Tricked into Supporting: A Study on Computational Propaganda Persuasion Strategies

The study reported in this paper aims to theoretically and empirically explore computational propaganda (CP) – a systematic process of political misinformation perpetrated on social networking platforms by automated agents with the aim of increasing support for specific political stances – focusing in particular on the factors determining its potential effectiveness. The claim maintained throughout this paper is that, among the possible factors determining this effectiveness, a pivotal one is represented by the design of CP messages themselves. Indeed, the hypothesis underlying this investigation is that the way CP content is created and presented is not casual, but deliberately designed to embed in it a set of persuasion strategies aimed at triggering a specific cognitive deliberation: considering misinformation as factual. Drawing from the Dual Process Theory of Cognition, the argument proposed is that info-cues contained in CP messages play a pivotal role in determining the likelihood of CP effectiveness. To test this hypothesis, a two-step analysis characterized by a mixed-method strategy has been implemented. To identify and collect CP messages, a machine learning algorithm able to perform bot-detection has been developed, while to analyze the content of those messages, a combination of qualitative and quantitative text analysis techniques has been employed. Lastly, preliminary results are presented and future work discussed.