Can we predict outbreaks of hate online ?

Add to my custom PDF

Cyberhate on Social Media in the Aftermath of Woolwich: A Case Study in Computational Criminology and Big Data

Specific events can influence the prevalence and severity of crimes with a prejudicial component. In particular, terrorist acts have been found to function as “trigger” events that “validate” prejudicial sentiments and tensions. These social responses contribute to the overall impact of an event. Contemporary online spaces, such as social media, offer new public spheres for expression and amplification. Cyberhate has recently been identified as a social problem on these platforms, one that must be addressed. However, the connection between these “trigger” events and the production of cyberhate remains largely anecdotal.

Williams and Burnap study the shape and size of cyberhate on Twitter, following the Woolwich terrorist attack in 2013. The sample consists of 427,330 tweets about the event, from a fifteen day period following the Woolwich attack. Due to the large sample size, the authors developed an interdisciplinary method called computational criminology. This method allowed them to analyse the data using advanced computing techniques. A supervised machine classifier that learned the features of hateful tweets toward minorities was employed.

Findings demonstrate that the event amplified deviant social reactions. An increase in cyberhate is evident within the first few hours following the terrorist event. Beyond the initial impact stage, when immediate responses to the event occur, the duration and diffusion of cyberhate is more inhibited. Cyberhate is limited in the reaction stage, where focus shifts to wider implications and issues of the event. The figure below shows the “half-life” of cyberhate tweets and retweets, evidenced by the rapid de-escalation immediately following the attack.

The study demonstrates that deviant reactions, in this case cyberhate, can form part of the social response to a trigger event. The study illustrates how social media presents a rich new form of data, which assisted by computational methods, can extract meaningful insights into social processes. Social media can act as early warning systems for the amplification of deviance beyond an event itself. Practitioners need to focus on interventions within the impact stage to encourage faster and more widespread de-escalation.

Social media can act as early warning systems for deviance beyond an event. Practitioners need to focus on interventions early on to encourage a faster and more widespread de-escalation.