Characterizing Arabic Computational Propaganda on Twitter

Abstract

Background: The rise of social media has nourished the "computational propaganda" phenomenon. Propagandists rely on advanced artificial intelligence methods to change their writing style with every new campaign, allowing them to evade detection methods easily. Detecting computational propaganda in Arab countries has become a trending topic in social media research. Most of the proposed methods have been focused on detecting propaganda based on the writing style. Unfortunately, these approaches were marred with significant limitations because propagandists' traits and behaviours must be considered. Objectives: This study aims to demonstrate the value of characterizing Arab computational propaganda on Twitter to close the research gap. It follows a data-driven approach to investigate the main characteristics that can distinguish Arab computational propaganda on Twitter. Method: It follows a scientific approach to obtain and combine data from reliable and propagandistic users who discuss the same topics. Then, it provides a deep analysis of two communities that discussed different topics. It characterizes the key features that can be used to differentiate between them. Finding: The findings show that around 70 per cent of propagandists rely on artificial amplification by retweeting to produce an echo chamber supporting their viewpoints. The propagandists' following-to-follower ratio is between 0.8 and 1, indicating they are a coherent army that supports each other. 98 per cen of the propagandists' users participate in diverse topic discussions, indicating that topic diversity and publishing volume are very important features for detecting propaganda on Twitter. Publishing periods can strongly help in detecting propagandists. Novelty: The study offers early evidence on social media regarding the behaviour of propagandists' users and messages. This study enlightens future research by identifying the important features needed to propose anti-propaganda detectors.