Home/Handbook/Our bias testing approach
CLINICAL AI

Our bias testing approach

Every model is evaluated for performance disparities across demographic groups before it goes to production. Disparities above a clinical-team-defined threshold block the launch.

We also monitor for drift in production. A model that launched fair can become unfair as the data distribution shifts, and that's a real failure mode.

We publish our methodology and the results internally. We expect to publish them externally when the work is mature enough to defend.

Curee AI

Ready to Deploy Your Digital Workforce?

See how Curee Workstation can automate your most time-consuming clinical, operational, and revenue cycle workflows. Live in days.