A human subject experiment was conducted to investigate caffeine’s effect on appraisal and performance of a mental serial subtraction task. Serial subtraction performance data was collected from three treatment groups: placebo, 200, and 400 mg caffeine. The data were analyzed by caffeine treatment group and how subjects appraised the task (as challenging or threatening). A cognitive model of the serial subtraction task was developed. The model was fit to the human performance data using a parallel genetic algorithm. How the model’s parameters change to fit the data suggest how cognition changes due to caffeine and appraisal. Overall, the cognitive modeling and optimization results suggest that the speed of vocalization varies the most along with changes to declarative memory. This approach provides a way to compute how cognitive mechanisms change due to moderators.

The success of simulation environments will depend partly on how realistically the models mimic human behavior. While human behavior is affected by various moderators (see Pew and Mavor [28], for an initial list), cognitive models typically do not take into account the effects of many of these moderators. We propose that cognitive models can be augmented to account for such effects by modifying either their knowledge or the parameters of the architecture that they are built with. To provide an example of the two ways in which cognitive models can be modified to capture the effects of behavior moderators, we present an ACT-R model that performs a cognitive task while being affected by the moderators of as anxiety and pre-task appraisal. These changes are validated in a preliminary way by comparison with human data, which shows us where these models can be improved and provides lessons for further work. Most importantly, we argue that more realistic models of human behavior reflecting these moderators and individual differences can be achieved by implementing similar modifications within other cognitive models and by reusing these modifications for an existing architecture as an overlay.

We describe a range of theories of how cognition is influenced by stress. We use a cognitive architecture, ACT-R, to represent these theories formally. The theories make suggestions for developing cognitive architectures, in that they nearly all of them require that time-on-task influence performance, and at least one suggests that workload and strategies are monitored to access and cope with stress. By examining the theories as a whole, we can also see that they are incomplete in that individually and as a group they do not make predictions that are consistent with data. For example, many of them do not predict that repeated serial subtraction (part of the Trier Social Stressor Task) will be affected by stress. We also see how the stress theories and the mechanisms that give rise to them can be tested. Acknowledgements

Simulating the effects of behavior moderators within a cognitive architecture is essential for building cognitive models that can realistically capture the full range of human performance. We demonstrate that some of these effects can be modeled by varying parameters of the cognitive architecture and some by modifying the knowledge that is built into the models. As an example of implementing the two approaches, we present an ACT-R model that performs serial subtraction under varying levels of task-appraisal and with and without anxiety realized as worry.

dental13.doc

Treating simulations as theories by not sampling their behavior

Protocol for an experimental investigation of the roles of oxytocin and social support in neuroendocrine, cardiovascular, and subjective responses to stress across age and gender