Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoning

2025/02/21 by Xuyang Wu, Wu, Xuyang, Jinming Nian +9 · 5 citations
Computer Science · Social Sciences · #Artificial Intelligence (cs.AI) #Artificial Intelligence in Law #Computation and Language (cs.CL) #FOS: Computer and information sciences #Multi-Agent Systems and Negotiation

paper · pdf · doi:10.48550/arxiv.2502.15361

openalex publication_date 2025/02/21 · openalex created_date 2025/10/16 · openalex updated_date 2026/07/28

Abstract

Recent advances in large language models (LLMs) have enabled automatic generation of chain-of-thought (CoT) reasoning, leading to strong performance on tasks such as math and code. However, when reasoning steps reflect social stereotypes (e.g., those related to gender, race or age), they can reinforce harmful associations and lead to misleading conclusions. We present the first systematic evaluation of social bias within LLM-generated reasoning, focusing on reasoning language models (e.g., DeepSeek-R1, OpenAI o1) that natively produce reasoning chains as part of their answers. Using the BBQ dataset, we analyze both prediction accuracy and reasoning bias across a broad spectrum of models, including instruction-tuned and CoT-augmented variants of DeepSeek-R1 (8B/32B), ChatGPT, and other open-source LLMs. We quantify how biased reasoning steps correlate with incorrect predictions and often lead to stereotype expression. To mitigate reasoning-induced bias, we propose Answer Distribution as Bias Proxy (ADBP), a lightweight mitigation method that detects bias by tracking how model predictions change across incremental reasoning steps. ADBP outperforms Stereotype-free Reasoning Pattern (SfRP) baseline in most cases, mitigating bias and improving the accuracy of LLM outputs. Evaluation and mitigation code is available at https://github.com/elviswxy/LLM_reasoning_bias.

Cited by

• Schoenfeld's Anatomy of Mathematical Reasoning by Language Models
• Investigating Thinking Behaviours of Reasoning-Based Language Models for Social Bias Mitigation
• Test-Time Reasoners Are Strategic Multiple-Choice Test-Takers
• FairReason: Balancing Reasoning and Social Bias in MLLMs
• Cognitive Chain-of-Thought: Structured Multimodal Reasoning about Social Situations

Related

• Evaluating Implicit Biases in LLM Reasoning through Logic Grid Puzzles
• BiasAlert: A Plug-and-play Tool for Social Bias Detection in LLMs
• Veracity Bias and Beyond: Uncovering LLMs' Hidden Beliefs in Problem-Solving Reasoning
• BiasCause: Evaluate Socially Biased Causal Reasoning of Large Language Models
• Investigating Thinking Behaviours of Reasoning-Based Language Models for Social Bias Mitigation