Criminal Justice Algorithms as Void Objects: Recidivism Prediction Opacity, Adaptive Enforcement, and False Positive Harm — A Within-Domain Spearman Convergence

Ten criminal justice algorithmic systems (COMPAS, PSA, SAS Risk Score, OASys, PredPol/Geolitica, ShotSpotter, Palantir Gotham, and others) scored on O (score calculation opacity — trade secret protection), R (feedback loop in predictive policing), and C (jurisdictional lock-in, legislative mandate binding). Spearman ρ between Void Index and documented harm: ProPublica racial false positive rates, Dressel & Farid accuracy vs. human baseline, Lum & Isaac predictive policing feedback harm, ACLU false arrest rates. EU AI Act Annex III §6 — PROHIBITED category for predictive policing. Highest adversarial value: framework survival in this domain answers the 'soft social domain' objection.