Net Worker

Where are your friends in cyberspace? Closer than you might think, says Internet researcher Jon Kleinberg

Cheryl Carlin

Jon Kleinberg helps us see the invisible networks that pervade our lives. A professor of computer science at Cornell, he teaches a class with the economist David Easley that covers, Kleinberg says, "how opinions, fads and political movements spread through society; the robustness and fragility of food webs and financial markets; and the technology, economics and politics of Web information and online communities." If it sounds like "Intro to How the World Works," that's the general idea.

Some of Kleinberg's research builds on social psychologist Stanley Milgram's famous 1960s experiments into the "small-world phenomenon." Milgram enlisted a random group of people in Omaha and asked each to forward a letter to one close acquaintance, with the goal of reaching a certain stockbroker in Massachusetts. By tracking the letters, he came up with his "six degrees of separation" theory: any two people on earth are connected by a string of five or fewer mutual acquaintances. Forty years later, Kleinberg runs his own tests on the small-world phenomenon sitting at his computer, poring over data from five million members of the blogging and social network Web site LiveJournal.

He was particularly curious to know how the physical distance between members of the online community affects the likelihood of their associating. He found that even in cyberspace, friendships depend on proximity. (In fact, the probability that people know each other is inversely related to the square of the distance between them.) "Why should it matter online if someone is 10 miles away, 50 miles away or across the globe?" he says. "You would think friends might be uniformly spread out around the world. That's not what happened. You still see heavy traces of geography."

Kleinberg's prominence is based partly on his work navigating the Web. In 1996, as a visiting scientist at the IBM Almaden Research Center in San Jose, California, he developed an influential search algorithm (not unlike one used at Google) that ranks the popularity of Web sites by measuring how other sites link to them. More recently, he's been intrigued by the possibilities of measuring "word bursts," spikes in Internet usage of a term that would, say, reflect new social trends or political concerns. In a test, Kleinberg analyzed State of the Union addresses since 1790, showing, for instance, that the word with the most "burstiness" between 1949 and 1959 was "atomic." 

Kleinberg, 35, says he expects to see machines, applications and Web sites become better at responding to users' past behavior and prompting them. Your computer might insist you reply to an important e-mail that's been waiting too long, scold you for procrastinating or, sensing that you're about to leave the office, remind you what's left to do. In a statistical sense, he says, computers "know much more about your behavior than you do."

Each advancement will likely be hugely profitable for whoever implements it, a prospect not lost on the students pouring into Kleinberg's classes or the standing-room-only crowds that fill his lectures at Yahoo! Research conferences. But he maintains that his temperament is best suited to academia. "I've now missed so many opportunities to make money off this stuff that I figure, why start worrying about it now?" Besides, he's an idealist. "It would be great if the consequence of getting the world hyperinformed is that we understand different cultures better, are more sympathetic to different points of view, get along better," he says. But as his own research would suggest, it's the online masses that will decide the extent to which those wishes are fulfilled.

Matt Dellinger runs the New Yorker magazine's Web site.

Get the latest stories in your inbox every weekday.