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

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.

💡 Research Summary

The paper addresses a critical gap in the study of online content moderation: the lack of comprehensive, standardized datasets that enable systematic, comparable, and reproducible analyses of moderation interventions. Existing research typically focuses on a single or a handful of interventions, requiring researchers to collect bespoke data for each study, which hampers cross‑study comparisons and limits the ability to draw generalizable conclusions.

To overcome these limitations, the authors introduce The Big Ban Theory (TBBT), a large‑scale, intervention‑centered dataset covering 25 distinct moderation actions that occurred on Reddit between 2015 and 2023, with some cases also tracking user migration to the alternative platform V‑oat. The dataset comprises activity from 339,125 unique users and 3,870,0732 messages (posts and comments). For every intervention, the authors collect three months of activity before and after the enforcement timestamp, and they organize the data along two orthogonal dimensions: time (pre vs. post) and space (inside vs. outside the moderated community). This yields four slices—IN‑BEFORE, IN‑AFTER, OUT‑BEFORE, OUT‑AFTER—which together capture the full context of the moderation event, even when the moderated space ceases to exist after a ban (IN‑AFTER unavailable).

The data acquisition pipeline consists of six reproducible steps:

1. IN data collection from public Reddit torrents and a Zenodo repository for V‑oat, using a ±3‑month window around the intervention timestamp.
2. Bot filtering by removing accounts that post two or more comments with identical timestamps, supplemented by manual validation and exclusion of self‑declared bots.
3. User selection that retains only those who posted at least ten messages in the target community during the pre‑intervention window, thereby focusing on genuinely affected users.
4. OUT data collection to capture activity of the same users in other subreddits or on V‑oat when IN‑AFTER data is structurally missing.
5. Field selection and standardization to harmonize heterogeneous schemas across years and platforms, adopting Reddit’s naming conventions for consistency.
6. Pseudonymization of all identifiers via a deterministic cryptographic hash, preserving internal linkage while protecting privacy.

The final data structure mirrors the four‑slice model: each slice is a top‑level folder containing CSV files for each intervention and a JSON metadata file. Metadata records the intervention ID, type (ban, quarantine, removal, etc.), violated policy, affected community, timestamp, and counts of affected users. The dataset adheres to the FAIR principles and is publicly released on Zenodo (DOI 10.5281/zenodo.18245670) under a CC‑BY‑4.0 license, enabling free reuse.

The authors illustrate several research use cases:

• Effectiveness assessment using descriptive statistics, ANOVA, or more sophisticated causal inference methods such as Difference‑in‑Differences and interrupted time‑series models.
• Bias and fairness analysis by linking intervention types to user attributes (e.g., political orientation, activity level) to detect over‑ or under‑enforcement.
• Predictive modeling where pre‑intervention signals (toxicity scores, linguistic features, network centrality) serve as inputs to forecast outcomes like user abandonment, migration, or changes in toxicity.
• Spill‑over and migration studies leveraging OUT‑BEFORE/AFTER slices to quantify how moderation actions affect activity on other communities or platforms.

Key insights include the value of a four‑slice design for constructing natural control groups, the importance of rigorous bot and low‑activity filtering for data quality, and the novel inclusion of cross‑platform migration data that enables analysis of de‑platforming effects. By providing a unified, well‑documented, and privacy‑preserving resource, TBBT paves the way for more transparent, fair, and evidence‑based moderation policies and for a new generation of reproducible moderation research.