The effects of a dynamic graphical model during simulation-based training of console operation skill

LOADER is a Windows-based simulation of a complex procedural task. The task requires subjects to execute long sequences of console-operation actions (e.g., button presses, switch actuations, dial rotations) to accomplish specific goals. The LOADER interface is a graphical computer-simulated console which controls railroad cars, tracks, and cranes in a fictitious railroad yard. We hypothesized that acquisition of LOADER performance skill would be supported by the representation of a dynamic graphical model linking console actions to goal and goal states in the 'railroad yard'. Twenty-nine subjects were randomly assigned to one of two treatments (i.e., dynamic model or no model). During training, both groups received identical text-based instruction in an instructional-window above the LOADER interface. One group, however, additionally saw a dynamic version of the bird's-eye view of the railroad yard. After training, both groups were tested under identical conditions. They were asked to perform the complete procedure without guidance and without access to either type of railroad yard representation. Results indicate that rather than becoming dependent on the animated rail yard model, subjects in the dynamic model condition apparently internalized the model, as evidenced by their performance after the model was removed.