A Noisy Model of Individual Behaviour.

This paper develops a model of individual adjustment subject to mistakes. In this case when mistakes are assumed i.i.d., this process produces a probability distribution of agents decision whose evolution is determined by Fokker-Planck equation. This distribution converges to the unique, globally stable stationary distribution. In the case when choice space is one dimensional this distribution satisfies Independence of Irrelevant Alternatives (IIA) property. In multidimensional case for IIA to hold strong additional conditions on payoff function or error structure should be satisfied. The paper also considers generalization of model for autocorrelated errors in one-dimensional case.