Wikipedia's Accuracy Compared to Britannica

Raul654 writes "Nature magazine recently conducted a head-to-head competition between Wikipedia and Britannica, having experts compare 42 science-related articles. The result was that Wikipedia had about 4 errors per article, while Britannica had about 3. However, a pair of endevouring Wikipedians dug a little deeper and discovered that the Wikipedia articles in the sample were, on average, 2.6 times longer than Britannica's - meaning Wikipedia has an error rate far less than Britannica's." Interesting, considering some past claims. Story available on the BBC as well.

Comments

Dooop

[MullerMn]

Slashdot Article Compared to Earlier Slashback: Found To Be Identical

Story available here.

Re:Dooop

[Prospero's Grue]

Slashdot Article Compared to Earlier Slashback: Found To Be Identical

Yeah, but the Slashdot Article is 1.4 times longer, so it's not as duped as you think...

Re:Dooop

[Anonymous Coward]

Please it took like me 2 seconds to find slander on Wikipedia...

Here this was up just yesterday and was just taken takendown. YES it was up on the web for a while before being noticed. I think the point is it should not have been up AT ALL. There is nothing inpressive in how long or how fast something slanderous and stupid was caught. Without Wiki it WOULD NOT HAVE BEEN UP AT ALL.

Under the rock group Dokken.

In 2005, Don Dokken and Jani Lane of a band called Warrant participated in a civil union ceremony to declare their love for one another.

Yeah yeah its funny and you can lie and Ad Hom away but this is what people see. No matter how much /dotters shout them down and browbeat them.

Other Encyclopedias don't have problems, anywhere even remotely close to Wiki with its slander and information athentication WARS.

I know, on dense heads and deaf ears. Wasting my time..

Re:Dooop

[ceejayoz]

Here this was up just yesterday and was just taken takendown.

So you left slander up on the Internet when you could easily have removed it? You're part of the problem!

Without Wiki it WOULD NOT HAVE BEEN UP AT ALL.

And neither would much of the useful content.

Other Encyclopedias don't have problems, anywhere even remotely close to Wiki with its slander and information athentication WARS.

Other encyclopedias don't have much of the more obscure information available in Wikipedia.

Re:Dooop

[Patik]

There is nothing inpressive in how long or how fast something slanderous and stupid was caught. Without Wiki it WOULD NOT HAVE BEEN UP AT ALL.

If you had a system where changes and additions had to be approved by other users before being applied to an article you would still get slander but in a different form -- slanderous people and trolls would simply watch the "waiting for approval" list and deny legitimate submissions while allowing their troll friends' slanderous submissions. Plus you'd have to worry about people making changes to an article that will be very different once a previously-made change gets approval, which would cause quite a headache.

Re:Dooop

[timster]

When I was in junior high, people used to slip pages from porn into the encyclopedia volumes in the school library. An annoyance to be sure, but until our society gains the sense to lock teenagers up in solitary confinement, we will still encounter stupid pranks from time to time.

Re:Dooop

[oleksaa]

I think you just gotta keep the fact that wikipedia is written by other people in mind when you are doing research. It is still a valuable tool. It may not always provide a complete account of a subject but usually gives a good overview and at the very least provides you with some common terminology related to the subject. I'm a CS grad. student and I use Wikipedia as a starting point for my research all the time. If i'm not at all familiar with a subject, I go there, get some keywords, and then use those to look around on the internet(usually google scholar, which is freakin awesome). Therefore, I think Wikipedia should never be used as a sole reference. Then again, if all of your information comes from any one place you're in trouble so that shouldn't be a surprise.

Not exactly

[bersl2]

However, a pair of endevouring Wikipedians dug a little deeper and discovered that the Wikipedia articles in the sample were, on average, 2.6 times longer than Britannica's - meaning Wikipedia has an error rate far less than Britannica's.

That part's new.

Re:Not exactly

[irote]

And it's also nonsense. The Wikipedia article is written flabbily, by a collection of authors, some experts, some not, some good writers, some terrible ones.

