Arresting Crime Before It Happens

In Chicago, cops try a ‘guardian’ approach armed with new crime-prediction methods and non-punitive social intervention.

Police investigate the scene of a shooting in Cornell Square Park on the South Side of Chicago. (Getty Images)

What if you could predict which people were highly likely to be part of violent crimes in the near future? What if you could then give those people concrete assistance in the form of social services, job training, childcare, and other help that might reduce the pressure on their lives? If Amazon can suggest books you might like to read, why can’t police suggest routes out of dangerous social contexts?

These are not rhetorical questions; local and national governments are not only asking them, but seeking answers. Those answers are desperately needed in the current troubled context of police-community relationships. Earlier this year, President Obama’s Task Force on 21st Century Policing released a report recommending that officers embrace a “guardian” rather than “warrior” mindset in order to “build public trust and legitimacy.” One form guardianship can take is intervening with offers of help before violence occurs.

Maybe some answers will come in Chicago, a city known for tough cops coping with tough jobs. (As I publish this, it’s also coping with the consequences of a damning video showing a policeman shooting Laquan McDonald, a black teenager who did not seem to present a fatal threat.) The Chicago Police Department has a long history of using data in the service of modern policing, and has been experimenting with intervention tactics for the last few years. A record number of homicides in Chicago — 56 alone in September 2015, the highest monthly number since 2002, and increases in recent yearly totals — has triggered intense interest in guardian-like intervention. At the same time, the Ferguson and Eric Garner stories — as well as many others — have made Chicago, like all American cities, sensitive to using race, ethnicity, or geography as a proxy for likelihood of involvement in violent crime. And the department’s limited resources make it essential that social-service interventions be targeted towards the people who need this assistance the most.

Now the Chicago Police Department believes it has found a method for predicting the likelihood that an individual will be involved in violence — as either victim or perpetrator — that relies solely on information drawn from that person’s criminal history. Using this model, developed by a team at the Illinois Institute of Technology led by Dr. Miles Wernick, the Chicago police can aim their non-punitive intervention efforts effectively at particular people.

The idea of intervening rather than punishing is not new. David Kennedy’s work with the group he co-founded in 2009, the National Network for Safe Communities, has been lauded for helping cities use strategic interventions to reduce violence and improve public safety — as well as to improve relationships between law enforcement and the communities it serves. But traditional interventions focused on particular geographies, mapping predictions of crime against neighborhoods, led to deep concerns about the use of predictive policing in communities of color.

What’s new about the current Chicago Police Department story is that it focuses on people rather than places. Backed by a series of National Institute of Justice grants over the last few years, the CPD and its academic partners have developed a dashboard for initial identification of people who are likely to be involved in a violent crime. The basic variables used in the model have changed over time, but they all relate to a subject’s criminal history. These days, they might include the pace of arrest activity — “is the arrest activity increasing over time?” — as well as the numbers of crimes in which the subject has been involved, arrest-record associations with other violent criminals, and the subject’s types and intensity of criminal history. The algorithm does not use any information about the subject’s home address, race, or gender.

According to Chicago Police Department Deputy Chief Jonathan Lewin, when tested against historical data, the model shows that if you are one of the top 20 subjects in your police district, “You have a one in four chance of becoming a party to violence within the next eighteen months.” He continues: “The model can’t necessarily tell if you’re going to be a victim or an offender. It just knows that you’re going to be a party. If you compare that prediction to one for another criminal, your chances are over five hundred times greater than they would be for a general member of the population or another criminal would be.” The CPD is continuing to build real-time information into the dashboard, for things like whether the subject has been the victim of violence recently or is wanted for questioning, to allow the model to evolve. And Rand is currently carrying out an in-depth evaluation of the program.

The model won’t directly trigger police action — its results are always vetted by human analysts who work with district commanders and staff. For a few years now, the CPD has used the lists generated by the model to bring subjects into meetings — “gang call-ins.” According to Chief of Crime Control Strategies Robert Tracy, the CPD tries to tell them during these meetings, “‘Leave the life of crime. Put the guns down. We want the community better.’”

