| Abstract: | In this paper we introduce a conception of learning which is a natural extension of economists' representations of learning and which is natural to develop using KBS technology. In the particular form of KBS we use, rule conditions and actions are well-formulated formulae of first-order predicate logic (FOPL). As a result, the simulation results obtained from these models are no less analytical than those of pure analytic models. Our results are further strengthened by an experimental design for simulations of competitive behaviour which eliminates implicit bias in the selection of possible behaviours. The system is applied to the Cournot duopoly model. We find that modest intelligence on the part of at least one duopolist systematically increases the profits of both. |
