ACT-RLearninginaReal-TimeDynamicDecision-MakingTask.docVIP

ACT-R Learning in a Real-Time Dynamic Decision-Making Task Cleotilde Gonzalez Center for Interactive Simulations Carnegie Mellon University Pittsburgh, PA 15213conzalez@andrew.cmu.edu Christian Lebiere Department of Psychology Carnegie Mellon University Pittsburgh, PA 15213cl@andrew.cmu.edu F. Javier Lerch Center for Interactive Simulations Carnegie Mellon University Pittsburgh, PA 15213fl0c@andrew.cmu.edu  We describe an ACT-R model of performance and learning in a real-time dynamic decision-making task. The task, called Pipes, is a simulation of a water distribution system. This simulation is an abstraction of a real-world resource management task performed in an organization with large-scale logistical operations (The United States Postal Service) (Lerch, Ballow and Harter, 1997). The system consists of a total of 23 tanks connected by pipes and controlled by pumps. Water arrives at pre-assigned intervals in the leftmost tank (source) and is to be pumped from left to right through a tree of tanks into the rightmost tanks (factories). Each factory has an associated deadline by which a certain number of water buckets must be delivered. To distribute the water, each tank has two pumps associated with it that can be turned on and off, with the restriction that at most five pumps can be active at any time. The goal of the experiment is to schedule the pumps to minimize the number of missed buckets over the course of the experiment. This task is a real-time dynamic decision-making task because: It requires a series of decisions rather than a single decision. The decisions are interdependent. Each decision affects all the following decisions. The environment changes continuously, both autonomously and from subject actions. Decisions need to be made in real time, at the correct time. In this experiment subjects were assigned to either of two situations that varied the speed at the simulation changed and thus the time pressure on the subjects. Since working memory