I am developing practical methods for evidence-based AI decision-making. My current research focuses on whether explicit evidence-admission boundaries can reduce unsupported consequential decisions while preserving useful decisions.