Wikipedia talk:Growth Team features#retention

This might not be the correct place and I'm sorry if it's not but I've tried looking everywhere and can't seem to find anyplace, any help? Chezetat (talk) 23:10, 18 April 2025 (UTC)

:Hi Chezetat, hopefully someone else can help on mentorship details, but if you have any questions about editing Wikipedia, you can ask them at the WP:TEAHOUSE. Just click "Ask a question", and fill out the "Subject" (the header/title) and "Description" (where you write your question, ignore the word "description"). CMD (talk) 01:07, 19 April 2025 (UTC)

::It won't let me ask a question without me being on for 4 days sadly :( Chezetat (talk) 02:35, 19 April 2025 (UTC)

:::Ah I see Chezetat, that is a temporary situation caused by vandalism. Hopefully it will end soon. CMD (talk) 02:41, 19 April 2025 (UTC)

::::Oh okay thank you, any tips on helping revert vandalism and making articles? Chezetat (talk) 02:43, 19 April 2025 (UTC)

:::::Best thing to do is find a page you're interested in that needs sources, and add what you can. CMD (talk) 04:56, 19 April 2025 (UTC)

:@Chezetat, you've got one, or you should, anyway. Do you see anything on Special:Homepage? -- asilvering (talk) 01:53, 19 April 2025 (UTC)

::No nothing at all. Chezetat (talk) 02:35, 19 April 2025 (UTC)

:::Strange. I thought you must have opted out by mistake, but if that were the case I don't know that I'd have been able to claim you. Can you see the mentorship module on Special:Homepage now? You might need to force a refresh of the page. -- asilvering (talk) 04:35, 19 April 2025 (UTC)

::: {{u|Chezetat}}, and now? Mathglot (talk) 16:55, 28 April 2025 (UTC)

::::Hmmm… I think this confusion might be related to this task:

::::T351234 Mentorship: Remove blocked users from Mentorship

::::We completed that task in response to multiple complaints from Mentors who didn’t want blocked accounts appearing on their Mentor Dashboard. However, I’m starting to wonder if the current logic needs refinement: T351234#10561866

::::An account that’s indefinitely blocked for clear vandalism doesn’t need a Mentor. But a newcomer who acted in good faith and received a temporary partial block might actually benefit from mentorship the most. - KStoller-WMF (talk) 19:21, 1 May 2025 (UTC)

:::::@KStoller-WMF, Chezetat has a clean block log, so that shouldn't be the problem here. -- asilvering (talk) 20:08, 1 May 2025 (UTC)

Hi, I'm not familiar with this initiative—although I will make an effort to change that. I'm here just to share that a first-time editor set off an anti-vandalism script for removing sourced content in an effort to complete a task presented here, and I rolled it back with a warning, not realizing I was biting a newbie (see User_talk:Burmaheaven). So the problem here might be just me! But if there's a way to help prevent this from happening in the future, that would be lovely. Many thanks, Patrick 🐈‍⬛ (talk) 22:36, 27 April 2025 (UTC)

:@Patrick Welsh, if you're worried about doing this in the future, take a look at the tags on the edit - you'll see it says "Newcomer task" and "Newcomer task: copyedit". But you're quite right to roll this edit back. Their edit was not productive, and looks like it was done with an LLM. -- asilvering (talk) 18:06, 28 April 2025 (UTC)

::Hi @Asilvering, Thanks for taking the time to respond. It never even occurred to me that someone would use an LLM to remove content in this way—although that would explain why I couldn't get any read on the underlying intention (which usually is obvious). Still, getting slapped with a revert and a warning template on literally your very first edit under a username is no way at all to greet a new editor. But, in any case, yes, I will take care to check tags for newcomer tasks. Cheers, Patrick 🐈‍⬛ (talk) 23:07, 30 April 2025 (UTC)

:@Patrick Welsh, thanks for taking the time to engage with the new editor—and for doing so in such a thoughtful and supportive way. It’s great to see experienced contributors helping newcomers learn.

:The "Copyedit" Newcomer task surfaces articles with specific maintenance templates, which communities can configure here: Special:CommunityConfiguration/GrowthSuggestedEdits. While the task includes guidance, it still gives newcomers a lot of freedom—so it's not uncommon for well-intentioned but imperfect edits to slip through. You handled the situation exactly right by reverting the edit while kindly explaining why.

:To help reduce frustration on all sides, the Growth team has been building features that give new editors more structure and support early on. One of these is the Add a Link task, which you may have seen in [https://en.wikipedia.org/wiki/Special:RecentChanges?hidebots=1&hidecategorization=1&hideWikibase=1&tagfilter=newcomer+task+add+link&limit=500&days=30&enhanced=1&urlversion=2 Recent changes]. This task presents link suggestions in a controlled way, so the only action a newcomer can take is to confirm or reject a suggested link. This keeps the editing experience simple and lowers the chance of mistakes or reverts. After completing a few of these easier tasks, newcomers are gradually introduced to more open-ended Suggested Edits.

