| Hybrid inspection systems, those in which humans and computers work cooperatively, have found increased application in today's complex manufacturing and service systems, as evidenced by their widespread use in inspection applications (e.g., printed circuit boards, aircraft components, consumer products and baggage checks in airports). While technological and economic imperatives have driven designers to automate whenever possible, knowledge of how operators and computers interact lags far behind. In particular, it is not yet completely understood how certain factors influence an operator's control decisions (that is, choosing between manual and automatic control). It is known that trust is an important factor in determining such behavior as it can directly impact function allocation decisions. If, for example, an operator overrides the automation too frequently or is too hesitant to take manual control, system performance will be compromised. Clearly, in such environments, the operator's moment-to-moment allocation of functions is a critical decision-making process, one that is important to understand and optimize. In order to improve inspection performance, this research addresses the issue of trust within the context of a manufacturing inspection task. It explores this dimension to establish a clear understanding of the term from both engineering and sociological perspectives. In addition, it reviews and evaluates current approaches related to the measurement of trust in automation in general, and hybrid inspection systems in particular. Specifically, this research develops a model of human trust for hybrid inspection systems using a quantitative approach that relates machine properties to an operator's perceptions and the consequent perceptions to decision-making and control actions.;The proposed framework was validated empirically in two studies that integrated and systematically varied a set of dimensions relating to hybrid inspection system errors. The results of these studies revealed that human trust is directly related to error characteristics: the location, severity, number, and, most importantly, uncertainties associated with system error. The results allow researchers and designers to predict human trust in automation based on pure quantitative measures, providing a better measure of human trust than the traditional qualitative dimensions. Although the synthetic task environment employed is in the domain of visual inspection, the approach outlined is not specific to this context. Moreover, the inspection task was relatively complex with substantial cognitive content, and as such, the results are transferable to other domains, such as anti-air warfare, air traffic control, and other complex manufacturing systems, providing a broad base of applications for both the theoretical human factors researcher and the practitioner. |
