It seems that social media is a veritable tinder box for arguments, with the slightest thing setting off a firestorm.  A recent study from a team of Finnish academics has set out to map just how arguments unfold on Twitter.

They trawled through the Twittersphere to try and understand the very basis of arguments online, including how they form and how we can characterize them.  They believe that, as a result of their analysis, they can actually spot controversies unfolding in real time and distinguish them from the usual chatter on the site.

The anatomy of an argument

The study builds on previous works that have explored how arguments look on Twitter, but it’s believed to be the first of its kind to try and predict them before then germinate.  Their central thesis was that arguments have a unique conversational structure that marks them out as different, and therefore possible to spot.

This includes things like the network of connections between participants, who agrees with whom, the structure of any endorsements involved, and of course the sentiment of the debate.

After studying a range of both controversial and benign conversations, the team identified a number of emerging trends.  For instance, in controversial topics, retweet and follow graphs would reveal clusters of users that suggest a polarized debate forming.

The end result is a beautiful firework like representation of controversial topics exploding onto the Twittersphere.  In the images below, the most controversial topics are shown in images a, b, e and f.

The polarization, or lack of, is clearly visible as people gather on both sides of controversial debates, or cluster more around the norm in more staid discussions.

The researchers went on to develop tools that can measure this level of polarization, which they believe can ably predict the controversy of a discussion as it unfolds.

They have tested these tools on a number of datasets and whilst they’re not perfect, they are an improvement on existing methods.

Whether you’re a company looking to spot controversy around a particular product or service, or even policy makers looking to predict civic unrest, this kind of tool would be a great asset to have.  Whilst more research is clearly needed to fine tune the tool, it seems likely that we will soon be able to spot controversies on Twitter at a very nascent stage.

The challenge then, of course, is managing to put out the firework, which is easier said than done!