Network toxicity analysis: an information-theoretic approach to studying the social dynamics of online toxicity

Abstract

The rise of social media has corresponded with an increase in the prevalence and severity of online toxicity. While much work has gone into understanding its nature, we still lack knowledge of its emergent structural dynamics. This work presents a novel method—network toxicity analysis—for the inductive analysis of the dynamics of discursive toxicity within social media. Using an information-theoretic approach, this method estimates toxicity transfer relationships between communicating agents, yielding an effective network describing how those entities influence one another, over time, in terms of their produced discursive toxicity. This method is applied to Telegram messaging data to demonstrate its capacity to induce meaningful, interpretable toxicity networks that provide valuable insight into the social dynamics of toxicity within social media.