Higher-Order Binding of Language Model Virtual Personas: a Study on Approximating Political Partisan Misperceptions

2025-04-16

AI Summary

Existing LLM-based user simulators primarily model individual attitudes, failing to replicate higher-order social perception biases—such as stereotyping and misperception across heterogeneous groups—observed in real-world political polarization. Method: We propose a novel “higher-order binding” paradigm that constructs embodied virtual personas via long-horizon, multi-turn, detail-rich synthetic interview-style backstories. These personas not only encode individual stances but also systematically emulate cross-group misperception mechanisms. Our approach integrates LLM-driven backstory generation, persona-conditioned reasoning, and a Wasserstein distance–based evaluation framework. Contribution/Results: Experiments demonstrate an 87% improvement in alignment between virtual persona response distributions and empirical human data; key political psychology effect sizes are reproduced with <5% error. This significantly enhances simulation fidelity and theoretical interpretability for complex political phenomena—including polarization and intergroup conflict.

Technology Category

• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Humans and AI: Game Design — Virtual Humans, NPCs and Autonomous Characters
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large language models (LLMs) are increasingly capable of simulating human behavior, offering cost-effective ways to estimate user responses during the early phases of survey design. While previous studies have examined whether models can reflect individual opinions or attitudes, we argue that a emph{higher-order} binding of virtual personas requires successfully approximating not only the opinions of a user as an identified member of a group, but also the nuanced ways in which that user perceives and evaluates those outside the group. In particular, faithfully simulating how humans perceive different social groups is critical for applying LLMs to various political science studies, including timely topics on polarization dynamics, inter-group conflict, and democratic backsliding. To this end, we propose a novel methodology for constructing virtual personas with synthetic user ``backstories"generated as extended, multi-turn interview transcripts. Our generated backstories are longer, rich in detail, and consistent in authentically describing a singular individual, compared to previous methods. We show that virtual personas conditioned on our backstories closely replicate human response distributions (up to an 87% improvement as measured by Wasserstein Distance) and produce effect sizes that closely match those observed in the original studies. Altogether, our work extends the applicability of LLMs beyond estimating individual self-opinions, enabling their use in a broader range of human studies.

Problem

• Approximating political partisan misperceptions using LLMs
• Simulating nuanced inter-group perceptions for polarization studies
• Improving virtual persona consistency with detailed synthetic backstories

Innovation

• Extended multi-turn interview transcripts for personas
• Higher-order binding for nuanced group perceptions
• Synthetic backstories improve response distribution accuracy