ShieldPST.ai · Technology Explainer Series

Predictive Policing

How statistical models and artificial intelligence use historical crime, location, incident, person, and environmental data to forecast where crime may occur or who may face elevated risk—and what agencies should understand about data quality, feedback loops, discriminatory effects, validation, transparency, explainability, officer discretion, civil rights, resource allocation, and the critical distinction between prediction and individualized legal suspicion.

Input Historical + Current Data
Output Forecast · Risk · Priority
Core Rule Prediction ≠ Reasonable Suspicion

What this explainer does

Predictive policing uses statistical analysis, machine learning, crime analysis, geographic information, or related analytical methods to forecast where crime may occur, identify people or situations believed to present elevated risk, or guide allocation of law-enforcement resources.

The concept ranges from relatively familiar crime forecasting— such as identifying locations where burglaries historically cluster— to substantially more controversial person-based systems that assign individuals risk scores or identify people believed more likely to become involved in violence.

The technology does not predict the future with certainty. It identifies statistical patterns in data and generates estimates about risk.

Prediction is not suspicion

A model may conclude that a neighborhood faces elevated burglary risk or that a person shares characteristics associated with prior violence.

That statistical prediction is not itself proof that any particular person has committed, is committing, or is about to commit a crime.

1. Overview

Predictive policing applies historical and current information to statistical or computational models designed to estimate future crime-related risk.

The concept is not one technology. A simple statistical forecast of burglaries by block is fundamentally different from an algorithm that assigns individuals a violence-risk score.

Some systems are primarily resource-allocation tools. Others may influence surveillance, investigative attention, field contacts, intervention programs, or decisions concerning particular people.

Central Concept Predictive policing is best understood as: past data → model → probability or risk estimate → law-enforcement decision . The final step—the decision made because of the prediction—is often more legally and ethically significant than the prediction itself.

2. Four Major Categories of Predictive Policing

Category Question Example
Place-Based Prediction Where is crime more likely to occur? Forecasting burglary risk within small geographic areas during specified periods.
Person-Based Prediction Who may be at elevated risk of offending or violence? Risk scores based on criminal history, associations, victimization, or other variables.
Victimization Prediction Who or what may be at elevated risk of becoming a victim? Identifying locations or people at risk of repeat violence or burglary.
Crime / Offender Analysis What patterns may connect offenses or help identify a perpetrator? Using characteristics of crime scenes to identify likely relationships among incidents.
Policy Rule An agency should not adopt one vague “predictive policing policy” and assume it adequately governs every category. Person-based risk scoring requires substantially different safeguards from place-based patrol forecasting.

3. How Predictive Policing Works

1. Collect Data Crime, location, person, environmental, or other records
2. Select Variables Determine which factors will influence the model
3. Train / Calculate Statistical model identifies patterns or relationships
4. Forecast System generates risk, location, priority, or probability output
5. Police Response Agency changes patrol, analysis, outreach, or investigation
6. New Data Police activity creates additional records that may return to the model
System Principle Predictive policing is a socio-technical system. Data, algorithms, analysts, officers, policy, deployment decisions, community behavior, and police-generated records all influence outcomes.

4. Place-Based Crime Forecasting

Place-based systems predict locations where particular types of crime may be more likely during a future period.

They build on longstanding criminological research showing that crime is not distributed evenly across geography and that relatively small locations can account for disproportionate numbers of reported incidents.

Hotspots

Identify geographic concentrations of prior crime.

Near-Repeat Patterns

Model whether a recent crime increases short-term risk nearby.

Risk Terrain

Analyze environmental features associated statistically with crime risk.

Place-Based Rule A predicted high-risk area may justify resource allocation or heightened awareness. It does not automatically make every person within the area reasonably suspicious.

5. Person-Based Predictive Policing

Person-based systems attempt to identify individuals believed to face elevated risk of committing violence, becoming victims of violence, or otherwise becoming involved in specified criminal activity.

Inputs might include prior arrests, convictions, victimization, age, co-arrests, gang information, associations, geographic exposure, or other variables depending on the system.

