I’m in the small camp doing visualization with R/Illustrator. A few thoughts.

Who Uses R?

The lack of time spent doing design may primarily be a factor of who uses R. Like many, I came to use it in graduate school, doing scientific research. Scientists (of any stripe) are not normally in the business of publishing graphics for public consumption. It may also be that their skillset and mindset place a low priority on design — they have never thought about design per se, just doing what they need to for academic/technical publication. Although many scientists leave to work in other fields, I imagine few did what I did, and went into graphic design.

Maybe Most R/Illustrator People Don’t Really Use Illustrator For Much

In the R/Illustrator combo, there is a natural split between initially getting the data mapped onto 2D space in a given manner — in R — and then doing more designery work like layout and color and typography — in Illustrator. For many people, they consider the R stage far more important, they see the Illustrator stage as merely about prettifying the result, and they probably have limited Illustrator skills. I knew many scientists who could “use” illustrator in the sense of adjusting text labeling and font. It’s part of their process, but not by much. And if they think of illustrator as being the “design stage” they wont’ self-report designing much either.

R is to JS, but D3 is not to Ggplot

There are also some natural impediments in R’s major libraries to doing design there — because as you say, Elijah, more design work could/should be done then.

ggplot2 seems to be dominant, while I have been a long-time lattice user. Both of these are great for pretty quick and dirty design work. But neither are very good for controlling layout, or making parallel arrangements of related graphs (infographic or dashboard style). And in general, these libraries are fairly easy because they make a lot of assumptions about what will be needed — far more assumptions than something like D3.

Yet this isn’t really a fair comparison. While D3 is a library, just as ggplot2/lattice are, D3 is lower-level and more powerful. People turn to Vega/others if they want something easier. What’s the R equivalent of D3? There actually is one: Grid, which runs lattice and ggplot2 both. With grid you can create arbitrary graphics in a window, scaled to x and y dimensions. To make original visualizations in R, you have to get down into grid territory. Grid isn’t that hard, but most people never touch it or know about it; and it may be a little too much like programming for some R uses, who are primarily statisticians/scientists, not programmers (even at the hobby level, such as myself). Grid is tiresome, but you can write your own grid libraries, of course — that’s what I did, to make custom vizzes for print in particular; still a weak spot for the JS world.