The Britannica, on the other hand, is written by someone with clear credentials as an expert, to a word limit, and is then edited for conciseness and clarity. That is to say, the Britannica piece will undoubtedly say more than the Wikipedia piece. The error per word rate in Britannica may be higher, but the error per fact rate is probably much more favourable to Britannica.

Easy example - compare the writing in a mainstream newspaper to a well-written one with tight editorial policies, like the Financial Times or the Economist. Your average Sidney Morning Herald, Guardian or San Francisco Chroncile article is probably longer, but it says less.

Re:Not exactly

[croddy]

So, what you're saying is that Britannica has a long way to go before it will be useful as a wiki?

Re:Not exactly

[Jugalator]

The Wikipedia article is written flabbily, by a collection of authors, some experts, some not, some good writers, some terrible ones.

Yes, and terrible contributions gets edited over time as the article stabilizes.

The error per word rate in Britannica may be higher, but the error per fact rate is probably much more favourable to Britannica.

So you have no idea or basis for this claim?

Easy example - compare the writing in a mainstream newspaper to a well-written one with tight editorial policies, like the Financial Times or the Economist. Your average Sidney Morning Herald, Guardian or San Francisco Chroncile article is probably longer, but it says less.

I don't know about you, but from the articles I've seen on Wikipedia, they've been quite rich in information.

Re:Not exactly

[irote]

What's the content unit? The fact or the word?

As you say, the quality of writing is not what's being examined. We turn to an encyclopedia, whether printed or online, for facts.

For this reason, it's the accuracy of these facts that is of interest to us.

Accept the (indubitably true) proposition that the fact-to-word ratio in Britannica is higher than in Wikipedia, then the submitter's 'argument' is false: dividing the length of an article by the number of errors in it does not give you an average error rate.

A word is neither true nor false, a statement can be.

Re:Not exactly

[Anonymous Coward]

The error per word rate in Britannica may be higher, but the error per fact rate is probably much more favourable to Britannica.

So you have no idea or basis for this claim?

He read it in Wikipedia, right after he wrote it.

Re:Not exactly

[irote]

I agree, they can be rich in information, no argument at all. I frequently use it as a resource.

My quibble is with the submitter's argument that the error per word ratio in Wikipedia is lower than in Britannica. I say this is meaningless: we're interested in the error per statement ratio.

Re:Not exactly

[CastrTroy]

Yeah, but you're paying for britannica. I'd really expect them to have less than 3 errors per article. Wikipedia is a free enclopedia by the people, for the people. It will get better if the community gets bigger. There's a lot of stuff you'll find in wikipedia that you won't find in britannica, because people can write about whatever they want.

Re:Not exactly

[iabervon]

That is to say, the Britannica piece will undoubtedly say more than the Wikipedia piece.

That's not actually true. Wikipedia's threshold for relevance is lower, so the articles say more, in addition to being less densely written. This is due, to a large extent, because Britannica has to print theirs, so they have pressure to keep things brief, whereas Wikipedia can go into lots of detail. I don't have access to Britannica, but I'm willing to bet that it doesn't explain the Reed-Solomon configuration for error correction on CDs. So chances as that Wikipedia articles have more information in them, although not by as big a factor as the increase in size. Of course, there's no way for us to know at this point the characteristics of the articles that Nature used for this comparison, because they seem to have merged related articles in both cases. For example, most of the content of the Wikipedia "Field Effect Transistor" is in the articles on particular types (MOSFET, JFET, etc.), and the article on Woodward in Britannica must have gotten sections from other articles (e.g., overviews of things he worked on) pulled in if Nature compared versions of remotely similar lengths or scope, since Britannica doesn't break up this topic into articles the same way.

Re:Not exactly

[pizzaman100]

Yeah, but you're paying for britannica. I'd really expect them to have less than 3 errors per article.

Note that study only picked 42 science articles. This does not mean that britannica has that rate of errors for other diciplines.

