| This study addressed a problem of human performance and system reliability for work settings characterized by supervisory control and high levels of automation (e.g., long-haul aviation, commercial trucking, etc.). Operational evidence indicates that extended periods of work underload, inherent in such human-machine systems, can induce short-cycle states of low alertness that erode the ability of operators to adequately cope with even well-trained tasks. A clear demonstration of this model has, however, been lacking in the research literature. The primary objective of this study was to generate evidence that could confirm or refute the existence of such a relationship and to provide insights into its dynamics. Subjects were given a supervisory control task--monitoring a radar in an air defense scenario--for four, two-hour sessions. They were required to query and classify each aircraft, and fire upon those determined to be enemies while letting friendly aircraft proceed. Task demands were minimal except during high workload trials involving large numbers of incoming aircraft. Three classes of measures presumed to index operator alertness--cardiac function, EEG power, and behavior (i.e., a secondary task)--were recorded during the one-minute of work underload preceding each trial, as independent variables. Five task performance measures were recorded as dependent variables during periods of high workload: changes in secondary task response rate, number of aircraft allowed through without processing, number of misclassified aircraft, frequency of repeated aircraft queries, and speed of task performance. Alertness and performance data were compared using cross-correlation methods. Significant relationships were identified for all performance measures, although not all alertness measures were equally sensitive. The primary effect of low alertness was a general reduction of cognitive resources, as evidenced by reduced performance on the secondary task. These effects were found to last over periods of several minutes, confirming the major hypotheses of the study. Results are discussed in terms of human performance models, operator strategies, and potential engineering interventions. |
