The video "Yes We Can," with its hiphop artists, earnest guitar, soul riffs, and montage of multiracial faces, quickly racked up millions of YouTube hits during the last presidential campaign. So did "Obama Girl," a video featuring a sexy actress lip synching lyrics about her "crush" on Obama. "Universal health care reform. It makes me warm," she mewed.
Why did these videos go viral? What powers propelled them into the political consciousness?
New research from a leading iSchool team, supported by a Google Research Award, shows that a predominant force behind the '08 election buzz was the blogosphere. Hyperlinks, discussions, and posts on high-profile blogs like the Huffington Post and Daily Kos regularly turned flurries of YouTube clicks into network blizzards. "The blogosphere was critical in creating the sparks. It was the engine of virality - not the mass media," says iSchool associate professor Karine Nahon. "The blogosphere led the way and the mass media followed."
Nahon, who worked in high-level, high-tech research and development positions before joining the iSchool, is director of the school's Virality of Information research (retroV) team. Earlier this year, retroV landed the highly-competitive $30,000 Google award, a one-year grant designed to seed innovative ideas and promote academic research aimed at improving information access.
"This was a great honor for the whole team. It's a very prestigious award," says Nahon, who holds a Ph.D. from Tel Aviv University and still advises the science and technology committee of the Israeli parliament. She comes to the iSchool from multiple disciplines: political science, sociology, computer science, management of information systems, and information science. The eclectic background serves her well at the interdisciplinary iSchool, she says: "Here, you can mingle those worlds. If I did the same research from only a political science point of view, or a computer science point of view, it would produce different findings. Information schools provide interdisciplinarity and breadth."
The Google award supports retroV's ongoing research into the phenomenon of virality, defined by team member Jeff Hemsley, iSchool Ph.D. student, as a word-of-mouth, person-to-person diffusion process that spreads messages into social networks at exponential speeds. "If 10 people forward links at each step of diffusion, on the fifth step, 100,000 people will see the message," says Hemsley. "And when something becomes popular very, very fast, it takes advantage of the attention dynamics built into the human system. We all look in the same direction at the same time."
In the course of their exhaustive research, retroV team members identified 10,000 blogs and around 13,000 posts linking to 83 of the most popular election-time videos and examined their behavior on the net for more than two years. The collection process alone was challenging. "It was a nightmare collecting data from reluctant sources - which is ridiculous because most of the data is public," says Nahon.
To process their work, RetroV developed pioneering methodology borrowed from econometrics and social network analysis. The new tools allowed them to measure, in real time, when information became viral. Graphs from the research - newly posted at retrov.org/blog/explore-the-retrov-dataset - track the life spans of election '08's top viral videos. Most flared a few days after they were uploaded, with one dramatic spike of internet activity. The No. 1 video "Yes We Can," produced by rapper will.i.am, had two spikes and a bump: the spikes when it was uploaded and on election day; the bump on inauguration day.
Tracking the links allowed the researchers to put numbers to a phenomenon called homophily - the tendency of the likeminded to seek the like-minded. The iSchool research showed that, in the blogosphere, liberals tapped liberal blogs, conservatives tapped conservative blogs. Even if users did cross-link to another viewpoint, it was typically not for the purpose of discourse, but to strengthen previously held political convictions or to attack those who didn't hold them.
Some scholars worry this birds-of-a-feather phenomenon enhances polarization and fragmentation on the net, and threatens its democratic nature. For Nahon, the pattern is a puzzle deserving serious investigation. "Why, when we have all these choices online, do we tend to close ourselves off? Why do we choose not to look for diversity?"
Also puzzling for Nahon was what she describes as an "extreme power law" at work in the blogosphere. On the Internet, the power law means very few information providers capture the attention of most users. "For example, 98 percent of Internet users use only four search engines, and Google captures around 73 percent of that audience. In the real world that would be a monopoly," she says. "It means we're all depending on the queries and results of four search engines."
The same imbalance holds in the blogosphere, where a few elite blogs capture most of the attention. The team identifies the most influential as "top blogs," as opposed to "tail blogs," the blogs of ordinary users and followers. Because the elites ignite the process of virality, they can frame messages and influence agendas, manipulating video content to enhance their own agendas and shape their own stories.
Nahon wants to increase awareness of these imbalances. "We need to look at the power law, ask what that means about us as a people, about the information we get. What influence do certain entities, certain bodies, have upon us that we don't even think about?"
The dynamics behind online virality are intriguing and complex, and research into the phenomenon is drawing keen attention from around the globe. And it's not just politicians interested in the virility of virality. Marketers are interested in it as a way to promote their products in social media networks. Conversely, some government agents want to learn how to block it, after seeing the role social networking played in Arab Spring uprisings, says Hemsley. "They want to find ways to cancel it out or counter it."
Hemsley will be carrying on the retroV research as part of his dissertation, expanding investigations into new mediums and contexts. In the '08 election, when 74 percent of Internet users logged onto political blogs, Twitter and Facebook played minimal roles in information spread. Now these networks, and others including Flickr and Meetup, are sharing center stage. Today, someone may be Tweeting a message and sending an update to Facebook and Twitter at the same time.
How does viral information cross-link and spread across these multiple social media networks? "We want to look at how they all work together to shoot these messages into the public consciousness so quickly," says Hemsley.