Carefully Considering Culture: Analyzing LLM Alignment in Single- and Multi-Cultural Settings using Cultural Consensus Theory
Carefully Considering Culture: Analyzing LLM Alignment in Single- and Multi-Cultural Settings using Cultural Consensus Theory
Title: 审慎考量文化：利用文化共识理论分析单一及多元文化背景下的大语言模型对齐
Abstract: Recent work in NLP has probed large language models for their understanding of cultural norms across countries. However, this work typically considers distributional patterns, ignoring group consensus or possible multicultural environments within a country.
摘要： 近期自然语言处理（NLP）领域的研究探讨了大语言模型对各国文化规范的理解。然而，这些研究通常仅关注分布模式，而忽略了群体共识或国家内部可能存在的多元文化环境。
In this work, we leverage cultural consensus theory (CCT) from cultural anthropology to model such multidimensional nuance. Applying CCT to the World Values Survey (WVS) across 10 countries and 12 domains, we demonstrate that models frequently misrepresent cultural structures by either failing to form cohesive consensus or severely over-regularizing consensus.
在这项工作中，我们利用文化人类学中的文化共识理论（CCT）来建模这种多维度的细微差别。通过将 CCT 应用于涵盖 10 个国家和 12 个领域的“世界价值观调查”（WVS），我们证明了模型经常会错误地呈现文化结构，表现为要么无法形成有凝聚力的共识，要么对共识进行了严重的过度正则化。
Through explicit representation of intra-group variance, CCT provides actionable diagnostics to evaluate when models reflect true人类 diversity versus algorithmic homogenization.
通过对群体内部差异的显式表征，CCT 提供了一种可操作的诊断方法，用于评估模型究竟是反映了真实的人类多样性，还是仅仅表现为算法同质化。
Paper Details:
- Authors: Krishna Pothugunta, John P. Lalor
- Subject: Computation and Language (cs.CL); Computers and Society (cs.CY)
- arXiv ID: 2608.09937
- Submission Date: 29 May 2026
论文详情：
- 作者： Krishna Pothugunta, John P. Lalor
- 学科： 计算与语言 (cs.CL)；计算机与社会 (cs.CY)
- arXiv 编号： 2608.09937
- 提交日期： 2026 年 5 月 29 日