Predictive Policing — the use of sophisticated data analytics to prevent criminal activity — is gaining traction in cities across the country. Just yesterday, The Atlantic Cities published a piece on how Twitter could help police predict crime.

As the practice has spread, concerns about potential misuses or abuses, particularly in low-income communities and communities of color, have also grown. The Chicago Tribune wrote last year of a young man visited at home by a police officer warning him that he was being watched because he is on a violent crime “hot list.” The Tribune story has recently been picked up by The Verge, the ACLU and others. New York City’s deeply controversial “Stop and Frisk” policy, curtailed this year under new mayor Bill de Blasio for disproportionately targeting minority communities, was defended by former commissioner Ray Kelly because “that’s where the data says the likelihood of crimes is highest.” Concerns have also arisen around the validity of the underlying data: for example, the police department in Milwaukee was alleged in 2012 to have downgraded offenses in order to create the appearance of a reduction in crime (see this provocative piece by Alex Howard for more on this and related issues.).

Accounts like these feed into a longstanding mistrust and resentment of law enforcement in low-income communities, and a painful narrative of navigating life as a person of color in America, described poignantly by Charles Blow last year in the New York Times. Police officers also suffer from these perceptions: no one wants to come to work feeling like they are in hostile territory, on guard for their lives. Addressing these concerns is vital to addressing disparities in Americans’ life chances, as well as to healing divisions between low-income city residents and those who’ve sworn to serve and protect them.

However, discarding predictive policing altogether, even if it were possible, would not only fail to solve the broader problem, but would also cost communities real gains in public safety. The Ceasefire model (now known as Cure Violence) has supported double-digit reductions in homicides in many cities by treating violence as an epidemic, like HIV-AIDS, and using highly targeted interventions to prevent outbreaks. Local Police departments nationwide have also used predictive analytics to prevent burglaries, vehicle thefts, and other crimes.

Beyond crime prevention, we here at Living Cities and others in our networks have seen powerful examples of data supporting meaningful improvements in the lives of low-income people. Local governments have used data analytics to prevent deaths from apartment fires, reduce rodent populations, and increase the use of city services in low-income communities. These same data have been opened to the public as part of a movement towards government as a platform for collaborative problem solving.

Parallel to the rise of government as platform has been the emergence of Collective Impact as an approach for large-scale change. Sparked by the Strive Partnership, which has measurably improved student outcomes in Cincinnati and Northern Kentucky over a period of six years and garnered the interest of hundreds of American cities, a Collective Impact field is growing to help leaders across sectors use data to align their investments in order to move the needle on seemingly intractable urban problems.

These experiences suggest that there may be benefits to “open-sourcing” predictive policing. That is, making it possible for law enforcement officials, communities and other relevant stakeholders to align their efforts, supported by data, towards both improved safety outcomes and a better citizen experience of law enforcement.

Open-sourcing predictive policing could involve the following actions:

Opening the data: Keeping predictive policing data behind a curtain not only excludes communities from decisions that affect their quality-of-life, but also denies police departments the wisdom of crowds, the “ground truth” communities bring to the data, and the ingenuity of innovators who could use the data and information technology to create entirely new approaches to public safety. One example of the potential of this approach is the work of Urban Strategies, a community-based data and advocacy organization, which used police data to challenge, and subsequently improve, the city’s preventive policing strategy.

Bringing all relevant stakeholders to the table: Police departments can join with communities, nonprofits, funders, public safety experts and even other city agencies to analyze and interpret relevant data, and to develop coordinated prevention strategies. New partners to the table can also bring new datasets, different context for interpreting the data, and new ideas for preventive approaches that together create new possibilities for insight and innovation. Not least of these new possibilities is the potential for action beyond enforcement activities, such as rehabilitating or demolishing “hotspot” vacant properties, renovating public parks that facilitate criminal activity, or offering likely offenders intensive supportive services. Cure Violence and LISC’s Community Safety Initiative demonstrate the power of these kinds of partnerships.

Building the “infrastructure” for shared accountability and continuous improvement: Open-sourcing predictive policing may require new capacities. For example, as John Tolva, Chicago’s former Chief Technology Officer, told me recently, the City of Chicago has opened vast amounts of crime data, but entities like newspapers, intermediaries like CrimeLab at University of Chicago, and civic app developers have played critical roles in making that data accessible to the public. Collective Impact partnerships, meanwhile, rely on “backbone organizations” like Strive to facilitate collaborative data analysis, help manage conflict between stakeholders, and use methodologies like Six Sigma to drive continuous improvement. Rigorous, participatory evaluation of predictive policing strategies can also shed light on whether preventive strategies are producing desired results, and at what cost.

When it comes to predictive policing, and to debates around law enforcement tactics more generally, low-income communities shouldn’t have to choose between feeling safe and feeling fairly treated. Learnings from the fields of municipal innovation and collective impact could help change the way cities and communities interact around public safety, and maybe even facilitate a little healing in the process.

Special thanks to John Tolva and Steve Spiker for feedback on earlier versions of this piece.