Gender Bias in Word Embeddings: A Comprehensive Analysis of Frequency, Syntax, and Semantics
Open accessThe DOI matches its Crossref record; 11 of the 11 references it cites were found in a registry or an index.
Abstract
Word embeddings are numeric representations of meaning derived from word co-occurrence statistics in corpora of human-produced texts. The statistical regularities in language corpora encode well-known social biases into word embeddings (e.g., the word vector for family is closer to the vector women than to men). Although efforts have been made to mitigate bias in word embeddings, with the hope of improving fairness in downstream Natural Language Processing (NLP) applications, these efforts will remain limited until we more deeply understand the multiple (and often subtle) ways that social biases can be reflected in word embeddings. Here, we focus on gender to provide a comprehensive analysis of group-based biases in widely-used static English word embeddings trained on internet corpora (GloVe 2014, fastText 2017). While some previous research has helped uncover biases in specific semantic associations between a group and a target domain (e.g., women - family), using the Single-Category Word Embedding Association Test, we demonstrate the widespread prevalence of gender biases that also show differences in: (1) frequencies of words associated with men versus women; (b) part-of-speech tags in gender-associated words; (c) semantic categories in gender-associated words; and (d) valence, arousal, and dominance in gender-associated words. We leave the analysis of non-binary gender to future work due to the challenges in accurate group representation caused by limitations inherent in data.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Enriching Word Vectors with Subword Information
Bojanowski, P., Grave, E., Joulin, A., & Mikolov, T. (2017) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00051 · Cited by 9,645 · cited 1×
“The current manuscript provides the first comprehensive investigation of the many, and often subtle ways that social biases are reflected in widely-used static English word embeddings trained on internet corpora (specifically GloVe 2014 [41] and fastText 2017 [8]).”
BackgroundStatistical Power Analysis for the Behavioral Sciences
Wassertheil, S., & Cohen, J. (1970) · Biometrics · doi:10.2307/2529115 · Cited by 1,383
The Single Category Implicit Association Test as a measure of implicit social cognition.
Karpinski, A., & Steinman, R. B. (2006) · Journal of Personality and Social Psychology · doi:10.1037/0022-3514.91.1.16 · Cited by 1,108 · cited 2×
“The method builds from the same logic as the Single-Category Implicit Association Test (SC-IAT) [32] used with human participants, and the relative associations can be interpreted roughly in comparison to statistics such as Cohen’s 𝑑 effect sizes.”
MethodWord embeddings quantify 100 years of gender and ethnic stereotypes
Garg, N., Schiebinger, L., Jurafsky, D., & Zou, J. (2018) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1720347115 · Cited by 1,065 · cited 1×
“[26] used word embeddings to study the evolution of gender stereotypes toward ethnic minorities in the United States during the 20 and 21 centuries.”
MethodBender, E. M., & Friedman, B. (2018) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00041 · Cited by 854 · cited 1×
“Furthermore, complementing computational methods by adding comprehensive data statements to language technologies as outlined by Bender and Friedman [5] will enhance transparency and raise awareness, which is the first step towards mitigating bias.”
MethodGonen, H., & Goldberg, Y. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1061 · Cited by 454 · cited 1×
“Gonen and Goldberg [27] showed that common debiasing methods for word embeddings fail to remove systematic gender biases, and that biases can be recovered from the debiased space.”
BackgroundAI-enabled recruiting: What is it and how should a manager use it?
Black, J. S., & van Esch, P. (2020) · Business Horizons · doi:10.1016/j.bushor.2019.12.001 · Cited by 403 · cited 2×
“Today, the vast majority of our daily tasks are facilitated and enhanced through the application of Natural Language Processing (NLP), from simplemachine translation to automated resume screening to auto-complete in emails [7].”
BackgroundBias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting
De-Arteaga, M., Romanov, A., Wallach, H., Chayes, J., Borgs, C., Chouldechova, A., Geyik, S., Kenthapadi, K., & Kalai, A. T. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287572 · Cited by 337
Bias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting
De-Arteaga, M., Romanov, A., Wallach, H., Chayes, J., Borgs, C., Chouldechova, A., Geyik, S., Kenthapadi, K., & Kalai, A. T. (2019) · arXiv · doi:10.48550/arxiv.1901.09451 · Cited by 337
Guo, W., & Caliskan, A. (2021) · Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3461702.3462536 · Cited by 180 · cited 14×
“Toney-Wails and Caliskan [46] employ the SC-WEAT to show that valence (pleasantness) norms are consistent across word embedding algorithms, time periods, and languages, while that social biases vary.”
BackgroundInfluentialSelf-citation
Cited by
Wolfe, R., Yang, Y., Howe, B., & Caliskan, A. (2023) · 2023 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3593013.3594072 · Cited by 42
Self-citationOn the independence of association bias and empirical fairness in language models
Cabello, L., Jørgensen, A. K., & Søgaard, A. (2023) · 2023 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3593013.3594004 · Cited by 14
Who's in and who's out? A case study of multimodal CLIP-filtering in DataComp
Hong, R., Agnew, W., Kohno, T., & Morgenstern, J. (2024) · Proceedings of the 4th ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization · doi:10.1145/3689904.3694702 · Cited by 11
“This difference is notable given that CLIP is not trained explicitly as a toxicity classifier, yet as a filtering method still obtains the same trends.”
Metadata
- Authors
- Caliskan, A., Ajay, P. P., Charlesworth, T., Wolfe, R., & Banaji, M. R.
- Venue
- Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society
- Publisher
- ACM
- Published
- 2022-07-26
- Pages
- 156-170
- ISSN
- 3065-8365 (ISSN-L)
- OpenAlex
- W4282026609
- Cited by
- 79
- References
- 11
- FWCI
- 15.92
- Type
- conference paper
- Language
- en
- Open access
- hybrid
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.