Is AI-powered Predictive Policing Good, Great or Ugly?
Anyck Turgeon
Digital Transformation, Cyber, Risk & Strategic Leader | GenAI | Agile, Finance & Transformational Coach | Board Member
Law Enforcement Transformation with Data-Driven Insights
No one wants to be a victim of robbery, exploitation, extortion, defamation, or abuse. Yet, few are aware that predictive policing has already been in use for decades globally in order to assess and prevent crimes pro-actively.?
As we evolve, distinct systems and services are being automated, integrated and orchestrated by different flavors of artificial intelligence (AI.) So we need to ask ourselves:
?
What is Predictive Policing?
To ensure alignment, let's define predictive policing as the use of data analysis, large datasets, and machine learning algorithms to anticipate potential criminal activities. By analyzing historical crime data, social media activity, weather patterns, and other relevant factors, these systems can identify patterns and trends that help law enforcement agencies allocate resources more effectively. This data-driven approach aims to improve public safety by predicting where and when crimes are likely to occur, thereby enabling preventive and corrective measures. [(Source: RAND Corporation)](https://www.rand.org/pubs/research_reports/RR1488.html).
?
Predictive Policing Globally Today: 8 International Case Studies
Predictive policing is transforming law enforcement worldwide by leveraging data analytics to forecast criminal activities. Below are some examples of benefits and challenges for utilizing predictive policing as experienced by different countries:
1. United States
?? - Benefits: In cities like Los Angeles, predictive policing has led to a 12% reduction in property crimes. The Los Angeles Police Department's (LAPD) use of PredPol software has helped allocate resources more efficiently, resulting in a notable decrease in burglaries and vehicle thefts.
?? - Challenges: Despite its success, the LAPD has faced criticism for potential racial bias in its algorithms. Studies have shown that minority communities were disproportionately targeted, raising ethical concerns based on historical data utilized. [(Source: Saunders et al., 2016)](https://www.rand.org/pubs/research_reports/RR1488.html) Should data analysts should re-calibrate their source data sets to reflect proportionality of current population demographics? .
?
2. United Kingdom
?? - Benefits: Kent Police reported a 9% reduction in burglaries after implementing predictive policing software. The system helped identify crime hotspots, allowing for better resource allocation and crime prevention.
?? - Challenges: The UK experience highlighted the need for continuous monitoring to ensure ethical use. Concerns about privacy and data protection were significant, necessitating robust legal frameworks. [(Source: Office of Justice Programs)](https://nij.ojp.gov/library/publications/predictive-policing-guide-using-data-and-technology-forecast-crime). As AI regulations, frameworks and guidelines are being enacted, how will agencies build AI ethical standards that enable them to still extract benefits from predictive policing and avoid penalties, reprisals or repatriation?
?
3. Germany
?? - Benefits: In cities like Munich, predictive policing has resulted in a 10% decrease in residential burglaries. Data-driven insights have enabled more targeted patrols and efficient use of police resources.
?? - Challenges: The German approach faced challenges related to public acceptance and transparency. Ensuring that the public understood and trusted the technology was crucial for its success. [(Source: Springer)](https://link.springer.com/article/10.1007/s10610-018-9384-7) As our world keeps on getting more complex, will humans be able to keep up with explainable AI or will it become too complex as we transition to different layers of AI maturity? .
?
4. Japan
?? - Benefits: The Tokyo Metropolitan Police implemented predictive policing to address rising cybercrimes, leading to a 15% reduction in reported cyber incidents. The system's ability to predict and prevent cybercrimes proved highly effective.
?? - Challenges: Japan’s focus on cybercrime highlighted the need for specialized training for law enforcement officers to ensure they could effectively use the technology. [(Source: Liberties.eu)](https://www.liberties.eu/en/news/japan-predictive-policing/16531). As technology evolves with new offerings continuously, what percentile of time will cyber-defenders have to spend on continuous learning? Already, many cyber-professionals are spending a large portion of their spare time continuously upgrading their skills and amount of certifications. As the pace of attacks keep on increasing, until when will agency representatives be able to keep up?
