Research may become less about finding a single answer and more about using AI to map the disagreement. The goal should be better questions, not faster certainty.