AI 中文总结
本研究分析59国的LLM跨国价值观模拟，发现富裕技术先进国家人群模拟更准确，且两种干预路径的准确率提升未必带来表征平等，为构建更具包容性的LLM模拟提供指导。
AI 中文摘要
研究人类观点的传统方法往往难以支持跨国家的代表性和可扩展性研究。大型语言模型（LLMs）可作为模拟人类观点的可扩展代理，实现更高效的观点分析。然而，将LLMs用于此目的不仅需要高平均准确率，还需要表征平等，即不同人群的模拟准确率具有可比性。模拟准确率不均可能在下游应用中复制或放大社会偏见。本研究系统调查了59个国家的国家级表征平等，发现存在显著且系统性的不平等：来自更富裕、技术更先进国家的人群模拟更准确。我们进一步比较了两种基础干预路径——上下文适配和参数修改，结果表明，平均或目标群体准确率的提升并不一定能转化为更高的表征平等。对于上下文适配，母语提示通常能提高准确率，但仍依赖于模型；而额外信息往往能同时提高准确率和表征平等。对于参数修改，特定语言的持续后训练可提高目标语言群体的准确率，但效果不均；偏好对齐则未在准确率或表征平等上产生系统性提升。人工标注的偏好数据通常比AI标注的数据更好地保留了准确率。这些发现强调了在准确率之外还需关注表征平等，并为构建更具包容性、对社会负责的基于LLM的模拟提供了指导。
英文摘要
Traditional methods for studying human opinions often struggle to support representative and scalable research across countries. Large language models (LLMs) can serve as scalable proxies for simulating human opinions, enabling more efficient opinion analysis. However, this use of LLMs requires not only high average accuracy but also representational equality, that is, comparable simulation accuracy across populations. Uneven simulation accuracy may reproduce or amplify societal biases in downstream applications. This study systematically investigates country-level representational equality across 59 countries and finds substantial, systematic inequality. Populations from wealthier and more technologically advanced countries are simulated more accurately. We further compare two foundational intervention pathways, contextual adaptation and parametric modification, and show that improvements in average or target-group accuracy do not necessarily translate into greater representational equality. For contextual adaptation, native-language prompting generally improves accuracy but remains model-dependent, whereas additional information more often improves both accuracy and equality. For parametric modification, language-specific continued post-training improves accuracy for targeted language groups but unevenly, while preference alignment yields no systematic gains in accuracy or equality. Human-annotated preference data generally preserve accuracy better than AI-annotated data. These findings highlight the need for representational equality alongside accuracy and offer guidance for more inclusive, socially responsible LLM-based simulations.
CommentsAccepted for publication in Computational Linguistics
