Unpacking Visualization Rhetoric

Note: An edited version of the following also appears on the Chart.io blog. 

Visualization can be useful for both more exploratory purposes (e.g. generating analyses and insights based on data) as well as more communicative ends (e.g. helping other people understand and be persuaded or informed by the insights that you’ve uncovered). Oftentimes more general visualization techniques are used in the exploratory phase, whereas more specific, tailored, and hand-crafted techniques (like infographics) tend to be preferred for maximal persuasive potential in the communicative phase.

In the middle ground is a class of visualizations termed “narrative visualization” – often used in journalism contexts – which tend to include aspects of both exploratory and communicative visualization. This blending of techniques makes for an interesting domain of study and it’s here where Jessica Hullman and I began investigating how different rhetorical (persuasive) techniques are employed in visualization. We were particularly interested in how different rhetorical techniques can be used to affect the interpretation of a visualization – valuable knowledge for visualization designers hoping to influence and mold the interpretation of their audience. (Here we defer the sticky ethical question of whether someone should use these techniques since in general they can be used for both good and ill).

We carefully analyzed 51 narrative visualizations and constructed a taxonomy of rhetorical techniques we found being used. We observed rhetorical techniques being employed at four different editorial layers of a visualization: data, visual representation, annotations, and interactivity. Choices at any of these layers can have important implications for the ultimate interpretation of a visualization (e.g. the design of available interactivity can direct or divert attention). The five main classes of rhetoric we found being used include: information access (e.g. how data is omitted or aggregated), provenance (e.g. how data sources are explained and how uncertainty is shown), mapping (e.g. the use of visual metaphor), linguistic techniques (e.g. irony or apostrophe), and procedural rhetoric (e.g. how default views anchor interpretation).

The maxim “know thy audience” points to another dimension by which a visualization creator can influence the interpretation of a visualization. While most visualizations concentrate on the denotative level of communication, the most effective visualization communicators also make use of the connotative level of communication to unlock a whole other plane of interpretation. For instance, various cultural codes (e.g. what colors mean), or conventions (e.g. line graphs suggest you’re looking at temporal data even if you’re not) can suggest alternate or preferred interpretations.

While the full explanation of the taxonomy and use of codes and connotation for communication in visualization is beyond this blog post, you can see a more complete discussion in a pre-print of our forthcoming InfoVis paper.  At the very least though I’ll leave you with an example which illustrates some of these concepts.

Take the following recent example from the New York Times where various aspects of the visualization rhetoric framework apply.

The choice of labeling on the dimensions of the chart “reduce spending” vs. “don’t reduce spending” leaves out another option, “increase spending”. The choice of the color green for “willing to compromise” connotes a certain value judgement (i.e. “go, or move ahead”) as read from an American perspective. The way individual squares are aggregated to arrive at an overall color is unclear, leading to questions that could be clarified through better use of provenance rhetoric. Moreover, squares cannot be disaggregated or understood as individual data, making it difficult for users to interpret either the magnitude of the response or the specific data reported in any one square. While compelling, applying the visualization rhetoric framework during the design of this visualization could have suggested other ways to make the interpretation of the visualization more clear.

Ultimately visualization rhetoric is a framework that can be useful for designers hoping to maximize the communicative potential of a visualization. Exploratory visualization platforms (like Tableau or Chart.io) could also be enhanced with an awareness of visualization rhetoric, by, for instance, allowing users to make salient use of certain rhetorical techniques when the time comes to share a visualization.

Those particularly interested in this space should consider participating in an upcoming workshop I am co-organizing on “Telling Stories with Data” at InfoVis 2011 in Providence, RI in late October.