Breaking Homogeneity: Diversifying Persona Sets for Creative LLM Outputs

2026-09-24

AI Summary

This study addresses the groupthink problem arising from homogeneous outputs of large language models in open-ended tasks by formulating persona diversification as a set-level conditional optimization problem. It proposes an orthogonal design space and an evolutionary persona generation method, combined with space-filling strategies such as coverage subset selection and uniform sampling, establishing the geometric distribution of persona sets as a universal mechanism for eliciting divergent outputs. Experimental results demonstrate that this approach increases response diversity by 78.8% and originality by 26.1% on the Alternative Uses Task (AUT). Furthermore, the proposed method is compatible with prompt engineering techniques, enabling further performance enhancements when integrated.

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.

Problem

• homogeneity
• persona diversification
• creative LLM outputs
• groupthink
• divergent thinking

Innovation

• Persona Diversification
• Evolutionary Persona Generation
• Set-level Conditioning
• Persona-set Geometry
• Creative LLM Outputs