Leveraging data for a healthier environment

The Lab partnered with the U.S. EPA to conduct one of the first large-scale field tests comparing AI-based inspection targeting with traditional decision-making in regulatory inspections. Using 15 years of historical data, the researchers developed a model to predict which hazardous waste facilities were most likely to have serious violations. When tested in the field, the model detected 82 percent more violations relative to the traditional approach. The U.S. EPA is now using the model to inform site selection for hazardous waste facility inspections nationwide.