The Signpost

News and notes

How many more hoaxes will Wikipedia find?

Contribute  —  
Share this
By The ed17 and Milowent

Another hoax on the English Wikipedia was uncovered this week—not by any thorough investigation, but through the self-disclosure of an anonymous change made when the editors were in their sophomore year of college. The deliberate misinformation had been in the article for over five years with plenty of individuals noticing, but not one suspected its authenticity. This leads to one obvious question: how many more are there?

Amelia Bedelia is a fictional character used by children's book author Peggy Parish and her nephew Herman Parish, who stepped in to continue the series after the former's death in 1988. Bedelia is over 50 years old and is literal-minded to the extreme. According to publisher HarperCollins, "When she makes a sponge cake, she puts in real sponges. When she weeds the garden, she replants the weeds. And when she pitches a tent, she throws it into the woods!" The New York Times Book Review noted that "No child can resist Amelia and her literal trips through the minefield of the English language—and no adult can fail to notice that she's usually right when she's wrong." Writer Cynthia Samuels continued:

However, Peggy Parish would likely be the first to tell her readers that her main character was not based on a maid in Cameroon.

Nor did she spend some "formative years" there.

Yet this is precisely what the Wikipedia article on Amelia Bedelia had said since January 2009: "Amelia Bedelia's character is based on a maid in Cameroon, where the author spent some time during her formative years. Her vast collection of hats, notorious for their extensive plumage, inspired Parish to write an assortment of tales based on her experiences in North Africa."

Related articles
News and notes

UK political editing; hoaxes; net neutrality
22 April 2015

Saving Wikipedia; Internet regulation; Thoreau quote hoax
15 April 2015

WikiWomen's History Month—meetups, blog posts, and "Inspire" grant-making campaign
11 March 2015

Gamergate; a Wiki hoax; Kanye West
11 March 2015

Monkey selfie, net neutrality, and hoaxes
13 August 2014

How many more hoaxes will Wikipedia find?
30 July 2014

Wikipedia's sexism; Yuri Gadyukin hoax
29 April 2013

An article is a construct – hoaxes and Wikipedia
11 February 2013

Hoaxes draw media attention; Sue Gardner's op-ed; Women of Wikipedia
28 January 2013

Rush Limbaugh falls for Wikipedia hoax, Public Policy Initiative, Nature cites Wikipedia
20 September 2010

Hoaxes in France and at university, Wikipedia used in Indian court, Is Wikipedia a cult?, and more
14 June 2010

Quote hoax replicated in traditional media, and more
11 May 2009

News and notes: Flagged Revisions and permissions proposals, hoax, milestones
10 January 2009

Media coverage of Wikipedia hoax results in article
17 April 2006

Hoax exposé prompts attempt to delete author
8 August 2005

Hoax articles on April Fool's rub some the wrong way
4 April 2005

Attempt to foist false article on Wikipedia revealed
14 February 2005

More articles

The hoax was only revealed when EJ Dickson, a journalist and one of the two original hoax editors, noticed a series of tweets including one from Jay Caspian Kang, an editor for the New Yorker, that highlighted the text Dickson wrote five years earlier. In her words, "It was total bullshit ... It was the kind of ridiculous, vaguely humorous prank stoned college students pull, without any expectation that anyone would ever take it seriously." Her co-editor Evan continued, "I feel like we sort of did it with the intention of seeing how fast it would take to get it taken down [by Wikipedia editors]".

Their edits were removed after Dickson publicized her edits in the Daily Dot.

Historical hoaxes

Hoaxes have a lengthy history on Wikipedia. The longest-lasting hoax was a two-sentence, obscure biography of Gaius Flavius Antoninus, who was supposedly a Roman politician who helped assassinate Julius Caesar in 44BCE.

At least 23 known hoaxes have lasted for five to six years, including an article on an equally obscure alleged war between Portugal and the Maratha Empire of modern-day India. Wikipedia editor A-b-a-a-a-a-a-a-b-a, who is now indefinitely blocked, wrote that this "Bicholim conflict" took place in 1640–41 and the resulting peace treaty played a major role in Portugal's keeping control over Goa until the 1960s. At the time it was exposed as a hoax, the meticulously created article had held good article status for five years. It was over 4300 words long, and had about 150 citations.

John Tyler, the tenth president of the United States, supposedly sent federal troops into Michigan in 1843 a secessionist movement spawned from an equally farcical Canadian-Michigander conflict over the Upper Peninsula of Michigan.

