An examination of complex human-machine system performance under multiple levels and stages of automation

The introduction of automation into highly complex systems has occurred under several guiding principles. The application of these principles has often resulted in tenuous interactions with regard to human performance within complex systems. With advances in technology increasing at an exponential rate it is no longer applicable to look at single automated tools but rather at how several automated tools fit together and affect system performance. Automation can change the nature of the demands on the operator and produce subsequent changes in performance not seen when automation is absent. Problems in human-automation interaction have included unbalanced workload, reduced system awareness, decision bias, mistrust, over-reliance, complacency, and reduced manual skills. Further, these problems can be exacerbated when the automation is less than perfectly reliable.; A common framework utilizing a model of human interaction with automation based on simple human information-processing stages was used in the design and analysis of four experiments. The model was used for tasks that varied in complexity and the amount of automation that was available to the operator. The first three experiments utilized a visual search paradigm and varied the stage the automation was present and the reliability of the automation that was used. For these studies, the automation that helped the operator locate the potential target (information automation) demonstrated a clear advantage over automation that recommended a course of action (decision-aiding automation) when the automation was perfectly reliable. Costs associated with unreliable automation generally were greater for the information automation stage, which was not congruent with the results of previous research.; The fourth study examined all of the possible combinations of manual and automated aiding for the four stages in an air-to-ground search and destroy mission that was carried out in a high fidelity combat flight simulator. By utilizing separate stage metrics, it was demonstrated that the automation in one stage influenced performance in subsequent stages and throughout the entire mission. These benefits were apparent in the primary task performance and the subjective ratings of mental workload, situation awareness, and trust in the automation.