Decision Justification Engineering: Evidence, Human Permission, and Continuing Authority

Decision Justification Engineering (DJE) explores what justifies delegating a defined task to an automated system—and what must remain true for that authority to continue. This release brings together a presentation, three-page executive summary, four conceptual companions, and an end-to-end case guide. A real-property valuation laboratory provides a concrete setting for examining evidence, quality control, repeated AI estimates, and the distinction between engineering sufficiency and human permission. The laboratory examines source-evidence representation, ten-qualifying-run ensembles, outcome-redacted retrospective predictions, one genuinely prospective prediction, and Leave-One-Out Cross-Validation characterization of two evidence states. The proposed governance architecture addresses continuing authority, its revocation, and controls on individual candidates. Measured findings are distinguished from illustrative standards and proposed deployment controls. The work does not demonstrate an end-to-end autonomous deployment. Start with the executive summary for orientation or the presentation for the full narrative. The website provides guided access to the material; the repository supplies supporting workbooks and experimental records. These linked resources are not all included in this seven-PDF deposit. Guided website: https://dje.truth-machine.org/ Supporting repository: https://github.com/truth-machine/decision-justification-engineering