A Wikipedia WIthout Graffiti

A Wikipedia WIthout Graffiti 290

Posted by CmdrTaco on Wednesday February 07, 2007 @01:00PM from the dog-without-a-bone dept.

Frequent Slashdot Contributor Bennett Haselton writes "I'm a Wikipedia junkie. There's nothing more fun than switching back and forth between reading about the history of human evolution, and following the latest speculation about the identity of the mysterious R.A.B. in the Harry Potter books, and Wikipedia is the best site to find it all in one place. But as a fan, it's always been frustrating for me knowing that Wikipedia could never improve beyond a certain point -- as it becomes more popular, it becomes more tempting to vandalize, and in turn becomes less reliable, a point that many have made already. That's why I'm excited that sites like Citizendium are approaching the same problem with a different model, one that could enable them to become what Wikipedia almost was, but which its intrinsic nature kept it from being: a central, reliable source of freely redistributable information about almost anything. The main difference is that Citizendium articles, after initially being built up through the same collaborative process that Wikipedia uses, will go into an editor-approved stage, at which point an editor (publicly identifiable on the article's history page) signs off on the accuracy of the article, and further edits also have to be approved by an editor."

Editor control over articles is controversial within the "radical collaboration" community; the Wikimedia foundation lists five "foundation issues that are essentially beyond debate", which includes "Ability of anyone to edit articles without registering". (In practice there are some safeguards in place to protect articles that are frequent targets of vandalism, like the George W. Bush entry.) But I'm fanatically results-oriented in my thinking, and I always ask: What are the purposes of this project, and how does this feature help achieve those purposes? It seems to me that a free online encyclopedia fills four main needs:

A source of information about pop culture that can be fun to read even without being 100% sure that it's accurate (like who R.A.B. is)
A source of information that can be freely and legally redistributed, e.g. by printing out copies for a class to read
A source of information on subjects where you need to be close to 100% certain that the information is reliable -- at least as certain, say, as you would be if you read the same fact in several books
A source of information that you can cite in a school paper as being reasonably authoritative and reliable

Anonymous authorship and real-time edits don't affect #1 or #2, and they actually hinder #3 and #4. Citizendium founder (and former Wikipedia board member) Larry Sanger said in a comment for this article, "There is no reason that [projects like Citizendium] even need be collaborative. What we're mainly interested in is reliable, independent/neutral, and free information." Such a statement -- "no reason that they even need to be collaborative" -- may be regarded by some Wikipedia devotees as heresy, but I think it hits the nail on the head. The purpose of such a project is defined by the quality of the information it produces. Collaboration is a possible means to that end, but collaboration itself is not the point.

For the reliability problem, I can't improve on this priceless sentence from Wikipedia's own "Citing Wikipedia" page:

For many purposes, but particularly in academia, Wikipedia may not be considered an acceptable source. ^{[ citation needed ]}

Wikipedia has actually done much better than I would have expected -- a study done in 2005 found that Wikipedia averaged about 4 errors per article compared to Britannica's 3, which is pretty good for a site where anybody can write that Columbus sailed to the New World in ships named the Ninja, the Pinto, and the Santa Fe. But for a site that harnesses the efforts of volunteers all over the world, I think the goal should be to surpass what has been done before, not just to tie with Britannica. And even if Wikipedia's error rate someday beats Britannica's, under its current model Wikipedia can never have the key property that Britannica has, which is that you can cite it as an authoritative source without sounding silly.

Citizendium's model of editor-approved articles, and editor approval of further edits to those articles, can help to achieve the benefits of collaboration, harnessing the efforts of volunteers, without falling into Wikipedia's traps. Assuming you can verify an editor's credentials (and we'll get to this in a minute), having an editor manage an article means two things: (a) you know the page wasn't vandalized in the last five minutes, and (b) you ought to be able to cite the work as a reference in a paper if your teacher isn't a total Luddite and you can explain to them how Citizendium works. Meanwhile, volunteers can still contribute without their own credentials being checked out; they can write as much as they want for an editor-approved article, as long as it's approved by the editor before going live.

There are still loopholes, of course. Currently Citizendium asks people to edit under their real name, but says that "we will use the honor principle to begin with", so anyone could claim to be a professor or a lunar astronaut. But the key words are "to begin with"; the difference between Wikipedia and Citizendium is that Citizendium views this as a loophole and not an intrinsic "community value", and loopholes can be fixed. To make the reliability as airtight as possible, I hope that Citizendium will eventually implement some sort of verification system, such as checking a professor's contact information on a Web page in the "faculty" section of an .edu Web server. I'm not instinctively thrilled by the thought of checking out volunteers' contact information, but it seems like the only way to achieve goals #3 and #4 above, so if it's as simple as sending a verification e-mail to an .edu address, that's a lot of gain for little effort. (Remember, this only has to be done for editors who sign off on articles, not for all volunteers. A non-editor volunteer could still ask to have their credentials checked out, so that they can be cited by their real name in the "end credits" of an article that lists volunteer contributors. But impersonation among regular volunteers is not likely to be a problem, since the editorial approval process ensures that only value-adding edits will be allowed, and it's unlikely that Alice would pretend to be Bob so that Bob can take all the glory of Alice's contributions to the project!)

