Using Computational Cognitive Modeling to Predict Dual-Task Performance With Sleep Deprivation

Abstract

OBJECTIVE: The effects of fatigue on multiple-task performance were explored through computational cognitive modeling. BACKGROUND: Fatigue typically has a negative impact on human performance. Biomathematical models exist that characterize the dynamics of human alertness, but the link between alertness and in situ performance on specific tasks is tenuous. Cognitive architectures offer a principled means of establishing that link. METHOD: We implemented mechanisms for fatigue, which produce microlapses in cognitive processing, into an existing model, adaptive control of thought-rational, and validated the performance predictions with Bratzke, Rolke, Ulrich, and Peters' data on fatigue and multiple-task performance. RESULTS: The microlapse model replicated the human performance results very well with zero free parameters, although the fit was improved when we allowed two individual differences parameters to vary. CONCLUSION: Increased frequency of microlapses as a result of fatigue provides a parsimonious explanation for the impact of fatigue on dual-task performance and is consistent with previous research. APPLICATION: Our results illustrate how using biomathematical models of fatigue in conjunction with a cognitive architecture can result in accurate predictions of the effects of fatigue on dual-task performance. Extending and generalizing this capability has potential utility in any safety-critical domain in which fatigue may affect performance.