Clickability tracking pixel

Camden, N.J., Police Use AI for Proactive Policing

The department is piloting crime forecasting software that promises to better direct police patrols to the places where certain crimes are most likely to occur, specifically using ShotSpotter to detect gunfire.

by Rebecca Everett, NJ Advance Media Group / February 19, 2020
New York City Mayor Bill De Blasio looks at data from ShotSpotter's gunfire tracking technology. AP Images.com

(TNS) — Camden County, N.J., Police Chief Joseph Wysocki said that when he first became an officer in the city nearly three decades ago, he was patrolling the streets “blind.”

“I had no intel. I didn’t know what I was going into” in a given neighborhood, he said.

It hasn’t been like that for Camden cops for years. Since 2013, the county-run force has relied on civilian crime analysts to direct patrols based on their predictions of where crimes might occur.

Now when you walk in the department’s command center all but the watch commander are civilians, contracted through a third-party to work as crime analysts and “virtual patrolers.” Each person sits before six computer monitors displaying video feeds from surveillance cameras, maps, real time police call data, and the intelligence they’ve collected to make sense of it all.

Wysocki, who succeeded J. Scott Thomson as chief in September, radiated pride as he discussed the evolving work of the analysts. But for him, there’s no reason to leave well enough alone.

That means trying something new: artificial intelligence.

“Our goal is to reduce crime. We’re at a 50-year low but if you look at 2019 numbers to 2018 numbers it’s single digit reductions,” Wysocki, 49, of Voorhees, said in an interview last week. “I don’t want to say we’re at a plateau, but what could we do to get them lower?”

To that end, the department is piloting crime forecasting software that promises to better direct police patrols to the places where certain crimes are most likely to occur. It is offered by ShotSpotter, the same company contracted by the county to use sensors placed around the city that detect gunfire, and alerts police in seconds.

“I’m interested to see how the computer will do against our analysts,” Wysocki said. “I have an open mind.”

In law enforcement, predictive policing has been touted as a smart way to be proactive and deter crimes before they happen, but groups including the ACLU have pushed back on the idea. Among their criticisms is that since the systems rely on data from past police responses, they can create a feedback loop, continuing the over-policing of low-income or minority neighborhoods.

It’s not clear what that might mean in a city like Camden, which is fairly unique in that its residents are mostly low income and approximately 96 percent minorities. The city is also dealing with a higher crime rate than most big cities, and has many abandoned properties.

ShotSpotter calls its crime forecasting technology Missions, but Wysocki has dubbed it “directed patrols” because he wants to avoid anything that sounds militaristic. Over the next three months, he will see whether the software can do a better job of suggesting when and where those patrols should be.

Sam Klepper, a senior vice president at ShotSpotter, said that in addition to directing patrols, the system also tracks the tactics used and the impact of the patrols, so they can be analyzed to see if they’re actually reducing crime.

ShotSpotter says that its system uses data including past police calls and crime data, the time, day and season, plus environmental factors like the density of vacant properties or bars.

It also uses “socioeconomic indicators” but the company would not be specific about what data it pulls from five-year census data estimates. Klepper did say the company does not use racial, educational or home ownership data and never uses personally identifiable information.

He said the company has taken steps it hopes will minimize bias or discrimination, such as randomizing patrol assignments “to avoid oversaturating high-risk areas.”

The company has also adopted a new policy to “not allow the system to create risk assessment models for crime types that are susceptible to enforcement bias. This can occur when police presence in an area can lead to more crime being recorded in that area which in turn can create runaway feedback loops that will continually return police to the same area.”

Wysocki said ShotSpotter asked his department for five years of police call and dispatch data, and the company isn’t using any information about specific residents, such as whether a neighborhood has a certain number of parolees or sex offenders.

It’s not clear what kind of investment it would be for the Camden County Police Department to use the crime forecasting software past the pilot period. Klepper declined to release pricing information and county spokesman Dan Keashen said ShotSpotter has not given the department any estimates.

The county currently spends roughly $350,000 a year for the company’s gunfire detection services.

While some police departments have tried data-based forecasting models that try to say who is likely to commit crimes, ShotSpotter’s focus on when and where to patrol is in line with Wysocki’s values. He said he wants cops out walking the beat, noting that a regular foot patrol can sometimes drive drug dealers off a corner better than an arrest can.

“Cops belong out on the street,” he said, even in the dead of winter. “That’s what we’re preaching. That’s what we’re asking them to do.”

And contrary to his experience as a 20-year-old rookie, Wysocki said his officers are hitting the street with the intelligence to prepare them for what they might face.

“They have the answers to the test,” he said.

©2020 NJ Advance Media Group, Edison, N.J. Distributed by Tribune Content Agency, LLC.

Looking for the latest gov tech news as it happens? Subscribe to GT newsletters.

E.REPUBLIC Platforms & Programs