My naive response, when I hear this worry expressed, that the clinician is obviously right and her patient wrong if the latter says (today) he is Napoleon (to take an example from Szasz) is usually greeted with a raised eyebrow and the question: but how do we know that? And once one begins to press this worry, it can seem that this meta-level asymmetry, which now has me and the hypothetical clinician, with perhaps other members of society, on the one hand and the supposed patient and possibly my more open-minded interlocutor (“Perhaps he really is Napoleon, travelled through time!”) on the other hand, itself only comprises difference in numbers.
I suspect that this is a case where the obvious and naive thought is given up not because it ceases to ring true, but because of an intuitive philosophical argument. The ‘who is to say?’ question is a philosophical question and seems to drive a conclusion that - like much philosophy - goes against what we had taken to be obvious. If that is so, it is a pity that disjunctivism isn’t more widely available since the worry Szasz implicitly trades on a form of the argument from illusion. I’d like to print out these passages from Sebastian Rodl’s book and hand them to all and sundry:
The argument (from illusion) is: Whenever I seem to know something (on the basis of perceptual. experience), I might have been fooled. Had I been fooled, I would not have known that I was. I would not have been able to tell my situation apart from one in which I am not fooled. This shows that my grounds do not place me in a position to exclude that I am in such a situation. They do not enable me to exclude that I am fooled. —The argument supposes that, had I been fooled, I would have believed the proposition in question on the same grounds on which I believe it now that I am not fooled. This straightforwardly entails that these grounds do not establish the truth of what I believe and therefore do not provide me with knowledge.
But when I know something on the ground that, say, I perceive it to be the case, then I would not, had I been fooled, have believed it on this ground, for, had I been fooled, I would not have perceived it to be the case. Hence, when I am not fooled, my grounds exclude that I am fooled: when I perceive how things are, I am not fooled with regard to how they are. One might object that this grants me grounds that rule out error at the price of making it impossible for me to know whether my belief is based on such grounds. For, when I am fooled, I do not know that I am fooled. So I can never know whether I am not fooled and my beliefs are based on grounds that [establish] their truth, or whether I am fooled and such grounds are unavailable to me.
This objection repeats the mistake: from the fact that, when I am fooled, I do not know that I am, it does not follow that, when I am not fooled, I do not know that I am not. When I know that p as I perceive it to be the case, then I know that I perceive that p. Thus I am in a position to distinguish my situation from any possible situation in which I would be fooled, for, in any such situation, I would not perceive that p, while in the given situation I do. [Rodl, S. (2007)Self-Consciousness, Cambridge, Mass.: Harvard University Press pp 157-8]
This is a good summary of disjunctivism in epistemology: the view that the epistemic significance of an experience is not restricted to what would be the highest common factor shared between a veridical experience and a mere illusion. There can be more to an experience than just an appearance, construed as merely as how things seem. It can be a matter of drinking in an aspect of the world itself. Rather than being limited to the highest common factor, experiences come in two distinct forms: Either, when all goes well, a taking in of how the world is. Or, when things go badly, a mere deceptive appearance. Hence ‘disjunctivism’. In the former case, experience is relational: a feature of the world itself forms part of the subject’s state.
Rodl’s passage makes two key points. First, in the good disjunct, one knows how one knows and hence, in the good disjunct, one can know that one knows (since how one knows is good enough because necessarily world-involving). Of course, had one been in the bad disjunct, one would have thought that one knew how one knew and that one knew but one would have been wrong on all counts: wrong that one knew (whatever fact about the world) and hence how one knew (it) and hence that one knew that one knew (it).
But, second, the fact that, in the bad disjunct, one does not know has no effect on the good disjunct. ‘[F]rom the fact that, when I am fooled, I do not know that I am, it does not follow that, when I am not fooled, I do not know that I am not’.
Disjunctivism helps highlight the real asymmetry in the case above. It is not that both clinician and patient have the same sort of cognitive state with the fact about Napoleon’s identity being an inaccessible extra. The real asymmetry in the case of the sane clinician and deluded patient is that one is right and the other wrong about who is and isn’t Napoleon.