The Big Ban Theory: A Pre- and Post-Intervention Dataset of Online Content Moderation Actions

TL;DR

The paper addresses the fragmentation and limited comparability of moderation research by introducing The Big Ban Theory (TBBT), an intervention-centered dataset that aggregates 25 moderation actions across Reddit (with Voat migrations) from 2015–2023, with three-month pre- and post-intervention windows. It defines a four-slice data model (IN-BEFORE, IN-AFTER, OUT-BEFORE, OUT-AFTER), standardizes metadata, and applies deterministic pseudonymization to create a FAIR-compliant, reusable resource. It provides both descriptive statistics and a range of use cases—ranging from effect evaluation and fairness analysis to predictive modeling and methodological benchmarking—supporting cross-intervention, cross-platform moderation research. By offering a publicly available, extensible foundation, TBBT enables systematic, reproducible analyses of moderation interventions while acknowledging current limitations such as platform bias and observational nature, paving the way for broader platform coverage and stronger causal inference in future work.

Abstract

Online platforms rely on moderation interventions to curb harmful behavior such hate speech, toxicity, and the spread of mis- and disinformation. Yet research on the effects and possible biases of such interventions faces multiple limitations. For example, existing works frequently focus on single or a few interventions, due to the absence of comprehensive datasets. As a result, researchers must typically collect the necessary data for each new study, which limits opportunities for systematic comparisons. To overcome these challenges, we introduce The Big Ban Theory (TBBT), a large dataset of moderation interventions. TBBT covers 25 interventions of varying type, severity, and scope, comprising in total over 339K users and nearly 39M posted messages. For each intervention, we provide standardized metadata and pseudonymized user activity collected three months before and after its enforcement, enabling consistent and comparable analyses of intervention effects. In addition, we provide a descriptive exploratory analysis of the dataset, along with several use cases of how it can support research on content moderation. With this dataset, we aim to support researchers studying the effects of moderation interventions and to promote more systematic, reproducible, and comparable research. TBBT is publicly available at: https://doi.org/10.5281/zenodo.18245670.

Figures (5)

• Figure 1: For an intervention occurring at time $t_0$, we identify four data slices: IN-BEFORE and IN-AFTER (if available) capture activity in the moderated space before and after the intervention, while OUT-BEFORE and OUT-AFTER capture activity by the same users outside that space.
• Figure 2: Overview of the six steps of the data collection and processing pipeline used to construct the dataset.
• Figure 3: Dataset structure, showing the top-level folders by intervention type, the intervention-level compressed archives, and the standardized JSON files corresponding to the available data slices for each intervention.
• Figure 4: Frequency of interventions in the TBBT dataset, by type (y axis) and year (x axis) (central heatmap), with marginal distributions (top and right bar charts).
• Figure 5: Number of comments (top panels) and active users (bottom panels) before and after each moderation intervention. For each intervention, bars compare pre- and post-intervention activity. The left panels show comparisons within the moderated space for interventions where IN-AFTER data are available. In case of bans with migrations (M), IN-BEFORE data is compared to OUT-AFTER data. The right panels show comparisons for community bans (B) using OUT-BEFORE and OUT-AFTER data. The y axis is on a logarithmic scale to accommodate the large variation in activity volumes across communities. Subreddits on the x axis are ordered from left to right by decreasing number of affected users.