The Department of Labor is embracing AI (www.dol.gov)

The Department of Labor announced a prototyping initiative—Artificial Intelligence Adjudicator Assistance (AIAA)—to explore how AI can support state Unemployment Insurance (UI) adjudication, a bottleneck exposed by the pandemic’s huge spike in claims. Framed by the White House Executive Order on trustworthy AI, the effort aims to test whether models can triage claims (distinguish cases needing substantial fact‑finding from routine ones) and assist non‑discretionary tasks to reduce back‑and‑forth, speed determinations, and relieve strained staffing and legacy systems. The DOL positions this as a low‑risk, research‑first approach to inform state policy and procurement rather than immediate deployment. Technically, AIAA trains and evaluates models on historical, high‑quality determinations in a locked environment: Colorado’s Department of Labor and Employment provides past claims data and real‑world adjudicator feedback while Stanford’s RegLab helps develop, test, and document methods. The initiative emphasizes closed‑model prototyping, rigorous testing against human decisions, and transparent documentation (blog series on prerequisites, trustworthy AI practices, risks, and infrastructure). Key implications include potential efficiency gains and workflow redesign for UI programs, but also well‑noted risks—data quality, bias, and misclassification—necessitating careful scoping, human oversight, and infrastructure investments before any operational use.