Content area
Full Text
Abstract
Adaptive automation systems have been proposed to automatically vary the level of automation based upon the user's workload. Towards this end, it is desirable to understand the relationships among four classes of information: 1) objective measures of workload, 2) user performance, 3) human physiology measures, and 4) user subjective ratings of workload. This paper proposes a framework for exploring the relationship among these variables, a method for utilizing these relationships to form a control signal for adaptive task allocation, and explores an existing data set to investigate the relationship. We suggest that objective workload metrics should be utilized to control the allocation of tasks to the human user, but that performance and physiology could be employed to modify task allocation. Within the data analysis, it is hypothesized that high correlations would be observed between objective workload and cardiac measures when the participants approach redline. This hypothesis was not strongly supported by the current analysis. We did find support for the premise that subjective workload and performance vary among participants and that objective workload can vary based upon the participants' ability to complete tasks in a timely fashion, providing one path for performance feedback to an adaptive automation system.
Keywords
Workload, Objective Workload, Performance, Physiology, Individual Differences
1. Introduction
Adaptive automation systems have been proposed which utilize changes in operator performance and physiological state to trigger changes in the level of automation [1, 2]. Although, previous research indicates that changes in each measure can correlate with changes in subjective workload across pools of participants in manned flight [3] and more recently in desktop applications [4], the relationship among these measures has not been robustly explored for individual participants [5]. As the level of automation must be adapted for individual operators, the understanding of the relationship among these variables derived while treating participants as a group may not be sufficient. Methods to quantify the relationship among these variables to account for individual differences will likely be necessary. Towards this end, the current paper proposes a method for exploring individual differences in the relationship among performance, physiological response, objective workload, and perceived workload.
To explore these relationships, the framework, depicted in the causal loop diagram of Figure 1, is proposed. This figure illustrates assumed, simplified, relationships...