Multi-Agent Simulation for Choice under Risk
Ichiro Nishizaki, Tomohiro Hayashida
In this paper, we develop a simulation system with artificial autonomous adaptive agents selecting one out of a given pair of binary lotteries which are represented by probability distributions over two outcomes. Agent’s decisions are made by a learning classifier system, and after classifying information of a given pair of binary lotteries, an agent chooses one out of them. The condition part of a classifier consists of conditions identifying probabilities and payoffs of a pair of binary lotteries and conditions identifying characteristics of the lotteries known by several models describing behavioral regularities of choices under risk. We compare the result of the simulation with that of the experiment by Selten et al. (1999), and demonstrating the similarity between them, we consider a mechanism of human choices under risk. Finally, we examine the possibility of controlling a subject’s preference with respect to risky events by the lottery ticket procedure.