The Relevance of Value Systems for Offensive Language Detection

Michael Wiegand, Elisabeth Eder, Josef Ruppenhofer


Abstract
We examine in how far a person’s value system has an impact on their perception of offensiveness. For instance, a scholar is likely to be offended by being accused of reporting unverified claims whereas many non-scholars would not feel that way. Thus, we move away from the assumption that offensiveness can be defined through a universal perspective. Ultimately, such research aims to support personalized approaches to content moderation. Our main contribution is the introduction of a dataset consisting of neutrally-phrased sentences on controversial topics, evaluated by individuals from 4 different value systems. This allows us to identify offensiveness patterns across value systems and conduct classification experiments.
Anthology ID:
2026.eacl-long.319
Volume:
Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
March
Year:
2026
Address:
Rabat, Morocco
Editors:
Vera Demberg, Kentaro Inui, Lluís Marquez
Venue:
EACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
6765–6789
Language:
URL:
https://preview.aclanthology.org/ingest-eacl/2026.eacl-long.319/
DOI:
Bibkey:
Cite (ACL):
Michael Wiegand, Elisabeth Eder, and Josef Ruppenhofer. 2026. The Relevance of Value Systems for Offensive Language Detection. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6765–6789, Rabat, Morocco. Association for Computational Linguistics.
Cite (Informal):
The Relevance of Value Systems for Offensive Language Detection (Wiegand et al., EACL 2026)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingest-eacl/2026.eacl-long.319.pdf