Modeling fatigue in the N-back task with ACT-R and the fatigue module

Fatigue, if not managed properly, can have dangerous consequences for cognition and performance. It has been well established that fatigue impairs cognition, but theoretical development is necessary to better understand this relationship and predict conditions when performance may be at risk. In the present work, we examine a theory of fatigue situated in the ACT-R cognitive architecture. The theory proposes that fatigue results in the reduction of activation of task-relevant procedural and declarative knowledge. However, the relative impacts of fatigue on these two types of knowledge remains unclear. Here we investigated a task that requires activation of both procedural and declarative knowledge and we examined the fit of models assuming different fatigue mechanisms. Thirty-nine participants completed a 2-back task across 8 sessions over a 24-hour period. There was a significant effect of time on reaction time, hit rate, and false alarm rate. Our ACT-R variants of the \textit{N}-back that included the fatigue module similarly showed an effect of time on those metrics. When comparing our variants to the behavioral data, the variant that included procedural lapses fit the data better than the variant that modeled fatigue as changes in activation strength and the variant that included both. These results provide information about the generalizability and boundary conditions of the mechanisms proposed by the ACT-R fatigue module.