In your discussion of contextualism, you are conflating "evaluator" and "attributor", I think.

I wasn't aware that there was an established distinction between the two. Thanks for the information! Though nothing of great weight can rest on it, since:

1. Every attributor is a (self-)evaluator. Asserting 'p' is equivalent to asserting 'p is true'.

2. Every evaluator is an attributor. To determine that some attributor's knowledge-claim is true or false, one must oneself attribute knowledge (or lack thereof) to the relevant agent.

if you say "I know I have hands" to a psychologist, you are the attributor, not the psychologist.

Yes. But if the psychologist evaluates your claim, by weighing in on its truth or falsehood, then the psychologist becomes a distinct attributor. I varied the psychologist or metaphysician as attributor, rather than paying much mind to self-attribution/self-evaluation, simply because I thought it would be less intuitive to talk about a person's knowledge-claims failing to meet his or her own psychological/social state. But, sure, strictly speaking I could have just varied the psychological state of someone self-attributing knowledge, and thereby made that person's own beliefs about his/her knowledge true or false. (At least, I think standard contextualist theories allow this.)

So if the psychologist hears about you making the claim in the context of a discussion of the Simulation argument, she should probably judge it false, irrespective of the context in which she is situated at the time she is making the evaluation.

Interesting. I think it's more complicated than that. For instance, I think contextualism predicts that the psychologist, in a later dinner conversation with a metaphysician friend, might say: 'Earlier I told a patient that he knew he had hands; but, of course, really he doesn't have hands.' The contextualist interprets this as meaning that the psychologist's state has importantly changed, hence her knowledge-attributions have changed, hence her knowledge-evaluations have changed. (Presumably part of the reason the psychologist's knowledge-attributions have changed in this case is that she's in a social context that includes a metaphysician with psychologically embedded 'higher standards'. I.e., contextualism predicts that social overlap produces synchronizations in correct knowledge attribution.)

Note that an invariantist might interpret the same data very differently. A relatively skeptical invariantist could suggest that the psychologist was speaking loosely, not-quite-correctly, when she said 'Yes, you have hands.' Or a relatively Moorean invariantist could suggest that the psychologist became too hyperskeptical in the face of social pressure from the metaphysician. On the other hand, a relativist would suggest that there isn't any determinate answer to whether 'Yes, you have hands.' was right, nor to whether 'No, he didn't have hands' was. Even stipulating all the contextual facts underdetermines whether knowledge is present.