Using Symbolic Regression to Infer Strategies from Experimental Data

Abstract

We propose the use of a new technique -- symbolicregression -- as a method for inferring the strategiesthat are being played by subjects in economic decisionmaking experiments. We begin by describing symbolicregression and our implementation of this techniqueusing genetic programming. We provide a brief overviewof how our algorithm works and how it can be used touncover simple data generating functions that have theflavor of strategic rules. We then apply symbolicregression using genetic programming to experimentaldata from the ultimatum game. We discuss and analyzethe strategies that we uncover using symbolicregression and we conclude by arguing that symbolicregression techniques should at least complementstandard regression analyses of experimental data.