Since the 2013 RfC on Wikidata Phase 2, using Wikidata in infoboxes has gone from the concept stage to more extensive mainspace experimental implementations. This had led to some controversies: some editors have been converting infoboxes to use Wikidata, believing use of Wikidata brings significant benefits, while other editors have been removing Wikidata from infoboxes, believing use of Wikidata brings significant problems. The opposing efforts are counterproductive. It is time to ask the broader Wikipedia editor community where to take it from here. 20:49, 6 April 2018 (UTC)

The poll is now closed, for the closing summary see #Discussion. But note:

At the bottom of that discussion is this summary of the summary: There is a consensus that data drawn for Wikidata might be acceptable for use in Wikipedia if Wikipedians can be assured that the data is accurate, and preferably meets Wikipedia rules of reliability. For the other issues raised within this RfC, there was no clear consensus.

See also: RfC: Wikidata Phase 2 (2013), RfC: Wikidata in infoboxes, opt-in or opt-out? (2016) and RfC: Linking to wikidata (2018)

Wikidata is a Wikimedia project that "acts as central storage for the structured data of its Wikimedia sister projects including Wikipedia, Wikivoyage, Wikisource, and others." In the same way that Wikimedia Commons hosts media files, Wikidata hosts claims (a piece of data in a named field in a Wikidata "item" about some person, place, thing, or concept). Most of the information is stored in a language-neutral format, using Q-codes in place of words, so that it can be used across the different language Wikimedia projects (with exceptions for language-specific phrases such as media captions).

Every Wikipedia article has a corresponding "item" on Wikidata. This includes links to pages on the same topic in other language Wikipedias and other Wikimedia projects, as well as labels and descriptions in different languages, and structured statements about the topic. Wikidata has much broader inclusion criteria than Wikipedia, and contains many items that do not currently have Wikipedia articles.

The possibility of using Wikidata in infoboxes was originally discussed in Wikidata Phase 2 back in 2012 (also see Phase 2 technical proposal and the Wikidata deployment Q&A on Meta), which led to the 2013 RfC.

Wikidata information is added both by individual editors and bot imports. Individual statements can be added by hand, or they can be imported from Wikipedia infoboxes, categories and articles (across all language Wikipedias), or from external databases.

Special code (magic words and modules) can request information from Wikidata. These are usually only used in a template. Every time you preview, save, or purge a page, all requested Wikidata values appear freshly on the rendered page. The information does not appear in the wikitext.

Changes to Wikidata values are stored in the Wikidata item history, rather than in the Wikipedia article history. This is the same as template code (where formatting, displayed parameters and template-stored data can change, although parameter values remain in the article) and inclusions of other pages into Wikipedia articles. Viewing a version of the page from history will retrieve and show current values from Wikidata (and current versions of templates), rather than the page as it appeared on that date.

We use Wikidata to provide most of the interwiki links to other language Wikipedia articles (although a small number of custom interwiki links remain locally, e.g., to sections of articles on other languages). We also use Wikidata information in templates such as {{Authority control}}, which provides links to catalogue entries on the article subject.

More recently, a number of infoboxes use information from Wikidata following from the Wikidata Phase 2 RfC in 2013.

Wikidata can be used to display all of the content in infoboxes, just some of it, or none of it. Images, captions, coordinates, locations, links, dates, units (including conversion between metric and imperial), websites, and maps can all be shown using Wikidata values. However, information that isn't structured/well-defined (such as free text or vague statements) can't be described in Wikidata.

Data from Wikidata is fetched and presented in Wikipedia using templates. This can be through existing templates, such as {{Coord}} and {{convert}} (which have been modified to provide access to Wikidata information), or it can be by using templates that provide access to Wikidata-specific "modules" such as Module:Wikidata, Module:WikidataIB and Module:Wd, that written using code called "Lua". These Lua modules can also evaluate the data and only pull data with certain qualities. For example, the module can only fetch data that has a reference (with caveats, see below), and it can be used to select specific fields or suppress fields.

We currently have a mix of opt-in and opt-out Wikidata infoboxes (e.g. {{Infobox person/wikidata}}, which is opt-in, vs {{Infobox artwork/wikidata}}, which is opt-out). A 2016 RfC VPP RfC: Wikidata in infoboxes, opt-in or opt-out? did not provide a consensus on which approach to take here, and the options below reflect this.

Infoboxes that are currently entirely from Wikidata can be found at Category:Articles with infoboxes completely from Wikidata.

In some versions of infoboxes, the data resides solely in Wikidata and is just displayed in the infobox in the article; the data itself is not in Wikipedia. If you look at wikitext in the edit window, you will see only a short template for the infobox like {{infobox_gene}}. The infobox will change whenever Wikidata is changed (it is rendered anew each time the Wikipedia page is re-cached).

These infoboxes have a link that sends you to Wikidata to change the data there. You can do this using your Wikimedia unified login. You cannot edit the data from within Wikipedia, although you can provide new values using Wikicode in the traditional way, which are then displayed instead. A prototype for editing Wikidata information from within Wikipedia is under construction by the Wikidata developers, although development work on this is planned for the end of 2018, with a beta version in 2019.

You can watch Wikidata pages using your Wikidata watchlist, and Wikidata also has a RecentChanges page. You can also follow changes in your Wikipedia watchlist and at Special:RecentChanges. This is not currently enabled by default: you need to go to the 'Recent Changes' and/or 'Watchlist' tabs of Special:Preferences and tick the check-box next to 'Show Wikidata edits in your watchlist' (they are turned on separately). Wikidata edits do not show up in the article history page, analogous to how edits to templates also aren't included in article histories.

Due to the high rate of edits on Wikidata (as each change to a value is saved separately), and the large number of articles that can be affected by an edit, including Wikidata edits on watchlists and RecentChanges causes technical problems. In order to reduce the dangerously high rate of watchlist changes, changes that affect many articles do not appear in the watchlist for more than the first few articles that are affected, the rest are simply dropped (phab:T177707), although changes to the item directly linked to the Wikipedia article are always displayed. Including any Wikidata changes in recent changes/watchlists was completely disabled on Wikimedia Commons and the Russian Wikipedia in October 2017 (phab:T171027) as they are currently the heaviest users of Wikidata; however this did not apply to any other Wikimedia projects. Additionally, there can be a lag between an edit being made on Wikidata and the change showing up here, which is tracked at d:Special:DispatchStats. Technical work to improve these issues is ongoing, and changes that do not affect the article are no longer shown.

The Wikidata community governs itself, just as the English Wikipedia community governs itself.

We (English Wikipedia) have a number of policies and guidelines that are not used by the Wikidata editing community - they have their own policies and guidelines.

• WP:V and WP:BLP: Wikidata has no sourcing policy. d:Wikidata:Verifiability has been proposed since January 2015. d:Wikidata:Living persons (draft) has been proposed since April 2013. d:Wikidata:Living people was created in September 2017; it is not tagged as a policy or guideline and consists of the Board of Trustees' 2009 resolution on living people. d:Wikidata:Requests for comment/Privacy and Living People is currently running.
• WP:RS
• WP:NOR: Wikipedia prefers secondary sources to primary sources, although both can be used on Wikidata.
• WP:NPOV: Where there are multiple sources with contradictory values, Wikidata stores all of values (and the references where available). The values can then be marked as 'preferred', 'normal-ranked' and 'deprecated'. However, more fine-grained controls are not available, and there is no Wikidata policy on this.

Each individual statement on Wikidata is referenced separately. References are stored in two different ways. In most cases they are stored in the 'references' part for each statement, where you can provide reference URL (P854), title (P1476), retrieved (P813), etc. for the given statement. Alternatively, the reference can have its own Wikidata entry (for example, if it is a notable book), in which case the reference can be an instance of stated in (P248), which is sometimes accompanied by additional information such as page(s) (P304). Wikidata also uses imported from Wikimedia project (P143), which is used to show where the imported information came from, however this is frequently used as "Imported from X Wikipedia", which is a circular reference (and as such, this is not treated as a valid reference here, and it is ignored when only importing referenced values). Additionally, in some cases that reference may be a circular reference to unsourced information within wikidata, and in some cases it is a reference to the Wikimedia Labs.

Wikidata stores references without any formatting, and different properties (title, authors, url, DOI, etc.) are stored separately. The reference information can then be reformatted as appropriate when it is fetched from Wikidata. However, we have a lot of different formatting styles here (e.g., see WP:CITEVAR), which complicates the process of reformatting the Wikidata information to follow the different formatting styles.

Titles, other names, descriptions and site links on Wikidata cannot be referenced. Some of these provide context, which can vary between languages. Others, like sitelinks, don't need to be referenced. Other pieces of information, such as images, can have references although they do not need them (and for these properties, referencing requirements can be selectively removed through the infobox code).

Various Wikipedia reference-related procedures, e.g. requesting a page number with {{page needed}}, indicating a failed verification with {{failed verification}}, defunct weblinks rescued by InternetArchiveBot, implementing the outcome of a WP:RSN discussion, etc., need to be adapted for Wikidata, and may need the editor to fix things immediately rather than adding them to backlogs.

References from Wikidata can be shown here using {{Cite Q}} and Module:Wd; however, they are opt-in in most Wikidata infoboxes (set refs=yes to show them; these exclude "imported from" references), and a recent deletion debate about {{Cite Q}} ended with no consensus (with caveats).

No. Wikipedia / Wikimedia projects are not required to use data from Wikidata anywhere. It is a decision to be made by each editing community. Wikis that decide to use Wikidata can determine how they want to use it. The decision is ours to make. Different options are discussed below.

This RfC has four questions.

Note: Wikidata references containing the word 'Wikipedia' (such as "Imported from: English Wikipedia") are specifically excluded from all options here to avoid circular references.

Some editors have deleted values from infoboxes so that identical or equivalent values will be pulled from Wikidata. Other editors have restored those local values to the wikitext. Both kinds of edits have little or no effect on the displayed infobox. To avoid fruitless edit warring, we should decide which kind of edit is appropriate and which is inappropriate.

.

That is, with references (preferably with a bit more than the basic url), if you want it to conform with our policies.

I think the above shows that, although Wikidata can be a good starting point for an article, it isn't always up to date, is prone to both good faith errors and vandalism, and that nothing beats some editor research and checking... Robevans123 (talk) 00:21, 9 April 2018 (UTC)[reply]

@Robevans123: All of that is true, but it's all true for Wikipedia as well. If we only correct the list of sister cities on Wikipedia, we have one project with accurate information. If we correct the information on Wikidata, we potentially supply accurate information to 280+ other Wikipedias. That's got to be worth trying for. I'm pleased you noticed that the results of pulling data from Wikidata is dynamic, and so it should be: if somebody makes the effort to add references on Wikidata, we want the referenced information to appear immediately. If you have a screen reader, you'll hear that the lists above are marked up as real lists (and that's why there's no "and"). As for alphabetically, how does this look:

• Changchun
• Chicago
• Frankfurt
• Johannesburg
• Leipzig
• Lyon
• Milan 

Or if you want to see even unreferenced claims:

• Changchun
• Chicago
• Frankfurt
• Guangzhou
• Johannesburg
• Leipzig
• Lyon
• Milan
• Xi'an
• Zaporizhzhia 

You can make another call that returns a citation from Wikidata, but there is no way of working out which citation style is in use in any particular article, so I haven't implemented it, as I'm not about to hand yet another excuse to complain to the usual suspects. --RexxS (talk) 01:20, 9 April 2018 (UTC)[reply]

• For those who advocate removing all references to WD, please consider data that acts as Help:Authority controls to articles, which is itself a form of referencing - we in anatomy use e.g. links to Terminologia anatomica, an international (and multilingual) listing of anatomical structures. We would have to re-port this and several other authority controls into WP and then manually defend it against vandalism, which is not uncommon given these terms are immutable but our articles are constantly the subject of defacement and the infobox is one of the first thing vandals see. Authority controls are a form of references linked to the article and I feel may require separate consideration given the emphasis that we shouldn't use wikidata owing to WP:V concerns--Tom (LT) (talk) 12:02, 9 April 2018 (UTC)[reply]
  • @Tom (LT): '... and then manually defend it against vandalism' .. but now you have to manually defend it also against vandalism on an external wiki where there are likely less eyes on the subject than here (except for the few that look at WikiData through their en.wikipedia watchlist). There is unfortunately no way to lock down immutable data on WikiData (after it is properly checked and verified to be correct) - it is either the whole page or nothing. --Dirk Beetstra ^{T C} 12:15, 9 April 2018 (UTC)[reply]
    • @Beetstra: Far easier to re-sync Wikidata to an external source than a myriad of Wikipedia articles. Also far easier to query for changes to a particular field over a particular period. For cases where a periodic re-check is enough, it's a lot easier to do on Wikidata.
      If you ban imports from Wikidata, are YOU going to undertake to do all the updates manually, if external IDs change? Jheald (talk) 12:32, 9 April 2018 (UTC)[reply]
      • Oh, it is rather easy to run a bot on en.wikipedia to update the fields or to add fields. We have been doing that for years. And since we are talking immutable data, lets just find our way to protect the really immutable and checked data. It does not make it in any form more difficult, and at least we can protect the data here. Unless WikiData will implement a way where the truly immutable data can be individually made immutable. --Dirk Beetstra ^{T C} 12:46, 9 April 2018 (UTC)[reply]
        • @Beetstra: Except that external IDs often aren't immutable. They get merged, split, updated. The rate varies from database to database, but if you don't keep them actively synchronised, they do rot.
          So once they get re-sync'd on Wikidata, then what? You bot-import them to here, you say. But that's not compatible with 1A. Which is why a blanket !vote for 1A is fundamentally clueless. Jheald (talk) 12:58, 9 April 2018 (UTC)[reply]
          • @Jheald: The problem is, that WikiData data, referenced or not, gets actively vandalized, spammed, &c. (yes, the same crap happens here). The data there is unprotected (same crappy problem as here), and part of it is unreliable (just as here). The sourcing requirements on WikiData are far, far less strict than Wikipedia, and there is no protection of the data that is there (same as here). But WikiData has way less people monitoring it (I know, this RfC also acts as a good reason to try and attract (well, force) more people to watch WikiData data if 1F is implemented). The data that is there and is 'referenced' either is fine, or it is not referenced at Wikipedia standards, or it is referenced but since vandalized, or it is correct but unreferenced, or incorrect and unreferenced. But we have NO clue at which level each P in each Q is (the handful that I ran in just a day or 3 ago contained 1 that was repaired only after 2 months, and one is still there after 3 1/2 months - a WP:POINT violation on my side if you wish).
          • No, I would not import them from WikiData, fully in line with my 1A - I would import PubChem numbers from PubChem, I would import CAS numbers directly from CAS (they don't allow that ..), I would import everything directly (that is what I have done in the past, with an attempt at protecting those datapoints years after we had the data .. an impossible and gargantuan task on en.wikipedia). I would not take the risk that the data in WikiData gets changed between one of their imports and the moment that we decide to take the data from them. And there is no reason to use WikiData as an intermediate if I can run a bot to do it directly (the result should be the same, with more certainty that the data is still correct).
          • The immutable data is immutable indeed until a database updates (though to take a chemical identifier, the CAS, those numbers are practically 'for life' except under some very rare circumstances). But there is no reason for 'the man in the street' to mutate those 'at will'. They are immutable until proven needed to be mutated due to a database update (which for some data may happen regularly, for others never). IMHO, it would be easy to implement a setting on data where one P-field per Q-set can be set to 'immutable', and have a group of editors promoted to 'mutators' who are the only editors who can mutate that immutable data. Then when you have a bot update the data from a verified database, they mark that P in each Q as immutable. Editors with a reviewing-type of right can do the same, they check if a certain manual P is properly referenced and correct against that reference, and can mark it as immutable. It is so simple (and yes, so un-wiki-like). At that point, if en.wikipedia trusts the referencing standard of WikiData, and we have that immutable, unvandalizable data, then we can have a new RfC to see whether opinions have changed. Until then, I remain a firm 1A supporter - even if the quality is the same, I prefer to maintain that data here then having to maintain it through an intermediate. --Dirk Beetstra ^{T C} 13:34, 9 April 2018 (UTC)[reply]
          • Note: if even only 0.1% (but growing) of the data on WikiData that we would have here on Wikipedia would be reliably correct and immutable I would have supported a laxer option (likely a 1D, but possibly a 1F) for that data, with support of removing local data if it is the same as the WikiData version (but still having the possibility to override a WikiData datapoint, but with an alerting categorization that the WikiData value is challenged). --Dirk Beetstra ^{T C} 13:52, 9 April 2018 (UTC)[reply]
            • @Beetstra: Sorry but I didn't your reasoning well grounded:

            - WD is vandalized/spammed, WP is vandalized/spammed, this a fact. But how could you assume that WD is more vandalized than WP ? Any statistical analysis about that ? And if WD is vandalized in the same extend than WP, would you agree that there is no reason to consider WP as better than WD ?

            - The point of the existence of non-sourced data in WD is not relevant to decide if WD can be used or not in WP bcause WD allows you to filter them. You don't like unsourced data ? Well, ignore them by using the filter tool.

            WD is full of sourced data from unreliable references ? Well, ignore sourced data from unreliable references by using the filter tool.

            Why do you take care of unsourced data/bad quality data if you can discard them from your data extraction ? You can add this feature in the code of your infobox once and that filter will be applied at each opening of the WP article. WD is full of unsourced data/bad quality data but WD gives you the way to get ride of this unwanted stuff, problem solved, chapter closed, unless you can demonstrate that unsourced data/bad quality data from WD are not correctly filtered by the filter tool.

            The only argument against sourced data in WD is than the amount of good quality and sourced data is representing a too low percentage of the data in a relevant field. Is it the case ? To demonstrate. Is it a permanent drawback ? No as sourced data in WD is growing. So based on that argument we could say WD is currently not interesting but not that WD will never be interesting.

            - You mention the fact that this is better to populate infoboxes directly from the original databases than using WD as intermadiate. You are right, but can you give us about that regular update ? Is there any plan or bot to perform that task ? When I see that contributions were never verified since several years (see the CHEMBL identifier in Hydrochloric acid were is spotted as unverified since more than 4 years), I have some doubt that such affirmation will be done.

            - I won't spend time about the affirmation that WP data in infoboxes are well sourced. If I take the boiling point of dichlorine, I can't find the link between the value and the source. Yes I know, there is a references list, but from which reference the value was taken ? Can you really affirm that all references have the same value for the boiling point ? Is it not possible that the value becomes more accurate with the time.

            - You affirmed that WD was not properly designed to fight vandalism or error. You completely skip tools like constraint violation reports which offers an always updated list of some deviations to some rules (format, single rule, unique rule,...).

            I could continue, but I think the main idea is there: WD is not ready for a full application in the diverse WP's fields but if a WP wikiproject is confident in a dataset in WD and maintains properly the dataset in WD, why do you restrinct the use of WD for that particular application ? Instaed of pointing the weakness of WD, why do you list the minimal requirements to allow the use of WD in WP ? And be fair: don't require stronger requirement for WD than the ones applied in WP:en. You mentioned the fact that the WD community was to small to monitore all contributions in WD, but can you show that all WP:en articles are in the watchlist of several contributors and that the contributions in those articles are regularly checked ? In not, how can you point that as a weakness in WD and not in WP ? Snipre (talk) 22:30, 10 April 2018 (UTC)[reply]

            • • "But how could you assume that WD is more vandalized than WP ? Any statistical analysis about that?" .. WikiData has a smaller editor base, it has a smaller admin base (the latter by a factor of 10). There are no clear BLP policies, spam policies, and I gave an example of spammed urls that stay there for months (one still being there). I do notice on meta that there start to be spammers that target WD in their very first edit (as that one that was there for over 2 months). Moreover, it is not my task to show that WD is more reliable than WP.

              "The point of the existence of non-sourced data in WD is not relevant to decide if WD can be used or not in WP bcause WD allows you to filter them." - I agree, I am talking however also about the referenced data on WD - taking our previous point: you have no clue whether the data is correct with respect to the reference, because even the referenced data can be changed without changing the reference. Then the policies and guidelines on WD are not the same as en.wikipedia, the data may be referenced to references that we would not use (and that is often on a case-by-case data, I have been discussing the French library - where the statement was that it was the French library, of course it must be reliable. However that specific item clearly states that it uses imdb as a source. That something is published by an established publishing house does not necessarily make a single datapoint reliable. And that with less stringent policies in the first place.

              Updating directly vs. indirectly. Yes, my experience with the chemical infobox and updating is exactly that - it takes an humongous amount of effort to do that with one - verify whether the datapoint is properly linked to the page where you are talking about (and that requires proper knowledge of chemical compounds ..). We have spotted errors in those databases or in the databases between each other. There are CAS numbers that are linked to multiple compounds, there are compounds with multiple CAS numbers. CAS takes an effort in repairing that. But a plain database upload 'we have hexane here, they have hexane there, import identifier' fails on many other compounds. En.wikipedia has always been reluctant to updating data directly from databases for a good reason. Do you expect me to trust that because it now happens on WikiData that those en.wikipedia concerns do not apply?

              Yes, there are datapoints that depend on the reference - however each of those datapoints is still immutable against the reference (ref A is saying 100, ref B is saying 100.1 - that 100 is NOT going to change with respect to A, that 100.1 is NOT going to change with respect to B). Yes, it is likely that there will be a more accurate reference C at some point, who comes to 100.094. That does not change that reference A said 100 and reference B said 100.1. It does mean that we would upgrade that we use 100 to 100.1 to 100.094, and change the reference. That does not make the datapoint mutable. And that is for physical data, the same reasoning goes with your birthday - it is unlikely that that will ever be defined more precise. There is absolutely 0 reason to change that data with respect to the references that are there.

