Johanna Drucker - Visualization and Interpretation: Humanistic Approaches to Display

In the several decades since humanists have taken up computational tools, they have borrowed many techniques from other fields. This has included the appropriation of visualization methods to create charts, graphs, diagrams, maps, and other graphic displays of information. But are these visualizations actually adequate for the interpretative approach that distinguishes much of the work in the humanities? To answer that question, we have to define the features of that interpretative work and identify the ways its assumptions and premises are distinct from those of other research methods. We also have to ask what kinds of attitudes toward knowledge can be expressed with current conventions of visualization and consider alternatives. If, as will be argued here, the activity of modeling interpretation is fundamentally at odds with current methods of information visualization (data display), then how would such interpretative work be structured within a computational environment and a user interface? The challenges are many.

We can start by considering what is being referred to in the phrase information visualization and how it contrasts with that of modeling interpretation. The standard approach to information visualization is to generate a graphic from live or static data. The process is relatively straightforward. A set of quantitative values is charted on a grid, plane, or space governed by a regular, standard metric. Columns, bars, pie charts, points, lines, network diagrams, and so on are all produced in this way. The interpretative work of shaping the data disappears from view in the final result. The image displayed on screen, in print, or through other output devices appears as a statement of fact. The interpretative dimensions of the activity that shaped the data are rendered invisible, not so much concealed as simply missing from view, absent without a trace.

Any statistician or quantitative analyst will freely admit to the interpretative aspects of data creation (choices about sample size, parameterization, and other aspects of statistical and quantitative manipulations). Visualizations are not characterized, for instance, as remediations (translations of documents or objects from analog to digital format). They are not described as the outcome of interpretative parameters (methods of quantification) that translate numerical values into graphical ones. These methods of translation use conventions that may or may not be logically consistent with the display of quantitative information (the scale or measure may have been altered). And often, the process of data creation produces results that are far from the original phenomenon from which it was extracted. But these processes of transformationfrom phenomenon to data and then to displayare rarely documented or noted. Instead, the output is simply called a data visualization, as if the process were unproblematic and as if the data and visualization were one and the same. In general, this is a one-way process, with visualizations generated from (that is,