Uma extensão baseada em características de redes sociais do modelo evolucionário de preferências sociais para autorregulação de agentes aplicada ao jogo do ultimato

Abstract

In order to find a better form of agents self-regulation, this work aims to propose an extension based on social networking characteristics to an evolutionary preference model for self-regulation of social agents applied to the ultimatum game. For this purpose we developed a model inspired in features of online social networks as used in sites like Ebay and Mercado livre. Thus we have as starting point to model a system that were based to a real environment of individuals and has a simpler approach to solve the problem of social trade in multiagent systems. Thus, the focus of this dissertation is the development of a hybrid model of agents inspired by social exchanges and discovery algorithms profiles on Online Social Networks, to self-manage their social exchanges (based on personality), negotiating and resolving conflicts dependencies in order to achieve and maintain social equilibrium. The model considers a set of personality traits that are invisible by other agents, but indirectly guide the success or failure of social exchanges. To validate the proposed model, we developed two scenarios of the Ultimatum Game (UG). A pure, where there was only the evolution of the model during iterations and another with the inspiration of exchanges in social networks. To analyze the efficiency of the approach, we took into account the total number of exchanges in UG, which in both cases could be zero exchanges, where neither of the two agents involved accepted a bid from another; one exchange, where only o ne of the agents involved accepted the offer and, two exchanges, where both agents involved accepted negotiations. What could notice was that the proposed model, unlike the original model, which tended to evolve significantly until around the 500th iteration, when has stabilized and in some cases even declined, the proposed model keep evolving the results of the exchanges showing up more efficient.