> The return on investment for a PhD student in Computer Science to society must be absolutely ridiculous.

Are you sure? Most of them produce a bunch of either theoretical papers with little hope of ever being applied (e.g. guys squabbling about second decimal point of the constant in theoretical computational complexity of some algorithm), or, if they're in an applied field, their papers are heavily gamified to beat state of the art in some meaningless way (or just fake the results). Most of it is useless BS and I'd say is actually worsening the society, because journals and conferences are now filled with useless cruft which make finding the actual valuable contributions very hard.