I kinda harp on the "belief is not sufficient" bandwagon, because I deal with dementia every day.
People with dementia genuinely believe some very bizarre things. They are not "Lying" when they report these things, and insist on their truthfulness-- Lying requires knowledge of the deception. Rather, they are relating non-factual accounts.
If I were to design a deep-cover reconnaissance probe, and had access to advanced AI and nanofabrication technologies, (and had precisely ZERO fucks given about ethics, either for AI or for sentience in general), the ideal design would be one that is indistinguishable from the studied subject population, and which is incapable of revealing any long term plans or details about its operation, because it itself does not know.
It would give completely unbiased return footage/data, because it does not choose what data to send-- it is always sending. It just "Lives its life" down there, and tells me all about it, all the time, completely unawares. Further, if it experiences all aspects of the subject population's culture-- such as birth and childhood, then its resultant "Personality" would be free of any meddling or bias on my part; It would be a member of the society it was sent to study covertly, and its own biases that it develops would be part of the dataset collected.
That of course, presumes that hyper-advanced aliens are up there, and that they have those qualifications. Russel's Teapot applies.
The philosophical implications about such robotic probes are fun to explore though.