Perhaps it helps to see imagined persons as controlled alternate personalities. Do we see people with multiple personality disorder as multiple persons? And why? How do we even define the identity of a person? Don’t we need to define the identity of a person in relation to a different person? Should we set it as totality of your dissimilarities with other persons? After all, if there are two perfectly exact copies of you, it doesn’t make any sense to say that they are two different persons – according to Ockham’s razor!
How would these thoughts contribute to solving the original questions? Hmm, let’s see it that way: A merely imagined character is not completely defined, because there are always aspects of them that would be defined for a real person. That’s why we can’t definitely how the character would differ from other persons. The imagined character has less “definiteness”, which is an attribute that would distinguish them from an actually simulated person. So, a virtual person needs to have a high degree of definiteness, otherwise it’s just a “mere imagination”.
How could outsiders tell the difference? Questioning might be an option, but asking questions about the presumed virtual person would increase their definiteness by the imagining mind coming up with more properties of that person. So, questioning is not a good idea. I guess there’s really no other reliable way than having some kind of device that tells you about the structure and content of the simulating mind. Some kind of “neuroscope”.
The characters we humans are able to imagine have a really low “definiteness”. Even our understanding of ourselves would probably only produce a laughable caricature of ourselves, because the “definiteness” of our model of ourselves still doesn’t have a really high resolution.
That is not surprising. The surprises come from the logical consequences of having a character with certain properties put in a novel situation. People usually take time for evaluating such logical consequences, and do them just in time, and not in advance for all possible situations. That’s what allows some consequences to be surprising, simply because we aren’t very good at predicting the results of the application of logic. That is also why mathematics can sometimes be surprising, even though all mathematics is already defined by basic axioms.
Good question! I guess in a more controlled simulation the virtual persons would have less autonomy or agency, and one might say that this decreases their degree of “personhood” – the more autonomous a being is, the more it counts as person, because persons are usually thought of as real agents.