Testing Impact of State Narrative Control Tools: Evidence from Twitter Data

• In Event: Social Media and Violence

Abstract

In this paper, we examine the impact of an emerging narrative control toolkit in authoritarian and weakly democratized settings. As part of the toolkit, powerful regimes use a combination of offline threats and violence against critical voices, censorship of the mainstream press, and counter-messaging campaigns to counter dissent, discredit rival political actors, and elevate pro-regime narratives. Existing theoretical perspectives offer competing predictions on the efficacy of this strategy. Prominent perspectives from comparative politics and democratization theory see the new media – such as Twitter and Facebook -- as a “liberation technology.” They suggest that the new media provides populations with the means to obtain the censored information, articulate preferences, and organize politically. Competing perspectives from cyber warfare, information warfare, and national security threat assessments see the new media as offering much less utility at best, and at worst being susceptible to sabotage. We assess these perspectives in Pakistan, a major weakly democratized setting with a history of military involvement in domestic politics. We have collected a new dataset of 1.2 million politically relevant twitter accounts and 9 million tweets from 2018, which we call Pakistan’s political twitter. We use supervised machine learning methods to classify the data for content and sentiment. The findings have the potential to inform scholarship and recent policy debates on the potential and limitations of mobilization on social media.

Authors

• Asfandyar Ali Mir, Stanford University

• Tamar Mitts, Columbia University