Abstract

Cognitive development on certain tasks (e.g., the balance scale task) is characterized by periods of relative stable, rule-like but suboptimal behavior that occur in a fixed order and alternate with short transition periods. Many models have been developed to capture the developmental phenomena associated with the balance scale task. However, most of these models do not account for important phenomena as discontinuous transitions or rely on questionable assumptions, and none of the models is able to predict improvement in behavior without feedback. We propose a computational model that is implemented in ACT-R and is based on the evaluation of success of applied knowledge structures combined with a mechanism to construct new knowledge by searching for differences in the presented balance scale problems. This model accounts for the empirical phenomena, including learning without feedback as is common in developmental tasks.

Citations