This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.
We shared our annual plan, for the period July 2023 - June 2024.
Our first project of the year will be Community configuration 2.0, which helps editors with extended rights transparently and easily configure important on-wiki functionality.
After we finish work on Community configuration 2.0, we will hope to fit in one of the following projects:
Article creation: This project aims to provide new editors with better guidance and guardrails in the article creation process, with the intention of lightening the load of new page reviewers.
Non-editing participation: This project aims to create low-risk ways for readers to participate in Wikipedia with the intention of funneling more readers into contributing to the Wikimedia movement.
Suggested Edits are now receiving topic predictions via the new Language-Agnostic Topic Classification. This change affects non-English Wikipedia wikis. It will ensure newcomers receive a greater diversity of task recommendations. Before, as this feature was a test, English Wikipedia was used to select topics. The change is gradual as lists of topics are refreshed when they become empty. The Research team will evaluate the impact in a few months. [1]
Starting on August 1, a new set of Wikipedias will get "Add a link": Georgian Wikipedia, Kara-Kalpak Wikipedia, Kabyle Wikipedia, Kabardian Wikipedia, Kabiyè Wikipedia, Kikuyu Wikipedia, Kazakh Wikipedia, Khmer Wikipedia, Kannada Wikipedia, Kashmiri Wikipedia, Colognian Wikipedia, Kurdish Wikipedia, Cornish Wikipedia, Cornish Wikipedia.
Mentorship
The Growth team provides dedicated features to establish a mentorship program for newcomers. Every newcomer gets a volunteer mentor who provides encouragement and answers questions. Communities can set up or join this mentorship system by visiting Special:ManageMentors. This mentorship system is configurable by the community at Special:EditGrowthConfig.
More communities have implemented mentorship. A Wikimedia Foundation data scientist will be looking at the impact of Mentorship. We will look at the impact on Spanish and English Wikipedia. [2]
The Growth team will also host a Mentoring new editors on Wikipedia session at Wikimania 2023 in Singapore. Workshop attendees will help brainstorm improvements to Growth’s mentorship features.
Positive reinforcement
We will share more complete experiment analysis for all the three parts of the Positive reinforcement project soon. At the moment, the new Impact module, Leveling up, and Personalized praise are still being A/B tested on the Growth team's pilot wikis.
In the meantime, initial leading indicators for the Personalized praise project have been published. Although this is still a relatively small sample, results seem healthy. They show that Mentors are indeed receiving notifications and clicking through to view their praise-worthy mentees.
Growth contributes to IP Editing migration
The Growth team is currently focusing on IP Editing: Privacy Enhancement and Abuse Mitigation. It is a project that touches many different Wikimedia Foundation teams. The Growth team will focus on temporary accounts through two main points:
the user experience of a logged-out user, that switches to a temporary account,
change Growth-owned extensions and features, so that they work as expected with temporary accounts. [3]
I noticed at Special:ManageMentors that a small number of mentors have been inactive for a few months. When I experimented by randomly editing one (EchidnaLives) to change them to "Is Away", I left the date field blank as I felt unable to enter a date when they would return. I clicked 'Submit' and was informed that "Mentorship-related options for mentor EchidnaLives were changed.".
However purging the cache at Special:ManageMentors still showed them as 'Active'.
Only when I made up a specific return date was the mentor marked as 'away'.
Suggestion: Mentors who have clearly not edited for a month or two ought to be able to have their status changed by other mentors without them having to state a specific return date. (Clearly they would have no idea when that would be). An automated message to the mentors talk page to inform them that their mentor status has been altered would then alert them to the need to change their status back once they return to active editing. I would be extremely reluctant to remove any mentor completely for taking a few weeks away (we're all permitted to do other stuff in real life and then return!). Alternatively, an automated process which detected that a mentor has been inactive for over a fortnight and then marked that person as 'away' might be another solution. The key issue is to avoid new users asking questions to a mentor that simply are never going to get answered.
I'm also not sure of the best way of searching for mentee questions that have gone unanswered for some time. I've added this search link to the Mentor List FAQ. It's based on assuming that the question was the 'latest revision' of the mentor's userpage over the last 30 days, but perhaps others can suggest more effective ways to look for unanswered questions. Nick Moyes (talk) 22:33, 29 August 2023 (UTC)
Those are good thoughts. We definitely don't want mentees asking questions to inactive mentors. Has this been brought to the Growth Team? {{u|Sdkb}}talk 23:30, 29 August 2023 (UTC)
There's two things to consider here: A way of marking someone as away when questions go unanswered, and a way to get those questions noticed and answered by someone else. The former seems like easy to scan a users' contributions, but the latter needs to be considered as well. After all, if someone goes inactive and they get 2 or 3 new questions, I don't want them to be unintentionally discouraged by being "ignored". Panini!•🥪 03:58, 30 August 2023 (UTC)
Change help page associated with "Copyedit" task?
While looking at Special:EditGrowthConfig in the context just above, I noticed that each task has an associated projectspace page listed as Destination page for learning more about [task].
For the "Copyedit" task, this page is set to Wikipedia:Basic copyediting, which talks about spelling, grammar, and punctuation; but many of the tags that are associated with the "Copyedit" task ({{Tone}}, {{Advert}}, {{Peacock}}) are not addressed in the basic copyediting guidance, but rather Wikipedia:Writing better articles.
This explains why so many edits tagged as "Newcomer task: Copyedit" are mere changes in punctuation and wording, without addressing the maintenance tag. Is this a feature? or should we change the associated help page to give newcomers a better idea what we're expecting to see cleaned up with the tags listed above? Folly Mox (talk) 18:01, 5 September 2023 (UTC)
I would absolutely support swapping this one out for WP:BETTER! -- asilvering (talk) 22:58, 5 September 2023 (UTC)
Mentee Engagement Feedback
Hello, I'm currently signed up to be a mentor and was hoping I could share some feedback here. I have been somewhat disappointed in the results of the mentor system so far; there don't seem to be many tools for reaching out, engaging with, or taking responsibility for one's mentees. The framework for a great system is there, but I have personally had very little interaction. Most of the few questions I've received have been nonsense, and the few positive, recurring users that are my mentees haven't reached out to me in any way. It's anecdotal, but out of 130 assigned mentors I've received constructive questions or messages from two. I am able to handle a greater volume, and would like to be able to reach out and offer that to individuals. One idea I had was a slate of welcome templates that mention the mentee-mentor relationship. If I notice a new user, I'd like to have a welcome message I can leave that explains how I am their mentor and strongly encourages them to reach out to me and use me as a resource. Any changes that focus on increasing engagement would be good in my book. Fritzmann (message me) 19:23, 11 September 2023 (UTC)
Regarding welcome templates, there's {{Mentor welcome}}, as well as some from users, such as Panini!'s here (which I based mine and its variants off of). Happy editing, Perfect4th (talk) 05:40, 12 September 2023 (UTC)
Usefulness of "Add links" task?
According to Special:NewcomerTasksInfo, most task types have tens of thousands of articles available for improvement based on the tagging: copyedit, expand, and update each have around 30,000 articles; references has almost 160,000.
In contrast, the links task has five eligible articles. A single person could knock out the entire task for the entire project in half an hour or so. This suggests to me that the links task is not a good match for en.wp, and we should consider disabling it and optionally replacing it. Folly Mox (talk) 18:19, 4 October 2023 (UTC)
Hello Folly Mox, and thank you for bringing this topic to the table!
At the moment, Add a link is based on templates. If an article has a template that says "add more links", then it is suggested. As you discovered, it is not optimal.
We have a solution: we can change that task to switch from template-based suggested articles to machine-suggested links.
We have a step-by-step explaining how that feature works. You can also test it at test.wikipedia.org (but don't expect great suggestions though).
With this system, users are suggested links to add. They can add them, reject them or skip. The idea is to show users that they can edit. With the prediction models we have to suggest links, the number of suggested links is almost endless. We also have boundaries to limit the number of added links and the number of article one user can edit each day.
That version of Add a link was tested at our pilot wikis, and, base don their feedback, some improvements were made. Since then, that new version of Add a link has been deployed to ~85% of Wikipedias and replaces the one you have at English Wikipedia. We are deploying at the remaining ones; English Wikipedia would be the last one to get it, but it can happen sooner if needed.