Risk ≠ Conduct

A person can share characteristics associated with elevated risk without engaging in criminal activity.

Association

Social or family relationships can become algorithmic variables even when the relationship itself is innocent.

Historical Records

Prior contacts or arrests may influence risk even when they did not result in conviction.

Heightened-Risk Use Person-based prediction presents substantially greater civil-rights, due-process, privacy, and fairness concerns than deciding where to deploy patrol resources.

6. Predicting Victimization

Predictive analytics need not be used only to identify suspected offenders. Some systems attempt to identify people or places facing an elevated risk of victimization.

That information may support outreach, violence interruption, victim services, burglary prevention, directed patrol, or other preventive strategies.

Purpose Matters A risk score used to offer voluntary assistance raises different concerns from the same score being used to justify surveillance, stops, searches, or enforcement attention.

7. Resource Allocation

Predictive systems can help agencies determine where limited patrol, investigative, or prevention resources might have the greatest effect.

That is one of the least controversial uses when the model operates at an aggregate geographic level and does not itself trigger coercive action.

Patrol Deployment

Allocate officers to locations associated with elevated forecast risk.

Problem Solving

Focus analysis on recurring locations or environmental factors.

Prevention

Direct lighting, outreach, environmental design, or other preventive resources toward identified risks.

8. The Model Is Built from Data

Predictive systems do not discover crime independently. They learn patterns from the data they receive.

Data Source What It May Represent Potential Limitation
Reported Crime Incidents reported by victims, witnesses, or police Not every crime is reported uniformly across communities
Arrests People arrested for suspected offenses Reflects enforcement decisions as well as underlying crime
Calls for Service Requests for police assistance Call patterns reflect reporting behavior and community-police relationships
Field Contacts Police encounters and observations Frequency can reflect prior deployment decisions
Geographic Data Locations, businesses, transportation, land use Correlations may not establish causation
Person-Level Records Criminal history, associations, victimization, age, addresses Errors or historical inequities can become model inputs

9. “Dirty Data” and Historical Policing Patterns

Historical law-enforcement data are not neutral measurements of every crime that occurred in a jurisdiction.

They reflect what was reported, where officers were deployed, what conduct officers observed, which complaints were investigated, whom officers stopped, whom prosecutors charged, and countless other institutional and community factors.

Data Principle If historical data contain error, selective enforcement, inconsistent reporting, outdated intelligence, or disparate patterns, the model can reproduce those patterns at scale.

10. The Feedback-Loop Problem

Predictive policing can create a self-reinforcing loop.

Historical Data Area shows high recorded crime or police activity
Model Prediction System labels area elevated risk
More Police Agency deploys additional officers there
More Observation Officers detect additional offenses or contacts
More Records Additional arrests, stops, reports, and field contacts are generated
Model Reinforced New police-generated data confirms the original risk pattern
Feedback-Loop Question Is the model predicting where crime occurs, or partly predicting where police previously looked for crime? Agencies should design validation capable of separating those effects.

11. Correlation Is Not Causation

Predictive models identify statistical relationships.

A model may find that certain locations, histories, relationships, or other factors correlate with future events without proving that those variables cause crime.

Analytical Rule A variable can be useful for prediction while still being a poor explanation of why the event occurs. Prediction and causation are different analytical tasks.

12. What Does “Accurate” Mean?

Predictive-policing claims should be evaluated against defined metrics.

Precision

Of events predicted, how many actually occurred within the defined parameters?

Recall

Of events that occurred, how many were captured by the prediction?

False Positive

How often does the system classify a location or person as high risk when the predicted event does not occur?

False Negative

How often does crime occur outside the model's predicted risk?

Calibration

Does a stated risk level correspond with observed outcomes?

Operational Effect

Does using the prediction actually reduce crime, improve response, or produce another desired result?

Accuracy Warning A model can predict historical patterns accurately and still fail to improve policing outcomes. Predictive accuracy and policy effectiveness are different questions.

13. Validation

Predictive systems should be tested before deployment and monitored after deployment.

