让对话者匿名，防止偏见影响AI辩论结果。
When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning
- 通过移除身份标识实现响应匿名化，打破自我与他人偏见。
- 实验证明群体辩论中盲目附和比固执己见更普遍。
- 提出可量化偏见的指标，适合改进AI推理系统的研究者。
多智能体辩论（MAD）通过多个智能体交换观点并聚合意见来提升大语言模型的推理能力。然而近期研究发现，智能体并非中立：它们易受身份驱动的奉承和自我偏见影响，盲目采纳同伴观点或固执坚持自身初始输出，从而损害辩论可靠性。本文首次提出一个统一框架，同时缓解并量化身份偏见。首先，将辩论过程建模为基于身份加权的贝叶斯更新；其次，提出响应匿名化策略——在提示中移除身份标记，使智能体无法区分“自己”与“他人”，强制对身份赋予相等权重，从而降低偏见、提升可信度；第三，定义身份偏见系数（IBC），作为衡量智能体倾向于跟随同伴还是自身输出的可解释指标。跨多种模型与基准的实证研究证实，身份偏见广泛存在，其中奉承现象远多于自我偏见。研究强调应确保MAD系统基于内容而非身份进行推理。代码已开源：https://github.com/deeplearning-wisc/MAD-identity-bias。
原文摘要 · Abstract (English)
Multi-agent debate (MAD) aims to improve large language model (LLM) reasoning by letting multiple agents exchange answers and then aggregate their opinions. Yet recent studies reveal that agents are not neutral: they are prone to identity-driven sycophancy and self-bias, uncritically adopting a peer's view or stubbornly adhering to their own prior output, undermining the reliability of debate. In this work, we present the first principled framework that joins sycophancy and self-bias to mitigate and quantify identity bias in MAD. First, we formalize the debate dynamics as an identity-weighted Bayesian update process. Second, we propose response anonymization: by removing identity markers from prompts, agents cannot distinguish "self" from "peer", which forces equal weights on agent identity, thereby reducing bias and improving trustworthiness. Third, we define the Identity Bias Coefficient (IBC), a principled bias metric that measures an agent's tendency to follow its peer versus itself. Empirical studies across multiple models and benchmarks confirm that identity bias is widespread, with sycophancy far more common than self-bias. Our findings highlight the need to ensure that MAD systems reason based on content rather than identity. Code is released in https://github.com/deeplearning-wisc/MAD-identity-bias.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。