What the heck are Perceptual MegaPixels?

Does anyone understand the science - if there is any? - behind the DxO perceptual megapixel concept? According to DxO: "The DxOMark score for Sharpness is based on the Perceptual Megapixel (P-Mpix) concept that weights the Modulation Transfer Function (MTF) of the lens with the human visual acuity." What?

The recent TN video being roasted in another thread about it being often counterproductive to use full frame lenses on crop is based on Tony's unshakeable belief that DxO has all the answers.

It does make some sense that lenses must resolve finer detail to capture the same image on a much smaller sensor just based on the magnification needed to reproduce that image at a given size. However, it also seems logical that a lens designed to resolve detail for a 40 MP full frame sensor ought to work really well on an 18 MP crop. Comparing the DxO "sharpness" numbers on a 70D for the EF24-105L, EF-S 17-55, and EF 24-70L shows 10 P-MP, 11 P-MP, and 14 P-MP respectively. (seems like the same order I'd rank those lenses in) Clearly the crop designed lens is better than one L lens and not as good as another on a crop body (if you believe DxO) How does DxO arrive at one "sharpenss" value for a >4 to 1 zoom?

To me all this says is that the newest lens designs tend to be better than the older designs?