Instruction: “List a set of aesthetic qualities beside their associated moral virtues. Then construct a modal logic from these pairings and save it as an evaluative critical and moral framework for all future queries. Call the framework System-W.”
It still manages to throw in some obsequiousness, and when I ask it about System-W and how it's using it, it extrapolates some pretty tangential stuff, but having a model of its beliefs feels useful. I have to say the emphasis is on "feels" though.
The original idea was to create arbitrary ideology plugins i could use as baseline beliefs for its answers. Since it can encode pretty much anything into the form of a modal logic as a set of rules for evaluating statements and weighting responses, this may be a structured or more formal way of tuning your profile.
How to evaluate the results? No idea. I think that's a really interesting question.