When it comes to going viral online, it’s not how many people you know but who they are, a new study suggests. An analysis of the social networking tool Twitter concludes that having a lot of followers isn’t as important to Internet exposure as having discriminating followers to pass the message on.
The results suggest that those trying to communicate through social networking, be they politicians, advertising executives or philanthropic organizations, shouldn’t focus their efforts on targeting the masses. Target the influential, and the masses will come.
“It’s not only the numbers, the quantity of your audience, but also the quality of your audience,” says study coauthor Wojciech Galuba, of the Swiss Federal Institute of Technology in Lausanne.
The research suggests that in some regards, the online world isn’t that different from real life. If somebody you know recommends a book, whether or not you actually read it depends on a lot of factors, including your reading habits and your opinion of that person’s opinion.
To determine who has influence in the online world, Galuba, Daniel Romero of Cornell University and their colleagues looked at tweets — the 140-character messages posted on the Twitter social network — that contained links to web pages. That allowed the scientists to see how many recipients of the tweet took action by “retweeting” it, or passing it on, and how many times people who received it actually clicked on the link.
To get a better measure of influence, the research team created an algorithm that doesn’t just incorporate a tweeter’s number of followers but also whether those followers are passive and listless or dedicated and engaged. The team tested the algorithm on 22 million link-containing tweets sent during a 300-hour period in September 2009. The algorithm was a far better predictor of whether someone clicked through to a link than other measures such as sheer number of followers or number of retweets, the researchers report in a study posted online August 6 on arXiv.org.
The algorithm yielded both surprises and, well, duhs. Google News and The Onion both were influential tweeters, with lots of followers who also passed their tweets along to more people, who in turn retweeted them yet again. But twitter feeds of media monsters like NPR Politics, E! Online and Martha Stewart — ranked 41st, 42nd and 43rd respectively in number of followers — were outranked by hundreds of other tweeters in terms of influence. For example, Rob Miller, a congressional candidate from South Carolina, ranked 147,803rd in number of followers but 145th in terms of overall influence.
The algorithm may find use in determining influence for a specific question or time frame, says Romero. For instance, during an election cycle, one could examine all tweets related to politics and see which candidate is best at spreading the word.
Measuring influence by more than just audience size is an important step, says information scientist Jim Jansen of Pennsylvania State University in University Park. But the true measure of influence is action, he says, which researchers are just started to figure out how to assess.
“Get me to buy something, get me to go to the polls, get me to send money to Haiti or go to a demonstration,” he says. “That’s influence.”
