Tufte refutes Tufte
In several places in the early pages of Visual Explanations, Edward Tufte seems to imply that the only good sort of explanation is a quantitative explanation. Consider the following passage (which is just one such example):
"More generally, when scientific images become dequantified, the language of analysis may drift toward credulous descriptions of form, pattern, and configuration — rather than answers to the questions How many? How often? Where? How much? At what rate?"
The funny thing is, Tufte's own book is a stellar refutation of the notion that the only good explanation is a quantitative explanation. While Tufte's book is about quantitative analysis, it itself is not an exercise in quantitative analysis. And yet it is filled with outstanding, non-quatitative explanations.
"More generally, when scientific images become dequantified, the language of analysis may drift toward credulous descriptions of form, pattern, and configuration — rather than answers to the questions How many? How often? Where? How much? At what rate?"
The funny thing is, Tufte's own book is a stellar refutation of the notion that the only good explanation is a quantitative explanation. While Tufte's book is about quantitative analysis, it itself is not an exercise in quantitative analysis. And yet it is filled with outstanding, non-quatitative explanations.
Well, the final chapter of the book is devoted to 'confections' which are almost devoid of quantitative information but convey narrative ideas by juxtaposition and analogy, so it's hard to imagine that tufte really thinks this. Visual Explanations is much more 'ecumenical' in the kinds of meanings and information display that it analyzes than Tufte's previous work.
ReplyDeleteThe way that Tufte talks about 'scientific' visualizations may be colored by his previous books, notably The Visual Display of Quantitative Information, and to a lesser extent, Envisioning Information. Both of these are more focused on maps, graphs, charts and tables, and how to give these a greater 'information density'.
I don't know if it's brought out by Tufte explicitly, but a there is a potentially useful distinction between information graphics that are intended to:
a: inform a decision, discover a pattern, or settle a contentious issue. For example, time series, scatterplots, or mapped data.
vs.
b: illustrate a concept or convey a narrative... such as a 'confection' or an anatomical diagram.
For example, Tufte would never demand that a scale be put on an anatomical diagram illustrating the human ear, but would object to a bar chart without a scale.
....
It also occurs to me that when Tufte uses the words 'become dequantified', that implies that potentially relevant quantitative information is removed or rendered illegible by distortion or lack of labeled scales.