Learning and Cognitive Modeling in Real-Time Dynamic Decision Making (Research Seminar, January 31st, 2002) F. Javier Lerch Director, Center for Interactive Simulations GSIA, Carnegie-Mellon University Abstract: A series of experiments were conducted to investigate learning in a dynamic, real-time task environment. We expected that time pressure and high workload would inhibit learning in participants with low Working Memory capacity and/or low fluid intelligence. We also expected that participants would improve performance by acquiring knowledge about what decisions work well under specific task conditions (learning situation-response instances), rather than learning to apply decision rules more closely and consistently. Our results indicate that the negative effect of time pressure and high workload was mainly on individuals with low fluid intelligence. Our results also indicate that decision makers improved performance by learning to follow rules less closely and more inconsistently. These results suggest that the main learning mechanism in dynamic, real-time environments is instance-based learning, and that the acquisition and refinement of instances require spare cognitive resources during task execution.