The Decision Reconstruction Performance Index™ (DRPI™) establishes the first formal measurement and scoring standard for evaluating execution-layer decision governance within AI-supported justice systems. As part of the Justice Decision Observability™ (JDO™) Framework Series, DRPI™ provides a structured methodology for assessing the completeness, integrity, and evidentiary quality of reconstructed decision pathways. While existing governance frameworks focus on system design, validation, and compliance, DRPI™ operates at the execution layer—where automated system outputs are interpreted and acted upon by human decision-makers. It introduces a standardized approach to evaluating how well decision pathways can be reconstructed and whether governance conditions are sufficiently documented to support institutional accountability. DRPI™ is built on four core measurement dimensions: reconstruction completeness, decision pathway integrity, evidentiary alignment, and governance consistency. Together, these dimensions provide a repeatable and comparable scoring structure that allows institutions to assess the quality of governance documentation across events, environments, and operational contexts. This standard enables the transition of Justice Decision Observability™ from a descriptive governance discipline to a measurable operational system. By introducing structured evaluation criteria, DRPI™ supports cross-case comparison, institutional benchmarking, and the identification of recurring governance conditions that influence decision outcomes. DRPI™ is descriptive, system-level, and non-adversarial. It does not evaluate algorithmic performance, determine legal liability, or assign individual fault. Instead, it measures the strength and reliability of decision pathway reconstruction as a foundation for transparency, oversight, and institutional learning. As automated technologies continue to shape decision-making within justice systems, DRPI™ provides the measurement infrastructure necessary to ensure that execution-layer governance is not only observable, but consistently and reliably evaluated.
Fleming et al. (Tue,) studied this question.