Besides verifying authors' credentials, the one change that I hope Citizendium considers in the future is to give authors and editors credit at the top of each article -- or, for articles with many contributors, perhaps editors would be listed at the top and the "end credits" would list all contributors, on a separate page if necessary. This is because credited authorship for an article can help improve the article's usefulness in two ways -- the article can be cited as a reliable source, and the "name up in lights" factor rewards people for contributing more and better articles. Having authors listed only on the history page of an article, as they are in the current model, achieves the credibility benefit but not the "name up in lights" benefit. Larry Sanger suggested that having authors listed at the top of each article might put off readers from submitting edits -- if an article is perceived as being "owned", then others might feel like it's rude for them to change it. For me personally, this could go either way -- on the one hand, I might not realize that I was welcome to edit an article, but on the other hand, I think I might be more inclined to submit edits if I knew there was an editor in charge to keep someone else from frivolously overwriting my edits later. But in any case, to address this problem, each article could carry a banner at the top saying "Readers are encouraged to submit edits and other suggestions", and each paragraph could be accompanied by an "Edit" link, similar to Wikipedia (except that edits would go into a queue to be reviewed by the editor instead of going live). This would address the ownership-intimidation problem without taking away from the "name up in lights" factor. Sanger says that the Digital Universe Encyclopedia -- comprising the Encyclopedia of Earth and an Encyclopedia of the Cosmos, under development -- has plans to join with Citizendium and will use the credited-author model on their version of the site.

You might say that editors having their "name up in lights" would be an ego thing for editors, and I think you'd be right -- but I don't think this would be a bad thing, inasmuch as ego would motivate more people to become editors and do their best work. Perhaps I'd be wrong about this. Maybe a limited experiment could be carried out with two sites that are similar in every respect except that one allows editors and authors to take credit for their work, as might turn out to be the case with Citizendium and Encyclopedia of Earth. The point is that I don't think such a suggestion should be judged by whether it goes against the "spirit" of the project (as it certainly does in the case of Wikipedia!), but rather whether it helps to achieve the projects goals, such as goals #1 through #4 listed above.

There are still some problems that Citizendium's differences from Wikipedia won't solve. Many schools discourage citing Wikipedia not because it's written anonymously or because it contains errors, but because it's an encyclopedia. Yale's guidelines for citing Wikipedia state:

As an encyclopedia, Wikipedia is written for a common readership. But students in Yale courses are already consulting primary materials and learning from experts in the discipline. In this context, to rely on Wikipedia -- even when the material is accurate -- is to position your work as inexpert and immature.

Presumably many academics would have the same objections to a student citing Citizendium. I understand what these teachers mean, but I think this is a case of not thinking in terms of results. If the purpose of an assignment is to collect and present information, then any means of accomplishing that goal should be valid, including the easiest method of looking up the information in an encyclopedia. To make a student look beyond the encyclopedia, an assignment can simply require depth of research that goes beyond what the encyclopedia would provide. (Students, if you're worried that your teacher will take this to heart and make your assignments harder, just be happy that your teacher is hip enough to be reading this in the first place.) Some things are hard, but they should only be hard if they're intrinsically hard, not because you handicapped yourself with arbitrary rules.

But there is another, more permanent problem -- even with verification of authors' credentials, how do we know that the information in Citizendium articles is accurate? How do we know the author didn't make a mistake, or lie? This gets into deeper issues because these problems exist no matter what source you're consulting. There are books in print that deny the Holocaust or the possibility of evolution, and they're printed on real paper, with ISBN numbers and everything. Some of them even make it into libraries. How skeptical should we be of we read in books? In January two advocacy groups presented a report to Congress in which many government scientists said they felt pressured by the Bush administration to downplay the global warming threat in their statements. Does that mean statements from government scientists are inherently suspect?

And almost anyone who has had more than two articles written about them, knows the feeling of reading the article and reacting, "Wow, I had no idea that I was a transgendered NRA member who volunteers with the Moonies!" The New York Times is hosting an article about me from 2000 claiming that I was fired from Microsoft, when I actually quit. I showed them a copy of my personnel file with "Voluntary resignation" printed on it, but they have still refused to change the article. (When I first wrote to the paper's "Public Editor" about the matter, created to restore "reader credibility" after the Jayson Blair scandal, they replied that they wouldn't change the error because it never appeared in the print version of the paper. Huh?) I put up my own webpage to tell my side of the story, but if you were a Wikipedia or Citizendium editor and you had conflicting information from different sources, who would you believe, the New York Times, or a Web site called PublicEditorMyAss.com?

And yet, I freely admit that even today, I would trust a fact from the New York Times more than a fact from Bob's Bait And Tackle Shop And Technology Blog. We instinctively trust sources because of their reputation; we figure that they must have gotten their reputation somehow. This is not a great algorithm for deciding trustworthiness, but it may be the best that we can do -- in a world where we can't verify every fact firsthand, what choice do we have but to rely on sources that have provided mostly-reliable information in the past? (Wikipedia vandals are able to hack this mental algorithm because we think of Wikipedia as "one source" with a high average reliability, when it's really comprised of many sources, some of whom are deliberately less reliable than others.)

So, I think the Citizendium model is a move in the right direction -- taking into account the limits of what we can know from third-party sources, and doing the best we can within those limits. The least we can do is to know who has signed off on the accuracy of an article, so we can factor that into our decision to trust it. Last month Citizendium released their first editor-approved article, a single article about Biology. It may not look like anything revolutionary right now, but the difference between that and the Wikipedia entry is that you can't change the title of the Citizendium article to LARRY SANGER IS A BUTT BRAIN HA HA. You have to go through an editor for that.

This discussion has been archived. No new comments can be posted.