Predictive Policing: Can a Crime Be Precisely Foreseen?

Predicting crime has always been the way to solve the ever fluctuating crime rate, especially since Minority Report hit the theaters. However, are there existing tools so advanced they can match the unprecedented crime prediction level? And, are they also reliable enough to do that? In this article lawless.tech takes a look at so-called predictive policing, and the way it works.

Predictive Policing as the Minority Report of Today

Generally speaking, predictive policing is a technology allowing one to identify potential criminal activities in certain areas. Algorithms used in predictive policing are based on various types of statistical crime data like locations, times etc. Based on them, the algorithm analyzes the recurrence possibility of a crime in the selected area. Generally, there are four categories of predictions that the system can make:

• Prediction of crimes.
• Prediction of offenders.
• Prediction of the criminal’s identity.
• Prediction of victims.

The talks over the predictive policing have been there for a while, especially in the pop culture. A widely known movie, Minority Report, based on the short story by Philip K. Dick, is probably the best example of the technology’s interpretation in quite an exaggerated manner. However, the current state of things differs from that in the movies. The study conducted by The RAND Corporation, a nonprofit institution that aimed at improving policy and decision-making process, suggests that “applying predictive policing methods is not equivalent to finding a crystal ball.” Therefore, the technology cannot actually foretell the future — its implications are limited solely to risk assessment.

The start of the public discussion of predictive policing dates back to 2008, when former Police Chief of the Los Angeles Police Department, William J. Bratton, first came up with a model for such a system. Back then, Chief Bratton together with James H. Burch II, the acting director of the Bureau of Justice Assistance, and Kristina Rose, the acting director of the National Institute of Justice (NIJ), worked on the concept of predictive policing and its possibilities for the law enforcement bodies. Their cooperation led to the organisation of a symposium a year later, during which the stakeholders were ready to discuss the issue with the public. In one of his speeches during the symposium, Chief Bratton said:

“Predictive policing is taking advantage of the evolution of [the community policing] concept, where we can gather information more quickly than ever in the past, analyze it, and from that, actually begin to predict that certain actions, based on intelligence, are going to occur and seek to prevent them. And so this is not a stand-alone type of concept. It is very much a significant enhancement of what has come before and what is still actually evolving.”

The concept has quickly become popular in the media. CBS News, NBC News have issued stories about the technology. The New York Times covered the experience of Santa Cruz Police Department that experimented with the technology. Their attempt was so successful that TIME Magazine even listed the technology developed by the SCPD amongst the 50 best inventions of 2011.

In 2012 , Dr. Jeff Brantingham and Dr. George Mohler set up a company named PredPol, which came about thanks to the cooperation between LAPD and the University of California.  Back then Chief Bratton wanted to find a way to use the data from CompStat, a system of organizational tools employed in many American police departments since the 1990s, differently than just for historical purposes. The goal of the project was to identify whether the CompStat data can be used to moderate recommendation of when and where a possible crime may occur. Using such an information, a police officer may have been deployed to monitor a certain area, which could become a way to prevent crime. The designed model was used in the SCPD during the 6-month trial of applying the technology in practice. The SCPD adapted the Hawkes Process model employed in seismology for forecasting earthquakes, to crime prediction.

Since then, the technology was implemented in many jurisdictions. Several states, namely Alabama, Arizona, California, Illinois, South Carolina, Tennessee, and Washington, used the technology in their police departments. Later on, the authorities of the Kent County in the U.K. also experimented with the tech. Several other countries have also developed their own unique prediction policing systems, like Crime Anticipation System (CAS) in the Netherlands, Pre Crime Observation System (PRECOBS) in Germany and Switzerland, and, finally, an unnamed predictive policing system which is being used in Xinjiang, China.

While predictive policing is called a technology that is capable of “stopping crime before it starts,” people should understand that it is still pretty far from the prospects shown in Minority Report.

Real Chance to Foresee the Crime

Predictive policing uses various types of data for the analysis. According to PredPol, the three most objective data points to be collected for a crime are the type, the location and the date & time. Police squads use the results of this analysis to optimize the patrolling strategy during specifics shifts. Such a strategy generally increases the chances of deterring or preventing a crime.

Generally speaking, such a technology’s purpose is to define whether different criminal offences have any patterns in time and space. It’s quite easy to assume that there can be districts in a city where it’s more likely that you get mugged, however, things aren’t that simple when you have to come up with a pattern for a city center. PredPol uses intricate data to predict a crime with better results highlighting the hot spots. Covering the four-months long experiment in Kent, the Economist observed that “8.5% of all street crime occurred within PredPol’s pink boxes, with plenty more next door to them; predictions from police analysts scored only 5%. An earlier trial in Los Angeles saw the machine score 6% compared with human analysts’ 3%.”

