HCC: AI assistance for human driver readiness and decision-time interaction: shared autonomy under compressed decision horizons

The National Science Foundation Division of Computing and Communication Foundations awarded Clemson University $600,000 under the Computer and Information Science and Engineering program (CFDA 47.070) on October 1, 2026.

The project develops computational methods for AI systems to coordinate with humans in safety-critical shared-autonomy environments where humans must interpret AI recommendations and take action within seconds. The work addresses the mismatch between current AI systems, which assume humans are always ready to respond, and research showing that human readiness varies with attention, workload, stress, situational awareness, and trust. The project comprises three thrusts: developing computational models that estimate human readiness from behavioral and physiological measurements including response time, eye gaze, heart rate, and simulator interaction data; developing decision-time interaction methods using alerts, prompts, and low-burden cognitive probes to reduce uncertainty about readiness while limiting cognitive burden; and developing deployment strategies for applying these principles across transportation, assistive robotics, healthcare, and other safety-critical applications.

Performance occurs at Clemson University in Clemson, South Carolina. The award runs through September 30, 2029.

Abstract

Artificial intelligence (AI) systems are increasingly used in shared autonomy, where humans and AI systems must coordinate under extreme time pressure, partial information, and rapidly changing risk. Examples include automated vehicles, robotics, emergency response, and other safety-critical systems. In these settings, humans may need to interpret warnings, decide whether to trust AI recommendations, and take action within only a few seconds. Current shared-autonomy systems often assume that humans are ready to respond at all times, even though research has shown that readiness can change quickly with attention, workload, stress, situational awareness, and trust. Even when an AI system detects a hazard correctly, poorly timed communication or a poorly matched request for human action can delay intervention, increase workload, or reduce trust. This project studies how AI systems can better coordinate with humans by adapting AI communication and assistance to changing human readiness under time pressure. The resulting principles can improve the safety, reliability, and trustworthiness of future shared-autonomy systems in transportation, assistive robotics, healthcare, and other safety-critical applications. The core idea of this project is to treat human readiness as a dynamic and uncertain condition that AI systems should consider when deciding when and how to interact with humans. The project has three thrusts. The first thrust develops computational models that estimate readiness from behavioral and non-invasive physiological measurements, including response time, eye gaze, heart rate, and simulator interaction data. The second thrust develops decision-time interaction methods that use alerts, prompts, and low-burden cognitive probes to reduce uncertainty about readiness while limiting cognitive burden. The third thrust develops deployment-time regulation methods that maintain stable human-AI coordination as humans adapt through repeated interactions. Human-in-the-loop driving simulator studies using advanced driver assistance systems (for example, forward collision warning) will evaluate the proposed methods under time-critical conditions. Expected outcomes include computational models for readiness-aware human-AI coordination, adaptive interaction algorithms, publicly available datasets and experimental resources, and evaluation methods that can inform the design of a broad class of shared-autonomy systems operating under uncertainty and time pressure. 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.- subawards are not planned for this award.