              So we have BLPs - where I think it fails, we have chemical compounds - where I think it fails. I am sure that there are many other fields where it fails. Many of our pages do not 'belong' to one WikiProject (where WikiProjects do not own pages) - so do we have to have fights between WikiProjects where one thinks that WD is good enough and the other doesn't, have lengthy discussions properly organized 'yes, for this group of articles a local consensus exists to use WikiData'. If I read the !voting above, there are many people who have concerns for the overall use, do you think that these editors could be convinced of a local consensus. IMHO, it is a question between 1A and 1F - either we don't, or we go full. Anything between that is a recipe for problems, discussions, fights. And no, WD is not ready for that.

              My minimal requirements are there - 3A - all data should be compliant to OUR policies and guidelines, which means that WD's policies and guidelines should be more strong than Wikipedia's, and that WD's protection of said data is (way) more stringent than ours. WD is by far not there yet, and I am sorry to see that the way WD started up (first shoot, then ask questions) does make it very unlikely that it will be up to that task in any near future. Until then, no. We are already fighting whether infoboxes should be included in the first place. --Dirk Beetstra ^{T C} 06:19, 11 April 2018 (UTC)[reply]

Hmm... Although topics what currently discussed in this page include important points and I share same concerns, it seems that more practical information about current state of WikiData is needed, especially about well-formatted data like identifiers. As far as I searched this page with CTRL+F, no one yet mentions the most powerful counter-vandal and quality-control tool in Wikidata. It's "Constraint violation report".

"Constraint violation report" is quality control tool which tells us things, for example, typo, duplicated data, weird data in context of surrounding data, and so on. Let me put here example of "constraint violation report" from my experience. For example, I can see "Unique value violation" list about Terminologia Anatomica ID (d:Property talk:P1323) at d:Wikidata:Database_reports/Constraint_violations/P1323#"Unique_value"_violations (ignore other sections. This is not well customized report. I simply put example from my editing region.)

This list is updated daily basis by bots. But what this list means? This list asks me question like "These two different pages have same TA98_ID. Is this OK?"

• This is sometimes OK (for example, if "Head" and "Human head" have same TA98_ID, that would be intentional. Not error or vandalism.)
• Sometimes not-OK (for example, if "Kidney" and "Brain" have same TA98_ID, that is weird.)

As additional information, these "constraint" is customizable and set at d:Property:P1323#constraints. You can see whole available "constraint" list at here (d:Help:Property constraints portal). As my personal experience, when I exported anatomy-related data to WiKiData from WikiPedia, I fixed many data entries which were suggested as potential errors by "constraint violation report" like as above example. And these days, I sometimes check constraint violation report page.Then I see that data is safe and OK even now.

This simple fact "I fixed many data" and "safe and OK even now" means, to me, that "Data in WikiData has better quality than data which was stored in infobox of WikiPedia". This is very simple point.

There is no substitutable way for "constraint violation report" in WikiPeida, as far as you don't programming your own software like Wikipedia:WikiProject Chemicals/Chembox validation (This is very great project. But it is difficult to imitate it for ordinary editors in other fields.)

As my experience, WikiData is safer and better place for data which has well-formatted style, like identifiers. From this reason, currently I personally feel "more reliable" if identifiers are retrieved from WikiData. Because "constraint violation report" pages guard us. --Was a bee (talk) 19:55, 10 April 2018 (UTC)[reply]

P.S. If you interested in "constraint violation", I think you also interested in more higher level tool "Complex constraint violations". For example, at d:Wikidata:Database reports/Complex constraint violations/P25, some data items which suffice certain criteria are automatically listed. For example, "Mother was born after her child was born". Such list is created by analyzing data of mother-child relation and birth/death in whole 40 million item pages. Although I haven't checked each items, if item is listed there, I can guess that something is wrong. This list is set at property talk page (d:Property talk:P25) using SPARQL query (diff). If you don't know well about SPARQL, you can request query at d:Wikidata:Request a query. This kind of automated error detection, or quality control is possible only in WikiData. --Was a bee (talk) 19:59, 10 April 2018 (UTC)[reply]

@Was a bee: So I see 31 violations on the 47 cases of P1108 (~75%). I know that that is just a one-item test group (others are staggering, I see 94477 for P1057 ('Too many results. 89476 records skipped.'), but how is this going to give me trust that those 31 cases are correct information? And how do you there monitor vandalism if you have already 94477 which have problems? Or when one is there a false positive (hence flagged but correct) and is changed so becomes unflagged and incorrect (Human head and head have the same TA98_ID you say, but they are flagged, if I change one of them they are different, one is incorrect, and they are not flagged ..). To me, the only way to know if something is correct is by locking it down in a checked and correct state. --Dirk Beetstra ^{T C} 06:39, 11 April 2018 (UTC)[reply]

@Beetstra you are mistaking the point of the TA links. Like medicine, anatomy infoboxes link to relevant databases and almanacs. We link relevant articles to the international standard nomenclature for that topic... in layman's terms, like linking to the Oxford or Collins dictionary definition. Thus your concern is about whether there is more vandalism on EN WP (greater views, greater ease of editing, greater edits) or Wikidata (less views, harder to edit, less edits). I don't think there's any data presented in the above discussions to validate either side of this point but it would be interesting to know.--Tom (LT) (talk) 11:35, 11 April 2018 (UTC)[reply]

@Tom (LT): I agree, we don't really know. Point in fact is that I just ran into another case of spamming where WD was targeted broadly (links were added to multiple Qs, where some were rather sought after), and specifically, broader than on any other Wiki where the links were spammed. As with the situation mentioned earlier, one of the links stayed for 14 37 days before being removed (addition 1st of March, removal 7th of April.

But besides that, I think my point is that the vandalism on WD should be significantly less than on en.wikipedia - sneaky number-changing is rather difficult to spot on en.wikipedia (many people don't see it even when it is on their watchlist, you have to either know, or being willing to check - you cannot blindly revert as you have to check whether the number got improved or not, and biting is an issue there), it will be less visible if someone does that through WD. That requires eyes on WD monitoring (especially since the en.wikipedia watchlist, at this moment, does not necessarily show that). And I think my deeper point is, that I really do not see the need why anyone would ever need to change the birthday of Donald Trump, there may be reasons to change the way we display it ('4-11-2018' (MDY, Gregorian), '11-4-1900' (DMY, Gregorian), '25-7-1439' (DMY, Hijri), '11th of April, 2018' (written out) .. the underlying number(s) do(es) not change (P(birthday_day)=4; P(birthday_month)=11; P(birthday_year)=2018; P(birthday_calendertype)='Gregorian'). We can discuss whether data is vandalized, or how much it is vandalized, we can discuss whether the data is generally correct or not (and to what percentage), we can discuss whether the data is preoperly referenced or not, we can discuss the chances whether references were added after the fact, or facts were changed after the reference, my point is that for a lot of data it plainly true that once it is correct, it is never ever going to change (and there is no reason to do change), some data has multiple 'corrects' (3 different papers describing the boiling point of water at different precisions - still each of those is hard-linked to the reference, that reference is not going to change, that number is not going to change, ever), there is data that regularly (but not continuously) changes (but is now correct and will only change upon a change of external database), and there is data that is regularly changing (soit, we have to live with that). But currently, we cannot even trust the truly immutable data. And if I have to check on en.wikipedia the reference supports that today is 11-4-2018, or I have to go to WikiData to check whether the reference that supports that today is the date displayed on en.wikipedia, then I prefer the former. As it stands now, material on WD is not even reviewed whether it is correct with respect to the reference that is there (or maybe multiple WD editors review it, but they don't know of each other that that has been done (since the last change), nor do we know it), let alone that it is locked-down in a correctly supported state.

I wonder how much bogus data there is on WD that is supported by an equally bogus reference, but which would be transcluded onto en.wikipedia because it is 'properly' referenced - point is that if on en.wikipedia a bogus piece of data is entered with a bogus reference it is properly visible, while that material could very well be added after a fully enabled WD infobox is added and hardly anyone notices (and now think BLP). --Dirk Beetstra ^{T C} 12:22, 11 April 2018 (UTC) (adapted --Dirk Beetstra ^{T C} 12:46, 11 April 2018 (UTC))[reply]

@Beetstra: "So I see 31 violations on the 47 cases of P1108 (~75%)." I see 94477 for P1057 This is not problem. There are various kind of constraint (see d:Help:Property constraints portal). And now what you mentioned is violations of "type constraint" (see d:Help:Property constraints portal/Type). As current state of usage, although type constraint is added to many properties, generally this violation is not treated seriously (no one cares so much).

Analyzing cases which you raised,,, In electronegativity (P1108) case, option parameter was not matched to current instance/subclass usage in "atom" items. So I updated that[5]. When next update, most "violation" would disappear. In chromosome (P1057), I fixed setting[6] for matching current main usage of property. It is item type "gene". When next update, number of "violation" would largely decrease.

After constraint setting is nicely configured, "type violations" lighting up "wrong usage of the property". For example, if electronegativity (P1108) is used in "chimpanzee" type page (e.g. Bubbles (Q997482)), it is strange ("What is electronegativity of monkey??"). In such case, I can guess that somebody mistakenly used electronegativity (P1108) at monkey item page. Or somebody changed (vandalized) item type at atom page from "chemical element" to "chimp". For example, as far as I see current type violations of electronegativity (d:Wikidata:Database reports/Constraint violations/P1108), there are no such case. So I can say that there are no vandal at least related to item type in electronegativity.

Your data locking feature is interesting idea. I like that. If locking functionality is discussed in future, I'll follow that discussion with supportive mind :) For example, atomic number of Helium is "2" (d:Q560#P1086). If this data "2" is locked, Wikidata can provide next level reliability. Various new WikiData feature discussion is on going at here (d:Wikidata:Contact the development team). --Was a bee (talk) 16:37, 11 April 2018 (UTC)[reply]

I can see how that might catch greater vandalism, but not the sneaky stuff. As mentioned earlier, if s.o. changes the boiling point of trimethylphosphine to 150 degC, hardly anyone but a hardcore chemist will know that it is wrong. Those who notice may want to check, but to all of them the data seems reasonable.

The locking of data is not something new. There are cases on Wikipedia of locked down infoboxes (template editor protected transclusions of data-filled infoboxes). That data is immutable and referenced, and locked down in that state. The only reason for change is if someone can make a case for change. —Dirk Beetstra ^{T C} 03:38, 12 April 2018 (UTC)[reply]

Template editor protection and transclusion! That locking system is interesting.

About nuance level data quality like examples what you raised (e.g. birthday of famous person, URL of certain official website, boiling point of certain chemical substance, etc) is basically hard to evaluate with automated algorithm in WikiData. About this point, I think, although technological aspect play roles in some degree (e.g. protection is possible or not), the most important factor would finally resulted in "human", not technology nor system. In other words, "Regarding that data, where is the most powerful contributor among all wiki around the world, now?" I put examples:

• 1. If most powerful contributor is in en.wikipedia, then en.wikipedia should not import data from WikiData. But it is better to export data periodically to WikiData and sharing that among other wikis. Although I don't know well about chemistry region, currently chemistry project in en.wikipedia seems such world's best.
• 2. If most powerful contributor is in WikiData. It is better to import data from WikiData. It makes en.wikipedia articles more reliable and excellent. For example, gene/protein data realm is kind of this. I suppose you already know that User:ProteinBoxBot project have shifted their data storage center from en.wikipedia to WikiData. So enormous quantity data related to gene/protein is now in WikiData. They are keeping data in WikiData anew by periodical updates. As fa as I know, this project is world's best project in gene/protein data realm among all wikis.

Because even if locking-system is implemented, if there were no good editor/maintainer, as you can imagine, system itself can't afford good things to users/readers. So "Where is good editor?" is finally matter. Blanket rule would not be efficient. --Was a bee (talk) 21:49, 12 April 2018 (UTC)[reply]

As usual the Wikipedia "discussion" process is plain old "politicking" and a "public" forum for criticizing, influencing, discrediting and trashing the "votes" and opinions and viewpoints of others. To some, at least. One has to wonder just what kind of "democracy" some who demand explanations for the opinions and votes of others are used to in the "real world". And why they react negatively when subjected to similar treatment from others. But then again, those who play the other "team" in the same game and submit to their demands and let themselves be influenced by them have their own role in creating and responsibility for the "democracy" they "create". — Preceding unsigned comment added by 68.234.100.169 (talk • contribs) 19:58, 9 April 2018 (UTC)[reply]

@68.234.100.169: It's WP:NOTAVOTE. The point of an RfC is to draw out why people take the positions they do; and it's entirely appropriate on both sides for people to raise questions, to see whether those positions stand up to challenge and fully consider all the angles. If you think this is "as usual", then you're right, this is "as usual", because this is how it is meant to work, when there are questions to thrash out. Jheald (talk) 20:44, 9 April 2018 (UTC)[reply]

IP 68.234.100.169, if you are referring to Mike Peel then I object. Mike and I are about as far apart on this issue as possible, and I could have asked the exact same question he did. There's nothing wrong with asking someone to expand/clarify on their !vote.

Actually it's not considered improper or unusual to counter-argue or try to change someone's !vote in an RFC. A RFC is discussion on the merits and not a vote. But Mike didn't even do that. Alsee (talk) 23:56, 9 April 2018 (UTC)[reply]

Just a side comment - perhaps not related to the RFC, but worth noting. I think a lot of the resistance to wikidata may stem from the fact that wikidata is not “user friendly”. To understand how wikidata works (and to even read a wikidata page), you have to have some idea of how all the parameters and coding behind wikidata. As someone who is very TEXT based, when I look at wikidata, I become quickly confused by all the parameters, P-numbers, Q-numbers etc. This makes it overwhelming (to the point of impossible) for a text based editor (like me) to figure out what’s going on... how to edit wikidata. Wikipedia is supposed to be the encyclopedia that “anyone” can edit... including technophobes like me. Currently, Wikipedia is relatively easy to edit... it is text based... just go to edit mode and type words ... then hit save. Unfortunately, the more we encorporate wikidata into Wikipedia, the harder it becomes to edit Wikipedia. If we roll this out fully, it will simply confuse and drive away text based editors like me. Wikipedia will become “the encyclopedia that SOME (those who can figure out wikidata) can edit... but others can’t.” Blueboar (talk) 12:16, 11 April 2018 (UTC)[reply]

• Yup, I noted that in my oppose of use of wikidata. I'm pretty techy, and it still took me atleast 10 minutes just to figure out vaguely what is going on and how to edit it and it is still much way more annoying to edit - only one field at a time, to add a statement, you have to do so many clicks and what not; everything is cumbersome, and i have no idea what half the things are about. For something as vital and widely used as infoboxes it should be very easy to edit, add a lot of information manually, or remove etc - there shouldn't be barriers to human editors. So maybe only things that should be updated en mass/with bot can be moved there (identifiers etc) but only very maybe.. Galobtter (pingó mió) 04:07, 12 April 2018 (UTC)[reply]
• Even with the bot thing, you'd have to ensure that the bot edits update correctly, which may not be the case, and one'd probably need to impose something close to enwiki bot policy on there.. Galobtter (pingó mió) 04:10, 12 April 2018 (UTC)[reply]

Well a major advantage of such a change could be fixing of Category:No local image but image on Wikidata (9,087) which has more than 8200 articles at present. If any of the Wikidata bashers want to do that manually, they're most welcome. But I don't see that happen anytime sooner (I mean whatever is not forever) without use of Wikidata (especially if they want to opt for manual editing). Capankajsmilyo (talk) 17:48, 11 April 2018 (UTC)[reply]

Are you now suggesting to automatically transclude those images? It still needs someone to check whether the image on WikiData is the right image. That set is just a (small?) subset of all pages with infoboxes without image. —Dirk Beetstra ^{T C} 03:20, 12 April 2018 (UTC)[reply]

The transcluded image will be checked by anyone and everyone reading that article. But this will add image to thousands of article which don't have one. I don't see how keeping the articles (for which images exist) without an image is better. Further, not every article is GA/FA. There are stubs and starts as well. Do you want to apply a blanket policy on all? Capankajsmilyo (talk) 03:31, 12 April 2018 (UTC)[reply]

Also it might be a good idea to have some statistical data of how many incorrect images were added vs how many correct were before arriving at a decision of blanket ban. Capankajsmilyo (talk) 03:34, 12 April 2018 (UTC)[reply]

Especially for those low view stubs, people won't know what the correct image is supposed to be Galobtter (pingó mió) 03:54, 12 April 2018 (UTC)[reply]

A bot could do it easily if we wanted...it wouldn't be approved though, as no human is checking the images, but if you're going to automatically add the image the same problem is there Galobtter (pingó mió) 03:40, 12 April 2018 (UTC)[reply]

The very first one I checked on was incorrect - it said there was a picture for Héctor Abad Gómez, but was actually was of Héctor_Abad_Faciolince. The second one was fine (Syamer Kutty Abba), so it'd be useful for someone manually/semi-automatically with checking to go through the category, but blanket adding images, definitely not. Looking at the history, the clearly incorrect image was added twice separately by different people, looking at the tag, appears automatically - no bot on enwiki would add an image (presumeably) because it had the same first matching first two words, and anyone who did it semi-automatically would be blocked very quickly. This clearly shows the different standards - of wikidata, anything goes as long as there is more, where on enwiki there is a much more emphasis that bots be very vetted for errors, and edits too. Galobtter (pingó mió) 03:54, 12 April 2018 (UTC)[reply]

That category isn't even right in many instances. The first one I checked was Sarah Onyango Obama, which has an image (the same as in Wikidata), just not in the infobox. Which is good as it is an image of her house, not of the person. Branko Oblak should, according to Wikidata, get the image File:Josip Katalinski at World Cup 1974 against Zaire.jpg (and many other language versions of his article already do, through Wikidata). It is an image of Katalinski though... Humphrey Ocean has the image, but above the infobox, not in it, which allows it to be shown a bit larger. Eddie Ockenden has as Wikidata image a picture of his whole team, which isn't really suitable for an infobox. Neither Wikidata nor the image at Commons gives any indication which person Ockenden is. A similar issue with George Odgers, but there the Wikidata item is included in the article outside of the infobox, with a caption identifying him.