1. Define Objective Identify the precise decision the model is meant to improve
2. Validate Data Review completeness, accuracy, representativeness, and legality
3. Test Model Evaluate performance using appropriate historical and prospective data
4. Test Groups Examine geographic and demographic differences in error
5. Evaluate Impact Measure what officers actually do because of predictions
6. Revalidate Repeat after major model, data, policy, or environmental change
Field Validation A model should not be considered validated merely because the vendor can demonstrate that it performed well on another city's historical data.

14. Bias and Disparate Effects

Bias can enter predictive policing through multiple stages.

Historical Bias

Prior policing patterns may shape the data used for training.

Sampling Bias

Available records may not represent all crime or all communities equally.

Measurement Bias

Variables may measure police activity rather than underlying criminal activity.

Model Bias

Statistical choices can distribute errors unevenly.

Deployment Bias

Police may respond differently to equivalent predictions in different areas.

Feedback Bias

Model-directed enforcement can create new data reinforcing existing predictions.

15. Proxy Variables

A model does not need to explicitly use race, religion, or another protected characteristic for those characteristics to be indirectly reflected.

Geography, housing, income, school, arrest history, associations, transportation patterns, and other variables can correlate with sensitive characteristics.

Proxy Warning Removing a protected characteristic from the input file does not automatically make a model neutral. Agencies should examine how variables operate together and who bears the system's errors.

16. Explainability and Transparency

Agencies should understand enough about a predictive system to explain why its outputs are generated and how those outputs are intended to be used.

Inputs

What information goes into the model?

Weights

What factors materially influence the prediction?

Output

What exactly does a score or hotspot mean?

Thresholds

What causes a person or location to be categorized as elevated risk?

Updates

How frequently does the model change?

Limitations

Under what circumstances should officers not rely on the output?

17. What Officers May Do with a Prediction

The constitutional and policy consequences depend heavily on how predictions affect officer conduct.

Action Prediction's Proper Role
Patrol Allocation May help determine where officers spend discretionary patrol time.
Crime Prevention May support environmental, outreach, or noncoercive interventions.
Investigative Prioritization May help analysts decide where to look for additional evidence.
Investigative Stop Prediction alone does not establish individualized reasonable suspicion.
Search Model output does not independently create search authority.
Arrest Prediction alone does not establish probable cause.
Use of Force Force remains governed by circumstances confronting officers, not an algorithmic risk score.

18. Fourth Amendment Considerations

Predictive analysis itself may involve information already lawfully held by government, but enforcement actions triggered by the prediction remain subject to ordinary Fourth Amendment standards.

Officers must still establish reasonable suspicion, probable cause, warrant authority, consent, exigency, or another recognized legal basis appropriate to the action taken.

Key Distinction A model can inform: where an officer looks. It does not automatically answer: what the officer is constitutionally allowed to do when someone is found there.

19. First Amendment Concerns

Person-based forecasting can raise significant First Amendment issues if models use associations, social networks, protest activity, social-media behavior, group membership, religion, journalism, or political activity.

Protected-Activity Rule Protected speech or association should not become a proxy for criminal risk. Agencies should require heightened review whenever predictive models rely directly or indirectly on constitutionally protected activity.

20. Due Process and the Right to Challenge Risk

Person-based systems can create difficult fairness questions when an individual does not know that government has assigned a risk score or placed the person on a high-risk list.

Errors may arise from mistaken identity, outdated criminal histories, incorrect gang information, stale addresses, inaccurate associations, duplicate records, or flawed data matching.

Notice

Does the person ever learn that the classification exists?

Correction

Is there a mechanism for correcting inaccurate information?

Consequences

What government action occurs because of the score?

21. Transparency

Meaningful transparency does not necessarily require publication of every technical detail or security-sensitive feature.

But policymakers and communities should generally be able to understand what problem the system addresses, what category of data it uses, what type of prediction it makes, what officers may do with the output, and how the agency evaluates error and disparate effects.

