Machine learning analysis has revealed patterns in hate speech online that point to complex – and sometimes counterintuitive – connections between real-world events and different types of hate speech. Yonatan Lupu of George Washington University in Washington, DC, and colleagues present these findings Jan. 25 in the open access journal PLOS ONE.
Previous research has provided important insights into hate speech publicly posted by users of online communities. Real world events can lead to an increase in hate speech online, and spikes in hate speech online have been associated with spikes in violent hate crime in the real world. However, most previous studies have focused on a limited number of moderated platform communities that have policies against hate speech.
Lupu and colleagues combined manual methods with a computational strategy known as supervised machine learning to analyze seven types of online hate speech in 59 million posts published between June 2019 and December 2020 by users of 1,150 online hate communities. Some communities were on the moderated platforms Facebook, Instagram or VKontakte, and others on the less moderated platforms Gab, Telegram and 4Chan.
This analysis revealed spikes in online hate speech that appeared to be related to particular events in the real world. For example, after a crisis involving Syrian refugees, hate speech against immigration surged.
Following the November 2020 US election, more sustained waves of increased online hate speech occurred. For example, there was an increase in the use of anti-LGBTQ slurs to describe various political causes, and Vice President Kamala Harris was a prominent target of increased gender-based hate speech.
Within the study period, the murder of George Floyd and subsequent protests were associated with the largest spike in hate speech, including race-based hate speech. However, other forms of hate speech also increased significantly, including hate speech targeting gender identity and sexual orientation – a topic that has little intuitive connection to the murder and the protests.
While the study cannot draw causal conclusions, the findings suggest a complex relationship between triggering events and hate speech online, with potential implications for strategies to reduce such speech. The authors call for additional research to further explore this relationship, especially given users’ tendency to migrate to unmoderated communities.
The authors add, “Hate speech remains a persistent and pervasive problem in the social media landscape and can increase in dramatic and sometimes unexpected ways following offline events.”
- Lupu Y, Sear R, Velásquez N, Leahy R, Restrepo NJ, Goldberg B, et al. (2023) Offline events and online hate. PLoS ONE 18(1): e0278511. DOI: 10.1371/journal.pone.0278511