====== Bias ======

===== Bias (Fairness, Society and Ethics) =====

==== Bias in General ====

  * **Overviews**
    * [[https://arxiv.org/pdf/2110.08527.pdf|Meade et al 2021 - An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre-trained Language Models]]
  * **Papers**
    * [[https://arxiv.org/pdf/2103.00453.pdf|Schick et al 2021 - Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP]]

==== In Large Language Models ====
    * [[https://arxiv.org/pdf/2311.04892.pdf|Gupta et al 2023 - Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs]]


==== Gender Bias ====
  * **Overviews**
    * [[https://arxiv.org/pdf/1906.08976.pdf|Sun et al 2019 - Mitigating Gender Bias in Natural Language Processing: Literature Review]]
  * **Papers**
    * [[https://web.stanford.edu/class/linguist156/Lakoff_1973.pdf|Lakoff 1973 - Language and Woman's Place]] Linguistics paper from 1973 by [[https://en.wikipedia.org/wiki/Robin_Lakoff|Robin Lakoff]], often credited for making language and gender a major debate in linguistics.
    * [[https://arxiv.org/pdf/1607.06520.pdf|Bolukbasi et al 2016 - Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings]]
    * [[https://arxiv.org/pdf/1707.09457.pdf|Zhao et al 2017 - Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints]]
    * [[https://arxiv.org/pdf/1903.03862.pdf|Gonen & Goldberg 2019 - Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Them]]
    * [[https://arxiv.org/pdf/1904.03310.pdf|Zhao et al 2019 - Gender Bias in Contextualized Word Embeddings]]
    * [[https://www.aclweb.org/anthology/P19-1161v2.pdf|Zmigrod et al 2019 - Counterfactual Data Augmentation for Mitigating Gender Stereotypes in Languages with Rich Morphology]]
    * BUG dataset: [[https://arxiv.org/pdf/2109.03858.pdf|Levy et al 2021 - Collecting a Large-Scale Gender Bias Dataset for Coreference Resolution and Machine Translation]]
    * [[https://aclanthology.org/2022.findings-acl.55.pdf|Gupta et al 2022 - Mitigating Gender Bias in Distilled Language Models via Counterfactual Role Reversal]]
    * Data Augmentation
      * [[https://aclanthology.org/P19-1161v2.pdf|Zmigrod et al 2019 - Counterfactual Data Augmentation for Mitigating Gender Stereotypes in Languages with Rich Morphology]]
  * **In Applications**
    * Coreference Resolution
      * [[https://arxiv.org/pdf/1804.06876.pdf|Zhao et al 2018 - Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods]] Introduces **Winobias** coreference dataset. They use data-augmentation and word-embedding debiasing techniques, to remove bias in coref systems //without reducing performance on existing coref datasets//.
        * SoWinoBias: [[https://aclanthology.org/2021.gebnlp-1.12.pdf|Dawkins 2021 - Second Order WinoBias (SoWinoBias) Test Set for Latent Gender Bias Detection in Coreference Resolution]]
      * [[https://arxiv.org/pdf/1904.03310.pdf|Zhao et al 2019 - Gender Bias in Contextualized Word Embeddings]]
    * Dialog
      * [[https://arxiv.org/pdf/2005.00614.pdf|Dinan et al 2020 - Multi-Dimensional Gender Bias Classification]] [[https://huggingface.co/datasets/md_gender_bias|dataset]]
    * Machine Translation
  * **Datasets**
    * MDGender: [[https://arxiv.org/pdf/2005.00614.pdf|Dinan et al 2020 - Multi-Dimensional Gender Bias Classification]] [[https://huggingface.co/datasets/md_gender_bias|dataset]]
  * **Resources**
    * Gendered word lists
      * [[https://huggingface.co/datasets/md_gender_bias|MDGender]]

===== Dataset Bias =====
Includes for example, annotation artifacts. For an introduction, read [[https://arxiv.org/pdf/1803.02324.pdf|Gururangan 2018]] and [[https://arxiv.org/pdf/1808.05326.pdf|Zellers 2018]].
  * [[https://arxiv.org/pdf/1805.01042.pdf|Poliak et al 2018 - Hypothesis Only Baselines in Natural Language Inference]]
  * [[https://arxiv.org/pdf/1803.02324.pdf|Gururangan et al 2018 - Annotation Artifacts in Natural Language Inference Data]]
  * [[https://arxiv.org/pdf/1902.01007.pdf|McCoy et al 2019 - Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference]]
  * [[https://arxiv.org/pdf/1908.07898.pdf|Geva et al 2019 - Are We Modeling the Task or the Annotator? An Investigation of Annotator Bias in Natural Language Understanding Datasets]]
  * [[https://arxiv.org/pdf/2204.12708|Schwartz & Stanovsky 2022 - On the Limitations of Dataset Balancing: The Lost Battle Against Spurious Correlations]]


==== Reducing Annotation Artifacts During Dataset Creation ====
  * Adversarial Filtering
    * [[https://arxiv.org/pdf/1808.05326.pdf|Zellers et al 2018 - SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference]] Introduces Adversarial Filtering to try to eliminate dataset bias
    * [[https://arxiv.org/pdf/2002.00293.pdf|Bartolo et al 2020 - Beat the AI: Investigating Adversarial Human Annotation for Reading Comprehension]]

===== Classes and Tutorials =====
  * [[http://web.cs.ucla.edu/~kwchang/talks/emnlp19-fairnlp/|Tutorial: Bias and Fairness in Natural Language Processing]]

===== People =====
  * [[https://scholar.google.com/citations?user=fqDBtzYAAAAJ&hl=en|Kai-Wei Chang]]
  * [[https://scholar.google.com/citations?user=6RxMYNEAAAAJ&hl=en|Mark Yatskar]]

===== Related Pages =====
  * [[Ethics]]
  * [[Robustness in NLP]]
  * [[ml:Fairness]]