– Not science fiction: Miami wants to predict when and where crime will occur (Miami Herald, April 23, 2015):
Armed with high-tech software and years of crime data, Miami police believe they will soon be able to stop crimes by predicting when and where they will occur.
It sounds a little like something out of a science fiction novel, but the department is in the process of adopting a system called HunchLab that produces maps showing small areas where specific crimes are likely to be committed during shifts. The probability program is a geographical version of “predictive policing” software, which more departments are using — even if, in the words of one supportive cop, it’s “kind of scary.”
Similar algorithm-based programs have been credited with lowering crime rates in cities around the country, and some South Florida departments recently have adopted their own systems. In Miami’s case, the department is funding the implementation of HunchLab and other software programs with a $600,000 federal grant doled out by the Bureau of Justice Assistance to encourage smart policing tactics.
Miami accepted the grant in November. On Thursday, the city commission will vote on a $120,000 contract with Florida International University to have Rob Guerette, an associate professor of criminal justice, study the department’s program and run tests ahead of its launch.
“It doesn’t replace actual police work,” said Lt. Sean MacDonald, who wrote the grant application. “It’s policing with smarter technology.”
MacDonald said Miami police want to use HunchLab mostly to tackle robberies, auto thefts and home burglaries — crimes in which the department has clearance rates about half the national average. With the program, officers or supervisors will be able to focus their attention on areas where software models tell them that data show there is a high probability that crime will occur.
Police, of course, have always tried to use data to identify crime trends, and for years Miami police have done that with the data-crunching system known as COMPSTAT. Except, now, instead of identifying where crime hot spots have occurred, they’re looking at where crime will occur.
“It goes beyond just looking at crime data,” MacDonald said.
Just how much — and whether — predictive software really works remains somewhat of a question. But officers in Los Angeles say a program known as PredPol developed by the department and college professors, and now sold to departments around the country, has helped prevent and stop property crimes, and is now being tested on gun crimes.
In Atlanta, crime dropped by close to 10 percent in two zones where predictive software by the L.A.-based PredPol was put into use by cops during a three-month pilot program, according to Lt. LeAnne Browning, who oversees the department’s video integration center. Supervisors handed out maps to officers at the beginning of their shifts, and when cops had spare time they spent it in the targeted areas. The department monitored where their cops were, how much time they spent in high-risk zones, and every report that came out of the highlighted areas.
“We all thought it was somewhat hocus pocus and Minority Report,” she said, referring to the Steven Spielberg sci-fi film in which police used psychic powers to stop murders before they happened. “We could see if PredPol was predicting fairly well. It’s kind of scary, because they were.”
It’s just one example of how police are using software to solve or prevent crimes. Miami-Dade’s robbery division, for instance, uses an IBM program called Blue PALMS to solve cold cases. The software has a database of every crime ever documented by Miami-Dade police, and detectives can punch in an old crime report and get a list of 20 suspects within one minute.
“This is not science fiction,” Major Arnold Palmer says in a promotional IBM video.
In Miami, MacDonald said that the software the department is using is a more elaborate version of PredPol, which uses only crime data. Miami police also are punching in everything from paydays to school calendars, weather reports and social media. The department also is using the federal grant to establish an offender database.
Larry Samuels, CEO of PredPol, cautioned about the use of multiple variables, saying the more information you pump into a system, the more the predictions will vary. He said with Miami, as with any policing tool, predictive software will be only as good as the officers using it.
“If you do it in a haphazard way you’ll get haphazard results,” he said. “The credit for success needs to rely in the department.”