Red state, blue state? On Twitter, never the twain shall meet

If you want to know how communities form on Twitter, you'd first need to know what the topic of conversation is.

If it's politics, you can almost be certain that people are talking to each other in partisan groups, mainly agreeing with each other, and pretty much ignoring the other side. If it's an entertainment or a professional subject, groups form tightly and talk deeply in distinct topic groups.

On the other hand, massive numbers of people may talk about brands, but have little interconnection between each other. And global news will likely generate large numbers of social hubs that each generate multiple conversation groups.

These were some of the findings of "Mapping Twitter Topic Networks: From Polarized Crowds to Community Clusters," the latest study from the Pew Research Center and the Social Media Research Foundation.

The study, released Thursday, found six distinct types of communities that form on Twitter, each dependent on what's being talked about. Each is functionally different and bears little resemblance to the others. Yet what was striking (though perhaps not all that surprising in retrospect) in the findings, suggested the report's authors, is just how much behavior in social media conversations -- on Twitter, at least -- mirror that of society in general.

Social media is increasingly home to civil society, the place where knowledge sharing, public discussions, debates, and disputes are carried out. As the new public square, social media conversations are as important to document as any other large public gathering. Network maps of public social media discussions in services like Twitter can provide insights into the role social media plays in our society. These maps provide insights in a similar way to aerial photographs of a crowd, showing the rough size and composition of a population.

The six types of communities were:

• Polarized crowds, which tend to form around politics, and are large and partisan, and which seem to essentially ignore opposing views rather than engage them in debate.
• Tight crowds, which form around professional, educational, or passionate interest, and which feature heavy interconnected conversation within each group.
• Brand clusters, which concentrate around products and celebrities, and which are massive in scale, but feature almost no interconnectedness.
• Community clusters, which form around global news, and which focus in social hubs at the center of multiple topic-oriented group clusters.
• Broadcast networks, in which prominent news sources tweet breaking news or commentary and large numbers of people re-tweet them in what Pew called an "inward" hub-and-spoke system.
• Support networks, such as those that form around company customer support accounts that respond to multiple disconnected user concerns and complaints, and which form an "outward" hub-and-spoke system.

But while these distinct groups seem to behave in a similar manner as society in general, they do not necessarily represent the entire Twitter user base or the Internet at large. That's because, the report found, those who actually bother to post on Twitter are a small and distinct subset of the much larger Twitter population. Further, according to the report, Twitter users make up just 18 percent of Internet users, and only 14 percent of the overall adult population.