Linguistic Bias in Language Models

Overview

The field examines how dialect, register, language, locale, and demographic cues affect what language models generate, understand, recommend, or judge. Researchers use controlled minimal pairs, multilingual benchmarks, human comparisons, and analyses of model representations to test differences across settings such as AAVE and standard English, low-resource languages, and high-stakes assessment. Mitigation approaches include balanced training and fine-tuning, decoding-time fairness checks, activation steering, and retrieval methods that preserve evidence across languages. A central tension is that alignment can reduce overt bias while latent associations and disparities persist or shift with prompt wording, task, language, and model architecture, making reliable measurement and generalization difficult.

Latest papers

92 papers4 papers in the last four weeks, down 43% on the four weeks before. 0.0% of all new papers.

Jan 4, 2025

Homogeneity Bias in Open-Weight LLMs Is Robust to Decoding Hyperparameters

Large language models (LLMs) reproduce homogeneity bias -- the tendency to portray marginalized groups as more internally similar than dominant groups -- but whether this bias is stable or an artifact of inference settings has only been studied in single proprietary models. We map homogeneity bias across a 5x5 temperature-by-top-p grid in seven open-weight instruction-tuned LLMs (7-20B parameters). Hispanic and Asian Americans are portrayed as more homogeneous than White Americans in at least 18 of 20 hyperparameter configurations across six of seven models, including at extreme sampling settings. African American and gender bias show model-specific variation in direction. A conservative cell-level re-analysis confirms Hispanic and Asian homogeneity as robust, while weaker African American and gender signals largely do not survive, establishing group-specific robustness. We also apply the same grid to a names-based paradigm in which group identity is signaled via racially distinctive surnames rather than explicit labels. The names paradigm corroborates Hispanic and Asian homogeneity bias, but Black-coded surnames elicit robustly less homogeneous outputs than White-coded names in every model tested -- a reversal absent from the label paradigm -- showing that how group identity is operationalized shapes which biases surface and in which direction.
Nov 23, 2023

General Phrase Debiaser: Debiasing Masked Language Models at a Multi-Token Level

The social biases and unwelcome stereotypes revealed by pretrained language models are becoming obstacles to their application. Compared to numerous debiasing methods targeting word level, there has been relatively less attention on biases present at phrase level, limiting the performance of debiasing in discipline domains. In this paper, we propose an automatic multi-token debiasing pipeline called General Phrase Debiaser, which is capable of mitigating phrase-level biases in masked language models. Specifically, our method consists of a phrase filter stage that generates stereotypical phrases from Wikipedia pages as well as a model debias stage that can debias models at the multi-token level to tackle bias challenges on phrases. The latter searches for prompts that trigger model's bias, and then uses them for debiasing. State-of-the-art results on standard datasets and metrics show that our approach can significantly reduce gender biases on both career and multiple disciplines, across models with varying parameter sizes.