Applying Intrinsic Debiasing on Downstream Tasks: Challenges and Considerations for Machine Translation
Not on the map. Why?
Abstract (arXiv)
Most works on gender bias focus on intrinsic bias -- removing traces of information about a protected group from the model's internal representation. However, these works are often disconnected from the impact of such debiasing on downstream applications, which is the main motivation for debiasing in the first place. In this work, we systematically test how methods for intrinsic debiasing affect neural machine translation models, by measuring the extrinsic bias of such systems under different design choices. We highlight three challenges and mismatches between the debiasing techniques and their end-goal usage, including the choice of embeddings to debias, the mismatch between words and sub-word tokens debiasing, and the effect on different target languages. We find that these considerations have a significant impact on downstream performance and the success of debiasing.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 20 of its 29 references that are in the map. First posted on arXiv 2 Jun 2024.
Builds on
20 of 29 references.
References · 20
- 2018SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processingcited by 1,973
- 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
- 2020Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillationcited by 591
- 2020Null It Out: Guarding Protected Attributes by Iterative Nullspace Projectioncited by 442
- 2021Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPcited by 336
- 2019Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Themcited by 268
- 2023LEACE: Perfect linear concept erasure in closed formcited by 254
- 2019Evaluating Gender Bias in Machine Translationcited by 247
- 2018Adversarial Removal of Demographic Attributes from Text Datacited by 166
- 2020Intrinsic Bias Metrics Do Not Correlate with Application Biascited by 148
- 2021Debiasing Pre-trained Contextualised Embeddingscited by 107
- 2022On the Intrinsic and Extrinsic Fairness Evaluation Metrics for Contextualized Language Representationscited by 75
8 more
- 2019Equalizing Gender Biases in Neural Machine Translation with Word Embeddings Techniquescited by 65
- 2019Attenuating Bias in Word Vectorscited by 64
- 2022Welcome to the Modern World of Pronouns: Identity-Inclusive Natural Language Processing beyond Gendercited by 47
- 2020OSCaR: Orthogonal Subspace Correction and Rectification of Biases in Word Embeddingscited by 40
- 2021Contrastive Learning for Fair Representationscited by 34
- 2022Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Informationcited by 33
- 2023Exploring the Impact of Training Data Distribution and Subword Tokenization on Gender Bias in Machine Translationcited by 7
- 2023Evaluating and Improving the Coreference Capabilities of Machine Translation Modelscited by 2
Built on by
Citing papers per year: 2025: 2.
2 papers.