But CPD has recently moved to an increased emphasis on what it calls “custom notifications.” The first step: a letter is drafted to each subject who remains on the list. It’s a long letter — it explains how the subject was identified, what the risk is, lists prior arrests, and describes intervention services that are available. “We actually show them what type of crimes they’ve been involved in and how this puts them at a greater risk [of involvement with gun violence] than a normal citizen in the City of Chicago,” says Tracy.

The letters are hand-delivered by district commanders with a knock on the door. Sometimes the individual on the list isn’t there when the delivery happens. Tracy says that can be helpful: “It’s a successful [custom notification] where you can get people, their loved ones, and people that have influence in their life or part of their life involved. Sometimes those people are oblivious to some of our concerns and the safety of the individual that’s involved and what gang and who they’re with. To bring them into the fold, it’s actually very good to give them options and maybe they can get that message to the individual.” He adds, “There are some family members that say, ‘I can’t do anything with them. Can you help us? We’ve tried and this person’s not listening.’”

And there are a lot of options on offer. Lewin lists the services: “Social services, job training, childcare services. If they’re an offender that’s coming back to society, offender re-entry programs.” There are mental health care services; drug-addiction treatment. Tracy says, “If you don’t have a GED we’ll get you back to school.” He adds, “So, even for the people that have the highest propensity for violence, we’re looking at alternative ways to try to ensure that we can try to keep them out of crime or at least let them know, ‘Hey, we’re looking at you, but we want to give you alternatives.’” And there’s a stick as well as a carrot; the letter describes the enhanced penalties that may apply if the subject is arrested again in connection with violent crime.

Tracy is hopeful that this approach, along with beat policing and other changes, will contribute to building trust of the CPD in communities where the police have not always been trusted. It’s a process. He says, “When we knock on a door, and we have a commander, and we have these services, they’re initially very wary.” Tracy believes that when the families involved start seeing that the CPD is not there to arrest the individual, and is offering services to the family as well as the individual, that has to help.

“Sometimes once they find out we’re not there to arrest, somehow that individual shows up and sits in and he might have been next door or somewhere else and someone’s able to text him that this is not about an arrest,” Tracy continues. “They will come to find out what’s going on.” And sometimes they’re surprised, he adds. “Some of the subjects don’t even realize how many times they’ve been arrested or what counted as an arrest, like narcotics possession. In some of their worlds it’s educating them as well.”

When asked whether the factors being used in the model are themselves proxies for race, Lewin is ready with a response: “This is about saving lives,” he says. “This is not designed to be a punitive effort. This is designed to be an effort that will save lives through interventions… It’s trying to do that in the most effective way possible.” And, he notes, use of the model will direct resources towards intervening in the lives of these subjects to fend off violence. “Every CompStat meeting, district commanders are shown a list of their riskiest subjects, and this list shows how many have been custom notified,” he says. This is the first time the department has implemented any kind of predictive effort about people. “I think this is state of the art for predictive policing,” Lewin says.

Tracy is proud of what’s happened so far: “Sometimes we get some very good results,” he says. But these notifications aren’t everything; Tracy says they are “one piece of a very comprehensive strategy that we have here in reducing violence and reducing crime in the city of Chicago.”

The question for the future: how will this form of predictive policing be received? Will custom notifications become part of the standard set of steps taken by police officers around the country? Or will they be perceived as intrusive and threatening? At this point, it seems mandatory to invoke Minority Report, the 2002 movie based on Philip K. Dick’s vision of police arresting people for crimes that have yet to be conceived, let alone committed.

The police brass in Chicago are certainly aware of this movie, and that’s why they take pains to explain that the algorithm alone doesn’t trigger actions. Overall, the CPD seems confident that it is doing the right thing. And it may be that if these models persuasively avoid reliance on race or geography while preventing yet another round of gun battles — maybe even events like the killing of Laquan McDonald— Americans will understand and appreciate their use.