Session IIa Report //// Information Visualization, Data Art/Design, Data Journalism: Revealing Hidden Worlds
This panel was a set of talks each predicated on a set of three assumptions:
1) that the visualization of data results in a clearer understanding of that data, but 2) only when it’s done “well”, so then 3) the visualization must be done in a way that accurately represents the data. Maneesh Agrawala began by showing a number of examples of charts where a redesign or supplemental information might improve the clarity of the piece. He used S.S. Stevens’ 1946 “On the Theory of Scales of Measurement”, an essay in which the psychologist outlines three types of data — nominal, ordinal, quantitative — used by Jacques Bertin in his 1967 evaluation of various visualization techniques. Eventually, Maneesh showed results from ReVision, a tool that he and his colleagues are developing at Stanford that seeks to supplement opaque graphs and charts by extrapolating, extracting, and overlaying data.
Jeff Heer spoke next, highlighting the work that he and his team have done to increase the interactive nature of data representations — something that he argues allows for better realization of multivariate patterns. Quizzing the audience to test our ability to evaluate the area of a shape on the fly, he revealed that bar charts are much more conducive to this task than circular pictograms or pie charts. Jeff wants to consider how we might “support more effective data exploration”, calling out data quality issues, confirmation bias, and other cognitive biases as hampering these efforts. He ended his talk by giving a demonstration of Voyager, a project that he and his colleagues have been developing to allow for deeper-dives of datasets via interactive visualizations.
Alyssa Goodman finished off the panel by introducing Glue, a package written in the Python programming language by her team at the Harvard-Smithsonian Center for Astrophysics. Described by its developers as, a “library to explore relationships within and between related datasets,” glue enables users to parse and present complex datasets with relative ease. Alyssa’s talk was themed around the oscillation between efforts to “explore” and “explain” data; she argued that the movement from explain to explore is much harder than going the opposite way. Glue, she suggested, directly aids in this transition, as it brings linked visualizations and multidimensional data exploration without significant knowledge of software development.
In all, the panel presented a thorough overview of what it means to put together visualizations that guide your viewer through the story you’re seeking to tell. Where I felt it lacked, however, was acknowledging — or even making room for — the fact that you (the scientist, designer, artist, lab assistant, etc.) are the one telling a story. Maneesh, for instance, left me wondering if he and his team considered the cultural meaning embedded within the charts that they claim ReVision can “classify, analyze, and redesign.” Color, for instance, has a wide range of meaning across various cultures. By reducing visualizations down to the data which drive them, we risk jettisoning the fact that authorship is embedded within the graphic. I felt similarly frustrated by Jeff’s assertion that “neuroscience really needs to get better at understanding how humans interpret area,” — another example of reducing reading and interpretation to a set of electrical transmissions in the brain, rather than a process highly influenced by one’s own experiences and background.
The visualization of data is the same process taken on by an author of a novel or a director of a film. Plot is arranged in such a way to make a case for an argument, shots are framed to inspire affect, data is visualized to prove a point. To suggest that there is a way to present data without attempting to make that point is to suggest a director does not consider their movie-goer’s gaze. I hope that a discourse surrounding data visualization that relies on the form’s seeming objectivity is not so dominant as to forget this.
