Maps

How Twitter Is Changing the Geography of Communication

New research suggests that location plays a smaller role now in who we talk to and what we talk about.

Since academics first began studying communication, they’ve been trying to figure out who we talk to and how those networks change with the invention of new mediums of interaction. Who you could talk to, and even what you might talk about, obviously differed between the eras of the covered wagon and the cell phone. And now we have an instantaneous, global and (mostly) free platform for talking to virtually anyone: the Internet.

So how has it altered the real-world geography of communication? Some previous efforts to address this question have come out of the workplace (researchers can’t query Google for all of our gmail data, but large international companies can do this with their own employees). There’s only so much to be learned, however, from the email correspondence between a company man in L.A. and his coworker in China.

"The holy grail has been to look at people in real life, to look at people outside the workplace, to see when people are on their own, communicating as part of their general course of life, how has the electronic revolution changed that?" says Kalev Leetaru, a University Fellow at the University of Illinois Graduate School of Library and Information Science. "That has been a difficult thing to look at because there hasn’t been much data."

Now, however, there is more data than most computers can process coming from public social networking platforms like Twitter. We wrote in late 2011 about some early research suggesting that many Twitter users in fact follow other people located within their same city, evidence, Richard Florida wrote, that the Internet is reinforcing the value of place instead of eliminating it.

But now that Twitter is a few years older – and considerably more global – Leetaru and several colleagues have conducted a massive new analysis of the site that suggests the opposite: "In effect," Leetrau says, "location plays a much lesser role now in terms of who we talk to, what we talk about, and where we get our information."

There are two ways to think about connectivity on Twitter: People "follow" each other (a fairly passive form of interaction), but they also actively engage other people through retweets or direct conversation (mentioning another user’s handle). This first kind of relationship is akin to subscribing to magazines. We may know, for instance, that you get Cat Fancy, TIME, and The Atlantic. But that information tells us nothing about which articles you actually read, or which magazines merely decorate your toilet.

"In reality on Twitter, even though you may be following a lot of people, you’re probably paying attention more to some than others. But there’s no way of recording that," Leetaru says. "What we can do is say 'well, here are the people you’re talking to. Here are the people you thought were interesting enough to retweet.'"

That earlier Twitter study looked at about 2,000 pairs of users back in 2009 who were connected through a "following" relationship. This latest project, with the findings published on First Monday, examined 39 days of the Twitter Decahose (which includes 10 percent of all tweets sent worldwide every day) from October and November of last year. In total, the cache included more than 1.5 billion tweets from more than 71 million individual users (averaging 38 million tweets a day).

Twitter’s data feed includes all kinds of meta-data you may never notice when you’re scanning the platform, including geographic location for some users (some people manually self-identify their city, while Blackberries and iPhones tag exact location latitude/longitude coordinates for tweets sent from a mobile device). In total, the researchers were able to geo-locate about 3 percent of the users in their database, at varying levels of preciseness.

Plot all of the exact locations on a map of the world, and it looks like this:

"Mapping the Global Twitter Heartbeat: The Geography of Twitter." Larger version here.

That picture looks similar to images from NASA’s Visible Earth City Lights project, which tracks global urbanization through the electricity that’s visible from a satellite. In fact, if you overlay the Twitter map atop the NASA one, they line up pretty neatly. Here, tweets are red, and night lights are blue. White areas have a roughly equal balance of the two:

Larger version here.

"Where there’s power, there’s Twitter," Leetaru says. "The more power, the more Twitter. That tells us that quite literally even in the most remote areas of Africa, if there’s power there, chances are there’s Twitter there at a relative density of power."

There are a few exceptions: Iran and China show fewer tweets than we’d expect given the electricity levels there because Twitter is censored. But for the most part, once we realize Twitter is nearly everywhere electricity will allow it to go, that picture allows us to examine how (and if) all of these people, all over the world, are interacting with each other.

On that front, this research produced several fascinating findings. Users tweet in numerous local languages. But those who self-identify a location in their bio almost always do so in English. “We see this across all languages of the world outside Asia,” Leetaru says. “That suggests that these people don’t care whether the rest of the world knows what they’re saying – but they do want the rest of the world to know that they exist.”

The researchers also mapped the distances – accounting for the curvature of the earth – between people who actively engage each other on Twitter, either through retweets or mentions (where both users had some geographic data available). And a curious pattern emerged there as well: The more two people interact, the closer they are likely to be geographically – but only up to a point. People seem to be physically closest to those users with whom they interact nine times in a given month, or about twice a week. But from there, the pattern reverses. Users who retweet or reference another person just once are on average about 800 miles apart. That drops down to about 600 miles – still a pretty impressive distance – before rising from there. But why nine? Leetaru isn’t sure (perhaps this is a new kind of Dunbar’s number?).

Most people on Twitter are simultaneously communicating with both people close to them and people on the other end of the globe. This map illustrates those connections using retweets between pairs of geo-referenced users:

Larger version here.

All of this raises one other question: What are all these people talking about? To address this, the researchers also examined the geographies associated with news (in the form of links) that people share on Twitter. The results: 26 percent of links were to stories related to the user’s city; 37 percent were related to events or news more than 100 miles away; and 46 percent came from stories about places more than 600 miles away. Someone in Chelsea, in the United Kingdom, for instance, linked to a story in the Times of India about a hanging in New Delhi. Someone in Murmansk Oblast, Russia, tweeted a German article about Antarctica. Lastly, from the paper:

Perhaps the most complicated geographic chain was a user near Dusseldorf, Germany who tweeted a link to a United Kingdom newspaper’s story about an article in the China’s People’s Daily about a story in the United States’ The Onion about North Korea (Associated Press, 2012).

This whole project was conceived as an initial survey of Twitter’s geography, from which plenty of other questions will obviously arise (why nine? And why are modern humans assembling such historically massive collections of “weak links”?). But this picture does begin to confirm that the electronic revolution in communication is, in fact, changing who we talk to in ways that at least partially unmoor communication from location.

“People are yearning for some way to move beyond geography, and Twitter is providing that to them,” Leetaru says, “and we see  that reflected in how Twitter is used.”

About the Author

  • Emily Badger is a former staff writer at CityLab. Her work has previously appeared in Pacific StandardGOODThe Christian Science Monitor, and The New York Times. She lives in the Washington, D.C. area.