Wikipedia talk:Reliable sources

RSQUOTE

WP:RS/QUOTE says:

The accuracy of quoted material is paramount and the accuracy of quotations from living persons is especially sensitive. To ensure accuracy, the text of quoted material is best taken from (and cited to) the original source being quoted. If this is not possible, then the text may be taken from a reliable secondary source (ideally one that includes a citation to the original). No matter where you take the quoted text from, it is important to make clear the actual source of the text, as it appears in the article.

I propose removing the word secondary from the middle. We need the text to be taken from a reliable source, but it doesn't really matter whether that source is secondary. There are times when another primary source (e.g., a press release by a company or politician that quotes and apologizes for a tone-deaf tweet they made) or a tertiary source is just as good. WhatamIdoing (talk) 00:45, 16 January 2023 (UTC)

WhatamIdoing made a bold change to remove "secondary" on 22 December 2022, I reverted, along with a reference to a disagreement on the WP:V talk page, Sources that summarize other sources. I can't think where "not possible" actually applies. I suppose the example of a "tone-deaf tweet" is for a tweet that was deleted, but the tweeter's press release is a discussion of what the tweet supposedly said and so is secondary. I continue to oppose this change. Peter Gulutzan (talk) 16:05, 16 January 2023 (UTC)

Press releases are primary sources. (They are also self-published and frequently non-independent of their subject matter.) Quoting someone doesn't transform the source into a secondary source. See WP:LINKSINACHAIN.

"Not possible" could apply to since-deleted recent contents, but it can also apply to lost works. Frequently what we know about works published a thousand years ago is only what someone claimed was in it 500 years ago (see "medieval and ancient works, even if they cite earlier known or lost writings" in the relevant policy).

Similarly, if a history textbook for schools quotes a bit of the US Declaration of Independence, or even reprints it in full, then that tertiary source is accepted by editors. Specifying that we must use secondary sources amounts to a ban on most textbooks available to students. WhatamIdoing (talk) 17:27, 16 January 2023 (UTC)

Citing a school textbook for a quote of that declaration, despite the fact that original or secondary sources could be found, would be against the current wording of WP:RS/QUOTE. Are you suggesting that your proposal would make it okay? Peter Gulutzan (talk) 15:57, 17 January 2023 (UTC)

• I have to agree with WhatamIdoing here… The key is to WP:SAYWHEREYOUGOTIT. If you looked at the text of a high school history textbook when quoting the Declaration of Independence, that textbook is what you should cite. Blueboar (talk) 16:50, 17 January 2023 (UTC)

  Isn't the question one of where you should take the quotation from - in your example, should you look at a high school history textbook for it? NebY (talk) 18:19, 17 January 2023 (UTC)

• More Background: Possibly wrong: The first edition of the section is due to an addition in April 2009 by Mangoe. The specific phrase "from a reliable secondary source" is due to an edit in November 2010 by Blueboar. Blueboar said "based upon talk", which I guess means talk page thread = Quotations. Peter Gulutzan (talk) 18:09, 17 January 2023 (UTC)

  Looking at that discussion, it appears that editors were thinking that there are only two available sources:

  • the original, authoritative source for the quotation, which is primary; and
  • sources that discuss the quotation, which are secondary.

  This is, however, a significant oversimplification of reality, especially for pop culture and recent events. WhatamIdoing (talk) 07:20, 20 January 2023 (UTC)

• As the instigator of the passage, I have two remarks here. First, the textbook sourcing for the Declaration is not a compelling case. We are on the internet; we can source it from the National Archives. Both are, technically, secondary sources, but the second is clearly more authoritative as a transcription. Either way I'm not seeing the relevance here, because the word "secondary" is implied by the context. The point of the prescription is that, in the quest for accuracy and authority, quotations need to be taken from as close to the original as possible. When it is not possible to use the primary source, we need to minimize the train of transmission. And when we're talking ancient "quotations", the primary source is actually the initial publication of the translation, and yes, we ought to get as close to that as possible. And again, this being on the internet, those primary translation sources are typically on-line. Mangoe (talk) 04:58, 18 January 2023 (UTC)