:Most new account holders on English Wikipedia don't have access to "Add a Link" yet, but we are gradually rolling it out to more newcomers. Thanks again for taking the time to support this new editor, and let me know if you have any questions, or if you have feedback on how to improve the Growth features for newcomers! Thanks, KStoller-WMF (talk) 16:56, 1 May 2025 (UTC)

Hello all – the Growth team has been reviewing feedback about the "Add a link" task. (Special thanks to @Folly Mox, @Asilvering, @WhatamIdoing for your thoughtful input so far!) While this task helps newcomers make their first edits—especially on mobile—we also understand it can add to the workload of those reviewing edits. My hope is that the Growth team can make a few key improvements to help us responsibly scale the “Add a link” task to 100% of newcomers on English Wikipedia.

We're moving forward with two changes aimed at improving task quality and reducing patroller frustration. These updates are planned over the next two months:

• Improving link suggestion quality – We're working on better filtering to exclude certain types of links (e.g. country names, which often violate MOS:OL on English Wikipedia). There are several possible technical approaches, and Growth engineers are currently evaluating the best path forward. (T386867)

• Limiting "Add a link" to new editors – We're proposing adding a Community Configurable limit that can limit the task to newcomers. We could do this either by capping the number of link suggestions per user or by making the task available only to newcomers with fewer than 100 or 200 edits. The goal is to reduce volume and help encourage progression to more meaningful editing. (T386079)

Together, these changes aim to make the task easier to scale to more newcomers, reduce cleanup, and better align with local standards. I have ideas for several other longer-term improvements as well, but I’ll save those for a future discussion to keep this post focused.

I would love your thoughts on a few questions:

1. Are there specific types of links you'd like to see excluded in addition to country names?
2. If we add a Community Configurable limit to the "Add a link" task, should it be based on a newcomer’s total number of edits, or only on how many "Add a link" edits they've completed?
3. We’ve been rethinking how Suggested Edits tasks are categorized by difficulty—especially since newcomers often don’t consider tasks like Copyedit "easy" (as it’s currently labeled). If you have thoughts on how tasks should be categorized, feel free to take a look at the mockups and ideas I just shared in this task: T390079#10784256

Thanks again to everyone who has shared feedback — we’re listening and continuing to improve Structured Tasks like “Add a Link” to support newcomers and foster long-term editor growth and retention. We welcome your thoughts and suggestions! - KStoller-WMF (talk) 19:43, 1 May 2025 (UTC)

:# Professions that are both fairly common and fairly generic (e.g., politician, actor, athlete, musician...) are probably unwanted, at least in the first paragraph. That is, we don't usually want Musician, but we do want Opera singer. You can probably get better advice in this area from someone else.

:# I think it should be based on the number of tasks they've completed.

:# Whatever decision is made about the right place to put copyediting, I think there should be more than one kind of "easy" task. Adding links won't appeal to everyone, and we need to give people a chance to find something that does appeal to them.

:WhatamIdoing (talk) 01:28, 2 May 2025 (UTC)

::Thanks for the feedback @WhatamIdoing!

::# That makes sense. Since the link suggestion model is trained on existing links across the wikis, it should generally avoid suggesting terms that aren't commonly linked. That said, since those common profession links likely appear in some articles, we know it won’t be perfect. I’ll keep this in mind as we start exploring technical approaches to filtering out certain types of suggestions (T390683).

::# I can see pros and cons of both approaches. I've been leaning toward limiting the task based on total edit count for a couple of reasons: when we eventually release "Add a link" to 100% (and stop the current A/B test), it will suddenly become available to any account that has the Homepage enabled. That could frustrate and overwhelm patrollers—especially if there’s no clear narrative that the task is limited to newcomers. Using a total edit count threshold helps communicate that. Also, while technical limitations shouldn’t drive user experience decisions, we currently don’t cache the number of "Add a link" tasks an editor has completed. Total edit count is easy to check, whereas tracking specific task types would require additional engineering work. It's definitely doable, but it's something I'm weighing in this decision.

::# I completely agree with your point about variety. This touches on Growth’s future strategy—we’re planning to expand the range of structured tasks so newcomers can find something that appeals to them and gradually build their skills. We already have another task type, "Add an image", that could be a good candidate for English Wikipedia in the future, if there’s support. Like "Add a Link" it helps more new accounts get started editing and it's already been used to add images to over 100,000 articles on other language wikis.

::Thanks again, and let me know if you have any other thoughts, I always appreciate your perspective! - KStoller-WMF (talk) 16:06, 2 May 2025 (UTC)

:::I'm glad that you're thinking about the patrollers' experience, as it is important for what newcomers do.

:::It sounds like logging the number of tasks would be expensive. My preference might not be strong enough to justify the costs. Could there be a phase-out system? For the first 100 edits, we offer lots of add-a-link tasks; from 100–200, we could offer just (e.g.,) two each day. A "cold turkey" system could leave someone adding links to 20 articles a day, and then suddenly nothing.

