A fuzzy system based approach to social modeling in multi-agent systems

The aim of this paper is twofold. In one hand, we want to make evident the necessity that agents operating in multi-agent environments have for a social model that somehow allows them to determine the trust they can put in the assertions made by the other agents in the environment. In the other hand, we want to introduce a quite general mechanism, fuzzy filters, based on fuzzy systems theory, which allows us to satisfy that necessity. We will study a problem where individual taxi driver agents compete in order to obtain passengers to transport in exchange of money. We have implemented a simulation in which the behavior of each individual agent is given by a set of fuzzy systems. Agents use one of them to determine the next passenger to go for. The other fuzzy systems, the fuzzy filters, allows the modeling of the other agents from the point of view of the confidence their assertions deserve. Agents will make use of those fuzzy filters to decide whether they must compete or not against the other agents for a passenger. A quite simple evolutionary method allows us to see how individual agents can improve their performance by refining its internal models of the competitor agents.