NSF AI institute for human-AI cooperation

The National Science Foundation Division of Computer and Network Systems awarded the University of Delaware a cooperative agreement valued at $5 million on July 15, 2026, under the Computer and Information Science and Engineering program (CFDA 47.070) to develop and deploy human-AI cooperative systems for dynamic rehabilitation settings.

The award funds the NSF AI Institute for Human-AI Cooperation, which will establish a symbiotic framework enabling AI systems to function as adaptive teammates in human-centered, personalized decision-making. The institute will conduct convergent AI research organized around three integrated thrusts: multiscale and multimodal modeling that unifies diverse data sources and knowledge graphs for grounded learning representations; ethical, human-aligned AI assistive technology; and approaches that align AI operation with the evolving goals of human decision makers across multiple timescales. The research integrates grounded sensing, multiscale modeling, and human-centric decision-making to develop AI systems that optimize solutions, adapt to context and instructions, and support more effective decision-making. Deliverables include research innovations in rehabilitation science, workforce development initiatives, stakeholder engagement, and expanded societal benefits from AI deployment.

The University of Delaware, a research institution headquartered in Newark, Delaware, serves as the prime recipient. The award runs through June 30, 2031, with performance at the Newark, Delaware campus.

Abstract

The NSF AI institute for human-AI cooperation promotes national leadership in artificial intelligence by transforming AI into an adaptive teammate allowing human users and AI systems to interact in dynamic collaboration. The institute will build a new synergistic framework for the development of robust intelligent agents that users can rely upon in human-centered personalized decision-making, with compelling impact in dynamic rehabilitation settings. The institute will conduct convergent AI research in real-world settings for human-AI interaction and will be enabled by state-of-the-art infrastructure and technology. By integrating grounded sensing, multiscale modeling, and human-centric decision-making, the institute will develop AI systems that optimize and personalize solutions, adapt to context and instructions and support more effective and responsive decision-making. These research innovations will be leveraged to strengthen workforce development, deepen stakeholder engagement, and expand the societal benefits of AI. The institute?S research vision is to develop a symbiotic framework for human-AI cooperation, enabling human-centered personalized models and decision-making that provide unprecedented performance by integrating data across timescales using grounded knowledge sources and evidence. This effort will turn fundamental AI advances into actionable inference, with contributions guided by pressing needs in rehabilitation science, where IT is essential that AI operation is aligned with the evolving goals of decision makers and users at multiple timescales. This innovative approach consists of three integrated thrusts: (1) multiscale and multimodal modeling that unifies diverse data sources and knowledge graphs to create grounded learning representations for analysis, hypothesis generation, long-term forecasting and decision-making support in time-evolving processes; (2) ethical, human-aligned AI assistive technology and semi-autonomous systems in digital, robotic and other embodied forms creating a representative state enabled by personalized models and verbal and non-verbal interfaces; and (3) novel algorithms and interfaces for instructing and aligning AI systems for more effective human-AI collaborative teaming, including new measures of reward, error and alignment informed by neuroscientific correlates of trust. The institute envisions AI systems that adapt to instruction and data-driven contextualization; ground their reasoning in heterogeneous, multimodal, multiscale data; provide population-wide and personalized predictions; and offer causal explanations for spatiotemporal forecasting and control. These systems will promote appropriate trust, maintain human autonomy, and support universal accessibility. The institute?S broader impact is to realize a new framework for human-AI symbiotic cooperation. The activities and operations will establish an essential nexus that links researchers, community stakeholders, and industry partners, ranging from computing hardware/software and medical equipment manufacturers to healthcare providers. The institute will develop the next generation of AI talent through inter-institutional training activities for workforce development and integration of the institute?S discoveries and developments across undergraduate, graduate and postdoctoral education. The institute?S long-term goal is to serve as a hub for human-AI collaborative research and education. An innovative discovery portal will provide education and research products built synergistically across partner institutions for public dissemination, supporting human-AI collaborative frameworks and making AI more accessible and impactful. 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 planned for this award.