But the alternative, waiting to apply new-but-actually-correct hypotheses until totally, absolutely proven, also causes lost lives. And as is usual with "Type

But the alternative, waiting to apply new-but-actually-correct hypotheses until totally, absolutely proven, also causes lost lives. And as is usual with "Type 1" vs "Type 2" errors, saving lives by avoiding one problem costs lives due to the other problem. The trick is to sit in the minimum of "lives lost due to Type 1 errors plus lives lost due to Type 2 errors". Unfortunately, that's not an analytic function with a known formula and computable derivative...