Understanding intergroup vs. intragroup toxicity in online communities




Tellakat, Mohini

Journal Title

Journal ISSN

Volume Title



The digital revolution has changed the world. People around the globe can stay connected with strangers, friends, and family, can learn about in an infinite number of topics, and can be a part of life changing movements and discussions. Anonymity on the internet, however, can lead to negative behaviors such as cyberbullying, trolling, and the formation of groups of highly discriminatory individuals. Internet-based interactions can lead to toxic (e.g. aggressive, hateful) online environments which can affect people’s sense of well-being and mental health in general. Toxic behavior comes in many forms, however, so the goal of this dissertation is to understand the psychological differences between two types of toxicity: Intergroup toxicity and intragroup toxicity. Intergroup toxicity is defined as expressing hatred, disgust, and general ill-will to people who are not within a community’s ingroup. Intragroup toxicity is defined as hatred, disgust, aggression, and general ill-will directed to people within a community’s ingroup. Using language analytic techniques on large datasets from Reddit, the current project builds on research done on toxicity in online communities to understand the differences between intergroup and intragroup toxicity. Specifically, the research analyzes psychological differences in terms of linguistic markers of mental health and style (Tausczik & Pennebaker 2010, Martens, Shen, Iosup, & Kuipers 2015). After analyzing the differences in the two kinds of toxicity, the research addresses how participating in a toxic group relates to the way people behave in other online groups they participate in. For the analyses, 10 subreddits with intergroup toxicity and 10 subreddits with intragroup toxicity were selected based on a network analysis (Datta & Adar 2019) and using Reddit’s controversiality measure. The language of the two groups was studied using random sampling and bootstrapping methodologies to compare the means between groups with intragroup toxicity and groups with intragroup toxicity. The full reddit histories of the people used in Study 1 were used to answer the research questions posed in Study 2. Authors were characterized into 4 major subgroups based on individual and group levels of toxicity: Low individual-level toxicity in both groups with intergroup and intragroup toxicity and high-individual level toxicity in both groups with intergroup and intragroup toxicity. Overall, the research found that there is a personality or trait-based form of toxicity among those who used toxic words at high rates in their home groups. If people used toxic words at high rates in the initial highly toxic 20 subreddits that were collected, they used toxic language in many of the other subreddits that they participated in. Additionally, if people had low individual-level toxicity, their toxic behaviors in groups other than the initial 20 subreddits tended to be related to which type of initially toxic group they posted in the majority of the time. Specifically, people with high individual-level toxicity behaved as though they had toxic personalities. The same patterns were found among people with low individual-level toxicity who participated in groups with intergroup toxicity. Finally, people with low individual-level toxicity who participated in groups with intragroup toxicity did not use toxic language as much, indicating that people who had low individual-level toxicity could be influenced by the groups that they chose to participate in. The differences in toxic language use showed that there cannot be a blanket statement or a blanket treatment of toxicity online. Though people with high-individual level toxicity seem to have trait-level toxicity, people who have low-individual level toxicity seem to be influenced by their communities. Being able to stage appropriate interventions for people who do have low-level toxicity becomes important when considering groups with intergroup toxicity promote more hatred than groups with intragroup toxicity do. Recognizing these group members and engaging with them before they spread toxic behaviors in the other groups that they participate in can be key to reducing proliferation of toxic behaviors online. With so many people engaging in important issues in online communities, the use of effective interventions for hate and for discontent become more and more needed. By helping to shed light onto a few aspects of online toxicity, the research from this dissertation can inform the understanding of the behavior of online communities.



LCSH Subject Headings