Bots on Twitter: Evaluative Analysis on non-authentic tweets

Abstract

Among the information manipulation strategies, inauthentic accounts have been gaining strength, especially when related to issues about politics. The social network that makes this action easier is Twitter, with its bots and hashtags system. With this in mind, in this article we intend to locate, analyze and categorize occurences of evaluation in inauthentic accounts that encourage the spread of beliefs and opnions about the current brazilian political scenario. Through use of the Bot Sentinel, which use machine learning based on a mathematical model (ZHANG, 2020) to predict the authenticity of a user and expose inauthentic accounts and their connections with the most commented themes, we collect 60 tweets posted between may and october of 2020. From that, we selected 10 tweets from non-authentic accounts containing the most popular hashtag in your month in its said period for each month of the gathering. The theoretical apparatus on which we rely is the appraisal system, more precisely the attitude subsystem (MARTIN; WHITE, 2005), to see how such evaluations operate to build relations of alignment and relationships between writers and their readers. The results indicate the use of evaluative standards of positive capacity for the president of the republic and of negative property to denigrate your opponents’ image, accentuating the idea of Us vs. Them (BORGES; VIDIGAL, 2018).