Alarm system trust - crying wolf versus missing the wolf

NEWSLETTER ARTICLE

The use of alarms generates two types of behavior: compliance and reliance. Compliance is taking an action when the alarm occurs. Reliance is the tendency to not take action until an alarm occurs, confident that the alarm system will warn of an impending problem. While the concepts are related, they appear to be independent phenomena (Meyer, J., Wiczorek, R., and Gunzler, T., Measures of Reliance and Compliance in Aided Visual Scanning, HUMAN FACTORS, Vol. 56, No. 5, August 2014, pp. 840 - 849).

The researchers varied both (1) the number of false alarms (called the positive predictive value, or PPV) and (2) the number of events for which no alarm actuated and the operator had to detect the fault (called the negative predictive value, or NPV). In one case, either 5% or 25% of the alarms were false alarms. In another case, 5% or 25% of the failures had no alarm associated with it. Fewer false alarms increased both reliance and compliance. The ability of the alarm system to detect most failures (only 5% of the events did not have an alarm) had no impact on compliance, but reliance on the alarm system decreased when the system failed to announce events.

Since we cannot be sure that we have alarms for all possible events, low reliance on the alarm system may be beneficial. This would further argue for fewer alarms, where those that are present are indicative of a need for operator response. In Beville-conducted alarm rationalizations, we argue that alarms should be used to avert serious consequences, while those resulting in process inefficiencies can and should be caught as part of the operator's normal surveillance and, therefore, not alarmed. This research adds credence to this approach.

Copyright 2015 Beville Operator Performance Specialists, Inc., All Rights Reserved