Representative Anthony Weiner (D-New York) needs to know about rumor centrality. If he’d known about it, he would know that sending a photo of himself in his knickers via Twitter will result in a surge of retweeting.
Rumor centrality is a key part of a system developed by scientists at MIT’s Laboratory for Information and Decision Systems, that can identify the most influential Twitter user on any given topic at any given time. A sort of search-engine-for-influence, the system is called Trumor and the researchers hope to launch a public web site this summer. The system was recently announced at MIT.
It’s not as easy as picking out the people with the most followers. The key ingredient is the number of links their followers might have, and how active the pathways (or discussions and retweets) are between those followers and their followers and so on.
Tauhid Zaman, a PhD candidate at MIT, spotted an interesting pattern while he was following tweets and retweets surrounding different events and topics, like the BET awards, a World Cup match, the LeBron James press conference. “Every single event had a similar structure on Twitter,” Zaman said in a phone interview. There was always a so-called “superstar” that would drive the entire conversation because their tweets were retweeted much more often and consistenly than any other tweeter. And by much more, Zaman means by a huge margin.
For example, let’s say 100 people actively tweet during the BET awards he might see a few people who would be retweeted a few times, but it was always one user who got retweeted over and over by at least 50 other users. Meaning, the superstar nearly always produces 50 to 60 percent of the tweets that are retweeted.
“There was always one superstar, over and over again across all events, it seemed uncanny,” Zaman said. “I wasn’t expecting this, and I knew it wasn’t an accident. I’d stumbled on a phenomenon.”
Just noticing this trend isn’t enough, however, so he and his colleagues developed a model. They’re scientists after all. Zaman then gathered data from retweets during specific events, stitched together the pathways between users who retweeted. Next he developed an algorithm to produce a score or influence ranking, called “rumor centrality,” to determine the superstar in each event or topic. What emerged shocked Zaman.
“It was a mind-blowing graph. I couldn’t believe that everything lined up perfectly. It’s so rare in a scientist’s life when the data matches your model,” he said. “Especially with a social thing. I mean you expect it more with physics or particles and atoms. Not with humans and social networks.”
Rumor centrality measures the amount of spread a user might have. This is not about number of followers, but more the quality of those followers. How many active followers a user has, how active those followers are in retweeting and how dedicated their followers are.
This summer Zaman hopes to develop the system into an online search engine where marketers, politicians, grass roots organizations, or anyone really, can find the top people of influence or “superstars” (i.e., those with high rumor centrality) on any topic, event or product. Zaman notes that the superstar is not always obvious. Sure there are the Lady GaGas that trump others in the number of retweets they get. But for the smaller companies or local organizations that want to know who their “superstar” is, Trumor could be hugely useful.
[via Technology Review]