Wikipedia:Village pump (miscellaneous)
Policy | Technical | Proposals | Idea lab | WMF | Miscellaneous |
Discussions are automatically archived after remaining inactive for a week.
100 graphs from the largest survey of the public's attitude to what actions should be taken on climate change
Hi all
I'm very pleased to say the EU's European Investment Bank (the largest not for profit bank in the world) has released its first batch of content under an open license. To the best of my knowledge this is only the second EU body to make content available under an open license, after the Commission.
They're released around 100 amazing graphs from the largest survey of the public's attitude to what actions should be taken on climate change (+some photos of their buildings). Broadly it shows widespread support for significant action on climate change.
Please help to encourage them to release more by adding them to articles.
https://commons.wikimedia.org/wiki/Category:Content_produced_by_the_European_Investment_Bank
Thanks
~~~~ John Cummings (talk) 19:45, 20 August 2021 (UTC)
- @John Cummings, I've added some of the photos they released to their own page. I'm glad they included those rather than just graphs, since I tend to find that photos are more useful than graphs for Wikipedia purposes (it's still great to have the graphs on Commons for other purposes). {{u|Sdkb}} talk 22:16, 28 August 2021 (UTC)
Universal Code of Conduct: Enforcement draft guidelines review
The Universal Code of Conduct Phase 2 drafting committee would like comments about the enforcement draft guidelines for the Universal Code of Conduct (UCoC). This review period is planned to begin 17 August 2021.
Community and staff members collaborated to develop these draft guidelines based on consultations, discussions, and research. These guidelines are not final but you can help move the progress forward. Provide comments about these guidelines by 17 October 2021. The committee will be revising the guidelines based upon community input.
Everyone may share comments in a number of places. Facilitators welcome comments in any language on the draft review talk page or by email. Comments can also be shared on talk pages of translations, at local discussions, or during round-table discussions and conversation hours.
There are planned live discussions about the UCoC enforcement draft guidelines:
- Wikimania 2021 session (recorded 16 August)
- Conversation hours - 24 August, 31 August, 7 September @ 03:00 UTC & 14:00 UTC
- Roundtable calls - 18 September @ 03:00 UTC & 15:00 UTC
The facilitation team supporting this review period hopes to reach a large number of communities. Having a shared understanding is important. If you do not see a conversation happening in your community, please organize a discussion. Facilitators can assist you in setting up the conversations.
Discussions will be summarized and presented to the drafting committee every two weeks. The summaries will be published here.
The full announcement and translations can be found here.
Please let me know if you have any questions. Best, JKoerner (WMF) (talk) 23:14, 20 August 2021 (UTC)
Britannica
Can i use Encyclopædia Britannica as reference? — Preceding unsigned comment added by GogoLion (talk • contribs) 07:18, 22 August 2021 (UTC)
- @GogoLion: yes, you can, but it is regarded as a tertiary source, and secondary sources are preferred if available. See WP:BRITANNICA for details. Cheers — Amakuru (talk) 07:38, 22 August 2021 (UTC)
Webcitation.org
The site is down (again, again..). The last time in 2019, it was down for 5 weeks. And when it came up, they no longer accepted new archives, and many archives no longer worked correctly (showing replacement ? characters). Enwiki currently has 239,069 webcitation.org links. In theory it is possible for bots to move to other providers, but in practice there is a fair amount of content-drift that means manual conversions are still best. An RfC was held in 2019 to deprecate the site which had SNOW support. If there is an article you care about, get rid of these links where possible. Two providers have good replacements available: Archive.today and Archive.org -- GreenC 15:50, 22 August 2021 (UTC)
Announcement
See User_talk:Secret for people that know/knew him who want to pass on best wishes, prayers or support Cas Liber (talk · contribs) 02:40, 23 August 2021 (UTC)
Copycat site
I'm not sure if this is the best place to add this, but I hope this is the right place. But hmoob.in has rather just copied the entirety of the English Wikipedia, with no attribution whatsoever. Unlike other copycat sites, this one is just rather everything translated into Hindi. So the main page ends up being a copy of our main page featured a couple of days ago. hmoob.in and ours. Similarly with the page Antandrus which is similar to their's, but in Hindi, with no attribution whatsoever. I checked Longhair's userpage (copycat site), which there seems to be a delay. I can't understand any non latin script so I'm not too sure what's on the website anyway, but I'm going off the images and the formatting. Thanks! SHB2000 (talk) 11:31, 24 August 2021 (UTC)
