Toxic speech is an endemic threat to civil discourse on social media; it’s marked by incivility, intolerance, and the intent to harm through threats, insults, and patronizing language. Further challenging is the fact that platforms’ content moderation policies often overlook this behaviour, as toxic speech drives user engagement and aids platforms’ bottom line: money. Toxic speech is usually conceptualized as an anti-social action and is associated with several social problems, including misinformation, bullying, and violence. However, this framing alone is reductive. It can also be an act of sociality in communities, marking to the world who belongs and who doesn’t. This means that we need to refine where we situate toxic speech on the spectrum of anti- to pro-social behaviour to understand both its harms and benefits in online settings.
In a recently published study in New Media & Society, “Online toxic speech as positioning acts: Hate as discursive mechanisms for othering and belonging,” researchers Esteban Morales (University of Groningen), Jaigris Hodson (Royal Roads University), Victoria O’Meara (University of Leicester), Anatoliy Gruzd (Toronto Metropolitan University), and Philip Mai (Toronto Metropolitan University) took a further look at how toxic speech is used by social media users to construct and maintain in- and out-group boundaries in online communities.
Methodology
The research team used Communalytic, a computational social science research tool, to collect 98,729 publicly available posts from the popular Colombian Telegram group Chismes Frescos Medellin (Fresh Gossip Medellin). Telegram is particularly useful for examining toxic speech for two reasons. First, the platform is continually gaining popularity, serving as a key space for socializing and sharing news. Second, Telegram is notorious for its stance on content moderation, rejecting censorship in favour of free speech. This conscious decision has made it a hotbed for violent content and communities.
Toxic posts were identified using the machine-learning classifier Detoxify, built into Communalytic’s Civility Analyzer module. The Civility Analyzer assigned a score from 0 (least toxic) to 1 (most toxic); only those posts assigned a score greater than 0.7 were retained (3,221 posts). A manual review of posts identified 66 cases of misclassification, leading to a final sample of 3,155 posts. Posts were then topically categorized, with those posts in the four most prevalent topics (1,258 posts, nearly 40% of the final sample) serving as the object for the study’s analysis. These topics included: security (652 posts, 20%); politics (252, 7.8%); migration (188 posts, 5.8%); and gender (166 posts, 5.2%).
To analyze the sample, our group used a thematic analysis to qualitatively tease out who was being positioned in each topic’s toxic posts — the self, the other, and the group — and how these positionalities are negotiated across the observed conversations.
Results
To answer our research question (“How do Telegram group members use toxic speech to position themselves and others in relation to narratives emerging from the group?“), we identified four key themes from an analysis of 1,258 posts:
Key Takeaways
Results indicate that toxic speech serves a paradoxical function: it both reinforces the boundaries of the group’s identity while also resisting these very boundaries. However, we find that the topical context in which toxic speech occurs informs whether it reinforces or resists the group’s identity.
- For instance, the topics of “(in)security” and “migration” in the Chismes Frescos Medellin group presented cohesion in members’ perspectives. For these topics, toxic speech worked to neutralize dissent and reinforce the group’s moral boundaries on issues of crime and migration.
- In contrast, “political (dis)affiliation” and “gender and sexuality” contained greater fragmentation in members’ viewpoints. In these contexts, members’ toxic speech worked to contest the group’s moral boundaries by asserting competing visions of the group’s “correct” values.
These findings signal the importance of contextualizing antisocial behaviours in the everyday digital cultures through which they emerge.
Conclusions
Overall, our study demonstrates that toxic speech is more than just a device for harming our opponents; it’s also a vital tool for sociability.
Toxic speech allows us to negotiate membership within online communities by sanctioning those who violate community norms and attack out-group members who threaten the in-group. Simply put, toxic speech can signal who belongs and who doesn’t.
Understanding toxic speech in online communities, therefore, warrants nuance in disambiguating the contexts in which it can (dis)enfranchise individuals. For example, in content moderation, the aim should not be to scrub every instance of harsh language. It’s to set fair, ethical boundaries around what’s acceptable while acknowledging that online conversations can be messy and emotional.
To learn more about this study, check out the full paper here.
Citation: Morales, E., Hodson, J., O’Meara, V., Gruzd, A., and Mai, P. (2025). Online toxic speech as positioning acts: Hate as discursive mechanisms for othering and belonging. New Media & Society. DOI: https://doi.org./10.1177/14614448251338493.