Using Visual Representations of Data to Enhance Sensemaking in Data Exploration Tasks

Abstract

This paper explains how visual representations of data enable individual sensemaking in data exploration tasks. We build upon theories of human perception and cognition, including Cognitive Fit Theory, to explain what aspects of visual representations facilitate sensemaking for the viewer. We make three primary contributions. First, we give a general characterization of visual representations that would be used for data exploration tasks. These representations consist of a scene, objects within the scene, and the characteristics of those objects. Second, we extend Cognitive Fit Theory into the data exploration task domain. We explain that the data exploration task has a number of spatial subtasks including observing data points, looking for patterns or outliers, making inferences, comparing observed facts or patterns to one’s own knowledge, generating hypotheses about the data, and drawing analogies from the context being observed to another context. Third, we offer a set of theoretical propositions about how visual representations of data can serve the sensemaking goal. Specifically, visual representations best facilitate sensemaking in data exploration tasks when they (1) support the four basic human visual perceptual approaches of association, differentiation, ordered perception, and quantitative perception, (2) have strong Gestalt properties, (3) are consistent with the viewer’s stored knowledge, and (4) support analogical reasoning. We propose that visual representations should possess several of these four aspects to make them well-suited for the task of data exploration.