?
5. South Africa
?? - Benefits: In Johannesburg, predictive policing helped reduce violent crimes by 8%. The system's ability to identify high-risk areas allowed for more proactive policing and crime prevention.
?? - Challenges: South Africa's implementation faced issues related to data quality and availability. Ensuring accurate and comprehensive data was essential for the system's effectiveness. [(Source: MDPI)](https://www.mdpi.com/2071-1050/11/21/6065). As several parties are required (e.g. data analyst, data scientist, AI architect, Big Data engineer, AI ethicist, etc.), when will the costs of producing reliable, near real-time predictive policing bypass its benefits?
?
6. Brazil
?? - Benefits: S?o Paulo's use of predictive policing led to a 7% decrease in homicides. The system's ability to forecast violent crimes allowed for better resource allocation and crime prevention strategies.
?? - Challenges: Brazil's experience highlighted the need to address socio-economic factors contributing to crime, as predictive policing alone was not sufficient to tackle the root causes of criminal activities. [(Source: RAND Corporation)] (https://www.rand.org/pubs/research_reports/RR231.html). Which other AI solutions do we need to develop to address the root causes of criminal activities detected in predictive policing and how can we build solutions that can make our world a better place?
?
7. India
?? - Benefits: In New Delhi, predictive policing initiatives resulted in a 6% reduction in street crimes. The system's ability to predict and prevent crimes in high-risk areas proved beneficial.
?? - Challenges: India’s implementation faced challenges related to infrastructure and technology adoption. Ensuring that law enforcement agencies had the necessary tools and training was crucial. [(Source: Office of Justice Programs)](https://nij.ojp.gov/library/publications/predictive-policing-guide-using-data-and-technology-forecast-crime). As India has a large tech-savvy population, how can we integrate more deep learning models to facilitate better predictive policing?
?
8. Australia
?? - Benefits: Melbourne's predictive policing efforts led to a 5% decrease in property crimes. The system's ability to identify crime hotspots allowed for more efficient use of police resources.
?? - Challenges: Australia's experience highlighted the importance of addressing ethical concerns related to data privacy and algorithmic transparency. Ensuring public trust and accountability was essential. [(Source: MDPI)](https://www.mdpi.com/2071-1050/11/21/6065). As privacy, security and AI regulatory requirements are evolving in regional, national and international specificities, how many AI GPRC & ethics professionals will we need in order to keep on trusting predictive policing, given its multiplying growth in complexity?
?Predictive policing offers significant benefits in terms of crime prevention and resource optimization. However, it also presents challenges related to bias, transparency, and privacy. By addressing these issues, predictive policing can become a valuable tool in enhancing public safety worldwide.
?
Which Technologies Have Been Mostly Used for Predictive Policing?
Predictive policing involves several advanced technologies, including:
1. Data Analytics: Analyzing large datasets from various sources, including crime reports, social media, weather patterns, and demographic information, to identify patterns and trends that could indicate future criminal activity.
2. Machine Learning: Employing algorithms that can learn from data and improve their predictions over time without being explicitly programmed, helping refine the accuracy of predictive models by identifying complex patterns.
3. Artificial Intelligence (AI): AI systems process vast amounts of data and make decisions or predictions based on that data. In predictive policing, AI might be used to identify anomalies or predict where crimes are likely to occur based on historical data.
4. Geospatial Analysis: Using geographic information systems (GIS) to map crime data and visualize where crimes are most likely to occur, helping police departments identify hotspots and deploy resources more efficiently.
5. Social Network Analysis: Analyzing relationships and interactions within social networks to identify key individuals or groups involved in criminal activities, aiding in understanding criminal networks and predicting potential criminal actions.
