mlsanityData to 7 Oct 2026

Does Thinking Help Fairness? Reasoning Tokens Resolve Some Biases but Create More

arXiv 2609.30768 ↗First posted 25 Sep 2026cs.AIcited by 0

On the map: LLM reasoning → Prompting and chain-of-thought reasoning also fits: Causal reasoning in LLMs

Abstract (arXiv)

Thinking in reasoning language models (RLMs) has been subject to debate on whether it resolves or amplifies bias. Prior works have shown competing conclusions in both directions. Using a within-model thinking-vs.-non-thinking ablation across QwQ-32B, DeepSeek-R1-Distill-Qwen-32B, and Qwen3-32B on three high-stakes decision tasks (Adult, COMPAS, Credit), we show that thinking has an asymmetric dual effect on counterfactual fairness: it both resolves counterfactual flips produced by the non-thinking baseline and creates new flips at near-saturating model confidence. In all nine (model, dataset) combinations, the created flips outnumber the resolved flips by roughly 5 times. To explain the effect, we treat the thinking trace itself as a measurable site of fairness change and study it through two dynamic instruments: 1) We propose Counterfactual Depth Probability Gap (CDPG) to track bias evolution along thinking depth, and observe that bias propagates and amplifies with thinking. 2) We also formulate the Bias Transition Matrix (BTM) to show how predictions of counterfactual pairs change from non-thinking to thinking, and find that the asymmetric dual effect originates in the pair-state joint transition.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 10 of its 25 references that are in the map. First posted on arXiv 25 Sep 2026.

Builds on

10 of 25 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024Chain of Thought Still Thinks Fast: APriCoT Helps with Thinking Slow16 Aug
  2. 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul

Other references · 8

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  3. 2023Evaluating and Mitigating Discrimination in Language Model Decisionscited by 117
  4. 2025Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoningcited by 20
  5. 2025Robustly Improving LLM Fairness in Realistic Settings via Interpretabilitycited by 16
  6. 2025Reasoning Towards Fairness: Mitigating Bias in Language Models through Reasoning-Guided Fine-Tuningcited by 5
  7. 2025Is Reasoning All You Need? Probing Bias in the Age of Reasoning Language Modelscited by 4
  8. 2026Inference-Time Reasoning Selectively Reduces Implicit Social Bias in Large Language Modelscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026The Long Road to the Same Answer: Cognitive Bias Under Escalating Reasoning Budgets in Large Language Models
  2. 2025Do Biased Models Have Biased Thoughts?
  3. 2026How Does Thinking Mode Change LLM Moral Judgments? A Controlled Instant-vs-Thinking Comparison Across Five Frontier Models
  4. 2026To Think or Not To Think, That is The Question for Large Reasoning Models in Theory of Mind Tasks
  5. 2026BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMs
  6. 2026Thinking Leakage: A Causal Audit of NoThink Post-Training in Hybrid Reasoning Models
  7. 2026Sparks of Rationality: Do Reasoning LLMs Align with Human Judgment and Choice?
  8. 2026Counting on Thinking: Tracing Evidence Integration in Language Models