  @Mangoe, what makes you think that an accurate transcription of a document is a secondary source? From the article: "Secondary sources involve generalization, analysis, interpretation, or evaluation of the original information." It's not a matter of counting up the number of WP:LINKSINACHAIN. An accurate transcription of a primary source is still a primary source. WhatamIdoing (talk) 07:11, 20 January 2023 (UTC)

Saying that it's an accurate transcription is begging the question: you don't know it is accurate unless you compare it against a more authoritative copy. And if such is available, we ought to prefer it as our source. That's really what my original version of this guideline is about. Mangoe (talk) 04:32, 22 January 2023 (UTC)

@Mangoe, let's stipulate that the National Archives has put an accurate copy of the Declaration of Independence on its website. Their accurate copy isn't a secondary source. Reproducing a primary source results in a copy of a primary source. It does not result in a secondary source. WhatamIdoing (talk) 00:02, 7 February 2023 (UTC)

I'm going to try to re-phrase this in a way that makes the problem clearer. Here are two versions of the disputed sentence. These two sentences mean exactly the same thing.

• If this is not possible, then the text may be taken from a reliable secondary source.
• If this is not possible, then the text may be taken from a reliable source that provides generalization, analysis, interpretation, or evaluation of the original quotation.

@Peter Gulutzan and others: Would you be satisfied with the second version? Keep in mind that all I've done is replace the word secondary with the standard definition of what that word means.

If, as I have suggested, you believe that the key point is that we need a source that is unlikely to have screwed up, rather than a source that provides some sort of transformative scholarly comment on the quotation, then perhaps you will see why I think this word should be removed. WhatamIdoing (talk) 06:37, 9 February 2023 (UTC)

I guess you're thinking of WP:SECONDARY when you say "the standard definition of what that word means", while I admit I think of non-Wikipedia definition e.g. Oxford Learner's Dictionary "book or other source of information where the writer has taken the information from some other source and not collected it himself or herself", I don't believe that's the wrong thing to do when Wikipedia's description doesn't match an ordinary meaning that an ordinary user might know. But if it is, then the other argument, that secondary excludes tertiary, is still in your way. Peter Gulutzan (talk) 15:08, 9 February 2023 (UTC)

Actually, the line I quoted comes from the Wikipedia article, Secondary source. For Wikipedia's purposes, a secondary source is something that intellectually transforms one or more original ("primary") sources into a new thing. Merely repeating it secondhand is not enough; see WP:LINKSINACHAIN.

(It sounds like that dictionary may be engaging in the practice of lies to children; secondary is not identical to secondhand.) WhatamIdoing (talk) 01:53, 13 February 2023 (UTC)

The grownup version is about the same. OED entry for "secondary", 3a: "Belonging to the second order in a series related by successive derivation, causation, or dependence; derived from, based on, or dependent on something else which is primary; not original, derivative." I had to go down a long way till I saw something remotely resembling what you called standard: "Draft additions September 2007 ... In the context of academic research or writing: designating or relating to analytical or critical commentary on material which forms the primary subject of study; designating a text with another text as its subject. Frequently contrasted with primary adj. 17." Peter Gulutzan (talk) 15:57, 20 February 2023 (UTC)

A secondary source is not quite the same thing as a source that is secondary. WhatamIdoing (talk) 04:51, 26 February 2023 (UTC)

RfC: The elephant in the room: AI sources

As AI is becoming increasingly prevalent and beginning to be used by sources considered reliable, I think it is time to ask: are sources written by an AI reliable, regardless of publication? If an AI article is written by a reliable publication, does it differ from an article written by the same AI from a situational or unreliable publication? Does that differ from an average person using that same AI? What about articles only assisted by an AI? (see also: the ongoing RSN discussion of CNET). DecafPotato (talk) 20:38, 9 February 2023 (UTC)

No. This is obvious and does not need and RFC. Headbomb {t · c · p · b} 21:26, 9 February 2023 (UTC)

@Headbomb, is it obvious? Most cases of AI writing aren't the CNET "running wild with tons of errors" type. Most of them are edited to contain factual information. This is asking about AI articles in general on reliable sources, not "is an unedited response I got from ChatGPT reliable".

