Justice Decision Observability™ (JDO™) establishes a new governance discipline focused on the execution layer of AI-supported justice systems where human decision-makers interpret, rely upon, and act on system-generated outputs in real-world operational environments. While existing governance approaches emphasize system design, model performance, compliance, and policy frameworks, they do not account for how decisions are actually formed at the point of execution. Between system-generated signals (such as alerts, classifications, or risk scores) and operational outcomes (such as supervision actions, custody decisions, or incident responses), a critical layer remains largely undocumented: the human interpretive and discretionary process that determines how those signals are understood and acted upon. This primer defines that missing layer as the domain of Justice Decision Observability™. JDO™ introduces the concept of the Moment of Authority the point at which a human decision-maker interprets available information, applies discretion within institutional context, and exercises authority to produce an outcome. It is at this moment that system outputs are translated into action, and where variation, ambiguity, and reliance behavior directly shape real-world consequences. The framework identifies a persistent governance gap between system visibility and decision accountability. While system signals are often well-documented, and outcomes are formally recorded, the interpretive pathway between them is typically unstructured, inconsistent, or entirely absent from the institutional record. This gap limits the ability of justice institutions, oversight bodies, and technology providers to reconstruct decisions, assess consistency, or establish defensible explanations following critical events. Justice Decision Observability™ addresses this gap by establishing a structured approach to documenting how decisions are actually made. It captures how signals are encountered within workflows, how meaning is assigned, how discretion is applied, how authority is exercised, and how that pathway is preserved in a reviewable and reconstructable format. This primer outlines the core components of the discipline, including: the conceptual architecture of JDO™ across the signal-to-outcome lifecycle the definition and operational significance of the Moment of Authority™ the identification of the decision visibility gap within justice system operations the role of reliance behavior, institutional context, and discretionary judgment in shaping outcomes the boundary conditions that distinguish JDO™ from adjacent domains such as model auditing, compliance review, and technical system evaluation Justice Decision Observability™ does not evaluate algorithmic performance, test for bias, certify compliance, or assess the technical integrity of systems. It operates after systems are deployed and focuses exclusively on documenting human interaction with system outputs at the point of decision. By making decision pathways visible, JDO™ enables: clearer reconstruction of critical incidents and decision sequences reduced ambiguity in internal reviews, investigations, and legal proceedings stronger alignment between policy, practice, and observed behavior improved transparency and defensibility in high-stakes operational environments more accurate attribution of responsibility between system design and human decision-making For justice institutions, JDO™ provides a structured method for understanding how decisions are actually made under real-world conditions. For technology vendors, it clarifies how system outputs are used in practice and distinguishes system performance from human reliance. For legal, oversight, and compliance bodies, it establishes a framework for evaluating whether decision pathways can be reconstructed and whether execution-layer governance is preserved. Justice Decision Observability™ represents the formalization of a previously unstructured domain of governance. It defines the human decision layer as a distinct, observable, and documentable component of AI-supported systems—establishing the foundation for a new standard in accountability, transparency, and operational governance across justice environments.
Fleming et al. (Sun,) studied this question.