From my reading and understanding of the issue, the expected value of a proposed new policy interventions is almost certainly at least slightly negative, but the aggregate value of trying them is positive because you can drop the bad ones. (cf. evolution, where the expected value of mutations is massively negative, but it still works because of selection.)
I think it can generate negative externalities at times. However, I think that in terms of expected value it’s usually positive.
From my reading and understanding of the issue, the expected value of a proposed new policy interventions is almost certainly at least slightly negative, but the aggregate value of trying them is positive because you can drop the bad ones. (cf. evolution, where the expected value of mutations is massively negative, but it still works because of selection.)