This task involves the work of making it possible to store data when a specific suggestions is declined/dismissed so that it can in turn be used to suppress these suggestions from being shown in the future.
Stories
| ID | Story | Ticket |
|---|---|---|
| 1. | As an editor who has declined/dismissed a suggestion, I don't want to see that same suggestion again in future edit sessions I initiate, so that my editing experience reflects decisions I've already made and doesn't repeatedly prompt me to reconsider something I've already evaluated. | T427236 |
| 2. | As an experienced editor who has declined/dismissed a suggestion, I want that action to influence how/if that suggestion is shown to other people in the future, so that my judgment as an experienced contributor: 1) Helps others avoid spending time on suggestions that aren't worth surfacing, 2) Prevents newcomers from making changes that do not support project policies and guideline, and 3) Can make the suggestion pool better over time | T433147 |
Open questions
- 1. At what point would a suggestion no longer be showed to anyone in the future? Thank you to @KStoller-WMF for prompting this.
- 2. To what extent (if any) should volunteers, on a per project basis, be able to configure the conditions/thresholds that need to be met for a suggestion to be suppressed/invalidated? E.g. invalidation based on reject button clicked, revision changed, text does not match etc. Thank you to @OKarakaya-WMF for prompting this.
- For context, addalink invalidates based on text match and llm generated suggestions invalidate based on revision etc
- 3. How might we handle situations in which it's believed that suggestion(s) have been invalidated in bad-faith? E.g., someone going through and rejecting every suggestion they see across many articles.
Potential approaches
Background
At present, suggestions people decline/dismiss within Suggestion Mode will be resurfaced to them (and other volunteers) in future edit sessions.
Without adjustment, this could lead unsuspecting volunteers to affect content in ways that do not align with project policies and guidelines. This could, as a result, create more patrolling work for experienced volunteers and, more broadly, this could cause people to lose trust/confidence in Suggestion Mode.
Related
- Proposal to move suggestions+invalidation to maintenance templates that are directly curated by editors so no back-end work is needed for managing invalidation: T432485: [prototype] User script that uses a LLM to help editors resolve cleanup templates
- Example of template for locally overriding newcomer tasks: https://en.wikipedia.org/wiki/Template:No_newcomer_tasks
- Sketch: invalidation framework via @Ottomata
- Draft proposal for modelling suggestion "sameness" in a way that enables efficient dismissal