User:Trizek (WMF), thanks for that link! I remember reading that with some interest awhile ago. I understand why en.wp would be the last project to enable it as well.You're probably aware of our style guideline MOS:OL, which advises against linking common terms. It's entirely unintuitive to new editors, who come here with the wholly reasonable expectation that articles link to one another as a core part of the wiki experience. I participated in a conversation recently with a newer editor confused about why their link to Tamil Nadu kept getting reverted.MOS:OL is well enforced, typically by User:Ohconfucius/script/Common Terms and its associated AWB module, but also manually. I think most of us are aware enough of WP:BITE to understand that reverting links to common terms added by low-edit accounts is not overall beneficial to the project, but it would be interesting to see what percentage of the AI suggested links were reverted outside the 48 hour window imposed by yall's methodology.At any rate, what I'm saying is that English Wikipedia's internal link culture is weird and maybe a significant outlier, in that we're more concerned with removing links than adding them. For practical purposes, I don't think enabling this feature would have a different effect on activation and retention than it has on other projects, but there might be more community pushback than anticipated. User:Ohconfucius/script/Common Terms.js seems to govern the broadest application of our MOS in this arena; if it could be parsed into an exclusion list then community concerns might be allayed.English Wikipedia is by no means fully internally connected. Without researching it, I'd guess that newer articles (age less than one year) are less well connected than they should be, particularly when they exist in another language project already.I'm not the biggest fan of MOS:OL, because why shouldn't we just link to articles we have, keeping in mind MOS:SEAOFBLUE? I've also cleaned up after the indicated script on a few articles where its insensitivity to context caused silly actions, like when it removed every link to People's Republic of China from History of China, or removed Finland from Finnish alcohol culture. But it has consensus and it's how things are here. Apologies if you already knew all this. Folly Mox (talk) 21:05, 4 October 2023 (UTC)
Thank you for the details, @Folly Mox. I was already aware of the outlines, but your detailed summary will help us preparing a potential deployment. Your link to Ohconfucius' list of terms will be of great help!
Add a link doesn't suggest random links. We have a model that ponders how many links can be included compared to the article length and to the number of already existing links (to avoid that sea-of-blue effect). It also excludes sections where links aren't needed (a list that can be edited by the community) and, of course, it doesn't create links to disambiguation pages. We can improve that model as we did for the wikis where we deployed Add a link previously.
Added links are less likely to be reverted, as you noticed; most reverts are due to the few users who don't read (or understand) instructions and then add nonsense links. It happens for any tool, no matter which instructions you put in from of users — humans are humans — and the only solution there is to explain again. The problem with these users is that they lead communities to highlight these bad edits, quite often with no comparison of all the benefits, and to blame the tool.
The revert rate outside of the 48 hours was so low that we haven't considered it.
Thank you again, Trizek_(WMF) (talk) 08:21, 5 October 2023 (UTC)
Mentorship change notifications
Mentees get a notification whenever their mentor is changed (e.g. when the previous mentor is removed after being blocked, becoming inactive, etc.). T336875 reduced the number of users getting these notifications somewhat, but I wonder whether we shouldn't consider going further. I just heard today from a new user who was confused about the notification he had automatically received after I removed his (blocked) mentor, and I'm sure there are many equally confused brand-new users we'll never hear from. Certainly there's some small benefit to letting users know about a change in mentor, but don't you think it's outweighed by the risk of confusion, frustration with unwanted notifications, etc.? Maybe the notifications could be limited to manual changes rather than the mass reassignments I do at Special:ManageMentors? Thanks. Extraordinary Writ (talk) 06:47, 10 October 2023 (UTC)
We try to balance the different point of view. I understand that getting a notification can cause confusion to some newcomers. I also understand the case where newcomers are happy to know that the mentor they already interacted with is no longer around.
We plan to improve the message text and tone, as the information in the current one is sometimes in the wrong order of importance (T327493T345635). We hope to reduce confusion there.
We also considered to have newcomers opting-in for these notifications, but checking on parameters is not something most newcomers do naturally.
I will document your idea of only notifying newcomers when a manual reassignment is performed, but I'd like to know your opinion regarding the options we have in hand first. :)
I think if newcomers are confused about having an initial mentor, and why they are assigned, then the messaging about this should be improved, and perhaps repeated when a mentor is changed. isaacl (talk) 14:18, 11 October 2023 (UTC)
I'm open to any suggestion regarding how the message is written! :) Trizek_(WMF) (talk) 14:30, 11 October 2023 (UTC)
Can you provide examples/pointers to the messages (preferably the entire message after having been assembled from its components) and context on when they are sent? It would be appreciated. isaacl (talk) 14:52, 11 October 2023 (UTC)
@Isaacl the notification says $USERNAME is your new mentor. Reason:$REASON (from these). The "reason" is whatever is in the log (examples here and here). — xaosfluxTalk 15:14, 11 October 2023 (UTC)
Thanks for the example. Is this message sent both when the mentor is initialy assigned and if it is changed? Is it a talk page message? If there is no additional message linking to more information about the mentorship program or the user's home page, then I can see why there is confusion. isaacl (talk) 15:19, 11 October 2023 (UTC)
@Isaacl it is an echo notification, not a talk page message. (I just tested it by changing the mentor/mentee relationship on one of my test accounts here). This is just when there is a relationship change, I'm not sure what is (or isn't) sent during initial load. It also says "Say hi to your new mentor" with a link to the new mentor's talk page. — xaosfluxTalk 15:36, 11 October 2023 (UTC)
Community Configuration 2.0 is a feature that will enable Wikimedia communities to easily customize and configure features to meet their unique needs. This approach provides non-technical moderators with more independence and control over enabling/disabling and customizing features for their communities.
Initial designs are drafted for two different approaches (see images). We will soon demo interactive prototypes to interested admins, stewards, and experienced editors (T346109). Please let us know if you have feedback on the design approach, or want to participate in prototype testing.
IP Masking
The Growth team has been working on several updates to ensure Growth maintained features will be compatible with future IP Masking changes. This work has included code changes to: Recent Changes (T343322), Echo notifications (T333531), the Thanks extension (T345679) and Mentorship (T341390).
Before December, the Growth team will initiate community discussions with the goal of migrating communities from Flow to DiscussionTools. This move aims to minimize the necessity for additional engineering work to make Flow compatible with IP Masking. (T346108)
Mentorship
We assembled some resources for mentors at Mediawiki wiki. This resource page is translatable and will be linked from the mentor dashboard.
We are working to resolve a bug related to mentors properly returning after being marked as "Away". (T347024)
We continue the deployment of the structured task "add a link" to all Wikipedias. We plan to scale the task to all Wikipedias that have link suggestions available by the end of 2023.
We plan to scale the new Impact Module to all Wikipedias soon, but first we are investigating a bug with the job that refreshes the Impact Module data. (T344428)
At some wikis, newcomers have access to the "add an image" structured task. This task suggests images that may be relevant to add to unillustrated articles. Newcomers at these wikis can now add images to unillustrated articles sections. (T345940) The wikis that have this task are listed under "Images recommendations" at the Growth team deployment table.
Other news
We disabled the “add an image” task temporarily (T345188) because there was a failure in the image suggestions pipeline (T345141). This is now fixed.
After a 2.5 years-long collaboration with Bangala Wikipedia, we have decided to start a collaboration with another wiki. Swahili Wikipedia is now a pilot wiki for Growth experiments.
I realize this isn't a very important bug, given that the feature is mostly intended for newcomers, but I've come across a small bug. The "impact" part only retains the 1000 last edits (even if more have been made in the last 60 days), leading to parts of the "recent activity" (including streaks) being missing after a few days of Special:RecentChanges patrolling. Again, not a big deal as I'm not really the intended public, just wanted to warn everyone that that was a thing. ChaotıċEnby(talk) 01:19, 3 November 2023 (UTC)
Suggested edits
Could we add "expand lead" to the "suggested edits" homepage widget? I just tagged an article with this and it occurred to me that this is a kind of edit that is extremely useful, has satisfying results for a new editor, and really isn't very difficult - you just need to read what the article says and summarize it, no outside research required. It seems appropriate for "medium" imo. -- asilvering (talk) 17:25, 5 September 2023 (UTC)
It looks like all that would be needed to do this would be to add {{Expand lead}} to the list of templates in the "Expand" task in Special:EditGrowthConfig. Folly Mox (talk) 17:51, 5 September 2023 (UTC)
Yes, though it would be even better if a new task was created altogether, since "expand" is down as "hard". imo this is a much easier task than, say, finding references. -- asilvering (talk) 17:54, 5 September 2023 (UTC)
Ah, good point. I hadn't noticed that because I got sidetracked with the thing I raised in the thread below. Folly Mox (talk) 18:07, 5 September 2023 (UTC)
Seconded. Not always so simple (things on Wikipedia never are...) but definitely an appropriate task for newcomers on the easier side. — Bilorv (talk) 19:33, 15 October 2023 (UTC)
Just... helplessly waving my arms at this thread in hopes an admin sees it and can change the configs... -- asilvering (talk) 16:07, 14 November 2023 (UTC)
Oh, I see that @Pppery was the last to edit this, so I hope you don't mind the ping here. Looks to me like no one objects to the idea, anyway... -- asilvering (talk) 16:10, 14 November 2023 (UTC)
If there is no objection at all, I can make the change. But I prefer to have a local admin to do the change. :) Trizek_(WMF) (talk) 17:32, 14 November 2023 (UTC)
In the Suggested Edits pane, images from articles are shown in inverted colour when using dark mode. Lotta blue faces. This can be fixed by adding class="mw-no-invert" inside the span tags that hold the image. Folly Mox (talk) 13:47, 13 November 2023 (UTC)
The community can decide which link is the most appropriate and change them in Special:EditGrowthConfig.
