ABSTRACT
This paper analyzes the use of a genetic algorithm (GA) for auto-fitting a non-linear, multivariable model to data. The example model and task used to illustrate the effectiveness of a GA is the Tower of Nottingham task. Empirical data collected from adults completing the task were used to create an initial model. This model’s variables were adjusted with a GA to simulate the performance of a child completing the task. The optimized fit was better than the previous fit generated by hand. We note the steps required to duplicate the process with other models.
