The Interactive Visualization Gap in Initial Exploratory Data Analysis

Abstract

Data scientists and other analytic professionals often use interactive visualization in the dissemination phase at the end of a workflow during which findings are communicated to a wider audience. Visualization scientists, however, hold that interactive representation of data can also be used during exploratory analysis itself. Since the use of interactive visualization is optional rather than mandatory, this leaves a "visualization gap" during initial exploratory analysis that is the onus of visualization researchers to fill. In this paper, we explore areas where visualization would be beneficial in applied research by conducting a design study using a novel variation on contextual inquiry conducted with professional data analysts. Based on these interviews and experiments, we propose a set of interactive initial exploratory visualization guidelines which we believe will promote adoption by this type of user.

Beyond roll-up’s and drill-down’s: An intentional analytics model to reinvent OLAP

“…In a similar vein, in [63], the authors propose a taxonomy of user tasks in exploratory data analysis that include (1) discovery (hypothesis formulation and determination of the data source that can answer it) , (2) data acquisition (and preparation), (3) exploration of the data, (4) modeling, via the construction of a model that explains the data, and, (5) communication of the results to other people via reports and presentations. The discussed taxonomy is very close to the one presented in [62] , but similarly suffers from the high-level of abstraction for the exploration part.…”

B2

“…B2's goal of bridging code and interactive visualization is motivated by recent surveys and interviews of data scientists [12,14,17,57]. In particular, Wongsuphasawat et al find that data scientists often switch between several tools including textual environments (e.g., MATLAB or Jupyter) and graphical interfaces (e.g., Tableau or Microsoft PowerBI) during their analysis sessions [57].…”

“…For many data scientists, this overhead is sufficiently prohibitive that they eschew visual analysis tools altogether and restrict themselves to working only in code [57]. Indeed, Batch and Elmqvist identify that visualizations "should be first-class members of the analytical process so that actions and transformations interactively performed in the component can be exported and passed on to the next component in the sequence" [14] and Alspaugh et al call for new systems that combine the expressiveness of programming and scripting languages, with the efficiency and ease-of-use of visual analysis tools [12].…”

“…Surveys and interviews with data scientists, however, highlight the impoverished use of visualization within computational notebooks. In contrast to visual analysis tools such as Tableau (née Polaris [49]) or Microsoft PowerBI, which offer rapid or automated specification of visualizations and direct manipulation interactions to coordinate multiple linked views, visualizations in notebooks are largely manually specified single, static views [14,57].…”

“…Recent work suggests that although notebook users could benefit from richer support for interactive visualization, the friction of switching between the two paradigms remains too high. For example, although notebooks allow code cells and visualizations to be interleaved, these two types of artifacts remain siloed [14]. Code cells cannot access the results of interactive operations performed on visualizations -for instance, after brushing a region on a scatter plot, data scientists cannot extract the selected points for subsequent analysis.…”