Neurobehaviorally Inspired ACT-R Model of Sleep Deprivation: Decreased Performance in Psychomotor Vigilance

Abstract

A Neurobehaviorally Inspired ACT-R Model of Sleep Deprivation: Decreased Performance in Psychomotor Vigilance Glenn Gunzelmann (glenn.gunzelmann@mesa.afmc.af.mil) Kevin A. Gluck (kevin.gluck@mesa.afmc.af.mil) Air Force Research Laboratory; 6030 South Kent Street; Mesa, AZ 85212 USA Hans P. A. Van Dongen (vdongen@mail.med.upenn.edu) Robert M. O’Connor (roconnor@mail.med.upenn.edu) David F. Dinges (dinges@mail.med.upenn.edu) Division of Sleep and Chronobiology, University of Pennsylvania School of Medicine 1013 Blockley Hall; 423 Guardian Drive; Philadelphia, PA 19104-6021 The model is inspired by neurophysiological findings from the sleep restriction community, and demonstrates the promise of this approach for understanding the impact of sleep restriction on performance. Abstract This paper describes how changes in architectural parameters in ACT-R can be used to understand and predict the effects of sleep deprivation on a fundamental aspect of human performance. In a sample task, the parameter manipulations produce changes in the model’s performance that closely resemble the neurobehavioral effects seen in human data. The parameter that is manipulated (G) influences a mechanism in ACT-R that is considered to be associated with the thalamus, an area that is sensitive to sleep deprivation. Neurophysiology of Sleep Restriction Research on sleep restriction has identified some of the ways in which lack of sleep impacts brain activity (e.g., Durmer & Dinges, 2005; Drummond & Brown, 2001; Lin, 2000; Portas et al., 1998). The effects that are observed depend on what task is being performed, especially at the cortical level (e.g., Drummond & Brown, 2001). At the subcortical level, the thalamus has been implicated in regulating arousal (Lin, 2000; Moruzzi & Magoun, 1949), particularly in modulating attention (Portas et al., 1998). Research on sleep restriction has found changes in the activation of the thalamus as a function of sleep debt (e.g., Portas et al., 1998). These findings point to the thalamus as a key neural structure in mediating the effects of sleep deprivation on cognitive performance. Introduction Sleep is essential for normal human functioning. When people are deprived of sleep or even experience restricted sleep schedules, their performance degrades. These performance drops are evident in everything from simple sustained attention reaction time tasks like the psychomotor vigilance task (Van Dongen, 2004) to complex, dynamic tasks like flying high-fidelity military aircraft simulators (Caldwell et al., 2004). The effects of sleep deprivation range from subtle increases in reaction times to “sleep attacks,” where an individual falls asleep while engaged in goal-directed behavior (Durmer & Dinges, 2005). These effects can have major consequences in settings where swing shifts or long or unusual hours are the norm, such as long-haul trucking, commercial aviation, and military operations. In order to improve our ability to predict how and when performance will decline as a result of restricted or deprived sleep, we must improve our understanding of how sleep deprivation impacts the cognitive system. If these predictions can be made, then actions can be taken a priori to mitigate the effects of sleep deprivation and minimize the likelihood of costly or tragic fatigue-related errors (Dinges, This paper describes our recent efforts along these lines. We first describe relevant research from the sleep restriction and cognitive modeling literatures. This is followed by a description of a cognitive model that makes predictions in a sample task used frequently in sleep restriction research. Math Models of Cognitive Throughput Research on the effects of sleep restriction also has resulted in a better understanding of how sleep and circadian rhythms interact to influence an individual’s ability to perform tasks. These findings have been incorporated into a variety of biomathematical models that are either commercially or publicly available (Mallis et al., 2004). A review of these models, their implementations, and their ability to predict novel empirical results was recently conducted and published (Aviation, Space, and Environmental Medicine, 75, 3, March 2004). All of the models produce some form of prediction of sleepiness or impairment in cognitive performance due to sleep loss, which is useful for quantifying the overall effectiveness of a person’s cognitive system relative to maximal. However, these models do not make performance predictions in specific task situations. For example a model may predict only that cognitive throughput is at 70%, leaving it unclear