Community Norms in the Spotlight: Enabling Task-Agnostic Unsupervised Pre-Training to Benefit Online Social Media

Liam Hebert, Lucas Kopp, Robin Cohen · 2026

Read Community Norms in the Spotlight: Enabling Task-Agnostic Unsupervised Pre-Training to Benefit Online Social Media by Liam Hebert, Lucas Kopp, Robin Cohen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-02525.txt.

Description

Liam Hebert and 2 co-authors make the case for pretraining social media models on community norms without task labels, in a 2026 arXiv paper. Discussion transformers model conversations as graph structures and show promise for fighting hate speech and misinformation. Their progress is throttled by dependence on high-quality human-labeled datasets, which are expensive and narrow.

The authors advocate a shift from task-specific fine-tuning to unsupervised pretraining grounded in community norms. Such pretraining could ease data scarcity across many downstream tasks at once. It could also make model decisions interpretable, by exposing the social norms the system used. The paper is a position piece rather than a system report. It maps opportunities for AI for social good that follow once norms become the organizing signal for social media modeling.

Catalog record

CREATORLiam Hebert, Lucas Kopp, Robin Cohen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-02525

Browse the full library · About the Tower of Babel