What makes a significant other different from a very dear friend?
Well, besides that.
Facebook data scientists have developed a novel method for identifying who among a user's friends is that person's partner—and their work puts an empirical stamp on something that is perhaps intuitive: A significant other occupies a unique place in a person's social network, one characterized not by "embeddedness"—the standard way of measuring a tie's proximity—but by what the researchers call "dispersion."
Here's what that means: The number of mutual friends ("embeddedness") is a reliable indicator of how close two people are. Simply put: You have more friends in common with your closest friends than with your acquaintances.
It might seem reasonable then if "embeddedness" could be used to predict people's plus-ones. But that's not what researchers Lars Backstrom of Facebook and Jon Kleinberg of Cornell University found. In a new paper, they write that embeddedness is an at best mediocre predictor of that special something. Relying on embeddedness, they were able to accurately predict Facebook users' significant others 24.7 percent of the time.
Another measure fared much better: "dispersion," or how many different networks of theirs a person's friend shares. In other words, your significant other won't just share many friends with you, but friends from all walks of life: your colleagues, your high school buds, your college friends, your family, and so on. Using dispersion, Backstrom and Kleinberg doubled their accuracy: 50 percent of the time, a person was romantic partners with the person who was the most dispersed across his or her social network. For married people, their accuracy rose to 60 percent, a figure which they say is more than 30 times higher than random guessing would produce (everyone in their sample had at least 50 friends).
The researchers were able to work with their findings to produce a telling observation about the health of a given relationship: The more dispersion a relationship demonstrated, the more likely it was that that relationship would still be around 60 days later (the researchers excluded married couples from this analysis, though that would be quite interesting to see too). This is not necessarily causal (no, you can't get your girlfriend to stick around by bringing her to the company holiday party), but couples who are building a life together, as demonstrated by their multiple and interconnected social circles, seem to have more staying power. (Presumably, though no measurement tools have been developed yet, the breakups of such relationships are also the most painful.)
For now, researchers relied on looking at Facebook users who identify their partners on the site, but, as this sort of analysis gets refined, it's easy to see how they could use their methodology to detect a relationship which is not quote-unquote "Facebook official." Such information will be useful to Facebook in prioritizing certain people in your newsfeed, showing you those—and especially the one—you care most about.
We want to hear what you think about this article. Submit a letter to the editor or write to email@example.com.