‘Eyes’ on the Street: How Computer Vision and Cognitive Psychology Can Help Us Get the Gist of Neighborhood Environmental Design and Explain Crime

• In Event: Using Computer Vision to Analyze Urban Images and Answer Crime & Place Questions

Abstract

For half a century, scholars of crime have theorized that the spatial attributes and layouts of places shape how likely people are to take action against perceived crime threats. The crime prevention through environmental design framework suggests people have heightened territoriality in areas with visually open, unobstructed spaces that are aesthetically pleasing. However, measures of these constructs have proved elusive, so studies testing this idea have generally been limited to data measuring the presence of objects such as vegetation or trash that impede sight-lines or degrade aesthetic value. This represents a substantial limitation, as research in cognitive psychology suggests that people make behavioral decisions by rapidly assessing the ‘gist’ of entire scenes rather than scanning specific objects. By training a computer vision AI model to rate several forms of scene gist for ~200k georeferenced Google Streetview images, this project introduces a strategy for measuring aesthetic value and natural surveillance quality across Chicago neighborhoods. Using data from Chicago’s 311 system to measure how likely neighborhood residents are to report man-made incivilities, this study explores the relationship between neighborhood-level visual characteristics, territoriality, and crime.

Authors

• Riley Tucker, University of Chicago

• Gaby Akcelik, University of Chicago Environmental Neuroscience Lab

• Nakwon Rim, University of Chicago Environmental Neuroscience Lab

• Alfred Chao, University of Chicago Environmental Neuroscience Lab

• Elizabeth Gaillard, University of Chicago Environmental Neuroscience Lab

• Hassaan Haq, University of Chicago Environmental Neuroscience Lab

• Marc Berman, University of Chicago Environmental Neuroscience Lab