Right, you might want to say that the distributions added to the representor should not have an epistemic interpretation, but should be thought of as a choice-theoretic representation. That is, we have a choice-theoretic reflection principle (something like, if I know that I’ll in future judge A and B to be permissible, I should judge them to both be permissible now), which along with other constraints forces choice behaviour that is representable by the larger representor.
Right, you might want to say that the distributions added to the representor should not have an epistemic interpretation, but should be thought of as a choice-theoretic representation. That is, we have a choice-theoretic reflection principle (something like, if I know that I’ll in future judge A and B to be permissible, I should judge them to both be permissible now), which along with other constraints forces choice behaviour that is representable by the larger representor.