Bias Association Discovery Framework for Open-Ended LLM Generations

2025/08/02 by Jinhao Pan, Pan, Jinhao, Chahat Raj +3
Computer Science · #Authorship Attribution and Profiling #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning in Healthcare #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2508.01412

openalex publication_date 2025/08/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/30

Abstract

Social biases embedded in Large Language Models (LLMs) raise critical concerns, resulting in representational harms -- unfair or distorted portrayals of demographic groups -- that may be expressed in subtle ways through generated language. Existing evaluation methods often depend on predefined identity-concept associations, limiting their ability to surface new or unexpected forms of bias. In this work, we present the Bias Association Discovery Framework (BADF), a systematic approach for extracting both known and previously unrecognized associations between demographic identities and descriptive concepts from open-ended LLM outputs. Through comprehensive experiments spanning multiple models and diverse real-world contexts, BADF enables robust mapping and analysis of the varied concepts that characterize demographic identities. Our findings advance the understanding of biases in open-ended generation and provide a scalable tool for identifying and analyzing bias associations in LLMs.

Citations

• Implicit Bias in LLMs: A Survey
• Unmasking Implicit Bias: Evaluating Persona-Prompted LLM Responses in Power-Disparate Social Scenarios
• AI generates covertly racist decisions about people based on their dialect
• SelfIE: Self-Interpretation of Large Language Model Embeddings
• Angry Men, Sad Women: Large Language Models Reflect Gendered Stereotypes in Emotion Attribution
• Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
• Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems
• BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset
• "I'm sorry to hear that": Finding New Biases in Language Models with a Holistic Descriptor Dataset
• Training language models to follow instructions with human feedback
• CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models
• Language (Technology) is Power: A Critical Survey of "Bias" in NLP
• Making Monolingual Sentence Embeddings Multilingual using Knowledge\n Distillation
• StereoSet: Measuring stereotypical bias in pretrained language models
• On Measuring Social Biases in Sentence Encoders

Related

• What's Not Said Still Hurts: A Description-Based Evaluation Framework for Measuring Social Bias in LLMs
• Ask LLMs Directly, "What shapes your bias?": Measuring Social Bias in Large Language Models
• Dissecting Bias in LLMs: A Mechanistic Interpretability Perspective
• Interpreting Bias in Large Language Models: A Feature-Based Approach
• Fine-Grained Bias Detection in LLM: Enhancing detection mechanisms for nuanced biases