Tapendra Pandey¹ (pandey@ou.edu), Ghulam Jilani Quadri¹; ¹University of Oklahoma

Individual differences in education, profession, visualization literacy, and familiarity shape natural and uncued interpretations of visualization, called high-level comprehension (HLC) (Quadri & Szafir, 2022; Quadri et al., 2023). However, how individual differences influence perception and comprehension of visualization remains unclear. Our empirical study offers a broader perspective on how they can guide and predict HLC. We propose a three-level hierarchical framework categorizing information depth of HLC: Basic Graph Knowledge(BGK; reading graphical elements), Conceptual Graph Understanding(CGU; capturing intended message), and Contextual and Interpretative Insights(CII; combining inference and reasoning), defined based on classifications provided by (Friel et al., 2001; Lundgard & Satyanarayan, 2022) and maps them to individual cognitive profiles. We recruited 20 participants from diverse backgrounds (9 professional domains, 4 education levels, 4 visualization familiarity groups) who completed two studies: (a) miniVLAT test (Pandey & Ottley, 2023) yielding literacy scores (mean = 8.2, SD = 1.7), and (b) main study focused on eliciting their understanding across 16 visualizations (visualization(4)× data(2)). Using open coding, two independent coders categorized 320 responses from (b) into three levels, ensuring high inter-rater reliability, revealing 15% BGK, 68% CGU, and 17% CII level responses. We calculated weighted comprehension-level scores, with 5%, 65%, and 30% of participants dominating at each level. Given the diversity of responses, participants reported nine statistical task results (Amar & Stasko, 2005) and often added more. High school students with frequent exposure to visualizations mapped to the same levels as students with STEM advanced degrees. Similarly, participants with identical literacy test scores showed varying levels of HLC. Individual differences showed complex interactions, and findings suggest standard literacy tests are insufficient to determine comprehension levels; i.e., interactions among all four shapes comprehension. Our resulting framework enables visualization designers to create visualizations that account for diverse cognitive profiles that emerge from a combination of individual differences.