If consensus is obvious, close it early, sure, but to me, it is not. And not a single person had a chance to reply when you reverted it. DecafPotato (talk) 21:44, 9 February 2023 (UTC)

One, there has been no WP:RFCBEFORE that I can find. Two, it clearly falls into WP:UGC. --Redrose64 🌹 (talk) 23:47, 9 February 2023 (UTC)

Taking a look at the Verge report over in the discussion about CNET, this is what stands out to me regarding AI assisted sources: "...editing its work took much longer than editing a real staffer’s copy. The tool also had a tendency to write sentences that sounded plausible but were incorrect, and it was known to plagiarize language from the sources it was trained on..."

Based on that report, AI assisted writing would therefore fall under WP:GUNREL. If a souce is using a specific byline for its AI writing then that byline can be deprecated much as we already to with Forbes: Columns written Forbes contributors are considered unreliabile opinion pieces but articles composed by Forbes staff writiers are reliable. But if we cannot differentiate between when and where AI assisted writing is being used then the entire source comes into question for reliability. Blue Riband► 01:01, 10 February 2023 (UTC)

• Here’s a question… how do we know whether a source was written by an AI? Suppose, for the sake of argument, a here-to-fore reliable source (say the New York Times) uses an AI to write a news article… how would we know? Blueboar (talk) 01:50, 10 February 2023 (UTC)

  In a nutshell we can't know until a pattern emerges. Then there will need to be a discussion on whether or not the previously reliable source has become questionable after a certain date. Blue Riband► 13:55, 10 February 2023 (UTC)

Whether AI was used or not isn't central for whether or not a source is trustworthy. In most cases it's just one tool used in the process of creating a source.

For each source we should look at it's actual reliability. Unedited ChatGPT obviously is not reliable. ChristianKl ❪✉❫ 15:47, 10 February 2023 (UTC)

• I just added a subsection on LLMs. I understand WP:BRD, but if it gets reverted, I won't be able to advocate for its reinstatement; someone else will need to champion that, or push for their own improved version of it, since I barely have any time to contribute to Wikipedia these days (and likely for the next few weeks).

I'd just like to address one thing: as Axios reported, other WP:RS have experimented with LLMs, but I don't think it's a problem as long as it's used in minimal, judicious ways, with proper oversight. But I think the subsection I added will cause minimal collateral damage. DFlhb (talk) 18:38, 2 March 2023 (UTC)

I disagree with the embargo on IMDB

A heckofalot of articles here on WP are about movies. A heckofalot of information about movies is on IMDB, and I would have to suggest it is pretty reliable, and should be considered as such. I get the feeling that most people there do their work in good faith. As an editor, as a person, I have to go on good faith. Same as Find a Grave, which I also contribute to. Has WP actually studied IMDB or FG and documented how frequently they perpetuates lies? or what sort of political bias they are governed by? Heck, I still advise people to worry about WP: the new/old adage still applies: don't believe anything you read on the Internet. check the sources; read'em if you have to. Because the IMDB is a valuable resource, a way should be found to allow its use. Here's a suggestion: a little disclaimer - just like the omnipresent BLP disclaimers at the top of bio pages: "This page cites IMDB for some of the information presented herein. While we believe IMDB to be generally reliable, please use your own judgement when reading it. Seek other sources if you are in doubt." I say all this because I was reverted for trying to answer some questions about TV appearances by the actress Tracy Camilla Johns. I might simply have gone to the editor who asked for the additions to go to IMDB. It certainly would have been easier than spending 20 minutes on lookups. And on this complaint. Verne Equinox (talk) 02:42, 15 February 2023 (UTC)

