When someone begins to rant on Twitter, we sometimes say that they’re tweet-storming, as if the successive messages were so many raindrops. Now, a new study suggests that that metaphor might be more apt than we knew. Twitter, it seems, has many storm systems, and from the right angle they’re surprisingly beautiful.
This revelation comes courtesy of a study by Kiran Garimella and other researchers at Helsinki’s Aalto University. Garimella and his colleagues set out to map controversies in social media settings. As the MIT Technology Review reports, the team “found a way to spot the characteristics of a controversy in a collection of tweets and distinguish this from a noncontroversial conversation.” In the process, they created a set of surprisingly beautiful charts that represent divergences of opinion in colorful bursts that suggest storm systems sweeping across the Internet.
Previous inquiries of this kind, Garimella claims, had focused primarily on individual issues such as elections and the deaths of world leaders, and used data sets that were selected in advance. Garimella’s work, by contrast, set out to more systematically detect and quantify controversy. By looking at patterns of hashtag usage and retweeting, the researchers were able to show how certain issues pull users apart, separating them into self-contained echo chambers.
Understood in these terms, the visualizations above make intuitive sense. Where the two colors diverge (a, b, and e), opinion on an issue is split, with little contact between the sides. Here, the researchers mapped topics such as debate over a beef ban in India and Russian protests after the death of opposition leader Boris Nemtsov. Those in which the colors meld—including conversations that occurred around the Germanwings crash or the SXSW conference in March 2015—are less divisive.
Garimella and his collaborators believe that this research isn’t just about producing cool images. “Ultimately,” they write, “we would like to understand how users perceive the world through the lens of their social media feed.” In the long term, this work may help push back against the ideological bubbles that social media helps to perpetuate.
For now, however, the applications of their work may be limited. As they acknowledge in their write-up, their methodology is limited to Twitter and can’t capture the complexities of controversies with “three or more competing views on the field.” Others might question their reliance on sentiment analysis, a frequently criticized method of evaluating linguistic expression of feeling that sometimes oversimplifies the real complexity of issues.
Nevertheless, the results of Garimella’s work are compelling, not least of all because they paint such striking pictures of the ways we argue today.