USING ALGORITHMS, A TEAM OF STUDENTS ANALYZED THE CLUSTERS OF PLACES THAT LIKE-MINDED PEOPLE FLOCK TO.
Every city is filled with different neighborhoods, but often, you won’t find these places on any map. They’re word-of-mouth zoning distinctions known only to locals. The boundaries are vague and arbitrary, based as much upon the way people eat and dress as real estate prices and income per capita.
Yet if these areas are distinctive to city culture, is there a way that we could measure them and analyze them--map them--scientifically?
A team of students (Justin Cranshaw, Raz Schwartz) and professors (Jason I. Hong and Norman Sadeh) from Carnegie Mellon’s Mobile Commerce Lab has done just that. Their research project is called Livehoods, which analyzed 18 million Foursquare check-ins to spot algorithmic relationships between the spots people frequent. The team tells Co.Design.
“Livehoods looks at the geographic distance between venues, but also a form of ‘social distance’ that measures the degree of overlap in the people that check-in to them. For example, if the algorithm notices that the people that visit a local bar are the same people that visit a nearby restaurant, these two places will be more likely to be grouped together.”
As more and more people and places are analyzed, Livehoods clusters this data into what becomes a collection of distinctive neighborhoods--places filled with people who enjoy going to the same restaurants, coffee shops, and music venues. And as calculating as the approach could seem, Livehoods’ scientific basis makes it extremely valuable as a social artifact: It defines local culture without the inherent judgement that comes along with human stereotyping.
With this scientific methodology in mind, the Livehoods team cross-checked their own findings of Pittsburgh with 27 resident interviews. What they found--the full results which will be shared in a paper presented this June--was “compelling evidence” neighborhoods as Livehood algorithms had defined them had “real social meaning to people in the city.” In other words, the digital map lined up with many residents’ own mental maps.
All of this said, Livehoods aren’t a perfect snapshot of humanity just yet. The datasets mined for the project are limited by the perspective of Foursquare users. A lot of us don’t use Foursquare (with a strong skew toward older adults, most likely). The team explains.
“Our technique, however, is agnostic to the specific source of the data, so as we get better, less biased sources of data, we should be able to produce more accurate views of the city.”
The young researchers also fear that we may take their boundaries a bit too literally. As much as Livehoods works to clarify invisible distinctions, the team, paradoxically, points out that these distinctions are more subtle than we might expect.
“In reality, neighborhoods tend to blend into one another.”
In which case, may I suggest a simple UI tweak? Maybe Livehoods should be rendered in gradients.
COMMENTARY: The type of location-based data visualization that the students at Carnegie Mellon are conducting is not all that new. In a blog post dated October 14, 2010, I commented on a location-based iPhone app from WeePlaces that also uses check-in data compiled from foursquare to create very revealing geodemographic maps of singles hotspots in the Manhattan and adjoining vicinities. However, having been a foursquare user, it is possible check into a merchant without actually being physically there. Users do this all the time to earn points towards foursquare badges. This could give false data about how many people actually frequent a particular merchant or hotspot. A better barometer are comments made by individuals that have actually visited a check-in spot. Having said this, users should always try visiting spots themselves to determine if they are really popular among actual visitors or individuals just checking in.
Courtesy of an article dated April 18, 2012 appearing in Fast Company Design