This summary report examines how safety models traditionally focus on crash frequencies or rates and do not always reflect the underlying causes of the crashes. Even models that accurately estimate crash measures and consist of multiple causal factors may not explain all causal relationships. Yet understanding what causes crashes is crucial to developing countermeasures and ensuring safety. To expand traditional modeling practices and results, the Federal Highway Administration’s Exploratory Advanced Research (EAR) Program sponsored a project to develop a framework that would generate realistic artificial datasets (RADs) that mimic the known causal relationships between contributing factors and crashes. The researchers applied the framework to generate RADs for ramp terminals and speed change lane facilities at diamond interchanges.

Recommended citation: Federal Highway Administration, Multidisciplinary Initiative to Create and Integrate Realistic Artificial Datasets (Washington, DC: 2023) https://doi.org/10.21949/1522003