Side-by-side Comparison Amplifies Dialect Bias in Language Models

• URL: http://arxiv.org/abs/2605.24384v2
• Date: Thu, 28 May 2026 23:41:28 GMT
• Title: Side-by-side Comparison Amplifies Dialect Bias in Language Models
• Abstract summary: Language models (LMs) can exhibit biases based on variations in their dialects.<n>We quantify covert dialect bias by evaluating how LMs associate stereotypical traits with intent-equivalent tweets.
• Score: 13.147713730335775
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Language models (LMs) can exhibit biases based on variations in their dialects, even in the absence of a dialect label, a behavior known as covert dialect bias. In this work, we quantify covert dialect bias in online discourse by evaluating how LMs associate stereotypical traits (derived from social psychology research on racial bias) with intent-equivalent tweets in Standard American English (SAE) and African-American Vernacular English (AAVE). While prior work shows that LMs associate more negative stereotypes with AAVE when evaluating tweets in isolation, we are surprised to find that this bias is significantly exacerbated when SAE / AAVE tweet pairs are compared side by side, a setting that more closely reflects high-impact decision making contexts in which models are used to rank candidates. The bias only worsens when dialect labels are explicitly specified. This is striking, given the extensive efforts from commercial developers to mitigate bias in their LMs. Encouragingly, we show that counterfactual fairness finetuning can mitigate covert dialect bias for some stereotypical traits, reducing average disparities when evaluating tweets in isolation, however, these improvements do not consistently hold across traits when evaluating SAE / AAVE tweets side by side. Our findings show that existing evaluation settings for covert dialect bias may underestimate its severity, specifically in contrastive settings. Additionally, overt dialect bias remains pronounced even after safety aligned finetuning, indicating that it remains an unresolved problem, and motivates the need for more robust evaluation and mitigation frameworks.

Related papers

• Reinforcing Stereotypes of Anger: Emotion AI on African American Vernacular English [46.47177439553625]
  This study examines emotion recognition model performance on African American Vernacular English (AAVE) compared to General American English (GAE)<n>We analyze 2.7 million tweets geo-tagged within Los Angeles.<n>We observe that neighborhoods with higher proportions of African American residents are associated with higher predictions of anger.
  arXiv  Detail & Related papers  (2025-11-13T23:13:08Z)
• Aligned but Blind: Alignment Increases Implicit Bias by Reducing Awareness of Race [14.700348476541684]
  We show that aligned language models (LMs) overlook racial concepts in early internal representations when the context is ambiguous.<n>We propose a new bias mitigation strategy that works by incentivizing the representation of racial concepts in the early model layers.
  arXiv  Detail & Related papers  (2025-05-30T21:41:44Z)
• Spoken Stereoset: On Evaluating Social Bias Toward Speaker in Speech Large Language Models [50.40276881893513]
  This study introduces Spoken Stereoset, a dataset specifically designed to evaluate social biases in Speech Large Language Models (SLLMs) By examining how different models respond to speech from diverse demographic groups, we aim to identify these biases. The findings indicate that while most models show minimal bias, some still exhibit slightly stereotypical or anti-stereotypical tendencies.
  arXiv  Detail & Related papers  (2024-08-14T16:55:06Z)
• The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Pre-trained Language Models [78.69526166193236]
  Pre-trained Language models (PLMs) have been acknowledged to contain harmful information, such as social biases. We propose sc Social Bias Neurons to accurately pinpoint units (i.e., neurons) in a language model that can be attributed to undesirable behavior, such as social bias. As measured by prior metrics from StereoSet, our model achieves a higher degree of fairness while maintaining language modeling ability with low cost.
  arXiv  Detail & Related papers  (2024-06-14T15:41:06Z)
• A Comprehensive View of the Biases of Toxicity and Sentiment Analysis Methods Towards Utterances with African American English Expressions [5.472714002128254]
  We study bias on two Web-based (YouTube and Twitter) datasets and two spoken English datasets. We isolate the impact of AAE expression usage via linguistic control features from the Linguistic Inquiry and Word Count software. We present consistent results on how a heavy usage of AAE expressions may cause the speaker to be considered substantially more toxic, even when speaking about nearly the same subject.
  arXiv  Detail & Related papers  (2024-01-23T12:41:03Z)
• What Do Llamas Really Think? Revealing Preference Biases in Language Model Representations [62.91799637259657]
  Do large language models (LLMs) exhibit sociodemographic biases, even when they decline to respond? We study this research question by probing contextualized embeddings and exploring whether this bias is encoded in its latent representations. We propose a logistic Bradley-Terry probe which predicts word pair preferences of LLMs from the words' hidden vectors.
  arXiv  Detail & Related papers  (2023-11-30T18:53:13Z)
• Disambiguation of morpho-syntactic features of African American English -- the case of habitual be [1.4699455652461728]
  Habitual "be" is isomorphic, and therefore ambiguous, with other forms of "be" found in both AAE and other varieties of English. We employ a combination of rule-based filters and data augmentation that generate a corpus balanced between habitual and non-habitual instances.
  arXiv  Detail & Related papers  (2022-04-26T16:30:22Z)
• The SAME score: Improved cosine based bias score for word embeddings [49.75878234192369]
  We introduce SAME, a novel bias score for semantic bias in embeddings. We show that SAME is capable of measuring semantic bias and identify potential causes for social bias in downstream tasks.
  arXiv  Detail & Related papers  (2022-03-28T09:28:13Z)
• UnQovering Stereotyping Biases via Underspecified Questions [68.81749777034409]
  We present UNQOVER, a framework to probe and quantify biases through underspecified questions. We show that a naive use of model scores can lead to incorrect bias estimates due to two forms of reasoning errors. We use this metric to analyze four important classes of stereotypes: gender, nationality, ethnicity, and religion.
  arXiv  Detail & Related papers  (2020-10-06T01:49:52Z)
• Hate Speech Detection and Racial Bias Mitigation in Social Media based on BERT model [1.9336815376402716]
  We introduce a transfer learning approach for hate speech detection based on an existing pre-trained language model called BERT. We evaluate the proposed model on two publicly available datasets annotated for racism, sexism, hate or offensive content on Twitter.
  arXiv  Detail & Related papers  (2020-08-14T16:47:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.