The Need for a Socially-Grounded Persona Framework for User Simulation

TL;DR

This paper introduces SCOPE, a socially grounded framework for creating and evaluating synthetic personas based on sociopsychological data, which significantly improves behavioral prediction and reduces bias compared to demographic-only personas.

Contribution

The paper presents SCOPE, a novel sociopsychological protocol for persona creation that enhances social simulation accuracy and reduces bias in large language models.

Findings

Demographic-only personas explain only ~1.5% of variance in responses.

Adding sociopsychological facets improves behavioral prediction.

SCOPE personas outperform default prompts and other models in alignment tests.

Abstract

Synthetic personas are widely used to condition large language models (LLMs) for social simulation, yet most personas are still constructed from coarse sociodemographic attributes or summaries. We revisit persona creation by introducing SCOPE, a socially grounded framework for persona construction and evaluation, built from a 141-item, two-hour sociopsychological protocol collected from 124 U.S.-based participants. Across seven models, we find that demographic-only personas are a structural bottleneck: demographics explain only ~1.5% of variance in human response similarity. Adding sociopsychological facets improves behavioral prediction and reduces over-accentuation, and non-demographic personas based on values and identity achieve strong alignment with substantially lower bias. These trends generalize to SimBench (441 aligned questions), where SCOPE personas outperform default prompting and NVIDIA Nemotron personas, and SCOPE augmentation improves Nemotron-based personas. Our results indicate that persona quality depends on sociopsychological structure rather than demographic templates or summaries.