Cognitive Architectures and the Driving Story
It is important to be able to put these architectures into context and to consider the places where they may be useful. Thus it may be useful to reflect on a few questions related to the story introduced at the beginning of this chapter. How might we use an architecture to model and simulate the effects of changes of the autopilot system on driver behavior and performance? How could we study the way sleep moderate this interaction? Which features of the interaction should we focus on (e.g., should we focus on the ways the autopilot alerts the user that it is enabled, or should we focus on the way we deliver the directions for use of the autopilot feature.) Perhaps the most pressing question involves the choice in our architecture for simulation of these interactions. Indeed, our answer to the previous questions would affect which architecture and simulation method we choose to employ. Consider how much money we can save by not paying multiple users to test out multiple instantiations of system designs through the design cycle!
Though modeling and simulation using cognitive architectures has clear advantages, collection of user data in the design of the system is still important wherever possible. One aspect of data collection that will increasingly complement using cognitive architectures in simulation is the collection of psycho-physiological measures. Reflecting on the story at the beginning of this chapter, an understanding of the ways that changes physiological measures related to sleepiness interact with the decision to run the autopilot in a risky manner may have helped us design a system that did not allow such an easy path to the risky behavior.