Numerous hoaxes have existed for shorter amounts of time. Among the most colorful was another painstakingly detailed entry on the Upper Peninsula War. Boasting 23 references in its bibliography, this fake article chronicled a struggle between the United States, Canada, and nascent separatists in Michigan spawned from a disputed territorial line in the Upper Peninsula. It ended with the massacre of numerous Canadian troops (along with 80–120 civilians suspected of being Canadian co-conspirators), and the arrest and execution of Michigan's governor.

This fantastical story turned out to be a success story for Wikipedia: the hoax, despite the effort that had been put into it, was caught, nominated for deletion within a week of creation, and disposed of.

What's left?

With this latest hoax revelation, how many more are out there? An op-ed published in the Signpost last year argued that studies show Wikipedia is very accurate and false information is near the level of statistical irrelevance. When hoaxes do occur, they "have reached great prominence, true, but they are small in number, and they can be caught." According to the author, "Wikipedia is generally fairly effective (if not perfect) at keeping its information clean and rid of errors."

Yet just by itself, the Bedelia hoax caused a number of others to be revealed in comment threads discussing the case, including false ghost stories and a new origin story for the corporate name Verizon. Dickson's article also referenced a prior hoax regarding the alleged inventor of S'mores; one of those claimed inventors even had their own biography article which was deleted last year, but not before being cited in a number of books. How many more remain hidden in plain sight?

Though not a defense, these problems of falling for false information are not new. John McIntrye, a copyeditor for the The Baltimore Sun and a noted critic of Wikipedia, also wrote about this latest hoax, and noted that those who were duped showed a "hardly novel" combination of laziness and gullibility, as demonstrated long ago by H.L. Mencken's 1917 Bathtub hoax.

Still, as EJ Dickson's article concluded, "I learned from my inadvertent Wikipedia hoax ... not that Wikipedia itself isn't reliable, but that ... many people believe it is." Numerous examples of Bedelia's alleged Cameroonian origins have been written about by scholars, bloggers, academics, and apparently even the current author of the series himself, who reportedly told a journalist in 2009 that the character was based on "a French colonial maid in Cameroon." The fact that these hoaxes are not caught for such a long time does not mean they cannot be caught—a discerning editor looking for questionable claims and lack of citations may spot them.

But the average reader using Wikipedia will likely not.

In brief

+ Add a comment

Discuss this story

These comments are automatically transcluded from this article's talk page. To follow comments, add the page to your watchlist. If your comment has not appeared here, you can try purging the cache.

The fallacy that "vandalism gets reverted immediately" is a long-standing one. It is why I keep a list of "Things that stayed too long" on my user page, as a salutary reminder to us all. There are steps we can take to catch much more of this sort of thing, by appropriate use of technology, but little appetite for doing it, it seems. All the best: Rich Farmbrough05:14, 2 August 2014 (UTC).

Generally speaking I find that obvious vandalism usually gets caught pretty quickly, where someone uses swears or the word "gay". But if it lasts five minutes (and thus drops off the bottom of Recent Changes), then it can last a long time. Likewise, subtle vandalism like the example provided here, which would need someone with a knowledge of the subject to identify. Lankiveil (speak to me) 06:26, 2 August 2014 (UTC).[reply]
Wouldn't the problem be mostly solved by focusing on IPs and accounts with only one edit? Do the vandal tools and bots already look for this? A lot (but not all) of sneaky vandals generally make one edit. Viriditas (talk) 09:26, 2 August 2014 (UTC)[reply]
I don't think so, unless it was also in a way that reduced the growth of the encyclopedia. IPs create a lot of article content (more than most accounts that I see), and my experience is that those creating convincing hoax entries do so with accounts with several edits. Of course, this can also make it easier to catch them, if you make sure to check all of someone's edits once you catch a single suspect entry. (talk) 15:58, 6 August 2014 (UTC)[reply]
Just last night, I came across an article with a section that has been clearly vandalized and only reverted after more than a month and 700+ views. This is just another downside of continuously losing long-term editors for past 6 years. How many of those 700+ views were stumbled upon by an editor who didn't pick up on the vandalism hint? Veteran editors would have spotted it in an instant. This change escaped recent changes patrol, bots and 6 page watchers (it's not a lot of page watchers, but better than a lot of stub articles who have 0 watchers). If you have more editors, they will catch these problems much quicker. Had I not stumbled upon this article and reverted the vandalism, I honestly don't know how long it will stay on the article until someone else picks that up. OhanaUnitedTalk page 20:00, 6 August 2014 (UTC)[reply]
What about applying Wikipedia:Pending changes to IP edits and accounts with less than, say 100 or 250 edits (new editors need guidance anyway). They can still edit, but the edits can be systematically reviewed. I don't want to discourage IPs like from contributing; hoaxes generally come from IPs or new accounts.--Milowenthasspoken 20:36, 6 August 2014 (UTC)[reply]