- While I don't speak much Hindi (or Sanskrit or whatever language this is), there does appear to be some kind of attribution in the page footer area. Third line up from the bottom on both pages mentioned here seems to be the source URL, or at least a diff in relation to my own userpage pointing to an older revision? -- Longhair\talk 14:44, 24 August 2021 (UTC)
- Don't really see it, unless I'm missing something (although I can not understand any non latin script language so I might be missing it). SHB2000 (talk) 01:27, 25 August 2021 (UTC)
- Oh wait, never mind. Saw the link although it just redirects to the article's header. SHB2000 (talk) 01:28, 25 August 2021 (UTC)
- Don't really see it, unless I'm missing something (although I can not understand any non latin script language so I might be missing it). SHB2000 (talk) 01:27, 25 August 2021 (UTC)
- Wikipedia:Mirrors and forks has some information on how to handle non-compliant sites. --Jayron32 16:08, 24 August 2021 (UTC)
Can we write articles about the Wikimedia movement?
Wikimedia movement has been tagged for "original research & "primary sources", and individual sources tagged as "third-party source needed".
Can works written by people who have volunteered on Wikimedia projects be cited on articles about the movement? What is the cut off point - a thousand edits? Ten? One? Running a meetup or editathon? A talk at Wikimania?
Are editors involved in the movement, in any way, conflicted in writing about it?
Otherwise, how can we write about Wikipedia? Or Earth? Or Human? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 22:10, 24 August 2021 (UTC)
- I think you'll find that in the discussion there, the 'cutoff point' was whether people who had been employed by or received funding from the WMF or one of its projects could be cited as third-party sources in order to establish notability of the topic (whatever the topic is - it is evident from discussions that there is a great deal of debate about this). This is a question of independent sourcing to justify articles - something Wikipedia is normally insistent on, and which some people might consider it to be highly inappropriate to make exceptions for.
- Nice canvassing effort, by the way... AndyTheGrump (talk) 22:19, 24 August 2021 (UTC)
- Pigsonthewing, there are editors who don't consider themselves part of the "Movement". Vexations (talk) 22:30, 24 August 2021 (UTC)
- Surely you can find sufficient high quality sources written independently of Wikimedia with which to write an article? --Malcolmxl5 (talk) 23:18, 24 August 2021 (UTC)
- With an article like this, I'd worry about how many sources are talking about a "Wikimedia movement" or just about individual communities. I would not categorically assume that the latter is part of the former. Jo-Jo Eumerus (talk) 09:54, 25 August 2021 (UTC)
- That is of course at the heart of much of the rather heated debate over this particular exercise in navel-gazing: The article concerned starts with an unequivocal assertion regarding a putative 'Wikipmedia movement' which it claims consists of a 'global community'. And while I'm sure that many of the more deeply-involved contributors to one or more of the many WMF-hosted projects may consider themselves as members of a 'global community', and/or a 'movement', it seems entirely unreasonable, to me at least, to assert that everyone posting stuff on any of the many projects, anywhere, can be reasonably described as a member of a 'movement' they may very well never have heard of: a position being advocated by some in the debate. A 'movement' which seems, depending on who you ask, to consist of the WMF's formal structure, or of the less-than-structured 'communities' within individual projects that it would seems self-evident often have little consistency over who they consider part of their own individual 'community', never mind any of the others. If we can't agree amongst ourselves exactly what the heck it is we are participating in, should we really be present to our readership an article which writes as if it is a tertiary source on the question? The article presents one perspective on the issue as if it were unequivocal fact, and, in my opinion does so in a manner which denies the autonomy of individual projects, the cultural diversity and political histories that shape such projects, and the social agency of individual participants. People use WMF-hosted user-editable-media for all sorts of reasons, and co-opting them into a global 'movement' without their knowledge or consent, even only rhetorically, isn't, in my opinion, something the English-language Wikipedia (or rather some of the participants within it) should be doing. AndyTheGrump (talk) 11:32, 25 August 2021 (UTC)
- It sounds like it might be useful to look at m:Wikimedia movement.
