Abstract
Making an evacuation decision for a hurricane requires one to weigh the probability of the hurricane hitting their city and the laborious nature of actually evacuating. Probablistic information about hurricane trajectory is often displayed via diagrams known as “cones of uncertainty”: a cone which represents the area that could be affected by any of the probable paths the hurricane could take, where variability in the computer models’ predicted locations of the storm is represented by the width of the cone. However, this design relies heavily on a forecaster explaining relevant aspects of the display as well as the potential evacuee’s understanding of variability (Broad et al., 2007). This design then requires evacuees to keep in mind their understanding of variability and the explanation given about the cone’s meaning in order to interpret; given of course, that either of these two things are/were present. In order to bolster understanding of, and overcome poor prior knowledge about, variability we created diagrams that consisted of a series of dots that represented places which the hurricane could touch down over time. These plots intend to leverage the Gestalt principle of proximity whereby, as variability increases, so does distance between dots. To test the efficacy of these plots a within-subjects experiment was conducted in which participants completed two blocks of questions regarding probability and variability while viewing either cone of uncertainty plots or our plots. The order of the blocks was counterbalanced between participants. Participants answered more questions correctly while viewing our plots (M= 0.42) than when viewing cones of uncertainty (M=0.34), t =2.0, p= 0.042, d=0.2. Generally, these results suggest that comprehension of visual displays is bolstered when the displays are designed in a way that parallels information we are attempting to display with basic principles of how visual perception operates.