It is shown that some economic phenomena cannot be studied through models based on the classic scheme of the agent with perfect rationality. They require the construction of models where the agents' bounded cognitive processes are explicitly represented. This goal can be reached through computational models based on a genetic algorithms and neural networks. The strength of this approach is shown through an oligopolistic model in which artificial agents with learning capacity, autonomously develop their price-setting procedures. The results of the simulations show that decision makers, endowed with limited cognitive resources, may evolve toward simple and robust decision rules using less information in more complex environments. Moreover they show how market-price can be strongly influenced by agents' cognitive processes.
Neural networks and genetic algorithms for the simulation models of bounded rationality theory - An application to oligopolistic markets
Editore SIPI., Roma, Italia
Rivista di politica economica 87 (1997): 107–146.
info:cnr-pdr/source/autori:Baldassarre Gianluca/titolo:Neural networks and genetic algorithms for the simulation models of bounded rationality theory - An application to oligopolistic markets/doi:/rivista:Rivista di politica economica/anno:1997/pagina_da: