Look Again: An Investigation of False Positive Detections in Combat Models; Master's thesis

摘要

This thesis investigates the role of false positive detections in simulated combat environments. Existing combat models tend to overlook or downplay false positive detections. Signal Detection Theory provides the framework for analysis of an observer's hits, misses, correct rejections, and false alarms. The experimenter hypothesized that false alarm rates are a function of observer experience, task instructions, and scene difficulty. In support of this thesis, the researcher developed 24 computer images containing varying numbers of human combatants in an urban environment. Sixteen students at the Naval Postgraduate School volunteered as observers for this experiment. Experimental results revealed that the factors significantly affecting false alarm rates were scene difficulty, task instructions, and the interaction of these two factors. Observer experience was not shown to be statistically significant. Observers given permissive instructions generated up to 3.5 times as many false alarms as did those given restrictive instructions. This experiment showed that the practice of modeling false alarms solely as functions of target and scene characteristics is inadequate. With respect to the generation of false alarms, future combat models also must incorporate an assessment of the instructions given to the observer.