The effects of threat frames in media reports have been observed across numerous contexts such as civil war, crime, international relations, and race relations. These studies have largely focused on the more simplistic role of media producers and how audiences receive and decode the information. I take a more complicated approach by analyzing the relationship between media threat framings and calls for violence in comments sections which may influence other readers. This analysis considers the various forms of online media reporting including mainstream, alternative and social media depictions of both left and rightwing protests in the United States between 2016-2020. I do not attempt to make a causal link between threat frames, calls for violence and actual violence – however this research helps to explain how violence becomes normalized as part of political discourse.

I create a large dataset by gathering online media reports related to high-profile political protests and their comments sections with custom Python 3 scripts. I then carry out a quantitative text analysis of this corpus using supervised machine-learning approaches. A variety of algorithms are used to classify reports based on the ideological position of the source, the group or cause behind the protest, the presence of threat framings, and calls for violence in the comments sections. I then analyze how calls for violence are influenced by the format of the media report (i.e., mainstream, alternative or social media), the ideological bias of the media source, and the protest’s cause.