Websites such as match.com often employ algorithms to judge how compatible you might be to someone for a romantic relationship. Typically these involve typical things you might expect such as common interests, opinions, hobbies etc. Facebook is no stranger to this game, and carries the pretty large advantage of having 1.1 billion people to average their data across.
Facebook and Cornell University have now decided to team up and use this huge amount of data to create an algorithm based on something different. They effectively use the network of mutual friends surrounding the couple to judge their "embeddedness and dispersion", basically how close their friendship circles are. Putting these numbers through a few equations gives a pretty good indication of whether the two are in a relationship, 60% of users who have a certain "optimal dispersion" as judged by the algorithm actually turned out to be married. Users in a relationship who don't meet this dispersion factor can be judged as likely to break up, the algorithm can even estimate when.
Of course needless to say, although this has the population of Facebook as a statistical sample, averages mean nothing to the individual. On top of that it doesn't account for anomalies such as users who are "in relationships" with their friends on Facebook as a joke.
Despite the ethical queries that might surround whether Facebook actually had permission to use its users data in this respect, this is a rather interesting, and surprisingly complicated piece of statistical research.
Image credit: Switched