Skip to main content



Facebook’s Friend suggestions

Nowadays, when someone talks about social network, our minds immediately roam to Facebook. This is not just because the movie on the site adopts the same two words as its title, that is merely a side effect. With a rapid rise to fame, Facebook has become THE social networking site in the world in just a couple of years. Yet, with all its success and glory, how can Facebook continue to grow and remain an important aspect of our lives? How can it outcompete other social networking sites like Google+? The only way is to find new and innovative ways to keep its users ever more engaged or “plugged-in” to their social lives on the site. One aspect of this is its friend suggestion feature. Finding a better algorithm to suggest the right friends to the user is essential in building their social network and would keep them much more engaged. As we have learned in class through the Strong Triadic Closure Property as well as graph theory in general, if our close friends have strong ties with a person, it is more likely that we will become friends with the person. Applied to Facebook, if the user’s friends have strong ties with a person, it is more likely that the user is a friend or at least an acquaintance (i.e. weak tie) with the person. Thus, an essential part of having a better friend suggestion algorithm is to figure out how strong one user is tied to another.

The first article published in 2009 on how Facebook adopts a at the time new feature, finding friends through E-mail. The users were not informed that FB is actually storing that information for future friend suggestions. The article explores the user privacy aspect of the “new” feature. We will ignore that aspect since it does not pertain to the topic of discussion and focus on why FB would want to do such a thing. Figuring out who has strong ties with whom on FB can be a very tricky thing. It requires a lot of information processing, thus, the more information you have the more accurate your suggestions are going to be. Generally, the people who you talk to the most are going to be on your e-mail contact list. With sites like Gmail, they might be people who you IM with regularly. These are very good indications of strong ties. By storing these strong tie edges from one user to the next, Facebook will be able to suggest friends to you more accurately. If one person is on the email list of many of a user’s FB friends/email list friends, then they are very likely to friends as well.

The second article published recently this year is about another one of FB’s new features, Timeline, which allows you to integrate your hobbies and interests (such as music interest, apps, games, articles liked, movies, food, etc..) with the site. With Timeline, FB can practically know things from what TV shows you like to your political stance. This brings us to another essential property in determining the strength of ties between two people. If two people share the same music taste and political belief, they are more likely to be strongly tied with each other than people who don’t have anything in common. This again helps in building a better, more accurate friends graph and help the friend suggestions feature.

As you can see, FB is constantly trying to acquire more information on its users to build a more accurate network graph, which will in turn lead way to better services like friend suggestions, event suggestions, and maybe even what shows up in your top news feed. These services will keep the user more informed and engaged in the site. However, with all the information flowing into FB, privacy of the user is becoming more and more of a topic of discussion. Although all this is good for the commercial interest of FB, it might not be so good for the user.

Links

http://www.insidefacebook.com/2009/06/12/facebook-now-suggesting-friends-found-in-imported- contact-lists/

http://arstechnica.com/gadgets/news/2011/09/facebook-wants-your-past-present-and-future-on- open-graphs-and-timelines.ars

Comments

Leave a Reply

Blogging Calendar

November 2011
M T W T F S S
 123456
78910111213
14151617181920
21222324252627
282930  

Archives