6. Natural Language Processing (NLP): Analyzing text data from sources like social media, police reports, and tip lines to detect potential threats or identify patterns in criminal behavior.
7. Big Data: Leveraging large volumes of data from various sources to find patterns and correlations that could predict criminal activity, including both structured data like crime statistics and unstructured data like social media posts.
8. Surveillance Technologies: Using cameras, drones, and other monitoring equipment to gather real-time data that can be analyzed for predictive purposes, such as facial recognition and license plate recognition.
9. Predictive Modeling Software: Software specifically designed to analyze crime data and produce forecasts or risk assessments, including tools like PredPol and HunchLab.
10. Sentiment Analysis: Evaluating public sentiment through social media, news articles, and other public communications to understand community tensions or unrest that could lead to crime.
11. Internet of Things (IoT): Using interconnected devices, like smart streetlights or security cameras, to collect and analyze data in real time, helping predict and prevent crime.
These technologies work together to help law enforcement agencies predict crime more accurately and make informed decisions on resource allocation and crime prevention strategies.
?
How Is Predictive Policing Used Today? 12 Case Studies
Here are 12 use cases of predictive policing, each with associated benefits, challenges, examples, and measures to avoid issues. These use cases are supported by referenced statistics and data:
1. Crime Hotspot Prediction
?? - Benefit: Predictive policing algorithms can identify crime hotspots, allowing law enforcement to allocate resources more effectively and reduce crime rates in those areas.
?? - Challenge: Crime data used to identify hotspots can be biased or incomplete, potentially leading to over-policing of certain neighborhoods.
?? - Example: A study in Chicago found that using predictive policing software to identify hotspots led to a 7.4% reduction in crime in targeted areas. However, this also resulted in increased police presence in predominantly minority neighborhoods, raising concerns about bias. [(Source: Saunders et al., 2016)](https://www.rand.org/pubs/research_reports/RR1488.html).
?? - Measures to Avoid:
???? - Regularly update and audit crime data to ensure accuracy.
???? - Use additional data sources, such as community reports, to provide a more balanced view of crime patterns.
???? - Implement oversight mechanisms to monitor the effects of predictive policing on different communities.
?
2. Predicting Burglaries Based on Detected Patterns
?? - Benefit: Algorithms can analyze past burglary data to predict when and where future burglaries are likely to occur, allowing police to deploy resources proactively.
?? - Challenge: Reliance on historical data may lead to false positives, with resources being allocated to areas where burglaries do not actually occur.
?? - Example: In Santa Cruz, California, predictive policing helped reduce burglaries by 19% within the first six months of implementation. However, the system also produced false positives that led to unnecessary patrolling in some areas. [(Source: Perry et al., 2013)] (https://www.rand.org/pubs/research_reports/RR107.html).
?? - Measures to Avoid:
???? - Validate predictions with on-the-ground intelligence and community input.
???? - Continuously refine algorithms based on new data and feedback from officers.
???? - Use predictive tools as a supplement to, rather than a replacement for, traditional policing methods.
3. Forecasting Illegal Gang Activities
?? - Benefit: Predictive analytics can identify patterns of gang activity, helping police prevent gang-related crimes before they happen.
?? - Challenge: Focusing on gang activity can lead to profiling and discrimination against certain groups, particularly young men of color.
?? - Example: A predictive policing program in Los Angeles used social network analysis to forecast gang violence, resulting in a 21% decrease in gang-related crimes. However, this also led to increased surveillance and tensions in certain communities. [(Source: Uchida et al., 2014)](https://nij.ojp.gov/library/publications/predictive-policing-guide-using-data-and-technology-forecast-crime).
?? - Measures to Avoid:
???? - Engage community leaders and organizations in discussions about predictive policing strategies to build trust.
???? - Ensure algorithms are designed to avoid profiling based on race, ethnicity, or other protected characteristics.
???? - Regularly review and adjust practices based on community feedback and crime data.
?
4. Reducing Violence
?? - Benefit: Predictive policing can help identify areas at high risk for gun violence, enabling targeted interventions that reduce shootings and save lives.
?? - Challenge: Over-policing in high-risk areas can strain community relations and lead to accusations of harassment or discrimination.
?? - Example: In Chicago, predictive policing focused on reducing gun violence led to a 23% reduction in shootings in targeted areas. However, it also resulted in increased stop-and-frisk incidents, which disproportionately affected minority communities. [(Source: Police Executive Research Forum, 2016)](https://www.policeforum.org/assets/predictivepolicing.pdf).
?? - Measures to Avoid:
???? - Balance predictive policing with community policing efforts to build trust and cooperation.
???? - Use data-driven approaches to guide, not dictate, police actions.
???? - Implement policies to ensure that stops and searches are conducted fairly and based on reasonable suspicion.
领英推荐
?
5. Identifying Repeat Offenders Faster
?? - Benefit: Algorithms can help identify individuals who are at high risk of reoffending, allowing for targeted interventions that prevent future crimes.
?? - Challenge: Predictive tools that focus on individuals can lead to ethical concerns about preemptive punishment and infringement on civil liberties.
?? - Example: A predictive policing system in Philadelphia identified high-risk offenders, leading to a 35% reduction in violent crime among those who received targeted interventions. However, the program faced criticism for potentially stigmatizing individuals based on statistical probabilities rather than concrete evidence. [(Source: Berk et al., 2009)](https://repository.upenn.edu/cgi/viewcontent.cgi?article=1340&context=psc_working_papers).
?? - Measures to Avoid:
???? - Use predictive data as one factor among many in decision-making, rather than the sole determinant of police actions.
???? - Ensure transparency in how individuals are identified and monitored.
???? - Establish ethical guidelines and oversight committees to review predictive policing practices.
?
6. Preventing and Addressing Domestic Violence
?? - Benefit: Predictive tools can help identify households at risk of domestic violence, enabling early interventions that protect victims and prevent escalation.
?? - Challenge: The use of predictive analytics in sensitive situations like domestic violence can lead to privacy concerns and unintended consequences.
?? - Example: In New York City, predictive policing helped identify high-risk households, resulting in a 15% decrease in domestic violence incidents. However, there were concerns about the potential misuse of data and violations of privacy rights. [(Source: New York Times, 2017)](https://www.nytimes.com/2017/05/22/nyregion/predictive-policing-cuts-crime-in-nyc.html).
?? - Measures to Avoid:
???? - Implement strong data protection policies to safeguard personal information.
???? - Use predictive tools to support, not replace, human judgment in sensitive cases.
???? - Ensure that interventions are carried out with the consent and cooperation of those involved.
?
7. Efficiently Monitoring Large Public Events
?? - Benefit: Predictive policing can help manage large public events by forecasting potential security risks and enabling preemptive measures.
?? - Challenge: Over-reliance on predictive tools can lead to excessive surveillance and restrictions on public freedoms.
?? - Example: During the 2012 London Olympics, predictive policing was used to monitor crowds and prevent potential threats, contributing to a safe and secure event. However, some critics argued that the level of surveillance was excessive and intrusive. [(Source: BBC, 2012)](https://www.bbc.com/news/technology-19011975).
?? - Measures to Avoid:
???? - Use predictive tools in conjunction with other security measures to ensure a balanced approach.
???? - Maintain transparency with the public about surveillance practices and data use.
???? - Ensure that security measures respect individual rights and freedoms.
?
8. Identifying and Disrupting Drug Trafficking Networks
?? - Benefit: Predictive analytics can help identify drug trafficking networks and disrupt operations before drugs reach communities.
?? - Challenge: Targeting specific individuals or groups based on predictive models can lead to accusations of bias and profiling.
