Whose Norms? Disentangling Cultural and Personal Alignment in Large Language Models

2026-06-05

AI Summary

This study addresses the challenge that large language models (LLMs) face in effectively balancing cultural norms and individual preferences in social decision-making, a domain lacking systematic evaluation of their interaction. The authors propose the PACT framework, which, for the first time, disentangles cultural norms from personal preferences and integrates cross-cultural human experiments, instruction fine-tuning analysis, human–AI alignment assessment, and response distribution correlation metrics to systematically evaluate model behavior. Findings reveal that LLMs are highly sensitive to national context—exerting an effect size of 7.8%, substantially surpassing demographic factors like age and gender—yet exhibit significant divergence from human judgment distributions, achieving a maximum human–model correlation of only 0.24. Consequently, current models fail to replicate the diversity and uncertainty inherent in human social decisions, highlighting critical limitations in cultural adaptability and individual preference modeling.

Abstract

Large language models are increasingly used for social decision-making situations that require balancing cultural norms with personal preferences. For example, a user preferring honesty might ask whether to correct a coworker publicly when local norms favor indirect feedback. Yet existing research studies cultural alignment and personalization largely separately. We introduce PACT, the Personal-Preference and Cultural-Norm Trade-off framework, which evaluates whether models choose to follow a cultural norm or allow personal preferences. We find that LLMs vary in how rigidly they enforce cultural norms, with behavior shifted more by country context (7.8%) than age (1%) and gender (0.7%) and shifting non-uniformly after instruction tuning. Furthermore, our five-country human study on PACT shows that culture-following in humans is mainly driven by scenario country, with the lowest agreement when participants judge their own cultural contexts, showing within-culture pluralism. Finally, human-LLM alignment experiments show that models can match majority choices, but fail to capture response distributions and uncertainty (with best correlations reaching only 0.24). Together, these findings motivate alignment evaluations that go beyond majority to capture cultural pluralism and disagreement in social judgment.

Problem

• cultural norms
• personal preferences
• large language models
• social judgment
• cultural pluralism

Innovation

• cultural alignment
• personalization
• PACT framework
• LLM social judgment
• cultural pluralism