Deciding when to switch tasks in time-critical multitasking

While cognitive modeling has begun to make good progress in accounting for human multitasking behavior, current models typically focus on externally-driven task switching in laboratory-task settings. In contrast, many real-world complex tasks, particularly time-critical ones, involve internally-driven multitasking in which people themselves decide when to switch between tasks. In this paper, we propose an adaptation of the ACT-R cognitive architecture that incorporates a notion of elapsed time for the current goal and uses time to determine when to switch away from the current task. We demonstrate the usefulness of this mechanism in an application to a dynamic, time-critical dual search task, showing how an ACT-R model can account for various aspects of human subjects' multitasking behavior.