Creating trust in fully autonomous vehicles through spatialized predictive sonification

• The impact of a sonic human-machine interface fostering trust in fully autonomous vehicles is empirically tested using an immersive prototype consisting of a sonification concept and a VR autonomous car ride. • Spatialized predictive sonification of upcoming driving maneuvers can provide intuitively understandable feedback to passengers. • Spatialized predictive sonification of upcoming driving maneuvers can improve general trust in autonomous vehicles and performance in non-driving related activities. The increasing automation of vehicles poses challenges in terms of public acceptance. In particular, trust in fully autonomous cars and their perceived benefits will play an important role in market acceptance. The present study investigates to what extent auditory feedback based on spatialized and predictive sonification can increase trust in autonomous vehicles while communicating their benefits. For this purpose, a sonification concept (Autonomous Vehicle Assurance, AVA) was developed, that anticipates upcoming driving maneuvers using predictive and spatialized sonification. An immersive prototype allowed users to experience AVA through a 22-channel Ambisonics loudspeaker system while participating in a virtual autonomous car ride. An experiment with 42 participants was conducted to measure the impact of AVA compared to conventional electric motor sounds. The comprehensibility of the auditory stimuli, the participants' trust in the reliability and safety of the vehicle in critical situations, as well as the context-independent trust in the vehicle were examined. In addition, the study investigated whether AVA helps occupants to better focus on secondary, non-driving activities and whether the perceived value of an autonomous car can ultimately be increased by additional auditory feedback. Using AVA as an example, the results show that spatialized predictive sonification of upcoming driving maneuvers can provide intuitively understandable feedback to occupants. Empirically, this led to improved performance on secondary activities, as participants were less likely to look away from the task at hand to ensure that the car was behaving correctly. The additional sonic feedback also led to greater overall trust in the car. However, the present results could not confirm the hypothesized positive influence of AVA on trust in specific safety-relevant situations and also did not lead to a higher perceived value of autonomous vehicles. In conclusion, the results of the present study indicate that spatialized, predictive sonification of upcoming driving maneuvers is an understandable and effective interaction design solution that should be considered when designing future human-machine interfaces for autonomous vehicles.