The rapid progress in Artificial Intelligence (AI) has expanded the capabilities of highly automated vehicles in many cities worldwide. These developments are paving the way for a future in which fully autonomous vehicles will handle all driving tasks, leaving humans as mere passengers. However, there are many driving situations that an automated system cannot handle correctly on its own. This project redefines the human-AI relationship and moves beyond viewing automation as either a simple tool to be commanded or a perfect system to be blindly trusted. Instead, the most effective and safest way to integrate humans and AI on a vehicle is to structure them as members in a collaborative partnership. This relationship will allow humans to share their intentions with the AI, help detect obstacles, enable the AI to adjust its driving behavior based on the human’s attention level, or allow the AI to ask for feedback on certain driving maneuvers, among other features. The fundamental discoveries from this research have significant potential to impact other disciplines in which human-AI teaming is relevant, such as robotics, manufacturing, emergency response systems, and healthcare. Furthermore, the project integrates interactive community workshops and an applied robotics challenge to educate high school students and adults about artificial intelligence and human-machine interaction. These educational activities will foster a greater understanding of automation while preparing the future workforce for careers in science and engineering.
The primary goal of this award is to improve the safety, performance, and resilience of human and artificial intelligence driving teams in highly automated vehicles. This research is based on the recognition that trust, shared situational awareness, and joint actions form a continuous feedback loop where their mutual interdependence is essential for effective team design. To achieve this goal, the research team will conduct various driving simulation studies to develop computational models that estimate both the trust humans place in artificial intelligence and the trust artificial intelligence has in human teammates. Simultaneously, the project will evaluate variations in the team’s situational awareness through adaptive information displays and reciprocal information exchanges. These computational models will be used to implement dynamic trust calibration strategies, such as adjusting the vehicle’s driving style, level of transparency, communication style, or intervention thresholds to directly influence human engagement. Moreover, the researchers will develop negotiation protocols for strategic joint decisions to manage troubling scenarios in dynamic driving environments. By evaluating this comprehensive framework through extensive simulation studies, the project aims to develop adaptive algorithms and innovative evaluation methods that enhance human factors in automated systems.
This award reflects NSF’s statutory mission and has been deemed worthy of support through evaluation using the Foundation’s intellectual merit and broader impacts review criteria.