Newsroom: Highlights

Bookmark and Share Email Print

01.08.10 - Researchers Surprised by Blog Network Analysis

The DNC-credentialed blog on top and the RNC-credentialed one on the
bottom form citation networks by linking to one another. Over time, connection
 strength and identity of key players change as bloggers react to news and
to each other. From left, the two blog networks are pictured at the start of the
DNC (far left), the start of the RNC (middle image) and on election day.

 



Like the latest installment of a workplace scandal, social networking is widely and thoroughly discussed.  Experts and lay users alike scrutinize the 21st-century media’s ability to increase business, promote higher education, facilitate job-hunting, offer timely commentary and provide ’round-the-clock communication.

All those applications require a robust network, something UCI social scientist and Calit2 affiliate Carter Butts wants to understand. 

Specifically, he was interested in what causes the ebb and flow in blog networks. Butts and a student researched these fluctuations by analyzing blog-blog citation networks in more than 1800 English-language blogs during the 2004 U.S. presidential election cycle, a period during which blogs were first widely accepted as a tool for mainstream political organization. They chose blogs from six distinct categories and analyzed them to see what, if any, external factors led to network changes.

The results, recently published in the Journal of Social Structure, were somewhat surprising.

Not only did political events occurring during the run-up to the election change the network structure, but researchers found there was also an unrelated, regular “heartbeat” to the changes, one that followed daily and weekly cycles of activity in the broader community.

Butts and co-researcher B. Remy Cross wanted to make sure the blogs they selected for the project represented a wide spectrum.  “We tried to pick groups that might be emblematic of different trends that would enable us to do some comparisons. At the very least, we wanted to obtain a highly diverse sample of blogs that wouldn't necessarily behave in identical ways,” he said.

The six categories comprised:
•    blogs credentialed by the Republican National Convention;
•    blogs credentialed by the Democratic National Convention;  
•    a group known as ‘the influentials,’ which were the most well-connected blogs at the time of the study, according to an Internet firm that measures the  popularity of online sites;
•    blogs chosen by a link-trace method. This starts with a ‘seed’ and expands outward to blogs linked to that seed, then to blogs linked to those blogs and so on;
•    another link-trace chain with a political bent;
•    and finally,  a random sample of blogs with RSS feeds.

The researchers took samples every six hours to gauge how links from one blog to others had changed during that time period. They were not measuring readership or the number of click-throughs from one blog to another, but rather how often and at what times the blogs updated the links they published to point readers to other online material.

“When you cite someone today and you didn’t cite them yesterday. Or, you cited them once yesterday and you’re citing them 10 times today. That’s the evolving structure we’re capturing,” Butts explained.

They discovered that blog networks were impacted not only by specific events in the election cycle – primaries, conventions, major announcements and the like – but by natural rhythms as well. “We found these periodic changes, or as we call them, seasonal changes,” he said. There was a diurnal cycle, which fluctuated from day to night, and a weekly cycle, during which more network changes occurred on certain days of the week, regardless of whether specific events had occurred in that timeframe.

They were taken aback by the strength of the periodic fluctuations. “We were looking for effects of the electoral cycle and we thought the daily stuff was just noise to be eliminated,” said Butts. “But when we began analyzing this more closely, we realized this wasn’t just a minor fluctuation, it was very systematic and very powerful, and it suggested something important about how this type of world evolves.”

That’s significant because it opens a window into behaviors occurring in other types of human networks. It could be relevant for understanding disease transmission, communication networks or workplace interactions. “There may be implications for everything from marketing to political dissemination,” he said, adding that it appears there are certain times of day or days of the week that may be ideal for using social networks to spread information or to slow the process down.

The findings also impact network architecture because changing usage patterns can affect infrastructure requirements.

Perhaps most important to a sociologist, however, is the value of Internet-based research as a tool for examining social behavior. “We tend to overlook it, but humans are animals and we have periodic behavior. Societies have institutionalized periodic behavior and that shapes not only our individual lives but our collective lives as well,” said Butts.

“This kind of work illustrates the value of online networks as a context for studying human interaction more generally.”

-- Anna Lynn Spitzer