The algorithm depends on a new metric the researchers are calling "disperson," which looks at connections between people who have different sets of friends. Close friends are likely to share a lot of friends in common, which social scientists call "embeddedness." But people in romantic relationships tend to connect each other to different spheres: for example, a husband is likely to introduce his wife to his work friends, college buddies, and members of his kickball league, with whom she may not have any other connections in common.

Relatives may also show this same "dispersion" dynamic, the researchers say.

The researchers were able to identify who was dating whom with 60 percent accuracy, much better than the 2 percent accuracy they'd get from random guessing. High dispersion also seems to be correlated with longer relationships. The study found that couples were 50 percent more likely to break up in the next two months if the dispersion algorithm failed to guess that they were dating.

The scientists also looked at metrics such as how many times a user viewed another's profile, attendance at the same events, and messages sent. Dispersion turned out to be the most overall accurate metric for determining romantic relationships. The researchers used multiple sets of anonymous data, including a large data set from 1.3 million Facebook users.

The study shows the enormous effort being made to crunch Facebook data into something meaningful for marketers. However, it's easy to see where the algorithm would fail. For example, many Facebook users list fake relationships as a joke, declaring themselves to be "married" or "engaged" to their best friends. The method also assumes that you, all your friends, your spouse, and all your spouse's friends are active on Facebook.

Of course, this is nothing new. Remember when Mark Zuckerberg wanted to predict who you would hook up with?

Update, 6:05PM: A Facebook spokesperson sent a statement about the study: "The specific findings from this study are not currently being used in making News Feed a better product. It has helped the team think about things and informed some of the features we use in ranking, but we are not using the specifics in the paper."