Do Persona-Infused LLMs Affect Performance in a Strategic Reasoning Game?

2025-12-07

AI Summary

It remains unclear whether persona-based prompting induces measurable behavioral differences or performance improvements in large language models (LLMs) during strategic reasoning—particularly in the adversarial game PERIL—due to the lack of a principled mapping from abstract personality descriptions to concrete decision strategies. Method: We propose a structured, mediation-based translation framework inspired by exploratory factor analysis, which systematically transforms LLM-generated persona responses into operationally defined, face-valid, and reliable heuristic strategies. Contribution/Results: Empirical evaluation demonstrates that strategically oriented personas significantly enhance LLM performance in adversarial settings—but only when translated into executable heuristics via our mediation process. The method substantially improves the reliability and interpretability of heuristic modeling, establishing a reproducible, causally grounded analytical framework for persona interventions in LLMs. This advances both the theoretical understanding and practical design of controllable, interpretable strategic behavior in foundation models.

Technology Category

• Game Theory and Economic Paradigms: Adversarial Learning
• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Although persona prompting in large language models appears to trigger different styles of generated text, it is unclear whether these translate into measurable behavioral differences, much less whether they affect decision-making in an adversarial strategic environment that we provide as open-source. We investigate the impact of persona prompting on strategic performance in PERIL, a world-domination board game. Specifically, we compare the effectiveness of persona-derived heuristic strategies to those chosen manually. Our findings reveal that certain personas associated with strategic thinking improve game performance, but only when a mediator is used to translate personas into heuristic values. We introduce this mediator as a structured translation process, inspired by exploratory factor analysis, that maps LLM-generated inventory responses into heuristics. Results indicate our method enhances heuristic reliability and face validity compared to directly inferred heuristics, allowing us to better study the effect of persona types on decision making. These insights advance our understanding of how persona prompting influences LLM-based decision-making and propose a heuristic generation method that applies psychometric principles to LLMs.

Problem

• Investigates persona prompting effects on LLM strategic decision-making
• Compares persona-derived heuristics with manually chosen strategies in game
• Proposes mediator method to translate personas into reliable heuristic values

Innovation

• Persona prompting with mediator translation improves strategic performance
• Structured mapping process enhances heuristic reliability and validity
• Applying psychometric principles to LLMs for heuristic generation