Basically this is a good example of why simply using Wikidata to fill our infoboxes is way too often a bad idea, and local control of the data is preferable. Fram (talk) 12:44, 12 April 2018 (UTC)[reply]

You do have control over local data in options other than 1A. You can always specify a local image in case you don't wanna use Wikidata image. Wikidata images can be fetched to reduce the pages in the category specified. If you want to do it by yourself, you're most welcome. But thousands of articles without any images (when we do have them) does not make sense to me. User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:17, 18 April 2018 (UTC)[reply]

Not with 4E and 4F. And if we have to override Wikidata all the time, then what's the use of a Wikidata-based infobox instead of using a local one immediately? Fram (talk) 04:27, 18 April 2018 (UTC)[reply]

2-3 instances won't qualify a descent sample size I guess. How many incorrect images vs correct images were imported in say 1000 imports? User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:43, 18 April 2018 (UTC)[reply]

If half the images are incorrect a sample size of 10 is enough really. I checked another two: File:UCV_2015-210_Harry_Abend,_1994.JPG was a image of Harry Abend's work which you wouldn't want for his infobox; (one image of which is already there in the article). Another the image was fetched from wikidata and was fine. The next had a sculpture image: File:June_Gordon_Marchioness_of_Aberdeen_and_Termair_bronze_bust_by_sculptor_Laurence_Broderick.jpg which was included in the article and shouldn't be in the infobox. The next one was Lauren_Ambrose; The image, File:Lauren_Ambrose_2000.jpg, was already included below and may not be good for the infobox (and definitely shouldn't be repeated both in the infobox and below). So 1/4 for usefulness, I suppose, 2/4 - no that isn't a suitable infobox image - and 1/4 already included. Galobtter (pingó mió) 04:59, 18 April 2018 (UTC)[reply]

The statistics could be very different for different classes of articles, for example BLP and localities. Even if 50% BLP images are incorrect (which still sounds very strange to me), it could be that only 1% of locality images is incorrect.--Ymblanter (talk) 05:13, 18 April 2018 (UTC)[reply]

The argument have to override Wikidata all the time, then what's the use sounds like if we have to monitor/edit the Wikipedia all the time what's the use. It also assumes that each and every image of Wikipedia is incorrect which is an absurd and unsubstantiated assumption. User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:43, 18 April 2018 (UTC)[reply]

The essential point is if we're overriding it half the time, then it isn't useful. If we introduce 4000 incorrect images for 4000 correct ones that doesn't seem overall improvement - there'd be a massive drop in quality. Galobtter (pingó mió) 04:59, 18 April 2018 (UTC)[reply]

How are you so sure that it's 50/50 and not 95/5 (correct vs incorrect) on a sample size of 8000. I don't think anyone has even assessed 100 images till date. If so, please share actual stats. Further Ymblanter's point carries weight that this will differ by Wikiproject. Telescopes already use it and are happy with it. BLP don't. So we might need to consider the breakup of such stats by Wikiproject before arriving on any conclusion. This might be suitable for some type of articles and not for others. That must be thoroughly assessed. User:Capankajsmilyo^{(Talk | Infobox assistance)} 05:24, 18 April 2018 (UTC)[reply]

Even a small sample size of 4 shows 4/6 - 66% - useless/bad. According to statistical significance calculation using an online calculator, it is 66%+-37.89 - so only <1% chance it is even 30% good. Bascially, itwould be extraordinarily unlikely if I just happened to find a very bad sample. I'm not sure why it would be better with non-BLPs. (you are welcome to disprove that though, if you wish to check some of the images, on BLPs and non-BLPs) Galobtter (pingó mió) 05:35, 18 April 2018 (UTC)[reply]

In essence, we have no reliable data on how reliable Wikidata is, and from personal experience many of us have found major problems with some of the data. It would be gross irresponsibility to massively import data that is unreliable and unchecked, but also excessively restrictive not to allow the use of data which is reliable and has been checked, providing that it can also be reasonably easily curated by the end users, ie. Wikipedians. Both the "no Wikidata under any conditions" and the "always use Wikidata" extremists are shutting doors unnecessarily. Wikidata has a long way to go, but it might get there. Some projects are more sensitive to bad data, they should not take the chance until it can be shown that Wikidata fits their requirements. Other projects find Wikidata useful, let them use it, they will test the systems and get changes made, and as time goes by Wikidata may become acceptable to more projects, and they can start to use it. Trying to force this either way will not work well, and will polarise the communities. This has parallels with WMF pushing unready software. When pushed, people push back, sometimes quite violently. The status quo allows us to bicker interminably at a sort of ground level, and get some work done. That is how it goes at Wikipedia. · · · Peter (Southwood) ^(talk): 05:55, 18 April 2018 (UTC)[reply]

Exactly that.--Ymblanter (talk) 06:05, 18 April 2018 (UTC)[reply]

I do understand that, but IMO I don't think the data quality is ever going to improve, nor is there a way to make sure to get the certain data that is reliable and checked, and thus the experiments cannot be limited to that; I think that is inherent in wikidata policy, of its desire to get as much information as possible, with minimal BLP and sourcing policy, acceptance of crappy sources et al. If it ever improves - and this will take a long time - then we can reevaluate, but currently allowing experiments/wikiproject use can/has easily allow bad data. There is nothing preventing bot importation of data directly from the high quality sources, as Beetstra points out - wikidata doesn't offer exclusive data, and thus it wouldn't be "excessively restrictive not to allow the use of data which is reliable and has been checked" (is data "checked" on wikidata - most of it is bot imports, which we can as easily do here if we wanted - often we don't want, because bots are bad with ambiguity, similar names et al, but problems with that, as I pointed out above about wrong images being added seem to exist on wikidata - so it is more like mass importation from wikidata circumvents our own community desires regarding data)

Not only that, I think many people will be confused by the extra steps to edit on wikidata, and I expect lot less editing of infobox fields if they are shifted to wikidata - from confusion of where the data is, to learning a new interface etc. Having the data on wikidata is basically a barrier to editing Galobtter (pingó mió) 08:34, 18 April 2018 (UTC)[reply]

Regarding directly importing, that is what I have been doing in a grey past on Wikipedia. It is an enormous task .. you have to check every single datapoint: is this identifier really describing this subject? That takes up to minutes per subject. For a lot of data you cannot say that you are correct by blind bot-importing directly from outside databases. I would be surprised if you get >95% correct (and that is partially because also Wikipedia is not correct in the first place). Below an editor talked about 95% correct, that is by FAR not satisfactory. --Dirk Beetstra ^{T C} 08:46, 18 April 2018 (UTC)[reply]

Regarding the difficulty of making sure that identifier is really describing the subject. Agree 100%, indeed that is what I was saying about "often we don't want, because bots are bad with ambiguity, similar names et al, but problems with that, as I pointed out above about wrong images being added seem to exist on wikidata - so it is more like mass importation from wikidata circumvents our own community desires regarding data". In Wikidata importations that checking isn't done. Galobtter (pingó mió) 15:45, 18 April 2018 (UTC)[reply]

Thanks to the enabling of automatic image inclusion from Wikidata (i.e. showing an image in an infobox without the image being defined here), we now have many articles with the same image twice, either above and inside the infobox (e.g. Eshin Nishimura), inside and below the infobox (James Preston Poindexter), in two separate infoboxes (e.g. Robert Little (minister)) or twice inside the same infobox(!) (e.g. William T. Dixon, John M. Brown, Christopher Payne, William R. Pettiford, Rufus L. Perry, Benjamin F. Lee, Joseph C. Price, Joseph Endom Jones, ...) These are clear examples of the implementation of such Wikidata imports making articles actively worse here. Fram (talk) 07:27, 18 April 2018 (UTC)[reply]

Do you really think that thise images should be outside infobox rather than being in it? As far as I see, in the examples given above (by you) Wikipedia editors placed those images (incorrectly) outside infobox, when they should have been inside it. And no-one, especially Wikipedia administrators opposing wikipedia did anything to fix that. User:Capankajsmilyo^{(Talk | Infobox assistance)} 08:01, 18 April 2018 (UTC)[reply]

You have noticed that the vast majority of examples I gave had images inside the infobox already? And as far as I can tell, there is no rule that says that images should be in the infobox and not above it (WP:MOSIMAGES and the linked page certainly don't mention it). Apart from that, can you please stop with the needlessly inflaming add-ons to your posts? From "Wikidata bashers" in your first post to this subsection to "especially Wikipedia administrators opposing wikipedia" (admins have no more or less authority over content than any other editor, and I think you mean "wikidata", not "wikipedia"), you insert comments which don't help this discussion at all and needlessly antagonize and divide the discussion. Fram (talk) 08:07, 18 April 2018 (UTC)[reply]

had images inside the infobox already who said that? Why are you trying to put words in my mouth? I said they should be inside infobox. Are you saying that Wikipedia prefer infoboxes without image + image above/below it. I see no common-sense especially no design-sense in that. User:Capankajsmilyo^{(Talk | Infobox assistance)} 08:13, 18 April 2018 (UTC)[reply]

I didn't put words in your mouth, I was asking you whether you noticed that most of them already had an image in the infobox before the automatic wikidata inclusion of an image created a duplicate image in it? The 8 examples I gave for "twice inside the same infobox" and the one example of "in two separate infoboxes". Fram (talk) 08:21, 18 April 2018 (UTC)[reply]

(edit conflict)No, most of them are because you didn't disable fetching from wikidata if the infobox was embedded, actually - those were the cases of two images in one infobox. You should make sure your changes to the infobox don't cause problems, Capankajsmilyo; I've reverted the edit for those problems. Those uses may need improvement, or they may simply be a different style- either way that doesn't mean problems should occur. Galobtter (pingó mió) 08:18, 18 April 2018 (UTC)[reply]

The stats given above for incorrect images don't confirm with the actual ones. A simple check on {{Infobox religious biography}} shows that out of 654 articles that have images on both Wikidata and Wikipedia, 75% were exactly same. Remaining 25% were the ones that were different, and out of those, I doubt that even 50% (i.e. 12.5%) are incorrect ones. So in total the experiment done on a sample of 654 shows >80% (I would say >95% in personal opinion) images confirm to Wikipedia standards. User:Capankajsmilyo^{(Talk | Infobox assistance)} 08:31, 18 April 2018 (UTC)[reply]

As opposed to what would be nearly 100% of the images that we have do comply with Wikipedia policies, which overall still means we go down. Now, you checked only one datapoint, if you score 95% on 1, you score 77% on 5, 60% on 10, 35% on 20. And that on BLPs and in an infobox that has even more fields. --Dirk Beetstra ^{T C} 08:37, 18 April 2018 (UTC)[reply]

80% is terrible... even 90% is very very bad - I'd expect >99% of our images to be correct. And of the extra ones it'll be even worse; the similar ones are likely imported from us, and give no benefit. Galobtter (pingó mió) 08:42, 18 April 2018 (UTC)[reply]

(ec)Many images on Wikidata are taken from them being used on the enwiki article in the first place. Taken these into the count of how many images on Wikidata are correct is circular reasoning. The question is whether, for images where we don't have an image in our article yet, or where the image on Wikidata is different to the one used here, how many of the images we find on Wikidata are acceptable for use in an infobox here, or are an improvement over what we already have in the second case. Fram (talk) 08:44, 18 April 2018 (UTC)[reply]

• I agree - I'd never heard of the category & the first 2 I looked at were wrong - #1 Kenneth Allen (physicist) had an image here (and in the infobox), #2 Harry Abend (an artist) had an image of a work on WP (not in infobox), but no portrait image either on WP or WD. Mind you, #3 Kirsty Alley worked ok, & has gone from the cat since I added a pic. But not an impressive argument for WD use. Johnbod (talk) 13:31, 21 April 2018 (UTC)[reply]

Just for clarity, the 1F&(4E|4F) combination still allows that if someone sees a value being transcluded from WikiData with which the observer disagrees (either it is wrong, because it does not have our flavour of display, etc.) to be overwritten with a local value (I know, if it is wrong it would be preferred that it is corrected on WikiData .. but that aside)? --Dirk Beetstra ^{T C} 12:27, 12 April 2018 (UTC) (adapted to include option 4E, which is the same, thanks User:Fram --Dirk Beetstra ^{T C} 14:00, 12 April 2018 (UTC))[reply]

That's not how I understand it. In 4E and 4F, Wikidata data always gets precedence, only when there is no Wikidata value can a local value be shown. If you see an error on Wikidata, you need to correct it there. How we should deal in this scenario with data which is acceptable on Wikidata, but unacceptable here (e.g. because of different sourcing requirements, different BLP policies, ...) is not clear. Simple things like genre warriors on music articles can then simply move to Wikidata, and then we would need to get them blocked or restricted there or the Wikidata item protected there, to get a stable infobox here.

Basically, we can no longer enforce our own policies if 4E or 4F gets accepted. Fram (talk) 12:55, 12 April 2018 (UTC)[reply]

Well we can. Its called 'Remove/replace the infobox with one that doesnt call wikidata' as I am not about to start messing around on another project to solve their problems. Only in death does duty end (talk) 13:00, 12 April 2018 (UTC)[reply]

@Only in death: but that would violate 1F 'Every infobox that is technically ready to convert from local content to Wikidata content may be implemented in mainspace' - if that passes everyone who would then re-replace the infobox with the WikiData would abide by the '1F' consensus here, and stand fully in their right. --Dirk Beetstra ^{T C} 13:55, 12 April 2018 (UTC)[reply]

Anyone who replaced an infobox (without fixing it) that contained vandalised/blatantly incorrect data would be equally at fault as whoever altered the data on wikidata and would end up sanctioned accordingly. 1F would not in any way negate our content and sourcing policies. It would just as Fram points out, make it hard to enforce without taking heavy-handed actions. Only in death does duty end (talk) 14:04, 12 April 2018 (UTC)[reply]

.. and that starts already by editorial choice, not only vandalism and blatantly incorrect data, think D-M-Y/M-D-Y, °F/°C/K, ise/ize, ... I cannot fathom what disagreements (my understatement of the day) this all would cause. --Dirk Beetstra ^{T C} 14:11, 12 April 2018 (UTC)[reply]

Wikidata only allows files from Wikimedia Commons. What would happen to non-free fair use logos used in infoboxes? Daylen (talk) 04:15, 14 April 2018 (UTC)[reply]

There are some technical options to do that (using local data, using different property, and so on). But I think simply that it is not suitable to replace such data field by Wikidata (e.g. logo field in company infobox, etc). Wikidata is useful, but only for limited data field. Not omnipotent. --Was a bee (talk) 06:00, 14 April 2018 (UTC)[reply]

Sure, fair use images must stay here, and the corresponding data can not be replaced.--Ymblanter (talk) 07:38, 14 April 2018 (UTC)[reply]

You can always keep and use local images. Wikidata images can be used when no local image is specified. User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:22, 18 April 2018 (UTC)[reply]

Not with 4E and 4F. Fram (talk) 04:25, 18 April 2018 (UTC)[reply]

Except for the extremists, 1A/1F, middle ground seems achievable if we include wikidata but keep an option to override the values with local data. I am not sure extremists are ever gonna change their views, especially 1A ones. But for others, there are ample workarounds and consensus options. A few arguments I've read was Wikidata is about "only bots" which is untrue. Most of the edits on Wikidata I have seen (for Property values) were by humans. Another was that by piping Wikidata control get over. Again untrue. By piping Wikidata, you just add another option to add data. You can always customise using |fetchwikidata=, |onlysourced=, etc. Plus you can override with local values too. Another untrue statement was that enwiki is a source at Wikidata. Although it's a source there, but the module and the template has been coded in such a way that it does not pipe in values from sources mentioned as enwiki, etc when |onlysourced= is set to yes. There seems to be a lot of misconception and misinformation prevailing about Wikidata which needs to be calmly looked at before arriving on any concensus. Also, it is worth considering that this discussion is not about Wikidata, but modules and templates fetching values from there which has been customised to enwiki policies. Capankajsmilyo (talk) 09:56, 14 April 2018 (UTC)[reply]

The only problem being that a middle ground is in this case a recipe for continuing fights. Note that infoboxes and WikiData both already have been looked at by ArbCom, this is material that could end up there again and again.

There are cases on Wikipedia where we have chosen an extremist stance - NFCC being one. Sometimes it is the better choice, and we can always revisit this at a later stage. At least for me, I can foresee scenarios where I can be convinced in the futere. —Dirk Beetstra ^{T C} 10:03, 14 April 2018 (UTC)[reply]

modules and templates fetching values from there which has been customised to enwiki policies. Some of the people working on these modules and templates have absolutely done an admirable job. However, it would be next to impossible to design a biographical infobox template that automatically pulls all of its information from Wikidata, and have it comply 100% with enwiki policies without human review and intervention. Plus, the implementation you describe with onlysourced etc is the current status; that's not the only option presented in this RfC and not what you yourself supported. Nikkimaria (talk) 12:55, 14 April 2018 (UTC)[reply]

Why should this template pull 100% info from Wikidata? I though this RfC is about 0% vs non-zero. I do not think there is a significant population of users who want to make every template pull all info from Wikidata.--Ymblanter (talk) 13:46, 14 April 2018 (UTC)[reply]

The 1F,4F combination would pull all the data that any implemented template requested. In the case of the {{Infobox person/Wikidata}} template, this would be 23 data items (some of which can have multiple values). By default, only items that are sourced are included. The only option for an Wikipedia editor would have would be to suppress individual items. Of the 26 !voters who expressed a preference, 3 went for the 1F,4F combination. Robevans123 (talk) 15:21, 14 April 2018 (UTC)[reply]

My position on this has been that a "one-size-fits all" approach is a bad idea. I can understand the concerns about BLP, but do those same arguments apply to, say, Template:Infobox road, which does use Wikidata for automatically copying over maps between wikis? --Rschen7754 17:46, 14 April 2018 (UTC)[reply]

Pardon the parody, but we should seek a "middle ground" about which side of the road to drive on. Either we can all drive in the middle, or we can drive on either side at random. The responses to this RFC cluster at the two ends for good reason. Aiming for some kind of "middle" is the worst possible outcome... and it's an unstable mess. Either we should fully embrace the positives of Wikidata and use it everywhere it makes sense (1F), or we should fully eliminate the negatives of Wikidata and not use it (1A). It's massively disruptive to have the two sides fighting over each template one by one, and it creates months disruption and staggering wasted work when even a single template or template-field is converted in one direction then re-converted in the opposite direction. No, no, no, and hell no. Flipping back and forth is unacceptably disruptive and wasteful. Alsee (talk) 14:02, 16 April 2018 (UTC)[reply]

That is completely not my experience with Template:Infobox road and its pulling the map data from Wikidata. There have been plenty of fights over that infobox, but none related to Wikidata. --Rschen7754 18:14, 16 April 2018 (UTC)[reply]

I doubt that traffic on two-dimensional roads is a good analogy here. Wikidata integration does not become magically better: for that it needs a pathway to success. We could discuss a bit more where we would like to end up eventually, provided that all technical challenges and quality assurances are being met. Most of the non-A options are a careful step towards the potential that Wikidata has to offer. We need some level of integration, even if it is only in some well-defined corners of our universe, to provide space for the technology to develop. If there are actual scenarios in play, we can learn what works well, and what doesn't. It's not as much on which side of the road we'd like to drive, but perhaps a question of whether we want to start thinking about permitting flying cars - and under which conditions. effeietsanders 21:48, 16 April 2018 (UTC)[reply]

There's a reason why question 1 isn't just yes/no, and that's because middle grounds are possible, providing they are agreed upon and are followed without editwarring. I think we're seeing groups at either end because people hold strong views here, but that doesn't rule out the middle options. Thanks. Mike Peel (talk) 21:50, 16 April 2018 (UTC)[reply]

I disagree, we don’t need to facilitate to help WikiData to develop, we do not need some level of integration. Moreover, we have run this experiment now for some time, but there was no drive to improve, there is a drive to implement and a drive to extract (and I can’t escape the feeling that the current fights about infoboxes and about wikidata are actually about this). Again like NFCC, we do not allow experiments, we have clear rules that cannot be negotiated. You can do your experiments in project namespaces, but 1A is simply a ‘no wikidata infoboxes in mainspace’. —Dirk Beetstra ^{T C} 03:46, 17 April 2018 (UTC)[reply]

@Beetstra: "no drive to improve" is harsh and unjustified. Just look at the work that's been going into Module:WikidataIB, and the various infobox templates, to improve support for different features. And bear in mind that South Pole Telescope, the example given above, didn't have an infobox before this work started! Mike Peel (talk) 22:10, 17 April 2018 (UTC)[reply]

@Mike Peel: Wikidata has run bots that imported spam, just approved of the data promise. WikiData runs bots at incredible import speeds. There is no mechanism to protect data with respect to their references. Lots and lots of material is unreferenced, and there is no guarantee that data is correct with respect to the references there are. It is an ocean of data.

We had, with a number of editors, a drive to verify a crude 250000 numerical, nearly immutable, identifiers, and that did never finish, and is in an inbetween state, now unmaintained. I have had hope on WikiData, and there have been some discussions around that, but seen how such drives go on en.wikipedia, I will not hold my breath. I am sorry, we have different goals, and those goals do not scale linearly with en.wikipedia’s goals.

(that is a red herring, right: en.wikipedia could just have added an infobox and get the data - that could very well have been done without any help from Wikidata). —Dirk Beetstra ^{T C} 03:32, 18 April 2018 (UTC)[reply]

The unsourced data or data having Wikipedia as source doesn't get piped with |onlysourced= set as yes. So why are you trying to imply that it does. User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:26, 18 April 2018 (UTC)[reply]

Sure, you can omit all data that is unsourced, problem is that there is no control over the data that is sourced. I have to assume that all is properly sourced and that it has not changed since. --Dirk Beetstra ^{T C} 07:08, 18 April 2018 (UTC)[reply]

Actually there is and it's implemented using abuse filter. User:Capankajsmilyo^{(Talk | Infobox assistance)} 07:51, 18 April 2018 (UTC)[reply]

Looking at the log (500 most recent entries), I see no filter actions for "changed sourced data" or something similar. "Adding unsourced ethnicity" is the only one. Can you please indicate which abuse filter you mean. Fram (talk) 08:01, 18 April 2018 (UTC)[reply]

There are no tags for this at all. 'new editor changing data' is the closest maybe. Quite some of those are either good faith or plainly correct, though. --Dirk Beetstra ^{T C} 08:33, 18 April 2018 (UTC)[reply]

• I want to give another general thought, and this "middle way" seemed to be a place to do it. There are some infoboxes where the content is really about data. Like Template:Infobox gene. I think the data-driven biology people have a nice relationship with the wikidata people and a bunch of them are wikidata people, and the data that gets presented in our infobox is technical stuff that to be frank general readers don't care about. I don't want to get in the way of that. But for infoboxes that have content that everyday people would care about -- that takes no special mojo to understand or think about -- I am very much against Wikidata being used. That is why I am supporting 3A above anything else, and explicit consent only, so that we can allow it where it is OK but rule it out where it has a risk of being harmful. Jytdog (talk) 20:11, 7 May 2018 (UTC)[reply]

Dates in Wikidata are defined as Universal Time without the possibility of specifying a time zone. Dates in English Wikipedia are generally local time. These are not compatible. Therefore, dates should not be imported from Wikidata. Jc3s5h (talk) 15:17, 19 April 2018 (UTC)[reply]

That is generally not an issue. From what I'm aware, there was a single telescope article affected by that - the first light time for Hale Telescope. Where it is an issue, then the dates can be locally defined, or perhaps a module could be written to do the time zone conversion (since it's mostly simple arithmetic). Thanks. Mike Peel (talk) 15:32, 19 April 2018 (UTC)[reply]

Sources normally give dates in the local time zone, but more often than not, do not state the time of day. It's impossible to correctly input such a date into Wikidata; Wikidata can't represent it. Thus the majority of Wikidata dates are garbage. Modules can't remove the stink from garbage. Jc3s5h (talk) 16:38, 19 April 2018 (UTC)[reply]

So what you're saying is that we shouldn't trust any date on a any website unless it also specifies a timezone - which we don't normally do here on enwp? GIGO? Thanks. Mike Peel (talk) 20:27, 19 April 2018 (UTC)[reply]

Most sources, including most websites, do not explicitly state time zones. This means the time zone must be determined from context. Most sources that Wikipedia uses place the responsibility to make whatever time zone determinations are desired upon the reader. Most Wikipedia articles do exactly the same, place the responsibility of determining the time zone upon the reader, if the reader needs that degree of accuracy. Wikidata does not permit entering a date with an unspecified time zone, so is adding a time zone statement (Universal Time) to statements from sources that were agnostic about time zones. In effect, Wikidata is misquoting sources. Jc3s5h (talk) 07:18, 20 April 2018 (UTC)[reply]

I do not really see the problem here - I agree that the source timing could be off by about 23 hours (as my DOB would be one day earlier or one day later depending on where you were observing my birth), but there is also no problem for WikiData to include next to years, months, days, hours, minutes and seconds a timezone qualifier (where needed) and a calendar qualifier (Gregorian / Hijri / other systems). Whether all of those need to be filled in on a date is up to the person who adds that data. All of that can then be calculated back. --Dirk Beetstra ^{T C} 07:37, 20 April 2018 (UTC)[reply]

Wikidata lacks the development bandwidth to fix the problem. The bugs have been reported years ago and only a few of the multitude of problems have been fixed. See T87764 Jc3s5h (talk) 07:47, 20 April 2018 (UTC)[reply]

@Jc3s5h: I am going to rephrase that: 'Wikimedia lacks the development bandwidth to fix any problem'. Anyways, the only requirement I see is that things should work per en.wikipedia standards (or better) on wikidata if we are to import their data. --Dirk Beetstra ^{T C} 08:41, 20 April 2018 (UTC)[reply]

Currently, consensus appear to be heavily skewed towards the 4 "A options". However, I personally think that 3A is a bit too excessive. Therefore, let me make the case for 3B.