Re:Not exactly

[Haeleth]

Note that study only picked 42 science articles. This does not mean that britannica has that rate of errors for other diciplines.

Note also that science is one of the few disciplines in which, by and large, there tends only to be one current scientific theory which is clearly stated and can easily be summarised encyclopedically. It is much, much, much easier to write a good article on the quark than on Hamlet.

And as you'd expect, Wikipedia's article on Hamlet is not very good. On the positive side, I can actually read it - unlike the Britannica article, which it seems you have to pay for. (They did let me read the first paragraph; it starts by presenting a contested theory as though it were an undisputable fact, an unimpressive opening that does not encourage me to part with $69.95 to see the rest.)

Re:Not exactly

[Haeleth]

Accept the (indubitably true) proposition

Your use of language is as careless as that you attribute to Wikipedia's editors. No proposition is "indubitably true", and no proposition can be proven by asserting its truth without providing any sort of argument to support the assertion.

It is plausible that Britannica presents facts more concisely. It is even likely. But unless someone actually

• Defines a "fact", in the context of an encyclopedia article, in an objective and measurable way;
• Devises a methodology for assessing the ratio of facts (thus defined) to words;
• Applies this methodology to a statistically significant selection of articles from Wikipedia;
• Applies the same methodology to a comparable set of articles from Britannica; and
• Publishes their definitions, methodology, and results,

then you simply can not describe the proposition as "true". And even if such a study existed, you would have to be pretty damn sure that its methodology was unassailable before you could consider describing the proposition it supported as "indubitably true".

Re:Not exactly

[EpsCylonB]

They should compare only articles of a certain age in wikipedia with the brittanica articles, and myabe wikipedia should warn if an article is either new or been hardly accessed.

Re:Not exactly

[flyingsquid]

I don't know about you, but from the articles I've seen on Wikipedia, they've been quite rich in information.

Of course, there's the issue of the type of information. Wikipedia has a dissertation-length discussions of Half-Life 2 and Babylon 5, for instance, and a meager couple screens devoted to Moby Dick (unless you count the discussions of Moby Dick's influences in Star Trek episodes, Japanese video games and comic books as a serious discussion of the novel).

Though I suppose you could make the argument that this is actually a strength rather than a weakness. Moby Dick may be a masterwork of American fiction, but today, video games and sci-fi soap operas have a vastly greater cultural influence than Herman Melville.

Re:Not exactly

[fbjon]

It really seems that Wikipedia is an good encyclopedia of things that actually matter to most people.

Re:Not exactly

[jdavidb]

I count five screens of information, not counting the "Selected adaptations and references" section, which certainly references more than Star Trek. Meanwhile, I went searching on britannica.com and found that there was no article at all on Moby Dick. There was an article on Herman Melville, though. It's 2845 words long. I admit that beats Wikipedia's, which is 883 not counting the bibliography. But combined with the text of the Moby Dick entry, that's 2672 words total, again not counting the (not just Star Trek) references section.

Re:Not exactly

[blackmagic1982]

This is all based on the assumption that there are loads of idiots out there that will randomly decide to edit articles of which they have no knowledge. True, this becomes a problem with more derisive subjects, but for, say, an article on the history of string theory or the description of last week's episode of Battlestar Galactica, why would a person who has no real knowledge of the subject go on this article in the first place...let alone be compelled to edit it? What is interesting about the wikipedia debate is that it goes into the how we view humanity as a collective body in general. Are most people idiots or experts? What qualifies someone as an expert? What is a bias? How does even introducing the concept of expert effect the dissemination and censorship of information? Debate is all well and good, but when it comes down to it, no one can prove all people (excluding you, dear /.er) are stupid, evil idiots and no one can prove that all people are intelligent, reasonable individuals. Besides, even now in our modern goggled age, school textbooks are riddled with errors, present vastly different views of histories depending on the region (even within a single country) and no one seems to mind or question. Wikipedia, even at it's worst, at least acknowledges the subjective of the idea of truth and expertise itself. And furthermore, unlike authoritative texts, it allows the reader to be an active participant and invites debate. Imagine the possibilities of a school system where all the students where encourage to help create the textbooks themselves instead of memorizing facts they assume are true without any thought? Of course this all IMHO. Because hell, I don't really know if wikipedia exists at all.

