Facebook Research

How does Facebook predict the outcome of a relationship ?

• • Researchers Draw Romantic Insights From Maps of Facebook Networks

• • [article] Romantic Partnerships and the Dispersion of Social Ties: A Network Analysis of Relationship Status on Facebook, Lars Backstrom (Facebook), Jon Kleinberg (Cornell University), 2013

How does it work ? Lets take a real example, which is of course anonymized, even though the outcome is encouraging ...

Below is the friendship graph of a young male. Below (1st figure) you can see the friendship graph with nodes colored and sized according to their degree. In the second figure the same friendship network highlights the four detected communities: an aggregation of the green, yellow and red communities and a newer blue community. The highest degrees are in the upper communities, so we would be inclined to thing that a best romantic match should be in one of those three communities.

That is not the case. Research shows that not degree (number of common friends) but betweenness is the key to having a good relationship. Betweenness is the amount of control a node has over inter-community communication, thus acting like a bridge between our social worlds. The intuitive idea is that partners are happy when they both bring their social life with them, in their backpack, and none of them cuts off ties with his previous friends. The presented friendship graph consists of the upper community (giant component), which is formed out of friends made from day 1 up to college (another region, city, school etc.), and the smaller community made out of new friends from college and the new hometown of the subject. So before, explaining more, let's look at the same graph in which we highlight the node color and size according to the betweenness (3rd figure, below).

There are a couple of nodes with high betweenness inside the upper community but there are some relevant nodes in the lower part too. One could argue that either one of the larger red nodes are a good romantic match according to the research findings. And that is completely true. If we do some coloring to help us, everything will become much clearer (4th figure, below).

The figure above highlights the nodes with high betweenness (node size) and also colors these nodes according to their sex (male, female). Considering the premise that the best romantic partner for a person is that node in his friendship graph with the highest betweenness placed so that it bridges two or more relevant communities (to the giant component) ... and is of course usually of different sex, brings us to a clear conclusion: there are four pink nodes in the lower part of the figure, out of which the lowest one is also the most relevant one in terms of betweenness. Guess what ?

It works.

October 2013

Adding a weighted context to friendships: what does the structure of the Facebook friendship graph tell us about friendship strengths?

• We at ACSA have developed an innovative FB app that can extract friendship graphs of an online user and corroborate them with the daily private messaging intensity of his/her friends. These "tie strengths" are mapped onto each node offering an innovative way of analyzing social graphs.

• Results to be follow soon.

August 2014

The fingerprint of educational platforms in social media: A topological study using online ego-networks

• Based on an article, with the same name, we recently presented at a conference, we have shown that educational-related platforms have a specific pattern of interconnection and clustering on Facebook.

• The interpretation using LigaAC's friendship network is described here.

January 2015

Cupid in your network

• Are matchmakers close friends or acquaintances? Sociologists have long studied the dating pool problem and have reached an interesting conclusion which you can read here.

It seems an acquaintance is more likely to introduce you to your mate than a close friend is. Your close friends tend to know all of the same people as you, while acquaintances have greater bridging capital, connecting you to diverse clusters of new people. On the other hand, your close friends might be more motivated to set you up; they care more about your happiness, and they know you better.

Matchmakers have a secret weapon: more unconnected friends. To start, matchmakers have far more friends than the people they're setting up. Before the relationship began, the people who later got matched had an average of 459 friends each, while matchmakers had 73% more friends (794 on average). Even compared to the couples' other friends, matchmakers are unusually well-connected.

Matchmakers were more likely to be close friends, rather than acquaintances. Matchmakers typically had 57 friends in common with either person in the future couple, amounting to about 6% of their pool of friends. Does that make them close friends? One good comparison point is the couple today: they typically have 52 friends in common, or about 7% of their pool of friends. If romantic partners are your gold standard for close friends, the matchmakers look like strong ties!

Couples who were introduced by a friend are more likely to have at least one friend in common a year prior to the start of their relationship (84% of them) compared to those who met in other ways (74% only), which is to be expected – this quantity increases over time, and a few months into the relationship, more than 99% of couples have a friend in common, regardless of how they met.

More interesting though, is that the fraction of mutual friends starts higher for couples who were *not* introduced by a mutual friend (around 4.4% overlap 18 months prior to the start of the relationship) compared to those who were (about 3.4% at that time). It is notable that that quantity increases faster for the latter, to the point that 18 months into the relationship the overlap of social networks of both members of the couple averages at around 7.7% whether or not they were introduced by a friend.

Finally, the sources of finding a partner are depicted below. Schools are the main contributor to relationships, but what is also interesting is that Facebook has an impact equal to that of meeting partners in bars and nightclubs.

February 2015