Human-to-AI Interfaces for Enabling Future Onboard Experiences

This paper presents a novel platform for supporting human-centric design of future on-board user interfaces. This is conceived to facilitate the interplay and information exchange among onboard digital information systems, autonomous AI agents and human passengers and drivers. Two Human-to-AI (H2AI) Augmented Reality (AR) interfaces, characterized by different degrees of immersivity, have been designed to provide passengers with intuitive visualization of information available in the AI modules controlling the car behavior. To validate the proposed user-centric paradigm, a novel testbed has been developed for assessing whether H2AI solutions can be effective in increasing human trust in self-driving cars. The results of our initial experimental studies, performed with several subjects, clearly showed that visualizing AI information brings a critical understanding of the autonomous driving processes, which in turn leads to a substantial increase of trust in the system.