Social-Group-Agnostic Bias Mitigation via the Stereotype Content Model
Published in ACL, 2023
Existing methods typically rely on word pairs specific to certain social groups, limiting their effectiveness to one aspect of social identity. This approach becomes impractical and costly when addressing bias in lesser-known or unmarked social groups. Instead, we proposed leveraging the Stereotype Content Model (SCM), a framework from social psychology. The SCM categorizes stereotypes along two dimensions: warmth and competence. By adopting this social-group-agnostic perspective, we demonstrated comparable performance to group-specific debiasing methods while offering theoretical and practical advantages over existing techniques.
Recommended citation: Ali Omrani, Alireza Salkhordeh Ziabari, Charles Yu, Preni Golazizian, Brendan Kennedy, Mohammad Atari, Heng Ji, and Morteza Dehghani. 2023. Social-Group-Agnostic Bias Mitigation via the Stereotype Content Model. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4123–4139, Toronto, Canada. Association for Computational Linguistics.