Demographic pluralism improves modeling diverse human opinions
Demographic Pluralism: Inference-Time Modeling of Pluralistic Human Preference Distributions
Artificial IntelligenceComputers and Society
Summary
People have different opinions about many topics, and these opinions vary even within groups of similar background. Existing AI tools often consider only broad groups and miss these smaller differences. The authors introduce a method called Demographic Pluralism that can estimate how opinions vary within different demographic groups without needing special training data. This approach better matches real opinion distributions when tested on various datasets and models.
What this means in practice
- •For content moderation teams: Improve content moderation by modeling diverse opinions within demographic groups to better respect cultural sensitivities.
- •For user experience designers: Design systems that reflect a wider range of user preferences by capturing opinion variation within demographic segments at inference time.
Authors
Meng-Chen Wu, Qipin Chen, Ansh Jain, Tess Wood, Zhe Du, Si-Chi Chin
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.