The Online Poisoning of the Minds: Misogynistic Radicalization through Social Media

This project investigates the relationship between the linguistic characteristics of short-form videos on social media and users’ misogynistic radicalization, as manifested through misogynistic attitudes in the comment sections. Focusing on the online ecosystems surrounding prominent figures in the "manosphere," I examine how both the specific lexicon used in the video content — violent, condescending, and therapeutic — and the dehumanizing language directed at women correlate with user responses. This research employs a self-collected dataset comprising 507 videos and 11,491 comments, evaluating for each video a Normalized Lexicon Score (NLS) (for each lexicon category) as well as a Normalized Women Dehumanization Score (NWDS) and a Normalized Question Score (NQS). I apply a computationally robust approach by using an LLM to categorize user comments into two dummy variables: misogynistic/non-misogynistic and aggressive/non-aggressive, allowing for a nuanced understanding of the responses. Using ordinary least squares regression, I aim to determine whether these linguistic features predict the presence of misogynistic and aggressive comments, and to what extent, controlling for video duration and engagement metrics, and clustering by influencer to account for potential audience variation.