Social Bias in Language Models

Overview

This topic studies how language models reflect or produce social disparities tied to traits such as gender, race, disability, religion, class, and language variety, in both text generation and consequential tasks. Researchers use counterfactual prompts, stereotype and decision-making benchmarks, human and community judgments, and audits across languages, dialects, modalities, and multi-turn interactions. Mitigation approaches include data augmentation and fine-tuning, prompting and retrieval, decoding-time controls, and interventions on model weights or internal representations. A central challenge is that measured bias changes with the task and context, while interventions can reduce visible disparities without removing underlying associations or can shift harms and degrade useful capabilities.

Latest papers

CardsList

217 papers20 papers in the last four weeks, level with the four weeks before. 0.2% of all new papers.

  1. FairCoder: Probing LLM Bias in High-Stakes Decision Making via Coding Tasks

    Jan 9, 2025Yongkang Du, Jen-tse Huang, Jieyu Zhao +1Social Bias in Language ModelsAlgorithmic Fairness

  2. BTBR: A Bayesian-Theory-Driven Probabilistic-Fuzzy Framework for Implicit Bias Removal in Large Language Models

    Aug 20, 2024Yongxin Deng, Xiaoyu Tan, Jing Pan +3Social Bias in Language ModelsLanguage Model Bias Evaluation

  3. BA-LoRA: Bias-Alleviating Low-Rank Adaptation to Mitigate Catastrophic Inheritance in Large Language Models

    Aug 8, 2024Yupeng Chang, Yi Chang, Yuan WuSocial Bias in Language ModelsLow-Rank Adaptation

  4. Fairness Definitions in Language Models Explained

    Jul 26, 2024Zhipeng Yin, Zichong Wang, Avash Palikhe +1Social Bias in Language ModelsDemographic Bias in Language Models

  5. MBBQ: A Dataset for Cross-Lingual Comparison of Stereotypes in Generative LLMs

    Jun 11, 2024Vera Neplenbroek, Arianna Bisazza, Raquel FernándezSocial Bias in Language ModelsLanguage Model Bias Evaluation

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

    Nov 23, 2023Bingkang Shi, Xiaodan Zhang, Dehan Kong +4Gender Bias in Language ModelsDebiased ML