# Asking For It: Question-Answering for Predicting Rule Infractions in Online Content Moderation

arXiv: https://arxiv.org/abs/2510.06350 · HTML: https://mlsanity.com/paper/2510.06350/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 53 references that are in the map.
- First posted on arXiv 7 Oct 2025.

## Builds on (most cited first, 11 of 11)

- ["Call me sexist, but...": Revisiting Sexism Detection Using Psychological Scales and Adversarial Samples](https://mlsanity.com/paper/2004.12764/) (2020)
- [SLM-Mod: Small Language Models Surpass LLMs at Content Moderation](https://mlsanity.com/paper/2410.13155/) (2024)
- [Detecting Community Sensitive Norm Violations in Online Conversations](https://mlsanity.com/paper/2110.04419/) (2021)
- [Let Community Rules Be Reflected in Online Content Moderation](https://mlsanity.com/paper/2408.12035/) (2024)
- [End User Authoring of Personalized Content Classifiers: Comparing Example Labeling, Rule Writing, and LLM Prompting](https://mlsanity.com/paper/2409.03247/) (2024)
- [PolicyCraft: Supporting Collaborative and Participatory Policy Design through Case-Grounded Deliberation](https://mlsanity.com/paper/2409.15644/) (2024)
- [CPL-NoViD: Context-Aware Prompt-based Learning for Norm Violation Detection in Online Communities](https://mlsanity.com/paper/2305.09846/) (2023)
- [Venire: A Machine Learning-Guided Panel Review System for Community Content Moderation](https://mlsanity.com/paper/2410.23448/) (2024)
- [The Unseen Targets of Hate -- A Systematic Review of Hateful Communication Datasets](https://mlsanity.com/paper/2405.08562/) (2024)
- [Governing online goods: Maturity and formalization in Minecraft, Reddit, and World of Warcraft communities](https://mlsanity.com/paper/2202.01317/) (2022)
- [The Future of Research on Social Technologies: CCC Workshop Visioning Report](https://mlsanity.com/paper/2404.10897/) (2024)

## Abstract (arXiv)

Online communities rely on a mix of platform policies and community-authored rules to define acceptable behavior and maintain order. However, these rules vary widely across communities, evolve over time, and are enforced inconsistently, posing challenges for transparency, governance, and automation. In this paper, we model the relationship between rules and their enforcement at scale, introducing ModQ, a novel question-answering framework for rule-sensitive content moderation. Unlike prior classification or generation-based approaches, ModQ conditions on the full set of community rules at inference time and identifies which rule best applies to a given comment. We implement two model variants - extractive and multiple-choice QA - and train them on large-scale datasets from Reddit and Lemmy, the latter of which we construct from publicly available moderation logs and rule descriptions. Both models outperform state-of-the-art baselines in identifying moderation-relevant rule violations, while remaining lightweight and interpretable. Notably, ModQ models generalize effectively to unseen communities and rules, supporting low-resource moderation settings and dynamic governance environments.