Measuring and maintaining performance in x-ray baggage inspection at security checkpoints: methodological and practical considerations

Inspecting X-ray images of passenger baggage for prohibited items at security checkpoints is crucial to ensuring aviation security. To prevent performance declines during inspection, the EU allows screeners to perform this task for only 20 min, although little is known about how this performance actually evolves over time. For many airports, longer screening durations would be practical, and this raises the question of the ideal screening duration in terms of both performance and screener well-being. To measure screeners’ performance, airports typically implement threat image projection (TIP). TIP projects fictional threat items (FTIs) onto the X-ray images of passenger baggage; and by recording the screeners’ decisions, it allows measurement of their detection rate. To draw meaningful conclusions from these data, it is essential for them to be reliable and valid.However, their reliability and validity are still poorly researched and not confirmed. This thesis addresses the question of how time on task affects performance in X-ray image inspection of cabin baggage, and it asks whether TIP performance data collected at airports provide a reliable and valid measure of operational threat detection. Manuscript 1 investigated how performance evolves with time on task in two groups of screeners who performed a 1-hr X-ray image inspection task in the laboratory. One group took 10-min breaks every 20 min; the other group screened continuously without breaks. To assess the validity of measurements of detection performance, we varied target prevalence. Results confirmed the typical target prevalence effect and showed that da is a valid measure of detection performance for X-ray images inspection. Manuscript 1 provides evidence that screeners were able to maintain performance for a full hour, and that breaks had no effect on performance. However, time on task caused a shift in response tendency and might cause more distress. In Manuscript 2, we investigated the effects of time on task on performance under real working conditions by analyzing performance data from a 4-month field study. A group of screeners at a European airport were asked to analyze X-ray images from a remote screening room for up to 60 min. Only when task load was high (number of images analyzed per min), did the screeners' hit rate decrease with time on task. The efficiency, in terms of the reject rate and processing time, increased with time on task. Screeners who conducted longer screening durations did not report more distress. Yet, there were marked individual differences in performance, in performed screening durations, and in preferred screening durations. In Manuscript 3, we examined the reliability and validity of TIP performance by analyzing a large data set from a European airport. We showed that TIP data can be a reliable and valid measure of operational threat detection, and that around 100 TIP events per screener should be considered to attain minimum reliability values of 0.7. The manuscript further provides recommendations on how to increase the reliability of TIP data.Taken together, these findings show that TIP data, which are in frequent use, can provide a reliable and valid measure of operational threat detection and that screeners can maintain performance for more than 20 min. Manuscripts 1 and 2 provide evidence that time on task in X-ray image inspection leads to a shift in response tendency rather than a decline in sensitivity. Based on performance and survey results, screening sessions could be designed more flexibly and an extension to 30–40 min could be considered. The manuscripts provide meaningful theoretical insights into performance in X-ray image inspection, especially with regard to the effect of time on task. They further provide methodological and practical contributions on appropriate detection performance measures, on how to measure performance reliably and validly, and on the design of screening durations.