Do We Blindly Trust Self-Driving Cars

Abstract

Trust is an essential factor in ensuring robust human-robot interaction. However, recent work suggests that people can be too trusting of the technology with which they interact during emergencies, causing potential harm to themselves. To test whether this "over-trust" also extends to normal day-to-day activities, such as driving a car, we carried out a series of experiments with an autonomous car simulator. Participants (N=73) engaged in a scenario with no, correct or false audible information regarding the state of traffic around the self-driving vehicle, and were told they could assume control at any point in the interaction. Results show that participants trust the autonomous system, even when they should not, leading to potential dangerous situations.