A Model of False Alarms in Target Acquisition by Human Observers.

Abstract : The modeling of target acquisition by human observers has focused on the problem of predicting whether targets will be detected. The closely related issue of false alarm prediction has received less attention. While predicting false alarms is secondary to true detection, it is nevertheless important to understand the effects of false alarms and to account for them in the development of doctrine. In this work we extend the scope of target acquisition modeling to the consideration of false detections. The model is based on the analysis of data obtained in a series of target acquisition tests. It is phenomenological in the sense that it seeks only to describe the results of the tests. An important finding from the analysis of the test data is that the dominant determinant of false alarm rate is the expectation of the human subject. A more general review of the test results reveals features that strongly suggest a description based on signal detection theory. Re-analysis of the test data in this context allows us to construct such a description and to extract the parameters that describe the observer ensemble. Finally, we demonstrate the correlation between the mean false alarm rate and a scene complexity statistic.