Teaching Mapping Apps to Pick the Prettiest Walk Over the Shortestby
Researchers at Yahoo! Labs in Barcelona have developed a kind of mapping algorithm that avoids the shortest route and the one with the least traffic in favor of the most beautiful way to go.
So far, the program has been tried only on London and Boston, according to the description in a research paper (PDF) that says it successfully used crowdsourced data on people-pleasing paths and incorporated them into walking routes between users’ current positions and their destinations.
The aesthetic judgments come from asking people online, through a website called UrbanGems.org, to compare photographs of different urban spots. The researchers used London as their test case. Users were asked which of two spots was more beautiful, which they found more quiet, and which made them happier. After running for four months, the study gathered information from 3,300 participants.
The algorithm would then create routes to optimize for each of the three pleasant characteristics, while ensuring that the routes were still pretty short—on average they turned out to be only 12 percent longer than the shortest option. The researchers then tested the routes with London residents, who validated the results. Here are the options, for example, for one walking trip in London.
Gathering all that information is time-consuming, which poses a major barrier to scaling this program beyond a couple of cities. To address this issue, the researchers tested whether metadata from posts on a photo-sharing site might be used to determine what spots people find pleasing. The Yahoo Researchers used a data set of 7 million geo-referenced Flickr photos and found that the “density” of photos (the number of photos taken in any one spot) was positively correlated with the UrbanGems beauty scores, as were Flickr tags that used positive words. Tags that used negative words, swear words, anxiety-related words, anger-related words, and sadness-related words were negatively correlated with the beauty scores. The researchers used this correlation to determine the most beautiful walking paths around Boston.
If the algorithm turns into an app that catches on, it could become a victim of its own success. Take quietness: The more people identify a particular spot as quiet, the more pedestrians the app would send there, undermining the very quality they sought.