Breaking Homogeneity: Diversifying Persona Sets for Creative LLM Outputs
Abstract
Language models often produce homogeneous responses to open-ended tasks; such homogeneity can spawn groupthink-the convergence of ideas toward a singular and potentially suboptimal decision. We formulate persona diversification as a set-level conditioning problem and study two orthogonal design choices: selecting versus generating personas, and space-filling versus frontier-seeking diversity. We instantiate this design space with four methods spanning coverage and dispersion subset selections, uniform-coverage sampling, and evolutionary persona generation. Evaluations on the Alternative Uses Task (AUT), Infinity-Chat, and Divergent Association Task (DAT) show the benefits of the proposed methods across tasks and creativity objectives. On AUT, evolutionary persona generation increases response diversity by 78.8%, originality by 26.1%, flexibility by 49.5%, and holistic creativity by 13.9% over task-only prompting, while maintaining 98.5% validity; on Infinity-Chat, it nearly doubles persona-induced response separation relative to random personas. Moreover, evolutionary personas compose with creativity-optimized prompting, further increasing its response diversity by 18.6% and creativity by 6.3%. These results establish persona-set geometry as a task-agnostic mechanism for eliciting divergent LLM outputs, and support persona diversification as a reusable complement to prompt optimization.
Related
- Augmented Hypothesis Testing with Persona-Based LLM Simulations
- Beyond Cultural Persona Bias: Response Geometry as a Measure for Language Models and People
- Diagnosing the Demographic Distributions of LLM-Based Synthetic Persona Data: Reference-Distribution Dependence and Post Hoc Adjustment
- Criteria for Weighted Homogeneity via Logarithmic Vector Fields
- A concise proof of homogeneity of an isoparametric hypersurface with (g, m)=(6, 2)
BibTeX
@misc{moon2026breaking,
title = {{Breaking Homogeneity: Diversifying Persona Sets for Creative LLM Outputs}},
author = {Sang Bin Moon and Nicole Cho and Daniel Borrajo and Sumitra Ganesh and Abolfazl Hashemi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.30492},
eprint = {2609.30492},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.30492},
}
RIS
TY - UNPB
TI - Breaking Homogeneity: Diversifying Persona Sets for Creative LLM Outputs
AU - Sang Bin Moon
AU - Nicole Cho
AU - Daniel Borrajo
AU - Sumitra Ganesh
AU - Abolfazl Hashemi
PY - 2026
DA - 2026/09/24
PB - arXiv
DO - 10.48550/arxiv.2609.30492
UR - https://arxiv.org/abs/2609.30492
AB - Language models often produce homogeneous responses to open-ended tasks; such homogeneity can spawn groupthink-the convergence of ideas toward a singular and potentially suboptimal decision. We formulate persona diversification as a set-level conditioning problem and study two orthogonal design choices: selecting versus generating personas, and space-filling versus frontier-seeking diversity. We instantiate this design space with four methods spanning coverage and dispersion subset selections, uniform-coverage sampling, and evolutionary persona generation. Evaluations on the Alternative Uses Task (AUT), Infinity-Chat, and Divergent Association Task (DAT) show the benefits of the proposed methods across tasks and creativity objectives. On AUT, evolutionary persona generation increases response diversity by 78.8%, originality by 26.1%, flexibility by 49.5%, and holistic creativity by 13.9% over task-only prompting, while maintaining 98.5% validity; on Infinity-Chat, it nearly doubles persona-induced response separation relative to random personas. Moreover, evolutionary personas compose with creativity-optimized prompting, further increasing its response diversity by 18.6% and creativity by 6.3%. These results establish persona-set geometry as a task-agnostic mechanism for eliciting divergent LLM outputs, and support persona diversification as a reusable complement to prompt optimization.
ER -