Demographic Pluralism: Inference-Time Modeling of Pluralistic Human Preference Distributions

2026-09-29

AI Summary

This study addresses the tendency of large language models to overlook intra-group preference diversity in culturally sensitive scenarios. To this end, it proposes an inference-time pluralistic preference modeling framework that requires neither fine-tuning nor access to distributional data. Specifically, the method leverages LLMs to generate multi-perspective distributions within demographic subgroups and evaluates them using the Jensen-Shannon distance, further revealing the underlying mechanism by which equal-weight aggregation outperforms weighted alternatives. Evaluated on benchmarks such as GlobalOpinionQA, the proposed framework reduces distributional errors by 8.4%–26.4%, significantly surpassing modular pluralism baselines.

Abstract

Large language models (LLMs) are increasingly used in culturally sensitive settings, where alignment requires representing diverse preferences within populations. Yet existing methods model populations at coarse demographic or community levels and overlook within-group variation. We introduce Demographic Pluralism, an inference-time framework that estimates population-level opinion distributions without opinion-distribution training data or task-specific fine-tuning by generating multiple perspectives within demographically grounded groups. Across four backbones on GlobalOpinionQA and VITAL, it reduces Jensen-Shannon distance by 8.4%-26.4% over Modular Pluralism. Among weighted, equal-weighted, and inverse-weighted aggregation, equal weighting performs best overall; group-level error also increases with group weight, helping explain weighted aggregation's weaker performance.

Problem

• Demographic Pluralism
• Large Language Models
• Preference Distributions
• Alignment
• Opinion Modeling

Innovation

• Demographic Pluralism
• Inference-Time Modeling
• Pluralistic Alignment
• Opinion Distribution Estimation
• Aggregation Strategy