Although I agree with its conclusion, I am not persuaded by the knowledge argument against physicalism. A physicalist can say that when Mary learns what it’s like to see red, she changes only by coming to bear to a truth she already knew a knowledge-relation that she never bore to that truth before. Or, so I have argued.
I do think that the knowledge argument is safe from the most popular objection to it, which is that Mary can learn what it’s like to see red merely by acquiring knowledge of a truth that she didn’t know before, where that truth is made true by some fact that Mary already knew (by knowing some other truth). Those who press this objection point to cases in which, for example, someone knows that there is water in the sink without knowing that there is H2O in the sink. This is an example of someone who knows a certain fact (i.e., that there is water a.k.a. H2O in the sink) by knowing some, but not all, of the truths which that fact makes true. But if physicalism is correct, Mary is an example of someone who knows all facts (since she knows all physical facts, which are the only facts there are, according to physicalists). But if Mary knows all facts, then she knows all facts about all truths, making it hard to see how she could be ignorant of any truth.
In this paper, I argue that G.E. Moore’s open-question argument is vulnerable to the same objection I raise to the knowledge argument.
I give a brief, non-technical summary of my position on the knowledge argument here.