Re: dark mode - can you share a screenshot please @Folly Mox? Trizek_(WMF) (talk) 14:00, 13 November 2023 (UTC)
Oh, for sure on the community configuration element ☺️ I just don't know what would be a more appropriate venue for discussion than this talk page. Maybe this and a lot of the above thread should head to one of the village pumps. I'll try to get a screenshot uploaded. I'm sure it's easy. Folly Mox (talk) 14:13, 13 November 2023 (UTC)
Re: dark mode -- Thank you for the image. It is quite self-explainatory. :) I documented the problem, but as the dark mode is at the very first steps (prototype testing). No idea when the fix will be done, as this dark mode is not promoted to newcomers. Trizek_(WMF) (talk) 15:21, 13 November 2023 (UTC)
Sure; that's understandable. I don't even know if the images are enclosed in span tags to accept the class="mw-no-invert" parameter, although I assume they are.As to the question of community discussion, it's clear this page doesn't get that much traffic (99 watchers), so maybe after work today I'll raise the above concerns at WP:VPI or WP:VPWMF. I'll be sure to drop a link here if I do 🙏🏽 Folly Mox (talk) 17:14, 13 November 2023 (UTC)
I have the same reply for every community: it is up to you. :)
The most recent case I'm aware of is French Wikipedia. They decided to change things on their welcome template, so that the mentor greets the newcomer. Before they had two lists: greeters and mentors. My only concern regarding English Wikipedia is that only 50% of newcomers get a mentor (tracked at T323048).
Another community (can't find which one now, sorry) decided to mention the Homepage on their welcome template. The homepage is available for all new accounts.
I'm currently using a modified version of this welcome template to welcome new mentees. --ARoseWolf 17:51, 21 November 2023 (UTC)
Nice! I see a lot of similarities with the French standard one: mention of the mentor's name, link to the mentor's talk page with a clear invite to contact them, link to the Homepage... :)
I'm curious to know why your wiki has multiple templates, even ones created by users. Couldn't it be confusing for users? Trizek_(WMF) (talk) 18:54, 21 November 2023 (UTC)
I don't see how it could be confusing, as a new editor would only get one of them. As to why, I don't know the whole history, but as a volunteer project, I imagine it grew organically. I think it's fine to have a choice. Mathglot (talk) 19:05, 21 November 2023 (UTC)
Honestly we could probably update some of these or mark some as historical. There are a lot of links to essays I'd never heard of, and several welcome templates link Wikipedia:The Missing Manual, which was originally written in 2008. I'm not sure how updated it is kept. Folly Mox (talk) 21:42, 21 November 2023 (UTC)
Probably, but it's hard to pick which. I picked one at random among those I have never seen used ({{welcome-ql}}) and it has 67 uses, most not too recent (three from 2023). Does that qualify it for removal? Who knows. One-by-one Tfd's might do it, but based on what criteria? It might be a gnomey task for someone who wanted to analyze them all and report on the current situation, but tbh I don't care enough to do it. Mathglot (talk) 22:11, 21 November 2023 (UTC)
@Trizek (WMF) a lot of the different templates have specific uses. There's one I use to welcome new users and specifically direct them at WP:WIRED, ones for students editing as part of a class project, etc. I've never seen anyone get more than one, so I don't think this is likely to have ever confused anybody. -- asilvering (talk) 14:20, 22 November 2023 (UTC)
I see a few cases where having too specific templates could have side effects. Having a too specific topic Like "please add sources" would "lock" the user to add sources. I have not idea if it is true, it is a genuine question! :) Trizek_(WMF) (talk) 15:08, 22 November 2023 (UTC)
Hm, I'm not sure, and I'm not sure we have any way to measure that. I don't think that's likely to be a problem, because in that case what would usually happen is someone would hand out a generic welcome template and then also one of the warning templates. So you'd get Template:uw-nor1 if you did original research, and something like Template:Welcome as your welcome message. There are a few "problem user welcomes" that are much longer and discuss something like COI editing at some length, like Template:welcome-coi. They have most of the same links regardless, just with an additional focus on whatever the problem is. -- asilvering (talk) 20:04, 22 November 2023 (UTC)
Not sure I know what you mean by "lock" the user to add sources, but if it means pushing a new user in the direction of always using sources, then that would be a positive outcome, very much to be desired. Mathglot (talk) 20:25, 22 November 2023 (UTC)
Oh, I like this one! I've never tried making a template before but I might try to create one like this. (It would have to be an actual template for me to actually use it regularly, since I am lazy and would want it in Twinkle.) -- asilvering (talk) 14:27, 22 November 2023 (UTC)
I think that is yet another reason to have multiple welcome templates; if someone likes a particular one, they are more likely to use it to welcome someone, and that aids editor retention, or at least, that is the whole point of the project. Anything that promotes more welcoming is fine in my book. Mathglot (talk) 20:25, 22 November 2023 (UTC)
Speaking of templates, I notices that Template:Mentor welcome says " Around 10% of new Wikipedia accounts receive a mentor randomly taken from a list of volunteers.". We are at 50% now, and hopefully more soon if we manage to recruit mentors. Trizek_(WMF) (talk) 15:10, 22 November 2023 (UTC)
Happy to help! Trizek_(WMF) (talk) 19:02, 22 November 2023 (UTC)
Several articles are being assigned as newcomer tasks very frequently
By "very frequently", I mean a level of newcomer edits that is disruptive in the aggregate. The articles are Parental abuse by children and Alpha Omega Alpha. Aoidh figured out that if you select "Medicine and health", those are the only two articles suggested. (This was discovered because the Parental abuse by children was posted on WP:RFPPI and I don't blame the editor for requesting page protection.)
I don't know whether any other articles are similarly affected, but it seems like something needs to be changed to avoid pummeling articles like this.
As a temporary measure, I've removed the templates from each article that were placing them in the "copyedit" task, and also removed the most obvious trigger of the {{Advert}} tag at Alpha Omega Alpha.The cleanup tags were present since 2018 and 2019, so presumably a newly implemented topic area selector is responsible for the recent influx of people changing a word or two in the lead without really addressing the substance of the cleanup concern. Folly Mox (talk) 11:41, 10 November 2023 (UTC)
A few months ago, I asked here whether people thought it was beneficial to point newcomers looking to learn more about what the "copyedit" task involved to Wikipedia:Basic copyediting rather than Wikipedia:Writing better articles, but the thread only garnered one response.It is definitely true that the types of edits being made by newcomers for the "copyedit" task are generally positive or at least not harmful, and it's engaging new people and letting them learn the fundamentals of Wikipedia editing, even if they're too inexperienced to help resolve the cleanup tag to any real degree.I think that's a good outcome, but I have experienced some mild annoyance when a page I watchlist gets tagged for copyedit cleanup and then sees frequent edits by accounts with single digit editcounts changing a single conjunction in the lead, adding a definite article, or adding or altering one or two punctuation marks, as if that's the kind of copyediting we add a tag about.A knock-on effect of high levels of trivial engagement by new accounts on pages so tagged may be experienced editors performing necessary cleanup, not so much because of their own concern for article quality, but to quell the roil of insignificant minor alterations, so they're not clicking through their watchlists to the same pages several times daily to review the edits for errors.So I think this system is still beneficial, both for new editor engagement and for encyclopaedic quality, although the first may be driving the latter indirectly, and the experience for the moderating force in that driver (mildly annoyed experienced editors) could be improved. Folly Mox (talk) 11:59, 10 November 2023 (UTC)
Daniel Quinlan and Aoidh, Special:NewcomerTasksInfo suggests there are more copyedit honeypots in the topic areas of "comics and anime", "fashion", "performing arts", "food and drink" (one article each), "history" (two articles), and "central america" (three articles). The linked page doesn't have an easy way to tell what these articles are. Most of the other topics have either zero articles tagged or hundreds to thousands, with a few in the double digits. Folly Mox (talk) 12:13, 10 November 2023 (UTC)
I was able to determine which articles were listed where by checking "Display newcomer homepage" at the bottom of the "User Profile" section of Special:Preferences, and then clicking on my user page, where a "Homepage" link is found to the left of "User page" and "Talk". From there I can view suggested edits and see, for example, that List of Kaizoku Sentai Gokaiger characters is the only article listed when I select "Easy edits" for Comics and anime. - Aoidh (talk) 12:58, 10 November 2023 (UTC)
I'm not questioning the utility of the system. I'm simply concerned about the rate of newcomer edits on a subset of articles. It overwhelms the ability of the community to conduct peer review and give timely feedback. For example, this large edit on Parental abuse by children removed references, a substantial amount of text and formatting, etc.—it needed to be reviewed, but instead it was buried under more and more newcomer edits.