Re:Not exactly

[rtb61]

Well, it would be more accurate to say that wikepdia is full of facts that people want to share.

More words == lower error rate?

[aborchers]

So if I go to Wikipedia and type the word "gibblefinch" a few thousand times into an article, I can reduce its error rate?

Re:More words == lower error rate?

[ThosLives]

Indeed. I think people need some education on how to establish proper metrics. There seems to be a misconception that just having metrics is sufficient. The significance and meaning behind most new metrics seems to be missing.

Re:More words == lower error rate?

[aborchers]

"Aw, people can come up with statistics to prove anything, Kent. Forty percent of all people know that." -- Homer Simpson

Re:More words == lower error rate?

[Phreakiture]

So if I go to Wikipedia and type the word "gibblefinch" a few thousand times into an article, I can reduce its error rate?

Only if that is what the article should say, and saying so is useful to someone looking up whatever topic it is you are looking up and finding the aforementioned gibblefinch storm. If, on the other hand, it is not useful or relevant, then not, it would tend to increase the error rate, or at lease lower the signal to noise ratio, rather greatly.

Re:More words == lower error rate?

[PhoenixK7]

OK. I can agree with this on one level, however these are a select number of articles that have been reviewed for inaccuracies. Presumably they don’t contain “gibblefinch”; repeated 5000 times to increase article length. What is indicated here is that per kilobyte (as a metric for length) there are fewer errors. Being more lengthy does not necessarily mean there’s really more information contained on the page, but given the gross difference in article length I’d hazzard a guess that the wikipedia articles don’t simply contain a bunch of fluff to make them look longer, they probably actually have more content.

To settle this issue, the metric should not be inaccuracies per kilobyte, but inaccuracies per idea/concept/fact or whatever, but those statistics are a little more of a pain to collect :-)

Re:More words == lower error rate?

[CMiYC]

Bleh. /me adds "PhoenixK7" to the list of people who should hit "preview" first.

Re:More words == lower error rate?

[TeXMaster]

If I write an article with a certain number of errors, and the rewrite the same article with the exact same conceptual content, but in a much more verbose manner, the article hasn't improved: it's still as right (or wrong) as the short version was. If the errors are conceptual or factual, the total length of the article is absolutely irrelevant. The only thing that is important is the number of concepts or facts expressed, and how many of them are right. So unless the additional length introduces more facts, more examples and similar (and is therefore 'valuable' additional content, and not just verbose verbiage) the length of the articles is not significant.

The only case in which error/length ratio is meaningful is when you are only considering grammar or typographical mistakes: badly formed phrases, missing or additional or misplaced letters, dates with inverted numbers, and so on and so forth.

Note how this post is rather verbose. I could come up with a much longer or a much shorter one, with the same content, and I would still be equally wrong (or right).

Careful with stats...

[erick99]

I am not sure that it is reasonable to consider error rate primarily as errors per unit of text. In that case, one could write a submission and then insert a lot of fluff to lower the "error rate." I would consider the absolute amount of errors per submission at least as important as the quantity of errors as a function of quantity of text. Just a thought.

Re:Careful with stats...

[Anonymous Coward]

Number of errors per article isn't that meaningfull of a measure either. What type of errors? Does one have a spelling error while the other says a whale is a fish?

Re:Careful with stats...

[shaitand]

We can safely assume the "experts" had moral concerns. Therefore they have corrected all the wikipedia errors leaving ZERO per article. Britanica on the other hand still has 3 per article.

Although, a difference of 1 error per article in lengthy science articles is not substantial enough to pass the margin of error of the experts themselves.