The results become clearer when put in comparison::

“Within six months of introducing predictive techniques in the Foothill area of Los Angeles, in late 2011, property crimes had fallen 12% compared with the previous year; in neighboring districts they rose 0.5% (see chart). Police in Trafford, a suburb of Manchester in north-west England, say relatively simple and sometimes cost-free techniques, including routing police driving instructors through high-risk areas, helped them cut burglaries 26.6% in the year to May 2011, compared with a decline of 9.8% in the rest of the city.”

Predictive policing in a sense is not only limited to making predictions.

“Rather, it is a comprehensive business process, of which predictive policing is a part,” reads the report on predictive policing by the RAND Corporation. Generally, the process consists of four stages: data collection, analysis, police operation and criminal response.

First things first, the relevant data is being collected and sorted out to form a database that is fed to the algorithm. The data is regularly updated to reflect the effect of the actions from the police and the subsequent criminal response. Here, not only the data directly related to crime is taken into account, but also that of the environment where the crime occurred. Then, the collected data set is analysed to set out some unique crime patterns for a selected region. The identified trends are then used to form hot spots on the map.

After the data is compiled and analyzed the police will have to do something about it. For instance, when the hot spots are defined, the number of patrols or the patrolling time may be increased to influence the situation. Constant monitoring of the effectiveness of such interventions taken by the police is important to evaluate the their influence as it gives the insight into the extent the policing measures have reduced the crime.

The interventions from the police will usually result in the behavioural patterns change of potential criminals. Thus, some hot spots may cool down, and the criminal activity may be transferred to another location. This will contribute to making the original data sets obsolete. Hence, new data will be collected, and the entire process will start anew.

While being a very hot topic in the media, predictive policing as any other innovative technology has gathered around many myths. It sometimes may lead to very unrealistic expectations from the end user, who is not quite familiar with how the technology exactly works.

The major myth would be thinking of predicting policing as a tool that can literally foretell the future. The reality is way more boring: he algorithms can help predict only the risk of a future event, not the event itself. In addition, the predicting will be only as good as the actual data used to make it.

One more interesting myth is that the predictive policing is the solution to all the crime there is. However, it is far from it as well. Predictive policing is just a tool with which the police can decide to take specific actions to improve a situation. Therefore, without it, any predictions are practically useless.

While so many myths have been constantly discussed by the developers as well as the law enforcement bodies, it may seem that there is still some lack of understanding on the general public’s part. Not only some of myths are still around, but so are many questions, especially related to civil liberties.

Are Civil Liberties Endangered?

Due to its very nature and numerous sci-fi connotations that pop up almost automatically, it seems inevitable that predictive policing has to catch the eye of digital rights advocates. A few concerns have been raised, namely the issues concerning privacy rights, discrimination and racial profiling, and overpolicing certain neighborhoods, among other things.

Most certainly, such an obvious concern has been subject to prolonged discussions ever since predictive policing has become a thing. Thus, during the first international symposium, Chief Bratton was asked as to how he sees the possible ways of preparing the community for the evolution of predictive policing tech The Chief responded:

“A concern of the community and rightfully so, is this issue of civil liberties. As we seek to use more and more data and information that are available through computer analysis, there’s a concern of privacy, there is a concern of civil liberties. And, that is something that needs to be discussed, needs to be discussed openly, needs to be researched, needs to be reviewed. I am comfortable that if we do that in a transparent way; if we admit up front that there are these fears and concerns, and potential issues; that we should not use that as an excuse to dismiss moving forward, but rather, find a way to ensure that the public is kept informed of everything we’re doing—[that it] is legal, is constitutional, and ultimately, will be for their benefit.”

Chief Bratton also addressed the expectation of privacy while discussing predictive policing. Specifically, he brought up the example of the usage of cameras by the police in public spaces. Respectively, he mentioned the position of the U.S. Supreme Court that permits such practice and that individuals have no legitimate expectation of privacy in the case. At the same time he also stressed that predictive policing in this regards cannot be abused. Mr. Bratton elaborated:

“As predictive policing moves forward, as community policing did, as problem-solving policing did, as broken windows policing did, we need to be mindful of legal issues, constitutional safeguards, and to discuss openly, the concerns of the public, of the media, of civil liberties groups, and be prepared to address it. And if in fact we find that we are going outside the guardrails, then we need to very quickly get back within the guardrails and not risk overturning what is an extraordinarily useful tool for policing to prevent crime.”

