公平的输出，偏见的内部：在高风险决策中LLM中的因果潜能与潜在偏见的不对称性
Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes Decisions
AI总结：
研究发现，尽管指令微调的语言模型在高风险决策中表现公平，但其内部表示仍保留并放大了种族相关的偏见，且这种偏见在不同群体中不对称，需通过双层测试框架进行治理。
摘要：
指令微调的语言模型在高风险决策中表现出行为公平性，同时保留了内部表示中的偏见关联。然而，这些被压抑的表示是否会影响模型输出，以及这种因果潜能是否在不同人口群体之间对称，尚不清楚。我们利用匹配的应用程序，仅通过种族相关的名字差异来测试开放式权重模型在抵押贷款审批中的应用，揭示了关键的不一致：模型在输出层面没有偏见，但跨模型层保留并放大了人口表示。通过激活引导和新颖的跨层干预，我们证明了这种被压抑的信息是决策相关：当重新注入关键层时，它会产生几乎完全的决策反转。关键的是，这种潜在偏见是不对称的——引导干预影响一个群体方向的决策，而在反向时产生最小影响，并且容易受到对抗性提示工程和参数高效微调的影响。这些发现表明，专注于输出的行为审计是不够的：公平的输出可能掩盖可利用的内部偏见。它们还推动了双层测试框架的发展，结合输出评估与表示分析，以促进高风险决策中的人工智能治理。
英文摘要：
Instruction-tuned language models exhibit behavioural fairness in high-stakes decisions while retaining biased associations in their internal representations. However, whether these suppressed representations can affect model outputs - and whether such causal potency is symmetric across demographic groups - remains unknown. We investigate the use of open-weight models for mortgage underwriting using matched applications that differ only in racially-associated names and reveal a critical disconnect: models show no output-level bias, yet retain and amplify demographic representations across model layers. Through activation steering and novel cross-layer interventions, we demonstrate that this suppressed information is decision-relevant: when reinjected at critical layers, it produces near-complete decision reversals. Critically, this latent bias is asymmetric - steering interventions affect decisions in one demographic direction, while producing minimal effects in reverse - and susceptible to adversarial prompt engineering and parameter-efficient fine-tuning. These findings demonstrate that behavioural audits focused on outputs are insufficient: fair outputs can mask exploitable internal biases. They also motivate dual-layer testing frameworks combining output evaluation with representational analysis for AI governance in high-stakes decisions.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
