Wikipedia:Wikipedia Signpost/2013-02-11/Op-ed


 * The views expressed in this op-ed are those of the author only; responses and critical commentary are invited in the comments section. The Signpost welcomes proposals for op-eds at our opinion desk.

Wikipedia gets quite a bit of press attention from drive-by vandalism, incoherent scribbles, rude gestures, and just plain page blanking perpetuated by Internet trolls and schoolchildren who take the site's free-to-edit model as an invitation to cause as much havoc as possible. The public perception that Wikipedia is riddled with errors and perpetually vandalized was a major retardant in the site's formative years, when it first engaged in its still-central battle for relevance and accuracy.

But this is a battle that, on the whole, Wikipedia has been winning for a lengthy amount of time. Years of nearly unchecked growth and explosive expansion have made Wikipedia not only the largest but also the most expansive information compendium the world has ever seen. Editing is tightly watched by users armed with tools like Twinkle, Huggle, rollback, semiprotection, and bots. Vandalism as we most commonly think of it is anything but dead—visible pages still regularly get as much as 50 percent of their edits reverted —but today's compendium of anti-vandalism tools have confined it in lesser form to the furthest and most overtaxed fringes of Wikipedia.

The dearth of vandalism lasting more than a few seconds has done much to improve our image. Five years ago, a project as enterprising as the Wikipedia Education Program could never have even existed, let alone thrived as it does today. The days when being a regular editor on Wikipedia was seen as unusual by others are slowly becoming more distant, its use ever more mainstream, and its editing body ever more academic. But another, subtler form of vandalism persists, and in the deterioration of its more visible cousin, may even be spreading—fabrication. Wikipedia has a long, daresay storied history with the spinning of yarns; our internal list documents 198 of the largest ones we have caught as of 4 January 2013. This op-ed will attempt to explain why.

It's frighteningly easy
Wikipedia's policy on vandalism is complex and extensive. Coming in at 41 KB, it is best remembered by the wrapper that adorns its introduction, stating that "Intentionally making abusive edits to Wikipedia will result in a block", a threat carried through more often than not. At just over 5k, the guideline on dealing with hoaxes is comparatively slim, and readily admits that "it has been tried, tested, and confirmed—it is indeed possible to insert hoaxes into Wikipedia". It is not hard to tell which is the more robust of the two policies.

First and foremost, this is a consequence of Wikipedia's transitional nature. The site has become mired somewhere between the free-for-all construction binge it once was, and the authoritarian, accuracy-driven project it is quickly becoming. The days of rapidly developing horizontal sprawl are long gone, swallowed up by the project's own growth; increasingly narrow redlink gaps and ever deeper vertical coverage are the new vogue, spearheaded by the plumping of standards and the creation of such initiatives as GLAM and the Education Initiative. Wikipedia gets better, but it also gets much more specialist in nature, and this has a major impact on its editing body. Explosive growth both in the number of articles and the number of editors, once the norm, has been superseded by a more than halved level of article creation and the declining number of active editors, both besides bullish, frankly unrealistic growth projections by the Wikimedia Foundation. The project has reached its saturation limit—put another way, there simply aren't enough new people out there with both the will and the smarts to sustain growth—and the result is that an increasingly small, specialized body of editors must curate an increasingly large, increasingly sophisticated project.

A sparser, more specialized editing body dealing with highly developed articles and centered mainly on depth has a harder time vetting edits than a larger, less centric one focused more on article creation. Take myself as an example: while I have the depth of field to make quality tweaks to Axial Seamount, I could never do as good a job fact-checking Battlecruiser as a Majestic Titan editor could, and I cannot even begin to comprehend what is going on at Infinite-dimensional holomorphy. This hasn't mattered much for pure vandalism: the specialization of tools has proved more than adequate to keep trollish edits at bay. But vetting tools have not been so well-improved; the best possible solution available, pending changes, has received a considerable amount of flak for various reasons, and has so far only been rolled out in extremely limited form. On pages not actively monitored by experienced editors, falsified information can and indeed does slide right through; with an ever-shrinking pool of editors tending to an ever growing pool of information, this problem will only get worse for the foreseeable future.

