広島大学大学院工学研究科研究報告 Volume 59 Issue 1
published_at 2010

エージェントベースシミュレーションを用いたムカデゲームにおける被験者の意思決定および学習に関する分析

Mechanism of decision making and learning of human subjects in centipede games by using agent-based simulation
Sugeo Yuya
Nishizaki Ichiro
fulltext
722 KB
BullGradSchEng-HiroshimaUniv_59-1_Sugeo-et-al.pdf
Abstract
This paper constructs an agent-based simulation model for behavioral analysis in a laboratory experiments of the centipede game using the human subjects. The centipede game is one of the games in extended forms which multiple players make decision in a predetermined order. In any game in extended forms, it is well known that the sub-game perfect equilibrium is a powerful solution concept to predict the behavior in the game. However, some experimental results are reported such that the equilibrium does not predict the behavior of human subjects, the centipede game is one of them. In the theoretical analysis approaches, it is assumed that players make decision are rational, though the decision making of human are not always rational but they make decision with the mechanism of trial and error. In this paper, a simulation model for behavioral analysis in the centipede game using artificial adaptive agents.
Keywords
centipede game
agent-based simulation
decision making
neural network