Comment by hamstergene

Saying biasing implies infinite possibilities to which the data can be made biased towards. It instantly raises the question why bias towards this and not something else. It almost sounds like a bad thing.

Saying debiasing implies there is a correct result which needs to be achieved by removing bias. It raises no questions, we want correct, we don’t want incorrect. Doing a good thing implied.

Don’t misinterpret me, I don’t think public models should spew commonly harmful content out of the box. Just explaining the PR trick, which is what the word “de”biasing de-facto is in this context.