Human-machine communication in self-driving cars: verbal feedforward and feedback increase trust, but not the intention to use it.

Abstract

Fully autonomous vehicles as cognitive agents will be one of the decisive use cases of artificial intelligence in the middle of modern societies. However, the potential positive effects of this technology will only apply if autonomous cars are accepted by a majority of the society the cars will operate in. To increase acceptance and trust, we created a virtual self-explaining car informing passengers prior to actions. In this study, we investigate the current attitude of the German and Austrian population (N = 8599) in a virtual reality experience. Participants experienced a ninety seconds virtual drive in one of three conditions of auditive feedforward and feedback, while head tracking data, and a simplified gaze vector were recorded. Our main finding is that a self-explaining car does have a positive impact on trust, but not influence the intention of using such a car. Additionally, we can show gender and age effects with female participants being generally less trusting in all conditions compared to male participants and a general decrease of acceptance with increasing age. Results are in line with previous human-machine research, supporting found gender differences as well as the fact that communications increase driving safety, but also negative emotional responses. These findings reveal a need for a well balanced self-explanatory artificial intelligence to enable human-machine interactions that foster safe traffic behavior and increase trust as well as the willingness to use such technology

Farbod Nosrat Nezami

Ph.D Candidate in Cognitive Science

My research interests is bringing human and machine closer to each other