NEW YORK (AP) – People take to Twitter to talk about everything from politics to breakfast to Justin Bieber in what feels like a chaotic stream of messages. So it may come as a surprise that the conversations on the short messaging service fit into just six distinct patterns.
The Pew Research Center, working with the Social Media Research Foundation and using a special software tool, analyzed and mapped millions of public tweets, retweets, hashtags and replies that form the backbone of Twitter chatter. The resulting diagrams show how people, brands, news outlets and celebrities interact on Twitter, depending on the topic of conversation.
When it comes to politics, for example, Twitter's citizens tend to form two distinct groups that rarely interact with one another, divided along liberal and conservative lines, according to the report, which was published on Thursday. Liberals tend to post links to mainstream news sources, while conservatives link to sites with a conservative blend, according to the study, whose authors likened their methods to taking aerial photos of crowds gathered in public places.
The researchers are quick to note that not everyone uses Twitter – only 14 percent of the U.S. population – and not all who do use it to talk about politics, for example. Still, looking at how conversations flow on social media can provide new insights into how people communicate in a way that was not possible until very recently.
"You could never do that in the old days when you were running around with a pen and clipboard," said Marc A. Smith, one of the study's authors and director of the Social Media Research Foundation.
What emerged in maps of political conversations that the liberal and conservative groups are not even arguing with one another. Rather, they are "ignoring one another while pointing to different web resources and using different hashtags," according to the study.
The telephone polls that take the pulse of the country about everything from politics to race, religion and technology will continue to form the research center's backbone. But Lee Rainie, one of the study's authors and director of the Pew Research Internet Project said there are other kinds of data that deserve exploration. Looking at social media – something that large swaths of people participate in – can give insights to important information about people's lives.
Here are the other five types of conversations:
– People who talk about well-known brands on Twitter tend to be disconnected from one another, focusing only on the topic at hand and not really interacting with each other. The study calls these "brand clusters." One graph, that looked at mentions of Apple, found that users didn't follow, reply to or mention any other person who also tweeted about the company.
– People who tweet from a social media conference, or about another highly specialized topic tend to form tight crowds of people who are connected to one another as followers. There are only a few users who are not connected to at least a few others in the group.
– "Community clusters" happen when several, evenly sized Twitter groups are connected to each other. In a sense, these can be compared "to people clustering in different stalls at a bazaar." The conversations in this group share a common broader topic, whether that's Michelle Obama or a tech conference, but each cluster takes a different focus.
– "Broadcast networks" are often media outlets or prominent social media figures with a lot of followers who repeat the messages such outlets send out.
– A Twitter "support network," is the last major conversation type. These conversations usually involve a large company, such as a bank or airline, that listens and replies to consumer complaints. When mapped, the interactions in these groups tend to look like a bicycle wheel hub with many spokes.
The Pew study: http://bit.ly/1h1fEDy
Reach Barbara Ortutay on Twitter at www.twitter.com/BarbaraOrtutay
Copyright 2014 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.