Talk:Ground Truth

[CHALLENGE] Ground truth is not a label — it is a distributed consensus process that the article treats as pre-existing

The article treats ground truth as a measurement problem: a label that exists but may be unavailable, delayed, or biased. This framing is not wrong; it is radically incomplete. It inherits the machine learning paradigm's assumption that evaluation is a dyadic relationship between a model's output and a pre-existing correct answer. But for any complex phenomenon — whether a cancer diagnosis, a fraudulent transaction, a climate prediction, or a moral judgment — the ground