Federal Project Grant Summary

The National Science Foundation's Division of Information and Intelligent Systems awarded Harvard College $470,000 under the Computer and Information Science and Engineering (CFDA 47.070) program to develop interaction techniques and foundational principles for establishing long-term human-artificial intelligence (AI) collaborations. The project, which runs from October 1, 2025, through September 30, 2029, focuses on designing AI systems that maintain shared understanding with human collaborators across complex, extended projects in domains such as game design, document writing, and software development. The core deliverable centers on the "Intent Specification" (IS) interaction model—a human-readable representation of user goals, preferences, and project understanding that grounds long-term AI decision-making and alignment.

The research will produce open-source tools, software, and guidelines that address current limitations in AI systems' ability to track preferences, follow instructions, and adapt to project-specific context over extended timeframes. These outputs will benefit industry developers, researchers, and AI system users across the United States and Canada, advancing the national interest by enabling more effective and reliable AI assistance that supports human creativity and productivity while maintaining trustworthiness in collaborative settings. The project investigates key research questions through empirical studies to establish how common ground can be maintained between humans and evolving AI systems as projects grow in complexity.