After putting the data of over a million Facebook users through careful analysis, researchers have found a reliable and intuitive measure for determining who a person's romantic partner is on Facebook — and even potentially how serious that relationship might be.
The study, by Lars Backstrom of Facebook and Jon Kleinberg of Cornell University, used 1.3 million users' data (carefully anonymized, of course); everyone was over 20, with between 20 and 2000 friends.
Even with so much data, it's not easy to approach the problem of sussing out romantic connections. As the paper's introduction describes it:
As people use on-line social networks to manage increasingly rich aspects of their lives, the structures of their on-line network neighborhoods have come to reflect these functions, and the complexity that goes with them.
The strength of relationships on social networks can be measured in many different ways, but a common one is what the researchers call "embeddedness." This is the amount of connections you and another person have in common — the more mutual friends and interests, the stronger the relationship tends to be.
But embeddedness turns out to be quite a poor predictor of romantic partners: When it was used as the primary measure to analyze users who have declare they're in a relationship, it only managed to identify the person's spouse or significant other a quarter of the time.
When you think about it, it makes sense. Think of all the friend groups you have — your hometown friends, your college buddies, your work chums — between a romantic partner likely met later in life and friends you've known for far longer, with whom are you likely to have more mutual connections? Embeddedness is a great way to identify older and stronger friendships, but it's barking up the wrong tree when attempting to predict your wife or boyfriend.
To do just that, the researchers proposed and tested a new measure: "dispersion." If embeddedness is the amount of mutual connections between friends, dispersion is the amount of connections a person has between unrelated groups of friends.
Take your college friends: you may have lots of friends in common with any given one of them, but it's not very likely that any of them is friends with the people you work with. So you have these two clusters full of strong relationships that don't really overlap with one another.
Now think about this: What kind of person is the kind that meets and befriends people from your family, your work life and your old school buddies? That's right: your significant other.
This is the theory behind the study's dispersion metric, and right away they found that their new method worked twice as well as previous ones: on married users, the friend with the highest dispersion score was their spouse 60 percent of the time. Factoring in other data raises the accuracy to 70 percent, and when the researchers looked at a specific demographic group — married white males — the accuracy reached 77 percent.
Not content to simply identify your partner, the researchers also investigated whether the same measure could tell them whether a relationship would last. As it turns out, it can; significant others with a high dispersion score (especially if that score was the highest among all your connections) were more likely to remain significant others over a 60-day period. That may not tell you whether it's Mr. or Mrs. Right, but it might be able to tell you whether it's Mr. or Mrs. Right Now.
Of course, all this is academic if someone's profile is constantly bedecked with "Hey boo ;)" messages from a certain someone. But there's value in being able to determine relationships simply from the shape of a person's network of friends: "Crucial aspects of our everyday lives may be encoded in the network structure among our friends," write the researchers, "provided that we look at this structure under the right lens."
The paper is available for free if you want to get the details on methodology and so on. Don't worry, though — you won't be getting any "recommended romantic partners" in your feed any time soon. This research was conducted strictly for scientific purposes.
Devin Coldewey is a contributing writer for NBC News Digital. His personal website is coldewey.cc.