Transparency Principle Agencies should be able to explain: what the system predicts, what it does not predict, what data it uses, what decisions it influences, and what safeguards constrain its use.

22. Procurement and Vendor Questions

Issue Agency Question
Model What statistical or machine-learning methodology generates predictions?
Training Data What data were used to develop or calibrate the model?
Local Data Which agency records become inputs?
Performance What independent validation supports the vendor's claims?
Bias Testing What geographic and demographic analyses have been conducted?
Explainability Can the agency understand why particular outputs are generated?
Model Updates Can the vendor materially alter the model without notice?
Audit Are predictions, user actions, data changes, and model versions preserved?
Data Ownership Who owns agency data, derived features, risk scores, and model outputs?
Termination What happens to agency data and predictions when the contract ends?

23. Agency Governance Framework

Define the Use

Specify whether the system predicts places, people, victimization, crime patterns, or resource needs.

Purpose Limitation

Define what decisions predictions may and may not influence.

Data Governance

Document sources, quality, errors, retention, corrections, and legal authority.

Bias Review

Assess whether predictions or resulting enforcement burdens fall disproportionately on particular communities.

Feedback Monitoring

Determine whether police deployment is generating data that reinforces the model.

Independent Validation

Test accuracy and operational effect rather than accepting vendor claims alone.

Legal Limits

State expressly that predictive outputs do not replace reasonable suspicion or probable cause.

Human Review

Require meaningful professional judgment before consequential use of algorithmic recommendations.

Person-Based Safeguards

Apply heightened approval, review, correction, and audit procedures to individual risk scoring.

Transparency

Explain the system's purpose, broad methodology, data sources, and safeguards where legally appropriate.

Audit Logs

Preserve model versions, predictions, searches, user actions, and consequential decisions.

Sunset / Reauthorization

Periodically require affirmative review of whether demonstrated benefits justify continued use.

24. Questions Every Agency Should Answer

What exactly does the predictive system forecast?
Is the system place-based, person-based, victim-based, or another type?
What operational problem is the agency trying to solve?
What decisions may the model influence?
What decisions is the model prohibited from making?
What data sources are used?
Does the system use arrests?
Does it use convictions?
Does it use calls for service?
Does it use field interviews or stops?
Does it use gang information?
Does it use associations or social-network information?
Does it use social-media activity?
Does it use location data?
Does it use commercially acquired information?
How are inaccurate source records corrected?
How old may input data be?
Have historical policing practices been evaluated as a source of bias?
Does the agency distinguish reported crime from police-generated activity?
Has the model been independently validated?
Was validation performed using local agency data?
What false-positive rate does the system produce?
What false-negative rate does it produce?
How is calibration measured?
Does performance vary geographically?
Does performance vary demographically?
What proxy variables may correlate with protected characteristics?
Does the agency test for feedback loops?
Are predictions influenced by prior police deployment?
Can the model distinguish underlying crime from enforcement activity?
Can analysts explain why a location receives a high-risk score?
Can analysts explain why a person receives a high-risk score?
Are model outputs preserved?
Is the model version used for each prediction preserved?
Are users told how predictions should and should not be interpreted?
Does policy state that a risk score is not reasonable suspicion?
Does policy state that a prediction is not probable cause?
May officers stop a person solely because the individual appears on a risk list?
May officers search a person solely because of predictive output?
May a prediction be included in a warrant affidavit?
If so, is the underlying methodology accurately disclosed?
How are First Amendment activities excluded or safeguarded?
Can political or protest activity influence a risk score?
Can lawful associations influence a risk score?
Does an individual have a method to correct inaccurate information?
How long do person-based risk classifications remain active?
Are person-based lists periodically reviewed?
What criteria remove someone from a risk list?
Are predictions shared with other agencies?
Are downstream agencies told the limitations of the prediction?
What independent evidence demonstrates that use of the model improves public safety?
Does the agency measure collateral consequences such as increased stops or community distrust?
Who audits model use?
How often is the model revalidated?
What event triggers suspension of the system?
Does the program have a sunset date or mandatory reauthorization requirement?