Non-contentious content which is well sourced within the article does not usually (actually, almost never) have an additional source (or a ref tag pointing to a source elsewhere in the article) just because it's in the infobox. For example, all info in this infobox is non-contentious and cited elsewhere in the article. No need for a repeat. Same for this other example and plenty of others. In the spirit of avoiding WP:CITEOVERKILL, I don't see why we should require duplicate/repeated sources simply because something is summarized in an infobox. 198.84.253.202 (talk) 16:17, 21 April 2018 (UTC)[reply]

This is moot is there is consensus for 1A, which means no Wikidata in the first place. {{3x|p}}ery (talk) 22:31, 21 April 2018 (UTC)[reply]

This does however show one of the problems with WikiData integration. If the data is the same as in the text, and in the text it is referenced, then I could indeed say that the data is properly referenced. But you get a ‘say where you got it’ type of problem in most other cases where the material is contentious: The material on WikiData may have another reference than used anywhere in the document, or material in infoboxes that is not ‘prose material’ and hence not referenced in the text. (And for the inverse, there is material that is properly referenced on en.wikipedia in the prose, and generally included in the infobox, but which would not be transcluded because it is currently not referenced on WikiData (e.g. data that was currently imported from an unreferenced infobox on en.wikipedia, because it is thoroughly referenced in prose). There is still so much to do on WikiData before en.wikipedia is ready for this. —Dirk Beetstra ^{T C} 03:59, 22 April 2018 (UTC)[reply]

I was just now editing Windmill Hill, Avebury (Infobox World Heritage Site) when I spotted the vandalism visible in the screenshot. Getting rid of the vandalism here isn't hard. But I am unable to find the vandalism on Wikidata, never mind correct it.

The same vandalism also affects our (enwiki) infoboxes in Beatrice oil field (Infobox power station), Sandys Row Synagogue (Infobox religious building), The Sharp Project (Infobox project), Half-Mile Telescope (infobox telescope), ... (each time, I only give one example, but as I am writing this Google finds 67 articles with this vandalism). So this is one bit of stupid vandalism, affecting a wide range of articles here, but which can't easily be found and reverted even by experienced enwiki editors.

It turns out that this is a bit of vandalism which is over 24 hours old, but was (only?) live at Wikidata for 25 minutes: someone changed the English label for "United Kingdom" to "jajaxedlol"United Kingdom. Although this has been reverted more than 24 hours ago, it has a lasting effect on dozens of pages here (Google lists 67, searching on enwiki directly gives 242 pages[7]).

So 242 (or more) pages, across multiple Wikidata-enabled infoboxes, have been vandalized in this most basic way (nothing clever or sophisticated) for more than a day, because a) vandalism detection and reversion on Wikidata is too slow and haphazard, and b) the changes on Wikidata get ported to enwiki at seemingly random intervals, causing shortlived vandalism to remain live here anyway. Fram (talk) 11:34, 25 April 2018 (UTC)[reply]

As I understand it, we have a parser that stores the parsed page. In that, it does not re-parse the infobox until the page is edited, and it gets only re-parsed after x amount of time. So if a vandalized page with a template transcludes a wrong value, it will put the vandalized WikiData data in the page on a refresh, and only remove it on the next refresh. With some bad timing that could indeed stay for hours while it is only live for seconds on en.wikipedia.

If a page on en.wikipedia gets vandalized, the page gets reparsed - when it gets reversed, it gets reparsed. Template-vandalism on en.wikipedia could stay longer, as that depends on the same refresh of parsed pages. --Dirk Beetstra ^{T C} 09:26, 26 April 2018 (UTC)[reply]

Which is why we have protected thousands of templates on enwiki, to avoid as much as possible that one vandal can vandalize hundreds of articles at once. In this case, an editor could actually change the label (the "title" of the article on Wikidata) of the United Kingdom without any problem and without being reverted for more than 20 minutes, affecting many enwiki templates at once. It's a kind of problem we finally mostly have gotten under control on enwiki, but where we would then open the gates again by relinquishing our control. Fram (talk) 09:43, 26 April 2018 (UTC)[reply]

And that is why it should be possible to protect immutable data ('individual Qs belonging to individual Ps') on WikiData (and probably also a whole set of Ps, Qs can already be protected). --Dirk Beetstra ^{T C} 09:54, 26 April 2018 (UTC) --Dirk Beetstra ^{T C} 09:54, 26 April 2018 (UTC)[reply]

• I can confirm Fram's observations on the "jajaxedlol" issue, which I saw yesterday while working through some Template talk:Infobox World Heritage Site#Step by steps. I couldn't find out what had caused it (as Fram apparently eventually did), so after thinking some time about how to report it here I eventually didn't while not knowing how to present it in a way that illustrated what exactly was going wrong. --Francis Schonken (talk) 10:47, 26 April 2018 (UTC)[reply]

Ah, every time I see I only see more problems with wikidata.. Galobtter (pingó mió) 15:15, 27 April 2018 (UTC)[reply]

We are now full week after this vandalism edit, but an en.wikipedia search results in hundreds of results for jajaxedlol. For the same reason, the catalan Wikipedia has hundreds of pages for gilipollas.

Although the same thing would be happening, likely, if Heart of Neolithic Orkney was changed to have those parameters, it would still only be one page that would be affected. That would indeed be worse if the template was changed (but the template in this case is protected ..).

The possibilities of this are infinite. --Dirk Beetstra ^{T C} 07:05, 1 May 2018 (UTC)[reply]

As of now, searching for 'neolithic orkney heart united kingdom' within Wikipedia does NOT find the article Heart of Neolithic Orkney, as that term was not contained in the (vandalized) article at the time of indexing. --Dirk Beetstra ^{T C} 08:44, 1 May 2018 (UTC)[reply]

@Beetstra: I can't reproduce either of these search results. The first search only returns this page, the second one does find the article you mentioned. Mike Peel (talk) 12:53, 1 May 2018 (UTC)[reply]

@Mike Peel: wow. For me now the same. Does performing the search update induce the reindexing? Anyway, this google search still worked for me at this moment]. --Dirk Beetstra ^{T C} 13:19, 1 May 2018 (UTC)[reply]

@Beetstra and Mike Peel: Oddly, I tried the first search earlier this morning (~10 am), and got 100+(?) articles, but the first few articles I looked at didn't actually contain "jajaxedlol". Go figure... Robevans123 (talk) 18:12, 1 May 2018 (UTC)[reply]

@Robevans123: the ones I checked also did not contain the vandalism anymore. All cached data I presume. —Dirk Beetstra ^{T C} 19:14, 1 May 2018 (UTC)[reply]

• WP:CIRCULAR example:
  • 10:16, 31 May 2017: the Madara Rider article is vandalized
  • 01:59, 9 July 2017: the vandalized version of the image caption in the infobox is transferred to Wikidata (obviously without checking the reference as given in the infobox, which is 43, at the UNESCO World Heritage Site website)
  • 01:59, 9 July 2017: local data of the infobox are deleted at Wikipedia, thus importing the erroneous image caption directly from Wikidata (= infraction on the WP:CIRCULAR policy).

--Francis Schonken (talk) 10:00, 1 May 2018 (UTC)[reply]

WP:CIRCULAR doesn't apply since no source is being claimed here. Thank you for fixing enwp vandalism that lasted for 11 months though! Mike Peel (talk) 12:50, 1 May 2018 (UTC)[reply]

"Also, do not use websites that mirror Wikipedia content or publications that rely on material from Wikipedia as sources. Content from a Wikipedia article is not considered reliable unless it is backed up by citing reliable sources. Confirm that these sources support the content, then use them directly". We are using material from a website that mirrored Wikipedia data, and that does not provide a reliable source for said information. Here, the mirror is unsourced as the person mirroring it did not copy a source, whereas our Wikipedia article has a reference regarding the naming. --Dirk Beetstra ^{T C} 12:55, 1 May 2018 (UTC)[reply]

No, you're misunderstanding what you're quoting. No-one is saying that we should use Wikidata as a reference for the information, which is what 'source' means in this case. Mike Peel (talk) 13:02, 1 May 2018 (UTC)[reply]

And I think that that is part of the problem that is being discussed here .. is taking data from WikiData 'sourcing' data, and is WikiData then at the same moment the reference to said data ("The word "source" when citing sources on Wikipedia has three related meanings: .. The piece of work itself (the article, book)", WP:V). If I use data from website X, then I say where I got it, 'Website X', if I use stuff from WikiData, then I should say where I get it, 'WikiData', which is hence my source of data. That is why I think that we should not use data from WikiData, sourced or unsourced, unless that WikiData can vow for the data, and that it is correctly referenced. I am sorry, the sourcing requirements on WikiData have to be more strict than on en.wikipedia, you may have to have reference for material that would not need references on Wikipedia. --Dirk Beetstra ^{T C} 13:19, 1 May 2018 (UTC)[reply]

@Mike Peel: what an abject nonsense. I say there's a WP:V infraction (as in the lead sentence of WP:V: "In Wikipedia, verifiability means that other people using the encyclopedia can check that the information comes from a reliable source"), that is, an infraction by not heeding the WP:CIRCULAR part of that policy. Then you start Wikilayering – please think through what you're trying to say: not a WP:V infraction? Then plain and simple vandalism... or did you mean that the 10:16, 31 May 2017 edit was not plain and simple vandalism? If you say that your error is not that you erroneously and unknowingly sourced your error to a Wikipedia vandal, then what you did is drawing in vandalised content from an unreliable source... In my book, if the WP:V transgression excuse is discarded, all what remains is plain vandalism. It is my strong opinion that you shouldn't be let near to a single Wikidata infobox, certainly if you start to prefer "vandalism" as an excuse over "WP:V infraction". --Francis Schonken (talk) 13:43, 1 May 2018 (UTC)[reply]

@Mike Peel: Actually, what you did there is a perfect example of one of my concerns regarding WikiData: the collection of data is (much) more important than any check that what is imported is correct, that there is no (sufficient) drive to have the data correct. The motto is: import, import, import. Combine that with the fact that the data, referenced or not, can be (and is) vandalised, and that it pays to spam WikiData as a shortcut to 800+ wikis. —Dirk Beetstra ^{T C} 19:14, 1 May 2018 (UTC)[reply]

Similar to #Jajaxdelol above:

• 14:19, 28 May 2018 – d:Q183 is vandalised
• 14:40, 28 May 2018 – vandalism reverted
• 07:13, 31 May 2018 (UTC) The "Trocolandia" spoof still appears on a few dozen English Wikipedia articles

--Francis Schonken (talk) 07:13, 31 May 2018 (UTC)[reply]

I started Wikipedia cleanup:

1. [8] – still 29 instances to go --Francis Schonken (talk) 07:49, 31 May 2018 (UTC)[reply]
2. [9] – still 28 instances to go --Francis Schonken (talk) 08:05, 31 May 2018 (UTC)[reply]
3. [10] – still 27 instances to go --Francis Schonken (talk) 08:32, 31 May 2018 (UTC)[reply]
4. [11] – still 26 instances to go --Francis Schonken (talk) 09:13, 31 May 2018 (UTC)[reply]
5. [12] – still 25 instances to go --Francis Schonken (talk) 09:24, 31 May 2018 (UTC)[reply]
6. [13] – still 24 instances to go --Francis Schonken (talk) 09:47, 31 May 2018 (UTC)[reply]
7. [14] – still 23 instances to go --Francis Schonken (talk) 09:58, 31 May 2018 (UTC)[reply]
8. [15] – still 2 instances to go (some refreshing finally seems to have kicked in) --Francis Schonken (talk) 10:12, 31 May 2018 (UTC)[reply]
9. [16] – all instances of the spoof "flushed" at English Wikipedia --Francis Schonken (talk) 10:45, 31 May 2018 (UTC)[reply]

In sum:

• This vandal act persisted on Wikidata for 21 minutes, in a single entry
• As a consequence of the short-lived vandalism of one Wikidata entry, this vandalism was multiplied at least 30 times on English Wikipedia articles, and it took around 200 times as long as the exposure period of the original vandalism to get rid of it in English Wikipedia.

Completely unacceptable MO, if you ask me. --Francis Schonken (talk) 10:45, 31 May 2018 (UTC)[reply]

•  Comment: This is basically technically avoidable effect. And I suppose in most cases, Lua is set in label-vandal-proof manner. The reason this happens is fetching "Label" as text in infobox. If fetching sitelink text, this doesn't happen. --Was a bee (talk) 08:14, 31 May 2018 (UTC)[reply]
  • @Was a bee: thanks for the suggestion on how to address this, could you please proceed with a WP:SOFIXIT? The affected infoboxes are afaik {{Infobox World Heritage Site}} and {{Infobox pyramid}}. Ask Mike Peel, who afaik set up both of these boxes with Wikidata functionality, for assistance if needed (I'm not an expert in Lua programming). Thanks. --Francis Schonken (talk) 08:42, 31 May 2018 (UTC)[reply]
    • @Francis Schonken: These seem to come from {{Wikidata location}}. So I posted idea feedback at there (Template_talk:Wikidata_location#Label_text_or_sitelink_text). As additional information, this type of phenomena (one edit in WD affects hundreds of pages in WP) happens basically only in certain properties of wikibase-item (one of 15 datatypes). Among those, I think country (P17) would be the most widely used case of such property. As you can see, this section (#Trocolandia about Germany) and above section (#Jajaxdelol about England) are both about country. In this sense, I think these are very good examples for thinking about this phenomena. I think there are various options about this issue, from simply "not-use-WD" to "replacing data once by WD", or some other technical ways.--Was a bee (talk) 15:57, 2 June 2018 (UTC)[reply]

What triggers updating the en.wikipedia cache to refresh data from WikiData - I presume that a WikiData edit does not trigger the cache-refresh on en.wikipedia for pages using that data. --Dirk Beetstra ^{T C} 08:56, 31 May 2018 (UTC)[reply]

I assume this works exactly in the same way as templates - refreshing a template does not trigger the page-refresh of the pages using the template; one needs to refresh them one by one.--Ymblanter (talk) 09:03, 31 May 2018 (UTC)[reply]

Changing a template does trigger the refreshing of pages (see here). Galobtter (pingó mió) 09:07, 31 May 2018 (UTC)[reply]

Strange. This is definitely not my experience.--Ymblanter (talk) 09:10, 31 May 2018 (UTC)[reply]

OK, what happens on en.wikipedia is that when a page is edited, it gets refreshed on the spot in the cache. If a template gets edited, ALL pages that are using said template get into the 'refresh queue', and their display will be refreshed in due time. My question is basically: if I edit a WikiData item, do all Wikipedia-pages that are depending on that item being added to the 'refresh queue', or does that only happen when either the page that is transcluding the data is refreshed, or when the template it is transcluding does? --Dirk Beetstra ^{T C} 09:32, 31 May 2018 (UTC)[reply]

(ec)It's not instantaneous, but it certainly is a lot quicker than when it comes down from Wikidata. See e.g. this and the previous change to a local template, which show up in all (12) articles already. This one as well is immediately reflected. Fram (talk) 09:35, 31 May 2018 (UTC)[reply]

On templates with a small number of transclusions it goes rather fast, on templates with thousands and thousands of transclusions it does take considerable time. --Dirk Beetstra ^{T C} 10:40, 31 May 2018 (UTC)[reply]

• “Trocolandia” does no longer appear any longer in any page, after I performed null-edits on all of the remaining ones from the search above. I also openend a phabricator ticket T196057 for the issue, in case this is not yet known. —MisterSynergy (talk) 10:26, 31 May 2018 (UTC)[reply]
  • Not on Wikipedia, but Google still happily finds pages with Trocolandia ... Now we have to wait for the Google cache to refresh. --Dirk Beetstra ^{T C} 10:51, 31 May 2018 (UTC)[reply]
  • For me, Google still finds Marmorpalais and GEO600 in Trocolandia. —Dirk Beetstra ^{T C} 03:24, 5 June 2018 (UTC)[reply]
    • Google says to me that Trocolanda is at Marmorpalais, Upper Harz Water Regale, File:Bautzen-nach1620-Merian.jpg, Commons-File:Aufgeschütteter_Fehlboden.jpg, and at a bunch of addresses reusing Wikipedia content. Rather significantly, I found that 12 days after the vandalism was fixed, Wikidata itself still displays the Trocolanda-vandalism at Q240190, Q882802, Q2348799, and Q1874400! This particularly concerned me, that Wikidata might continue to feed Trocolanda to Wikipedia via these items. I tested it, and I'm thankful to report that it doesn't. When a page is linked to one of those contaminated items Wikidata bypasses the garbage item-value and imports it as "Germany". Alsee (talk) 23:43, 9 June 2018 (UTC)[reply]

Wikidata is a useful and expanding resource which has the potential to be a useful tool to find information to create and improve articles on Wikipedia. I just don't believe that including information automatically in infoboxes is the best way to make use of this resource. However, in Wikipedia, we have various policies and guidelines regarding the use of bots and semi-automated tools which we can use to make edits. I would like to propose a mechanism using similar tools and processes that could be used to utilise more information from Wikidata, while meeting Wikipedia's policies and guidelines.

Page stalkers of RexxS might have seen a suggestion I posted on his talk page a while ago. At the time I thought it might be a way of monitoring what has changed on Wikidata. I now believe that it should be re-purposed to make suggestions to Wikipedia editors (using article talk pages) about useful information that could be used in the article. I've amended the proposal slightly.

So, this suggestion uses two bots (one working on Wikidata, and the other on Wikipedia) although the two might possibly be combined into one:

1. The Wikidata bot (WD BOT) regularly monitors wikidata items used in a Wikipedia infobox template, for example, the {{Infobox person}} template.

A list of template parameters and their matching wikidata properties needs to be generated when setting up the bot. Also, a list of wikidata items that need to be monitored needs to be generated. This could be generated from the What Links Here page, or possibly be adding a property to the Wikidata item indicating which infobox is used in an article.

1. If any of the watched properties of the item have changed since the time of the previous check, WD BOT generates a report detailing the changes in a log file somewhere on Wikidata.
2. The Wikipedia bot (WP BOT) regularly checks the log file on Wikidata.
3. If the log file has been updated with changes since the last check, WP BOT writes a report of the changes on the talk page of the appropriate Wikipedia page. The report will also include details of the reference information on Wikidata, preferably in a cite template such as {{Cite Book}}.
4. Any Wikipedia editor who has the page on their watchlist will see that the talk page has been updated (provided they haven't hidden bot edits).
5. The Wikipedia editor can then check whether the changes to the infobox are appropriate and referenced and if the information is from a reliable source. They can then make any appropriate changes to the article; adding to infobox, adding text to the article to show the new/changed data in the article, adding a citation to the text using an existing source, or creating a new reference using the information in the WP BOT report.

Any changes are recorded in the page history of the article, so appear on its watchlist and can be easily reverted/challenged/discussed.

I've not said anything on the frequency of the bot operations. RexxS pointed out that the {{Infobox person}} has roughly 270,000 transclusions, and with 20+ parameters would require roughly 6,000,000 properties to be checked. Bearing in mind that most information in infoboxes is static and that Wikipedia is a work in progress, the bot work could be split into manageable chunks, for example, 40,000 checked each day completing a full check each week, or 10,000 checked each day completing a full cycle each month.