Would it be possible to simply not show categories with fewer than say twenty articles? Or perhaps sparse categories could be aggregated into a general catchall category?
Another option would be a more direct implementation of rate limiting. For example, if someone makes a newcomer edit on an article, don't show it again for 12 hours or some other period of time.
Side note: It seems like a lot of newcomer edits are being mismarked with the minor edit flag when the edit is definitely not a minor edit. Would it be possible to coach this a bit better in the newcomer system? Daniel Quinlan (talk) 21:45, 10 November 2023 (UTC)
I think I was the one questioning the utility of the system, and talked myself into an opinion about it.The issue we're seeing here seems tied in with the fact that the "copyedit" task (and currently inoperable "links" task, see § Usefulness of "Add links" task? above) is the only type of suggested edit that is assessed in difficulty as "easy" (for which see this ANI and this related thread on mw).It might be getting close to time for another community discussion about this: in my own estimation, the "update" task seems easier, but part of the problem comes from the maintenance templates used to include an article under the "copyedit" task. {{Inappropriate person}} and {{In-universe}} are easy to fix. {{Advert}} and especially {{Tone}} are much more difficult, and not generally suitable for newer editors. Splitting the copyedit task into tiers may be of use, although it would further reduce the number of articles suggested to very new editors looking for easy tasks.We could also potentially wait about it: Pppery's change of the help page associated with the copyedit task may help (which would be emotionally validating to me personally). Growth Team may also implement the AI-assisted version of the "links" tasks soon, which should give more easy edits.Another step any editor can take is, when we run across articles in the course of our gnoming that could use copyediting, be more liberal in our application of tags like {{Tone}}, especially if the articles are in the sparse–easy-edits topics, to spread out the newcomer edits more broadly, so they can be more easily reviewed. We could also do the reverse, and fix up any copyedit-task–eligible articles in sparse topics, to ensure that newcomers looking for "easy" tasks are always presented with many article choices or none at all. Folly Mox (talk) 17:50, 11 November 2023 (UTC)
Also, there are more cleanup tags we could add to the "copyedit" task in Special:EditGrowthConfig. {{essay-like}}, for example, is just a more specific (often more easily addressed) version of {{tone}}. Folly Mox (talk) 18:55, 11 November 2023 (UTC)
I've excluded very low-transclusion cleanup templates as being of limited utility. For example {{cleanup tense}} should always be an easy fix, but only has 27 transclusions.{{essay-like}} may not actually be "truly easy" (but likely easier than {{tone}}), and falls into a category of cleanup templates which require some knowledge of what we expect in terms of encyclopaedic tone, but are likely placed on articles in a worse initial state than {{tone}} might be. Similarly with {{story}} (256 transclusions), {{manual}} (492 transclusions), {{how-to}} (332 transclusions), {{over-quotation}} (1325 transclusions), and {{travel guide}} (478 transclusions).There are more templates available for other structured tasks too. In general I think we're underutilising en.wp's established cleanup template structure within the suggested edits feature, probably mainly due to lack of engagement by experienced editors. User:Trizek (WMF), is there an upper bound on the number of cleanup templates that can be associated with each structured task? Folly Mox (talk) 20:19, 11 November 2023 (UTC)
I don't think {{all plot}} is what you want here - it generally means more content on non-plot subjects need to be added, not that copyediting is required. I agree the others can be added to the existing copyedit task, but I'll hold off doing so myself for a day or two to allow for other comments. Creating a separate "easy copyedit" vs. "hard copyedit" task does not appear to be something that can be done via Special:EditGrowthConfig and would need a Phabricator task. * Pppery *it has begun... 20:37, 11 November 2023 (UTC)
I'm sure some of my selections are not actually easy, which is why the emphasis on easy-seeming. I probably also missed some genuinely easy tasks. I'm not proposing these templates be included without further discussion (my methodology was: hastily skim WP:TC while having a half-conversation with my roommate).I do think that additional structured tasks are locally configurable per-project, without phab input, but have to be defined manually in MediaWiki:NewcomerTasks.json rather than through the interface. Perhaps Trizek or xaosflux knows with greater certainty. Folly Mox (talk) 21:26, 11 November 2023 (UTC)
Last I checked, you can't just add them directly in to the json. It's been a while though. — xaosfluxTalk 22:54, 11 November 2023 (UTC)
I don't think "essay like" is a good newcomer task; often these articles need deletion or redirection to better articles, not copy editing. I've been around the block quite a few times, and often find myself scratching my head over what to do with them. Espresso Addict (talk) 02:20, 14 November 2023 (UTC)
I agree. I'm not sure deletion or redirection is that frequent of an outcome, but it is one of the more difficult tags to resolve. Daniel Quinlan (talk) 02:45, 14 November 2023 (UTC)
Getting zero edit accounts to make a single not entirely unhelpful copyedit to an article isn't a bad start to their editing career; I wonder if maybe part of the problem is that things like fixing typos and obviously incorrect grammar is just something we already do ourselves so instinctively that cleanup tags aren't a great fit for it. I almost think it might be better for the Suggested Edits goals for the "easy edit" copyedit task, if instead of using cleanup tags it just gave a random assortment of articles in subcats of Category:Start-Class articles (2,021) Folly Mox (talk) 02:53, 14 November 2023 (UTC)
Adding wikilinks is pretty harmless, as long as they are steered away from linking countries, linking all instances of a term or linking adjacent words. (Pity we ditched the "wikify" template.) Typo fixing too, as long as they don't start changing English flavo(u)r. I don't think anything else I've seen truly new users attempt provides them with a good experience. Removing sections of text will get substantial pushback, even where totally merited. Espresso Addict (talk) 03:37, 14 November 2023 (UTC)
Trizek has the unofficial wikilink blacklist, and assures us above that the AI-backed "link" task should be avoiding SEAOFBLUE. I agree that typo fixing is a fine first edit (it was mine, iirc, as an IP lo these many years ago). I do also think the five pages of guidance in the modal at the first clickthrough in the Suggested Edits flow should definitely mention ENGVAR, which is a super common stumbling block.Probably my own experience is just the cognitive dissonance of seeing people drawn to clean up an article without doing any article cleaning, like showing up to a garden party and alphabetising the buckets. The tight bunching associated with the sparse topics is properly a separate issue. Folly Mox (talk) 06:00, 14 November 2023 (UTC)
I just went and unearthed my first IP edits and they are mostly substantive additions (though my first edit after creating an account is a typo fix). I wonder if it would be more satisfying to suggest more creative work? Or would it just be off-puttingly hard? Could we suggest stub or start articles where there are usable free online sources (Guardian obits, f'rex, or RSoc memorials) that are already referenced in the article but have scarcely been used? Espresso Addict (talk) 11:34, 14 November 2023 (UTC)
@Folly Mox, we aren't using any sort of rejection links lists. The algorithm we use for Suggested link (which is not yet deployed at your wiki) has been proven reliable at the Wikipedias where it is deployed. When the deployment will happen, we will carefully observe the feature, and make the necessary fixes if any.
The other tasks are less structured, and they will be more opened to the creative work @Espresso Addict mentions.
The Editing team works on Edit check, to prompt users to add citations when they add a certain amount of text to an article. It is a work in progress that just started. Trizek_(WMF) (talk) 11:55, 14 November 2023 (UTC)
Is there anything we can do to help collate a heap of articles of the form I suggest? I come across them every day and don't have time to help them myself, and I'm sure other editors do too. Some sort of hidden category we could add? Espresso Addict (talk) 12:03, 14 November 2023 (UTC)
Yeah. My experience as a new editor trying to remove absolute obvious garbage from articles was to be instantly reverted. -- asilvering (talk) 13:07, 14 November 2023 (UTC)
User:Espresso Addict, that sounds like a computationally fuzzy problem. Guessing whether an already cited source is underutilised from a purely numeric standpoint would involve getting the prose length of external pages, and determining if it contains more information that would be appropriate for and is not yet present in the article citing it seems to me like such an advanced AI task that it might represent the point where human editors are obseleted by technology.User:Trizek (WMF), that sounds like a fair plan. I mentioned in the earlier thread how the WP:OL scripts lack contextual awareness, and trust the backend for Growth's "links" task is a more advanced piece of software than a regex engine. Since I've pinged you already, I had a couple questions elsewhere that got lost in the thread (mostly my fault for loquacity): what is the upper bound on templates that can be used to include an article in a structured task? and what is involved, technically, in adding a new task (for example if we wanted to subdivide "copyedit" into "easy" and "not easy" varieties, based on separate maintenance tags)? Third question, with apologies: have Growth ever considered adding functionality to allow maintenance categories to place articles in structured tasks? This seems like a natural marker for eligibility. Fourth question, which might be for someone at Machine Learning: is there a publically accessible script / API / gizmo that will surface an article's topic(s)? I'm presuming Structured Tasks uses the LiftWing "articletopic" model, but I don't understand the documentation. Folly Mox (talk) 13:02, 14 November 2023 (UTC)
Folly Mox -- I meant human curated; if every time I (and all the other dunnahowmany "experienced" editors) came across an article that fits the description one could pop on a hidden category, then there'd soon be thousands of them in the heap waiting for the new editors to work on. Espresso Addict (talk) 13:14, 14 November 2023 (UTC)
@Folly Mox, regarding Growth tasks, any template can be used anywhere (but not categories). We can't add new tasks, for now.
