ACT-R Models of Cognitive Control in the Abstract Decision Making Task

This paper discusses a method of modeling individual differences in cognitive control by developing models that differ in control structure. Such a strategy for modeling behavior is necessitated when tasks are complex and individual differences in performance vary on many measures of performance. For these tasks, merely adjusting a parameter to fit various groups of subjects may be impractical or impossible. Two such models for the Abstract Decision Making task are implemented in ACT-R to fit the performance of high and low-control subjects in the first experiment. These models are then used to accurately predict performance on a second experiment that involves novel “games” unrepresented in the prior experiment.