I've come across section blanking that stood for almost a year and bogus edits uncaught for even longer but it doesn't help when Jimbo himself feeds the bears by praising the "funny" vandalism. It only encourages more bears to come pawing through our campsite looking for tasty morsels of attention. - Dravecky (talk) 05:45, 2 August 2014 (UTC)[reply]

  • The mother of all hoaxes was the whole Siberian Wikipedia based on the non-existent Siberian language. To my best memory, it took nearly a year of a heated shutdown discussion to convince meta.wikimedia bureaucrats that it is a fake. Notably, it truthfully claimed to contain thousands of articles. I accidentally discovered that 95% were year articles and other not even stubs, probably generated by bot, and half of the rest was rabid Muscovitephobia. (Did I forget to mention it was 99% unreferenced?) It was shut down only after the perpetrator declared in some message board post that he got bored with the prank. The perpetrator herded numerous sock- and meat- puppets and was supported by some wiki-Russophobes, as well as by well-meaning "fellow travelers" — enthusiastic conlang fans. I was planning to write a "sign post" about this curiosity, but got distracted and forgot it, and now most of the traces faded away. And you are telling us here about subtle vandalism... I bet there are several thousand of deliberate bullshitting in areas nobody cares. -No.Altenmann >t 09:12, 2 August 2014 (UTC)[reply]
Wow, that's one debacle I've never had cause to look at. Ed [talk] [majestic titan] 17:59, 2 August 2014 (UTC)[reply]
  • See also the recent Chicken Korma case on Reddit: [1]. I think it is safe to assume that Wikipedia's demographics have a bearing on which hoaxes are likely to last longest (cf. different outcomes of the Bicholim conflict hoax, set in India, vs. the Upper Peninsula War, set in the US). Pending changes would help; the German Wikipedia has had its hoaxes, but generally fewer than the English one. Andreas JN[[Special:Contributions/Jayen466|466]Wikipedia] 10:12, 2 August 2014 (UTC)[reply]
  • That's because we have no sense of humor. And . Paradoctor (talk) 16:57, 2 August 2014 (UTC)[reply]
  • Earlier this week I stumbled across yet another hoax, which had lasted for more than 7 years. I would be extremely surprised if there were not many many more like this which have yet to be detected.
David Wilson (talk · cont) 10:43, 2 August 2014 (UTC)[reply]
  • Wouldn't it be possible to see how many pages (and which ones) have no watchers? I'd think those are the pages most likely to have false information. kosboot (talk) 01:34, 3 August 2014 (UTC)[reply]
  • That is correct, I cannot see that page. Can you see how many articles are listed on it?--Milowenthasspoken 02:32, 3 August 2014 (UTC)[reply]
  • Well, dusting off my sooper 37173 Admin powers, I looked at that page & found there are ... a lot. It shows the pages in alphabetical order, but only shows the first 2000, ending at 1975–76 Israeli League Cup‏‎. -- llywrch (talk) 06:07, 3 August 2014 (UTC)[reply]
    • User:Llywrch, don't you mean it shows the first 200, rather than the first 2,000? Andreas JN466 08:32, 3 August 2014 (UTC)[reply]
      • I selected the view by 500 option. I was able to view 4 pages in the category, which allowed me to view a total of 2000 articles. They were arranged in alphabetical order.

        Now I know that sometimes I express myself in a very verbose, rambling & awkward manner, so I'm not surprised at your question. However, it should be clear from my report that this tool is useless. Unless you are interested in protecting articles beginning with the word "1975". IIRC, I'm not the first Admin to point out how this tool is handicapped, so I assume this limit was done for performance reasons. But there is a pressing need for Wikipedians to learn which pages lack watchers -- probably more pressing than some of the software projects the Foundation have taken on. -- llywrch (talk) 17:52, 3 August 2014 (UTC)[reply]

