Audience segmentation restores heterogeneity in LLM social simulations, with moderate granularity and data-driven selection often improving structural and predictive fidelity on U.S. climate-opinion data while no configuration dominates all evaluation dimensions.
J., Jung, S., & Salminen, J
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Restoring Heterogeneity in LLM-based Social Simulation: An Audience Segmentation Approach
Audience segmentation restores heterogeneity in LLM social simulations, with moderate granularity and data-driven selection often improving structural and predictive fidelity on U.S. climate-opinion data while no configuration dominates all evaluation dimensions.