Hierarchical Organization Simulacra in the Investment Sector
Chung-Chi Chen, Hiroya Takamura, Ichiro Kobayashi, Yusuke Miyao
TL;DR
The paper investigates whether hierarchical, multi-agent LLMs can mimic professional investment decision-making by using news articles as triggers. It introduces a Hierarchical Organization (HO) framework with Analyst, Trader, and Head Trader roles, and compares HO against Single Trader and chain-of-thought baselines across 115,223 Taiwan news articles and 15 years of trading data. Key findings show that HO improves alignment with professional decisions and profitability relative to simpler setups, but prompts and perceived seniority introduce biases that affect outcomes; longer horizons and cross-LLM collaboration can further modulate alignment with practitioners and markets. The work highlights both the potential and the limitations of LLM-driven hierarchical decision-making in finance, emphasizing design choices such as hierarchy depth, prompt formulation, and cross-LLM interaction for practical applications and future research.
Abstract
This paper explores designing artificial organizations with professional behavior in investments using a multi-agent simulation. The method mimics hierarchical decision-making in investment firms, using news articles to inform decisions. A large-scale study analyzing over 115,000 news articles of 300 companies across 15 years compared this approach against professional traders' decisions. Results show that hierarchical simulations align closely with professional choices, both in frequency and profitability. However, the study also reveals biases in decision-making, where changes in prompt wording and perceived agent seniority significantly influence outcomes. This highlights both the potential and limitations of large language models in replicating professional financial decision-making.