Parts Measured, Persons Missed: A Profile-Centered Account of Visualization User Experience
Authors
Anjana Arunkumar (Northeastern University), Lace M. Padilla (Northeastern University)
Presentation
- Session
- Data, Meet Human: Vis That Cares
- Time
- Wednesday, Nov 11, 10:18 – 10:27 (US/Eastern) · session 10:00 – 11:30
- Location
- Hall America north
Keywords
Latent profile analysis, individual differences, visualization user experience
Abstract
Studies have made steady progress identifying individual differences that predict how people use charts, such as graph literacy, spatial ability, and need for cognition. Yet, these measures rarely account for more than a fraction of the outcome variance, and their utility varies considerably across viewers and contexts in ways that are difficult to explain. A potential reason for this variability is that studies often examine these traits as independent or additive predictors, not accounting for how much they co-vary, interact, and suppress one another. We argue that a visualization interpretation is a configuration of how participants perceive, attend to, and recall visual information, and that it is not recoverable from any single dimension or their sum. To test this claim, we use latent profile analysis (LPA), developed in psychology for this purpose, on a dataset of 100 real-world communicative visualizations (n = 311) across measures of perceptual orientation, attentional engagement, and chart recall. The LPA revealed four visualization interpretation configurations, which we contextualize as viewer profiles: the Analytical Engager, Aesthetic Appreciator, Affective Integrator, and Detached Processor, which are substantially better predictors than individual differences. This work illustrates a more comprehensive approach that offers a richer account of the visualization user experience, with implications for how we design studies, evaluate visualizations, and build systems that adapt to viewers.
For Practitioners
This work is most relevant to practitioners who design visualizations for broad, non-captive audiences: data journalists and news graphics teams, government and public-health risk communicators, and UX researchers evaluating charts and dashboards in industry. The four viewer profiles give these practitioners a concrete audience taxonomy: rather than designing and evaluating for an "average" reader, they can design defensively for Detached Processors (redundant annotation, reduced complexity), add narrative anchoring for Aesthetic Appreciators, and hook topic relevance to chart content early for Affective Integrators. For evaluators, the results caution that aggregate metrics obscure qualitatively different audience responses; segmenting evaluations by engagement pattern yields more actionable results. The proposed short classification instrument offers a practical path to profile-aware design, evaluation, and viewer-adaptive systems.