{
  "id": "2206.11484",
  "version": "v1",
  "published": "2022-06-23T05:30:47.000Z",
  "updated": "2022-06-23T05:30:47.000Z",
  "title": "Towards WinoQueer: Developing a Benchmark for Anti-Queer Bias in Large Language Models",
  "authors": [
    "Virginia K. Felkner",
    "Ho-Chun Herbert Chang",
    "Eugene Jang",
    "Jonathan May"
  ],
  "comment": "Accepted to Queer in AI Workshop @ NAACL 2022",
  "categories": [
    "cs.CL",
    "cs.CY"
  ],
  "abstract": "This paper presents exploratory work on whether and to what extent biases against queer and trans people are encoded in large language models (LLMs) such as BERT. We also propose a method for reducing these biases in downstream tasks: finetuning the models on data written by and/or about queer people. To measure anti-queer bias, we introduce a new benchmark dataset, WinoQueer, modeled after other bias-detection benchmarks but addressing homophobic and transphobic biases. We found that BERT shows significant homophobic bias, but this bias can be mostly mitigated by finetuning BERT on a natural language corpus written by members of the LGBTQ+ community.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2022-06-23T05:30:47.000Z"
    }
  ],
  "analyses": {
    "subjects": [
      "I.2.7"
    ],
    "keywords": [
      "large language models",
      "natural language corpus written",
      "significant homophobic bias",
      "measure anti-queer bias",
      "extent biases"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}