Elamur Research⌕
☰
home / papers / 2024 / Disagreeable, Slovenly, Honest and Un-named Women? Investigating Gender Bias in English Educational Resources by Extending Existing Gender Bias Taxonomies

Disagreeable, Slovenly, Honest and Un-named Women? Investigating Gender Bias in English Educational Resources by Extending Existing Gender Bias Taxonomies

Paper · 2024 · Haotian Zhu, Kexin Gao, Fei Xia et al.

Summary

The paper asks how gender bias in English educational materials can be measured automatically and at scale, rather than through the manual coding that education researchers usually rely on. The authors start from existing text-bias taxonomies (Hitti et al., 2019; Doughman et al.) and add new categories: distributional bias, namedness, and definitional bias. They then build an NLP pipeline that extracts person mentions and assigns them gender, partly by using coreference to link common nouns to gendered pronouns. The pipeline also collects the linguistic contexts of each mention.

Bias is quantified mainly with odds ratios. The method is applied to two kinds of data: K-12 content and exam corpora, and the lexical resource WordNet. Every bias type examined shows up in at least one source:

  • male characters are far more frequent than female ones and are much more often named;
  • WordNet contains sex-marked entries and generic "he" in its definitions;
  • female-specific WordNet entries carry more negative sentiment;
  • some "-woman" definitions simply point back to their "-man" counterparts.

The findings give textbook authors, test designers and lexicon maintainers concrete targets to revise.

Key points

  • The extended taxonomy has six types. B1 and B2 are structural (explicit sex marking, generic he), B3 is contextual (societal or behavioral), and B4–B6 are new (distributional, namedness, definitional). The conclusion counts 7 bias types examined in total.
  • The corpora fall into two groups:
    • Content: textbook (32,626 documents), CCS_doc, wee_bit, OneStop.
    • Exam: pisa, naep_science, naep_math.
  • The lexicon is WordNet 3.0.
  • In the textbook data, 32,884 male mentions were extracted versus 14,308 female mentions. About 73% of male mentions are named, compared with roughly 32% of female mentions.
  • In WordNet, vaderSentiment scores averaged −0.141 for female-specific entries and 0.056 for male ones. The difference is significant (t = −2.15, p = 0.035).
  • Of the 62 WordNet entries that have both "-man" and "-woman" forms, 10 have a detailed "-man" definition, while the "-woman" gloss merely refers back to it (e.g., horsewoman → horseman).

Interesting ideas

  • I1. Turn a qualitative bias taxonomy into categories that can be counted mechanically (mention frequency, named versus unnamed, derived definitions), so manual-coding concepts become automatic statistics.
  • I2. Assign gender to role nouns such as "the teacher" by following coreference chains to the pronouns that later refer to them, instead of relying on a fixed lexicon.
  • I3. Compute odds ratios of gendered mentions over many linguistic contexts rather than a predefined stereotype list, so that unexpected associations can surface.
  • I4. Audit paired lexicon entries for asymmetric definitions, where one member of a pair is defined only by pointing at the other.
  • I5. Compare sentiment across sets of lexicon entries grouped by a property, as a way to detect skewed connotations in a reference resource.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): The text-world builder could tag each person entity as named or unnamed and keep its attributes, giving cheap, explainable statistics on who in a story gets names, roles and actions.
  • I2 (frames): The coreference step could pass pronoun-derived features back to antecedent common nouns, enriching entity records with evidence that can be traced.
  • I3 (tree): Odds ratios over UD relations (e.g., which verbs take female versus male nsubj) could audit both the treebanks and parser outputs for skewed role patterns.
  • I4 (world): When compiling the ontology from WordNet, a check could flag glosses that depend on a sibling entry, so the knowledge graph receives full definitions instead of circular references.
  • I5 (learn): A similar grouped-sentiment comparison could act as an evaluation control on compiled knowledge or LLM-teacher outputs, revealing systematic connotation drift before the data is used for training.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of WashingtonSeattle · United States353 works · 200,043 citations

 Fields, topics and concepts

📚 Shelves 3The field of science this work belongs to (OpenAlex fields).Social Sciences · 1,772 worksSafety Research · 543 worksGender Studies · 52 works

