St. James Ballroom
A Statistical Design of Experiments Approach to Address Uncertainty in Human Error Modeling in Security Screening Operations (303831)
Robert Brigantic, Pacific Northwest National Laboratory
Jan Irvahn, Pacific Northwest National Laboratory
*Andrea Trevino Gavito, Northwestern University
Keywords: Design of experiments, Bayesian Networks, Security Screening, Human Error Modeling
In conducting security screening operations, humans play a vital role in successfully interdicting illicit objects from entering a secure space or venue. At the same time, humans can make inadvertent errors or deliberately violate procedures while conducting screening. In this poster presentation, we combine human reliability analysis and probabilistic graphical models, such as Bayesian networks, to characterize and quantify human factors and their impact on decision outcomes in security screening. Sensitivity analysis is conducted via statistical design of experiments (DOE) to assess model uncertainty. We estimate the relative influence of factors, as well as their effects and interactions providing a means to identify, visualize and communicate the most relevant variables in our model to organizational stakeholders. Moreover, we provide a general applied statistics framework that can be tailored for application to different venues where security screening is necessary, such as airports, stadiums, or international borders.