Accuracy

[penguinoid]

Wikipedia has less errors, you say? We'll be fixing that shortly...
-- The Britanica Team

Re:Accuracy

[LiquidCoooled]

Don't you mean 'Britannica'?

Sincerely,
A Wiki editor.

ps, we don't hold grudges and most of us will gladly help clean up your mistakes :)

Re:Accuracy

[colinbrash]

Wikipedia has less errors, you say? We'll be fixing that shortly...
-- The Britanica Team

Modded as "Funny," but this is actually very insightful. The problem (and advantage) of Wikipedia is its volatility. Anyone can go change something to be incorrect (whether maliciously or not), at any time.

This study, unfortunately, tells us almost nothing. The average number of errors per entry is really not a valuable statistic. How bad were the errors? How long are the errors there? Wikipedia, because of its volatility, really cannot be instanced in the way this study has done. It would be more revealing to do a study of the past X months/years/whatever, to determine how many errors there were, what kind of errors there were, and how long these errors were around.

Of course, then the Encyclopedia Britannica wouldn't be studied as *it* should. Because it is *not* such a volatile resource. In reality, the two resources are not as similar as people think.

And regarding the average number of errors per length of text: this statistic is downright worthless. If someone states something incorrect in one sentence, how is it any better to state the same incorrect thing in 10 sentences?

Versatility

[soulsteal]

Sure they found errors in Wikipedia and Britannica, but which one can you go back to and correct?

Game, set, match!

Re:Versatility

[BushCheney08]

...but which one can you go back to and correct?

Both. Doing it to one of them is likely to get you kicked out of the library, though...

Re:Versatility

[mcgroarty]

Which one is more likely to grow links to goatse.cx between the time you cite it and the time your professor reviews your paper?

Re:Versatility

[bogado]

Simply cite the revision you used. By this I mean, "Article on frogs on wikipedia on the revision made in 06:31, 17 November 2005 : http://en.wikipedia.org/w/index.php?title=Frog&old id=28572307". If any new information like links to goatse.cx are added the person who is reading you work will know that it was not from that version that you based your work.

Re:Versatility

[petermgreen]

if you are citing an encyclopedia then you are probablly doing something wrong.

if something matters enough that you are citing sources then you REALLY should be citing primary sources from the appropriate field.

Re:Versatility

[vertinox]

Which one is more likely to grow links to goatse.cx between the time you cite it and the time your professor reviews your paper?

Lets just say I'm banned from using the color copier at my local college library.

Evolving vs. Static

[mattwarden]

I think you would also need to take into consideration the maturity of the chosen articles, since Wikipedia's content evolves continuously rather than on set publication dates. Newer articles probably would have a higher error rate.

Re:Evolving vs. Static

[krgallagher]

"Newer articles probably would have a higher error rate."

I think the choice to use scince-related articles slants the results. There are not a lot of people who feel capable of writing about Epitaxy. On the other hand, those subjects that are more accessible to a large group of people, such as Ethanol or Thyroid have significantly higher error rates. I think it is probable that more popular subjects would have a higher error count due to 'urban myth' being included as fact.

Accuracy - Good, Writing Poor

[ehaggis]

As the article states, the writing style in Wikipedia can be poor. Low diction, poor grammar and bad structure contribute to the chaos.

Most research I do on Wikipedia does not depend on good writing, but accurate information, especially on pop culture items or obscure "geek" subjects. Wikipedia does well in this. I have seen defaced articles "heal" with ten minutes of the incident.

As a contributor to Wikipedia, I am glad it is gaining widespread notoriety and validation.

Informative

[drewzhrodague]

I find Wikipedia quite informative, and easy to get to. I don't see what the problem is, or why those people want to class-action Wikipedia. I've learned a bunch of things by browsing, and investigating things mentioned in the articles. Even if Wikipedia were a little bit innacurate, it would certainly beat out my first 8 years of education, where I've found almost all of the science I've learned is actually wrong (by talking to scientists, and reading books, and wikipedia).

