Adrian Liu

An ‘‘expectation reflection’’ principle, requiring epistemic agents to match their future rational opinions on average, is often thought to be a requirement of epistemic rationality. Dorst (2023) posits a mechanism wherein ambiguity asymmetries in evidence can lead to rational violations of expectation reflection. He further argues that this mechanism could rationalize real-world belief polarization. I show that asymmetries in when evidence is ambiguous do not themselves generate rational polarization. I then show that Dorst’s argument equivocates between two types of ambiguity asymmetry. Only the second type of asymmetry leads to polarization, but only the first type is plausibly generated by the relevant mechanism. A realistic story of rational polarization would require a mechanism that generates asymmetries of the second type, not the first, and characterizing such a mechanism is no simple matter.