Epistemic bias anti-licenses NPIs in polar questions

There is general agreement that the distribution of "any" is unrestricted in polar questions. I argue that this is not the case: in contexts where there is epistemic bias in favor of the prejacent of a polar question, the question exhibits the same behavior as a declarative with respect to the licensing of "any". I provide an account for this observation in terms of intervention: epistemic bias forces polar questions to be parsed as having a silent modal E which intervenes between "any" and the question operator "whether" that otherwise licenses "any".