The relative decline in editor vetting capacity is paralleled by the ease with which falsehoods can be inserted into Wikipedia. Falsified encyclopedic content can exist in one of three states, by its potential to fool editors examining it: inserted without a reference, inserted under a legitimate (possibly offline) reference that doesn't actually support the content, and inserted under a spurious (generally offline) reference that doesn't actually exist. While unreferenced statements added to articles are often quickly removed or at least tagged with or, editors who aren't quite knowledgeable about the topic at hand passing over a page are extremely unlikely to check newly added references, even online ones, to make sure the information is legitimate. This is doubly true for citations to offline sources that don't even exist. Taking citations valeur faciale is standard operating procedure on Wikipedia: think of the number of times that you have followed a link through or looked up a paper or fired off an ISBN search to ascertain the credibility of a source in an article you are reading; for most of us, the answer is probably "not many". After all, we're here to write content, not to pore over other articles' sourcing, a tedious operation that most of us would rather not perform.

This is why complex falsifications can be taken further than mere insertions: they can achieve the kinds of quality standards that ought to speedily expel any such inaccuracies with great prejudice. The good article nominations process is staffed in large part by two parties: dedicated reviewers who are veterans of the process, and experienced bystanders who want to do something relatively novel and assist with the project's perennial backlog. In neither case are the editors necessarily taking up topic matters they are familiar with (most of the time they are not), and in neither case are the editors obligated to vet the sourcing of the article in question (they rarely do; otherwise who would bother? ), whatever the standards on verifiability may be. And when a featured article nomination is carried through without a contribution of content experts (entirely possible), or the falsification is something relatively innocent like a new quote, such articles may even scale the heights of the highest standard of all in Wikipedia, that much-worshiped bronze star! Nor are hoaxes necessarily limited to solitary pages; they can spread across Wikipedia, either through intentional insertions by the original vandal, or through the process of "organic synthesis"—the tendency of information to disseminate between pages on Wikipedia, either through copypaste or the addition of links.

Then why aren't we buried?
Readers of this op-ed may well take note of its alarmist tone, but they need not be worried: studies of Wikipedia have long shown that Wikipedia is very accurate, and, by derivation, that false information is statistically irrelevant. Well, if as I have striven to show manufacturing hoaxes on Wikipedia is so strikingly easy, why isn't a major problem?

Answering this question requires asking another one: who are vandals, anyway? The creation of effective, long-lasting hoaxes isn't a matter of shifting a few numbers; it requires an understanding of citations and referencing and the manufacture of references to sources, the positing of real intellectual effort into an activity only perpetuated by unsophisticated trolls and bored schoolchildren, and as it turns out the difficulties involved in making believable cases for their misinformation are a high wall for would-be vandals. And even when real hoaxes are made, studies have shown that Wikipedia is generally fairly effective (if not perfect) at keeping its information clean and rid of errors. Hoaxes have reached great prominence, true, but they are small in number, and they can be caught.

But there is nonetheless a lesson to be learned. Wikipedia is extremely vulnerable. If some sophisticated wash wants to launch a smear campaign on the site, falsification would be the way to do it; and that is something that should concern us. The continual unveiling and debunking of hoaxes long after they have been created is a drag on the project's credibility and on its welfare, and when news breaks out about hoaxes on the site in the media it takes a toll on our mainstream acceptance. This is not a problem that can be easily solved; but nor is it one that should be, as it is now, easily ignored.

