Talk to me!: exploring stereoscopic 3D anthropomorphic virtual assistants in automated vehicles

Trust in self-driving cars is considered to ensure appropriate usage of such vehicles. Representing the automation system as an anthropomorphic virtual assistant that communicates the system's behavior could help to build and maintain trust. To explore such assistants, a prototype of an anthropomorphic virtual assistant was designed and integrated in a conditionally automated vehicle. In a driving simulator experiment with a between-group design (n = 56), participants experienced either an anthropomorphic 2D or stereoscopic 3D virtual assistant. While driving, distrust towards the automation system was gradually evoked. We analyzed trust and driving behavior. The current study yielded findings that the implementation of stereoscopic 3D impacts overtaking behavior, potentially fosters safety-consciousness and mitigates inappropriate usage of the system. We further report on the design of the virtual assistant.