The findings of the RAND Corporation’s research tend to agree with the ambiguity of the  issues of privacy rights in question. Particularly, the difference is in the potential risk that predicting policing tools may pose depending on one of four categories of the prediction. For example, if it is about predicting a crime, no personal data is actually involved. However, it’s not that simple when it comes to predicting the perpetrator.

When it comes to PredPol, the company’s stand regarding the issue is quite straightforward. For prediction policing, PredPol collects a limited amount of data points for further analysis that are not private data, but are inherent to the crime itself. In its White Paper on the Theory and Practical Deployment of the technology the project states:

“PredPol is about predicting where and when crime will occur, not who will commit a crime. PredPol is not criminal profiling. It does not use any information about individuals or populations and their characteristics. The patterns inherent in the crimes themselves provide ample information to predict where and when crimes will occur in the future. Predictive policing disrupts the short-term, situational causes of crime. It does not solve criminality, or the propensity for individuals to commit crime. Predictive policing is therefore not a replacement for policy and community engagement strategies needed to steer people clear of criminal careers in the first place.”

However, it seems like the general public thinks that the ambiguity in terms of privacy issues is a negative characteristic of predictive policing. In the Statement of Concern about Predictive Policing, the American Civil Liberties Union (ACLU) and 16 more civil rights organizations raise many questions about predictive policing. In particular, the signees draw the attention to the lack of transparency from the vendors and the agencies. They highlight that “all stakeholders must understand what data is being used, what the system aims to predict, the design of the algorithm that creates the predictions, how predictions will be used in practice, and what relevant factors are not being measured or analyzed.”

The RAND’s report also agrees with the signees:

“Transparency about the types of information collected and the uses of that information may further help allay fears of invasion of privacy.”

The next major issue highlighted by the media and digital rights advocates concerns the discrimination and racial profiling.

The issue was heavily addressed in the Statement of Concern by civil rights organizations:

“The institution of American policing, into which these systems are being introduced, is profoundly flawed: it is systemically biased against communities of color and allows unconscionable abuses of police power. Predictive policing tools threaten to provide a misleading and undeserved imprimatur of impartiality for an institution that desperately needs fundamental change. Systems that are engineered to support the status quo have no place in American policing. The data driving predictive enforcement activities — such as the location and timing of previously reported crimes, or patterns of community- and officer-initiated 911 calls — is profoundly limited and biased.”

Seems like the attention is mostly focused on profoundly flawed data that is fed to the predictive policing algorithms. For example, the signees of the aforementioned statement stress that law enforcement bodies use only the records of the response to the situations they actually encounter rather than the complete records of all the crimes that occur. As a result, they say that the information in the output is essentially biased, which leads to the “unwarranted discrepancies in enforcement.”

The application of the technology can also entail a certain bias when it comes to signifying hot spots that require attention from law enforcement bodies, according to an Equalfuture research. For example, it may create a vicious circle of overpatroling certain neighbourhood with negative crime history records.

In addition, many other implication have been made regarding the alleged racial profiling of predictive policing systems. One of the most vivid examples of such was the experience of Robert McDaniel. He had neither committed any violent crime, nor had any gun violations. However, he still was placed on the Chicago city’s heat list, which was designed to signify potential high-risk violent offenders. As the list was specifically designed to record such offenders, there was no place for McDaniel to make it there, since he had only multiple arrests on suspicion of minor offenses but only one misdemeanor conviction. “I haven’t done nothing that the next kid growing up hadn’t done. Smoke weed. Shoot dice. Like seriously?”, McDaniel told in an interview to Chicago Tribune.

Conclusion

One of the ways to solve any problem is to try to tackle it before it emerges. This is probably the most important thesis when it comes to the crime prevention. A lot of methods have been used to achieve the desired result – community policing, intelligent policing, behavioral influences and many more. Hence, predictive policing emerged as a next evolutionary step in crime prevention. However, as innovative as it may seem, there is still balance between privacy and crime preventing to be found.

Predictive policing tools are widely considered to be efficient. They show better results in controlled experience than living data analysts and strategists. They also have an influence over the crime rate in real neighborhoods.

At the same time, these tools are widely criticized by the public. The community considers it a devolutionary step from the broken windows policing, and draws significant attention to the allegedly inherent flaws, like biased data that is crunched to achieve the results. In turn, this reportedly leads to quite negative outcomes, like privacy issues or discrimination.

What stakeholders in question can agree upon is that predictive policing tools can be used properly in future. What they need is a tweak to work properly for the good of the society.

Follow us on Twitter to stay tuned on the recent developments in regulation of new technologies, and be the first to read expert opinions.