Your perception of any city or neighborhood is largely determined by things you can't quantify, like the vague feeling that a place just seems friendly, or clean, or well-lit. So much of our experience of cities is subjective like this. But if officials could figure out how to measure otherwise emotional or intuitive perceptions of, say, safety, they might be able to better intervene to make neighborhoods actually feel safer (add more park benches, turn up the lights, wash off the graffiti?).
"I had always assumed that I liked some [places] more, but I hadn’t really given much thought as to why," says Phil Salesses, one of the authors on a new PLOS ONE paper studying the question. "For me it was always a kind of overall gut instinct, an overall feeling. But once I started this project, and I started paying attention to details, I realized that something like trash on the street can flip a bit in my brain."
Salesses and collaborators Katja Schechtner and César A. Hidalgo built an online comparison tool using Google Street View images to identify these often unseen triggers of our perception of place. Have enough people compare paired images of streets in New York or Boston, for instance, for the scenes that look more "safe" or "upper-class," and eventually some patterns start to emerge.
"We found images with trash in it, and took the trash out, and we noticed a 30 percent increase in perception of safety," Salesses says. "It's surprising that something that easy had that large an effect."
This also means some fairly cost-effective government interventions – collecting trash – could have a significant impact on how safe people feel in a neighborhood. "It’s like bringing a data source to something that’s always been subjective," Salesses says.
We've previously written about a similar, non-academic project to use Street View images to measure "beautiful" streets. This PLOS ONE paper models how the tactic might more broadly be applied, producing the kind of data that cities could use to influence our subjective feelings about neighborhoods.
Salesses and colleagues used 4,136 geo-tagged images from four cities, Boston, New York, and the Austrian cities of Salzburg and Linz (the latter two cities were photographed by Salesses himself instead of by Google). Users were then asked which of the images looked safer, more upper-class, or more unique.
Not surprisingly, places perceived as "safe" were also more likely to be perceived as "upper-class." These maps from New York show the results of the three questions based on where the images were drawn, with green areas having positive perceptions (higher Q scores) and red areas having lower perceptions:
The researchers also looked at homicide data in New York City and found that, while controlling for things like income and population, these perceptions of safety correlate strongly with incidence of violent crime (from the paper: "We note from the start that our intention is not to make a causal statement, but simply to use this correlation to validate the value of the information contained in our measures of urban perception").
Unlike earlier Street View comparison projects, these researchers also photo-shopped individual elements out of a scene, refining their ability to isolate the role of individual objects, street art or trees on how people feel about place. The same tool could clearly be used to measure responses to other questions: Does this street look like somewhere you'd want to live? Is it family-friendly? Would you go shopping here? And the results might turn up concrete factors like flower planters, bike racks, pedestrians or potholes.
"We came up with the hammer and we were looking for things to whack with it," Salesses says, referring to why the project focused on perceptions of safety.
Ideally, in the future, he envisions officials using a tool like this to do cost-benefit analyses of removing trash or planting new trees: Where should cities spend scarce resources to produce the greatest improvements in how people perceive a place?