Understanding representations of uncertainty, an eye-tracking study – Part 2: The effect of expertise

As the ability to make predictions regarding uncertainty information representing natural hazards increases, an important question for those designing and communicating hazard forecasts is how visualizations of uncertainty influence understanding amongst the intended, potentially varied, target audiences. End-users have a wide range of differing expertise and backgrounds, possibly influencing the decision-making process they undertake for a given forecast presentation. Our previous, Part 1 study (Mulder et al., 2023) examined how the presentation of uncertainty information influenced end-user decision making. Here, we shift the focus to examine the decisions and reactions of participants with differing areas of expertise (meteorology, psychology, and graphic-communication students) when presented with varied hypothetical forecast representations (boxplot, fan plot, or spaghetti plot with and without median lines) using the same eye-tracking methods and experiments. Participants made decisions about a fictional scenario involving the choices between ships of different sizes in the face of varying ice thickness forecasts. Eye movements to the graph area and key and how they changed over time (early, intermediate, and later viewing periods) were examined. More fixations (maintained gaze on one location) and more fixation time were spent on the graph and key during early and intermediate periods of viewing, particularly for boxplots and fan plots. The inclusion of median lines led to less fixations being made on all graph types during early and intermediate viewing periods. No difference in eye movement behaviour was found due to expertise; however, those with greater expertise were more accurate in their decisions, particularly during more difficult scenarios. Where scientific producers seek to draw users to the central estimate, an anchoring line can significantly reduce cognitive load, leading both experts and non-experts to make more rational decisions. When asking users to consider extreme scenarios or uncertainty, different prior expertise can lead to significantly different cognitive loads for processing information, with an impact on one's ability to make appropriate decisions.