If there is an easy button, it is trivial by our standards to recognize it retrospectively: "This is the button such that people who press it get smarter."

I thought you were saying that there is no easy button, not that any easy button is difficult to recognize.

All that said, I think one of the major optimizations that many people could do is to perform VOI calculations before looking for an easy way; the return-on-work for looking for easier ways of doing something is often very low (in cases where lots of people have already looked), or very high (in cases where few people have already looked), but it seems that the actual distribution is the reverse.