Maintenance categories are usually paired with a maintenance template. We relied on templates first, as they provide guidance. But we know the limitations there: categories are more versatile, and sometimes templates aren't helpful (full of jargon, with no action items nor ressources). At some wikis, there is no maintenance categories or templates. This is why we worked on Suggested links and Suggested images.
We considered to have a different way to tag the difficulty of a task (adding links could be 1-easy/ the link is suggested then you yes/no it, 2-medium/ the link is suggested then you search for the right one 3-hard/ you add links by yourself) but it wasn't raised as a priority issue.
with a possible 2024 project around Article creation for new editors (as more than 20% of newcomers create an account to create an article).
Regarding suggested links, the research for the model is on Meta, and it was published as a research paper. For Suggested images, the system is totally different, as we look for existing usages of the same image on the same article (or section) at other wikis.
@Espresso Addict, if your idea is to highlight sources that can be used to improve an article, it should be through a maintenance template. This way, anyone would know what to do, with the source handy.
I expect that most queues experienced contributors bang away against have maintenance categories with no associated maintenance template, but I don't have data for that. As an example (not for a newcomer task), I've recently started in on fixes in Category:CS1 errors: periodical ignored (25,974). This maintenance category has no associated cleanup template (appropriately), and repairing the error automatically removes the article from the category. Additionally, any article can be a member of as many maintenance categories as apply, whereas more than two-ish cleanup templates is typically seen as overkill. Anything based on a detectable error rather than editorial judgement is more likely than not to be in a maintenance category with no associated cleanup template. Folly Mox (talk) 18:52, 14 November 2023 (UTC)
We can offer tasks that aren't selected with templates. Suggested links and Suggested images are of this type: links and images are suggested based on prediction algorithms or usages at other wikis.
I agree with you : we have cases where templates are useful, and sometimes a category is sufficient; it is the case of detectable errors. I don't think they are suitable for newcomers though: often they are too complex.
The point of having a template is to tell something readers or other users. "This article lacks citations" is a clear warning, and Growth features can use this template to suggest to add a citation. However, the missing connexion are often instructions: the template doesn't clearly says what have to be done.
@Trizek (WMF), do you think it would be helpful to put a line like "Still not sure how to fix this? Ask at the Teahouse for help!" in there somewhere? We get "I don't understand this tag" or "can I remove this tag?" questions all the time there and it's really no problem. Plus, it can be encouraging to talk to another person. -- asilvering (talk) 21:57, 21 November 2023 (UTC)
@Asilvering, as it requires a significant effort to convey the meaning of a maintenance template and the actions required to remove it, redirecting users to a place where they can ask for more details makes perfect sense. It is worth testing it, at least with one template. Trizek_(WMF) (talk) 14:19, 22 November 2023 (UTC)
@Trizek (WMF) ah sorry, I meant at the end of one of the explanations for newcomer tasks, in the little pop-out window. (Sorry, I don't remember what this feature is called.) -- asilvering (talk) 14:31, 22 November 2023 (UTC)
Oh, you mean editing the message shown when a user works on a suggested edit, to encourage them to contact their mentor or the Teahouse at the end of the task? Trizek_(WMF) (talk) 15:00, 22 November 2023 (UTC)
@Trizek (WMF) Yeah, that one. We've definitely got a few people at the Teahouse explicitly saying something like "the tutorial thing told me to copy edit this page but it looks fine, I don't understand what I'm supposed to do, help?" So I assume there are a lot of other people who have the same experience but don't actually come looking for help and just stay confused. -- asilvering (talk) 19:57, 22 November 2023 (UTC)
Right. As you can have users who will not go through Suggested edits to fix something, they might not see the edited message. What about adding a line on the template itself "please remove me when you fixed the issue"? Trizek_(WMF) (talk) 20:04, 22 November 2023 (UTC)
@Trizek (WMF) I think they all have that already. They link to H:MTR. The additional confusion arises, I think, because the newcomer homepage or suggested edits window is telling new users to edit those pages. We know that the maintenance templates might be dated and that no human curator is making a list of "newcomer articles" specifically to be featured in those widgetss, but new users don't know that, or at any rate often act like they think these are specifically created newcomer tasks. So it's an issue with the newcomer homepage / suggested edits more than the maint tags themselves. -- asilvering (talk) 20:09, 22 November 2023 (UTC)
Got it. I re-checked a few templates, and there is the link you mention: "(template removal help)". Honestly, I didn't paid attention to it until you mention it. Maybe I was looking at something more descriptive, like "How can I remove this banner?".
If more people are okay, we can consider the change. I advise to start by counting the number of requests regarding template removal, and at a defined date experiment one change. Then, you compare the number of requests, as way to measure if the change had an impact.
That was recently changed from "Learn how and when to remove this template banner" per a request at Module talk:Message box. Feel free to discuss there. * Pppery *it has begun... 16:31, 23 November 2023 (UTC)
Oh yikes, I don't like that change at all, it's clear as mud now. Thanks for the link. -- asilvering (talk) 16:44, 23 November 2023 (UTC)
Oh, I forgot to mention: there have been several recent discussions about the "minor edit" flag, and consensus seems to be that newcomers marking non-minor edits as minor is the correct use case for the modern meaning of "minor edit": "requires review". Folly Mox (talk) 17:54, 11 November 2023 (UTC)
I have created {{no newcomer task}} and configured all newcomer tasks to exclude articles from it, since that seems like a better solution than removing valid maintenance tags as Folly Mox did to Parental abuse by children. * Pppery *it has begun... 20:45, 11 November 2023 (UTC)
Thank you. I was aware the action was suboptimal and not a real solution when I made the choice. Folly Mox (talk) 21:18, 11 November 2023 (UTC)
Add a link, a guided task for newcomers.Permalink for the Noticeboard request. Apparently, the request was a "mixed bag", to quote one of the comments. I agree on the fact that some edits aren't perfect, or are of the vandalism sort. But they have to be compared to the reste of edits who are apparently okay.
Regarding added links (related to #Usefulness of "Add links" task?), we have a solution: we can set boundaries by suggesting appropriate links, with a yes/no choice. The goal is to encourage newcomers to edit first, and then go deeper. We plan to deploy it at English Wikipedia at the beginning of 2024. Trizek_(WMF) (talk) 13:58, 13 November 2023 (UTC)
My apologies if I've made you feel like this is your problem to solve. I just thought you'd be a good person to loop in. Folly Mox (talk) 14:45, 13 November 2023 (UTC)
@Folly Mox, you are right: I'm the right person to ping! :) We want to understand the incident, like we don't want our features to cause frustration at any community.
I feel that case is a misuse of the tools, by not following the prompts, more than a technical issue. We always have had, and will always have problems with editing tools and users coming with an agenda; it is true since we have an edit button. :) Trizek_(WMF) (talk) 11:48, 14 November 2023 (UTC)
As to the interpretation of the ANI, Bob (the IPv6, an experienced contributor) was trying to bring awareness to and gain understanding about the tight bunching of edits by low-edit accounts, having previously spoken with a functionary on his talkpage about it. The first instinct of en.wp admins (like the functionary previously contacted and Uncle G in the ANI), was to check for malicious intent: UPE, sockfarm, GAMING user access levels, etc.I think I've finally managed to distill what I think the actual experience of the problem feels like: the prompts the newcomers may be following (1, 2, although in fairness see also 3) do not align with the type of edits requested at the article (This article contains content that is written like an advertisement). This seems to encourage unnecessary / potentially unwanted types of copyedits. Folly Mox (talk) 13:25, 14 November 2023 (UTC)
I see. I'm not trying to put the problem on someone else. I'm just looking for the most common answer.
My first call there is: we have always seen people not respecting the rules. You named a few: UPE, sockfarm, GAMING user, etc. Any point of entry will be used by these users to achieve their goals.
Then we can have the case of users getting a task that is not suitable for them. With time and experience across wikis, I identified a few reasons why a task is not done the right way:
the user don't read the instructions (90% of cases!)
the user don't understand the instructions
instructions haven't been translated in the wiki's language (not a concern here, but I give you all reasons!)
the user goes beyond what is expected (they add sources on a add a link task)
the instructions given in the task aren't the same as on the maintenance template
a template added to suggest the edit is not the right one (you select the copyedit task and the maintenance template is all about sources)
My advice: have a good look at the templates associated with the tasks that were quoted at ANI as some might fall under the reasons I listed.