25. Where Predictive Policing Is Going

Real-Time Prediction

Models may increasingly update continuously using live CAD, ALPR, video, gunshot alerts, and other sensor information.

RTCC Integration

Predictive systems may become embedded directly into real-time operational platforms.

Multimodal AI

Models may combine text, video, audio, location, vehicles, networks, and historical records.

Automated Recommendations

AI may move from predicting risk to recommending specific police responses.

Generative Analysis

Language models may summarize predicted risks and explain patterns to officers and commanders.

Cross-System Risk Profiles

Agencies may increasingly combine crime records, intelligence, sensors, vehicles, social networks, and commercial data.

Future-Looking Principle The next generation of predictive policing may not look like a map covered with red boxes. It may look like an RTCC system that continuously analyzes people + places + vehicles + cameras + ALPR + gunshot alerts + social networks + police records and then recommends where police should go and whom they should examine. At that point, agencies will need to govern not merely prediction but automated government decision support.

26. Key Terms

Predictive Policing Use of statistical, analytical, or machine-learning methods to forecast crime-related risk and inform law-enforcement decisions.
Place-Based Prediction Forecasting geographic areas where crime is expected to be more likely.
Person-Based Prediction Assigning or estimating risk concerning particular individuals.
Hotspot Geographic area exhibiting elevated concentration or forecast risk of crime.
Risk Terrain Modeling Spatial method examining environmental features associated with crime risk.
Near-Repeat Concept that certain crimes may increase short-term risk of similar offenses nearby.
Risk Score Numerical or categorical estimate of predicted risk produced by a model.
Training Data Data used to develop or fit a predictive model.
Feature Variable or input used by a statistical or machine-learning model.
Proxy Variable Variable that may indirectly represent or correlate with another characteristic.
Feedback Loop Cycle in which model-directed policing creates new data that reinforces future predictions.
False Positive Prediction of elevated risk when the predicted event does not occur.
False Negative Failure to identify an event or risk that later occurs.
Calibration Degree to which predicted probabilities correspond with observed outcomes.
Model Drift Degradation or change in model performance as conditions or data change over time.
Explainability Ability to understand or communicate meaningful reasons for a model's output.
Algorithmic Bias Systematic error or unequal impact associated with model design, data, deployment, or use.
Automation Bias Human tendency to place excessive trust in machine-generated predictions.
Human-in-the-Loop Workflow requiring meaningful human judgment before consequential decisions.
Algorithmic Decision Support Systems that provide predictions, classifications, recommendations, or other outputs intended to influence human decisions.

27. Related ShieldPST.ai Resources

Real-Time Crime Centers

Integrated data, sensors, cameras, analysts, and AI-assisted operational decision support.

Open explainer →
Facial Recognition Technology

Algorithmic identification, candidate lists, accuracy, human review, bias, and governance.

Open explainer →
Social Media & OSINT

Public-source intelligence, online identity, automated analysis, and First Amendment considerations.

Open explainer →
ALPR & Vehicle Intelligence

Vehicle-location data, automated alerts, historical searches, and investigative analytics.

Open resource →
AI for Criminal Investigations

AI-assisted evidence analysis, lead development, human verification, and investigative safeguards.

Open resource →
Technology Explainers

Return to the Shield Technology Reference Library.

