Updating for Externalists
Externalists believe that it is possible to have the total evidence E without having the evidence that E is your total evidence. Externalism is inconsistent with the conjunction of the principle of conditionalization and the principle of reflection. So externalists must reject at least one of these principles. Adam Elga affords an argument that the externalist should reject reflection. However, Bernhard Salow has recently argued persuasively that this leads to indefensible consequences: the externalist conditionalizer must condone deliberately designing experiments which you can recognize in advance will make you arbitrarily confident of a falsehood. Deliberate self-delusion like this is not rational. So the externalist must reject conditionalization. There are two extant proposals for which update rule externalists should adopt in its stead: one defended by Matthias Hild and Miriam Schoenfield; and one which I have defended for the special case of experiments whose outcomes are theory-dependent. Pace Hild and Schoenfield, the first proposal is inconsistent with externalism. And the second proposal is insufficiently general to rescue the externalist from all cases of deliberate self-delusion. I therefore show how to generalize the second update rule to a wider class of externalist experiments, and show that, so generalized, the update will always satisfy the principle of reflection (and, therefore, will never permit deliberate self-delusion).