The Visual Search Environment: Are Two Heads Better Than One?

Abstract

Visual search tasks such as luggage screening are characterised by high error rates. Screeners are required to find illegal objects among distractors which are often difficult to tell apart. Different kinds of distractors amidst cluttered backgrounds, and the low prevalence of targets also impede the search process. Therefore, researchers suggest that adding a second screener to view the same images may improve the detection performance. After all, it is commonly assumed that two members of a team will be more effective than an individual working alone. However, the superiority of a two-member team over a one-member team is not always guaranteed because advisors (human or automation) are not perfectly accurate. Any over-reliance and over-acceptance of an inaccurate advisor’s choices in safety-critical situations can have catastrophic consequences.

Past researchers have identified that the dyadic relationship is affected by the perceived ability of the advisor. This perceived ability is demonstrated by the advisor’s perceived status and accuracy (i.e., performance). People tend to defer to people of higher perceived ability as they believe that such people are more trustworthy than someone of lower perceived ability. This deference is commonly seen in hierarchical organisations such as medical and aviation industries. However, what remains unclear is how this dyadic relationship is affected by external task characteristics. Safety-critical situations such as luggage screening and medical screening are especially susceptible to time pressures and multiple external distractions. These external task characteristics, coupled with the strong emphasis on error-free decisions, further compound the difficulty of attempting such visual search tasks. It therefore becomes imperative to consider how the dyadic relationship is influenced by the demanding environment, and how performance and agreements are impacted.

The present thesis was designed to investigate the effects of advisor’s perceived ability (e.g., status) and external (e.g., time pressure) variables on performance in a simulated luggage screening task. Participants were informed that they would be assigned to a human advisor and were given a choice to rely on their own decisions or to rely on their advisor’s decision. Studies 1A and 1B first examined the dyadic relationship and how participants were influenced by advisors of different statuses and performance. Studies 2, 3, and 4 were then designed to extend upon Studies 1A and 1B’s findings through the addition of external factors such as time pressure, accountability, and distraction.

Several interesting themes emerged from the present results. Participants were highly sensitive to the advisor’s performance rather than the advisor’s perceived status. However, this sensitivity did not lead to better accuracy and participants showed some over-reliance on the moderately accurate advisor. The present findings also suggest that prior-training on the luggage screening task may mitigate any negative effects of distraction and the low prevalence effect. Participants showed little to no awareness of the time pressure and accountability manipulations as well, suggesting that more research is needed to discover more effective manipulations. Overall, the present results show that screening performance may be improved with proper training and by increasing awareness of the screening task’s challenges. The present thesis provides a valuable starting point for incorporating more realistic task characteristics into research on the visual search environment.