The six types of conversations on Twitter

A new report from the Pew Research Center and the Social Media Research Foundation says that Twitter conversations have distinct shapes -- at least six of them with differing structures reflecting their social activity. The research relied on a specialized software tool aimed at making big data perspectives more broadly accessible and represents a new kind of study for Pew, long known for public opinion surveys.

"For years and years telephone surveys have been the the anchoring methodology for how we understand the world" at Pew, says Lee Rainie, the director of its Internet Project, and it's likely to stay that way. But Rainie, who co-authored the new report, says the Internet has opened up a new world of data that can be used to add depth to our understanding of public opinion research.

"While the physical world has been mapped in great detail, the social media landscape remains mostly unknown," reads the report. "A more complete map and understanding of the social media landscape will help interpret the trends, topics, and implications of these new communication technologies."

That's why Social Media Research Foundation Director Marc Smith, University of Georgia Professor Itai Himelboim, and University of Maryland Professor Ben Shneiderman came together with Rainie to figure out the digital topography of Twitter.

The six shapes of Twitter networks

The result? Six recurring shapes: "Polarized crowd," "tight crowd," "brand clusters," "community clusters," "broadcast network," and "support network." Each has its own specific structure, and represents a visual way of classifying the type of interactions occurring within localized parts of Twitter. Some are more obvious, such as the polarized crowd, which shows two separate groups discussing the same topic but rarely interacting. But others are more novel -- displaying complex behaviors with varying inward and outward hubs.

The support network, for instance, was observed in Twitter accounts attempting to resolve customer complaints. The main brand account replies to many "otherwise disconnected users, creating an outward hub." This differs from the broadcast network, more likely to be found in news, where the main account gets replied to or shared by disconnected users to create an inward hub.

Here's what they look like:


(Pew Research Center)

The researchers believe the models will be helpful to businesses who are interested in being more strategic in their social media interactions.

Some of the most common questions Rainey receives about the Internet is just how entities should navigate the new social media world. And with this research, he thinks that will start to become easier. "These tools give the latitude and longitudes for organizations to see where they fit and where they might want to go," he says. "Just giving people an orienting point is one of the most important parts of this work."

But, Rainie says, "this is strictly observational work," noting that while six is a rather modest number of models, these same shapes continually resurfaced in their research.

The tool

NodeXL, the data tool used to identify those shapes, was first developed by Smith's team while he worked at Microsoft and is hosted on the company's open source project hosting Web site Codeplex, but is now a project of the  Social Media Research Foundation. The software package is an add-on for the Office 2007 or 2010 version of Excel released under the open source Microsoft Public License. It helps users collect social network information and create more complex visual representations of networks from spreadsheets.

Smith likens the images created by NodeXL in the Pew report to aerial photographs of crowds online -- and says that's where civil dialogue is increasingly heading. "I'm a sociologist and I think it can be argued that a lot of society moved online in the past 10 years," he explains. "Facebook, Twitter, and the like are the new global square."

"The Internet lets us get the pulse of culture as it's happening as has never before been possible," says Shneiderman. But as social engagement moves online, some researchers have found it harder to capture and explain social behaviors in their full context. NodeXL, the researchers say, helps solve that problem by translating larger data sets into networked visualizations.

"We don't see all of Twitter, but we are seeing neighborhoods of Twitter," Smith admits. The report notes Twitter  is not reflective of the full population -- with only 18 percent of U.S. Internet users on the service and reactions that are sometimes at odds with the general public opinion. But Smith says it represents an interesting area to gather data from, because it seemed to have a "more important role in public affairs" than some other social networks due to its rapid pace, and the data was more openly available.

Why the network matters

In his field, Smith said some researchers are limited from harnessing the power of larger datasets by their technical abilities. But with NodeXL, his team believes they can offer a way to use data to create structures that convey more information, with less difficulty for users.

"We're bringing the power of big data analysis to a bigger audience," he said. And Smith does believe that more complex representations are the next natural progression in how we visualize information.

"Every culture gets a data structure equal to their computational power," Smith argues, saying that society has reached the point where visualizations of networks are more in line with our capabilities than the pie charts, bar charts, or even linear regressions of the last century. "It's all going to be about networks. "

The team that did the project itself is a unique network -- bringing together a variety of researchers from across the globe and across research fields.


NodeXL Team Network (Social Media Research Foundation)

"One of the rich things about this project is the interdisciplinary aspect," says Shneiderman, who teaches computer science and says his students frequently use the tool. NodeXL has been developed over the past seven years and downloaded over 240,000 times to date.

Exploring digital topographies

For Pew, this research represents a first step in using this tool to study online communities as a supplement to its traditional public opinion methodology. And the researchers recognize there are limits to where their observational research may apply.

For instance, "it's certainly the case that different social media platforms may show different shapes," says Smith. The researchers say there is some initial studies that suggest Wikis and blog formats map differently, and other evidence suggests that the shapes they have identified might show up in message boards and other places with a reply feature.

But the discovery of the landscapes of those platforms is part of the adventure. "Whether they are different or similar will be fascinating, if for no other reason than the uniqueness of the relationship," says Himelboim.

"I liken this to the Darwin exploration or a botanist going into the Amazon noting there are certain kinds of birds and animals," agrees Shneiderman. Only instead of discovering new species of fauna and flora, the researchers hope the tool helps uncover the social dynamics of networks.

"If we created the digital camera for cyberspace, perhaps Pew is becoming the national geographic of cyberspace," jokes Smith.

Andrea Peterson covers technology policy for The Washington Post, with an emphasis on cybersecurity, consumer privacy, transparency, surveillance and open government.
Comments
Show Comments
Most Read Business
Next Story
Brian Fung · February 20