The New York Times recently reviewed a study by Yosh Halberstam and Brian Knight that finds social media users live in an echo chamber where they are more likely to hear people with whom they already agree. Interestingly, among the expert the Times interviewed for the article was an NYU doctoral student, Pablo Barbera, who has also researched this topic. Reading Barbera's comments in the Times makes you think he agrees with the reporter' stake on the featured report. But in a related Wall Street Journal article about Barbera's research, we learn that he draws the opposite conclusion from his own study than Halberstam and Knight drew from theirs.
Barbera's research looks at who people follow on Twitter over time, focusing not on what people say but what they read. He found that while people initially follow like-minded tweeters, establishing an echo chamber in their timeline, as time passes they follow more people of increasingly diverse opinion. From my own experience on Twitter, this pattern rings particularly true.
Perhaps the different conclusions by the authors are due to their diverse methodologies. Studies observing a snapshot of subjects' followers are incapable of identifying shifting exposure to more diverse opinions. And studies that look only at subjects' tweets over time would still miss the evolving balance of viewpoints subjects would read in their timelines.
If we think through the pattern Barbera describes for the evolution of tweeters' timelines, it makes a lot of intuitive sense. When we start out on Twitter we follow news sources we like, our personal friends and celebrities we like. These all tend to share the same perspective.
Over time we connect with conversations about various topics. In these conversations we are exposed to the back and forth of opposing and agreeing tweets on the topic, alike. Many of the tweets from people with other perspectives will inevitably catch our eye. They may be provocative, informative, funny or otherwise engaging enough to elicit us to follow the tweeters who wrote them.
Similarly, people we follow mention other tweeters, including people with whom they disagree. But when people we enjoy following mention other people, our curiosity is aroused. Again, exposure to these other voices inevitably leads us to follow some of them.
Over time our initially one-sided collection of friends diversifies. As a result, we will see more and more perspectives reflected in our timelines. Even if what we say in our own tweets is slower to reflect these moderating forces, what we read still exerts influence over us, over time.
Further, if these moderating forces did not exist, we would expect that polarization towards the extremes would occur. But even the snapshot and message studies seem not to suggest this. Not only are these other studies incapable of showing changes in exposure to moderating voices, they tend to show the existence of polarization rather than increasing polarization. And given the intuitive logic of Barbera's study based on the aspect of Twitter that drives opinion formation, namely exposure to information, it seems more likely that over time, Twitter helps us find common ground.