The city is using a type of predictive policing called risk-terrain modeling, which aims to help police identify places that attract crime and intervene to make them less attractive to criminals.
(TNS) -- Atlantic City is not where you might expect to find the next policing success story.
It’s in the throes of a financial crisis that triggered a state takeover. The police force has been pared down to 267 officers from 374 in 2010. Even more severe cuts could be on the way.
And yet, lately, Police Chief Henry White Jr. finds himself optimistic.
“The first six months of this year, our violent crime is down about 20 percent compared to the same time last year. But at the same time our arrests are also down 17 percent,” he said. “We were able to reduce crime without contributing to mass incarceration.”
That’s a first for White, an old-school cop who started on foot patrol in 1985 and rose through the ranks.
What gives him hope is a high-tech intervention his department has adopted with help from Rutgers University criminologists. It’s a type of predictive policing called risk-terrain modeling that aims to help police identify places that attract crime, and intervene to make them less attractive to criminals.
Predictive policing has become a buzzword in recent years, as a growing number of startups have marketed competing software options to police forces around the country. But civil-rights advocates have raised alarms. A coalition of 16 groups including the ACLU and the Brennan Center for Justice last August joined in a statement warning that such tools exacerbate racial biases, ignore community needs and contribute to the over-policing of poor minority neighborhoods.
But Joel Caplan, a wiry, bespectacled professor from the Rutgers School of Criminal Justice, believes he’s cracked that problem with risk-terrain modeling (RTM for short), which he developed with another professor, Leslie Kennedy. And, unlike tech companies that charge tens of thousands of dollars, Caplan’s giving away his software to any department that will use it.
White had experimented with data-driven policing before. But looking back, he said, “we were just playing games.”
Now, after less than a year using RTM, even intractable hot spots like Stanley Holmes Village, a housing project that has seen numerous shootings, are starting to calm, he said.
“We’ve had certain neighborhoods in town that have been hot spots since I was on patrol. We made a ton of arrests. But, you know what? They were still hot spots until recently,” he said. “Before, we would clean up an area temporarily but all we were doing was displacing crime.”
Part of what’s radical about RTM is that many interventions it advocates are not about policing at all — or, at least, not policing as we normally think of it.
Police still have to investigate crimes and make arrests, but the focus of RTM is creating conditions that deter crime from occurring.
The first step is to identify those conditions. Caplan often uses an analogy to make this point.
“If you noticed kids playing at the same place over time, you might identify that as a hot spot of playful behavior. But if you take your focus off the kids and look at the environment, you might see slides and swings and open fields — what we might call a playground. If we can identify places with similar characteristics — swings, slides, open fields — we can expect playful behavior to occur.”
Adding or removing those characteristics would in turn influence the behavior.
So, when it comes to hot spots of criminal behavior, he said, “the response isn’t to assume that people who are located there are bad, or criminals” — the answer wouldn’t be, say, to stop-and-frisk passersby — “but to do things to make the environment less attractive to crime. That’s how we can reduce crime, and we can do it without arrests.”
An early success applied RTM to cellphone robberies in Glendale, Ariz. It found they clustered near convenience stores, which had kiosks offering instant cash for phones. Police asked store owners to move the kiosks to the front of the store, near windows and surveillance cameras. Robberies fell 42 percent.
Assessing Burglary Risk
Software developed by Rutgers University uses a type of analysis called risk-terrain modeling to assign areas a relative risk value based on the presence of landmarks and features such as vacant lots, schoolyards, and take-out restaurants. Police in Atlantic City used this software to map burglary risk after a number of break-ins this spring. Two suspects were arrested for a series of burglaries that were committed mostly in an area that the software showed had the highest risk for that crime.
SOURCE: Atlantic City Police Department
Adoption of RTM has been slow, Caplan said, in part because he doesn’t market it. It’s word-of-mouth. But he believes it could be used to respond to everything from Philadelphia’s opioid crisis to traffic accidents to border security.
Atlantic City police applied RTM to shootings and homicides, mapping the crimes, then all the landscape features around them. Then they began holding monthly meetings with police and community stakeholders.
“We found that convenience stores, laundromats and vacant properties — where all those things were located together, they increased the risk,” Caplan said. Because most shootings were drug-related, police theorized a narrative: Convenience stores were where buyers were solicited, unsupervised laundromats became host to transactions, and vacant properties doubled as stash houses.
Then, he said, “we addressed that narrative.”
That didn’t just mean sending in more patrols. Instead, they treated the landscapes.
They put sign-in sheets inside convenience stores and laundromats, so patrols could log their visits. They helped business owners get security cameras. And they shared the information about problematic vacant properties with Atlantic City code enforcement, so they could prioritize which vacant buildings to board up, which lots to clean, and which negligent property owners to push into compliance.
Even if no crime has occurred in a given spot, White said, “we’re getting there before the crime does to make sure that now it is not the next hot spot. We’re no longer playing Whack-a-mole.”
