WinoQueer: A Community-in-the-Loop Benchmark for Anti-LGBTQ+ Bias in Large Language Models

WinoQueer: A Community-in-the-Loop Benchmark for Anti-LGBTQ+ Bias in Large Language Models. Felkner, V., Chang, H. H., Jang, E., & May, J. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9126–9140, Toronto, Canada, July, 2023. Association for Computational Linguistics.

Paper doi abstract bibtex

We present WinoQueer: a benchmark specifically designed to measure whether large language models (LLMs) encode biases that are harmful to the LGBTQ+ community. The benchmark is community-sourced, via application of a novel method that generates a bias benchmark from a community survey. We apply our benchmark to several popular LLMs and find that off-the-shelf models generally do exhibit considerable anti-queer bias. Finally, we show that LLM bias against a marginalized community can be somewhat mitigated by finetuning on data written about or by members of that community, and that social media text written by community members is more effective than news text written about the community by non-members. Our method for community-in-the-loop benchmark development provides a blueprint for future researchers to develop community-driven, harms-grounded LLM benchmarks for other marginalized communities.

@inproceedings{felkner-etal-2023-winoqueer,
    title = "{W}ino{Q}ueer: A Community-in-the-Loop Benchmark for Anti-{LGBTQ}+ Bias in Large Language Models",
    author = "Felkner, Virginia  and
      Chang, Ho-Chun Herbert  and
      Jang, Eugene  and
      May, Jonathan",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.acl-long.507",
    doi = "10.18653/v1/2023.acl-long.507",
    pages = "9126--9140",
    abstract = "We present WinoQueer: a benchmark specifically designed to measure whether large language models (LLMs) encode biases that are harmful to the LGBTQ+ community. The benchmark is community-sourced, via application of a novel method that generates a bias benchmark from a community survey. We apply our benchmark to several popular LLMs and find that off-the-shelf models generally do exhibit considerable anti-queer bias. Finally, we show that LLM bias against a marginalized community can be somewhat mitigated by finetuning on data written about or by members of that community, and that social media text written by community members is more effective than news text written about the community by non-members. Our method for community-in-the-loop benchmark development provides a blueprint for future researchers to develop community-driven, harms-grounded LLM benchmarks for other marginalized communities.",
}

Downloads: 0

{"_id":"aJAezTgeMuif6WWFk","bibbaseid":"felkner-chang-jang-may-winoqueeracommunityintheloopbenchmarkforantilgbtqbiasinlargelanguagemodels-2023","author_short":["Felkner, V.","Chang, H. H.","Jang, E.","May, J."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"WinoQueer: A Community-in-the-Loop Benchmark for Anti-LGBTQ+ Bias in Large Language Models","author":[{"propositions":[],"lastnames":["Felkner"],"firstnames":["Virginia"],"suffixes":[]},{"propositions":[],"lastnames":["Chang"],"firstnames":["Ho-Chun","Herbert"],"suffixes":[]},{"propositions":[],"lastnames":["Jang"],"firstnames":["Eugene"],"suffixes":[]},{"propositions":[],"lastnames":["May"],"firstnames":["Jonathan"],"suffixes":[]}],"booktitle":"Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","month":"July","year":"2023","address":"Toronto, Canada","publisher":"Association for Computational Linguistics","url":"https://aclanthology.org/2023.acl-long.507","doi":"10.18653/v1/2023.acl-long.507","pages":"9126–9140","abstract":"We present WinoQueer: a benchmark specifically designed to measure whether large language models (LLMs) encode biases that are harmful to the LGBTQ+ community. The benchmark is community-sourced, via application of a novel method that generates a bias benchmark from a community survey. We apply our benchmark to several popular LLMs and find that off-the-shelf models generally do exhibit considerable anti-queer bias. Finally, we show that LLM bias against a marginalized community can be somewhat mitigated by finetuning on data written about or by members of that community, and that social media text written by community members is more effective than news text written about the community by non-members. Our method for community-in-the-loop benchmark development provides a blueprint for future researchers to develop community-driven, harms-grounded LLM benchmarks for other marginalized communities.","bibtex":"@inproceedings{felkner-etal-2023-winoqueer,\n title = \"{W}ino{Q}ueer: A Community-in-the-Loop Benchmark for Anti-{LGBTQ}+ Bias in Large Language Models\",\n author = \"Felkner, Virginia and\n Chang, Ho-Chun Herbert and\n Jang, Eugene and\n May, Jonathan\",\n booktitle = \"Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)\",\n month = jul,\n year = \"2023\",\n address = \"Toronto, Canada\",\n publisher = \"Association for Computational Linguistics\",\n url = \"https://aclanthology.org/2023.acl-long.507\",\n doi = \"10.18653/v1/2023.acl-long.507\",\n pages = \"9126--9140\",\n abstract = \"We present WinoQueer: a benchmark specifically designed to measure whether large language models (LLMs) encode biases that are harmful to the LGBTQ+ community. The benchmark is community-sourced, via application of a novel method that generates a bias benchmark from a community survey. We apply our benchmark to several popular LLMs and find that off-the-shelf models generally do exhibit considerable anti-queer bias. Finally, we show that LLM bias against a marginalized community can be somewhat mitigated by finetuning on data written about or by members of that community, and that social media text written by community members is more effective than news text written about the community by non-members. Our method for community-in-the-loop benchmark development provides a blueprint for future researchers to develop community-driven, harms-grounded LLM benchmarks for other marginalized communities.\",\n}\n\n\n","author_short":["Felkner, V.","Chang, H. H.","Jang, E.","May, J."],"key":"felkner-etal-2023-winoqueer","id":"felkner-etal-2023-winoqueer","bibbaseid":"felkner-chang-jang-may-winoqueeracommunityintheloopbenchmarkforantilgbtqbiasinlargelanguagemodels-2023","role":"author","urls":{"Paper":"https://aclanthology.org/2023.acl-long.507"},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://jonmay.github.io/webpage/cutelabname/cutelabname.bib","dataSources":["j3Qzx9HAAC6WtJDHS","5eM3sAccSEpjSDHHQ"],"keywords":[],"search_terms":["winoqueer","community","loop","benchmark","anti","lgbtq","bias","large","language","models","felkner","chang","jang","may"],"title":"WinoQueer: A Community-in-the-Loop Benchmark for Anti-LGBTQ+ Bias in Large Language Models","year":2023,"downloads":1}