Addendum: some highlights
Sorted by date of discovery, here is a selection of what I consider to be fifteen of the most impactful and notable hoaxes known to have existed on Wikipedia.
 * November 6, 2003 – February 23, 2004: Uqbar. One of the earliest hoaxes to have been debunked, the kingdom of Uqbar is a historical hoax (a story within a story) that was passed off as real early in Wikipedia's history.
 * December 2004 – April 2005: Roylee. A referral for comment on four months of activity from a user who "has carried out a sustained introduction of fringe theories and original research into a large number of articles (145 listed at User:Mark Dingemanse/Roylee [defunct]) since December 2004."
 * May 26 – September 22, 2005: Wikipedia biography controversy. To quote from the article: "a series of events that began in May 2005 with the anonymous posting of a hoax article ... about John Seigenthaler, a well-known American journalist. The article falsely stated that Seigenthaler had been a suspect in the assassinations of U.S. President John F. Kennedy and Attorney General Robert F. Kennedy. Then 78-year-old Seigenthaler, who had been a friend and aide to Robert Kennedy, characterized the Wikipedia entry about him as "Internet character assassination". The hoax was not discovered and corrected until September 2005... after the incident, Wikipedia co-founder Jimmy Wales stated that the encyclopedia had barred unregistered users from creating new articles."
 * October 5 – 26, 2005: Alan Mcilwraith. A former call center worker who created a new identity for himself as a decorated military man on Wikipedia, complete with an in-uniform portrait (now known to have been bought on eBay). The story hit headlines in April 2006, and the article was recreated—now about the hoax he perpetuated (see Signpost coverage).
 * ? – March 3, 2007: Essjay controversy. The only fabrication on Wikipedia major enough to have a 39k Good article to call all of its own, this was a hoax not in the classical sense—that is, not carried out across the mainspace—but in an extremely prominent editor's falsified credentials; when combined with a poorly timed promotion to ArbCom, the result was a spectacular fireworks display.
 * November 2005 – 21 June 2007: Baldock Beer Disaster. A disaster in more ways than one; the article appeared on the Main Page as a Did you know? entry on November 25, 2005, and was not rooted out until more than a year and a half later.
 * November 18 – December 18, 2008: Edward Owens hoax. A fisherman turned pirate who never really existed, created by students as part of a class exercise at George Mason University; now has its own article.
 * September 13–14, 2010: Roger Vinson. An addition was made claiming that the man in question, a federal judge in Florida, is an avid taxidermist who displays mounted bear heads in his courtroom. When Rush Limbaugh used this erroneous information on his talk show, it sparked a media reaction—a demonstration of how even relatively short-lived pieces of vandalism can be damaging.
 * Spring 2009 – October 2011: Cohen-Cruse Ruse. "A number of apparent sock puppets seem to be creating an elaborate set of fake pages around a few members of a "Cohen" and a "Cruse" family. It involved a number of completely (very carefully) faked biographies, other faked things (like synagogues) and a lot of associated edits to real pages that attempted to justify and contextualize those fake people." It lasted two years, and a major community clean-up followed.
 * ? – February 15, 2012: Legolas2186. Allegations of impropriety were brought against Legolas2186, a prolific (and supposedly trustworthy) writer with a large number of Madonna-related article credits to his name. As was eventually discovered, Legolas had been manufacturing sources, inventing information, and generally doing as he damn well pleased with his sourcing. A permanent ban and months of clean-up by the community followed (see Signpost coverage).
 * March 8, 2006 – March 21, 2012: [//en.wikipedia.org/w/index.php?title=Brierfield,_Lancashire&diff=42863927&oldid=35815037 Brierfield, Lancashire]. An addition was made claiming that the small town was the primary inspiration for Tolkien's Mordor. By the time it was [//en.wikipedia.org/w/index.php?title=Brierfield,_Lancashire&diff=483136210&oldid=483129402 removed] in March 2012, it had been on the page for a good six years.
 * June 9, 2004 – July 13, 2012: Gaius Flavius Antoninus. Created on June 9, 2004 and lasting eight years and one month before discovery, this purported assassin of Julius Caesar has the honor of being the longest-lasting hoax ever created on Wikipedia. Given the level of dissemination that happened in that time and the prominence of Caesar's (historically classical) assassination, it's also probably one of the most illustrative of the failings of Wikipedian vetting.
 * September 25 – November 19, 2012: Chen Fang. Chen Fang was the mayor of a small town in China, but he was also a student at an American university who created a fictional article about himself to make a statement about Wikipedian inaccuracy, and his case was cited in a Harvard University writing guideline on the topic. It took seven years and two months for someone to notice.
 * July 4, 2007 – January 28, 2013: Bicholim conflict. The primary inspiration for this op-ed, the Bicholim conflict is (was) one of the most complex and well-crafted hoaxes to have existed on Wikipedia, and spent half a decade, most of its life, as a supposedly verified Good article. A complete fabrication, in 4,500 words it described a clash between colonial Portugal and the Indian Maratha Empire in an undeclared war that supposedly helped cement Goa's independence (see Signpost coverage).
 * ? – February 1, 2013: Bonō Pusī Kalnapilis. A hoax created on our sister project, the German Wikipedia, that was not discovered to be a hoax until it was selected as a Did you know? entry, spending two hours on the main page before being caught.