NSF Grant Backs Cornell AI Models of Human Choices in Complex Systems
Cornell researchers receive a three-year, $600,000 NSF grant to build interpretable AI models of human decisions in networked systems like grids and transit.
The story in focus
Cornell researchers have won a three-year, $600,000 National Science Foundation grant to develop AI models that better represent human decision-making in cyber-physical systems, such as power grids and transportation networks, while making uncertainty and trade-offs explicit.
- The project is led by Ricardo Daziano, professor of civil and environmental engineering, with Anna Scaglione, professor of electrical and computer engineering at Cornell Tech.
- The research combines Bayesian deep learning, microeconometrics and graph-based data to model choices that feed back into networked systems.
- A stated objective is building models that can recognize when a prediction is weakly supported by available data.
What the project covers
The grant supports work on AI that predicts how people behave in cyber-physical systems — environments where physical infrastructure meets digital information, automated recommendations and connected technology. Examples named include transportation networks, electric-vehicle charging systems and grid infrastructure.
Cornell said the project combines Bayesian deep learning, microeconometrics and graph-based data to produce interpretable choice models that account for interactions across heterogeneous networks. Ricardo Daziano, the lead investigator, said decisions depend on the physical environment, information, what others are doing and feedback from the system, and can then change the system for everyone else.
Why the grid is a focus
The researchers point to the electric grid as a key case: it is shifting from a few large power plants toward coordinating millions of distributed resources while absorbing new demand from electric vehicles and data centers. Anna Scaglione described the grid as too distributed, dynamic and complex to manage with existing tools.
Daziano said the work aims beyond predictive accuracy, toward understanding the trade-offs behind a prediction and whether a model stays reliable when asked about a new policy or technology. He traced the project's origins to doctoral research by his former student Daniel Villarraga, now at Universidad de los Andes in Colombia, on deep-learning discrete choice models with an interpretable behavioral core.
Who is leading the Cornell research?
Ricardo Daziano, professor of civil and environmental engineering, leads the project with Anna Scaglione, professor of electrical and computer engineering based at Cornell Tech.
Sources & further reading
- Grant aims to improve AI models of human choices in complex systems | Cornell Chronicle ↗news.cornell.edu
Help keep the record accurate. Send a correction.
