An information market for multi-agent decision making: Observations from a human experiment

Abstract

Intelligent decision support systems increasingly embrace agent-based mechanisms to cope with decentralized decision problems. In this paper, we focus on the economic market model commonly used as a blueprint for the design of autonomous multi-agent systems. We present an experimental information market in which the human traders have only limited information. We analyze the traders' private preferences and their actual behavior in the market as they exchange and acquire information. We observe that while the traders' individual preferences show a consistent deliberative pattern throughout the market experiment, their actual decision behavior in the market appears to be reactively driven by the decisions of the other traders. These observations from human traders may have important implications for the design of market-oriented multi-agent systems to address decision problems characterized by incomplete information.