ between the groups during the drive. We observed a temporary decrease in trust after the taxi ran a red light, but trust recovered at the end of the drive. Trust in a human driver was higher than in a self-driving car before the simulation but did not differ anymore at the end of the experiment. This suggests that trust in automation develops the same as interpersonal trust after experiencing a dangerous situation.Show less
This study investigated the development of trust in automated vehicles. To do so, we set up an online experiment with two conditions. The first condition is called ‘Perfect’, and the second...Show moreThis study investigated the development of trust in automated vehicles. To do so, we set up an online experiment with two conditions. The first condition is called ‘Perfect’, and the second condition is called ‘Poor’. In addition, we measured the risk-taking attitude of the participants in this study. We performed multiple statistical analysis to determine the link between trust in the automated vehicle and the phase (in each condition there are eight phases). Furthermore, we utilized regression analysis and correlations to see whether risk taking could predict the level of trust. The results showed a significant main effect for phase and a significant interaction effect for phase and group. Especially the trust in Phase 4 decreased in the perfect vehicle condition and increased again after Phase 5. However, there were no significant correlations between risk-taking attitude and trust in the automated vehicle. Together, these results provided insight into the development of trust over time in automated vehicles, especially after one critical incident. To conclude, we suggested several improvements for future research. These suggestions included taking gender differences into account, replicating our study in a driving simulator and providing the participants with information about the system. Furthermore, we advised investigating the link between risk taking attitude and trust in a sample with high risk-taking participants.Show less
