Characterizing Stereotypical Bias from Privacy-preserving Pre-Training

Arnold S, Gröbner R, Schreiner A (2024)

Publication Type: Conference contribution

Publication year: 2024

Publisher: Association for Computational Linguistics (ACL)

Pages Range: 20-28

Conference Proceedings Title: Proceedings of the Fifth Workshop on Privacy in Natural Language Processing

Event location: Bangkok TH

ISBN: 9798891761391

URI: https://aclanthology.org/2024.privatenlp-1.3

Abstract

Differential Privacy (DP) can be applied to raw text by exploiting the spatial arrangement of words in an embedding space. We investigate the implications of such text privatization on Language Models (LMs) and their tendency towards stereotypical associations. Since previous studies documented that linguistic proficiency correlates with stereotypical bias, one could assume that techniques for text privatization, which are known to degrade language modeling capabilities, would cancel out undesirable biases. By testing BERT models trained on texts containing biased statements primed with varying degrees of privacy, our study reveals that while stereotypical bias generally diminishes when privacy is tightened, text privatization does not uniformly equate to diminishing bias across all social domains. This highlights the need for careful diagnosis of bias in LMs that undergo text privatization.

Authors with CRIS profile

Stefan Arnold Lehrstuhl für Wirtschaftsinformatik, insbesondere IT-Management René Gröbner Lehrstuhl für Wirtschaftsinformatik, insbesondere IT-Management Annika Schreiner Lehrstuhl für Wirtschaftsinformatik, insbesondere IT-Management

How to cite

APA:

Arnold, S., Gröbner, R., & Schreiner, A. (2024). Characterizing Stereotypical Bias from Privacy-preserving Pre-Training. In Ivan Habernal, Sepideh Ghanavati, Abhilasha Ravichander, Vijayanta Jain, Patricia Thaine, Timour Igamberdiev, Niloofar Mireshghallah, Oluwaseyi Feyisetan (Eds.), Proceedings of the Fifth Workshop on Privacy in Natural Language Processing (pp. 20-28). Bangkok, TH: Association for Computational Linguistics (ACL).

MLA:

Arnold, Stefan, René Gröbner, and Annika Schreiner. "Characterizing Stereotypical Bias from Privacy-preserving Pre-Training." Proceedings of the 5th Workshop on Privacy in Natural Language Processing, PrivateNLP 2024 - Co-located with ACL 2024, Bangkok Ed. Ivan Habernal, Sepideh Ghanavati, Abhilasha Ravichander, Vijayanta Jain, Patricia Thaine, Timour Igamberdiev, Niloofar Mireshghallah, Oluwaseyi Feyisetan, Association for Computational Linguistics (ACL), 2024. 20-28.

BibTeX: Download