Re:Informative

[drewzhrodague]

You see, I had none of that. No chemistry, no calc, no trig. Instead, I got random insignifigant battles, simple science things that were wrong (blood cells break open on the sharp edges of broken vessels to clot), no physics, no astronomy. I did have some random teachers that were sense-making, and actually educational -- I've read Shakespeare, and some classics that stuck. Lots of teachers stressing memorization, which I found useless. I tell my SO that I would've been fine coming out of 6th grade, and going off to college. I wasn't some special smarty wizz kid, but a normal interested human. I found my education at 8 different schools a bit lacking, and left to get my GED when I was legally allowed.

Speaking of poor writing....

[Anonymous Coward]

Writing style

Nature said its reviewers found that Wikipedia entries were often poorly structured and confused.

The Encyclopedia Britannica declined to comment directly on the findings.

But a spokesman highlighted the quality of the entries on the free resource.

"But it is not the case that errors creep in on an occasional basis or that a couple of articles are poorly written," Tom Panelas, director of corporate communications is quoted as saying in Nature.

This confused me, until I realized that single-sentence paragraphs 2 and 3 should be a compound sentence.

Nature editorial asks scientists to contribute

[AxelBoldt]

Nature also published an editorial which asks scientists to contribute to Wikipedia: "Nature would like to encourage its readers to help. The idea is not to seek a replacement for established sources such as the Encyclopaedia Britannica, but to push forward the grand experiment that is Wikipedia, and to see how much it can improve. Select a topic close to your work and look it up on Wikipedia. If the entry contains errors or important omissions, dive in and help fix them. It need not take too long. And imagine the pay-off: you could be one of the people who helped turn an apparently stupid idea into a free, high-quality global resource."

Re:Nature editorial asks scientists to contribute

[jjthe2]

So did Nature fix the errors it found?

Re:Nature editorial asks scientists to contribute

[StupendousMan]

And will you then ask them to visit the site
every week to remove the changes made by
people who aren't experts in the field?

That's why I gave up on it -- it's like
trying to build a sandcastle too close to the
water's edge. I'd rather use my time to
create something that won't be destroyed
after a month or two.

Re:Nature editorial asks scientists to contribute

[swillden]

If you as a scientist find a glaring problem that is extensive, then decided to delete it or change it totally, other Wikipedians will "revert" your changes as "vandalism."

There's an easy workaround to this, though. If you make a large change (and, frankly, large changes *are* likely to be vandalism) and it gets reverted, just post to the talk page, explaining your credentials and highlighting the nature of the problems, then post your edits again, with a comment that the changes are legitimate and to please refer to the talk page before reverting.

As for deleting an article because it's wrong... just don't do that. Replace it with an accurate stub at the very least.

Another thing

[Ostien]

Does Britannica have extencive articles on Lightsaber combat?

Wikipedia: 1
Britannica: 0

Re:Another thing

[laughingcoyote]

That's all just made up shit, dude. Why would you want that in an encyclopedia??

While I don't have a set of Brittanicas right here, I would guess that you can find references in Brittanica to the plays of Shakespeare, Aphrodite, Zeus, Thor, and The Odyssey.

All of that is "made up shit", but a culture's fiction and mythology is still relevant to a discussion of the culture in question. So why shouldn't Wikipedia, with its quicker-changing nature, have information on more modern fiction and myth?

Re:Another thing

[smallguy78]

...or the fine art of pegging ??!

Hah! "Science" articles!

[ceeam]

What does Britannica say about "Goatse"?

http://en.wikipedia.org/wiki/Goatse

Re:Hah! "Science" articles!

[Citizen of Earth]

What does Britannica say about "Goatse"?

More telling is what Britannica says about Wikipedia:

Sorry, we were unable to find results for your search.

Re:Hah! "Science" articles!

[Dirtside]

What does Britannica say about "Goatse"?

Finally! An on-topic link to Goatse! Will wonders never cease?

Did they fix them?

