I'll adjust it a little bit. I don't think it's specifically human performance. I think it's that the expectation level we have for air carrier accidents, which is now zero, seems to be in conflict with the fact that people sometimes make mistakes. The fact that there are usually at least two people checking things isn't good enough for a zero accident rate.
The math on this should be pretty easy to figure out. How many mistakes per thousand flights does the average controller make? Of the potential mistakes, what percentage could lead to a collision? So if it's 1 in 10 thousand for pilot or ATC, and they both have to miss to have a hit, then that's average of 1 in 100 million flights. Or whatever the real math is.
Based on the math, they can change the procedures, increase the distances, or whatever. Worst case? They have to build more runways, figure out how to get the carriers to user larger planes so there are fewer overall flights, get people to fly less, get the numbers down.
I say all the above because I while I believe training can help with performance, you're talking about 10%-20% increases in things, not double. Especially with something like alertness. If you want to make meaningful differences in things, you need different procedures, not better training, just from my perspective.
Or we can say that things are fine, and that we're catching these almost accidents. I don't think that's a bad argument, really. To me, airline travel is very safe, and we're just used to zero accidents, so the next one is going to hit pretty hard.