I don't think there's anything really new here for long-time LWers (we all know Goodhart's law/Lucas critique, Ioannidis-style results, 'extraordinary claims require extraordinary evidence' etc), but some of the points about cost-benefit of statistical precision might be novel to some of us.
I agree with this completely. A lot of the times we simply don't have information beyond broad generalizations or estimates more accurate than within a order of magnitude or something, and yet people still insist on trying to formulate and act on precise quantifications, even though such information doesn't even exist.
Brian Tomasik's latest article, 'Quantify with Care', seems to be of sufficient interest to readers of this forum to post a link to it here. Abstract: