Large Language Models Substantially Compress Well-Being Inequality but Largely Preserve Its Socioeconomic Structure
Not on the map. Why?
Abstract (arXiv)
Research using large language models (LLMs) to generate synthetic populations has repeatedly shown that model outputs compress the diversity of human experience. This has raised doubts about whether LLM-generated data can capture meaningful differences within populations. We show that such compression does not necessarily erase the social structure of human heterogeneity. Using 93,901 respondents from 66 countries and territories in Wave 7 of the World Values Survey, we ask six LLMs to predict respondents' life satisfaction from demographic, socioeconomic, and attitudinal profiles. All six models substantially understate the overall dispersion of life satisfaction. Yet after normalizing for these differences in scale, they largely reproduce the human income gradient in well-being inequality: lower-income groups remain relatively more heterogeneous than higher-income groups. The pattern is robust to country fixed effects, equal-country weighting, WVS survey weights, and observed demographic composition, and it extends directionally to employment, education, and perceived control. Fidelity is weaker for extreme outcomes and country-specific gradients. These results show that the amount of heterogeneity preserved by an LLM and the way that heterogeneity is distributed across social groups are distinct properties. LLM-generated populations can therefore substantially compress human variation while retaining meaningful information about where that variation is concentrated.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 12 of its 35 references that are in the map. First posted on arXiv 27 Sep 2026.
Builds on
12 of 35 references.
References · 12
- 2023Whose Opinions Do Language Models Reflect?cited by 773
- 2022Out of One, Many: Using Language Models to Simulate Human Samplescited by 752
- 2024Quantifying the Persona Effect in LLM Simulationscited by 169
- 2024Large language models that replace human participants can harmfully misportray and flatten identity groupscited by 144
- 2024Evaluating Large Language Model Biases in Persona-Steered Generationcited by 78
- 2024Benchmarking Distributional Alignment of Large Language Modelscited by 61
- 2023Diminished Diversity-of-Thought in a Standard Large Language Modelcited by 45
- 2025Large Language Models as Virtual Survey Respondents: Evaluating Sociodemographic Response Generationcited by 7
- 2026Large Language Models Are Bad Dice Players: LLMs Struggle to Generate Random Numbers from Statistical Distributionscited by 6
- 2026Can Persona-Prompted LLMs Emulate Subgroup Values? An Empirical Analysis of Generalisability and Fairness in Cultural Alignmentcited by 5
- 2026Large language models accurately predict public perceptions of support for climate action worldwidecited by 1
- 2026Beyond Averages: Evaluating LLMs on Human Survey Replication at the Distributional Levelcited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Individual utilities of life satisfaction reveal inequality aversion unrelated to political alignment
- 2026Beyond Marginal Distributions: A Framework to Evaluate the Representativeness of Demographic-Aligned LLMs
- 2024Uplifting Lower-Income Data: Strategies for Socioeconomic Perspective Shifts in Large Multi-modal Models
- 2024One fish, two fish, but not the whole sea: Alignment reduces language models' conceptual diversity
- 2025Distributive Fairness in Large Language Models: Evaluating Alignment with Human Values
- 2026The Limits of Simulated Societies: How Post-Training and Survey Fine-Tuning Erase Cross-Cultural Variance
- 2025Can Finetuing LLMs on Small Human Samples Increase Heterogeneity, Alignment, and Belief-Action Coherence?
- 2025AlignSurvey: A Comprehensive Benchmark for Human Preferences Alignment in Social Surveys