Using Signal Detection Theory to Examine the Effects of Workload and Likelihood Displays on Human Responses to Alarms

The purpose of this study was to examine the effects of workload and likelihood displays on human responses to alarms through the use of Signal Detection Theory (SDT). Thirty students performed compensatory tracking, resource management, and system monitoring tasks. They interacted with a Binary Alarm System (BAS) or a Likelihood Alarm System (LAS) during varying workload levels. Overall Response Rate (ORR), False Alarm Response Rate (FARR), True Alarm Response Rate (TARR), Sensitivity (d'), and Bias (c), were used as alarm response measures. Root Mean Squared Tracking Error (RMSTE) and Root Mean Squared Resource Management Error (RMSRMF) reflected primary-task performance. Proportion of Correct Resets (PCR) reflected secondary-task performance. As expected, participants responded significantly more often when they used the BAS during low workload. However, they responded significantly more often to true alarms when they used the LAS when compared to the BAS regardless of workload level. Participants were significantly more sensitive when they used the LAS, and they were significantly more biased during high workload. Last, participants performed the system monitoring task significantly better when they used the LAS. The main theoretical contribution of this study was to demonstrate the usefulness of adapting SDT to study human response to alarm signals. The main practical contribution of this study was to show the ability to enhance knowledge-based databases and fault diagnosis methods used by alarm systems by incorporating likelihood information as a fundamental aspect of their decision-making algorithms. Future research should focus on refining the methods used to develop LASs, which may ultimately serve as an adequate way to mitigate the cry-wolf effect.