Cornell University computer scientists and a Facebook engineer have found in their latest research that it is possible to predict whether users are in a romantic relationship based on their shared social circles, and whether they will break up soon.
Using data from 1.3 million anonymous Facebook users who list themselves as being in a relationship, the two researchers developed an algorithm based on common social connections. They discovered that social embeddedness — meaning that two people have a large number of mutual friends — does not help much in predicting if they are in a romantic relationship.
In fact, a better predictor is dispersion. The theory from the Facebook researcher is that high dispersion means that a couple's mutual friends are not strongly connected, while low dispersion means that a couple's mutual friends are closely connected. Couples with low dispersion are less likely to have separate social spaces and more likely to break up.
Facebook claims that healthy romantic relationships should be characterized by each person having multiple distinct social circles, even if those friends are not particularly close.
The algorithm based on dispersion developed by the researchers has a relatively high accuracy in predicting whether two people are in a romantic relationship (60% for married couples, and 33% for those dating). More interestingly, the study found that romantic relationships with low dispersion are 50% more likely to end within two months.
As for the significance of this study, Facebook data scientist Lars Backstrom said, "If we can do a better job of figuring out who the most important people are in your life, then we have a good shot at making Facebook better." In other words, Facebook wants to deeply mine your personal data to make money.