I think it's image quality that people generally associate with CG, namely lack of aliasing, which high downsampling provides.
I've seen someone say that "downsampling" is the wrong name to call it, since it typically means discarding all information above a certain threshold, which is exactly what happens here. Perhaps supersampling is the better term? That's the term devs seem to use when their games support it too.
Edit: Getting geeky about it now, looks like the correct technical term for it is
Decimation
So GeDoSaTo and DSR are decimators.