The logic is wrong because you've assumed a negative or zero correlation between societal usefulness and earnings when the correlation is actually a positive one.

Maybe a social work degree has positive externalities that far exceed the earnings. But there's also a lot of crap that doesn't, and incentivizing the landslide of crap and disincentivizing the skills that are in shortage will do a lot of damage.

Stuff that pays well tends to be stuff that's actually valued by society. Structural engineers and surgeons get paid lots of money because they actually value add and their skills are in short supply.