What does an operator need to learn?

Using a simulated geosynchronous satellite relocation task, three types of training schemes, namely, in-the-loop, out-of-the-loop, and a composite of these two methods were evaluated. Verbal protocols in addition to performance and strategy measures were used to understand learning in this complex task. The results point toward an amplitude hypothesis of learning where two distinct phases are evident. In the first, large amplitude fluctuations exist due to the lack of a good mental model of the system dynamics. In the second, the amplitude fluctuations are low, and the performance improvements are dramatic suggesting the end of the mental model development phase and a gradual improvement in the system optimization parameters leading to the traditional power law learning curve. Based on the results, it may be concluded that to learn a system or process well, the operator needs to: 1. Develop a good mental model of the system dynamics to minimize the large fluctuations in performance, and 2. Understand the optimization criteria to improve performance with low amplitude variations.