If such a process can be made to work then it may be possible to increase the level of automation, and also to feed back on reference information to Wikidata from the sources used in Wikipedia. Robevans123 (talk) 15:06, 27 April 2018 (UTC)[reply]

I suggested this at the preparatory phase of this RfC. There was zero interest in developing this proposal.--Ymblanter (talk) 15:17, 27 April 2018 (UTC)[reply]

@Ymblanter: Thanks - I don't think I was watching the talk page for this Rfc back then, but I've just had a dig in the archives and seen your proposal. Great minds and all that... Well, there's some interest now. We'll see if anyone else bites! Robevans123 (talk) 15:47, 27 April 2018 (UTC)[reply]

Sure, and if the material is wrong since the beginning, we will still never notice. And this still requires bots on two sides to monitor. I have this EXACT system running on en.wikipedia alone - it has shown impossible to maintain, and impossible to get correct values in it due to the humongous amount of work needed to sift through tens of thousand of datapoints, and it is because of that utterly unreliable. When these bots start flagging thousands datapoints of which hundreds are actually correct volunteers will quickly start ignoring as the work is simply too much to finish. When this RfC sits out and closes, I will propose a completely different solution to that problem, likely obsoleting any bots. But if Wikipedia is going to get any credibility in the future, we’ll have to take some drastic steps (which some wikiprojects already took). Unfortunately, WikiData missed that boat, and will now face the problem on their own. —Dirk Beetstra ^{T C} 19:06, 27 April 2018 (UTC)[reply]

Hi all, we have written a grant proposal called GlobalFactSync. In core, it is quite similar to the third way described here. Coming from DBpedia, the backend side is our stronger suit as we have been monitoring and extracting data from Wikipedias for 10 years. We also have a mapping from Infobox properties to Wikidata properties, done by volunteers (we are trying to merge and improve it in the course of the grant project). What I noticed in the discussion above is that doing a maintenance via Bot seems to have some drawbacks, i.e. the bot needs to monitor both changes on Wikipedia and Wikidata day and night. We have implemented such a system called DBpedia Live which process all edits on several Wikipedias (It broke three weeks ago, since there was a switch to Kafka for updates and it still usesd the very old OAI-PMH). The English WP has 130k mainspace edits per day on average. While this can be done, it is a lot of work upfront. Hence, we were suggesting a User Script/Gadget, that does it on a request basis, i.e. each time an editor uses the gadget to get suggestions and compare enwiki to wikidata and other languages, the data is accumulated.

We were also discussing adding info on discussion pages. It can be done, but there is an extra overhead in keeping the discussion page in good shape, i.e. once the infobox is updated the bot needs to remove or update the flag from the talk page, so it causes an extra edit for each edit. We didn't propose this in the grant proposal, as it needs a greater consensus in the WP communities than a user script / gadget. However, if such a great consensus could be reached, we could provide API and backend for the information necessary and help with the bot. Also collaborate on the template parameter to Wikidata mapping. We can also provide the list of infobox template used for each Wikidata item as mentioned above for all Wikipedias. That is already available.

Anyhow, the main idea here is that automation does not mean taking the control from the editor, but automate the way information change is displayed and accessible to get a better handle on the data. SebastianHellmann (talk) 12:11, 3 May 2018 (UTC)[reply]

In response to Beetstra's contention above that data is sourced from Wikidata, it's important to clear up some misconceptions. First, the data that arrives in infoboxes on Wikipedia is not transcluded from Wikidata. Transclusion is defined at Transclusion as "the inclusion of the content of a document into another document by reference". There is no document on Wikidata that content can be transcluded from, nor is the data that arrives passed by reference. The action of importing the data takes values that are individually filtered to meet the needs of Wikipedia. Each value that is imported will have an associated reference that can be checked, just as any fact in any infobox should have. Further, it is nonsensical to call data "reliable" or "unreliable", as those are properties of the sources that verify the data. The reliability of the source cited on Wikidata is as easily ascertained as a source found on Wikipedia. In Wikipedia, which is not considered a reliable source itself, we are able to take sourced content (along with its references) and add it to another article. We do not reject the content as "unreliable", nor do we reject the sources because they were cited on Wikipedia. This is because Wikipedia is an unstructured repository of content and references. Wikidata is a structured repository of content and references. While Wikidata has policies less rigorous than Wikipedia's, it would indeed not be appropriate to indiscriminately import content from Wikidata. But that is not what happens in our Wikidata infoboxes: they discriminate between content that is sourced and that which is not; they allow the editor at the article level to check that source; and they allow that editor to reject, replace, or amend the content of the infobox. The sourcing requirements for content on Wikidata are not relevant to the question "Can Wikidata infoboxes be used in mainspace?" because our infoboxes only see the fraction of Wikidata content that has an associated source. If someone feels it necessary to verify the content against its source and the source's reliability, they do that in exactly the same way as they would verify content on Wikipedia, that is, by checking source and the source's reliability. Content and its sourcing that is found on Wikidata is not somehow mystically "tainted" by Wikidata's policies or data model. That content and its source is independent of the repository where it has been collected. --RexxS (talk) 00:23, 21 May 2018 (UTC)[reply]

• I'm sorry, but I disagree strongly with almost everything that you write above. I'll mention just one detail. Checking the original sources of stuff that gets into WP via WD is absolutely not the same as checking sources on WP. If I want to check the source of something imported from WD (and I don't care how you call that), I first have to go to WD, then find the thing I want to check, then find it's source, then check that source. On WP, all I need do is go to the reference, then go to the source. In addition, if something on WP is changed, I see that in my watchlist. If, say, an official website is "hijacked" by some spammer on WD, I won't see that. Etc etc etc. --Randykitty (talk) 02:42, 21 May 2018 (UTC)[reply]

  You can import data together with the source (and actually block data which have no source or a bad source from being imported).--Ymblanter (talk) 07:29, 21 May 2018 (UTC)[reply]

  @Ymblanter: You keep on saying (and others do the same): "You can import data together with the source (and actually block data which have ... a bad source from being imported)". Can you show me a real example of data that is being transcluded (please use the right term, this is transcluded, not imported) from WikiData where a 'bad source' is filtered. Does 'bad source' equate 'from Wikipedia'? Because we have more 'bad sources' than Wikipedia itself. And how does said template that is to transclude data from WikiData decide that the data actually NEEDS a source (see WP:BLUE in the specific instance of transclusion? --Dirk Beetstra ^{T C} 10:55, 21 May 2018 (UTC)[reply]

  I am not a right person to answer this. I once inquired whether this is possible, and was told it is possible. We need a more tech-oriented person to explain how this is possible.--Ymblanter (talk) 10:59, 21 May 2018 (UTC)[reply]

  I know exactly how what can be done: I can filter data on BLP's by having a reference, or not. If it has a reference, I can see whether that reference is from some typical unreliable sources (say, imdb), and exclude those. After that, it becomes impossible, as it becomes judgement. There is, as we already discussed on WP:RS, no way more accurate than artificial intelligence that can decide whether a certain source is a reliable source for the fact. For some data it is a case-by-case judgement. That means that every transclusion (at edit or when added later) needs to be manually checked whether the transcluded fact is reliably sourced. The only way that I see around this would be a change in data model. But at this time, plain filtering by domain (which is limited and heavy) is the only LUA can do. --Dirk Beetstra ^{T C} 11:41, 21 May 2018 (UTC)[reply]

  No, this is incorrect. One can make a list of sources which were used by bots to source the statements in Wikidata (which is not so long) and to see in advance which sources are acceptable and which are not. This is the only manual work which can be done (plus possible vandalism on Wikidata as I explained above).--Ymblanter (talk) 11:45, 21 May 2018 (UTC)[reply]

  @Randykitty: Checking the original sources of stuff that gets into WP via WD absolutely is the same as checking sources on WP.

  If you want to check the source of something imported into an infobox from WD, you click the pen-icon link and look at the reference; if it's online you follow the hyperlink, otherwise you find the print version.

  If you want to check the source of something manually added to an infobox directly on Wikipedia, you first have to find the text in the article that has the same information, then click superscripted link and look at the reference; if it's online you follow the hyperlink, otherwise you find the print version.

  The only difference is that the fact imported from WD to the infobox is guaranteed to have a reference by default; the fact added to an infobox directly on Wikipedia has no such guarantee.

  {tq|"if something on WP is changed, I see that in my watchlist."}} Yes, and if the same thing is changed on WD, I see that on my watchlist as well. You're imagining problems that don't exist. --RexxS (talk) 14:17, 21 May 2018 (UTC)[reply]

  Yes, I can see WD edits in my enWP watchlist, I know that. Originally, I switched that feature on, only to have my watchlist flooded with inconsequential changes (adding of a label in Swahili, all kinds of irrelevant database identifiers, and whatnot). So I switched that off and won't switch it on again. So, no, I won't see any WD changes. --Randykitty (talk) 14:44, 21 May 2018 (UTC)[reply]

  So let's be accurate. You choose not to see WD changes, but you choose to see WP changes. And that is your basis for rejecting WD and accepting WP. The only difference between the two cases is that you choose not to look at changes occurring in one of them. I expect the closer will give that reasoning the weight it deserves. --RexxS (talk) 15:16, 21 May 2018 (UTC)[reply]

  Yes, that's what I wrote, isn't it? The way WD works, if I add it to my watchlist it's flooded with irrelevant stuff, making it difficult or almost impossible to filter out the stuff that's important. If that means that I now have to accept unreliable stuff that is prone to vandalism in WP articles, so be it. --Randykitty (talk) 15:24, 21 May 2018 (UTC)[reply]

@RexxS: I am sorry, I am wording this stronger than Randykitty: This is utter, pertinent bullshit.

The material is on WikiData, and a piece of code is making it appear in en.wikipedia. That is exactly the same as how templates work, a mechanism that we call transclusion. We ‘refer’ to the data we want to see, and the code makes it appear. It is not imported, it stays on WikiData. En.wikipedia does NOT have control over what is on WikiData,

Regarding the filtering. Yes, we have LUA code that filters data to define whether data has a reference. That code cane, to a certain extend filter also the type of reference (exclude ‘from Wikipedia’). It is impossible for the LUAcode to filter whether the attached reference is a correct source (does it actually state the statement), whether the value has been altered without touching the reference, does the reference actually constitute a reliable source (imdb for a BLP, though I could write code to exclude imdb, I cannot filter ALL unreliable sources known to mankind in our LUA code, especially knowing that often context matters). No, RexxS, you cannot assure you hat all material on WikiData that IS referenced is suitable for transclusion. And I know you are now going to suggest that we should then by hand ovverride all data that we can’t use, which just negates the ease of use, and makes infoboxes more confusing to editors: transclude all from Wikidata, hiding where data is, but not this, that and the other and for such use this local value.

As I wrote elsewhere, it stretches my AGF that every editor who is transcluding an infobox NOW is going to check whether all data is correct and whether all data is referenced according to our referencing standards. Especially for newbies. Moreover, I can NOW add an wikidata infobox with field ‘day of birth’ empty on both wikipedia and wikidata. Then someone can come next week to WikiData and add a birthday referenced to an unreliable source (but stating the source). That results that THEN en.wikipedia is starting to transclude that data. Am I NOW responsible for enabling the transclusion of BLP sensitive data with a bad reference that is going to be added in the future. Do I HAVE to return to check when WikiData changes? If that happens multiple times do I get blocked because I am not adhering to the BLP policy. Or are you going to block any editor on WikiData if that editor repeatedly adds BLP sensitive data with references that fail en.wikipedia reference standards? I am waiting for our banned editors to use WikiData as a proxy.

Going on on references, there is WP:BLUE material on WikiData .. that we don’t transclude because it not referenced. So more local settings: only referenced data from WikiData, but this is finebecause it does not need a reference.

The whole ‘you can filter by references’ argument is a red herring, a carrot. There is not enough control (until we teach our LUA infoboxes artificial intelligence).

Yes, data is tainred because it is on WikiData: it is an unreliable source for our data, not following our referencing standards, we do not have a reliable system for filtering data that we want or not. Until WikiData is changing its datamodel it is not suitable foruse on wikipedia.

@RexxS: as to your definition of Transclusion: In computer science, transclusion is the inclusion of part or all of an electronic document into one or more other documents by hypertext reference. Transclusion is usually performed when the referencing document is displayed, and is normally automatic and transparent to the end user.[1] The result of transclusion is a single integrated document made of parts assembled on the fly from separate sources, possibly stored on different computers in disparate places - the data that we display in a fully wikidata enabled infobox arrives in the article by transclusion of said infobox, where the data displayed is transcluded WikiData data. That means that if the template changes, or the data on WikiData changes, the displayed article is changing.

In the end it does not matter - the editor who is importing data from WikiData, or the editor who is adding code resulting in the transcludion of data from WikiData has sourced information from an unreliable source, and is responsible for what is displayed in the article (and, IMHO, also for data that is added later to WikiData if that is resulting in transclusion onto the article that they added the automatic transclusion to). —Dirk Beetstra ^{T C} 04:15, 21 May 2018 (UTC)[reply]

@Beetstra: I'm also sorry, but the only bullshit is your fake claims of problems that don't exist.

"The material is on WikiData, and a piece of code is making it appear in en.wikipedia. That is exactly the same as how templates work. Nonsense. Templates dumbly transclude all of the marked content from a document. Data from Wikidata database is examined and filtered by the code that imports its values. We can control what is imported and reject unsuitable data or choose to replace it with a local value as we see fit. We can't do that with transclusion. The control is just a strong as the control we have over locally added data.

You complain that "It is impossible for the LUAcode to filter whether the attached reference is a correct source ... whether the value has been altered without touching the reference, does the reference actually constitute a reliable source ... especially knowing that often context matters. Yet you wilfully ignore the fact that you can't know that a reference manually added to infoboxes (in the few cases they exist) is a "correct source", or whether the value in the infobox has been altered without touching the reference, or that the reference actually constitutes a reliable source. It's just the same set of problems whether the infobox content and reference comes from Wikidata or from somewhere else in the article (or from an editor's OR). And the solution remains the same: you check the fact against the reference that supports it, a job made easier in practice by using content from Wikidata. But you continually choose to ignore that fact.

I do not assume that "all material on WikiData that IS referenced is suitable for transclusion". It's not transcluded. It's filtered to remove the most egregious stuff, which is more than can be said of material added locally. What is left may or may not meet our sourcing requirements, but it still stands a better chance than what can be added locally by any editor, which completely unfettered.

Our infoboxes are designed by default so that fetching Wikidata has to be enabled on each article. Why should we not ask editors who add such an infobox to be responsible for checking that the data they are enabling comes from a reliable source? We ask the same of editors who add a non-Wikidata infobox (in theory). It stretches my AGF that you believe that happens every time.

I can add a non-Wikidata infobox with the with field 'day of birth' empty. If someone comes next week to Wikipedia and adds a birthday referenced to an unreliable source (but stating the source), are you going to start complaining about me adding the infobox that allowed the editor to add unreliably sourced information? Because that's exactly analogous with what you're demanding of me if I add a Wikidata-enabled one. In every case, it's the same whether we grab the sourced content from Wikidata or from a editor locally. You might as well claim that "local editors re not a reliable source", because it's just as nonsensical as your fake claim about Wikidata being a source.

Your admission that you think Wikidata taints a source is telling. If the same content and same source were found on es-wiki, that would be fine with you. But the moment that content and source is fetched from Wikidata, you're up in arms. Your argument is nothing more than IDONTLIKEIT.

It's not my definition of transclusion. It's Wikimedia's. If you're going to use the word, use it properly. The infobox template is transcluded into an article, of course, but the content of each field does not come from that template document. The date of death of George Harris (theologian) is no more "transcluded" from Template:Infobox person/Wikidata than the date of death of George Harrison is "transcluded" from Template:Infobox musical artist. While we're looking at those two, tell me how long it takes you to verify the fact in each case. The Wikidata one is verified in a few seconds; a much longer time is needed to verify the non-Wikidata one. Let anybody who has an open mind do the same. They will soon see whether your fears are real. --RexxS (talk) 15:11, 21 May 2018 (UTC)[reply]

We are again running in circles, RexxS. At least for the third time. To me, WikiData is at the very best the same as en.wikipedia, and with what I see going on, I don’t think it is. —Dirk Beetstra ^{T C} 16:11, 21 May 2018 (UTC)[reply]

You know, Dirk, I can agree with your assessment that "WikiData is at the very best the same as en.wikipedia" – at least when we're thinking about sourcing. I do not now (nor ever will, I suspect) claim that Wikidata has sourcing any better than English Wikipedia's. For me, the value of making use of Wikidata is what it brings to the smaller Wikipedias, and the convenience of having a single place to update facts as they change. Where we disagree, I guess, is that I think there's enough of Wikidata that has decent sourcing to make the effort of fetching infobox data worthwhile; and you don't. I'm prepared to leave it at that. Feel free to have the last word. Cheers --RexxS (talk) 21:12, 23 May 2018 (UTC)[reply]

@RexxS: and I fully agree with that. There is value in WikiData. The system is great, ir allows for significantly increasing reliability of what we do here. But, in my opinion, not in the current state/data model. I currently see two, heavy pro-WikiData editors heavily bickering (almost an edit war) about transcluding or not transcluding a WikiData infobox because there is badly sourced datafrom WikiData (or not). Both WikiData and infoboxes have been in front of ArbCom (both sent back).

If this RfC closes as 1A, I hope that WikiData starts thinking. Very much unlike any other Wiki, WikiData has the potential to be a reliable source. In my opinion, a true wiki-way does NOT make sense on WikiData (and I have been saying similar a couple of times). It is late, but not too late, and it will take a humongous effort. —Dirk Beetstra ^{T C} 03:30, 24 May 2018 (UTC)[reply]

• Guys. It really doesn't matter whether we call the process by which the Wikidata data arrives here "transclusion", "importation", or "Abra Kadabra Presto!". Look at the conditions and outcomes, and make your arguments for or against those rather than quibbling over terminology. Nikkimaria (talk) 13:37, 21 May 2018 (UTC)[reply]

Exactly, Nikkimaria. --Dirk Beetstra ^{T C} 14:32, 21 May 2018 (UTC)[reply]

I haven't read this whole thread, but in case it hasn't been mentioned before, there needs to be better automatic notification of when a Wikidata edit affects a Wikipedia article. E.g. if you edit something on Wikidata, it should show up as an edit/change to any affected Wikipedia pages. Implementing this will be a significant technical challenge however. SharkD  Talk  08:33, 21 May 2018 (UTC)[reply]

That feature has been available for some time. If you look at a standard watchlist, there's a line prefixed "Hide:"; if you untick the "Wikidata" box, you'll see changes to Wikidata items associated with the pages on your watchlist. However, up until recently, the only way to get information from a Wikidata item into a Wikipedia infobox was to read the entire item and pick out the property you want. That meant that Wikidata could not determine which property is used on Wikipedia, so every change to the Wikidata item showed up in watchlists, which was too much for many editors, and they switched it off. I have an updated version of the Lua module:WikidataIB waiting in its sandbox which utilises newer developments that allow Wikipedia to read just the properties required for the infobox, not the whole Wikidata item. That means the Wikidata part of the watchlist can now be filtered selectively to only display changes that will affect the infobox. That is still being refined (for example, we want to see changes to the English label, but not the Portuguese label, etc.). I'm loathe to roll-out a major change to code while the RfC is still ongoing, but I assure you that the Wikidata developers are keen to implement precise watchlist notifications, and I expect a big reduction in false positives soon. --RexxS (talk) 10:38, 24 May 2018 (UTC)[reply]

A quick summary, 1A vs non-1A is 27/25. User:Capankajsmilyo^{(Talk | Infobox assistance)} 04:03, 17 April 2018 (UTC)[reply]

But the vote isn't 1A vs non-1A. Please don't summarize 6 options into "1 vs. all" (using your method gives "1F vs non-1F is 10/42", any reason you don't add that result?). Either give a votecount for all options (plus the "others"), or don't bother as it isn't a vote anyway. Fram (talk) 11:32, 17 April 2018 (UTC)[reply]

If you're interested in the statistics, see this spreadsheet. By my count, 1A vs. non-1A is currently 22.5 to 26.5, but please let me know if I've miscalculated (you can see the counting on the second sheet). And as I said on the talk page when I first posted that sheet, "Of course, that's the crudest possible measure of this RfC, and doesn't take into account people's comments, but it might be a useful first-look." Thanks. Mike Peel (talk) 13:25, 17 April 2018 (UTC)[reply]

Something that will have such a profound effect on Wikipedia should be more widely advertised. 188.28.163.10 (talk) 15:16, 17 April 2018 (UTC)[reply]

@anon any suggestions of additional places to advertise it? Thanks. Mike Peel (talk) 22:10, 17 April 2018 (UTC)[reply]

@Mike Peel: Capankajsmilyo brought this up just now in Wikidata's Project chat. mahir256 (talk) 05:04, 18 April 2018 (UTC)[reply]

WP:CANVASS anyone? Are we now going to inject WikiData editors here to make a decision on how Wikipedia is going to use WikiData? --Dirk Beetstra ^{T C} 08:05, 18 April 2018 (UTC)[reply]

It should be not difficult to figure out if we have votes of editors normally not active on the English Wikipedia.--Ymblanter (talk) 08:12, 18 April 2018 (UTC)[reply]

Now, since this is a quite important feature and situation that we are talking about, should this RfC not be on a site-notice or similar? --Dirk Beetstra ^{T C} 08:23, 18 April 2018 (UTC)[reply]

Agree with Beetstra on this. User:Capankajsmilyo^{(Talk | Infobox assistance)} 08:33, 18 April 2018 (UTC)[reply]

I mean if individual RfAs are on a watchlist notice, seems like this should have a watchlist notice... Galobtter (pingó mió) 08:46, 18 April 2018 (UTC)[reply]

An interesting this to see is that Portal discussion has got more than 500 votes and much more comments. This discussion seem to have accumulated 50-60 till now. Some of the locations/strategies used there for advertisement can be used here as well. User:Capankajsmilyo^{(Talk | Infobox assistance)} 06:01, 24 April 2018 (UTC) Another thing that can be done is including the link to discussion in {{Infobox}} so that it can be seen on every page / template using infobox. User:Capankajsmilyo^{(Talk | Infobox assistance)} 06:03, 24 April 2018 (UTC)[reply]

