From Surveys to Narratives: Rethinking Cultural Value Adaptation in LLMs

Adapting cultural values in Large Language Models (LLMs) presents significant challenges, particularly due to biases and limited training data. Prior work primarily aligns LLMs with different cultural values using World Values Survey (WVS) data. However, it remains unclear whether this approach effectively captures cultural nuances or produces distinct cultural representations for various downstream tasks. In this paper, we systematically investigate WVS-based training for cultural value adaptation and find that relying solely on survey data can homogenize cultural norms and interfere with factual knowledge. To investigate these issues, we augment WVS with encyclopedic and scenario-based cultural narratives from Wikipedia and NormAd. While these narratives may have variable effects on downstream tasks, they consistently improve cultural distinctiveness than survey data alone. Our work highlights the inherent complexity of aligning cultural values to guide task-specific behavior. Code: https://github.com/faridlazuarda/ from-surveys-to-narratives . Introduction Recent research in Large Language Models (LLMs) suggests LLMs align closely with the cultural values of Western, Educated, Industrialized, Rich, and Democratic (WEIRD, Henrich et al. 2010) societies without adaptations (Johnson et al., 2022; Ramezani and Xu, 2023; Cao et al., 2023, among others). The WEIRD-centric bias can harm specific groups and limit the model's usefulness to a diverse global audience. Indeed, culture is a distinct and vital aspect of human society, influencing behavior, norms, and worldviews (Geertz, 2017). However, current research lacks robust mechanisms to adapt LLMs' outputs in ways that reflect different cultural value systems (i.e., culturally adapt LLMs). 1 1 For this paper, we focus on "culture" at a linguisticregional level (e.g., Iraq and Jordan represent Arab culture 5 0 5 10 15 Dimension 1 (UMAP)