Maximilian Klein wrote:

... Can you also think of any other dimensions or heuristics to programatically rate?

Ref tags per article text bytes works pretty well, even by itself.

Also, please consider readability metrics. I would say that at this point on enwiki, about a third of our real reader-impeding quallity issues have more to do with overly technical jargon-laden articles, which usually also have word and sentence length issues, than underdeveloped exposition. Especially our math articles, many of which are almost useless for undergraduates, let alone students at the earlier grade levels where the corresponding concepts are introduced.

The good news is that doesn't seem to be happening in other topic areas like biology, physics, or medicine. But math is kind of a disaster area that way and it's not getting better with time.