NSF AI grant $1.25M: "Explainable AI" decision support for flood/hurricane evacuation sheltering (George Mason University)

The U.S. National Science Foundation (NSF) awarded about $1.25 million ($1,250,000) to George Mason University's "SCC-IRG: Resilient Sheltering Decision Support for Emergency Evacuations using Explainable AI" (NSF Award 2531369; program: S&CC — Smart & Connected Communities; October 2025–September 2028).

Per the abstract, evacuation and public sheltering move people out of harm's way and are common life-saving strategies for severe weather such as flooding and hurricanes. However, some citizens are less inclined to evacuate or use public shelters because of transportation challenges, past experiences, risk perceptions, and concerns about whether critical services will be available at shelters. From the emergency-management planning side, choosing which shelters to open and when based on infrastructure risk, optimizing resource allocation in operating shelters, and estimating shelter demand are all challenging.

Current decision-support systems rely primarily on weather forecasts, flood-risk assessments, retrospective knowledge of shelter usage, and past public behavior. But such inputs cannot fully account for the dynamic, evolving needs and movement of the public, nor the failure risks of the infrastructure needed to run shelters given uncertain, dynamic interdependencies such as transportation and power. This project addresses that gap with explainable AI.

Why it matters: in disaster response, it is decisive that humans — emergency managers and local governments — can understand and trust why the AI recommends a given action, not merely what it outputs. That is the point of "explainable" AI here. By trying to capture not only external weather conditions but also public behavior and infrastructure interdependence (if transport fails people cannot evacuate; if power fails shelters cannot function) dynamically, it advances beyond static support systems. It is a concrete example of Smart & Connected Communities — using AI for societal resilience — and points to a practical direction for AI in disaster management and public policy.