        • Thanks, User:Llywrch. If we look at the alphabetical listing in Category:All articles needing cleanup, we find that the category contains 23,725 articles in total, and 1975–76 Israeli League Cup‏‎ would be the 58th article in that list. Extrapolating from this we can guess that the articles up to 1975–76 Israeli League Cup‏‎ in an alphabetical category listing of Wikipedia articles represent about 58/23,725 of the total x, and 58/23,725 = 2000/x. This gives us an estimate of well over 800,000 Wikipedia articles that no one is watching, or more than one in six. I guess we could do the same exercise with some other large categories that contain hundreds of thousands of articles, to see whether these yield similar estimates. Do you agree with the maths? Andreas JN466 23:41, 4 August 2014 (UTC)[reply]
          • I'm not 100% comfortable about using that category as a base for extrapolation, but I don't have a better one to suggest. (Many obvious ways to create large numbers of articles are limited to prevent an unsustainable load on the servers.) And it is likely accurate as to the magnitude: the number is doubtlessly somewhere between 100,000 & one million. Until someone comes up with a better way to arrive at the number, it'll do. -- llywrch (talk) 16:36, 5 August 2014 (UTC)[reply]
  • One week later... "Welcome to Wikipedia, the free encyclopedia that anyone can edit. Now with 73,296 articles in English!" - Dravecky (talk) 07:09, 4 August 2014 (UTC)[reply]
  • The percentage of articles with hoaxes is no doubt small compared to the entire size of the project. From July 2010-September 2011 the WP:UBLPR project reviewed every BLP tagged as unreferenced (I wish I could remember the number; it was over 20,000 articles) - we did find some hoaxes, but very very few percentage-wise.--Milowenthasspoken 12:27, 4 August 2014 (UTC)[reply]
  • PROD is hopelessly clogged with crap and should not be the suggested solution to any problem. Any system that fails deadly is, by definition, broken. And who cares anyway? There is no problem to solve here. Graffiti isn't a hoax in the first place, and every single statement in the entire corpus is an act of faith anyway. And that's a good thing. Maury Markowitz (talk) 18:11, 4 August 2014 (UTC)[reply]
  • Considering how long it took just to get support to replace the April fools front page "jokes" with misleading but technically correct "jokes," the number of narcissistic editors who think it's funny to mislead people from cultures who don't celebrate April fools day and who can't afford expensive encyclopedic sources, are pretty high. Olive — Preceding unsigned comment added by (talk) 06:29, 4 August 2014 (UTC)[reply]
Considering the kind of abuse and trolling that I received on my talk page over briefly blocking the IP that created this hoax and semi-protection of the article, I feel vindicated. Yes, the amount of hoaxes is small compared to the size of the project. But trolls, vandals, and pro se defamation plaintiffs don't understand the meaning of the word compared, as in "compared to what?" I don't know what the ultimate remedy is -- certainly I would not mass-prod poorly-sourced articles -- but we need to take it more seriously. In the meanwhile, brief blocking of IPs and semi-portection are the only tools we have. I'd love to see a bot that could search out for some more obvious vandalism. Bearian (talk) 17:17, 4 August 2014 (UTC)[reply]
You mean, like ClueBot NG? Paradoctor (talk) 18:05, 4 August 2014 (UTC)[reply]
Bearian, you're putting the cart before the horse. You simply state that such and such a problem exists and it has to be fixed. Really? What is it that needs to be remedied, exactly? What is the scale of the problem, in numbers? Why do we need to take "it" more seriously? You don't propose any sort of metric to any of these obvious questions. Maybe we should start there before jumping to conclusions? Maury Markowitz (talk) 18:17, 4 August 2014 (UTC)[reply]
This is a good discussion, but first of all, IMO it is in a wrong place. If the problem is serious it must be discussed in an appropriate WP venue (WT:V? Otherwise it is just keystrokes straight into bit bucket. Now quick answers: what? - (a) false info is bad (b) reputation. in numbers: (a) probably tiny (b) large: lost of noise is generated, high visibility for tiniest of tiny fraction of hoaxes detected by non-wikipedians. Why? - Perfectionism? Why we bother with FA/GA/, etc.? In any case, the first step in an approach to any problem is to decide whether the effort is manageable. Solution: a team of of bots one tags every added unreferenced statement with {{cn}}, another removes all bot-tagged pieces in, say, a week passes, third one watches tag not removed without ref added. Problem: vandals will find a way; eg., adding fake refs. Theory: This is but a special case of the problem of database integrity. Judging by a primitive implementation of whatever DB handles interwiki now, the problem of wikipedia integrity will not be solved any time soon. 'New buzzword: Wikipedia:Integrity (WP:INTERGITY): is anybody willing to jump-start it? Staszek Lem (talk) 19:19, 4 August 2014 (UTC)[reply]