@Verne Equinox: Have you read WP:IMDB? This was discussed and agreed some years ago. It's not likely to be overturned unless IMDb seriously change their policy regarding user-submitted content. --Redrose64 🌹 (talk) 19:33, 15 February 2023 (UTC)

It would take logic and intelligence to solve for Wikipedia and IMDB, you are looking at the wrong place. Editors favor themselves being right over including useful information. user-submitted content oh that gross, kill it with fire. 2601:248:C000:3F:9D65:D6B4:F90F:280E (talk) 22:01, 28 February 2023 (UTC)

One small example. I had to submit a change request to IMDB last year because they hadn't even got the title of the show correct. - X201 (talk) 22:19, 28 February 2023 (UTC)

No, IMDB isn't 'generally reliable'. It is full of nonsense placed their by people trying to promote themselves and/or whatever they are paid to plug. It utterly fails to meet WP:RS standards. AndyTheGrump (talk) 22:52, 28 February 2023 (UTC)

are death certificates and marriage certificates/licenses reliable sources?

are death certificates and marriage certificates/licenses reliable sources? Ladyluck58 (talk) 19:13, 24 February 2023 (UTC)

No, in most cases. See WP:BLPPRIMARY. Masem (t) 19:41, 24 February 2023 (UTC)

I guess it depends on what you're relying on them for. What could be more authoritative about a person's date of death than their official death certificate?

From my own experience in writing Margaret Sibella Brown#Death, I had sources differing on the spelling of her name and her date of death. I went with the death certificate for the date, but not for the spelling. I explained the discrepancies and why I went which way on each fact. That gives the reader all the information, and they can draw their own conclusions. -- RoySmith (talk) 15:25, 25 February 2023 (UTC)

When editors have used official birth or death certificates in the past, they have sometimes made the unpleasant discovery that they relied on the wrong person's certificate. They're probably reliable for a statement that "a" person named Margaret Brown died; they don't, however, tell you whether the Margaret Brown in the death certificate is the same Margaret Brown that you're writing an article about. WhatamIdoing (talk) 06:05, 26 February 2023 (UTC)

That. Gråbergs Gråa Sång (talk) 15:48, 12 March 2023 (UTC)

Chewing the citations

Over time I have noticed a tendency that I call "chewing the citations" (by analogy with chewing the scenery) where statements are introduced with an attribution to the cited source even though they are not evidently controversial or are not unique to the source. While it is important to cite the height of the Eiffel Tower, for example, this is not controversial and so there would be no need to say "According to a Reuters report…." When this is done repeatedly, it results in tedious he-said/she-said text that may suggest nonexistent controversy, or fails to identify the poles of controversy and can invite the inclusion of fringe views. I wonder if some comment about this might be added. 67.180.143.89 (talk) 19:38, 26 February 2023 (UTC)

This appears to be nonsense. Eiffel Tower, for example, doesn't hedge. It simply says, "The tower is 330 metres (1,083 ft) tall" and provides a citation. A citation absolutely should be provided in that case. Do you have specific examples of what you claim? If so, provide them. --Yamla (talk) 19:49, 26 February 2023 (UTC)

You are refuting assertions that I did not make. I have commented on the phenomenon before at the Jerome Powell article, and came across one today at the "Banking in Canada" article, but I have not kept a running dossier of examples to provide immediately upon demand. 67.180.143.89 (talk) 19:57, 26 February 2023 (UTC)

Please give a specific example. --Yamla (talk) 20:02, 26 February 2023 (UTC)

See Jerome Powell comments here and "Banking in Canada" example here. These are not the only examples I have come across and I doubt that either of us have such a comprehensive knowledge of WP's content to judge how extensive the practice might be. The policy page can be changed or not, but I did not imagine that cautioning against superfluous attribution of statements might be so objectionable that it would be promptly denounced as "nonsense." 67.180.143.89 (talk) 20:26, 26 February 2023 (UTC)

