Shape-Up

Urban gun violence is a persistent public health crisis, especially in the United States. Researchers have found that simple physical improvements, such as cleaning up abandoned properties, can reduce nearby gun assaults by up to 40%. While community organizations and local governments routinely conduct maintenance, repair and beautification, these efforts are almost never strategically directed to prevent gun violence.

Shape-Up is a web application for prioritizing and coordinating physical cleanups to maximize gun violence prevention. Shape-Up’s analysis combines (1) artificial intelligence, which learns to spot high-risk landscapes using urban imagery and violence data; and (2) community reports of unfavorable neighborhood conditions.

By finding the high-risk landscapes where residents have requested fixes, Shape-Up's insights are actionable, data-driven, and responsive to community members. Scaled widely, Shape-Up will recast gun violence prevention as a shared endeavor among community residents and across city agencies, bringing solutions within reach.

Within cities, gun violence risk varies substantially between neighborhoods and even from one block to the next, due in part to differences in physical condition. For instance, gun assaults are more likely to occur near vacant lots and abandoned housing. Vacant spaces are convenient places to store guns and conduct drug transactions. Disrepair conveys that a place is not well-guarded or cared-for, weakening the social forces that keep violence under control. Still, most cities treat violence prevention as a policing issue, ignoring these upstream factors.

Cities have data on where violence occurs and where disrepair is worst.  Moreover, addressing disrepair is cost-beneficial: one study found that turning unkempt vacant lots into small, simple parks returned $26 to taxpayers and $333 to society for every dollar invested, in gun violence reductions alone. Nonetheless, few cities have adopted strategies to stop gun violence through physical maintenance, remediation and beautification. One barrier is analytic: even in the most data-driven cities, no validated tool exists for identifying locations where physical improvements will best address violence. Another barrier is organizational: strategic cleanup decisions rarely occur, and when they do, it is behind closed doors, where community members are neither engaged nor informed.

Urban gun violence disproportionately affects young men of color. For instance, African-American children are 10x more likely to die by firearm assault than their White counterparts, and homicide accounts for half of all deaths among African-American males ages 15-24. The effects of community violence extend far beyond the individuals injured by firearms and their loved ones. For children, exposure to community violence causes negative short- and long-term health effects; for elders, fear of violence can severely restrict daily activities. By helping to rebuild safe neighborhoods, Shape-Up aims to benefit all members of the communities most affected by gun violence.

Shape-Up directly engages these community members in gun violence prevention. Most larger cities already solicit resident reports of maintenance problems such as overgrown weeds, illegal dumping, and sidewalk disrepair, through non-emergency reporting (“311”) systems or outside providers. Ordinarily, these reports are routed to the appropriate agency and addressed in turn. Shape-Up amplifies reports from places where the physical environment is most conducive to gun violence--i.e., where an enhanced response will matter most to safety. Moreover, the platform enhances accountability to residents by displaying project roles, objectives and progress.

Shape-Up is a web app that identifies high-priority locations for physical improvements and creates a space for coordinating the community-wide response. In brief, Shape-Up analyzes data from multiple sources, then produces a Hot List of locations where physical improvements are most urgently needed to address gun violence risk. Shape-Up highlights these locations on an interactive map. For these Hot List locations, Shape-Up displays project goals, roles and progress.

Shape-Up provides two resources that cities currently lack: an analytics tool linking gun violence and neighborhood environments, and a platform for sharing information about ongoing efforts to prevent violence through physical improvements. The analytics tool employs artificial intelligence algorithms, specifically convolutional neural networks (which derive environmental information from aerial/satellite imagery) and machine learning regressions, which determine associations between landscape features and previous gun violence incidents. The algorithm assigns an environmental risk score to every street block in the city; high-risk locations are those where the landscape is particularly conducive to gun violence. Next, the analytics tool cross-references these scores against recent reports from residents of problems requiring cleanup, such as overgrown weeds, illegal dumping, or sidewalk disrepair. Locations with high scores and serious reports are selected as Hot List locations.

Our goal is to catalyze effective, coordinated action at these Hot List locations. Improvements could include mowing, repainting, or boarding-up in the short term; in the longer term, they could include demolishing an abandoned structure or installing public art. Many cleanups will be complicated undertakings; while Shape-Up provides a Comments section for each Hot List location, most coordination and decision-making is expected to occur offline. Once decisions are reached, Shape-Up will display the lead partner(s), the action steps that are planned, and progress against those benchmarks. Once these action steps are complete, Shape-Up will mark the cleanup as “complete” and will continue to track resident reports and violence incidents to ensure that the area has experienced a lasting improvement. 

Shape-Up uses AI to derive information from urban imagery on a large scale and to identify patterns in the physical environment associated with gun violence risk. Other applications of AI in this realm, such as crime prediction and facial recognition, raise ethical and constitutional concerns because they are used punitively. In contrast, Shape-Up uses machine learning to foster collaboration among city agencies, community members and other partners. 

We are eager to join the ranks of Solve teams using AI and machine learning to develop sophisticated models and address complex problems. We will use the AI Innovations Prize to continue refining our techniques, incorporating new data sources and new analytic techniques. One particular area of research interest for us is the possibility of using deep learning to reimagine how specific places could be made to look, using techniques such as convolutional autoencoder models, in order to reduce violence risk.

We will use data and technology to address the upstream drivers of urban gun violence, specifically physical deterioration in the built environment. This approach is grounded in peer-reviewed research studies from several U.S. cities finding that physical improvements reduce nearby gun violence. By reducing community violence, these interventions will also reduce exposure to violence among children and teens. If selected for the Everytown prize, we will use grant funds to improve our user interface and continue to seek partnerships to expand access to this product.