Detecting Mass Protest through Social Media

Abstract

Building on the existing understanding of the dynamics of mass protest, this paper argues that social media data can be used to detect the occurrence of such protests. It outlines a theoretical framework arguing that during times of mass mobilization, network central actors will geographically converge upon city centers, and that the relative magnitudes of social media activity in the center and periphery of the state can be used to detect the occurrence of protests. This article presents a new dataset of 2.2 million geocoded Ukrainian tweets, which are used to empirically test the theory against the observed 2014 Euromaidan protests in Ukraine. By relying completely on count and network data rather than keywords, hashtags, or other contextual clues from the content, this technique is portable across language barriers and national borders.