In particular, I thought I should take issue with the quote he gives from David Smith:
good charts merely present data, and leave the analysis (obvious though it may be) to the viewer
To me, it seems that a charts which "merely present data" are just not possible. The designer of the chart needs to make all sorts of descisions: the order of the data, importance of axes, labels, colours and so on. Few of these are without consequence to the interpretation of the graph, and the futher the chart is from a simple table, the greater the potential for bias (or, more charitably 'interpretation') there is.
I say 'potential', because not every opportunity for bias need be taken. Having said that, not all bias is intentional either. Some of it might be just accidental, or plain confusing. But each additional line on the graph has the potential to introduce spurious anaysis.
In fact, I would venture so far as to say that this is a problem even for the simpler charts that Nathan Yau advocates via (here and here) which are used to convey rational argument without appeals to emotion. There is still ample opportuinity to mislead others (and onesself) by unintentionally misusing a graphic. Indeed, the impression that a graphic is unbiased because it is simple might make us let our guard down.
It's perhaps unfair to assimilate the use of perceptual tricks to appeals to pathos: such techniques could be used in rational arguments too. But they are subtle and easy to get wrong.
I suspect that the fundamental reason for this potential to mislead is also what makes data visualizations useful in the first place: we are phenomenal pattern-matching machines, and we're particularly good at visual patterns. We would struggle to interpret 5 million rows of a database, but can easily identify trends when the same data are presented in a different format.
(This post was edited on 07/10/2011 to properly attribute a quote to David Smith.)
Is this how I start all my blog posts? ↩