A growing number of law enforcement agencies, in the US and
elsewhere, have been adopting software tools with predictive analytics,
based on algorithms that aim to predict crimes before they happen. The concept sounds like something out of science fiction thriller
"Minority Report", the 2002 Steven Spielberg film based on a Philip K. Dick story.
Without some of the sci-fi gimmickry, police departments from Santa
Cruz, California, to Memphis, Tennessee, and law enforcement agencies
from Poland to Britain have adopted these new techniques.
The premise is simple: criminals follow patterns, and with software
-- the same kind that retailers such as Wal-Mart and Amazon use to determine consumer purchasing trends, police can determine where the
next crime will occur and sometimes prevent it.
Colleen McCue, a behavioral scientist at GeoEye, a firm that works
with US Homeland Security and local law enforcement on predictive
analytics, said studying criminal behavior was not that different from
examining other types of behavior like shopping. "People are creatures of habit," she said. "When you go shopping you go to a place where they have the things
you're looking for... the criminal wants to go where he will be
successful also." McCue states that "the technology could help in cities where tight
budgets were forcing patrol reductions. When police departments are
laying (off) more sworn personnel, they can do more with less."
The key to success in predictive policing is getting as much data as
possible to determine patterns. This can be especially useful in
property crimes like auto theft and burglary, where patterns can be
"You can build a model that factors in attributes like the time of
year, whether it is hot and humid or cold and snowy, if it is a payday
when people are carrying a lot of cash," says Mark Cleverly, who heads
the IBM unit for predictive crime analytics. "It's not saying a crime will occur at a particular time and place,
no one can do that. But it can say you can expect a wave of vehicle
thefts based on everything we know."
IBM has worked with dozens of agencies such as London's Metropolitan
Police, the Polish National Police and a number of US and Canadian
cities. In Memphis, officials said serious crimes fell 30 percent and violent
crimes declined 15 percent since implementing predictive analytics in a
program with IBM and the University of Memphis in 2006.
The program known as CRUSH -- Criminal Reduction Utilizing Statistical History -- targeted certain "hot spots" to allow police to
deploy more efficiently.
John Williams, crime analysis manager for the city's police, said the
system has had a dramatic impact, allowing Memphis to get off the list
of worst US cities for crime. "If the data is indicating a hot spot, we are able to immediately
deploy resources there. And in a lot of instances we are able to make
quality arrests because we're in the right area at the right time," he
Although beat officers can use their instincts for similar results,
Williams said the software could be far more precise, such as predicting
burglaries in a small geographic area between 10 pm and 2 am.
In one case, the software was able to help police break up a group
that was committing armed robberies on the city's Hispanic population. "There were 84 robberies, but we had no idea it was so organized," Williams said.By crunching the numbers, police were able to pinpoint the zone and
time of likely holdups: "We caught a group of robbers in progress, we
had leads on additional robberies," he said. Williams said police officials from as far away as Hong Kong, Rio de
Janeiro and Estonia have come to review the experience in Memphis.
In Los Angeles, another program developed by scientists at the
University of California-Los Angeles and Santa Clara University was
tested in a single precinct, and resulted in a 12 percent drop in crime
while the rest of the city saw a 0.2 percent increase. That test and others led to the creation of a company called PredPol.
Los Angeles will expand its use of the program under contract with PredPol, said CEO Caleb Baskin who said, "the system is based on a model from mathematician George
Mohler which is very effective in predicting the time and location for
crimes that have not yet taken place."
PredPol had begun working with other cities in California and "we've
had inquiries from a lot of places in the US and international
locations," Baskin said. "The science that underlies the tool will work anywhere. The question
is does the agency maintain a database that we can plug into."
While use of such analytics generally wins plaudits for helping
"smarter" policing, it does raise concerns about Big Brother-like snooping, Andrew Guthrie Ferguson, a law professor at the University of the
District of Columbia, said the use of technology could be positive but
that it could lower the threshold for constitutional protections on
"unreasonable" searches. "To stop you and frisk you and search you, a police officer needs
reasonable suspicion, so my question is how will this affect reasonable
suspicion?" he said. If the search is based on a computer algorithm, Ferguson said, and
the case comes to court, "How do you cross-examine a computer?"
IBM's Cleverly said the technology can in many cases improve privacy. "You can pinpoint the record of who has access to information, you
have a solid history of what's going on, so if someone is using the
system for ill you have an audit trail," he said.
As for "Minority Report" and its predictive software, Cleverly said,
"It was a great film and great short story, but it's science fiction and
will remain science fiction. That's not what this is about."