Browse explainers →

28. Selected Primary and Authoritative Sources

National Institute of Justice — Overview of Predictive Policing
NIJ overview describing predictive policing as the use of advanced analysis and intervention models to forecast where or under what circumstances crime may occur and guide police resources.
Review NIJ overview
National Institute of Justice — Predictive Policing: Understanding and Applying Analytical Techniques
NIJ-sponsored research describing major predictive-policing categories, including place, offender, victim, and perpetrator prediction approaches.
Review research brief
National Institute of Justice — From Crime Mapping to Crime Forecasting: The Evolution of Place-Based Policing
NIJ discussion of the development of place-based crime analysis, forecasting, research limitations, and evidence concerning proactive policing.
Review NIJ article
U.S. Department of Justice — Artificial Intelligence and Criminal Justice
DOJ report examining AI applications across criminal justice, including predictive policing, facial recognition, risk assessment, fairness, privacy, transparency, and responsible-use considerations.
Review DOJ report
National Institute of Standards and Technology — AI Risk Management Framework
NIST framework addressing AI governance, validity and reliability, transparency, accountability, privacy, security, and harmful bias.
Review NIST AI RMF
NIST Special Publication 1270 — Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
NIST guidance addressing how bias can arise from data, algorithms, human decision-making, societal structures, and deployment contexts.
Review NIST publication
NIST — Field Testing of Law Enforcement AI Tools
NIST recommendations emphasizing the importance of testing, auditing, and field evaluation when law-enforcement agencies deploy AI systems producing predictive or analytical outputs.
Review NIST recommendations
RAND Corporation — Predictive Policing: The Role of Crime Forecasting in Law Enforcement Operations
Foundational research guide examining predictive methods, data, policing interventions, implementation, and evaluation.
Review research

29. Key Takeaways

Bottom Line
  1. Predictive policing is not one technology. It includes place-based forecasting, person-based risk prediction, victimization prediction, resource allocation, and other analytical approaches.
  2. Predictive systems use historical or current data to estimate future risk; they do not literally know what will happen.
  3. Place-based forecasting generally presents different legal and governance risks from assigning risk to named individuals.
  4. Historical police data are not neutral measurements of all criminal activity. They reflect reporting patterns, police deployment, enforcement decisions, recordkeeping, and community behavior.
  5. Errors and historical inequities contained in source data can be reproduced or amplified by an algorithm.
  6. Predictive policing can create feedback loops when model-directed enforcement generates new police records that reinforce future predictions.
  7. Agencies should distinguish whether a model predicts crime or partly predicts where police have historically looked for crime.
  8. Correlation does not establish causation. Variables associated with crime risk do not necessarily explain why crime occurs.
  9. A model's accuracy should be measured through defined metrics such as false positives, false negatives, precision, recall, calibration, geographic performance, and demographic performance.
  10. Predictive accuracy does not by itself establish that deploying the system improves public safety.
  11. Agencies should independently validate predictive systems using conditions and data relevant to their actual jurisdiction.
  12. Removing race or another protected characteristic from the input does not automatically eliminate bias because other variables may function as proxies.
  13. Agencies should understand enough about a system's inputs, outputs, thresholds, methodology, and limitations to govern it responsibly.
  14. Most importantly, a predictive risk score does not constitute individualized reasonable suspicion or probable cause.
  15. Predictive information may help determine where officers deploy or where investigators look for evidence, but enforcement action still requires the independent legal basis applicable to that action.
  16. Person-based systems require heightened safeguards concerning First Amendment activity, inaccurate records, associations, notice, correction, retention, dissemination, and due process.
  17. Vendor contracts should address model methodology, data sources, validation, bias testing, model changes, auditability, data ownership, and termination.
  18. Agencies should evaluate the entire system: data → algorithm → prediction → officer decision → police activity → new data.
  19. The next generation of predictive policing will likely become deeply integrated with RTCCs, sensor networks, video analytics, ALPR, artificial intelligence, and automated decision-support systems.
  20. The central future governance question will therefore be not merely “Can the computer predict risk?” but “What may government do to a person because the computer says risk is elevated?”

ShieldPST.ai · Technology Explainer Series

This explainer is provided for training and general informational purposes. It is not legal advice and does not replace current review of controlling federal and state law, state constitutional provisions, First Amendment requirements, Fourth Amendment requirements, civil-rights law, anti-discrimination law, criminal-intelligence requirements, public-records law, discovery obligations, agency policy, collective-bargaining obligations where applicable, vendor contracts, model documentation, validation studies, prosecutorial guidance, or consultation with agency counsel. Predictive analytics, artificial intelligence, algorithmic decision-support technology, civil-rights standards, and governing law remain dynamic.

© 2026 Shield Public Safety Training. All rights reserved. · Reviewed August 10, 2026.