SKILL TRANSFER RESEARCH: THE ISSUE OF PROCESS SIMULATION FIDELITY

NEWSLETTER ARTICLE

Simulator fidelity (the accuracy with which process simulators reproduce the appearance and behavior of the system being modeled) has surfaced as a major topic area in training circles throughout the petrochemical industry. Does having a high fidelity simulator mean the transfer of skills and knowledge will be enhanced? Will a high fidelity simulator that costs nearly ten times that of a lower fidelity simulator actually increase refinery operators ability to control the process? The answers to these questions are difficult to come by and may not be completely understood for a long time. However, recent research into the transfer of skills and knowledge from training to the real world environment indicates that the answer may be NO (An Informational Perspective on Skill Transfer in Human-Machine Systems, Human Factors, 1991, 33(3)).

The research indicates that the high fidelity approach to training is not always synonymous with enhanced skill transfer. Instead, data are available that show better skill transfer if the training task differs on specific dimensions from the actual real world task. That is, not all elements of the training tasks need to be exactly like those of the real world tasks for operators to "learn" what is being taught. For example, pilots perform better on crosswind landings if no crosswind is used in training. Instead, display system feedback is used to train the pilots for crosswind situations. Other studies have shown that by using part-task training in teaching pilots how to land on aircraft carriers, as opposed to whole-task simulation, skill transfer is enhanced (i.e., carrier landings are better). In each of these experiments the researchers deliberately made the training tasks partially different than the actual task.

The current research maintains that skill transfer can only occur when critical similarities between the training tasks and the actual tasks are maintained and that informational invariants constitute the properties that define those critical similarities. Informational invariants are defined as critical relationships that remain invariant while the operator is performing correctly, but change to different values when performing poorly. For example, with starting-up a process unit, one or more relationships will be invariant if the operator properly brings the unit up. Those relationships will change however, if the operators fails to follow normal procedures. It is informational invariance (which is a small subset of the total information available to the operator) that requires high simulator fidelity to enhance skill transfer, not aspects that are obvious and/or easy to understand.

Copyright 1992 Beville Operator Performance Specialists, Inc., All Rights Reserved