Political online astroturfing in the 2020 United States presidential election campaign

After the 2016 United States presidential election and the Cambridge Analytica scandal, digital politics with the use of cryptic algorithms was revealed to the wider public but certainly did not stop but rather developed and increased, culminating in Donald Trump’s 2020 fierce election campaign. Since the early 2010s, one of the growing concerns for the public sphere has been the use of social bots on Twitter during election campaigns around the world, an estimated share of bots ranging from 5 to 20% of all Twitter accounts. Such bots can quickly flood Twitter with propaganda, are difficult to detect by humans, and can be orchestrated in a deceptive campaign of online astroturfing – the use of bots in social media to imitate a bottom-up movement and create the impression that the action is supported by a large group of non-related people (social proof). Examples of such behavior have been found in the ongoing Russian-Ukrainian conflict, the Brexit referendum, and the 2016 United States presidential election, among others, and thus was expected to take place during the 2020 United States presidential election as well. The purpose of this paper is to prove the use of political online astroturfing on Twitter during the 2020 United States presidential election as well as to measure its characteristics in order to reveal patterns that may be useful for detecting and studying online astroturfing in future political campaigns. We analyzed Twitter accounts that had been using the most popular hashtags related to the election campaign during two periods: at the beginning of 2020, before the COVID-19 outbreak and within two weeks before the election date. These accounts were tested with the Botometer tool – an algorithm assessing the probability that a Twitter account is in fact a social bot, proving that 8.64% of active, politically-oriented Twitter accounts were algorithm-controlled, not actual humans. We also analyzed a wide spectrum of the Twitter accounts’ metadata, noticing particularly interesting patterns regarding features such as replies to tweets ratio, retweets to tweets ratio, the share of retweeted tweets, numbers of followers and following, tweets per day ratio, and various distributions of botscores. Big data analysis and comparisons conducted between bot and human accounts, different periods, and with outcomes previously obtained in presidential election campaigns in other countries allowed us to propose a distinctive model of political online astroturfing in the United States, including features such as high automation of Twitter communication, high retweet ratio, and large follower and following networks, among others, suggesting a relationship between astroturfing patterns and the political-media system of a country as well as its level of democracy and political competition development.