Educational companion dossier · Fact, interpretation, lived experience, clinical education, fiction, and mechanics are labeled separately. Scope & safety
PsychologicalWar.orgHistory before hype.
REAL-WORLD INTERPRETIVE

PREDICTIVE LAW ENFORCEMENT AND BEHAVIORAL THREAT ASSESSMENT

Predictive Policing in the United Kingdom

NDAS, HART, the Gangs Violence Matrix, the Violence Harm Assessment successor, and emerging AI-supported crime mapping—separated by function, deployment state, and consequence.

Read the working brief Choose a learning path
REAL-WORLD INTERPRETIVE

ANALYTICAL & SAFETY BOUNDARIES

Keep related capabilities, evidence, and authority states separate.

  • Retrieval, matching, identification, association, forecasting, prioritization, threat assessment, and coercive intervention are different functions.
  • Place-based and person-based systems have different targets, evidence requirements, and rights consequences.
  • A source describes a system or claim; it does not automatically establish current deployment, accuracy, effectiveness, legality, or fairness.
  • A human review step is meaningful only when the reviewer has enough time, evidence, authority, independence, and an effective way to reject or correct the output.

Analytical boundary: “Pre-crime” is an analytical label for anticipatory decision architectures, not a single technology or proof that future conduct can be known. Record retrieval is not prediction; identity matching is not prediction of conduct; association is not guilt; a map is not necessarily a forecast; a concern category is not a statistical probability; a score, tier, list, or alert is not lawful grounds for coercive action; formal human presence is not meaningful review without time, evidence, authority, and an effective intervention path. Exact source bytes are preserved, but external claims remain source claims until independently verified.

Simulation safety boundary: Educational and non-operational. No real-person scoring, surveillance deployment, list construction, targeting, coercive workflow, evasion, or system-defeat guidance.

Source basis: 1 exact owner-supplied source instance(s) preserved in the WIP.75 predictive-law-enforcement collection.

LEVEL 1

ORIENTATION

Why this matters

REAL-WORLD INTERPRETIVE

One-sentence brief

British examples show why machine learning, intelligence watchlists, data platforms, and geographic analysis require different validation and governance even when they all redirect police attention.

REAL-WORLD INTERPRETIVE

Three key points

  1. HART was a documented person-based risk model used in custody-related decision support.
  2. NDAS was a multi-use data and analytics capability; its serious-violence prediction model was not deployed.
  3. The Gangs Violence Matrix was a consequential watchlist/intelligence system, not a machine-learning forecast.
LEVEL 2

WORKING BRIEF

Evidence, context, and limits

REAL-WORLD INTERPRETIVE

HART: operational person-based prediction

HART used a random-forest model to classify future offending risk after arrest and informed eligibility for a deferred-prosecution program while retaining officer discretion. Public evidence of its later operational status is bounded by the report cutoff.

  • Document model target and features.
  • Separate advisory output from practical influence.
  • Evaluate postcode proxies and subgroup effects.
REAL-WORLD INTERPRETIVE

NDAS: platform and use cases, not one algorithm

NDAS supported distinct analytics use cases. A serious-violence model was rejected after low reported accuracy, while other work used language and network analysis to surface cases and patterns. Platform existence does not establish deployment of every contemplated model.

  • Evaluate each use case separately.
  • Preserve ethics-review outcomes.
  • Do not infer deployment from research.
REAL-WORLD INTERPRETIVE

Matrix and successor governance

The Gangs Violence Matrix ranked named people using police records and intelligence and could influence enforcement and partner treatment. It ended after legal and data-protection challenges; the successor introduced clearer thresholds and governance but remains a consequential watchlist requiring continuing scrutiny.

  • Track disproportionate impact.
  • Audit data sharing and retention.
  • Assess functional continuity after replacement.
LEVEL 3

COMPLETE DOSSIER

Limitations, game links, and review context

DISPUTED / MULTIPLE ACCOUNTS

Known limitations and gaps

  • The owner-supplied reports preserve their own evidence classifications, research cutoffs, and uncertainties; the site does not silently upgrade them.
  • Public information about thresholds, training data, model coefficients, operational modes, interventions, retention, and outcomes is frequently incomplete.
  • Program labels change. Functional continuity must be evaluated rather than inferred from a name alone.
  • Observed arrests, contacts, alerts, or declining crime after deployment do not by themselves establish predictive validity or causal effectiveness.
REAL-WORLD INTERPRETIVE

Related PsychologicalWar.org analysis

No related public page is required to understand this analysis.

LEVEL 4

RESEARCH EDITION

Sources, methods, and stable links

REAL-WORLD INTERPRETIVE

Linked reports

  • Predictive Policing in the United Kingdom: NDAS, HART, the Gangs Violence Matrix, and AI Crime MappingProtected research reference; no public report route in this release.
REAL-WORLD VERIFIED

Method and corrections

This page follows the public method for provenance, confidence, source independence, alternative accounts, limitations, review state, and visible correction.

Read the methodView corrections
NEXT

CONTINUE

Related learning

Predictive Law Enforcement and Behavioral Threat AssessmentBehavioral Threat Assessment or “Pre-Crime”?A source-bounded examination of BTAC, NTAC, structured professional judgment, targeted-violence prevention, low base rates, services, intervention, and civil-liberties boundaries.Predictive Law Enforcement and Behavioral Threat AssessmentBehavioral-Science Investigative UnitsHow specialized behavioral-science units support investigations, threat assessment, interviewing, research, training, consultation, and case management without becoming mind-reading engines.Predictive Law Enforcement and Behavioral Threat AssessmentDefinitions and Analytical BoundariesDistinguish retrieval, identity resolution, association, mapping, forecasting, person scoring, threat assessment, watchlisting, and coercive action.Predictive Law Enforcement and Behavioral Threat AssessmentEurope and the EU AI Act: Prohibited and High-Risk BoundariesPRECOBS, SKALA, KrimPro, the Netherlands Crime Anticipation System, and the distinction between prohibited individual crime prediction and regulated law-enforcement uses under the EU AI Act. Learning pathPredictive Law Enforcement: Evidence, Psychology, Governance, and RedressA sixteen-step path through terminology, behavioral threat assessment, behavioral-science units, person and place prediction, traveler risk, watchlisting, regional cases, technology lifecycle, evaluation, rights, and redress.
Page complete Predictive Policing in the United Kingdom Page label: REAL-WORLD INTERPRETIVE