Learning through social interactions

SEMESHENKO VIKTORIYA; GORDON MIRTA B; NADAL JEAN-PIERRE

Dresden, Germany

Conferencia; Physics of Socio-Economic Systems - AKSOE06; 2006

We consider a simple market model in which a population of heterogeneous individuals, subject  to local positive externalities, must choose repeatedly either to buy or not a unit of a single homogeneous good, at a price posted by a monopolist. If the weight of externalities is strong enough, there are multiple Nash equilibria, revealing coordination problems. We assume that individuals learn to make their choices repeatedly using deterministic and trembling hand decision  rules. We study the performances along the learning path and the reached equilibria for different learning schemes. In the presence of multiple equilibria, coordination on the optimal one  through learning is shown to be reached only with some of the learning schemes. The result  of the learning rules depends crucially on the learning parameters and the agents? initial beliefs.  With a trembling hand dynamics the system is shown to converge to a stationary state, in which the decisions fluctuate close to the optimal ones. This equilibrium has the flavor of (but is not identical to) what is known as a Quantal Response Equilibrium in the economic literature.