DPRF: A Generalizable Dynamic Persona Refinement Framework for Optimizing Behavior Alignment Between Personalized LLM Role-Playing Agents and Humans

2025-10-15

AI Summary

Current LLM-based role-playing agents (RPAs) suffer from low behavioral fidelity and pronounced cognitive biases due to reliance on manually crafted, uncalibrated persona profiles. To address this, we propose the Dynamic Persona Refinement Framework (DPRF), an automated approach that iteratively identifies and corrects persona misalignments across models and diverse scenarios—including debate, social media interaction, interviews, and film review—via synergistic free-form generation and theory-informed structured analysis. Evaluated on five mainstream LLMs, DPRF achieves significant improvements in behavioral prediction accuracy relative to ground-truth individual behavior, outperforming all baseline persona modeling methods. Crucially, DPRF demonstrates strong generalizability across tasks and models, enabling robust applications in user simulation, computational social science, and personalized AI systems.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• 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: Agentic search

Abstract

The emerging large language model role-playing agents (LLM RPAs) aim to simulate individual human behaviors, but the persona fidelity is often undermined by manually-created profiles (e.g., cherry-picked information and personality characteristics) without validating the alignment with the target individuals. To address this limitation, our work introduces the Dynamic Persona Refinement Framework (DPRF).DPRF aims to optimize the alignment of LLM RPAs' behaviors with those of target individuals by iteratively identifying the cognitive divergence, either through free-form or theory-grounded, structured analysis, between generated behaviors and human ground truth, and refining the persona profile to mitigate these divergences.We evaluate DPRF with five LLMs on four diverse behavior-prediction scenarios: formal debates, social media posts with mental health issues, public interviews, and movie reviews.DPRF can consistently improve behavioral alignment considerably over baseline personas and generalizes across models and scenarios.Our work provides a robust methodology for creating high-fidelity persona profiles and enhancing the validity of downstream applications, such as user simulation, social studies, and personalized AI.

Problem

• Optimizes LLM role-playing agents' behavioral alignment with target individuals
• Identifies cognitive divergence between generated behaviors and human ground truth
• Refines persona profiles to mitigate behavioral misalignments across scenarios

Innovation

• Iteratively identifies cognitive divergence between agents and humans
• Refines persona profiles to mitigate behavioral misalignments
• Generalizes across diverse models and behavioral scenarios