I got a signpost today regarding portal RFC, but this RFC was not mentioned there. I think, this should have been in there. User:Capankajsmilyo^{(Talk | Infobox assistance)} 03:15, 26 April 2018 (UTC)[reply]

Yup, people only seem to care when something is up for deletion, not when say something could affect ~3 million pages not in a deletion related way. Any way of a watchlist notice? The only reason it has 400+ !votes is because it had a notice on ~1500 portal pages. Galobtter (pingó mió) 06:11, 26 April 2018 (UTC)[reply]

The large intro in probably is dissuading many people too and the generally boringness of it, I suppose too. Like the bike-shed effect. Galobtter (pingó mió) 06:17, 26 April 2018 (UTC)[reply]

• Threaded discussion moved down from voting section:

  I only see four users below who did not vote 1A; these four users have 8K, 1K, 243K, and 46K edits on the English Wikipedia, respectively.--Ymblanter (talk) 11:06, 25 April 2018 (UTC)[reply]

  Since you want more detail, of the four (current) such !votes: The first has 832k Wikidata edits vs 8k EnWiki edits (99%), and was in the middle of a simultaneous Wikidata-run while they !voted. The second has 74k Wikidata edits vs 1k EnWiki edits (98.4%), and had been active on Wikidata more recently than EnWiki before !voting. The third is not in question, they were locally aware of the RFC during drafting. The fourth has a significant percentage of activity on both sites, and could credibly have arrived to !vote here via local advertizement or via Wikidata canvass. The origin of future !votes is an open question. Alsee (talk) 11:43, 25 April 2018 (UTC)[reply]

  Right, but you yourself only have 6K votes here. I would prefer you to not judge users based on solely bad faith assumptions, in particular if these users have contributed to the English Wikipedia (and, for that purpose, to Wikidata) more than you did.--Ymblanter (talk) 11:58, 25 April 2018 (UTC)[reply]

  Ymblanter it is unhelpful to gratuitously attack my edit count. It is unhelpful to incorrectly assume that I "judge[d] users based on solely bad faith assumptions". I did not judge the users at all. The sole issue is how the !votes arrived on this page. Biased participation can manufacture a fictional-majority in any direction on any issue. It is perfectly appropriate for me to notify the closer that some !votes may be present due to de facto notification of users with a predetermined position. It is then up to the closer to determine how to handle that concern. BTW I invite you (or anyone else) to move my replies and yours down to the discussion section, starting from your 11:06 25 April comment. We don't need to clutter the !voting section with this dialog. Alsee (talk) 12:56, 25 April 2018 (UTC)[reply]

  I don't see why this needs to be in the !vote section at all, so I've moved the whole lot down. Note that the mention on Wikidata's project chat was not posted as an invitation to participate, but as a FYI/start of discussions there about some of the concerns raised here. Thanks. Mike Peel (talk) 14:20, 25 April 2018 (UTC)[reply]

  First, intent is irrelevant as far as the RFC itself goes. The post at Wikidata had a de facto effect of canvassing, and it is reasonable to alert the closer to the concern. As an independent matter, the sequence and timing of events makes your interpretation of intent rather implausible. The user has previously been warned for canvassing. They !voted on April 7. Ten days later they posted a vote-count that the RFC was going against them. In a matter of hours this sequence played out: IP replied to the vote-count by suggesting more advertizing, someone asking where to advertize it, then then the user posted at Wikidata with the section title Enwiki RFC raises concerns and a link for Wikidata users to click over to here, then they "agree"d with wanting more advertizing. It strains credulity to suggest that they were utterly unaware that their actions were going to have the effect of a WP:Votestacking advertizement in their preferred direction, when they were in the middle of noting that the RFC was going against them and discussing their desire to advertize. Alsee (talk) 01:03, 27 April 2018 (UTC)[reply]

Why is this posted on my talk page by you? Are you an administrator? Cos if you are not, I would like to know what's happening. Further, what is it I'm being accused of? You people raised concerns about Wikidata verifiability, and you want me to not even tell the people there that there policies are flawed (in the views of editing community here) ? Capankajsmilyo^{(Talk | Infobox assistance)} 01:44, 27 April 2018 (UTC)[reply]

Discussed on usertalk. Alsee (talk) 02:02, 27 April 2018 (UTC)[reply]

How can you assume that this RFC is going against my direction? As far as I can see there's 50% votes on both sides. From the limited experience of RFCs of mine, this leads to a state of no consensus even if 5-10% up/down. And how is your opinion to not advertise on English Wikipedia via signpost and/or infobox in good faith? Capankajsmilyo^{(Talk | Infobox assistance)} 02:19, 27 April 2018 (UTC)[reply]

How the RFC is going is irrelevant to whether or not something is canvassing. However your posting of your votecount here shortly before posting at Wikidata:Project_chat suggests that you were concerned with the votecount when you posted at Wikidata:Project_chat. Alsee (talk) 02:27, 27 April 2018 (UTC)[reply]

Is it an infobox which pulls all fields from Wikidata (option i) or some fields from Wikidata (option ii)? My understanding was that we are discussing option ii, and if 1A passes that outlaws any infoboxes pulling anything from Wikidata (for example, fields like Commons Category or Official Website - we have plenty of those). This is not the choice I support, but I only have one vote, fine. I see however that some people oppose use of infoboxes clearly under understanding that we are talking about option (i) - all or nothing. This needs to be clarified, otherwise it makes the whole RfC even more confusing.--Ymblanter (talk) 05:24, 19 April 2018 (UTC)[reply]

@Ymblanter: This is about infoboxes pulling some, or all, fields from Wikidata (not your option i or ii, but both). This is about whether infoboxes should pull any data from WikiData. And Official Websites are just as vandalism prone (if not more) than others, as I see now spammers targeting WikiData first, and that those websites stay vandalized for months is not helping their cause. --Dirk Beetstra ^{T C} 07:40, 19 April 2018 (UTC)[reply]

The I believe this vote is invalid since it is based on a false premise.--Ymblanter (talk) 07:59, 19 April 2018 (UTC)[reply]

It would still be true if there is one local field and one magically appearing WikiData field (though the editor did not express that). And see the confusion when someone is locally removing a wrong field, and sees another, (or even the same) magically staying there. Or when fields are added they don't do anything because they are overridden by a WikiData value.

The case this editor describes, however, is explicitly mentioned in the lede of this RfC: '[[File:South Pole Telescope infobox from Wikidata.jpg|thumb|right|An example infobox built entirely from Wikidata information. This is used in the [[South Pole Telescope]] article and if you look at the edit window, all you will see is {{tl|Infobox telescope}}.]]'. So this is about option 'í' (what this editor apparently !voted on), and about your 'option ii' (which encapsulates option 'i' as an extreme case). --Dirk Beetstra ^{T C} 08:18, 19 April 2018 (UTC)[reply]

I believe that your argument is incorrect, but I have no intention of entering into a long argument (which in the past never lead to anything good), so that I will leave this for a closing admin to decide on.--Ymblanter (talk) 08:20, 19 April 2018 (UTC)[reply]

@Ymblanter: I would suggest that you take up this question with the caster of that !vote then to get it clarified. --Dirk Beetstra ^{T C} 08:22, 19 April 2018 (UTC)[reply]

I'm coming to this discussion late and currently have no time to go through all the different proposals and !votes. So I'll throw out my idea here, apologies if somebody else has already brought up something similar. I share the concern of those people who feel that WD is much more vandalism-prone than WP. As far as I see, the templates most involved are authority control, infoboxes, and the {{official website}} template. Would it be possible to have WD pre-fill these templates on demand when an editor adds them to an article? The editor would then have to verify the data and be responsible for including them into the article. Once that is done, subsequent editors could have a button that would cross-check info in the article with the WD item and, again, editors would have the choice of including those data or not. In no case would WD (or a bot) add data directly from WD into WP. Just a thought, thanks for listening... --Randykitty (talk) 13:13, 11 May 2018 (UTC)[reply]

• See Template:Infobox UNESCO World Heritage Site#Pre-loading Wikidata values, which has sort of an implementation of that idea for a particular template. --Francis Schonken (talk) 13:21, 11 May 2018 (UTC)[reply]
• The problem is that such a setup only works in one direction: the content from Wikidata comes here, but then any changes to it aren't fed back to Wikidata. The system works best when it's bi-directional: we gain information from Wikidata in infoboxes here, and we share back new information into Wikidata where it's then used in other language Wikipedias and elsewhere. For comparison, it's like saying "let's keep all images used in enwp here, and have a script to copy new ones over from Commons", which would be a nightmare. Thanks. Mike Peel (talk) 13:43, 11 May 2018 (UTC)[reply]

• The problem with possible vandalism from Commons is minimal compared to the damage that WD can do. And I don't see why WD cannot regularly scan changes to an infobox and "leech" whatever changes were made. But that's their problem, not ours. --Randykitty (talk) 14:13, 11 May 2018 (UTC)[reply]

I still don’t see why we need to import from WD. They have downloaded data from external sources or manually added, and there is no quality control (as the example by user:Fram from the vandalism that Mike Peel copied over shows), nor sufficient vandalism control. Whether there is a reference to said source is no guarantee that the data has not been vandalised on WD since (forgetting that said source would be a reliable source for that data on en.wikipedia in the first place). It is much better to use the same external database or careful editors directly, and control the data here -our vandalism control outnumbers WD by at least an order of magnitude. No, user:Mike Peel, it is not ‘much better’, it is orders of magnitudes worse, and I would, at best, allow careful single pulling of data from WD at discretion of single editors (who can be ‘blamed’ for being not careful enough). The bidirectional route is, at the moment, a recipe for disaster. —Dirk Beetstra ^{T C} 18:29, 11 May 2018 (UTC)[reply]

The way changes to Wikidata show up in your watchlist on English Wikipedia has recently been improved (happened a couple of days ago). Previously if a Wikidata property was used within a Wikipedia page any change to any property in the entire item on Wikidata would create a line in your watchlist.

Now only if a change occur to the specific property used (or the aliases) will a line appear in your watch list. In my opinion this is a significant step forwards so congrats to the Wikidata team. I have requested the option not to show changes to "aliases". In my opinion this significantly improves quality assurance for material used from Wikidata and thus makes me more comfortable supporting Wikidata use within Wikipedia. Doc James (talk · contribs · email) 21:12, 13 May 2018 (UTC)[reply]

Well done to Ladsgroup and the rest of the folks who have been working on that. Following improvements to the Wikidata–Lua interface, I've been able to do an overhaul to Module:WikidataIB (in the sandbox while testing) that ensures it now reads only the property related to the infobox field, rather than the whole Wikidata entry, so that should also reduce the number of false positives in watchlists when Wikidata is changed. --RexxS (talk) 21:47, 13 May 2018 (UTC)[reply]

@Doc James: this helps in keeping an eye on the data, but this does not take away my other concerns. --Dirk Beetstra ^{T C} 05:17, 14 May 2018 (UTC)[reply]

Given that this RFC has been open for over a month, and that discussion seems to be dying down, is it time to request that an uninvolved admin review and close? Blueboar (talk) 12:10, 11 May 2018 (UTC)[reply]

The request has been sitting there for about a week. We clearly have a shortage of admins in a suicidal mood.--Ymblanter (talk) 12:14, 11 May 2018 (UTC)[reply]

Since 6 May [17]--Ymblanter (talk) 12:18, 11 May 2018 (UTC)[reply]

Any closer claiming a consensus would be expected to be able to defend that decision. I would not wish to be that person. A no consensus close might also be challenged, but then the onus would be on a challenger to show how this collection of diverse opinions could be construed as a consensus, and for what. If nobody will touch it we can eventually draw the conclusion that no-one is willing to claim a consensus. If no-one steps forward to close by the end of May, I suggest that we accept no consensus as the default. Cheers, · · · Peter (Southwood) ^(talk): 10:07, 12 May 2018 (UTC)[reply]

We can still post once directly at AN (the current announcement is in a transcluded template) and explicitly ask for three closers.--Ymblanter (talk) 10:10, 12 May 2018 (UTC)[reply]

NO. This has to be independently closed. I can agree with a closing committee, but not with a ‘by default’ close. —Dirk Beetstra ^{T C} 12:01, 12 May 2018 (UTC)[reply]

Really, Dirk? What sort of closure are you expecting?

• Q1: Can Wikidata infoboxes be used in mainspace? - only a minority voted "No".
• Q2: What can Wikidata infoboxes display? - an even smaller minority voted "Nothing".

Are you really going to insist that there is consensus for those two outcomes? Because it sure doesn't look like any consensus to me. If it lies open forever, it will just as clearly signal "no consensus" as would a default close. --RexxS (talk) 12:35, 12 May 2018 (UTC)[reply]

And that is exactly why I want this to be properly closed. ArbCom is ready for more fights about infoboxes and wikidata, and leaving this open makes that clear that they need to look at it. If there is no consensus then have that properly codified. —Dirk Beetstra ^{T C} 13:09, 12 May 2018 (UTC)[reply]

Assuming a “no consensus”... the real question that the closers will have to address is: what does that “no consensus” MEAN? Does it mean there is no consensus FOR using Wikidata in infoboxes (default to not using it) or does it mean there is no consensus AGAINST using Wikidata in infoboxes (default to allowing it) or does it mean no consensus AT ALL ... for OR against... (With no default position. In which case the debate remains unresolved, and will require additional RFCs). Blueboar (talk) 14:41, 12 May 2018 (UTC)[reply]

The status quo is defined by Wikipedia:Requests for comment/Wikidata Phase 2: "It is appropriate to modify existing infoboxes to permit Wikidata inclusion when there is no existing English Wikipedia data for a specific field in the infobox". A "no consensus" close will not alter that. --RexxS (talk) 14:54, 12 May 2018 (UTC)[reply]

Although it's probably fair to say that that particular RFC would have turned out differently with the experience that we have now, 5 years later... --Randykitty (talk) 15:18, 12 May 2018 (UTC)[reply]

To be fair, it wouldn't turn out differently. The majority of editors then, as now, were in favour of some Wikidata integration, carefully implemented. We've had an explosion of content on Wikidata leading to many of the observed problems, but we've also created the tools to make infoboxes far more flexible, putting the decision on whether to import Wikidata in the hands of the editors at the article level, and ensuring that unsourced data is filtered out by default. We also now have the experience of very successful opt-out infoboxes such as {{infobox gene}} and {{infobox telescope}}, as well as offering opt-in alternative versions of other templates by creating a "/Wikidata" version. Editors have all the control they need to ensure that they can choose in any given article, by local consensus, to allow or disallow Wikidata integration right down to the individual field level. That's a perfectly tenable position that is no different from the consensus-driven process of adding (or removing) any other content to an article, and the majority recognise that. --RexxS (talk) 18:26, 12 May 2018 (UTC)[reply]

Sigh, and that is why I think a committee should look at the results and not have a by default close. The whole sourcing question is a big issue, and if we do end up in the middle, we will need an RfC on that whole sourcing question - there is IMHO a whole discrepancy in the whole definition of sourcing with respect to WikiData that cannot be addressed within the constraints of this RfC (it is too late ... I should have thought of this earlier). —Dirk Beetstra ^{T C} 15:04, 13 May 2018 (UTC)[reply]

Why will we need an RfC? The sourcing of infobox information from Wikidata is subject to just the same requirements as the sourcing of infobox information locally on Wikipedia, except that in the Wikidata case, we can arrange to automatically exclude any data that is not sourced. The necessity of being able to verify the accuracy of the sourcing is identical in the two cases, except that in the Wikipedia case, you often have to search through the article text to find the reference to check. The downside of getting information from Wikidata is that the watchlisting is still crude, but it's improving steadily. What's the question you want to ask? Perhaps someone can answer it in a satisfactory manner for you, without needing another 30 days of inconclusive debate. --RexxS (talk) 17:42, 13 May 2018 (UTC)[reply]

No... The information from Wikidata ISN’T subject to the same strict sourcing requirements. That’s one of the primary reasons why so many editors oppose incorporating it. Blueboar (talk) 20:48, 13 May 2018 (UTC)[reply]

Yes. You misunderstand. All of the information that is used in Wikipedia is subject to the same strict sourcing requirements, regardless of whether it comes from Wikidata or from a local edit – by definition. And the same mechanism is available to verify the compliance of that sourcing: check the reference. If that misunderstanding is one of the primary reasons for opposition, it's a pretty clear indication that there's a need to inform folks better. --RexxS (talk) 21:15, 13 May 2018 (UTC)[reply]

Exactly, Blueboar, it IS NOT. That is why anything but 1A requires follow up RfCs. —Dirk Beetstra ^{T C} 02:36, 14 May 2018 (UTC)[reply]

Yes it IS. The Wikipedia community creates the sourcing requirements for its content, and nobody else. There is no need for follow up RfCs, especially when you can't even articulate what questions need to be answered. --RexxS (talk) 11:52, 14 May 2018 (UTC)[reply]

Yes, we have our sourcing requirements, RexxS, and WD fails that. And I have articulated that in my !vote, WikiData does not meet our sourcing requirements (with the whole sourcing on WikiData being a massive red herring). --Dirk Beetstra ^{T C} 14:12, 14 May 2018 (UTC)[reply]

No, you're wrong. The text brought in from Wikidata is filtered by default to ensure it has a reference, so Wikidata's internal policies are irrelevant. You might as well complain that we shouldn't include text from Encyclopedia Britannica 1911, because EB's sourcing requirements don't match ours. We ensure that content on Wikipedia meets WP:V by checking the reference. That is how we ensure our sourcing requirements are met. It is just as simple (and often simpler) to check that a fact in an infobox is supported by its reference when it comes from Wikidata as when it has been added by an editor locally. The only red herring is your insistence that the sourcing requirements on Wikidata have any effect on the sourcing requirements here on Wikipedia. --RexxS (talk) 14:45, 14 May 2018 (UTC)[reply]

And that is here the discrepancy in understanding is. Let’s first see what this RfC brings, shall we. I am waiting for the closing committee. —Dirk Beetstra ^{T C} 20:25, 13 May 2018 (UTC)[reply]

Once again, that is a red herring. Material can be referenced against an unreliable source, still it would be incorporated if not filtered out (especially if the data+ref is added after the fully enabled box is inserted and the editor decided to opt-in for all referenced data -you can’t opt-out on unsuitably referenced data when it is not there yet). Data can be altered without changing the reference (meaning that an editor could have checked his BLP data to be correct, only to see it vandalised later). Data however that is unreferenced and suitable to be transcluded is not, because it is unreferenced, even if that data does not need a reference (sky, color=blue). But it is all besides the point, WikiData is an unreliable source, it does not matter whether a datapoint is referenced (properly) or not, it is unreliable by nature. —Dirk Beetstra ^{T C} 16:21, 14 May 2018 (UTC)[reply]

Your argument is the crimson clupeid. Material added locally on Wikipedia can also be referenced against an unreliable source just as easily. The only difference is that checking for verifiability is somewhat easier in a Wikidata infobox because: (1) the link to check the reference is always present in a field added from Wikidata, unlike the multitude of unreferenced fields added locally; and (2) you are guaranteed that a reference actually exists when the information comes from Wikidata, unlike the locally added values which may not be referenced anywhere, and you have to search through the entire article to establish that. In the cases where a non-Wikidata infobox actually has references, it is just as simple, nay simpler, for the data to be changed without changing the reference – with exactly the same consequences for the editor who sees his contribution vandalised later. As for the case of data that is suitable for transclusion without filtering, it is a simple matter to turn off the filtering for fields like "image". You're making up a problem that doesn't exist. You can't manufacture a case that distinguishes Wikidata from local edits when exactly the same considerations apply to each. Your characterisation of Wikidata as "an unreliable source" is misguided. Wikidata is not now, and never has been a source, unreliable or otherwise. It's merely a container. You might as well claim that paper is an unreliable source because some books contain falsehoods. The information contained in Wikidata is no more reliable or unreliable than the data in the current infoboxes on Wikipedia - and that's not surprising because a lot of it originated there anyway. What is beside the point is this big fuss over a pretence that we don't have control of what we show our readers. We do, and that's a fact, no matter how inconvenient to your theories. --RexxS (talk) 16:49, 14 May 2018 (UTC)[reply]

I give up, you simply don’t get it. We’ll see each other in a next discussion after this RfC closes I guess. —Dirk Beetstra ^{T C} 17:35, 14 May 2018 (UTC)[reply]

I think you'll find you're the one not "getting it", but no doubt we will explore that another time. --RexxS (talk) 19:30, 14 May 2018 (UTC)[reply]

Why is it that Wikidata proposers see this 4-question, six-option-for-each RfC (which they designed) now as a "Wikidata vs. no Wikidata" RfC? If that was the question you wanted to ask, then you should have asked that question. To now claim that the option of the 6 you proposed which got the most support is the one that has been rejected is disingenious. If you don't ask a yes-no question, then don't interpret the results as a yes-no outcome. Fram (talk) 06:52, 14 May 2018 (UTC)[reply]

@Fram: There were a wide range of options to see where people were on the distribution, and that means that the options aren't exclusive/uncorrelated (someone that picks one answer is likely to be happy with an adjacent answer, possibly except when they're at one end of the scale). It's not a simple yes/no in the same way that it's not a simple pick-the-one-with-the-most support (plus it's a !vote), but given the distribution (a lot of people going for option A) it makes sense to also look at that vs. the other options (as per Nouill's comment above). Also, you yourself had input into the design of this RfC, as did many others, so you can't now claim it's just been designed by Wikidata proposers! Mike Peel (talk) 11:07, 14 May 2018 (UTC)[reply]

