Effects of false alarms and misses on reliance, compliance, and attention when using an automated warning system

Automation support has been integrated into many complex and dynamic systems as a way to enhance human performance and decrease operator workload. How the human operator uses the automation is related to their trust in the system. Trust in automation is often compared to trust in human-human relationships. Two concepts that are components of this trust are reliance on, and compliance with, the automation.; The three studies discussed in this paper examined the effects of automation false alarms and automation misses on task performance, reliance and compliance. Studies one and two utilized the MAT battery, and study three used the X-Plane flight simulator to test performance, reliance and compliance. Study two also used eye tracking to examine attention allocation.; The first study found that the automation improved performance over baseline, but did not find that the rate of false alarms or misses affected reliance and compliance on the automation. Sensitivity seemed to have more of an effect on performance than did the level of false alarms or misses, and also effected compliance.; The second study found that workload had more of an effect on reliance and compliance than did the false alarm and miss rate. Compliance increased with low workload, and reliance increased with high workload. There were no significant differences found in the amount of attention allocated to the areas of interest based on false alarm or miss rate.; The third study used 'experts' to perform the flight task. All participants were certified pilots. The third study found that the false alarm rate had an effect on compliance with the automation, with slowing response times to automated false alarms, and the miss rate affected reliance on the automation, resulting in a decrease in reliance on the automation with a high rate of automated misses.; The 'experts' who participated in the study showed more effects of false alarm and miss rates on reliance and compliance than did the 'novice' participants from studies one and two. This suggests that training and experience may change how the operator uses the automation, and calibrates their trust in the automation. The implications of this are discussed in the document, along with future research directions.