A method linking institutional baseline conditions to AI system evaluation to assess whether deployment repairs or compounds existing failures.