My input was to remove the most obvious bits of bias. I gave my misgivings about the current RfC (and my opinion about what kind of RfC should be had instead) before the start of the RfC. My input into the design of this RfC was not a stamp of approval for it. I note, from the final discussions before this went live, that Beetstra, Francis Schonken, and others had misgivings about it, and that e.g. Beetstra explicitly said that starting with a straightforward yes-no question would have been better. You didn't go with it, but want to twist the results into a straightforward yes-no anyway, because otherwise you have one option with a lot of supports, and 5 with each a lot less support, and that gives a different impression than your grouping now of 2-6 together. Fram (talk) 11:25, 14 May 2018 (UTC)[reply]

You had your chance to frame the question Can Wikidata infoboxes be used in mainspace?, Fram, so it's no good whining over the outcome now where only a minority answered that question in the negative. There's no way that a minority view can be a consensus, so accept it and move on. --RexxS (talk) 11:52, 14 May 2018 (UTC)[reply]

I had my chance to frame the question? Where? Anyway, I'm not whining over the outcome, I'm not the one here repeatedly claiming how the closers should read the results or that option X or Y has been voted out. There is a minority for all questions, which is logical if you divide an RfC into so many questions; I tried to frame it as a yes-no question. You had your chance to suppport that question at the time, but decided against it. To claim the result as a yes-no vote now and dismiss anyone who sees through that obvious redefining of the question as "whining" is not helpful. If the outcome is so obvious, then why the need from at least three people not voting for A to put this here and pre-decide the outcome? The only things obvious is that there is no consensus for B, C, D, E or F, but for some reason none of you feel the need to dwell on those, only focusing on A again and again. Fram (talk) 12:29, 14 May 2018 (UTC)[reply]

Sure, you had your chance. And you are whining about the outcome, even though the question was your preferred Can Wikidata infoboxes be used in mainspace?. Let's face it, you're only complaining because the majority didn't agree with you. You don't see me complaining that the majority didn't agree with me, do you? This RfC was problematical, not because of Question 1 (which asks a simple question that has a simple answer); and not because the answers were pre-packaged into a multiple choice; but because of the colossal amount of misinformation and misunderstanding spread about beforehand. My aim now is to examine what works well and build upon that, and discuss what concerns folks have and find ways of resolving them. Your position has been akin to Ian Paisley's "No surrender" stance that blocked progress for years. It's time for you to start seeking compromise and working with those who are doing their best to ensure that Wikidata integration produces results that the majority can be happy with. --RexxS (talk) 13:03, 14 May 2018 (UTC)[reply]

I see you reframing the RfC to suit your purposes. You still sprout the same "Wikidata is progress, anyone opposing integration of Wikidata is hampering progress" nonsense. I am more than happy to work with "those who are doing their best to ensure that Wikidata integration produces results that the majority can be happy with", but I haven't found them so far. I have found people doing all they can to ensure that the Wikidata version of infoboxes remain, or the Wikidata info in infoboxes, like Mike Peel with the World Heritage Site infobox, or you in multiple instances of the biography infobox. I don't need your patronizing holier-than-thou words when your deeds speak for themselves, and I have seen often enough what "progress" Wikidata infoboxes have brought after 5 years of experimenting. But apparently this is due to my stance that blocked this "progress" for years, even though only one Wikidata infobox (which was really dreadfully implemented) has been reverted due to my "stance" and you and others were free to continue developing and showing the progress. But of course, if you want to dismiss all people not wanting more of these infoboxes as being duped by the "colossal amount of misinformation" and not as having genuine concerns and enough valid arguments to oppose the continuation of this experiment, then it is easy to ask for compromise. The majority agreed with no one here, not just with me, as this is an extremely divisive issue, which has been shown again and again. The smart thing then is to let such an RfC run its course and let uninvolved admins close it without trying to influence them. All my "whining" has been in response to people like you who just can't let this happen and feel that they need to pre-summarize the outcome for some unclear reason. A great basis for compromise and collaboration you are forming there. Let's just drop this silly conversation and let people simply close it without "helping" them with these counts, shall we? Fram (talk) 13:17, 14 May 2018 (UTC)[reply]

You're still hoping that this RfC will be closed in your favour, aren't you? Well, what will your response be when it isn't? Will you actually stop insulting me ("holier-than-thou" indeed), and start working with me to improve how Wikidata is brought into Wikipedia? You obviously need reminding that I've created no infobox templates other than cut-down ones for demonstration purposes, like {{Infobox video game series/Wikidata}}. I have worked for five years refining the tools that editors have asked for: ensuring that every call will defer to any locally supplied value; implementing whitelisting and blacklisting of fields to enable opt-in infoboxes where control is at the individual article level; ensuring that by default only Wikidata that has a reference is imported; providing convenient links to the Wikidata entry to edit it or to check references; and more. You only have to look at the "1A, 1B, etc." votes to see that many are made under false assumptions, with no experience or knowledge of the conditions governing how Wikidata information is able to be imported. So I don't think I'll be giving you a free run to misdirect the closers in the vain hope that they might think that a minority somehow represents a consensus. --RexxS (talk) 13:45, 14 May 2018 (UTC)[reply]

No, I'm hoping that whoever closes this will see the attempts from you and others to influence his close. If you don't want to be insulted, then drop the personal comments and you won't be replied to in kind. I have no interest in working with anyone with the attitude you display here though. Can you give a link to any post by me instructing the closers how to close this? Or are all my posts on this subject simply replies to others proclaiming triumphantly that A hasn't gotten a majority, and keeping silent on the much lower scores for B, C, D, E, or F? "You obviously need reminding that I've created no infobox templates other than cut-down ones for demonstration purposes" No, I don't need reminding of this, both because it's information that doesn't interest me, and because I didn't claim that you created infoboxes or not. It would be nice if you could reply to what has actually been said (or just not reply at all). Not to what hasn't been said, not to what you think people hope or believe. I could just as well claim that you still hope that whoever closes this will dismiss most "A" votes "because they were made under false assumptions". Fram (talk) 14:04, 14 May 2018 (UTC)[reply]

And regarding your request as to what needs to be done if this one closes as no-consensus, then we will need a follow up RfC. And that will be the case for anything that is not 1A or 1F. --Dirk Beetstra ^{T C} 14:12, 14 May 2018 (UTC)[reply]

@Fram: Keep insulting me then - it just shows the weakness of your arguments, as does your disdain for working with other good-faith editors like myself. Here's the link to you trying to influence the closers by whinging about Q1, Can Wikidata infoboxes be used in mainspace?. Now, can you give a link to any post by me instructing the closers how to close this? Or are all my posts on this subject simply replies to others proclaiming that there is a consensus while ignoring the fact that their preference is not shared by the majority? After all, I didn't start this thread. ... your deeds speak for themselves, and I have seen often enough what "progress" Wikidata infoboxes have brought after 5 years of experimenting. - it seems you did create the false impression that I had been creating "Wikidata infoboxes". So yes, you do need reminding of the facts when you have the nerve to criticise my efforts to accommodate every concern that's been raised by now. It must be really galling for you to see how much real progress has been made in meeting genuine concerns, but of course it will never be enough for you, will it? As for my expectations, I'll certainly explain to you the false assumptions that several A votes were made under if you ask. --RexxS (talk) 14:32, 14 May 2018 (UTC)[reply]

I guess both Beetstra and myself (and many others presumably) are then unable to read, write, and comprehend arguments. If you read my reply to this thread as an instruction to closers how to close this, then so be it. To me, something like "Are you really going to insist that there is consensus for those two outcomes? Because it sure doesn't look like any consensus to me. If it lies open forever, it will just as clearly signal "no consensus" as would a default close. Your post of 12:35, 12 May 2018 (UTC) is the one focusing solely on the "A" option for some reason, as if it was an A vs not A question. Followup replies by you have "A "no consensus" close will not alter that. " and " the majority recognise that." and "another 30 days of inconclusive debate.". Adding to that the claim I already repeated above, about "A" votes being made "under false assumptions". Oh, and "it's no good whining over the outcome now where only a minority answered that question in the negative. There's no way that a minority view can be a consensus, so accept it and move on." I don't think I'll bother explaining the discrepancy between what I said about your work on Wikidata infoboxes and your defense of them, and your "false impression" statement above. If you can't see the difference between what I actually said, and what you read in it, then it is another piece of evidence that it is pointless continuing this discussion. Fram (talk) 04:36, 15 May 2018 (UTC)[reply]

I indeed do not see anymore how I have to explain that WikiData data is not reliable. As I said, anything else than a full 1F or a full 1A closure will need follow up discussion, and that is best performed in the form of an RfC. How that follow up RfC is going to be is depending on how this RfC is closed - if it is a default 'no consensus' close, then that will be an RfC similar to this one but likely more direct, or an RfC that is going to specify how to interpret specific questions regarding the 'intermediate' close. --Dirk Beetstra ^{T C} 06:34, 15 May 2018 (UTC)[reply]

And I don't see why you insist on trotting out patent falsehoods like "Wikidata is not reliable". That is pure nonsense – propaganda for those who wish to resist progress in any form. Anyone who looks at Wikidata can find huge amounts of reliable information, and I find it astonishing that you could think that folks are so gullible that they might fall for your claims. What unreliable data is present in the infobox at South Pole Telescope? Twenty pieces of information, and every one of them pulled from Wikidata. How about a more modest effort: François Barbé-Marbois whose infobox has only six pieces of information, each of which can be easily verified by following the icon-link. There is more information available on Wikidata, like the name of his wife, but it's unreferenced and so does not appear. By default we don't bring in unsourced claims, and that means that the editors on Wikipedia are in control of what comes from Wikidata. It matters not a jot what Wikidata's sourcing policies are, because we can enforce our own. Until you realise that, there won't be a common starting point for further discussion, but if you want to waste another month arguing from an unreal position, I guess that's your prerogative. I would prefer to spend my time improving the encyclopedia. --RexxS (talk) 17:21, 15 May 2018 (UTC)[reply]

It is an open wiki, and an unstable one, the very definition of an unreliable source. Wikidata gets spammed, vandalised, filled with vandalised data because the editor does not care to check. WikiData has the same disclaimer as Wikipedia. You have no clue whether data is correct. It is not propaganda, it is s8mply a site that fails blindly our sourcing requirements. I don’t know how often I have to repeat that. I don’t trust em.wikipedia to be a reliable source, I don’t trust es.wikipedia to be a reliable source, I don’t trust a sentence on en.wikipedia that ‘’has’’ a reference without checking a reference. I don’t trust WikiData, and you expect that Wikipedia will be more trustworthy after blindly and automatically importing data from another site ‘because we only use data that is referenced’. WikiData IS an unreliable source. I spend my time on improving this encyclopedia, not by making it worse with this blind importing of unreliable data. Have you read WP:V recently. Don’t you see that WikiData fails that? That the material is sourced does not matter, a lot of material on en.wikipedia is sourced, but WikiData has no control over how well it is sourced (just like en.wikipedia), or whether the statement has not been changed since.

I know I m not convincing you, and probably most of the other W8kiData fans, and tha5 is likely why we need a follow up RfC on any close that is not pure 1A or pure 1F.

I said earlier that I was going to stop, but I am afraid that that we will need to repeat that any claim of reliability on WikiData is utter nonsense. —Dirk Beetstra ^{T C} 20:13, 15 May 2018 (UTC)[reply]

@RexxS: In case you think that I am the only person saying that WikiData is an unreliable source, then please read the !vote of the last two !voters. From the very first all the way to the now last, it is a recurring theme: WikiData is an unreliable source. —Dirk Beetstra ^{T C} 20:30, 15 May 2018 (UTC)[reply]

Wikidata is not a source. At all.--Ymblanter (talk) 20:41, 15 May 2018 (UTC)[reply]

What? So where is Wikipedia getting that data? —Dirk Beetstra ^{T C} 21:02, 15 May 2018 (UTC)[reply]

Editors add it, just like they do on Wikipedia. Wikipedia is not a source, either. --RexxS (talk) 21:21, 15 May 2018 (UTC)[reply]

(edit conflict) And I don't know how often I have to repeat that Wikidata is not a source. It is a repository. The facts stored there are either sourced or not. The sources for those facts may be reliable or not. You know exactly if the data is correct in just the same way as you do for data stored on Wikipedia: you check the reference. Did you read what you wrote? "I don’t trust a sentence on en.wikipedia that ‘’has’’ a reference without checking a reference." and that's exactly the same for a fact imported from Wikidata. Why are you claiming that there is any difference? The steps you go though to meet WP:V are just the same. There is nothing about Wikidata that inherently makes the information found there unreliable, any more than that sentence on Wikipedia that you are willing to check. Why not check the fact imported into an infobox from Wikidata and test whether it meets our sourcing policies? If it doesn't, you can block it, fix it, or supply a value that you know to be correct. That's no different from any other fact in any non-Wikidata infobox. I also spend my time improving this encyclopedia, by making sure that editors have the tools to make proper use of information from Wikidata without having to do it blindly. You can stop any time you like, but you've still not refuted my demonstration that Wikidata contains reliable information in the two infoboxes I quoted to you. Here's another two at random: Very Large Telescope and Robert Axelrod. You'll see that the first has some local parameters which override the data on Wikidata; the editor has control. In the second, Wikidata gives his occupation as "mathematician", but the infobox doesn't show that because on Wikidata it's only sourced to the English Wikipedia, so it never arrives in our article. What's left is information that can be checked against the sources, just the same as in any other infobox. How many more examples of reliable data on Wikidata do I have to give you before you concede that the way we draw information from Wikidata into an infobox means that such information is made just as reliable as what we would find in an infobox that was populated completely by locally supplied values? --RexxS (talk) 21:21, 15 May 2018 (UTC)[reply]

”It is intended to provide a common source of data which can be used by Wikimedia projects such as Wikipedia ..”. Every place that contains data is a source. WikiData is a repository for data, a common source of data. If I use it, I sourced it on WikiData, and WikiData sourced it elsewhere. CNNis a source, PubChem is a source, my mother is a source, any place you get data from is a source. WikiData is a source. —Dirk Beetstra ^{T C} 21:49, 15 May 2018 (UTC)[reply]

From WP:RS:

Definition of a source ... The word "source" when citing sources on Wikipedia has three related meanings: The piece of work itself (the article, book); The creator of the work (the writer, journalist); The publisher of the work (for example, Random House or Cambridge University Press)

. Which of those three possibilities does Wikidata fit? It is not the work itself, as that is the facts that are imported. It is not the creator of those facts, as they come from external authors. It may well be considered the publisher of the facts, but like a Wikipedia, it is only ever a republisher of information that has already been published elsewhere. The reliability of the facts that arrive in Wikipedia are not affected by their transmission via Wikipedia – they were either reliable or not where they were originally published. Again, I'll make it clear: to judge the reliability of any fact on Wikipedia, whether added locally or imported from Wikidata, you have to look at the real source, the one you find by examining the reference. And the result is identical, regardless of whether the source is cited at the bottom of the article or on Wikidata. No matter how you look at it, Wikidata is not a source in any meaningful sense of the word for any of the facts that arrive in our infoboxes. --RexxS (talk) 22:26, 15 May 2018 (UTC)[reply]

of course it is A piece of work. It is a database, a list of information, it has information aggregated. That is a laughable argument. Let me dumb it down: it is an unreliable database of information where wikipedia gets part of the information it displays gets from, and for every infobox that displays correct data transcluded from WikiData I give you 50 correct statements on Wikipedia (with references). That however, does not make wikipedia, nor wikidata, a reliable place to withdraw information from. Both are open wikis. Both do not have the sufficient protection mechanisms that are needed to make it reliable (even is mostly correct). Unlike Wikipedia, however, WikiData has the possibility to to become a reliable database with information (if they change their model and work on it), in which case I will fully support transclusion of data as that would (massively) increase the reliability of wikipedia. Until then, as you say, it makes, at best, no difference, and think that WikiData is just bringing the quality of en.wikipedia down. —Dirk Beetstra ^{T C} 03:34, 16 May 2018 (UTC)[reply]

If we apply this reasoning consistently, information which is added to Wikipedia articles from other Wikipedia articles is sourced to Wikipedia (and there is quite a lot of non-trivial information of this kind, for example, in the articles on railway stations the whole railway lines are imported via templates and can not be modified within the article). Even more, since the templates typically do not reansclude the sources, by this logic, this information is not reliable.--Ymblanter (talk) 06:38, 16 May 2018 (UTC)[reply]

Wow ... You just confirmed Wikipedia:General disclaimer and WP:VERIFY. Wikipedia is not a reliable source, WikiData is not a reliable source, external open wikis are not a reliable source. Whether it is referenced or not only gives you a chance to verify, it does not make it a reliable source. --Dirk Beetstra ^{T C} 07:08, 16 May 2018 (UTC)[reply]

Sure, but I do not see anybody around on a crusade to eliminate templates in Wikipedia because they transclude data from an unreliable source and are difficult to edit. The discussion on what to do with vandalism on Wikidata, and how Wikidata-transcluded fields in the template can be edited is a valid discussion. The full rejection of Wikidata because it is not a reliable source is not a valid viewpoint. It is not a source, and should not be viewed as such.--Ymblanter (talk) 07:22, 16 May 2018 (UTC)[reply]

What, is that what I am doing, being on a crusade to eliminate things? Wikipedia is not a reliable source, and we are all working to get it better. Wikidata is a source, it should be viewed as such. You are aggregating data from other sources, but by that it becomes a source - what else is the use of WikiData, just aggregate data . If I create an open Wiki on Wikia, and I only add information with references there, I am still an open wiki and I am still not a reliable source, even if all material is true.

Now secondly, Wikipedia has a massive editor base and a significant number of admins. They have significant capabilities in vandal fighting and all those edits happen here. And we do have capabilities of vandal fighting on templates (and by the way, many templates use data within the page format the display, so vandalism happens on the article, not on the template - many templates with higher risk are protected so they cannot be vandalized, all measures to increase the reliable use of templates).

But now we have established that Wikipedia is not a reliable source, in part because we transclude templates that can be altered behind the scenes, how do you think that becomes if we transclude data from an external source through templates that can be vandalized into our articles.

Again, if WikiData would be more stringent on the correctness of their data than Wikipedia itself is, then it would make sense to transclude their data (as that would make the data on Wikipedia more reliable), but since that is not the case, transcluding that data will only make Wikipedia less reliable. I am willing to change my mind when that happens, but that needs a lot of work to be done on WikiData. --Dirk Beetstra ^{T C} 07:49, 16 May 2018 (UTC)[reply]

Right, we are back to our discussion of half a year ago. You think we should forget about Wikidata and bot-import data from databases directly to the English Wikipedia (and other Wikipedias and sister projects, which you do not have to care about, should themselves organize a similar process). I disagree with that, and I think the positions are pretty much clear. (I still insist that saying that Wikidata is used as a source is incorrect - we do not have a single piece of data outside of Wikidata article sourced to Wikidata). They are probably not going to change. The problem is that stupid bot owners for whatever reason are unwilling to perform this export directly to the English Wikipedia (possibly because many of them just do not care about the English Wikipedia) and they perform this import of data - reliably sourced - to Wikidata. Try to convince them to come here, or perform your own bot import, may be the problem is solved then.--Ymblanter (talk) 08:00, 16 May 2018 (UTC)[reply]

I don't need to convince them, our editors are more than capable. And I see scenarios where it is possible to use WikiData, but that needs a change in how WikiData operates, it needs to become significantly stricter in sourcing ánd protection so that the reliability (not correctness) is better than Wikipedia.

Indeed, no data is sourced to WikiData, because then you would show that it is wrong - but if the infobox would tell you where the data was from, it is from WikiData, and that is the source of the information ... there should be a citation that the data is from WikiData, or, since WikiData is not a reliable source, the data should not be there in the first place. --Dirk Beetstra ^{T C} 09:06, 16 May 2018 (UTC)[reply]

We are going in circles. By the same logic, template transclusion at Wikipedia should not exist.--Ymblanter (talk) 09:15, 16 May 2018 (UTC)[reply]

No, because templates can be protected and/or templates display information that is in the article. Moreover, if you transclude a template, you know what you transclude, unless the template is altered (of which many are protected) the display does not change. Here you transclude a template, and when you add data on WikiData that has a reference, the page changes. Yes, we are unreliable, that we already know. But WikiData is less reliable at the moment and not in the article - not even on this wiki.

The point is, we have reliable data from an non-MediaWiki source, we include the information and add a reference to the reliable data .. done. But if you import that data into WikiData, you have it stored on a unreliable source (even if correct), and then Wikipedia transcludes it from the unreliable (even if correct) source (WikiData). By your logic, when Wikipedia is putting a statement with a reference, Wikipedia becomes a reliable source. And that is not so. The very fact that anyone can change the fact, change the reference, or change both makes it, even when correct, still unreliable.

But I agree, we are going in circles. --Dirk Beetstra ^{T C} 10:46, 16 May 2018 (UTC)[reply]

@Beetstra: Sorry, but your behaviour is totally unfair: "it needs to become significantly stricter in sourcing and protection so that the reliability (not correctness) is better than Wikipedia." You can't require that WD has an higher reliability than WP, you can just require than WD has the same requirement than WP and have expectation than WD can provide better reliability.

