@ct_bergstrom just tried Bard and it seems slightly more objective than GPT. It's important to remember these models are supposed to steer away from negativity and "bad thoughts". That's got to "leak" into character judgments.