Researchers from Cornell University have developed an algorithm that can help to identify a Facebook user's romantic partner with a high degree of accuracy by examining data collected from the vast online social network.
Prof. Jon Kleinberg and Facebook employee Lars Backstrom developed an algorithm to assess the nature of a user's relationships with others on Facebook. They found that the algorithm could help correctly identify a user's spouse, fiancé or other romantic partner about 70 per cent of the time.
The researchers looked at two factors in their algorithm:
- How many mutual friends a Facebook user shares with another (called "embeddedness")
- How many of those mutual friends are also friends with each other (called “dispersion”)
According to the paper, the more mutual friends two Facebook users have, and the more dispersed those mutual friends are to one another, the more likely it is that the two users are in a romantic relationship.
The algorithm was tested on data from 1.3 million randomly selected Facebook users who listed their relationship status as either "married," "engaged" or "in a relationship." The data was made anonymous, and lists of the users' friends were also examined.
It was assumed by the researchers that a romantic partner would be heavily embedded in the Facebook user's life – meaning the couple would share many of the same friends. This assumption resulted in correctly identifying the user's partner about 25 per cent of the time.
In order to boost the accuracy of the algorithm, Kleinberg and Backstrom decided to examine the couple's mutual friends based on how disperse they are in relation to one another. It was assumed that a couple's mutual friends are not highly connected among themselves, but instead are dispersed in many different areas of the central user's life.
For example, your romantic partner may know many of your contacts, but these contacts all play a different role in your life and do not necessarily know each other. Some may be your co-workers, some may be family members and others may be members of your recreational sports team or club.
The researchers determined that dispersion in particular was a good indicator, and was able to help identify a user's spouse 60 per cent of the time.
Finally, the researchers examined how often the contacts interact with each other, looking at factors such as how often people visited each other's Facebook profiles, went to the same social events and how often they took photos together.
Altogether, these factors helped researchers correctly identify the user's romantic partner 70.5 per cent of the time.
On the flip side, researchers found that if the algorithm did not correctly identify a user's partner, there is a high chance the actual relationship may soon end.
The authors write that in cases where the algorithm failed to correctly identify the user’s partner, the actual relationship is “significantly more likely to transition to ‘single’ status over a 60-day period.”
The team will be presenting their paper at a conference on social computing in Baltimore that starts on Saturday.
