CMU and Meta Build AI Tools to Track Population Movement in Disasters

Carnegie Mellon University's National Science Foundation AI Institute for Societal Decision Making is partnering with Meta's AI for Good program to build artificial intelligence tools that help emergency responders understand and respond to natural disasters in real time.

The collaboration will transform aggregated mobility and connectivity data into clear visualizations that emergency managers can use on the ground during wildfires, hurricanes, and severe winter storms. Researchers will combine Meta's existing data streams with satellite imagery and open-source AI models—including Segment Anything, DINO, and large language models—to create what the partners call dynamic situation reports.

Tracking evacuations and returns

The tools aim to answer critical questions during disasters: Are residents complying with evacuation orders? Where do populations remain in affected areas? When are communities beginning to return after an event?

"This NSF AI-SDM-Meta collaboration is an excellent example of how academic-industry partnerships can impact social good," said Rebecca Nugent, head and Fienberg University Professor of CMU's Department of Statistics and Data Science. "This outreach project aims to help emergency responders better understand population mobility during disasters as well as provide additional related resources to the public."

The NSF AI Institute for Societal Decision Making brings together researchers in artificial intelligence and social sciences to develop technologies for complex, high-stakes human decision-making. The institute examines how people perceive risk and trust, and how uncertain, dynamic, and resource-constrained circumstances shape decisions in critical situations.

Testing during 2026 disaster season

The project will evaluate the tools during natural disasters occurring in 2026 to determine which approaches work best in emergency situations. NSF AI-SDM already collaborates with several state and local emergency management agencies and will share the work with them to gather operational feedback.

"Our partnership with Meta solidifies an important informational piece relevant to AI-SDM's effort on designing effective disaster risk communication by understanding human mobility and networking behavior," said Aarti Singh, NSF AI-SDM director and FORE Systems Professor in CMU's Machine Learning Department.

Laura McGorman, director of Meta AI for Good, said leveraging AI and real-time mobility data "has the potential to transform the way governments and nonprofits respond to natural disasters."

The feedback collected during 2026 will inform future development of more interactive and potentially automated data tools for emergency response.

Why it matters

Emergency managers often make life-or-death decisions with incomplete information about where people are located during disasters. Real-time population movement data could help responders allocate resources more effectively, verify evacuation compliance, and identify communities that need immediate assistance—capabilities that become increasingly critical as climate change intensifies the frequency and severity of natural disasters.

Once finalized, the tools will be made available through the Humanitarian Data Exchange and the NSF AI-SDM website, according to details first reported by Carnegie Mellon University.