Personality Structured Interview for Large Language Model Simulation in Personality Research

• DOI: 10.48550/arxiv.2502.12109
• Publication Date: 2025-02-17
• Authors: Pengda Wang; Huiqi Zou; Hanjie Chen; Tianjun Sun; Ziang Xiao; Frederick L. Oswald
• Keywords: FOS: Computer and information sciences; Computer Science - Computation and Language; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Computation and Language (cs.CL)

Although psychometrics researchers have recently explored the use of large language models (LLMs) as proxies for human participants, LLMs often fail to generate heterogeneous data with human-like diversity, which diminishes their value in advancing social science research. To address these challenges, we potential theory-informed Personality Structured Interview (PSI) a tool simulating responses personality In this approach, simulation is grounded nuanced real-human interview transcripts that target construct interest. We provided growing set 357 structured from representative sample, each containing an individual's response 32 open-ended questions carefully designed gather theory-based evidence. Additionally, psychometric research, summarized evaluation framework systematically validate LLM-generated data. Results three experiments demonstrate well-designed interviews could improve heterogeneity LLM-simulated and predict personality-related behavioral outcomes (i.e., organizational citizenship behaviors counterproductive work behavior). further discuss role LLM-based outline general designing simulate