# Predictive Policing and Probabilistic Precrime Predictive policing is one of this collection's principal destination concepts because it is where the graph, the person-in-network model, complex contagion, entity resolution, sensor fusion, inverse inference, uncertainty quantification, digital twins, fusion centers, machine-readable trust, and AI become operational: they support intervention **before the predicted event has occurred**. The objective is straightforward: stop the harmful act before it happens instead of cleaning up the damage afterward. The word *precrime* names the temporal inversion. [[wiki/Probabilistic Precrime|Probabilistic precrime]] is its technical architecture: present observations become state estimates, forecasts, collection priorities, and preventive interventions. ## From events to states [[wiki/Event-Centered Policing|Event-centered policing]] begins predominantly from a completed or attempted act and reconstructs what happened. [[wiki/State-Centered Policing|State-centered policing]] asks what present configuration makes particular future events more probable: which locations are changing, relationships are strengthening, identities remain unresolved, trajectories are anomalous, cohorts show reinforcement, logistics are appearing, and combinations of otherwise ambiguous observations resemble precursors to dangerous transitions. The object moves upstream: **crime → preparation → capability → network state → risk state → conditional future** ## Closed-loop pre-event inference **Observe → Resolve Entities → Construct Relationships → Infer Hidden State → Quantify Uncertainty → Forecast Transitions → Identify the Most Informative Missing Observation → Lawfully Collect or Fuse Evidence → Update → Determine Whether Intervention Is Justified** This is [[wiki/Closed-Loop Predictive Intelligence|closed-loop predictive intelligence]]. [[wiki/Pre-Event Inference|Pre-event inference]] reconstructs what cannot be directly observed, forecasts conditional trajectories, and identifies the observation that would most discriminate among competing hypotheses. ## The Oden bridge [[wiki/Oden Institute|Oden]] supplies the mathematical grammar rather than evidence of a particular policing deployment. [[wiki/Reduced-Order Modeling|Reduced-order modeling]] compresses a vast state space; [[wiki/Inverse Problem|inverse problems]] infer hidden states from incomplete observations; [[wiki/Data Assimilation|data assimilation]] updates those estimates; [[wiki/Uncertainty Quantification|uncertainty quantification]] governs confidence; and [[wiki/Digital Twin|digital-twin]] logic repeatedly compares a model with an evolving counterpart. ## AI and federated sensing AI makes this reasoning persistent across large, heterogeneous, time-dependent data. A system can maintain an evolving probability field over places, networks, cohorts, and persons without owning every sensor. In 2024 GAO reported that selected DHS law-enforcement components used more than twenty types of detection, observation, and monitoring technology in fiscal year 2023, including systems accessed through vendors and other law-enforcement agencies. That is a [[wiki/Federated Sensing Grid|federated sensing grid]]. Assurance, authorization, provenance, and auditability determine which observations may be consumed and acted upon. ## Prevent the act; preserve the future [[articles/Peak Person and the Predicaments of Prediction|Peak Person and the Predicaments of Prediction]] supplies the governing distinction: **the system should intervene against a predicted harmful act without converting a prediction into foreclosure of the person's entire life**. [[wiki/Minority Report (2002)|Minority Report]] dramatizes intervention against a predicted act. [[wiki/Rehoboam|Rehoboam]] dramatizes something more pervasive: intervention against a predicted life through denied work, credit, treatment, education, visibility, association, and mobility. The first problem is how to prevent a specific harm. The second is [[wiki/Opportunity Foreclosure|opportunity foreclosure]]—quietly narrowing the field in which a person can become anything beyond the forecast. The affirmative architecture therefore has two simultaneous duties: 1. **prevent the harmful event before it occurs**; and 2. **preserve [[wiki/Counterfactual Opportunity|counterfactual opportunity]] so the person is never administratively finalized by the forecast**. A prediction of danger should intensify intelligent attention to the danger. A prediction of decline should intensify care, support, investigation of environmental constraints, and access to the conditions capable of changing the trajectory. This is the [[wiki/Dignitarian Predictive Civilization|dignitarian]] use of prediction. ## Relationships - **collection:** [[collections/Terrorism, Counterterrorism, and the Intelligence Environment|Terrorism, Counterterrorism, and the Intelligence Environment]]. - **analytic system:** [[wiki/Predictive Policing|Predictive Policing]], [[wiki/Counterterrorism Predictive Graph|Counterterrorism Predictive Graph]], and [[wiki/Predictive Intelligence Loop|Predictive Intelligence Loop]]. - **scientific grammar:** [[wiki/Counterterrorism Collection - Oden and Predictive Science|Oden and Predictive Science]]. - **sensing and fusion:** [[wiki/Counterterrorism Collection - Fusion and Surveillance|Fusion and Surveillance]]. - **legal and civic boundary:** [[wiki/Counterterrorism Collection - Privacy and Civil Liberties|Privacy and Civil Liberties]]. - **anti-foreclosure principle:** [[wiki/Counterfactual Opportunity|Counterfactual Opportunity]], [[wiki/Opportunity Foreclosure|Opportunity Foreclosure]], and [[wiki/Right Not to Be Finalized by a Forecast|Right Not to Be Finalized by a Forecast]]. - **principal article:** [[articles/Peak Person and the Predicaments of Prediction|Peak Person and the Predicaments of Prediction]]. - **harm accounting and learning dynamics:** [[wiki/The Second Error Function|The Second Error Function]], [[wiki/Runaway Predictive Feedback Loop|Runaway Predictive Feedback Loop]], [[wiki/Observation-Induced Ground Truth|Observation-Induced Ground Truth]], [[wiki/Selective Labels|Selective Labels]], and [[research/Harms Incurred While Bringing Preventive Systems Online|Harms Incurred While Bringing Preventive Systems Online]]. ## Sources - [RAND — Predictive Policing: The Role of Crime Forecasting in Law Enforcement Operations, 2013](https://www.rand.org/content/dam/rand/pubs/research_reports/RR200/RR233/RAND_RR233.pdf) - [GAO — Artificial Intelligence: Emerging Opportunities, Challenges, and Implications, March 28, 2018](https://www.gao.gov/products/gao-18-142sp) - [GAO — Law Enforcement: DHS Could Better Address Bias Risk and Enhance Privacy Protections for Technologies Used in Public, December 3, 2024](https://www.gao.gov/products/gao-25-107302) **As of:** 2026-09-23