:::"Add an image" sounds like a good idea. We have ~15 years of research showing that readers want more pictures, but many (most?) articles have no images or only one. If it's easy/cheap to filter the results, I'd suggest excluding stubs and WP:BLPs, at least to start. WhatamIdoing (talk) 20:08, 2 May 2025 (UTC)

:Hi! I have a suggestion on #3. In publishing, an accepted manuscript undergoes two separate kinds of edits:

:* First, the manuscript goes to a copy editor. Their job includes attending to grammar, spelling, and house style. But much of their focus is on the quality of the prose at the sentence and paragraph level. It involves a fair amount of judgment that you can only acquire from reading lots of high-quality writing and from experience interacting with other editors and authors.

:* Second, the manuscript is composed, that is, rendered on the page exactly as it is to appear in print. This version goes to a proofreader whose focus is only on correcting unambiguous errors. They work more at the level of the word and the punctuation mark.

:Could this distinction be implemented here to restrict "easy" to just the rules-based, proof-editing work, and then opening things up to more general copy-editing work at the "intermediate" level? You could probably use the same filter: anything flagged as needing a copy edit would almost certainly benefit from the attention of someone focused more narrowly on proofing tasks.

:Cheers, Patrick 🐈‍⬛ (talk) 17:43, 2 May 2025 (UTC)

::Maybe this could use the WP:CHECKWIKI lists of typos, to find spelling errors? WhatamIdoing (talk) 18:31, 2 May 2025 (UTC)

::I'm delighted to hear that Growth is looking into the "copyediting isn't easy" issue. I think @Patrick Welsh has good ideas here, and also I'd like to reiterate (as I have many times before...) that what is called "copy edit" on wikipedia is really not very much at all like what anyone outside of wikipedia thinks of when they see "copy edit" (described by Patrick above). I think we should avoid using the term "copy edit" wherever possible. "Proofreading" is a good one to use for the very simple stuff. "Content editing" might work for some of the more complicated tasks? -- asilvering (talk) 18:05, 4 May 2025 (UTC)

:::I'm glad you think this might prove helpful! The team here is welcome to adopt it, adapt it, or discard it—whatever works best for the project.

:::I completely agree, btw, that we should not attempt to impose terminology from the publishing industry onto Wikipedia. I don't even adhere to it in my own edit descriptions. I usually reserve "ce" for the correction of grammar, spelling, and formatting mistakes (i.e., proofreading stuff). I use "style edit" when I've made a change to improve the flow or general readability of the prose without changing its meaning. For content edits or larger organization changes, I try to just describe the change without assigning it to any particular level, which would serve no purpose.

:::Cheers, Patrick 🐈‍⬛ (talk) 18:25, 4 May 2025 (UTC)

:Regarding limiting the "add a link" task - it would be great if communities got to decide which of these they preferred. But if that adds far too much bother for the folks who actually have to implement that, limiting based on edit count is fine imo. -- asilvering (talk) 18:08, 4 May 2025 (UTC)

We are looking for a few volunteers to review a model we plan to use to identify peacock/puffery wording in new edits. Our idea is to show new users in-context help so that they can improve their edits.

If you are interested to help us reviewing a few edits, please add your name to the table at mw:Edit_check/Peacock_check/model_test, for enwiki. This task has more context, also feel free to ask your questions in response. :)

Trizek_(WMF) (talk) 14:31, 15 May 2025 (UTC)

In February, Mentorship was successfully rolled out to 100% of newcomers on English Wikipedia. Following this milestone, we collaborated with Spanish Wikipedia to expand Mentorship coverage to 70% of new accounts, with plans to reach 85% soon unless concerns are raised by mentors. (T394867)

Our efforts to improve and scale the “Add a Link” structured task continued across multiple fronts:

• Community Feedback & Model Improvements: We’ve responded to community concerns with targeted changes:
• Restricting access to newer accounts (T393688)
• Some links types were removed to align with recommendations written in the English Wikipedia Manual of Style (T390683)
• Allowing communities to limit “Add a Link” to newcomers (T393771)
• The model used to suggest the links was improved to ease its training (T388258)
• English Wikipedia rollout and A/B test: We increased the rollout to 20% of newcomers, with analysis underway. Preliminary data suggests this feature makes new account holders more likely to complete an unreverted edit. (T386029, T382603)
• Surfacing Structured Tasks: An experiment where we show “add a link” suggestions to newly registered users while they are reading an article is running on pilot wikis (French, Persian, Indonesian, Portuguese, Egyptian Arabic). Initial results are under analysis. (T386029)

• “Get Started” notification: Engineering is in progress for a new notification (Echo/email) to encourage editing among newcomers with zero edits. Early research shows this type of nudge is effective. (T392256)
• Confirmation email: We are exploring ways to simplify and improve the initial account confirmation email newly registered users receive. (T215665)

Communities can now manage which namespaces are eligible for Event Registration via Community Configuration. (T385341)

The Wikimedia Foundation’s 2025–2026 Annual Plan is taking shape. The Growth and Editing teams will focus on the Contributor Experiences (WE1) objective, with a focus on increasing constructive edits by editors with fewer than 100 cumulative contributions.

We value your insights and ideas! If you would like to participate in a discussion, share feedback, or pilot new features, please reach out on the relevant Phabricator tasks or at our talk page, in any language.

Growth team's newsletter prepared by the Growth team and posted by bot • Give feedback • Subscribe or unsubscribe.