I agree this is (mostly) an individual wiki administration problem and not a WMF problem. I'm not really sure how best to solve the individual wiki administration problem, though, or even where to start a discussion about it. Personally, I think the difficultly/approachability level of many of the tasks is out of step with what the newcomer module uses (tasks listed as "hard" that are relatively easy and vice-versa), so I'm wondering how the categories were arrived at in the first place? -- asilvering (talk) 16:04, 14 November 2023 (UTC)
Categorization of difficulty was based on several community conversations we had with a handful of wikis. We knew that they were imperfect classification, but it was better than providing a pot-pourri of tasks with no clue on where to start.
Copyediting is about making a small fix to the way an article is written, and it is a valuable and easy way to get used to editing Wikipedia. Copyedits help articles be more professional and trustworthy.
Maybe some communities put more than just that under it?
I think copy editing is quite a difficult task, that requires knowledge of many different Wikipedia guidelines in all but the most obvious of cases. I say that as someone who has copyedited professionally. -- asilvering (talk) 19:04, 14 November 2023 (UTC)
asilvering, have you read this discussion from May? This was the last major conversation I am aware of that we had about the difficulty levels of various Suggested Edits. Fortunately we have a lot more granularity in terms of configurability than I think people recognised at the time. I'm thinking our job now might be to find the cleanup templates that are actually easily resolved, and/or actually represent requests for minor copyediting, but it's just one idea. Folly Mox (talk) 19:02, 14 November 2023 (UTC)
I think that's a very good idea, @Folly Mox. I wasn't aware of that discussion in particular but I did know from previous discussions on this talk page that it is a matter of selecting individual maintenance tags (at least, when it comes to giving new users copy editing tasks that are actually approachable for them). I asked Trizek how the categories were arrived at in the first place because I wanted to know if there was any fact-gathering work that was done beyond community conversations (since it would be pointless to do that work). Personally, I think community discussion is not a good way forward in this case, since the community at large is broadly hostile to newcomers. A more focused discussion in a place that self-selects for people who are (or at least try to be) less hostile, like this talk page for example, is still, well, people who aren't newcomers mostly talking about what they imagine being a newcomer is like, so probably full of blind spots. What I'd really like is to have survey data. In a hypothetical universe in which I was running one, I'd round up a bunch of month-or-so-old newbies and have them rank various maint tag tasks from 1-10 on three scales: "For someone in their first week of wikipedia editing: 1) how difficult do you think this task is? 2) how daunting do you think this task is? 3) how clear is it? ie, is it evident to a new user what needs to be done?" -- asilvering (talk) 19:20, 14 November 2023 (UTC)
I don't think editors tag for minor copy editing, they either just do it, or don't bother tagging. Things I've seen tagged for copy editing tend to be where the article is either a translation or something written by a non-native English speaker; the latter class can be close to unintelligible; what is wanted is rewriting, not copy editing. The former class it is often worth identifying the source article when there's any ambiguity. Neither of these scenarios seem positive for newcomers, particularly if they don't have copy editing or language skills already.
Agree it is difficult to imagine being a newcomer. My first experiences were 17 years ago, and I had worked in publishing and scientific writing for years beforehand. A survey sounds useful but month-old editors would be selecting for those who stuck at it, rather than walking away when the tasks presented were inappropriate. I'd suggest a 4th question, how enjoyable or satisfying did you find it? Because fundamentally, we got hooked on Wikipedia because it was fun, hard though that is to remember sometimes. Espresso Addict (talk) 23:24, 14 November 2023 (UTC)
One of my first experiences as a newcomer was posting to this very talk page, since the newcomer homepage gave me copyediting as an easy task, and I found it both bewildering and demoralizing. -- asilvering (talk) 00:47, 15 November 2023 (UTC)
I'm inclined to agree that it's altogether possible no maintenance template is appropriately addressable by a brand new editor. Maybe trying to find the "truly easy" ones is uh folly. Maybe it would be better just to cast a wide net ({{cleanup}} has 10377 transclusions). This won't help guide newcomers towards making genuinely helpful improvements, but it should at least reduce the tight bunching that led to this thread and the ANI.And if there's no cleanup template that is appropriately addressable by a brand new account, and no cleanup tag specifically just for minor copyediting, why limit ourselves to cleanup templates for the "copyedit" task? Why not just use {{unreferenced}}? Those articles are plentiful, usually low-traffic, almost never in any kind of refined state, and most have probably been content-stable for a years without improvement.As a side benefit, any experienced contributor who arrives to clean up after misguided good faith newcomer edits is pretty much going to have to add a reference or AfD the article as a matter of principle.asilvering, that's disquieting, about your own experience as a newcomer. I'm glad you stuck around. Folly Mox (talk) 03:01, 15 November 2023 (UTC)
Setting new editors on unreferenced is an interesting notion. In my experience many of them fall into the translation camp and could use copy editing, and at least no one is going to complain about a newbie inadvertently changing the reference styles. And they're usually short, which is probably a bonus. On the other hand, they're not much watched, so even a truly excellent edit isn't going to get the thanks/interaction it might in higher-traffic areas (says she after having worked there intermittently sans communication from anyone).
One suggestion is some sort of bot assessment of all newbie edits, from vandalism to excellent, coupled with encouraging established editors to look at the bot-graded excellent edits and thank/greet/advise the newbie as appropriate. Espresso Addict (talk) 03:16, 15 November 2023 (UTC)
Unreferenced is already one of the tasks. The full list of the ones we use on en-wiki is at Special:EditGrowthConfig. -- asilvering (talk) 05:12, 15 November 2023 (UTC)
I'm aware we already use {{unreferenced}} for the references task. I have the impression we can use it for copyedit as well. I could be wrong, but it seems like the individual tasks probably don't care what the others have in their templates arrays.I admit something I hadn't thought of is the end user experience, when they've finally made enough copyedits that they feel comfortable enabling "Find references", only to see the "articles found" counter fail to increment, as if Wikipedia has no unreferenced articles. We could ameliorate this somewhat by excluding {{unreferenced section}} from copyedit, which we should anyway. We could also change the templates for the {{reference}} task adding for example {{more citations needed}}, {{more citations needed section}}, {{non-primary source needed}}, {{primary sources}}, {{one source}}, {{better source needed}}, {{unreliable source}}, and {{citation needed}}. Folly Mox (talk) 06:05, 15 November 2023 (UTC)
The copyedit templates were also discussed here in August 2022 (where a newcomer described the copyedit task as "overwhelming"), but I don't think any action was taken at that time. Perfect4th (talk) 15:57, 16 November 2023 (UTC)
And by me the year before: [7]. I wonder who will bring it up in 2024! -- asilvering (talk) 01:32, 17 November 2023 (UTC)
I notice even that the particular article I mentioned, Ichabod Crane, is still being targeted, and is in no better shape than it was two years ago. -- asilvering (talk) 01:34, 17 November 2023 (UTC)
I was surprised today at an edit summary to the article Guru Gopinath Natana Gramam (where I converted an IP's G11 to a prod) where the newish (~70 edits) editor stated "Major grammar changes to save this article" while making a few cosmetic edits. These articles may present too much difficulty for a new editor (I tried salvaging Guru Gopinath Natana Gramam but gave up and prodded it instead). Espresso Addict (talk) 01:51, 17 November 2023 (UTC)
Yep. I learned how AfD worked in my first week here for this reason. Here's wiki-baby me hand-coding an AfD nom (no Twinkle!) on my second day of editing, prompted, as the edit summary tag makes clear, by the newcomer homepage widget: [8]. I wouldn't say it's "too difficult", but I would certainly say it's "not helpful". Though, I did stick around, so maybe there's some usefulness to throwing people into the deep end off the word go. -- asilvering (talk) 02:05, 17 November 2023 (UTC)
Deleted edits disappear, which I would have thought was a bit discouraging. As a tangent, I think it would be really useful to try to gather editors like you, Asilvering, who have joined since the newcomer project was initiated on enwiki, and try to understand what factors are associated with successful recruitment while most other new editors' contributions soon dry up. (Background? Character traits [persistence, obstinacy, robust self-belief...]? Something to do with the newcomer tasks suggested? A burning urge to contribute on a topic? Or just plain luck?) Espresso Addict (talk) 02:32, 17 November 2023 (UTC)
Personally, I think the kind of person who devotes their free time to editing an open-source encyclopedia is a rare sort of person (judging by the people who are face-out on this site, a disproportionate number of us are academics, librarians, programmers, and/or retirees), and so it isn't so much an issue of converting newcomers into that kind of person, so much as ensuring that the newcomers who are don't get turned off and leave. That's a much harder problem to fix, unfortunately. I will say that being here got immediately nicer the moment I became extended-confirmed, to the point that whenever anyone mentions editing wikipedia, I try to sign them up on the spot, even if they have no intention of editing in that moment - the sooner that 30-day timer runs out, the better. I suspected this was a problem almost immediately upon beginning to edit, received confirmation of that bias also quite early (see: [9]), and was the very next day accused (by an admin!) of being a sockpuppet. Not a very welcoming experience! I stayed for a few reasons, one of which was the kind reception I got on this very talk page. I won't deny (it's quite clear in my edit history) that another major reason was that I discovered shortdesc helper and hotcat. Knowing that I could escape noob hell in 30 days (rather than the year+ it would have taken me to make the more meaningful edits I had signed up to make) made it more bearable. -- asilvering (talk) 03:12, 17 November 2023 (UTC)
I agree. I try hard to help that kind of newbie editor, but by the time I encounter them (usually in the wake of a G11/A7 speedy on an academic, though I've had a series of clashes with the AfC folk over academics as well) they're often so turned off that I can't recover the situation. The default reception for anyone whose first edits look competent used to be "hail and welcome" and is now "eff off, paid spammer". Espresso Addict (talk) 03:24, 17 November 2023 (UTC)
The other option is to promote drive-by tagging. Maybe make it into a newcomer task: check random pages and see if they are too promotional, have too few references, have bad prose, and other easy-to-identify problems. Sungodtemple (talk • contribs) 13:35, 15 November 2023 (UTC)
There is a technical way to make certain elements visible only to certain user permissions groups, and the idea of big, approachable versions of cleanup templates only visible to accounts without autoconfirmed (the default displaying to logged out readers and all other user access levels), is an idea.I don't think promoting driveby tagging (a bad practice) to new accounts is a healthy choice, and also we can't currently add newcomer tasks (as I mistakenly thought until earlier this thread).Above, Espresso Addict suggested a curated selection of newcomer-edit-friendly articles, and this idea I feel falls into the same sphere, of a not quite walled garden for new editors to edit in, without the potential to break anything important. My crazy idea above to change the copyedit template to {{unreferenced}} is similar in that respect, now that I think about it. Folly Mox (talk) 14:03, 15 November 2023 (UTC)
With my admin hat on, I'd strongly suggest not promoting a newcomer task of tagging; new editors in my experience make absolute havoc of the various types of deletion, and there's no better way of annoying an article creator than driveby tagging.
