Abstract

The presence of illegal, harmful content, rumors, misinformation, and Twitter bots has consistently brought the social media platforms such as Twitter into the spotlight. Therefore, it is advisable to exercise caution when analyzing tweets. To establish the credibility of any patterns and findings derived from tweets, it is essential to thoroughly investigate the source and authenticity of the tweets in question. This paper advances in this direction by introducing a novel approach involving bot detection and a comparative analysis of human and bot-generated tweets related to the farmers' protest. A framework for knowledge differentiation is deployed to accomplish this goal. The framework unearths the global perspectives of people about Indian farmers’ protests, in the form of stances, the results of which serve as nuggets of knowledge derived at the lower level of abstraction. Unexpected results of stance detection motivated the study of bot detection in each tweet of each stance. Knowledge discovered by bot detection and characterization studies was thus built over stance detection and yielded higher-order knowledge nuggets, which identified the widespread presence of bots in tweets both for and against the protest, thus establishing the misuse of social media platforms like Twitter to influence and control the narrative of the social events that significantly impact people’s lives. Characterization of issues being tweeted by humans vs. bots in favor of and against farmers’ protests was accomplished by conducting a comparative analysis of N-grams in each category. Vocabulary analysis established that texts tweeted by bots mimicked the vocabulary pattern of the tweets by human users. Research inferences such as these can be invaluable for policy makers, enabling them to gain a macro-level understanding of the situations on the ground level and leverage such information for making policy decisions, in order to be prepared to handle similar situations in the future. 

Statistics