I'm interacting with an AI, helping it to perfect its interaction with me, but the changes are not universal human changes that would correct the interactions. A thought: there should be superusers whose interactions, examined, could be used to make universal changes to an AI's interactions or the user interface, which seems to be flawed, at least for the one I'm using.