Then always the same argumentation which doesn't take care of the reality:

"no data is sourced to WikiData": wrong some data are sources, in the same way than WP, and it is possible to filter data according to their source

"the infobox would tell you where the data was from, it is from WikiData": wrong, you can display the source data beside the value in WP. And if you don't trust WD as repository for source data, how can you continue to work in WP as the same reality exists: everyone can modify WP in the same way than in WD. And we have the same way to monitor modification than in WP using watchlist. Snipre (talk) 11:18, 16 May 2018 (UTC)[reply]

Totally unfair? That is what we see with commons, they are stricter on fair use than we are. And I possibly could live with a similar reliability .. but we are far away from that.

Again the same red herring, 'it is possible to filter data according to their source' - yes, but a) you don't know if the data changed under the source, b) whether the data is actually properly supported by the source, or whether the source is a source that we would use (for that data). Then there is data that does not need a source in some cases, and does in others. "sky:color=blue".

No, the data is from WikiData, it is sourced from WikiData, and WikiData sourced it somewhere else - and I can get that data as well. I agree that it helps that we can now see more of the changes, but we don't know whether the data that is transcluded was right in the first place. Guys, WikiData is an unreliable source, it is an open wiki. --Dirk Beetstra ^{T C} 12:03, 16 May 2018 (UTC)[reply]

• We seem to be re-arguing everything already stated with the !votes in the threads above. No need to repeat ourselves. The closers will read all our comments and make a determination. Let’s leave them to it, shall we? Blueboar (talk) 12:29, 16 May 2018 (UTC)[reply]

  Indeed. I'm a bit disturbed by all these "1A is definitely losing" attempts, above, to cloud an incoming closer's decision-making and assessment process, especially since the statistical technique, if you can even call it that, is childishly faulty, comparing 1A against all other options combined. The air is thick with desperate bullshit in here, so I feel impelled to counter that manipulation by blowing some holes in it. 1A is clearly the single leading outcome, by a very wide margin. As of this writing, and to just take question 1 (since if it goes in the direction of 1A, it makes the other questions moot), I see a 41 : 3.5 : 5.5 : 13.5 : 3 : 17.5 ratio (in A through F order, counting things like "1D or 1F" as 0.5 each). Even if you do add it up "all against 1A" (which is patently manipulative, and if turned around each in turn into "all against 1B", etc., looks very bad for every option except 1A) it's 41 : 42, which in any sane assessment will come out as a consensus for 1A – a single solidly supported outcome with consistent rationales, versus a bunch of random stuff with incompatible rationales – or as "no consensus" if the closer is spineless.

  However, there are more important considerations here that the raw numbers (what with this being not a vote, which is part of why I changed the heading that implied that it was one). First off, 1F is irrational. I reads, "Every infobox that is technically ready to convert from local content to Wikidata content may be implemented in mainspace", but the entire nature of the dispute is that many in the community do not believe any infobox is "ready" to be converted to WD, because WP's core content policies do not apply on WD, so we have no en.WP content control over what appears on en.WP via WD. That is, 1F is a textbook example of the fallacy of begging the question. Second, 1B has no bearing whatsoever on 1A versus any other option(s), because it's a sandbox. Indeed, 1A could proceed as the consensus, and 1B could happen anyway, because we can do anything reasonable in a sandbox that isn't presented as an actual article (otherwise the entire "Draft:" namespace and much of userspace and the project namespace would be deleted immediately). 1C is essentially identical in actual meaning to 1A, other than it concedes that a future RfC might decide otherwise on a case-by-case basis. Given the overwhelming solidarity behind 1A, an RfC producing a 1C exception outcome for any given template is unlikely. Though 1A doesn't state that a future RfC might overturn it or make limited exception to it, and we all know it, and it's implicit in 1A's meaning, because no RfC or other decision here is Forever, and every consensus can change. Lastly, 1D and 1E are basically more rational takes on the vague sentiment behind 1F. If you're going to do math, add 1A and 1C versus 1D, 1E, and 1F, and ignore 1B as noise. The 1A+1C outcome is clearly strongly dominant. This is another case of "too soon", as so many of us have been saying about WD for several years now. Rather than fix the core (namely core content policy) problem, by rejiggering WD to support tagging of data with different sourcing standards, so that data can be extracted on sourcing-filtered basis, WD's fans keep re-re-re-RfCing this stuff in hope of WP:WINNING by exhausting the opposition (another pair of classic fallacies). It's patent forum shopping, and it needs to stop.  — SMcCandlish ☏ ¢ 😼  15:58, 18 May 2018 (UTC)[reply]

  @SMcCandlish: While I'm not sure about the discussion in this section, you can hardly claim that the RfC is on the whole forum shopping. There hasn't been an organized RfC on this in ages, and this RfC has been directly requested by the Arbitration Committee in order to resolve the content disagreement within the community. I'm not sure about the conduct in the section, but the RfC itself can hardly be called forum shopping. Tamwin (talk) 19:13, 23 May 2018 (UTC)[reply]

  The WD fans pushing, pushing, re-pushing, the same basic "let us replace WP content with WD content, never mind your concerns" thing, in essentially the same terms over and over again, without the concerns being addressed (yet the concerns remaining the same and being widespread among editors, i.e. representing a consensus that they're legitimate issues that need to be resolved first), well, yes, it is forum shopping. The sad thing is, a lot of us see the potential of WD, but the WD crowd don't want to accede to anything, they just want it their way, and now. But it's not ready for primetime, not when you're talking about directly integrating WD content into en.WP and other sites with their own verifiability and other policies.  — SMcCandlish ☏ ¢ 😼  22:17, 23 May 2018 (UTC)[reply]

  " ... without the concerns being addressed ...". That's a lie. Every legitimate concern raised with me about the code I write to import Wikidata into infoboxes has been addressed, and resolved. You should be ashamed of yourself for spreading such patently untrue propaganda. --RexxS (talk) 11:00, 24 May 2018 (UTC)[reply]

  "The sad thing is, a lot of us see the potential of WD, but the WD crowd don't want to accede to anything, they just want it their way, and now." I don't think this is true at all. This to me mostly seems like a "we will take the problems in the short term, to solve them in the long term" vs. a "we don't want any problems in the short term, call us when it works" type of debate. And neither is acceding on the opposing viewpoint. It's actually not so different from inclusionists vs deletionists debates we have had in that regard. —TheDJ (talk • contribs) 11:39, 25 May 2018 (UTC)[reply]

Posted closure request on AN Galobtter (pingó mió) 13:58, 14 May 2018 (UTC)[reply]

How is it that I found this page only now, because today I happened to casually took a look at a infobox talk page ? --Robertiki (talk) 15:30, 27 June 2018 (UTC)[reply]

Robertiki the RFC was unusually well advertised, including but not limited to announcement on Centralized Discussion.[18] There will always be people on both sides of an issue who don't notice an appropriately advertised RFC until it's over. If you want to more actively participate in RFCs in general, you can always join Feedback request service. You could also watchlist Village Pump (proposals) and/or {{Centralized discussion}}. Alsee (talk) 16:19, 27 June 2018 (UTC)[reply]

Above, there are some discussions about using the watchlist or recent changes on enwiki to keep an eye on Wikidata changes affecting enwiki articles. Which sounds all fine and useful, until one realises that once again, the delay between a change on Wikidata, and its appearance on our wtachlist or on recent changes, has now become more than 15 hours (you can check the current situation on this page). Which means that no Wikidata changes appear in recent changes, and in your watchlist they only appear way down your list, long after you have checked the other (enwiki) changes made at that time.

This is an unworkable situation and sets the door wide open for vandalism (or good faith mistakes) which can't be spotted by regular vandalism watchers or most watchlist users on enwiki. This isn't the first time this happened, I have dropped notes about such delays and longer ones regularly in 2017 as well. Fram (talk) 10:59, 24 May 2018 (UTC)[reply]

That's unusual. Pinging @Lydia Pintscher (WMDE):. Mike Peel (talk) 11:42, 24 May 2018 (UTC)[reply]

Like I said, not really unusual at all. There are better periods (although even then the delay is often too much to be useful for recent changes, but not for the watchlist), but these bad tims happen too often as well. Fram (talk) 11:49, 24 May 2018 (UTC)[reply]

It's unusual enough that you felt it necessary to mention it here and now. From my checking of it in the past, it's normally minutes, not hours. Mike Peel (talk) 13:38, 24 May 2018 (UTC)[reply]

Thanks a lot for the ping. That is definitely not normal as you can see on https://grafana.wikimedia.org/dashboard/db/wikidata-dispatch?refresh=1m&orgId=1. Something broke and my team is looking into it right now. --Lydia Pintscher (WMDE) (talk) 15:21, 24 May 2018 (UTC)[reply]

Ok we found the issue. It was a case of trying to improve something and breaking it instead. Fixed now. The lag is still high but going down now. Sorry again for the issue. --Lydia Pintscher (WMDE) (talk) 18:25, 24 May 2018 (UTC)[reply]

I mentioned it "here and now" because multiple edits on this page right before my post were discussing the use of the watchlist and recent changes to watch relevant Wikidata changes. As this discussion was already spread over different sections, it seemed easier to add a new section especially for this. Fram (talk) 04:30, 25 May 2018 (UTC)[reply]

• At 07:07, 30 May 2018 the article Diamela del Pozo was created. It has {{Twitter}} and {{Facebook}} on them.
• The WikiData item at that time is NOT linked to the new Wikipedia article. See box 'Wikipedia' in top-right, only showing an 'es' entry in this revid of 04:47, 13 February 2018‎ on WikiData
• That results in the article NOT transcluding the Twitter and Facebook, but displaying "{{Twitter}} template missing ID and not present in Wikidata." and "{{Facebook}} template missing ID and not present in Wikidata.".
• At 7:08the item is linked to the en.wikipedia article. Thát action shows as "07:08 Q21127480 (diff | hist) . . Nick Number (talk | contribs) (A Wikidata item has been linked to this page.)"
• That not only results in the appearance of the interwiki in es in our article, but also in the transclusion of the twitter and facebook (they now display as "Wikidata/2018 Infobox RfC on Twitter" and "Wikidata/2018 Infobox RfC on Facebook". (the watchlist-message from WikiData does not say that it actually resulted in the transclusion of the facebook, and twitter).

This shows, that we data appears in articles after an edit is performed, and that an editor who edits here does not know what data will be transcluded in the future. This is a new BLP, which in the minutes after creation (and probably for a relatively unknown person in the hours/days after creation - in about 2 hours the number of watchers is still '2', me now being 1 of them to be able to test above, the other I assume is the page creator) editors could introduce all sorts of data into BLPs that are not noticed by anyone on en.wikipedia but the page creator (and only if they both watchlist their new article, and have WikiData changes turned on). With editors regularly being away for several hours, with 2 page-watchers edits on WikiData are easily missed. --Dirk Beetstra ^{T C} 06:23, 30 May 2018 (UTC)[reply]

I'm not sure what the problem is that you're trying to point out here? The same editor who created the enwp article added the sitelink on Wikidata (@Nick Number:), so it would be reasonable to assume that the Wikidata-enabled templates were added with the intention of showing the info from Wikidata after they added the sitelink. If you're watching the article, then you'll see "A Wikidata item has been linked to this page." as a line in your watchlist. If you're doing new page patrol then you either see the error messages or the post-sitelinked version of the article. Do you somehow want the Wikidata sitelink to be added before the article is created? Mike Peel (talk) 11:27, 30 May 2018 (UTC)[reply]

No, this is to show that data is/can be added after a transcluding template is being placed (which is obvious). On low-watched articles (as this article is), that means that only very few editors will see material being added to such an article. There are above continuing arguments that people who add a WikiData-transcluding template to an en.wikipedia article are responsible for checking whether the material that is transcluded at that time is correct/verified/appropriate - but if that material is added later (as this example technically shows - there are no further issues in this example) there may be no-one to check whether the material is appropriate for en.wikipedia. --Dirk Beetstra ^{T C} 12:31, 30 May 2018 (UTC)[reply]

So...I'll just keep doing what I'm doing then. If I'm understanding correctly, there's no great harm in having broken templates on the article for the <1 minute between creating it and linking it. Nick Number (talk) 15:45, 30 May 2018 (UTC)[reply]

No Nick Number, what you did was very correct, you just gave me an example of something that I think could give problems elsewhere. I’ll wait/search for such an example, but I guess people get the idea. —Dirk Beetstra ^{T C} 16:24, 30 May 2018 (UTC)[reply]

But surely if an editor adds a template to a page with only two watchers and somebody later alters how that template renders in the article, very few editors will see that change. That's true whether the template is connected to Wikidata or not. Why are you touting this as a problem for Wikidata-infoboxes, but not for any other type of template? Are you going to be asking editors who place any sort of template to be responsible for any future changes to the template? As usual, this is a general issue that is being misrepresented as something peculiar to Wikidata-infoboxes. I will assume that anyone observing this thread will get what the actual situation is. --RexxS (talk) 20:38, 30 May 2018 (UTC)[reply]

I agree that this is a problem with using templates in general (and why I don’t think we should use them) ... however, also think that the problem is harder to spot and fix when WD is involved. Blueboar (talk) 23:44, 30 May 2018 (UTC)[reply]

a) templates transcluded on pages generally have way more watchers, b) edits to the new BLP here are to the page here, and indeed will not have many watchers. BLP violations through edits on WD are just way harder to spot, WD does not have the vandal fighting capabilities that en.wikipedia has. —Dirk Beetstra ^{T C} 03:15, 31 May 2018 (UTC)[reply]

The point that I am trying to make here is though: there are arguments being made basically saying: "if I add an infobox to en.wikipedia, then it is my responsibility to check whether all the values that are being transcluded are properly referenced, factual, correct, etc. etc.". The point is, that I can add an infobox with transcluded information that is thoroughly checked by me, but that someone else can later, on WikiData, add data that results in it being transcluded without having been checked by me. The editor who added the infobox does not have responsibility for what is being added later on WikiData, and the editor who adds data to WikiData has no responsibility for what is transcluded on en.wikipedia. I know that the situation is the same as with changing a template on en.wikipedia that results in changing articles, but that editor has an en.wikipedia responsibility, and has to follow en.wikipedia policies (or at least, en.wikipedia standards), whether he edits the article or the template. I let others extrapolate what that means in areas where there are legal policies applicable (BLPs), or where we have sanctions applied (arbitration enforcement). --Dirk Beetstra ^{T C} 07:39, 31 May 2018 (UTC)[reply]

Most widely used templates on enwiki are restricted to very few users (admins plus template editors), meaning that the chance that someone changes e.g. something in "template:infobox biography" as vandalism is very, very low (accounts get compromised, and editors do stupid stuff, but these are exceptions). On the other hand, if you have a template in your article which is protected here, but fetches data from Wikidata, then the values displayed here can be vandalized much more easily (see the examples of changes to the labels of countries). Basically, on enwiki you can change template data locally (directly in the article), which will show up in the page history, watchlist and recent changes immediately. For most templates, you (where "you" is a random vandal) cannot change the template directly though. But with a Wikidata-filled template, random vandals again can easily vandalize articles, and it doesn't show up in the page history, it often doesn't show up in recent changes (due to the lag, e.g. at the moment WD changes take more than 6 minutes to appear here), so at best it shows up in the watchlist. And after it has eventually been reverted on Wikidata, it may still linger in enwiki for days. Fram (talk) 08:46, 31 May 2018 (UTC)[reply]

Stronger than that, we even protect single use, fully prefilled templates (infoboxes) so that the data they display does not get vandalized. With implementation of WikiData on that, we would lose that capability. --Dirk Beetstra ^{T C} 10:33, 31 May 2018 (UTC)[reply]

First, my thanks to the closers. Now... the next question is how to implement what they say. If the consensus is that Wikidata can be used ... but only as long as Wikipedians can be assured that Wikidata’s info is reliable... how do we achieve that assurance? I would say the most obvious way would be to require an in-line citation to a reliable source along with any info that is populated via Wikidata. It would be up to the developers at Wikidata to figure out how to make that happen. Blueboar (talk) 22:29, 13 June 2018 (UTC)[reply]

@Blueboar: Most of the Wikidata templates already have the code built in to show the in-line references (set refs=yes), it's easy enough to change the default for that back to showing them if we want that. Thanks. Mike Peel (talk) 22:41, 13 June 2018 (UTC)[reply]

@Blueboar: What have the developers at Wikidata got to do with how we import data into Wikipedia? The code to do that job resides here, under our control, not on another project. Why should we have to have an inline citation for "sky is blue"-type statements or for images that are imported from Wikidata? We already have the guarantee that each piece of information likely to be challenged has a reference, and that it can be verified by following the link to that reference. What more must be required to meet Wikipedia's policies regarding factual information? I understand that data subject to MEDRS will need stronger sourcing, but the means of verification, even of that, can rely on the same mechanism. --RexxS (talk) 23:42, 13 June 2018 (UTC)[reply]

Well, according to the closers, the consensus of the community is: “what WP does now is not enough” ... so my question is essentially asking: “OK... then what more needs to be done? What do we here at WP want WD to do so we can be assured of reliability?” Blueboar (talk) 02:20, 14 June 2018 (UTC)[reply]

I would say there are two main problems: 1) unsourced and badly sourced data and 2) vandalism on Wikidata. For 1) we need to require that all data we pull out of Wikidata is sourced reliably for our standards, as discussed above. 2) is at this point more difficult to address. I would say we need (possibly with the help of Wikiprojects) to identify the items which gets vandalized too often and are not safe at this point to show here - for example, country names (or, possibly, another way round - which items are relatively safe to show, or, conversely, where the benefit of having them on Wikidata overweighs the risks related to vandalism), and also we need some technical discussion about the Google caching issues identified during this RfC,--Ymblanter (talk) 05:37, 14 June 2018 (UTC)[reply]

I am afraid that this is an unclarity in the closing statement (or in the RfC). I guess a follow up question is, whether the community thinks that the material sourced from wikidata is sufficiently sourced on wikidata (when it carries sources, or when it does not need to be sourced). —Dirk Beetstra ^{T C} 06:01, 14 June 2018 (UTC)[reply]

@Ymblanter: I've taken two significant measures with Module:WikidataIB following the close of this RfC: (1) because the Wikidata label has shown itself vulnerable to casual vandalism, where a Wikidata item has a sitelink, that is now used instead of the label (thanks to Was a bee for suggesting that); (2) the code now only fetches the Wikidata statement required, rather than the whole entry, which should help eliminate a lot of the irrelevant changes reported in watchlists. The aim is to eventually make it easier for more editors here to use their watchlists to spot vandalism on Wikidata. If anyone has any suggestions for further improvements in how we import from Wikidata, please express them. --RexxS (talk) 10:34, 14 June 2018 (UTC)[reply]

Thanks, it definitely is a good direction.--Ymblanter (talk) 10:38, 14 June 2018 (UTC)[reply]

There appears to be a problem with the overall summary. In the fourth paragraph, after finding no consensus on other issues, the closers find

Despite this clear bifurcation, there is a consensus on one point: if Wikipedia wants to use data from Wikidata, there needs to be clear assurances on the reliability of this data.

In the final summary, the first of these conditions is dropped:

There is a consensus that data drawn for Wikidata might be acceptable for use in Wikipedia if Wikipedians can be assured that the data is accurate, and preferably meets Wikipedia rules of reliability.

That seems to have a quite different connotation. Kanguole 10:59, 14 June 2018 (UTC)[reply]

How so? They seem to have the same connotations to me. Blueboar (talk) 11:58, 14 June 2018 (UTC)[reply]

No, the problem is that with question 3, there was no "no data, even sourced, is acceptable" option, so everyone(?) who answered 1A and 2A also answered 3A, not meaning "data drawn for Wikidata might be acceptable for use in Wikipedia if Wikipedians can be assured that the data is accurate, and preferably meets Wikipedia rules of reliability." but "we need the most strict rules about using data from Wikidata, and preferably don't use it at all. Basically, some 40 votes who supported "Wikidata is never acceptable" have now been counted with the "data might be acceptable" consensus simply because no stronger option existed for Q3.

This doesn't mean that it perhaps for now isn't the best possible consensus position, but it is too simplistic to look at 3A in isolation and say that most people supported it. Fram (talk) 12:05, 14 June 2018 (UTC)[reply]

That is reflected in the fourth paragraph, where there is the qualification if Wikipedia wants to use data from Wikidata. However this was omitted from the overall summary, which says that there is a consensus that using data drawn from Wikidata might be acceptable under some conditions. Such a consensus is not supported by the main body of the closure (or the RFC). Kanguole 12:13, 14 June 2018 (UTC)[reply]

I think the point is that if it "might" be acceptable under some conditions, it is certainly unacceptable without them. In other words, the consensus is unclear if we include the conditions, but clearly against the use otherwise. No inconsistency arises from that reading. Tamwin (talk) 18:33, 14 June 2018 (UTC)[reply]

so what the naysayers should do now is spell out what those conditions actually are. Give the good folks at Wikidata some goals to achieve. Blueboar (talk) 20:01, 14 June 2018 (UTC)[reply]

Blueboar most wikidata-proponents have already been willing to do virtually anything&everything technically possible to "improve" how wikidata is used. The problem is that all of those efforts don't help. Those "improvements" do exactly zero to address the fundamental objections of most critics. The fundamental problem is the automatic-import of content. The only improvement that actually resolves those concerns is one that shuts off automatic-import. That is effectively equal to 1A. That leaves us at an impasse. We can't resolve the situation by proponents offering to work harder at what they're already doing. We need X% of people "in the middle" to either get on board with the side saying what we've already got is good enough, or we need X% of people "in the middle" to get on board with the side saying that automatic-import creates issues that are literally impossible to fix. Alsee (talk) 23:33, 14 June 2018 (UTC)[reply]