On the walled-garden question, that's a good point. My intention was to create a fairly safe but useful space in which newbies might have a satisfying experience for their earliest edits, but obviously at some point the new editor must be encouraged to edit more broadly and confront the dragons long-established editors. Lots of satisfying easy tasks that come to my mind actually require library access, and so are not possible for newcomers (unfortunately).
I also keep thinking whatever happened to WikiProjects? Certainly that's where I learned the ropes, and my back-of-the envelope is that active newcomer-friendly projects such as Women in Red are the most valuable training ground. Espresso Addict (talk) 00:24, 16 November 2023 (UTC)
I'm confused by the assumption that newcomers don't have library access? Loads of wikipedia editors are current university students (or staff...) and so have access to academic libraries. Public libraries also allow access to all kinds of academic and newspaper databases. Plus there's always the Internet Archive and libgen. -- asilvering (talk) 03:21, 16 November 2023 (UTC)
Sorry, Asilvering, I meant specifically the WP:Wikipedia Library. Obviously new editors may well have access to all sorts of content via academic libraries or public libraries, but one can't assume that they have, in the way one can assume that any established editor does. Espresso Addict (talk) 03:45, 16 November 2023 (UTC)
Er, what easy tasks require Wikipedia Library access? -- asilvering (talk) 03:50, 16 November 2023 (UTC)
Almost anything involving seeking sources benefits greatly from WL access, as would, say, suggesting that slightly more experienced editors might start to flesh out existing stub biographies from ODNB, Grove, or Times/NYT obits, &c&c. Espresso Addict (talk) 04:18, 16 November 2023 (UTC)
I really don't think this is a relevant concern, given the numerous possibilities for access outside of the wikipedia library. -- asilvering (talk) 04:21, 16 November 2023 (UTC)
This first meeting language will be English, but we plan to host conversations in other languages, and about other topics. Please visit the conversation page on-wiki for the details on how to join. You can also watch the page, or suggest ideas for upcoming conversations there.
Impact Module
At the beginning of November 2023, the Growth team deployed the New Impact Module to all Wikipedias. We recently released a follow up improvement to how edit data was displayed based on editor feedback. [10]
Developers can find some initial proof of concept code shared on gitlab.
Mentorship
When a mentor marked themselves as "Away", they were not getting their name assigned to new accounts when they returned. This has been fixed. [14]
We improved the message received by newcomers when their mentor quits, to reduce confusion. [15]
We worked on ensuring that all mentees are assigned to an active mentor. This required reassigning mentees with no mentors to a new mentor. We paused this as the clean-up script confused some editors. We will resume it when the identified blockers are resolved. [16]
It is now possible to create an Abuse Filter to prevent one user from signing up as a mentor. [17]
@MMiller (WMF): Perhaps the following is not one for the growth team, but could you point it in the right place?
I think it's long been a problem that talk pages can be extremely verbose, especially on issues on which many people are interested. This sorta limits future participation to either editors who are able to read through all existing content, or those who don't bother read existing content. I think the phenomenon tends to discourage talk page participation.
I wonder if it's possible to use LLMs to summarise talk page sections, and integrate this into Wikimedia, so that there's something like a "summarise" button at the top-left of each section? I presume you'd not want to plug into the OpenAI API, but there are open source LLMs these days. ProcrastinatingReader (talk) 17:39, 5 December 2023 (UTC)
Kinda reminds me of the old abandoned Summary: namespace. Seems like doing it at runtime would solve some of the old problems, while creating less serious ones. I do agree that exceptionally long conversations are a turn-off from additional participantion, at least for me and my slow phone. Folly Mox (talk) 18:53, 5 December 2023 (UTC)
Hi @ProcrastinatingReader and @Folly Mox -- thanks for getting in touch about this idea; we've heard it a couple times from a few different places. I've experienced the challenge of digesting a huge talk page, too -- like especially a long RfC. I'm pinging my colleague @MPinchuk (WMF), who is the WMF product manager thinking most directly about using LLMs. While we wait for her response, could you tell me more about this idea? I think it would be wonderful if it worked well, but I also can imagine ways it can go wrong. Do you think editors will trust an automated summarization to capture the nuance of a long discussion in a balanced way? MMiller (WMF) (talk) 19:41, 5 December 2023 (UTC)
FWIW, I also think there could be a lot of benefits to summarizing discussions, but we'd probably need Wikipedians experienced in wading through long RfCs to help us evaluate the quality of the output... so be careful what you wish for, lest it comes back to me asking you to help check the diffs (ETA: thinking about this one situation in particular, which I've talked with @L235 about...)Maryana Pinchuk (WMF) (talk) 00:24, 6 December 2023 (UTC)
@MMiller (WMF) and MPinchuk (WMF): elaborating a bit more: I think discussions vary in terms of parsability. eg: this section so far is not terribly long and can be skim-read quickly. Conversely, there's also discussions that are really long and repetitive. Immediately before posting this I was reading an archived talk discussion in which someone wrote [these] discussions have generated more than 500 kilobytes of text. Or put another way, that's roughly 140 pages at a standard 12 pt font. ... Having dumped an unreasonable amount of my time and energy into this during the last two weeks, I am going to take a wikibreak for a while.
Some examples, along with examples of how LLMs could be used appropriately:
Talk page discussions on highly visited pages or RfCs on matters that many editors are interested in. Comments are often repetitive, and most comments tend to span just 2-3 major themes. Once someone gets good at skimming, it's easy enough to ascertain these themes in a well-structured discussion. Unstructed discussions can be harder to parse.For instance, Talk:Donald_Trump#Include_mention_of_Abraham_Accords_in_opening_section is a section with 127 comments. If I'm an editor wanting to opine on the issue, it'd be nice to get a summary of what's under discussion and what the main arguments are so far. An RfC closer would need to read the actual comments rather than rely on an LLM, but for the people entering the discussion an LLM summary is probably sufficient. (Realistically, anyone entering now, assuming they even want to post in it, likely wouldn't read all the existing content anyway.)
Summarising archived discussions. It's quite typical to need to refer to discussions in the archives to get an idea for what was discussed.
Complex disputes at WP:ANI. Once it gets too long, few editors not already involved in the discussions will want to get involved. Usually the length is not due to lots of new material, it's due to people going back and forth in circles. These are often very summarisable.
High-level (two-sentence) summaries of all ongoing discussions on talk pages with lots of sections/ongoing discussions.
