Many multi-agent settings require that agents identify appropriate partners or teammates with whom to work on tasks. When selecting potential partners, agents may benefit from obtaining information about alternative possibilities through gossip (i.e., by consulting others) or using a reputation system (a centralized repository of information about past behavior). This paper defines a statistical ' model, the "Information-Acquisition Source Utility model" (IASU) by which agents operating in an uncertain world can determine the amount of information to collect about potential partners before choosing one and which information sources they should consult (gossip, reputation system, or additional personal interaction with the agent). The IASU model explicitly represents the cost of information, which may vary by information source. To maximize the expected gain from a choice, it estimates the utility of choosing a partner by iteratively estimating the benefit of additional information. The paper reports empirical studies that compare the effectiveness of the IASU model with a baseline in which only prior experience with a potential partner is used as the basis of the decision and with a model that determines in advance both the amount of information and its allocation among the different sources. Two different application domains are used in these empirical studies, the Surrogate Venture Game model, which concerns choosing an optimal partner for a business venture, and a restaurant domain. The results of the experiments show that the use of the model significantly increases the agents' overall utility. Copyright © 2008, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.
|Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS
|Published - 1 Jan 2008