-
SAFE-T LOGO
SAFE-T: Safety Algorithm Fairness Evaluation for Transportation
Durham, North Carolina, is using AI-driven tools such as StreetLight Data and crash prediction models to prioritize where sidewalks, bike lanes, and other safety projects are funded. These systems shape Durham’s Road Safety Audits under the $5 billion federal Safe Streets and Roads for All program (2022–2026), which has already awarded $3.9 billion to more than 2,000 communities and will conduct a final competitive funding round in FY2026. Decisions made in this cycle will determine local infrastructure for decades
The consequences are immediate. An average of 23 people die on Durham’s roads each year, and most victims are pedestrians and cyclists. Fatal and serious injuries concentrate in low-income and historically Black neighborhoods that still lack basic protections such as sidewalks despite decades of documented crashes. Smartphone and fitness app data overrepresent affluent and recreational users, and crash records reflect uneven reporting. Without independent evaluation, these systems can misidentify need, direct resources away from high-injury corridors, and entrench existing disparities under the appearance of objectivity.
SAFE-T (Safety Algorithm Fairness Evaluation for Transportation) introduces an open, replicable benchmark to audit whether transportation AI equitably detects, predicts, and prioritizes risk. SAFE-T evaluates five dimensions: (1) demographic equity in volume estimation, (2) parity in fatal-crash prediction accuracy, (3) alignment between recommendations and high-injury corridors, (4) modeling of suppressed (latent) demand, and (5) representativeness of input data sources. Given that funding decisions are underway now, SAFE-T is urgently needed in Durham and broadly applicable to other cities before algorithmic bias becomes permanently built into public infrastructure.
Log in or sign up for Devpost to join the conversation.