Luckily, first impressions aren’t everything.
It was mid-July when the team of researchers from the University of Chicago arrived at the Charlotte-Mecklenburg Police Department. Their goal: To work with CMPD in devising a better way to predict and head off police misbehavior that put the public and officers at risk.
The two had been paired by a White House data initiative for safer policing. But the timing could not have been worse.
Departments nationwide were answering for a series of controversial deaths of black men at police hands. In Charlotte, the city tensed under the strain of the approaching manslaughter trial of Officer Wes Kerrick, charged with killing unarmed Jonathan Ferrell. While the trial ended in a hung jury and the charges were dropped, Chief Kerr Putney says the case almost tore CMPD apart.
In short, it was far from the best time for a group of strangers to expect cops to talk openly about how they did their jobs.
“I wasn’t very warm to the idea,” says Putney, an assistant chief at the time. The rank and file? “They were much more hesitant than I was.”
But something began to click. Over the two-day visit, the research team led by Lauren Haynes and Joe Walsh of the Center for Data Science and Public Policy began having better conversations with CMPD personnel. They rode patrol shifts, talked with Internal Affairs. And once CMPD was convinced their officers’ identities would be protected, the department began turning over more and more info on arrests, traffic stops, dispatches, discipline and just about everything else CMPD gathers, which is a lot.
“Our partners are light years ahead of just about every department in the country in terms of data,” Walsh says.
All went into the team’s new model on how to predict and prevent “negative interactions.” As of now, the model has some 300 different predictors factored in. Before, CMPD’s “Early Intervention System” was built around time thresholds – three uses of force by an officer within a 180-day period would set off an alarm. All sides say the criteria was too broad, leading to too many officers who had done nothing wrong being placed under suspicion.
The new model factors in everything from part of town to time of day. Even the weather or the officer’s private life is considered. The most telling factor so far: Has an officer used force in the past?
So far the test results have been good – with more high-risk officers and fewer low-risk ones being flagged.
Putney appears on board.
“This gives us a more holistic picture,” he says. “We remain on the cutting edge trying to make sure we are able to weed out bad behavior but we’ve cut down on false positives. You don’t want to give the impression that officers are doing something wrong when they really aren’t.”
Police already are considering a real-life benefit from the partnership. As of now, CMPD sends two officers to most domestic abuse calls. But based on how many of those incidents end up in a fight, Putney says the department may soon add a third or fourth officers in hopes of deescalating the situation from the start.
The goal is to make all sides safer, not winning the fight, Putney says. “We hope we don’t have the fight at all.”
Haynes says researchers hope their data will influence training, identifying areas where officers need more expertise.
Meanwhile, the group continues to tinker – to bring their model in line with the complexities of an officer’s job. During the ride-alongs, the research team was surprised by how many things the officers had to deal with – and how much time they spent helping the public.
“We had nothing but professional interactions with all of them” she saidys. “CMPD really is the gold standard.”