When the Echo Chamber Shatters: Examining the Use of Community-Specific Language Post-Subreddit Ban

TL;DR

This study addresses whether community-level subreddit bans on Reddit effectively reduce harmful language and activity, and how individual users respond. It introduces an unsupervised pipeline that identifies in-group vocabulary via Jensen-Shannon Divergence against a baseline of millions of Reddit comments, enabling scalable analysis across 15 banned subreddits. The results reveal heterogeneous ban effects: top users tend to reduce activity more than random users, with substantial variation across subreddits and content categories (dark jokes, anti-political, mainstream right, and extreme right). The method provides a scalable, language-based lens for moderation assessment and highlights that ban efficacy is not uniform across communities, underscoring the need for nuanced, context-aware moderation strategies.

Abstract

Community-level bans are a common tool against groups that enable online harassment and harmful speech. Unfortunately, the efficacy of community bans has only been partially studied and with mixed results. Here, we provide a flexible unsupervised methodology to identify in-group language and track user activity on Reddit both before and after the ban of a community (subreddit). We use a simple word frequency divergence to identify uncommon words overrepresented in a given community, not as a proxy for harmful speech but as a linguistic signature of the community. We apply our method to 15 banned subreddits, and find that community response is heterogeneous between subreddits and between users of a subreddit. Top users were more likely to become less active overall, while random users often reduced use of in-group language without decreasing activity. Finally, we find some evidence that the effectiveness of bans aligns with the content of a community. Users of dark humor communities were largely unaffected by bans while users of communities organized around white supremacy and fascism were the most affected. Altogether, our results show that bans do not affect all groups or users equally, and pave the way to understanding the effect of bans across communities.

Figures (5)

• Figure 1: Example plots comparing user behavior after a subreddit ban. Users from the top 100 and random samples are displayed in terms of their relative change in activity and change in in-group vocabulary usage. Distributions are displayed along each axis for convenience.
• Figure 2: Comparison of top and random user behavior changes across fifteen subreddits banned after a change in Reddit content policy in January, 2020. (a) Top users show more significant drop-offs in posting activity after a ban, but have around the same change in in-group vocabulary usage as a uniform sampling of subreddit participants. (b) Ban impact on eleven subreddits categorized by content. Each subreddit appears twice, representing top and random users. Four uncategorized subreddits are excluded from the plot. Trends are summarized in \ref{['table:ban_impact_by_category']}.
• Figure 3: Scatterplot showing differences in activity and vocabulary shifts between top and random users of each subreddit. Each axis shows the statistical significance, expressed as -log(FDR), of either activity (x-axis) or vocabulary (y-axis) shift. Dashed lines indicate significance at a threshold of 0.05, such that subreddits with greater values show significant differences between top and random users.
• Figure 4: Visualization of GLMM results showing differences between subreddits in postban behavior. For each row, blue cells indicate that the subreddit in a given column had a lower proportion of postban activity/ingroup vocabulary use than the subreddit in that row, while red cells indicate that the subreddit in a given column had a higher proportion of postban activity/ingroup vocabulary use than the subreddit in that row. · indicates p $<$ 0.10. * indicates p $<$ 0.05. ** indicates p $<$ 0.01. *** indicates p $<$ 0.001.
• Figure 5: Comparison of top and random user behavior changes under different keyword selection methodology. The subplot on the left corresponds to \ref{['fig:summary_top_bottom']} in the main text. The differences in activity shift between the two plots are minute and only due to omission of slightly different users for having no in-group vocabulary usage before or after the ban. The relative positions on the vocabulary shift axis remain largely the same except for a wider distribution and several subreddit user-type pairs exhibiting the maximum possible negative shift as the median.