  • Superb points Staszek. But let's not forget that the difference between medicine and poison is the dosage. Given plummeting participation, largely due (IMHO and being guilty of exactly the same thing) to rampant perfectionism, I'd hate to burn down the Wiki to save the Wiki. And I strongly believe this is a far more dangerous problem than minor graffiti in an unimportant article. Maury Markowitz (talk) 13:30, 5 August 2014 (UTC)[reply]
I've always felt that one way to cut down on vandalism is to allow editing only by registered users. Sure, determined vandals will register, but the point is requiring registration will still reduce the number of vandals. I know the majority disagrees. -- kosboot (talk) 19:41, 4 August 2014 (UTC)[reply]
The topic is not vandalism, but hoaxes. The motivation to perpetrate a hoax is higher than to simply add "asshole" into text, so that registering an account is not really an obstacle for a "true" hoaxer. Staszek Lem (talk) 20:01, 4 August 2014 (UTC)[reply]
  • If all IP edits, and edits from new users with less than 250 edits (who need guidance even when acting in good faith!), were subject to Pending Changes, quality would be improved, and new hoaxes minimized.--Milowenthasspoken 02:28, 5 August 2014 (UTC)[reply]
  • I think this is the sort of thing that is a realistic improvement. However, as someone that sometimes edits from an IP because I'm on my phone or at someone's house, let's put a little effort into being able to link back to your account after you realize your edits will otherwise be anon. I wouldn't want to have to do anything but click a button before seeing that edit. Maury Markowitz (talk) 13:30, 5 August 2014 (UTC)[reply]
@Maury Markowitz:, I think the simplest solution is to consider the harm a hoax might do. If the Foundation publicly reminds everyone that as a public carrier, it is not legally responsible for the content of Wikipedia, & announces it will not assist in the legal defense of anyone who adds a hoax that defames someone, that will limit the harm as much as can be done. That is, beyond the usual browsing of articles & dealing with any questionable material. -- llywrch (talk) 16:43, 5 August 2014 (UTC)[reply]
Indeed! This seems like the best suggestion among many excellent ones. Maury Markowitz (talk) 16:47, 5 August 2014 (UTC)[reply]
  • Should hoax editors be sanctioned? Should those who step forward after the fact be cut some slack? How about those whose hoaxes are long in the past? Reasons for sanctioning: 1) It sends a message that hoaxing Wikipedia is a bad thing, and 2) should anyone who perpetrated a hoax be able to have a "clean" Wiki-reputation? Reasons for not sanctioning for obviously-ex/reformed hoaxters: 1) If the hoaxters are trolls masquerading as reformed hoaxters, it just feeds them, and 2) if they truly are reformed, it would be WP:POINTy to sanction them especially years later. Sometimes forgiveness[2] is the order of the day. My personal take - if the person hasn't been disruptive recently, let the past be the past, but allow any "buried" past to be resurrected if the person becomes disruptive in the future. This doesn't have to be specific and it doesn't have to even name the account or IP address, but it should be clear enough to indicate the scope of the damage. For example, an administrator can step in and say "in 2006, this editor created a hoax which stood for 1-2 years before being corrected. There was some off-Wiki publicity which arguably damaged Wikipedia's reputation in a minor way" rather than "On March 2, 2006 at 13:45:56 UTC, the editor used the IP address to edit Foobar. This edit stood until April 3, 2007 at 15:06:07. It was the subject of discussion at Slashdot. The resulting publicity arguably hurt Wikipedia's reputation." Recommend but don't necessarily require that the person self-disclose the broad outline of any hoaxes he's done in the past and what he's learned since then if he asks for any administrator-assigned user-rights that normally require trust (e.g. template-editor rights) or if he becomes heavily involved in an area of Wikipedia where people look to him for wisdom and advice (e.g. some project-space areas). Very strongly recommend or even require that he make a similar disclosure if he is the subject of a community discussion and the past or the perception that he is hiding the past is likely to be considered relevant by at least some discussion participants. The latter can typically be done as a positive: "In 2006, I perpetrated a hoax on Wikipedia. Since then, I've learned that this was a bad idea because .... I've also learned that .... and I strongly recommend against perpetrating hoaxes on Wikipedia for these reasons and because ...." Where sanctions are warranted against active hoaxters, they should be done pretty much as they are now against any disruptive editor: The more often the disruptive behavior is repeated and/or the more likely it looks like a "disruption-only account" or an editor who is "not here to build an encyclopedia" the more severe the sanction, the more likely it is that this is a one-off thing and that the behavior will not be repeated, the less severe the sanction or warning. davidwr/(talk)/(contribs) 18:42, 5 August 2014 (UTC)[reply]


The Signpost · written by many · served by Sinepost V0.9 · 🄯 CC-BY-SA 4.0