Change Details

This task involves the work of reconsidering what happens when people elect to decline taking action on a Peacock Check that is presented to them. The need for this exploration emerged during a 19 March Editing Team discussion wherein we came to align on the following disconnect: 1. Publishing an edit that introduces text written in – what experienced volunteers are likely to consider – a non-neutral tone is relatively disruptive 2. Dismissing a Peacock Check, at present, is relatively lightweight/seamless The disconnect between "1." and "2.", we think, will send people encountering Peacock Check a conflicting message. === Stories - As a newcomer/Junior Contributor who has explicitly chosen to forego revising the tone of what they've written, I would like to know what information/content I can offer to the experienced volunteers who are likely to review/patrol this soon-to-be published edit, so that I can increase the likelihood that the contribution I consider to be useful and worthwhile remains on the wiki. - As an experienced volunteer who is reviewing an edit in which someone has a) added new text written in a tone considered to be non-neutral and b) decided NOT to revise the tone of said text, I would like to know why this person made this decision so that I can decide what – if any – action (e.g. revert the edit, post a message on the person who made the edit's talk page, etc.) to take in response === Decision(s) to be made - [ ] 1) What - if any – modifications will we make to the Peacock Check's "decline path" to ensure it reinforces the importance of considering tone when adding new content to Wikipedia and the potential impact of //not// doing so? - [ ] 2) How/if might we use the signal the decline path ideally captures/elicits could be useful for the model retraining the ML Team is planning in T393103 - [ ] 3) Related to "2)," how might we surface the signal the decline path [ideally] captures to volunteers so they can use it as an input to improve moderation processes? //E.g. maintain an allow-list of sorts to avoid false positives.// -- One idea, might it be possible to look at the SHAP values the model returns for edits where some declines to revise what they've written and log this in some place? //@ppelberg to talk with @SSalgaonkar-WMF about this.// - [x] 4) How – if at all – will we pass on the responses volunteers offer to the model/ML Team? -- TBD; see T393103. === Approaches //Ideas needed...// - **Approach #1**: prompt people to indicate //why// they've declined to revise the tone of what they've writen? NOTE: any approach needs to scale to a world where people are seeing ≥3 Checks/edit session. === Done - [ ] Potential "Approaches" are documented - [ ] An "Approach" is chosen and implemented

This task involves the work of reconsidering what happens when people elect to decline taking action on a Peacock Check that is presented to them. The need for this exploration emerged during a 19 March Editing Team discussion wherein we came to align on the following disconnect: 1. Publishing an edit that introduces text written in – what experienced volunteers are likely to consider – a non-neutral tone is relatively disruptive 2. Dismissing a Peacock Check, at present, is relatively lightweight/seamless The disconnect between "1." and "2.", we think, will send people encountering Peacock Check a conflicting message. === Stories - As a newcomer/Junior Contributor who has explicitly chosen to forego revising the tone of what they've written, I would like to know what information/content I can offer to the experienced volunteers who are likely to review/patrol this soon-to-be published edit, so that I can increase the likelihood that the contribution I consider to be useful and worthwhile remains on the wiki. - As an experienced volunteer who is reviewing an edit in which someone has a) added new text written in a tone considered to be non-neutral and b) decided NOT to revise the tone of said text, I would like to know why this person made this decision so that I can decide what – if any – action (e.g. revert the edit, post a message on the person who made the edit's talk page, etc.) to take in response === Decision(s) to be made - [ ] 1) What - if any – modifications will we make to the Peacock Check's "decline path" to ensure it reinforces the importance of considering tone when adding new content to Wikipedia and the potential impact of //not// doing so? - [ ] 2) How/if might we use the signal the decline path ideally captures/elicits could be useful for the model retraining the ML Team is planning in T393103 - [ ] 3) Related to "2)," how might we surface the signal the decline path [ideally] captures to volunteers so they can use it as an input to improve moderation processes? //E.g. maintain an allow-list of sorts to avoid false positives.// -- One idea, might it be possible to look at the SHAP values the model returns for edits where some declines to revise what they've written and log this in some place? //@ppelberg to talk with @SSalgaonkar-WMF about this.// - [x] 4) How – if at all – will we pass on the responses volunteers offer to the model/ML Team? -- TBD; see T393103. === Requirements //WIP// - The {nav Decline} response someone submits needs to be logged so that we can aggregate these responses -- Based on what the Editing Team discussed offline on 12 May, this logging should happen automatically via VEFU. -- Although, we'll need to talk with the ML Team to learn what – if any – adjustments/enhancements will need to be to this data so that it can useful for retraining purposes === Approaches //Ideas needed...// - **Approach #1**: prompt people to indicate //why// they've declined to revise the tone of what they've writen? NOTE: any approach needs to scale to a world where people are seeing ≥3 Checks/edit session. === Done - [ ] Potential "Approaches" are documented - [ ] An "Approach" is chosen and implemented - [ ]

This task involves the work of reconsidering what happens when people elect to decline taking action on a Peacock Check that is presented to them. The need for this exploration emerged during a 19 March Editing Team discussion wherein we came to align on the following disconnect: 1. Publishing an edit that introduces text written in – what experienced volunteers are likely to consider – a non-neutral tone is relatively disruptive 2. Dismissing a Peacock Check, at present, is relatively lightweight/seamless The disconnect between "1." and "2.", we think, will send people encountering Peacock Check a conflicting message. === Stories - As a newcomer/Junior Contributor who has explicitly chosen to forego revising the tone of what they've written, I would like to know what information/content I can offer to the experienced volunteers who are likely to review/patrol this soon-to-be published edit, so that I can increase the likelihood that the contribution I consider to be useful and worthwhile remains on the wiki. - As an experienced volunteer who is reviewing an edit in which someone has a) added new text written in a tone considered to be non-neutral and b) decided NOT to revise the tone of said text, I would like to know why this person made this decision so that I can decide what – if any – action (e.g. revert the edit, post a message on the person who made the edit's talk page, etc.) to take in response === Decision(s) to be made - [ ] 1) What - if any – modifications will we make to the Peacock Check's "decline path" to ensure it reinforces the importance of considering tone when adding new content to Wikipedia and the potential impact of //not// doing so? - [ ] 2) How/if might we use the signal the decline path ideally captures/elicits could be useful for the model retraining the ML Team is planning in T393103 - [ ] 3) Related to "2)," how might we surface the signal the decline path [ideally] captures to volunteers so they can use it as an input to improve moderation processes? //E.g. maintain an allow-list of sorts to avoid false positives.// -- One idea, might it be possible to look at the SHAP values the model returns for edits where some declines to revise what they've written and log this in some place? //@ppelberg to talk with @SSalgaonkar-WMF about this.// - [x] 4) How – if at all – will we pass on the responses volunteers offer to the model/ML Team? -- TBD; see T393103. === Requirements //WIP// - The {nav Decline} response someone submits needs to be logged so that we can aggregate these responses -- Based on what the Editing Team discussed offline on 12 May, this logging should happen automatically via VEFU. -- Although, we'll need to talk with the ML Team to learn what – if any – adjustments/enhancements will need to be to this data so that it can useful for retraining purposes === Approaches //Ideas needed...// - **Approach #1**: prompt people to indicate //why// they've declined to revise the tone of what they've writen? NOTE: any approach needs to scale to a world where people are seeing ≥3 Checks/edit session. === Done - [ ] Potential "Approaches" are documented - [ ] An "Approach" is chosen and implemented - [ ]