Federal Project Grant Award Summary

The University of Wisconsin-Madison received a $250,000 project grant from the National Science Foundation's Division of Mathematical Sciences under the Mathematical and Physical Sciences (CFDA 47.049) program, effective July 1, 2026, through June 30, 2029. This award supports the development of reliable statistical methods for estimation, prediction, and causal inference when using multiple AI (Artificial Intelligence)-generated synthetic datasets. The research addresses a critical gap in scientific practice by establishing the statistical foundations necessary to safely integrate heterogeneous and potentially low-quality synthetic data generated by machine learning models and large language models into rigorous scientific analysis, ensuring that AI-generated data serves as a calibrated scientific resource rather than a liability in high-stakes applications such as clinical research and public health.

The project will deliver a unified statistical theory and methodological framework that treats synthetic data as auxiliary information, automatically calibrating its contribution to prevent losses relative to labeled-data baselines while exploiting informative sources when available. Complementing the core research outcomes, the award supports workforce development initiatives including graduate and undergraduate student training, curriculum integration at the University of Wisconsin System, and the public release of open-source software that will make the methodology broadly accessible to the scientific community, federal agencies, and industry partners. The research advances national priorities in responsible AI deployment and scientific integrity while building U.S. research capacity through educational and technology transfer components.