These Algorithmic Visualisations of Twitter Controversies Look Like Fireworks

Twitter controversies can be confusing for the average user. They blow up without warning, and you might not even realise that you've been embroiled in a frenzied hashtag war until it's too late. But that could be set to change if researchers can spot controversies developing in real-time.

MIT Technology Review highlights an Arxiv paper by a team headed up by Kiran Garimella from the Department of Computer Science at Aalto University in Finland, who have found a way to distinguish polemical conversations from non-polemical ones by visualising controversy in a cluster of tweets.

The researchers tested out their theory by comparing ten "neutral" conversations with ten conversations associated with controversial hashtags. They mapped out the structure of these conversations by analysing the activity around the tweets.

In order to build the visualizations, they took activity related to a single topic of discussion and fed it into a graph partitioning algorithm which produced two "partitions" (red and blue) that represented sets of Twitter users that agreed or disagreed with each other. The level of controversy is defined by how separate the two partitions are from each other.

Just take a look at the picture below: On the left, visualizations (a) and (e) represent the controversial #beefban hashtag, which blew up after beef was banned in Maharashtra, a western state in India. The level of polarization in the debate is reflected by the space between the red and blue partitions. A Twitter conversation that didn't generate controversy, #sxsw, looks like a more cohesive firework in (c) and (g).

The researchers argue that the structures of controversial clusters of tweets are different to others. Controversial conversations, they say, can be deduced by analysing how tweets are endorsed; who agrees with who in the thread; the network of connections between all involved in a discussion; and whether the overall mood of the conversation is positive or negative.

"Identifying these topics is a first step towards creating systems which pierce echo chambers," write the researchers in the Arxiv paper. Echo chambers refer to the repetition and endorsement of certain beliefs within an enclosed system.

"Our work is motivated by interest in observing controversies at societal level, monitoring their evolution, and possibly understanding which issues become controversial and why," they continue.

Controversies on Twitter have been studied before, but these are usually based on pre-identified disputes, and are mostly political in nature. The researchers from Finland want to study controversies that develop around political, cultural, and economic topics, and to create a method of spotting a social media kerfuffle more easily.

The researchers ultimately want to understand how social media users perceive the world. They believe that spotting potential controversies as they appear and evolve could also allow them to balance polemic opinions by injecting contrarian voices into people's news feeds in the future.

"Quantifying controversy can provide a basis for analyzing the 'news diet' of readers, offering the chance of better information by providing recommendations of contrarian view, or trying to deliberate debates and connect people with opposing opinions," they write.