- Reasonable people could disagree about whether the people who participate in any social movement "count" if they don't personally self-identify as a part of that movement. It is the nature of loosely organized group that Alice can say that Bob is part of the group, and Bob can say that he's not, and neither of them are wrong. WhatamIdoing (talk) 17:56, 25 August 2021 (UTC)
- I'm not sure how looking at a Mediawiki website would help resolve the issue of our own article lacking third-party sources. As for your comment about personal self-identification, I'm fairly certain that where there is a disagreement amongst 'reasonable people', Wikipedia policy isn't to assert definitively that one person is right and the other is wrong. Which is why I object to the article making sweeping statements about what this nebulous 'movement' consists of, based entirely on what some participants and/or the WMF say. More so when it is making statements about real people, rather than hypothetical Bobs and Alice's. AndyTheGrump (talk) 18:20, 25 August 2021 (UTC)
- If you read the page at Meta-Wiki, you might discover that there are multiple possible definitions of "the movement", some of which do not line up with the editor-focused first sentence in the article.
- Movements are always nebulous. That's what makes them be movements. (Alice and Bob are real enough to be notable.
;-)
) WhatamIdoing (talk) 19:44, 27 August 2021 (UTC)
- I'm not sure how looking at a Mediawiki website would help resolve the issue of our own article lacking third-party sources. As for your comment about personal self-identification, I'm fairly certain that where there is a disagreement amongst 'reasonable people', Wikipedia policy isn't to assert definitively that one person is right and the other is wrong. Which is why I object to the article making sweeping statements about what this nebulous 'movement' consists of, based entirely on what some participants and/or the WMF say. More so when it is making statements about real people, rather than hypothetical Bobs and Alice's. AndyTheGrump (talk) 18:20, 25 August 2021 (UTC)
- That is of course at the heart of much of the rather heated debate over this particular exercise in navel-gazing: The article concerned starts with an unequivocal assertion regarding a putative 'Wikipmedia movement' which it claims consists of a 'global community'. And while I'm sure that many of the more deeply-involved contributors to one or more of the many WMF-hosted projects may consider themselves as members of a 'global community', and/or a 'movement', it seems entirely unreasonable, to me at least, to assert that everyone posting stuff on any of the many projects, anywhere, can be reasonably described as a member of a 'movement' they may very well never have heard of: a position being advocated by some in the debate. A 'movement' which seems, depending on who you ask, to consist of the WMF's formal structure, or of the less-than-structured 'communities' within individual projects that it would seems self-evident often have little consistency over who they consider part of their own individual 'community', never mind any of the others. If we can't agree amongst ourselves exactly what the heck it is we are participating in, should we really be present to our readership an article which writes as if it is a tertiary source on the question? The article presents one perspective on the issue as if it were unequivocal fact, and, in my opinion does so in a manner which denies the autonomy of individual projects, the cultural diversity and political histories that shape such projects, and the social agency of individual participants. People use WMF-hosted user-editable-media for all sorts of reasons, and co-opting them into a global 'movement' without their knowledge or consent, even only rhetorically, isn't, in my opinion, something the English-language Wikipedia (or rather some of the participants within it) should be doing. AndyTheGrump (talk) 11:32, 25 August 2021 (UTC)
- This strikes me as being less a content issue and more a titling issue. There is a notable topic in there somewhere… but we need to more clearly determine what that notable topic actually is. The dispute appears to be over using the term “movement” in the title, and whether that term is used by anyone outside of the various projects under the wikimedia umbrella (ie a source that is independent of the topic). Simply removing that term from the title would help. There are lots of independent sources that discuss Wikimedia without calling it a movement. We can still mention that many within the various projects see what we do as being part of a broader “movement” (the current non-independent sources are reliable for that). Blueboar (talk) 20:05, 25 August 2021 (UTC)
Zoophoria
The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.
