This research program will bring together elements of neural networks, stability analysis, and experimental economics. The purpose of this research is to develop and test theory for the evolutions of decision-making in dynamic oligopoly environments. Imagine that agents learn about each other and their environments through repeated institutional interactions, and adapt their decisions in light of the information they acquire. So guided by institutional rules and informational restrictions, an agent's decision provides a local gradient to make predictions about statistical regularity of decisions made. Such neural network models can be shown to be generalized versions of dynamic oligopoly models based on adaptive decision-making rules. A set of appropriate experiments with human subjects to validate and/or aid in amending the theory will be carried out. This project will focus on simple quantity setting or price setting oligopoly environments. Each agent involved in the learning process makes a simple decision and gets information and a reward before making another decision. These environments are simple enough that the computation of theoretical equilibria (e.g., noncooperative Nash equilibria) is achievable (though sometimes requiring numerical methods), and is testable through controlled experimentation with cash motivated human subjects. Understanding the dynamics of decision-making in these simple,environments portends the successful examination of more complex scenarios where policy issues may be at stake. For example, how do long term cost reduction (R&D) investments interact with short term pricing decisions; how can a few influential (large) buyers alter the decision-making influence the evolutions of market concentration and the distribution of market shares?