?? - Example: A predictive policing program in Miami used data analytics to identify and dismantle several drug trafficking rings, leading to a 22% reduction in drug-related crime. However, it faced criticism for disproportionately targeting minority communities. [(Source: Miami Police Department Annual Report, 2018)](https://www.miamipolice.org/docs/Annual_Report_2018.pdf).
?? - Measures to Avoid:
???? - Use predictive tools to identify patterns and trends rather than targeting individuals.
???? - Ensure that drug enforcement efforts are guided by evidence-based practices and not solely by algorithmic predictions.
???? - Regularly review practices to ensure compliance with anti-discrimination laws and policies.
?
9. Optimizing Patrol Routes for Improved Response Times
?? - Benefit: Predictive policing can optimize patrol routes based on crime patterns, increasing efficiency and reducing response times.
?? - Challenge: Relying solely on predictive tools for patrol routing can lead to gaps in coverage and missed opportunities for community engagement.
?? - Example: The Los Angeles Police Department (LAPD) used predictive policing to optimize patrol routes, reducing crime by 13% in targeted areas. However, this approach sometimes resulted in less coverage in non-targeted areas, leading to increased crime elsewhere. [(Source: Police Foundation, 2013)](https://www.policefoundation.org/publication/predictive-policing-in-los-angeles-the-los-angeles-police-department-experience/).
?? - Measures to Avoid:
???? - Use predictive tools to supplement, not replace, officers' knowledge and experience.
???? - Balance data-driven patrols with community policing efforts to maintain public trust.
???? - Regularly review and adjust patrol strategies based on crime trends and community feedback.
?
10. Forecasting and Combatting Cybercrime
?? - Benefit: Predictive analytics can help identify potential cyber threats and vulnerabilities, allowing law enforcement to take preemptive action.
?? - Challenge: The rapidly evolving nature of cybercrime can make it difficult for predictive models to keep up, leading to outdated or inaccurate predictions.
?? - Example: In 2019, a predictive policing initiative in the UK helped identify several potential cyberattacks, preventing significant financial losses. However, some predictions were based on outdated data, leading to missed opportunities. [(Source: UK Home Office, 2019)](https://www.gov.uk/government/publications/cyber-crime-prevention).
?? - Measures to Avoid:
???? - Regularly update predictive models to reflect the latest cybercrime trends and threats.
???? - Use predictive tools in conjunction with real-time monitoring and intelligence gathering.
???? - Foster collaboration between law enforcement and cybersecurity experts to stay ahead of emerging threats.
?
11. Predicting Human Trafficking and Dismantling Networks
?? - Benefit: Predictive policing can help identify potential human trafficking operations, enabling early interventions and protecting victims.
?? - Challenge: Human trafficking is a complex and hidden crime, making it difficult for predictive models to accurately identify patterns without comprehensive data.
?? - Example: A predictive policing program in San Francisco used data analytics to identify trafficking hotspots, leading to the rescue of several victims. However, the program struggled with a lack of reliable data, resulting in some false positives and missed opportunities. [(Source: San Francisco Chronicle, 2020)](https://www.sfchronicle.com/bayarea/article/SF-program-fights-human-trafficking-by-predicting-14997374.php).
?? - Measures to Avoid:
???? - Work with non-governmental organizations and community groups to gather reliable data on human trafficking.
???? - Use predictive tools as part of a broader strategy that includes victim support and community outreach.
???? - Continuously refine models based on new data and feedback from frontline workers.
?
12. Anticipating Terrorism Threats and Mitigating High-Risk Scenarios
?? - Benefit: Predictive analytics can help identify potential terrorism threats, allowing law enforcement to prevent attacks and protect public safety.
?? - Challenge: Predictive policing in counter-terrorism can lead to excessive surveillance and infringements on civil liberties, especially for certain ethnic or religious groups.
?? - Example: After the 9/11 attacks, the New York City Police Department (NYPD) used predictive policing to prevent several potential terrorist plots. However, the program faced criticism for targeting Muslim communities and infringing on civil rights. [(Source: The Intercept, 2015)](https://theintercept.com/2015/04/27/nypd-surveillance-post-911/).