It’s not just Atlantic City. Across the country, big data is already transforming crime-fighting.
Departments began mapping hot spots a few decades ago, identifying areas where crime spiked and flooding them with patrols and, often, a zero-tolerance approach.
“They would go out there and stop everybody for everything: riding your bike without a light, or jaywalking. Sure, you’re going to make some arrests: you stop a hundred people and chances are somebody’s going to have drugs,” White said. “But it turned the community off. There was a high cost.”
Today, those maps are being replaced by slick software programs fed with real-time data and marketed by for-profit companies.
The leader is PredPol, which generates heat maps based on past incidents.
But one study found that if deployed in Oakland, Calif., it would concentrate forces in low-income communities of color. That potential was concerning enough to Oakland police that, after consideration, they decided not to adopt the software.
Still, other cities are pressing ahead, about a dozen of them with HunchLab, a software created by a Philadelphia company, Azavea. Philadelphia Police recently tested HunchLab here, but declined an interview request, saying the research is not yet complete.
Cities pay between $20,000 and $100,000 per year for the service, depending on their size.
In return, the computer model combines crime data (police incident reports and calls to 911 for service), geographic traits like nearby bars or vacant buildings, and factors like the weather and school schedules. It generates a map dotted with boxes — risky zones — and a screen encouraging an officer to visit a box and perform some intervention: for example, walk a 15-minute foot patrol.
In Greensboro, N.C., police tested HunchLab by giving officers on one shift the software and providing another group with their existing hot spot maps. Major crimes fell by about 30 percent with HunchLab.
That impact was startling, given that the local officers didn’t think the software had any effect. “The officers didn’t see any value in it,” said Jeremy Heffner, HunchLab product manager. What it did do, he concluded, was force officers to spend 15 minutes patrolling every side street and back alley of a compact area.
“The HunchLab boxes broke them out of their routine. It shows that such nuanced changes, even that we don’t know we’re making, can have an impact.”
Heffner is keenly aware of civil-rights concerns; he believes he’s addressing them. For one thing, HunchLab doesn’t use arrest data, or any data that would create a feedback loop by interpreting elevated police action as elevated risk. For another, it moves officers around, so residents don’t feel harassed.
But, unlike RTM, HunchLab’s approach to overcoming bias is to provide less information, not more. It doesn’t tell police why a box is selected or even whether it’s a high- or average-risk zone. And while RTM focuses on the why, HunchLab works off the notion it’s better for officers not to know too much.
“This shouldn’t be justification to stop someone,” Heffner said. “That’s why we actually hide a lot from the officers. We don’t tell them the likelihood that a robbery will happen, because people get hung up on probabilities. The goal isn’t to go into these places and make a bunch of arrests. The goal is to have nothing happen.”
To Caplan, who was a Cape May police officer before he went to the University of Pennsylvania to study criminology, concealing information is a poor workaround.
He believes the way forward is transparency: “It’s knowing where to go, knowing what to focus on, knowing why you’re doing what you’re doing, being able to explain it to the community and get feedback — it’s that transparency that helps to reduce bias and improve community relations.”
White is trying to build those community ties by combining RTM with old-fashioned policing tactics. He has added foot and bike patrols, which are popular with residents, using RTM to prioritize patrol locations.
There are hurdles, though.
“The biggest challenge we’re having here is getting the rank-and-file to buy in,” he said. “We’re trained to focus on the bad guys, and it’s hard for them to make that shift.”
Then there’s the hard work of winning over Atlantic City residents.
Kellie Cors-Atherly, who runs Peace Amongst Youth, a support program for crime victims in the city, said she hasn’t noticed the impact of RTM yet.
“We’re having active shootings in Atlantic City once a week. That doesn’t seem to be changing so much. And it’s still a division between the police and the community. There’s still a big trust factor,” she said.
And for activists like Steven L. Young, who is affiliated with the National Action Network, community-police relations can’t be solved without more sweeping reforms.
“They have a track record of so many officers abusing and beating down the community, and they’re still allowing them to be on the streets doing the same thing. The relationship remains the same.”
Still, some business owners have noticed an impact.
Sammy Nammour, who manages 13 Cedar Food Markets in Atlantic City and Pleasantville, said the daily police visits have helped restore order.
“There’s been multiple occasions where someone would come in and scream and shout and to walk out with something,” he said. “It happens a lot less often now.”
He thinks the kids who frequent his store also benefit from casual, friendly interactions with officers.
After all, the dream of predictive policing is that it can change the climate, and eventually overcome persistent police biases.
White sees that promise in RTM.
“A good example is the perception that housing projects are connected with crime,” he said. “What we found in Atlantic City is housing projects are not in fact correlated with crime, and if police were to focus their patrols in those areas, they would be in the wrong place.”
©2017 Philly.com Distributed by Tribune Content Agency, LLC.
Looking for the latest gov tech news as it happens? Subscribe to GT newsletters.