I think, if you go ahead with this project, it would be good to get a consultant in. An expert on current AI with a good prognosis of how they'll work in the future. With my limited understanding of this kind of science, I wouls assume, what they need to measure, is comprehension beyond what a machine can understand and in fact being flawed. A non artificial or purely randomized pattern of flawdness so to speak. Also, may, the ability to create something that a human can appreciate as harmonious better than an AI... which i where his tone deafness would make things interesting. (I'm better at getting characters into trouble than out lol)
Just wondering, is the AI in your world self-conscious? Would that be a game changer? Maybe an additional inner-journey or even deciding to free all those artificial ?people??