Parametric Social Identity Injection and Diversification in Public Opinion Simulation

2026-03-17

AI Summary

This work addresses the challenge that current large language models (LLMs) struggle to capture social diversity in public opinion simulation, often leading to flattened inter-group differences and homogenized intra-group responses. To overcome this limitation, the authors propose the Parameterized Social Identity Injection (PSII) framework, which enables fine-grained control over synthetic agents’ social identities by explicitly injecting parameterized representations of demographic attributes and value orientations into the model’s hidden layers. PSII is the first approach to identify and mitigate the “diversity collapse” phenomenon inherent in LLMs, surpassing the constraints of conventional prompt engineering. Evaluated on World Values Survey data, the method significantly enhances the realism and diversity of simulated opinion distributions, reduces KL divergence from real-world data, and strengthens both inter-group differentiation and intra-group heterogeneity.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Humans and AI: Learning Human Values and Preferences

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large language models (LLMs) have recently been adopted as synthetic agents for public opinion simulation, offering a promising alternative to costly and slow human surveys. Despite their scalability, current LLM-based simulation methods fail to capture social diversity, producing flattened inter-group differences and overly homogeneous responses within demographic groups. We identify this limitation as a Diversity Collapse phenomenon in LLM hidden representations, where distinct social identities become increasingly indistinguishable across layers. Motivated by this observation, we propose Parametric Social Identity Injection (PSII), a general framework that injects explicit, parametric representations of demographic attributes and value orientations directly into intermediate hidden states of LLMs. Unlike prompt-based persona conditioning, PSII enables fine-grained and controllable identity modulation at the representation level. Extensive experiments on the World Values Survey using multiple open-source LLMs show that PSII significantly improves distributional fidelity and diversity, reducing KL divergence to real-world survey data while enhancing overall diversity. This work provides new insights into representation-level control of LLM agents and advances scalable, diversity-aware public opinion simulation. Code and data are available at https://github.com/halsayxi/PSII.

Problem

• public opinion simulation
• social diversity
• large language models
• Diversity Collapse
• social identity

Innovation

• Parametric Social Identity Injection
• Diversity Collapse
• Public Opinion Simulation
• Large Language Models
• Representation-level Control