A classification of Antifa Twitter accounts based on social network mapping and linguistic analysis

Abstract

Where machine learning and other quantitative methods have been used for classification purposes in helping researchers manually identify extremist Twitter accounts, they have chiefly been used to classify far-right and ISIS/Jihadi accounts. There are no large-scale quantitative studies of far-left (specifically Antifa) accounts. Since the election of Donald Trump as the 45th President of the USA, there has been a rise in public acts of violence, vandalism, de-platforming and harassment of political speakers by Antifa. Twitter plays an important role in helping Antifa to organise, radicalise individuals, harass other users and share propaganda. This paper attempts to provide a classification of Antifa accounts based on network mapping and textual clues. We found that approximately 1.1% (644) of Antifa sympathetic accounts can be classified as Cognitive Extremists. Violent Extremists (313) accounted for 0.5% (313) of all Antifa sympathetic accounts. However, members of that group had an average following of 5370 followers compared to an average of 3804 for Cognitive Extremists and 707 for all Twitter users, indicating widespread support from extreme and violent action from all three categories of Antifa accounts.