Liechtensteiner Volksblatt

The Liechtensteiner Volksblatt newspaper has gone out of business after 145 years. Every link (example) now redirects to Liechtensteiner Vaterland, the only other newspaper in the country. There are around 70-ish citations of this newspaper on en.wp, so archive links will need to be placed in. Curbon7 (talk) 23:56, 10 March 2023 (UTC)

Photographs

Many WP articles are descriptions of artworks, buildings, series of industrially produced vehicles and similar things. Actual photographs of such things can be verified by many visitors of WP in reality. That prevents falsifications. Historic photos are not less reliable than historic verbal descriptions. Why aren't photographs mentioned in this guideline as reliable sources?--Ulamm (talk) 09:40, 12 March 2023 (UTC)

@Ulamm: Have a look at the article Cottingley Fairies. --Redrose64 🌹 (talk) 11:08, 12 March 2023 (UTC)

WP:OR may be relevant, depending on context. Gråbergs Gråa Sång (talk) 15:41, 12 March 2023 (UTC)

Reply on the second answer at first: Original research of the intolerable kind are private records that cannot be verified. Measurements of physical values such as temperatures, pressures and flows at a certain time. The same thing is with chemical analyses.

The truth of photos of objects with public access that remain unchanged or almost unchanged for years can be controlled by the public, including the Wikipedia/Wikimedia communities.

This leads to my reply on the first answer.

There is an immense difference between photos of situations and photos of constant objects:

Photos of situations are always doubtful. The art of retouching is as old as the art of photography (last not least because just the first photographers wanted to fit some principles of presentations, people were accustomed to from paintings)

But, as I already wrote in my first post, photos of stable objects have a protection against alterations that violate the truth;

the publisher of an essentially altered photograph has to face the possibility that next day (or next week or next year) somebody else will publish a true photo of the same object.

Especially in WM Commons, the correct photo will be placed in the same category as the wrong one.

This way, the falsification can be unmasked much easier than a wrong description in a scientific manual, which is reviewed once in a decade (or more rarely).--Ulamm (talk) 17:35, 12 March 2023 (UTC)

"*Prefer* secondary sources" seems inaccurate

There may be instances where primary sources were acceptable. However in practice i observe that the applied rule is "primary sources are never acceptable". The wording "prefer" implies that the article writer gets to make a mostly autonomous choice on using primary sources, when that is unlikely to be permissible. The correct phrasing should be something like "Avoid primary sources almost always", and then be followed with documentation on when primary sources are even acceptable at all. Xenofur (talk) 15:54, 14 March 2023 (UTC)

There is a myth that primary sources are forbidden. We should not strengthen the myth by further entrenching it in the policy. Zero^talk 01:02, 15 March 2023 (UTC)

Question about “self-published” sources

So, what if a source has some, but not all, of their sources published? Does the publishing of some account for the self-publishing of others? QuicoleJR (talk) 23:58, 15 March 2023 (UTC)

Reliable yet profringe

I see we have some back and forth regarding a recent addition about profringe sources being unreliable. This is not always the case. Profringe sources CAN be reliable… yet excluded because discussing/citing them can be UNDUE. In other words, the addition is addressing a valid problem, but using the wrong policy to address it. Blueboar (talk) 14:07, 16 March 2023 (UTC)

No, you're wrong. The well-established consensus is that white supremacist sources by authors such as Arthur Jensen, Charles Murray and Heiner Rindermann are never RS regardless of who the publisher is. That's why guidelines such as WP:PARITY do not apply to these sources, and they are kept out of articles even when published in the same journals as their critics. This guideline needs to be updated to reflect the current consensus and standard editing practices. Sovkhozniki (talk) 14:35, 16 March 2023 (UTC)

Note: Sovkhozniki's suggested change appears to be related to this RFC: https://en.wikipedia.org/wiki/Talk:Eyferth_study#Request_for_comment_on_hereditarianism_subsection