The study of human-AI interaction in criminal justice has been constrained by an access problem: the most consequential interactions happen inside institutional black boxes. Justice Decision Observability provides a governance documentation methodology that produces structured, replicable data about how humans interact with AI outputs in high-stakes justice settings.
The Research Opportunity
Academic research on AI in criminal justice has primarily addressed two domains: the technical performance of algorithmic systems (fairness, accuracy, bias) and the policy frameworks governing their deployment (regulation, ethics, standards). The human-behavioral layer, meaning what decision-makers actually do with AI outputs, remains underexplored.
This is not for lack of interest. It is for lack of data. The interactions that matter most (a parole officer responding to a risk score, a judge weighing an algorithmic assessment, a corrections administrator acting on an automated alert) are typically undocumented. Without documentation, there is no data to study.
Justice Decision Observability produces that data.
Research-Relevant Frameworks
JB-DOF™ as Research Architecture
The five pillars of the Justice Beacon Decision Observability Framework map to established research domains: Signal Integrity connects to information processing and decision support literature. Reliance Behavior Mapping connects to automation bias and technology acceptance research. Discretion Governance connects to street-level bureaucracy theory. Institutional Pressure Mapping connects to organizational behavior and institutional isomorphism. Outcome Integrity Monitoring connects to program evaluation methodology.
Operationalizable Variables
Each JB-DOF pillar generates documentable, measurable variables: override rates, reliance frequencies, information completeness scores, discretion exercise patterns, institutional pressure indicators, and outcome concordance metrics. These variables are designed to be suitable for quantitative analysis while the governance narratives provide qualitative depth.
Cross-Disciplinary Relevance
JDO sits at the intersection of multiple research fields:
- Human-Computer Interaction (HCI): how AI interfaces shape decision-making
- Organizational Behavior: how institutional dynamics mediate individual responses to AI
- Criminology: how algorithmic tools reshape discretion in justice settings
- Social Psychology: automation bias, authority compliance, and decision-making under uncertainty
- Public Administration: governance compliance, street-level implementation, accountability structures
- Science and Technology Studies (STS): sociotechnical systems, technology-in-practice
Collaboration Opportunities
JBS welcomes collaboration with academic researchers studying human-AI interaction in institutional settings. Stephanie Fleming holds a PhD in Social Psychology and has held adjunct faculty positions at Florida State University, Saint Leo University, Tallahassee Community College, and the University of Phoenix. JBS is interested in research partnerships that advance the empirical study of governance documentation effectiveness, human reliance behavior in justice settings, and the institutional factors that mediate AI-informed decision-making.
Research Collaboration
For research partnerships, data access inquiries, or academic consultation, contact JBS directly.
Contact JBS