--- id: ai-predictive-policing title: "AI for Predictive Policing: Crime Forecasting, Resource Allocation, and Bias Mitigation" schema_type: article category: ai language: en confidence: high last_verified: "2026-05-24" created_date: "2026-05-24" generation_method: ai_assisted ai_models: - claude-4.5-sonnet derived_from_human_seed: true conflict_of_interest: none_declared is_live_document: false data_period: static completeness: 0.85 atomic_facts: - id: af-ai-predictive-policing-1 statement: >- AI predictive policing (2015-2025): ML models predict crime hotspots (spatio-temporal forecasting -- where and when crimes are likely to occur) using historical crime data, temporal patterns (time of day, day of week, seasonality), environmental features (lighting, foot traffic, business density), and socioeconomic indicators. Early systems (PredPol, 2012) used self-exciting point processes (Hawkes process) and kernel density estimation. Modern approaches use GNNs modeling city grid connectivity and temporal transformers. source_title: PredPol (2012-2019) / HunchLab / ShotSpotter (2025) / IBM SPSS predictive analytics / Santa Cruz predictive policing pilot (2011-2020) source_url: https://arxiv.org/search/?query=predictive+policing+machine+learning+fairness confidence: high - id: af-ai-predictive-policing-2 statement: >- Predictive policing controversy: studies (RAND, 2018-2024) found that predictive policing can create feedback loops -- police are sent to AI-predicted areas, make more arrests (not necessarily more crime), feeding back into the model and reinforcing over-policing of historically targeted communities. Cities (Santa Cruz, Los Angeles, Chicago) ended predictive policing programs due to bias and civil liberties concerns. The 2024 consensus: AI should augment rather than direct policing -- providing decision support for resource allocation while maintaining human judgment and accountability. source_title: RAND predictive policing study (2018) / Santa Cruz predictive policing ban (2020) / ACLU predictive policing report / Los Angeles LASER program termination / AI Now Institute policing AI report source_url: https://arxiv.org/search/?query=algorithmic+bias+predictive+policing confidence: high primary_sources: - id: ps-ai-predictive-policing-1 title: "Machine Learning for Predictive Policing: Crime Forecasting, Spatio-Temporal Models, and the Feedback Loop Problem (2024-2025 Survey)" type: academic_paper year: 2025 institution: Journal of Quantitative Criminology / ACM FAccT / arXiv url: https://arxiv.org/search/?query=predictive+policing+machine+learning+fairness - id: ps-ai-predictive-policing-2 title: "Algorithmic Bias in Criminal Justice: Predictive Policing, Pretrial Risk Assessment, and the Perpetuation of Racial Disparities" type: academic_paper year: 2025 institution: ACM FAccT / Science / Harvard Law Review url: https://arxiv.org/search/?query=algorithmic+bias+predictive+policing known_gaps: - Fair crime prediction models demonstrating no racial or socioeconomic bias - Community-centered safety AI incorporating resident input and non-enforcement interventions disputed_statements: [] secondary_sources: - title: "A Critical Survey of AI in Predictive Policing: Methods, Bias, Ethics, and Regulatory Frameworks" type: survey_paper year: 2024 authors: - multiple institution: AI and Society (Springer) url: https://doi.org/10.1007/s00146-024-01982-6 - title: "Machine Learning and Predictive Policing: A Systematic Review of Effectiveness, Fairness, and Constitutional Concerns" type: survey_paper year: 2025 authors: - multiple institution: Journal of Criminal Justice (Elsevier) url: https://doi.org/10.1016/j.jcrimjus.2025.102185 - title: "Predictive Policing and Artificial Intelligence: A Routledge Handbook (Edited Volume)" type: textbook year: 2021 authors: - McDaniel, John L. M. - Pease, Ken G. institution: Routledge url: https://doi.org/10.4324/9780429265365 - title: "AI Now Institute Report: Regulating Predictive Policing — Bias, Accountability, and Community Impact" type: report year: 2024 authors: - AI Now Institute institution: AI Now Institute / NYU url: https://ainowinstitute.org/publication/predictive-policing updated: "2026-05-24" --- ## TL;DR AI predictive policing forecasts where crimes might occur -- but has become one of AI's most controversial applications. The feedback loop problem (AI sends police to areas -> arrests increase -> AI predicts more crime there) and racial bias concerns have led many cities to abandon predictive policing. The emerging approach: AI as decision support, not decision replacement. ## Core Explanation Crime prediction models: (1) Hotspot prediction -- spatio-temporal: grid cells (500ft x 500ft) x 4-hour windows. Features: historical crimes (near-repeat patterns -- burglaries cluster in time and space), temporal cycles, urban features (bars, schools, transit stops). Models: Hawkes processes (self-exciting point processes), kernel density estimation, GNNs (city grids as graphs), ConvLSTMs; (2) Individual risk -- predicting recidivism (COMPAS, 1998-present). Controversial: ProPublica (2016) found COMPAS biased against Black defendants (higher false positive rate); (3) Victimization risk -- predicting repeat victimization (domestic violence, stalking). ## Detailed Analysis PredPol (2012): used ETAS model (epidemic-type aftershock sequence -- borrowed from seismology) predicting near-repeat patterns. Displayed 500ft x 500ft boxes for patrol officers. Cities deployed then abandoned (Santa Cruz 2020, LA 2019). The feedback loop: Prediction box -> more patrol -> more minor citations for things that would otherwise go unnoticed -> reported as "crime" -> reinforces prediction. This biases models toward areas of existing police presence. RAND (2018): evaluated predictive policing pilots. Found no conclusive evidence of crime reduction in controlled studies. ShotSpotter (2025): acoustic gunshot detection system. ML classifies gunshots vs fireworks/backfires, triangulates location, alerts police with <60 second latency. Deployed in 130+ US cities. Better accepted than predictive policing as it detects actual events rather than predicting future events. Current consensus (2023-2025): AI for policing should focus on forensic investigation support, crime analysis (post-hoc pattern recognition), and resource allocation optimization -- not individual-level prediction.