Overall: I imagine cumulatively reading talk discussions probably takes up a disproportionate amount of editing time, and for those more new to editing, it's probably very off-putting. So I think LLM usage on talk pages would make editors more productive, and make entering ongoing discussions more accessible (to newer editors, to those with less time to contribute, those with slower reading abilities, etc)
Do you think editors will trust an automated summarization to capture the nuance of a long discussion in a balanced way? I suspect it depends on how it's used and the quality of the predictions. If it's used by RfC closers to close discussions, or admins to decide the outcome of ANI discussions and levy sanctions, or bureaucrats when closing WP:RFAs, then I think people would object, regardless of the quality of summarisation. Contrarily, if the tool is used by editors to summarise archived discussions, I can't imagine anyone would object and think people would be thankful it exists. I'm less sure how summarising ongoing discussions would be received; could be uncontroversial or not. If this is something the team would be able to dedicate effort towards then I could ping a few people here to get thoughts. ProcrastinatingReader (talk) 13:10, 6 December 2023 (UTC)
PR has clearly put more thought into this than I have, and I agree with most everything just above. As to nuance specifically, it will be inappropriate for people to use AI tools to close discussions, but human laziness and excess trust towards software will ensure it happens eventually. We recently had an RfC close successfully overturned because the closer admitted to not having read the entire enormous discussion, and left a paltry closing statement that did not adequately summarise people's discussion points.To the question of prioritisation, I would say that the en.wp community does have use cases for AI discussion summarisation already (see above), and not for article summaries. Also, I feel like I read somewhere that LLMs have already read all of Wikipedia as a training corpus, and examining their ability to summarise articles might not give great data due to interactions with their priors. The project and talk namespaces are likely a bit better as targets for new research, especially as conducted by the Foundation. Folly Mox (talk) 13:54, 6 December 2023 (UTC)
@Folly Mox & @ProcrastinatingReader Chiming in to the discussion as Maryana linked to our ongoing work on summarization in the WMF Research Team. The question of using LLMs for summarization has come up in several places recently; for example, I took part in some of the discussions at this year's Wikimedia Hackathon (see, e.g., this blogpost). While I generally believe that LLMs could work well in those tasks, one of the main challenges is that it is very challenging to systematically evaluate when the model is doing a good or bad job in a setting like this (lets say summarizing talk pages). Among others (and in addition to what has been mentioned by others here) the task is perhaps not very well-defined, we dont have readily available ground-truth datasets for modeling, etc. Therefore, in our research we started to approach a different but similar problem: can we use LLMs for automatic text simplification. The idea is to make existing content more accessible to readers with varying levels of education/literacy. The advantage is that the task of simplification is pretty well-scoped (there is at least some agreement of what makes an article easier to read) and there are some agreed-upon recipes how to evaluate with benchmark datasets. This will help us better understand what these models are capable of and figure out some of the technical challenges. Ideally, the learnings will then help us approach other use-cases around summarization (such as talk pages) as the underlying models are very similar; though each of them brining other challenges as described above. If you have thoughts around that project feel free to also follow up on the talk page of the research project on meta. MGerlach (WMF) (talk) 18:05, 15 December 2023 (UTC)
I would caution the exploration of this idea. LLMs cannot ensure factual accuracy, even when rewriting existing factual material. There's a lot of jargon that an LLM would likely struggle to parse (e.g. when someone says "per WP:OR", another "per SYNTH", another "per original research", how does it know that's the same thing three times?). And I think conflict and maximalist discussion would follow from editors who believe they are misinterpreted. (I can already see the comments reading: "Note: as of this revision, the LLM summary mistakenly says X when I never said that—what I said was [another 200 words]".) — Bilorv (talk) 21:40, 10 December 2023 (UTC)
Quit mentorship but still receiving questions
I quit my mentorship on December 10th, but for some reason I am still receiving mentorship questions [18], [19], [20]. ––FormalDude(talk) 21:18, 24 December 2023 (UTC)
According to this, you are still assigned a number of mentees. May be related to phab:T330071; maybe see also phab:T351464? Pppery probably knows. Folly Mox (talk) 22:34, 24 December 2023 (UTC)
No idea what caused this, but rejoining and quitting again may help. When the WMF gets around to finishing phab:T330071 any lingering mentor assignments will be cleared, and phab:T351464 from November is not likely to be relevant. Whatever bug caused this looks inconsistently reproducible, since other mentors have quit later than you and had their mentees properly reassigned. * Pppery *it has begun... 23:58, 24 December 2023 (UTC)
Thanks, I just tried rejoining and quitting again. ––FormalDude(talk) 00:15, 25 December 2023 (UTC)
Ah, darn. Sounds like I just need to wait out phab:T330071 then, right? ––FormalDude(talk) 19:29, 25 December 2023 (UTC)
@FormalDude Sorry to hear about this issue! Extraordinary Writ's hypothesis sounds quite plausible. I've pinged an engineer on T330071, so hopefully we can determine if it's the same issue or something unrelated.
Thanks for being a mentor previously, and sorry to hear you were receiving so many questions unrelated to editing. Please let us know if you have any suggestions for further improvements we should consider for the Mentorship tools or improvements to new editor onboarding to limit the number of basic / repetitive questions Mentors end of fielding. Thanks, - KStoller-WMF (talk) 18:21, 2 January 2024 (UTC)
@FormalDude A Growth team engineer investigated and unfortunately it appears there is a bug in the reassignment system. Indeed, it seems to relate to having so many mentees that the job to reassign mentees doesn't finish in the allotted time. We've logged a task to fix the issue here: ReassignMenteesJob is not able to finish in time when a mentor has too many mentees assigned (T354222).
And we are working on a one-off fix for reassigning your mentees (T354220) so that you don't need to wait until the underlying bug is fixed to officially step away from Mentorship.
Thanks for reporting the issue, and for all the time you spent helping mentor newer editors! - KStoller-WMF (talk) 22:05, 2 January 2024 (UTC)
Thank you @KStoller-WMF! My only feedback would maybe be to make it more clear that mentors are for help with editing Wikipedia. It looks like mentees see a button that says "Ask your mentor a question", perhaps this could be changed to "Ask your mentor a question about editing". It may also be helpful to specify in the prompt that mentors can only help with questions about Wikipedia editing, and possibly give an example in the placeholder text. ––FormalDude(talk) 23:54, 2 January 2024 (UTC)
I'd support that change – I've run into this issue some as well. Perfect4th (talk) 04:03, 3 January 2024 (UTC)
I would recommend adding another sentence to the end of the prompt that says something like Your mentor is only able to answer questions about Wikipedia editing. For the placeholder text, you could keep the current wording and add an example, so: Say hello and ask your question. E.g. How do I create a citation? ––FormalDude(talk) 05:11, 3 January 2024 (UTC)
Personally, with the first modification in place, I don't think it's necessary to add another sentence saying that an English Wikipedia mentor is only able to answer questions about Wikipedia editing. It starts to feel like talking down to the editor in question. isaacl (talk) 06:32, 3 January 2024 (UTC)
I don't think basic instructions constitute talking down to someone. There doesn't appear to be a ton of guidance to mentees on what mentors are actually for, so I really don't think adding some would hurt. Also, have you ever participated as a mentor? ––FormalDude(talk) 10:05, 3 January 2024 (UTC)
I agree with asilvering that someone who is asking non Wikipedia-related questions to their English Wikipedia mentor, and isn't following the prompt to "Ask your mentor about editing" isn't likely to be dissuaded by a sentence saying that the mentor is only able to answer questions about Wikipedia editing. isaacl (talk) 19:48, 3 January 2024 (UTC)
The added guidance may not dissuade the majority of people who ask off-topic questions, but if it can prevent even a few people, it is worthwhile. ––FormalDude(talk) 07:45, 4 January 2024 (UTC)
I'd agree with this too: banner blindness and related phenomena mean that a person is more likely to ignore text the longer it is. Bad questions are caused by low internet, media or English-language literacy, all of which are worsened by longer text. — Bilorv (talk) 21:21, 3 January 2024 (UTC)
I do like the idea of adding a short example question, though. "How do I create a citation" is a pretty good one. Clear and short. -- asilvering (talk) 23:12, 3 January 2024 (UTC)
Added the example since that seems harmless, although that field only accepts pure text so the italic formatting you used wouldn't get rendered properly. Will let the discussion above resolve before editing the prompt, although I'm inclined to agree with FormalDude that the guidance should be edited, although I'm not, and have never been, a mentor either. * Pppery *it has begun... 17:39, 3 January 2024 (UTC)
Maybe change the text in the help panel that appears on articles correspondingly? Not to be nitpicky, but I think something like "Ask your mentor about editing" would be more concise. Sungodtemple (talk • contribs) 19:27, 3 January 2024 (UTC)
Done For future reference see {{Edit protected}} if you need to get the attention of an admin. * Pppery * it has begun... 20:54, 14 November 2023 (UTC)
(ETA: thinking about this one situation in particular, which I've talked with @L235 about...) Maryana Pinchuk (WMF) (talk) 00:24, 6 December 2023 (UTC)
