Predictive Policing

Predictive Policing: Could Infusing Law Enforcement With Data Science Stop Crime In Its Tracks?

Photo courtesy of IBM Media via Flickr

Healthcare industries use science and data to not only treat human ills, but predict and prevent them. Law enforcement, currently, is more of a band aid than a vaccination for crime, so to speak.

But what if law enforcement — in partnership with scientists and criminologists — were able to collect and combine data to generate insights that could actually predict and prevent crimes before they occur?

That’s what many hope for the future of crime prevention. According to Elizabeth Glazer, director of New York Mayor Deblasio’s Office of Criminal Justice, data could be used to society’s advantage by moving “the study of crime from the realm of astrology into the realm of science.”

The problem

Law enforcement has gathered a lot of data over the years, and has since the 90s used intelligence-led policing programs to look back at crimes in a historic way, so that police departments can deploy resources accordingly. This form of data-driven policing has been helpful in analyzing past events, but is on a whole inadequate in predicting the future.

Then, there’s crime itself. Though crime is at a historic low in New York City (along with the entire country), the reason is amazingly elusive, even to experts. This drop in crime corresponds with an even larger spike in incarceration, but there is little hard evidence to determine correlations — what’s working, and what’s not.

Data also shows that recidivism is a major contributor to crime: about two thirds of those enter the criminal system are arrested again within three years of their release, and mass incarceration is as such that 25 percent of the world’s prisoners are American.

Poverty, poor education, and other elements of inter-city culture can foster criminal behavior, which is reinforced as some continue to commit crimes and bounce to and from prison without successful reform.

Ignoring such problems, or addressing them only in hindsight, has had very limited success. But with predictive policing, through which officers use data to predict and prevent crime, new solutions could lead to radical changes.

Data-driven solutions

Data has been used to the police’s advantage for several decades, and some of it is even available to the public:  we can now view data on past crimes, for example, through NYC’s crime map — an interactive that marks the location and details of historic crimes.

But there is greater sophistication at work behind the scenes, as city police departments, companies, and organizations push new strategies that aim to prevent and predict crime, and not just examine it in retrospect.

Here’s a look at how it works, from the RAND Corporation:

Screen Shot 2015-03-30 at 11.22.39 AM

In the U.S., such predictive policing techniques have already been implemented in California, Washington, South Carolina, Arizona, Tennessee, and Illinois to notable success.

PredPol

The American company PredPol provides cloud-based software to law enforcement agencies to algorithmically generate predictions on places and times future crimes are most likely to occur.

The software has been used to proven and significant crime reduction in Los Angeles, Santa Cruz, Atlanta, Richmond, and elsewhere.

PredPol takes just three datasets — past place, type, and time of crime — to develop unique algorithms for specific locations and map the result in boxes to assist officers in patrolling. To the relief of privacy advocates, it never collects personal information about law enforcement, victims, or criminals.

PredPol was named one of the top 50 inventions in 2011, and is the leader in the emerging marketplace of predictive software.

The Crime Lab

But some seek to go even further back, to stop people from engaging in criminal behavior to begin with. The Crime Lab, which was announced to partner with New York City’s Office of Criminal Justice in January of 2015, uses evidence-based measures to prevent arrests and incarceration.

With a staff of statisticians, academics, and behavioral economists, the lab will employ “sophisticated data research methods” to test practical interventions in targeted locations based on on computer science.

In the past, Crime Lab has found success in Chicago in its employment of cognitive behavior therapy (C.B.T.) for troubled teens, which yielded a 44 percent reduction of crime when in effect as well as academic betterment.

The program, identifying that 70 percent of youth homicides were caused by ill-handled alterations in the presence of a gun, found when students were encouraged in C.B.T to think more rationally and control impulses, crime decreased significantly.

Such efforts went to show that teenagers are not a lost cause, as many have previously been eager to assume. They also show that focusing on behavior and controlling risk is worthy of rigorous focus.

What lies ahead

The efforts of companies, tools, programs and softwares demonstrate that when law enforcement is infused with a dose of data science, it has the potential to become even more effective, and make the world a bit safer all the while.

This isn’t necessarily news. Back in 1997, Cambridge criminology professor Lawrence W. Sherman’s widely cited paper postulated that “Police can prevent robbery, disorder, gun violence, drunk driving and domestic violence” but only through proactive approaches to arrests, smartly directed patrolling, problem-oriented policing, and respectful community interactions.

Essentially, we’ve known for years how to prevent crime and how not to; now, we finally have the science and the experts to apply such data practically. We also have a landscape safer than ever before, on which we can finally find out — scientifically — what causes and assists criminal behavior, and where to focus energy.

But predictive policing is not without its critics: many worry about privacy issues, including the possibility of ex-cons being tracked. As the RAND corp rightly point out in a 2013 study, predictive policing is not a crystal ball. Data-driven policing could be a slippery slope, especially if personal data and excessive surveillance were to become inputs.

Others are concerned about the impact of predictive and proactive efforts on racial profiling, while others suppose it could actually curb prejudice.

The truth may be somewhere in between. Though there is good evidence to support the efficacy of data science in preventing crime so far, there will always be worries of a Minority Report-style future, in which the prevention of crime would take precedence over civil liberty.

We measure success by the understanding we deliver. If you could express it as a percentage, how much fresh understanding did we provide?
Jennifer Markert