Marina Rizzi (CCA and UniTO)

Self-Regulation of Social Media and the Evolution of Content: a Cross-Platform Analysis

Abstract: This paper investigates the effectiveness of Twitter’s policy against racist hate speech, introduced in December 2020, in reducing abusive content. Using a dataset of 8 million tweets from users discussing race-related topics, and machine learning techniques to identify instances of hate speech and racist hate speech, I find a significant reduction in abusive content in Twitter, especially from racist users, after the policy’s implementation. Additionally, using a dataset of users with accounts on both Twitter and Parler, a less regulated social media platform, I examine whether abusive behavior shifted to Parler following the policy change. Findings suggest an overall increase in hate speech on Parler post-policy, in particular for users that were active in Twitter in those months. Text analysis methods indicate a possible shift in topics between the two platforms. These results highlight the complexity of online content regulation and the potential for spillover effects to alternative platforms.