Generated User Interfaces for Human-AI Interaction in Cars: Evaluation of AI-Generated Motion Sickness Notifications in Automated Cars

Generated user interfaces use generative AI models to generate important parts of the human-machine interaction. The advantage is a context sensitive, personalized and learning interaction. Such a highly adaptive interaction was generated to realize a motion sickness support system for automated cars. In automation level 3, 4 and 5, drivers and passengers may not look outside of the front window but get engaged in non-driving related activities (NDRA) that cause motion sickness (MS) in combination with accelerations caused by curves or braking. A camera-based occupant monitoring system was applied in a user study to detect activities that cause motion sickness on curvy roads. A generated user interface notified the participants ahead of curvy road sections. The study was conducted in a full cabin driving simulator. 21 participants were involved in a in-between subjects design with permutated order. The independent variable changed the adaptiveness of the notification. A) A notification was triggered once before any curvy road section, B) a notification was triggered constantly during the curvy road section, and c) a notification was triggered at coincidence of curvy road and NDRA-MS. The dependent variables (DV) included motion sickness-induced and non-induced track segments, as well as specific non-driving related activities (reading, turning around, using a smartphone) that were identified as contributing to motion sickness. The results show that adaptive, generative interfaces improve response times and reduce mental workload, while also providing users with a sense of control and security. Adaptive notifications led to a reduction in motion sickness symptoms compared to static notification strategies. Participants also reported greater comfort and safety during the experiment when adaptive notifications were used. We conclude that optimizing AI-generated interfaces in automated vehicles is crucial for minimizing confusion, mental load, and time pressure.