Modeling Democratic Elections via Congestion Networks

摘要

We address the problem for selecting a representative via a democratic election simulated by a multi-agent system. Due to lack of analytic tools for simulating electoral tendencies, we propose here, a model based on a set of non-cooperative agents which are competing for obtaining a maximum number of votes from a population. A key element in this model is to recognize the 'profiles' that characterize in a specific election how the voters make their political decisions. In order to influence the voters, the agents apply different strategies. In our model, an agent's strategy is represented via a vector of weights which indicate the amount on each profile the agent utilize to persuade the voters. We have programmed a simulation system of the voting tendencies by sectors, in agreement with the changes of strategies that the candidates perform throughout their political campaign. As the profiles used for characterizing the voters, and forming the strategies of the agents have a limited nature, then a congestion network is built. The use of a congestion network allows us to search the singular points of the competition system, enabling us to predict the possible winner of that competition according to the strategies used by the agents.