What is Predictive Policing?

Predictive PolicingRecent news stories about Orange County Sheriff and Los Angeles PD are using techniques to “stop crime before it happens” has people asking about the tech wonder known as predictive policing: what is it, how is it being used in California and what it means for criminal investigations. Let’s take a look.

Predictive policing may also be referred to as “proactive policing.” Actually these two approaches are quite different, but both are used with the hope of preventing or mitigating crime. Let’s get a quick idea of each one.

A post card definition of predictive policing is using statistics, machine learning and data analytics techniques (artificial intelligence) to identify potential crime activity, potential offenders, or even possible victims.

Predictive policing depends on software and data supplied to it for analysis. The software analyzes data sets including past criminal data, environmental data, and arrest records and makes forecasts about specific crime activity likely to take place within a short time horizon.

According to suppliers of predictive policing systems the programs detect patterns in data and direct police to “promising” locations or people for immediate police intervention. The promises in using these systems are that they will help law enforcement stop crime by pointing officers to places where crime is likely to happen, and will help police to manage personnel resources more efficiently.

While large law enforcement agencies are enthusiastic about the deployment of predictive policing tools, they raise concerns and controversy about civil rights, individual privacy and high-tech profiling.

Proactive policing is different than predictive policing. These two can be used in combination, though. Proactive policing uses police presence and community engagement programs to reduce crime.  These include increased patrolling of high-crime locations, addressing problems that may cause crime, employing programs targeted at high-risk groups, and community participation in crime prevention.

How Predictive Policing Tools Work

In simple terms, predictive tools use computer software and past crime data that map the data to individual small geographic areas of a city. They then estimate, in near real time, the risk of a crime occurring in each area of a grid usually for an immediate upcoming time period. The most sophisticated systems combine additional data from multiple sources to create criminal activity forecasts, such as the terrain of the city (like location to freeways), weather forecasts, type of business establishments in the area, and artificial intelligence techniques to identify invisible patterns.

Predictive Policing in Orange County and Los Angeles County

The Orange County Sheriff’s Department, Orange County cities and LAPD have been using variations of predictive policing software and analysis for a few years.

The LAPD was among the first to embrace predictive policing operations in early 2011 with operation LASER. That system assigns “points” every time an individual has an interaction with police such as an arrest, a stop, or a complaint; or you get points for your criminal history such as being on parole or probation. There is a “Catch-22” factor in these systems; the higher the points you have the more future “attention” you are likely to get from police, earning you more points for each interaction. The list of people on the LAPD’s LASER “Chronic Offenders” list is not public, so you have no way to know if you are on the list.

The LAPD also employs a system called PredPol to predict which neighborhood blocks will see vandalism, a burglary, break in, or a property theft. The idea is that police show up in a detected neighborhood “hot spot” before a crime occurs and that their presence will deter the crime, or they could even catch someone in the act.

In Orange County, the Sheriff’s department uses a mix of predictive analytics, including PredPol. Officers are encouraged to spend time in targeted areas identified by the computer system.

Garden Grove Police use another system that gives officers and the public detailed real time information about crime in the city. Officers on duty see reports and graphs of crime statistics, trends and incident reports. They use the information to plan where be at any given time of day. The makers of this software claim it is so easy to use it doesn’t require the expertise of data analysts but can be used by beat officers on patrol for real time intelligence.

Departments across California are employing new kinds of data driven policing methods, and almost always without citizen’s being aware of how data is driving police activities.

What Could Possibly Go Wrong?

A gang case in New Orleans based in part on evidence gathered using data-driven systems is being appealed on the basis that the evidence was gathered in violation of the defendant’s rights. In this case, almost no one knew, including city officials, that an artificial intelligence system was in use by police.

Lots of other red flags are being raised about predictive policing: forecasts are only as good as the data fed into them, the algorithms themselves may contain bias, people in heavily policed areas are more likely to get have police interactions or get into trouble, and the public and police themselves have no real visibility into how the software algorithms work and decide the locations or people to target.

While a RAND corporation study casts doubt on how effective these high-tech systems are on a local level, other studies say predictive policing can accentuate inequalities in the justice system because of the role of race, class, group affiliation, or even political forces in shaping the power equation.

Controversy abounds over how the systems affect officer’s perceptions of “reasonable suspicion” in stopping people. The data that algorithms rely on may reinforce systemic bias. For example, New Orleans PD used field interview cards, information sheets officers fill out after stopping or coming in contact with a citizen, as part of their predictive database. Defense attorneys are charging this violates privacy, is unconstitutional, and causes racial or other bias to seep into the decision-making system.

Challenging Predictive Policing Evidence In Criminal Cases

Challenges to the accuracy, viability and effectiveness of predictive policing methods are coming from researchers, advocacy groups, lawsuits, and ethical questions.

But, what about immediate options for criminal defense when an arrest has been made after police and investigators rely on predictive policing tactics? Your first line of defense is hiring a defense attorney who is experienced with high-tech crime fighting tools and techniques.

Evidence that relied on these systems may be able to be suppressed so it cannot be used against you in court. Predictive policing programs have been accused of introducing bias, generating false positives, or containing programming and algorithm flaws. Your defense attorney may be able to file a motion with the court to exclude such evidence.

As we learn more about predictive policing systems and as more prosecutions are filed using evidence and police activity based on them, the more cases are being challenged in courts by defense attorneys across the country. Rulings will impact cases that are now moving into the justice system. Stay tuned.

David Stein is available to consult with you at no charge on any criminal matter. He is an experienced Orange County criminal attorney, working in all courts. Contact our law offices at (949) 445-0040 today to talk about how we can help you.