A Model of Flexible Control in Task Switching

The paper presents an ACT-R model of task switching, which implements flexible mechanism of control over task rules, that adapts to the level of task change predictability. In predictable task switching situations, the model loads relevant task rules into an easily accessible focus of attention (simulated with an ACT-R goal buffer). However, when tasks change randomly, such strategy would lead to long recovery from incorrect rules already loaded into the focus. So, during unpredictable task changes the model always retrieves relevant task rules from its declarative memory. Two experiments were administered to test assumptions of the model: one required predictable switching while the other imposed some level of unpredictability. The model aptly simulated the pattern of switch trial RTs from both studies, and it also replicated constant RTs of repeat trials in fully predictable conditions, as well as decreasing RTs of repeat trials in less predictable conditions.