I raised an AfD for Zoophoria on the 15th. I can see it in some places but not others and there has been no activity.Slimy asparagus (talk) 06:59, 25 August 2021 (UTC)
- @Slimy asparagus It looks like the issue is the AfD wasn't transcluded to the log (WP:AFDHOWTO step 3 for reference). I have done that. —Danre98(talk^contribs) 14:28, 25 August 2021 (UTC)
- Thanks. Slimy asparagus (talk) 14:39, 25 August 2021 (UTC)
Origins of bread
in List of breads, how are origins of white bread and whole wheat bread usa and canada? -- RZuo (talk) 21:41, 28 August 2021 (UTC)
What is the Best AI model for Content Moderation on Wikipedia?
Imagine you’ve just spent 27 minutes working on what you earnestly thought would be a helpful edit to your favorite article. You click that bright blue “Publish changes” button for the very first time, and you see your edit go live! Weeee! But 52 seconds later, you refresh the page and discover that your edit has been reverted and wiped off the planet.
An AI system - called ORES - has been contributing to this rapid judgement of hundreds of thousands of editors’ work on Wikipedia. ORES is a Machine Learning (ML) system that automatically predicts edit and article quality to support content moderation and vandalism fighting on Wikipedia. For example, when you go to RecentChanges, you can see whether an edit is flagged as damaging and should be reviewed. This is based on the ORES predictions. RecentChanges even allows you to change the sensitivity of the algorithm to "Very Likely Have Problems (flags fewer edits)" or "May Have Problems (flags more edits)”.
In this discussion post, we want to invite you to discuss the following *THREE potential ORES models* -- Among those three models, which one do you think presents the best outcomes and would recommend for the English Wikipedia community to use? Why?
ABOUT US: We are a group of Human–computer interaction researchers at Carnegie Mellon University and we are inviting editors to discuss the trade-offs in AI-supported content moderation systems like ORES; your input here has the potential to enhance the transparency and community agency of the design and deployment of AI-based systems on Wikipedia. We will share the results of the discussion with the ML platform team which is responsible for maintaining the ORES infrastructure. However, the decisions of the discussion are not promised to be implemented. More details are available at our research meta-pages: Facilitating Public Deliberation of Algorithmic Decisions and Applying Value-Sensitive Algorithm Design to ORES.
Model Card One: High Accuracy
- Performance table
Group / Metrics | Accuracy Percentage of edits that are correctly predicted
|
Damaging Rate Percentage of edits that are identified as damaging
|
False Positive Rate Percentage of good edits that are falsely
identified as damaging |
False Negative Rate Percentage of damaging edits that are
falsely identified as good |
---|---|---|---|---|
Overall | 98.5% | 3.4% | 0.5% | 26.3% |
Editors that have registered more than two months | 99.7% | 0.2% | 0.0% | 61.2% |
Editors that have registered only less than two months | 95.7% | 10.7% | 1.8% | 23.0% |
Editors that have not registered | 94.8% | 12.7% | 2.4% | 22.8% |
- Explanation: this model has the highest overall accuracy.
Model Card Two: Fair Treatment
- Performance table
Group / Metrics | Accuracy Percentage of edits that are correctly predicted
|
Damaging Rate Percentage of edits that are identified as damaging
|
False Positive Rate Percentage of good edits that are falsely
identified as damaging |
False Negative Rate Percentage of damaging edits that are
falsely identified as good |
---|---|---|---|---|
Overall | 97.2% | 1.2% | 0.1% | 69.9% |
Editors that have registered more than two months | 99.6% | 0.0% | 0.0% | 94.0% |
Editors that have registered only less than two months | 91.2% | 4.4% | 0.8% | 68.5% |
Editors that have not registered | 90.7% | 4.5% | 0.0% | 67.2% |
- Explanation: Compared to Model One, this model treats experienced editors, newcomers, and anonymous editors more similarly, but it has lower overall accuracy.