?? - Measures to Avoid:
???? - Ensure that counter-terrorism efforts are guided by evidence and not solely by algorithmic predictions.
???? - Implement oversight and accountability mechanisms to protect civil liberties.
???? - Engage with community leaders and organizations to build trust and cooperation in counter-terrorism efforts.
By understanding the benefits and addressing the challenges of predictive policing through careful implementation, continuous evaluation, and community engagement, law enforcement agencies can enhance the effectiveness and fairness of their strategies while maintaining public trust and ethical standards.
?
The Future of Predictive Policing
Although already utilized for decades, as public awareness grows, predictive policing is very likely to play a more significant role as crimes waves continue to rise and predictive policing can be used for crime deterrence .
Yet, as governments are implementing governance laws, regulations and frameworks, are exploding, we are already seeing strong prohibitions like with the EU AI Act whereas social credit scoring systems and specific predictive policing applications are strictly prohibited.
With the rise of LLMs in multiple languages, increased demands for transparency in AI usage, significant advances in computer vision, red teaming approaches, universal usage of multimodal models which multiply complexcity, predicitive policing organizations will have to acquire a much larger army of AI Professionals, which definitely creates a barrier to entry for nations not already utilizing and augmenting AI predictive policing daily.
Cross-collaborative agreements and practices in clean data sharing is also anticipated but is complex given differences in norms, beliefs, practices and objectives.
Simple concepts like freedom, human rights and liberties are being perceived and implemented differently so it may be hard to get to a consensus as humanity is testing different paths towards social innovation.
Our future can be extraordinary and keep on including predictive policing in more advanced forms but, we need to ensure that related AI solutions are being developed and orchestrated to correct root problematic causes and aim towards our overall betterment. At all times, human oversight and participation will have to be performed so predictive policing can serve our continuously improved lives whilst being fair, reliable, subservient and trustworthy.
?
References
1. Brennan Center for Justice. “Predictive Policing Explained.” [Link](https://www.brennancenter.org/our-work/research-reports/predictive-policing-explained)
2. Office of Justice Programs. “Predictive Policing: The Future of Law Enforcement?” [Link](https://nij.ojp.gov/library/publications/predictive-policing-guide-using-data-and-technology-forecast-crime)
3. RAND Corporation. “Predictive Policing: Forecasting Crime for Law Enforcement.” [Link](https://www.rand.org/pubs/research_reports/RR1488.html)
4. MDPI. “Predictive Policing and Crime Prevention.” [Link](https://www.mdpi.com/2071-1050/11/21/6065)
5. Springer. “Predictive Policing in Germany.” [Link](https://link.springer.com/article/10.1007/s10610-018-9384-7)
6. Liberties.eu. “Predictive Policing in Japan.” [Link](https://www.liberties.eu/en/news/japan-predictive-policing/16531)
7. BBC. “Technology and Security: How Technology is Changing Police Work.” [Link](https://www.bbc.com/news/technology-19011975)
8. The Intercept. “How the NYPD’s Post-9/11 Surveillance Program Impacted Muslim Communities.” [Link](https://theintercept.com/2015/04/27/nypd-surveillance-post-911/)
9. San Francisco Chronicle. “SF Program Fights Human Trafficking by Predicting Risk.” [Link](https://www.sfchronicle.com/bayarea/article/SF-program-fights-human-trafficking-by-predicting-14997374.php)
10. New York Times. “Predictive Policing Cuts Crime in NYC.” [Link](https://www.nytimes.com/2017/05/22/nyregion/predictive-policing-cuts-crime-in-nyc.html)
11. Police Executive Research Forum. “Using Predictive Policing to Enhance Crime-Fighting.” [Link](https://www.policeforum.org/assets/predictivepolicing.pdf)-*