LLM investor simulations lack individual-level fidelity, study finds

A new study published on arXiv investigates the ability of large language models (LLMs) to accurately simulate individual investor behavior over time. In a paper-trading simulation involving 80 participants, researchers found that current LLMs struggled to reliably predict trade occurrences, buy-sell structures, and asset selections. While some LLMs achieved similar overall trading activity to participants, their portfolio trajectories differed significantly. The study highlights a critical gap between general behavioral accuracy and true individual-level fidelity in LLM simulations, suggesting a need for separate evaluations of behavioral accuracy, explanation alignment, and evidence dependence. AI

IMPACT Highlights limitations in LLM simulation capabilities, suggesting a need for more nuanced evaluation methods for personalized user modeling.

RANK_REASON Academic paper published on arXiv detailing research findings. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• arXiv
• Behavioral Accuracy
• cs.IR
• Individual-Level Fidelity
• Investor Simulation
• Large language models

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →