Reply to revised OP: Policy debates should not appear one-sided (because there's no systematic reason why good policies should have only positive consequences) but epistemic debates are frequently normatively one-sided (because if your picture of likelihood ratios is at all well-calibrated, your updates should trend in a particular direction and you should almost never find strong evidence on more than one side of an epistemic debate; "strong evidence" by Bayesian definition is just that sort of evidence which we almost never see when the hypothesis is wrong).