Criminal Calculus - Algorithm Week


By Nicole Stinson

Image for illustrative purposes, courtesy of Brett Lider.

Crime fighting through numbers, once popularized by the criminal drama television series, NUMB3RS, is now becoming a reality through new studies. Faced with tight budgets and limits on hiring, police departments have sought the aid of computer programs with the ability to predict the outbreak of local crime and violence on an international scale.

Kalev Leetaru, the creator of the Culturomics 2.0 project, tells BTR, “everyday society just puts out incredible amounts of information … it certainly seems that we should leverage that to understand human society in new, unique and powerful ways.”

Culturomics 2.0 was designed in 2011 as a pilot and prototype for graphing geographical patterns from news media sources. Until Culturomics 2.0, “no one had ever used a computer to look at hundreds of articles to extract information and look for patterns,” says Leetaru.

Feeding news media data to the supercomputer Nautilus from The National Institute of Computational Sciences at University of Tennessee, he found that significant patterns and relationships could be made through network diagrams. These diagrams could predict strong relations between countries through their representative nodes. The more connected the modes, the more those countries were linked in news media.

“We include all news media and the reason for that is all news media is bias,” says Leetaru. “If you scoop up everything you can get your hands on you start to be able to triangulate things.”

In his research, Leetaru also used Nautilus to graph the tone of these news articles. He found that when the graph was negatively correlated that instabilities often occurred in those countries soon after.

“In the case of the Egyptian revolution, it was a war of progression, you don’t just get people saying, “hey I think I will go and protest today,” he says. “Measuring this progression through news data allows us to predict these instabilities.”

It has never been used on active cases, however, the study has demonstrated how easily it could have revealed the location of Osama Bin Laden. In April, Leetaru released his new project, Global Database of Events, Language and Tone (GDELT). This initiative built on Culturomics 2.0; however, shifts away from predicting crime and violence.

In the same year as Culturomics 2.0’s release, the Los Angeles Police Department in conjunction with the University of California’s Mathematical and Simulation Modeling of Crime (UC MaSC) Project finished the technology Predpol.

“PredPol’s technology generates predictions for areas and windows of time that are at highest risk for future crimes,” Donnie Fowler, the Director of Business Development for Prepol, tells BTR.

It took six years to develop Predpol and it is an on-going project with co-founders Jeffrey Brentingham, a professor of anthropology and George Moehler, an assistant professor of mathematics and computer sciences, now at Santa Clara University. Patrol officers are delivered Prepol’s predictions before each shift and they are produced for all types of crimes. The aim is to allow officers to focus on high-risk areas.

As an educational tool, new police have also come up to speed on local crime faster, Fowler says. Predpol relies on only three pieces of data: the type of crime, the location, and when time the crime occurred.

“The magic of this technology is in the math more than in the amount of data crunched,” says Fowler.

In the last two years, the technology has developed to include predictions for gang activity, traffic accidents, and gun violence.  It has also been used to calculate roadside bombs in Iraq.

A report by The Economist questions this targeted approach, stating that the positioning of “police in hotspots discourages opportunistic wrongdoing, but may encourage other criminals to move too less likely areas.”

However, the LAPD are praising the results from Predpol.

“The best news is that the technology works, with declines in crime in every place where PredPol is being used,” says Fowler.

Since the implementation of Predpol, crime has continued to fall in Los Angeles with Chief Charlie Beck announcing that violent crime was down by 8.2 percent from 2011. Gang homicides were also down by 12 percent and overall gang-related crime has decreased by more that 10 percent in the year.

Predpol technology is now also used in the United Kingdom and Australia.

New York City also has its own crime fighting supercomputer in a joint venture with Microsoft. Unveiled last year, the Domain Awareness System (DAS) uses data from security cameras, license plate readers and crime reports to predict future crime outbreaks.

“Part of the reason we have been able to continue driving down crime to record lows is our heavy investment in technology and our willingness to develop new, cutting-edge solutions to keep New Yorkers safe,” outgoing Mayor Michael Bloomberg told the press.

As possibilities of technology continue to grow, so will the research into crime preventive measures and the power to predict crimes before they happen. The integration of these new crime fighting super computers is proving to be the differential in the algorithms of lowering crime.