Background
Presently, the Android app uses the ORES scores of Goodfaith and Damaging on the diff screen and filter view to provide patrollers a quality score for an edit. As ORES is deprecated, there is guidance from research to start using the Revert Risk Model instead.
Task
- @scblr Update designs on diff screen and filter view (Edit Patrol and Watchlist) to reflect using the revert risk model
- @ARamadan-WMF check in with Moderator Tools CRS
- Engineers change update interface and API after checking in with the team responsible for the new API and Deigo
- Engineers work with Shay to ensure data wiring is up to date with change
Requirements
- Instead of buttons that read Quality and Damage, there should be a single indicator that reads "Revert Risk Score"
- Take users that want to learn more about the Revert Risk score to https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk
Documentation
- https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk
- https://api.wikimedia.org/wiki/Lift_Wing_API/Reference/Get_reverted_risk_language_agnostic_prediction
- https://wikitech.wikimedia.org/wiki/ORES
- https://lists.wikimedia.org/hyperkitty/list/wikitech-l@lists.wikimedia.org/thread/EK65B7QCQHEG37C2ERPIUSP64OX3ZEUJ/
- T348298
- T352217
Designs (Figma)
1. Diff view  | 2. Tooltip on tap  |
- Uses M3 chip component
- Tooltip is only shown when tapping the chip (not on first use)
- Tooltip contains a link to https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk