Search papers, labs, and topics across Lattice.
The paper introduces Parametric Social Identity Injection (PSII), a framework to address the "Diversity Collapse" problem in LLM-based public opinion simulation, where social identities become indistinguishable in hidden representations. PSII injects explicit, parametric representations of demographic attributes and value orientations into the intermediate hidden states of LLMs. Experiments on the World Values Survey demonstrate that PSII significantly improves distributional fidelity and diversity compared to existing methods, reducing KL divergence to real-world data.
LLM-based simulations of public opinion suffer from "Diversity Collapse," but injecting explicit social identity representations into hidden states can fix it.
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.
Tsinghua AI