Kwan-sun Hong, Jeong-ho Eom, Journal of Internet Computing and Services, Vol. 26, No. 3, pp. 145-152, Jun. 2025
10.7472/jksii.2025.26.3.145, Full Text:  HTML
Keywords: LLM, ProfieAgent, Korean War, Decision-Make

Abstract

This paper highlights the significance of human cognitive irrationality, an aspect often overlooked in conventional war decision-making models, and introduces the ProfileAgent framework, a novel methodology employing Large Language Model (LLM)-based AI agents to represent this irrationality. Using the Korean War (1950–53) as a case study, the research profiles key leaders' personalities and biases, enabling agents to accurately replicate historical decision-making processes and explore predictive simulations for similar future scenarios. The core finding demonstrates strong semantic similarity between simulated dialogues and actual historical records. By integrating behavioral economics and political psychology theories to model cognitive biases inherent in war decision-making and embodying human irrationality through AI language models, this research provides a new methodological approach to the study of military history. Additionally, it empirically verifies the reconstruction capability and predictive potential of the proposed ProfileAgent framework through experimentation with Korean War scenarios.

Statistics
Cite this article