🧭 Topics 2Research subjects and languages the work is about.English · 2,263 worksBias, fairness & safety · 517 works

Cite

Cite as Haotian Zhu, Kexin Gao, Fei Xia and Mari Ostendorf. 2024. Disagreeable, Slovenly, Honest and Un-named Women? Investigating Gender Bias in English Educational Resources by Extending Existing Gender Bias Taxonomies. Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing (GeBNLP).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2024.gebnlp-1.14

@inproceedings{zhu2024disagreeable,
  title = {{Disagreeable, Slovenly, Honest and Un-named Women? Investigating Gender Bias in English Educational Resources by Extending Existing Gender Bias Taxonomies}},
  author = {Haotian Zhu, Kexin Gao, Fei Xia and Mari Ostendorf},
  booktitle = {Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing (GeBNLP)},
  year = {2024},
  doi = {10.18653/v1/2024.gebnlp-1.14},
  url = {https://doi.org/10.18653/v1/2024.gebnlp-1.14}
}

References

38 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Steven Bird, Ewan Klein, and Edward Loper. 2009. Natural language processing with Python: analyzing text with the natural language toolkit. " O’Reilly Media, Inc.". find
  2. Su Lin Blodgett, Solon Barocas, Hal Daumé III, and Hanna Wallach. 2020. Language (technology) is power: A critical survey of “bias” in NLP. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5454– 5476, Online. Association for Computational Linguistics. ↗
  3. Tolga Bolukbasi, Kai-Wei Chang, James Y. Zou, Venkatesh Saligrama, and Adam Tauman Kalai. 2016. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. In Neural Information Processing Systems. ↗
  4. Shikha Bordia and Samuel R. Bowman. 2019. Identifying and reducing gender bias in word-level language models. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop, pages 7–15, Minneapolis, Minnesota. Association for Computational Linguistics.
  5. Carole Brugeilles, Sylvie Cromer, and Centre développement. 2009. Analysing gender representations in school textbooks. find
  6. Aylin Caliskan, Joanna J. Bryson, and Arvind Narayanan. 2017. Semantics derived automatically from language corpora contain human-like biases. Science, 356(6334):183–186. ↗
  7. CK12. 2007. CK-12 Free Online Textbooks, Flashcards, Adaptive Practice, Real World Examples, Simulations. find
  8. Chandler May, Alex Wang, Shikha Bordia, Samuel R. Bowman, and Rachel Rudinger. 2019. On measuring social biases in sentence encoders. ArXiv, abs/1903.10561. ↗
  9. Michigan. 2014. Michigan Open Book Project. find
  10. Jenna Cryan, Shiliang Tang, Xinyi Zhang, Miriam J. Metzger, Haitao Zheng, and Ben Y. Zhao. 2020. Detecting gender stereotypes: Lexicon vs. supervised learning methods. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. find
  11. George A. Miller. 1992. WordNet: A lexical database for English. In Speech and Natural Language: Proceedings of a Workshop Held at Harriman, New York, February 23-26, 1992. ↗
  12. Jad Doughman and Wael Khreich. 2022. Gender bias in text: Labeled datasets and lexicons. CoRR, abs/2201.08675. ↗
  13. Lydia Namatende-Sakwa. 2018. The construction of gender in ugandan english textbooks: a focus on gendered discourses. Pedagogy, culture society, 26(4):609–629. ↗
  14. Jad Doughman, Wael Khreich, Maya El Gharib, Maha Wiss, and Zahraa Berjawi. 2021. Gender bias in text: Origin, taxonomy, and implications. In Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing, pages 34–44, Online. Association for Computational Linguistics. find
  15. Zahra Fatemi, Chen Xing, Wenhao Liu, and Caim-ming Xiong. 2023. Improving gender fairness of pre-trained language models without catastrophic forgetting. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 1249–1262, Toronto, Canada. Association for Computational Linguistics. find
  16. Alison Flood. 2023. Fresh call for oxford dictionaries to change ’sexist’ definitions. The Guardian. find
  17. Yasmeen Hitti, Eunbee Jang, Ines Moreno, and Carolyne Pelletier. 2019. Proposed taxonomy for gender bias in text; a filtering methodology for the gender generalization subtype. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing, pages 8–17, Florence, Italy. Association for Computational Linguistics. find
  18. Ji Ho Park, Jamin Shin, and Pascale Fung. 2018. Reducing gender bias in abusive language detection. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2799–2804, Brussels, Belgium. Association for Computational Linguistics.
  19. Anh Ngoc Quynh Phan and Tho Xuan Pham. 2021. Gender stereotypes as hidden curriculum: A case of vietnamese english textbooks. International Journal of Education. ↗
  20. OECD Pisa. 2015. Draft science framework. find
  21. Maarten Sap, Marcella Cindy Prasettio, Ari Holtzman, Hannah Rashkin, and Yejin Choi. 2017. Connotation frames of power and agency in modern films. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2329–2334, Copenhagen, Denmark. Association for Computational Linguistics. find
  22. Beatrice Savoldi, Marco Gaido, Luisa Bentivogli, Matteo Negri, and Marco Turchi. 2021. Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9:845–874. ↗
  23. C. J. Hutto and Eric Gilbert. 2014. Vader: A parsimonious rule-based model for sentiment analysis of social media text. find
  24. Siyavula. 2014. Open Textbooks | Siyavula. find
  25. Izzuddin, Reza Pahlevi Dalimunthe, and Sulistiyono Susilo. 2021. The portrayal of women in arabic textbooks for non-arabic speakers. SAGE Open, 11. ↗
  26. Steve Stemler. 2001. An overview of content analysis. Practical assessment, research & evaluation, 7(17):1– 6. find
  27. Lucy Li, Demszky Dorottya, Bromley Patricia, and Jurafsky Dan. 2020. Content analysis of textbooks via natural language processing: Findings on gender, race, and ethnicity in texas u.s. history textbooks. AERA Open, 6. ↗
  28. Terry Locke. 2004. Critical discourse analysis. Continuum research methods series. Continuum, London. find
  29. Christopher Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven Bethard, and David McClosky. 2014. The Stanford CoreNLP natural language processing toolkit. In Proceedings of 52nd Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 55–60, Baltimore, Maryland. Association for Computational Linguistics. ↗
  30. Paulina Toro Isaza, Guangxuan Xu, Toye Oloko, Yufang Hou, Nanyun Peng, and Dakuo Wang. 2023. Are fairy tales fair? analyzing gender bias in temporal narrative event chains of children’s fairy tales. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6509–6531, Toronto, Canada. Association for Computational Linguistics. ↗
  31. Sowmya Vajjala and Ivana Lučić. 2018. OneStopEnglish corpus: A new corpus for automatic readability assessment and text simplification. In Proceedings of the Thirteenth Workshop on Innovative Use of NLP for Building Educational Applications, pages 297–304, New Orleans, Louisiana. Association for Computational Linguistics.
  32. Sowmya Vajjala and Detmar Meurers. 2012. On improving the accuracy of readability classification using insights from second language acquisition. In Proceedings of the Seventh Workshop on Building Educational Applications Using NLP, pages 163–173, Montréal, Canada. Association for Computational Linguistics.
  33. Francisco Valentini, Germán Rosati, Damián Blasi, Diego Fernandez Slezak, and Edgar Altszyler. 2023. On the interpretability and significance of bias metrics in texts: a PMI-based approach. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 509–520, Toronto, Canada. Association for Computational Linguistics. ↗
  34. Phuong Anh Vu. 2008. Gender stereotypes in story textbooks for primary school students in vietnam. find
  35. Sandra Waxman. 2013. Building a Better Bridge. In Navigating the Social World: What Infants, Children, and Other Species Can Teach Us, pages 292–296. Oxford University Press. ↗
  36. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Ryan Cotterell, Vicente Ordonez, and Kai-Wei Chang. 2019. Gender bias in contextualized word embeddings. In 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), pages 629–634, Minneapolis, Minnesota. Association for Computational Linguistics. ↗
  37. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2017. Men also like shopping: Reducing gender bias amplification using corpus-level constraints. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2979–2989, Copenhagen, Denmark. Association for Computational Linguistics.
  38. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2018. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 15–20, New Orleans, Louisiana. Association for Computational Linguistics. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhu-2024-disagreeable-slovenly-honest-un-named/