Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model

Abhijith Chintam, Rahel Beloch, Willem Zuidema, Michael Hanna, Oskar van der Wal


Abstract
Language models (LMs) exhibit and amplify many types of undesirable biases learned from the training data, including gender bias. However, we lack tools for effectively and efficiently changing this behavior without hurting general language modeling performance. In this paper, we study three methods for identifying causal relations between LM components and particular output: causal mediation analysis, automated circuit discovery and our novel, efficient method called DiffMask+ based on differential masking. We apply the methods to GPT-2 small and the problem of gender bias, and use the discovered sets of components to perform parameter-efficient fine-tuning for bias mitigation. Our results show significant overlap in the identified components (despite huge differences in the computational requirements of the methods) as well as success in mitigating gender bias, with less damage to general language modeling compared to full model fine-tuning. However, our work also underscores the difficulty of defining and measuring bias, and the sensitivity of causal discovery procedures to dataset choice. We hope our work can contribute to more attention for dataset development, and lead to more effective mitigation strategies for other types of bias.
Anthology ID:
2023.blackboxnlp-1.29
Volume:
Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP
Month:
December
Year:
2023
Address:
Singapore
Editors:
Yonatan Belinkov, Sophie Hao, Jaap Jumelet, Najoung Kim, Arya McCarthy, Hosein Mohebbi
Venues:
BlackboxNLP | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
379–394
Language:
URL:
https://aclanthology.org/2023.blackboxnlp-1.29
DOI:
10.18653/v1/2023.blackboxnlp-1.29
Bibkey:
Cite (ACL):
Abhijith Chintam, Rahel Beloch, Willem Zuidema, Michael Hanna, and Oskar van der Wal. 2023. Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model. In Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP, pages 379–394, Singapore. Association for Computational Linguistics.
Cite (Informal):
Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model (Chintam et al., BlackboxNLP-WS 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.blackboxnlp-1.29.pdf
Video:
 https://aclanthology.org/2023.blackboxnlp-1.29.mp4