Model Card Three: Balanced
- Performance table
Group / Metrics | Accuracy Percentage of edits that are correctly predicted
|
Damaging Rate Percentage of edits that are identified as damaging
|
False Positive Rate Percentage of good edits that are falsely
identified as damaging |
False Negative Rate Percentage of damaging edits that are
falsely identified as good |
---|---|---|---|---|
Overall | 96.1% | 7.6% | 4.0% | 2.4% |
Editors that have registered more than two months | 99.9% | 0.4% | 0.0% | 17.9% |
Editors that have registered only less than two months | 91.8% | 19.8% | 9.1% | 1.0% |
Editors that have not registered | 82.7% | 30.8% | 19.9% | 0.8% |
- Explanation: Compared to Model One and Two, Model Three attempts to achieve a better balance between false positive rate and false negative rate. The false negative rate is the best among the three models. But this model has lower accuracy and higher damaging rate.
If you are not satisfied with any of the models described above, you can try out this interface, pick a model on your own, and share your chosen model card in the discussion by copying and pasting the wikitext offered in the interface.
Bobo.03 (talk) 15:32, 20 August 2021 (UTC)
- Honestly, after looking at them and comparing them, in my opinion I think Option 2 would be a better option. While it has a
nicehigh false negative rate of 69.9% overall, looking at the individual percentages show that it's mostly experienced users that would have their edits identified as false negatives. However I think that percentages don't always show the full picture. Instead I think it would be better for it to be shown as like 1 in every thousand edits is identified as a false negative, that way we can actually see what scale we're looking at. However, I think a combination of models 2 and 3 would be best as Option 2 has a low False Positive rate but a high False Negative rate and Option 3 has a high(ish) false positive rate and a low false negative rate. Blaze The Wolf | Proud Furry and Wikipedia Editor (talk) 16:32, 23 August 2021 (UTC)
Discussion Break
Hi @Bobo.03: I'm sure this has come along since my very early engagement with it. I know you are well-aware of CluebotNG, but I'd like to draw a highlight that although it once accepted a false positive rate of 0.25%, it has been changed to use 0.1% as its threshold. That hit 55% and 40% of vandalism (thus 45% and 60% false negative). That, I think, gives a pretty clear marker that Wikipedians are way more willing to accept it missing something than an unwarranted hit. Unwarranted hits kill off new users, and irk experienced users, while many issues missed can be caught by alternate means. I tried to have a fiddle with the interface but couldn't figure out how to make it apply different tolerable false positive rates to different groups. Nosebagbear (talk) 20:43, 20 August 2021 (UTC)
- I would add my opinion that the false positive rates reported for option 3 are way too high for me, and I suspect for most other editors. Phil Bridger (talk) 23:08, 20 August 2021 (UTC)
- I'd like to add that IMO, whatever happens to 'Experienced' editors is pretty irrelevant to me, so that leaves newcomers and anonymous edits. False positives rates above 1% are unacceptable from the outset IMO. So the 'fair treatment' table approach is the most viable one, IMO. Since this only flags, but doesn't revert, I'm OK with a higher false positive rate than ClueBot NG, but it should be sub 1% on any given categories, and lower would be even better. Headbomb {t · c · p · b} 02:24, 22 August 2021 (UTC)
- Hi @Nosebagbear: Thanks for your questions! The interface, based on ORES, indeed does not allow users to apply different FPR tolerance level to different groups! Would you mind to share that in your opinion, which level of FPR is tolerable for different groups? The information would be extremely useful to improve the interface and the ORES model itself. Thank you! — Preceding unsigned comment added by Bobo.03 (talk • contribs) 18:20, 22 August 2021 (UTC)
- I'd like to add that IMO, whatever happens to 'Experienced' editors is pretty irrelevant to me, so that leaves newcomers and anonymous edits. False positives rates above 1% are unacceptable from the outset IMO. So the 'fair treatment' table approach is the most viable one, IMO. Since this only flags, but doesn't revert, I'm OK with a higher false positive rate than ClueBot NG, but it should be sub 1% on any given categories, and lower would be even better. Headbomb {t · c · p · b} 02:24, 22 August 2021 (UTC)