Understanding the root causes of crime in urban areas is critical to effectively addressing this serious problem. Variation in crime levels across and within cities offers important clues. Some heterogeneity may be related to physical and social characteristics of urban environments, such as social cohesion, physical disorder, and greenspace. In addition to sociological and economic factors (e.g., racial tension, job opportunities, etc.), aspects such as the amount and character of street activity also impact crime. This project proposes new methods including causal inference modeling and machine learning (ML), combined with new data sources such as imagery and sound from the NSF-funded Array of Things (AoT) project, to explore three time-varying neighborhood factors on crime: 1) the amount and usage of greenspace, 2) the amount and character of street activity and related social cohesion; 3) the level of visual and auditory disorder; as extant research suggests their importance in relation to crime and their potential for intervention. These discoveries could transform social science research by measuring complex social and physical environment variables at scales never before investigated. This will also push the boundaries of ML algorithms embedded in intelligent distributed sensor networks. This project will develop novel ML algorithms that quantify the quality of social interactions, which thus far have not been explored. Convergence among scientists, residents, and organizations that comprise neighborhoods will increase community understanding and acceptance of "smart" technologies for connected communities to address social science questions. Further, the project has direct impacts on increasing diversity in STEM fields through partnerships with community organizations Colony 5 and MAPSCorps, which provide hands-on training of minority youth across Chicago to use Internet of Things (IoT) and urban data to understand and improve their neighborhoods. Results from this project could lead to new smart city technologies and the ability to map important social behavioral variables such as social cohesion across entire cities. This work may then inform interventions that could be used to increase social cohesion, potentially through urban greenspaces, which could lead to reductions in crime and overall increased well-being for urban residents.

Measuring street activity and social cohesion is difficult, typically requiring human observation or surveys, which are taxing on observers and limit the number of neighborhoods that can be studied. This project proposes to use recent technological advances to measure human behavior and social interactions en masse. Smartphones can provide mobile testing labs, facilitating new forms of surveys measuring the cognitive impact of a person's urban exposure over time as well as tracking mobility anonymously. Combined with advances in ML, this opens new potential for exploring human interactions in cities. AoT will provide images and sound from some 200 Chicago locations, using technology that can be readily extended to more fully instrument selected experimental venues, exploring the use of ML "at the edge" to quantify and characterize street activity, disorder, and greenspace use across diverse neighborhoods. This project will relate these measures to crime, using Chicago's open crime database, to examine the physical and social environmental factors that may influence crime and its mitigation.