Computer scientist predicts your next Facebook friends

A Stanford University researcher wins a Microsoft fellowship for analysis of social networking activity

Half of the friends you will add on Facebook in the future can be predicted, said Stanford University's Jure Leskovec. He has been elected as one of this year's recipients of the Microsoft Research Faculty Fellowships.

Leskovec is an assistant professor of computer science at Stanford University. The focus of his research is on the traces human activity leaves on the Internet. Someone reads an article on a news website, writes a blog post or forwards someone else's tweet on Twitter: It is data like this that Leskovec collects to analyze human behavior on the Web -- and even predict it. He does the same with data from social networks.

"Data shows that who will be our next friend on Facebook is not so random as we think," he said. He has just finished a project with the company that runs the social networking site. Based on information about the personal networks of users and their communication he was able to tell in advance half of the new contacts they would add shortly after.

In the future the rate of correctly predicted new friends could be even higher, he said. "We are able to train the analyzing methods," Leskovec said.

Findings like this could be used to develop models of how online groups grow over time. "It will soon tell us how healthy a community is," said Leskovec. So far his analysis shows that a social network should neither be too sparsely populated nor too dense. "Our research suggests it may not be good to saturate a network," he said. This means too many contacts and too much communication could someday thwart vitality and growth in a network.

Another project Leskovec just finished was an analysis of the user community of Microsoft Instant Messenger. The research project in cooperation with Microsoft proved the hypothesis of the "6 degrees of separation." Leskovec found that people using Messenger were in general 6.6 steps apart from each other. While this was fundamental research, dealing with questions like this could also enable practical applications, Leskovec said.

"The interesting question is how to find the right connections, for example, when you want to know who to ask to get introduced to the queen of England," said Leskovec. A solution to this question could help in finding efficient ways for routing through the Web, he adds. This could, for instance, be useful for finding the shortest path between two computers on the Internet.

Apart from social networks, Leskovec analyzes the use of online media as well, and he lets his computers dig through 30 million articles every day. One of his goals is to design algorithms to find patterns that show what happens to these news items. This could show, for example, how information changes gradually. "It could reveal that your political attitude affects how you treat certain information. Maybe you forward a very long Obama statement while you shorten the quotes of other people in a text," Leskovec said.

Recently he found out in a study that news spreads quite differently depending on the platform where it is first published. One finding was that material published by newswires gained the highest attention shortly after being published. Blog posts in contrast very often got a number of attention peaks over time.

Leskovec has already made some plans on how to spend the money coming with the fellowship (he will receive $100,000 this year and the same amount in 2012). Leskovec said part of the money will go into "risky projects or startups" that without the grant would not have been possible to do. He said he particularly appreciates the fact that he can use the money at his own discretion. "It is a gift without strings attached and we did not promise anything in return," he said.

Leskovec also plans to buy new equipment and use part of the grant to organize seminars. He wants to send his students to work with Microsoft, too. "That is a good opportunity for them to get introduced into new fields of research," he said.

Leskovec received his Ph.D. in machine learning from Carnegie Mellon University in September 2008 and spent a year as a postdoctoral researcher at Cornell University. He did his undergraduate studies in computer science at the University of Ljubljana, Slovenia.

Insider: How the basic tech behind the Internet works
Join the discussion
Be the first to comment on this article. Our Commenting Policies