Abstract
This study explored the impact of people’s working memory capacity (WMC) on COVID-19 real data visualization (DV) processing when DV is represented (a) as counts versus percent growth, and (b) with COVID-19 related versus unrelated seductive details versus no details. 70 undergraduates of US university explored six stimuli each comprising two DVs, one question, and four answer options. Additionally, they completed ACCES test of reading span (rspan), visuospatial span (sspan), and operations span (ospan). The following novel approach was proposed that leverages eye movement data for understanding people’s visual processing. First, gaze-maps were created as color-based probability distributions of people’s eye fixations on regions for each stimulus. They were saved as images and clustered with Agglomerative algorithm of unsupervised machine learning. Two, three, and four hierarchical clusters were identified as different types of people's attention distributions. The examination of images in two-cluster model revealed that people’s attention was relatively more distributed around DVs and less around verbal sections in cluster 1 as opposed to cluster 2. More clusters specified around which DVs and verbal areas attention was relatively more distributed. Second, multiple mixed-effects logistic regression analyses were performed with rspan, sspan, and ospan scores as predictors of the clusters for different DV representations. The findings indicated that people's WMC was significantly associated with different clusters of attention distribution. Specifically, there were higher chances that people with higher visuospatial WMC had their attention distributed around (a) DVs, when they were represented with COVID-19 related seductive details, or data was represented as counts, and (b) verbal areas, when DVs were represented with COVID-19 unrelated seductive details. On the contrary, people with higher operational WMC had their attention distributed around verbal areas, when data was represented as percent growth. The proposed method motivate designing personalized technologies sensitive to the differences in users’ WMC.