facebook relationships

The Facebook section that predicts relationships and when they will end.

There is talk of a scientific paper, which was authorized by some computer scientist and a senior engineer at Facebook. The paper has shown that our online social network not only reveals whom we are going out with, but also when the relationship will end. All this without checking your relationship status.

The study, which was done by Lars Backstrom and Jon Kleinberg, used data from not less than 1.3 million Facebook users. The data is used to analyze the shape of our friendship networks. This is information like whose friends with who, and how many friends individuals had in common, who were friends with their friends, you get the picture.Sna_large

 

The information looks at people aged 20 and above who have described themselves as being in a relationship. The data also looked at the number of friends that they have, so between 50 and 2000 friends. When you add this up it comes to around 379 million nodes and 8.6 billion individual links. This is pretty compelling evidence.

The two top engineers established that the dispersion was key factor in determining whether the pair was in a relationship. You see, the number of mutual friends between any two individuals, termed as embeddedness, was not enough.

The thing about dispersion is that it not only measures friends, but also the network structures that connect the individuals together. And for what they were looking for, a feature called ‘Low dispersion’.  A quality that was associated with couples; it indicated not only did the two people have mutual friends, but also that these friends knew each other.

DISPERSION ALGORITHM: RELATIONSHIPS THAT DIDN’T USE IT WERE 50% MORE LIKELY TO BREAK UP WITHIN TWO MONTHS.

facebook relationships

 

To add some light into this, just picture romantic partners as social bridges between individuals’ networks. They introduce people to each other and create new friendships.

The dispersion algorithm is able to correctly identify who someone’s spouse was about 60% of the time, and guess somebody someone’s partner a third of the time. This isn’t exactly perfect but much better than the two percent achieved by pure guessing.

I can see how this would work, considering that couples would be likely to reduce the dispersion between their friends over time, this happens as social functions gradually introduce the pair’s social networks to one another.

For the killer, the study also revealed that it was possible to predict, with some reasonable amount of accuracy, whether couples were likely to break up.  Back to the dispersion algorithm, relationships that didn’t use it were 50% more likely to break up within two months. This is as compared to those that had been identified by looking at friends of mutual friends.

leave a comment, tell us what you think.

Featured images courtesy of creativecommons.org