Two Ways to Model the Effects of Sleep Fatigue on Cognition

We compare how the same cognitive model completes a task within two alternative modifications to a cognitive architecture to represent sleep deprivation. One modification (ACT-R/F) has a module that uses a biomathematical model of the effects of sleep deprivation on performance to drive parameter changes in the architecture that impact behavior and performance. The second, new, modification (ACT-R/Φ) represents the effects of sleep deprivation on physiological systems and has these systems modulate cognition. The model completes the psychomotor vigilance task (PVT) within both ACT-R/Φ and ACT-R/F. We found that the two implementations produced similar response times (means) in simulated days one and two. However, the distribution of the response times across the two days of sleep deprivation varied between models. The ACT-R/Φ model shows a wider distribution in both days 1 and 2 due to an increased and modulating production utility noise that affects its ability to select the correct rules consistently. Though they represent sleep deprivation in different ways, and on different levels, both of these implementations lead us towards a more unified understanding of how sleep deprivation affects our bodies, how we think and behave over time, and how to represent these effects.