On Simulating the Propagation and Countermeasures of Hate Speech in Social Networks

Abstract

Hate speech expresses prejudice and discrimination based on actual or perceived innate characteristics such as gender, race, religion, ethnicity, colour, national origin, disability or sexual orientation. Research has proven that the amount of hateful messages increases inevitably on online social media. Although hate propagators constitute a tiny minority with less than 1% participants they create an unproportionally high amount of hate motivated content. Thus, if not countered properly, hate speech can propagate through the whole society. In this paper we apply agent-based modelling to reproduce how the hate speech phenomenon spreads within social networks. We reuse insights from the research literature to construct and validate a baseline model for the propagation of hate speech. From this, three countermeasures are modelled and simulated to investigate their effectiveness in containing the spread of hatred: Education, deferring hateful content, and cyber activism. Our simulations suggest that: (1) Education consititutes a very successful countermeasure, but it is long term and still cannot eliminate hatred completely; (2) Deferring hateful content has a similar although lower positive effect than education, and it has the advantage of being a short-term countermeasure; (3) In our simulations, extreme cyber activism against hatred shows the poorest performance as a countermeasure, since it seems to increase the likelihood of resulting in highly polarised societies.