发表机构
University of Oxford; Burning Glass Institute; University College London（牛津大学; 燃烧玻璃研究所; 伦敦大学学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究揭示大型语言模型后训练阶段加剧招聘中的年龄歧视，导致模型决策高度相关，使全球系统性排斥率从5.6%升至17.3%，引发对边缘群体不平等的担忧。
AI 中文摘要
雇主越来越多地使用大型语言模型（LLMs）来自动化其招聘流程。本文研究了单一文化偏见（monocultural biases）的风险，即大型语言模型的广泛部署使得整个劳动力市场中的偏见趋于同质化，从而导致某些人口群体面临更大的系统性排斥。针对十个大型语言模型，我们测量了其基础版本和后训练版本中的招聘偏见，以确定哪个阶段（预训练还是后训练）导致了单一文化偏见。我们发现，与其基础模型相比，后训练模型对年长申请者发出回电的可能性降低了3.6%。这一负面转变在我们评估的十个模型中有八个出现。后训练模型的决策比基础模型具有更高的相关性，这很可能由技能或大学专业等人力资本特征驱动。然而，模型间更大的共识将全球系统性排斥率从5.6%提高到17.3%，并加剧了人口统计学上的不平等，后训练模型的交叉系统性排斥率介于12.2%至21.7%之间。我们发现，这种不平等主要由年龄歧视驱动，且在后训练阶段被加剧。这些结果表明，尽管后训练技术可能提升模型选拔最佳申请者的能力，但通过统一引入新的偏见，它们可能对处于边缘的群体增加系统性不平等风险。
英文摘要
Employers are increasingly using large language models (LLMs) to automate their hiring process. This paper investigates the risk of monocultural biases, in which the widespread deployment of large language models homogenizes biases across the labor market, leading to greater systemic exclusion for certain demographic groups. For ten LLMs, we measure hiring biases across their base and post-trained versions to identify which stage, pre-training or post-training, lead to monocultural biases. We find that, compared to their base models, post-trained models are 3.6% less likely to callback older applicants. This negative shift occurs in eight of the ten models that we evaluate. Post-trained models have much more correlated decisions than base models which is likely driven by human capital traits like skills or college major. However, greater consensus among models increases global systemic exclusion rates from 5.6% to 17.3% and exacerbates demographic inequalities, with intersectional systemic exclusion rates ranging from 12.2% to 21.7% for post-trained models. We find that this inequality is primarily driven by age-based discrimination that is exacerbated in post-training. These results indicate that while post-training techniques may improve models' abilities to select the best applicants, they may raise systemic inequality risks for those at the margin by uniformly introducing new biases.
CommentsAccepted at COLM 2026