[Ctrl-Z]

So, since they found these inaccuracies in the article, I would like to know whether they edited them and fixed them as they went, or just played the part of the silent observer. To me, this is the great thing about Wikipedia; if you know the subject and you find an inaccuracy, be bold and fix it already.

Longer article...

[everphilski]

... doesn't mean a better article. Encyclopedias are meant to be concise and to the point. A starting point for research, not a be-all and end-all. And I don't agree with normalizing errors to the length of the article, it should be the number of errors per article. Just because you wrote more stuff it doesn't give you the leeway to screw up more...

Re:Longer article...

[barawn]

Just because you wrote more stuff it doesn't give you the leeway to screw up more...

Uh, so if the Brittanica has an article which says "Bill Clinton was the 41st President of the United States" and that's all, and Wikipedia has a 12-page entry on Clinton which gets his date-of-birth wrong by one day but is perfectly accurate everywhere else, that's okay?

Look at some of the articles listed. The Wiki article (on Robert Burns Woodward) has a detailed breakdown of his life, his career, discoveries, and Nobel prize. The Wiki article's section on his honors and awards - which is just a list - is longer than the entire Britannica article!

Get your facts info from more than one source

[nysus]

No resource, no matter who it's written by, is absolutely definitive. Any thorough research will require going to many different sources to arrive at the best approximation of the "truth." Any person who relies on just one source for their information any topic is making a mistake. Wikipedia, Britannica, and other reference works should be considered only as starting points for further research. They should be considered nothing more than signposts for finding your way to other ideas and avenues to explore a topic.

Use Wikiagra - Increase Length and Girth!

[ScentCone]

But much of the extra length in the WP articles is often more commentator-ish, or blocks of material containing links, etc. Things that more traditional encyclopedias wouldn't want to include. And a lot of lengthier WP articles tend to get repititive, or have summaries and details that come close to being mutally unnecessary. Not a bad thing, just a different thing. Saying that WP articles are longer, and thus represent a lower real error rate is pretty misleading, I think. It's not the length of your article, it's how you use it.

12 % of Nature authors consult Wikipedia weekly

[AxelBoldt]

Note also that they "surveyed more than 1,000 Nature authors" and found that "more than 70% had heard of Wikipedia and 17% of those consulted it on a weekly basis." I wonder what percentage of Nature authors consult the Encylopaedia Britannica on a weekly basis.

There is still one critical difference -

[old_skul]

Britannica is authored by an entity which takes responsibility for its errors and has a long history of accuracy. Its content is "vetted", meaning that there is a measure of academic validity to what was written.

Some Wikipedia entries are far more detailed and far more accurate than Britannica's - however, that doesn't change the fact that the content was written by unknown persons with unknown source material for their entries.

Since when?

[Anonymous Coward]

"Wikipedia articles in the sample were, on average, 2.6 times longer than Britannica's"

Since when does longer mean better? If anything, Britannica's conciseness could be the result of several revisions and reviews for impact per word. Encyclopedias are about bang for the buck -- you can't fit everything into an article. It's meant to be a starting point.

That's where Wikipedia is supposed to excel -- the amount of live links available to primary web sites in addition to bibliography.

Two questions

[Colgate2003]

"Accuracy per word," or whatever you want to call it, may be greater, but are those words as well-written or necessary in the Wikipedia article?

Also, less than 3 errors/article compared to about 4 errors/article gives us more than 33% more errors/article Wikipedia. Many people (including Nature) are calling this close. Since when is 33% close? "Closer than expected," maybe but not close.

Man with one watch ..

[OzPeter]

One of my fav sayings (which also translates well into a coding practice when people want multiple copies of the same data in separate locations)

"A man with one watch always knows what time it is, but a man with two watches never knows."

Unless of course one of the watches is a nixie watch and that the batteries have run out after 2 days usage, or the cathodes have busted from all that shaking.

Can't reference Wikipedia because it changes

[nincehelser]

Wikipedia seems fine for informal use, but how can you possible cite sources with something that is constantly changing?

Re:Can't reference Wikipedia because it changes

[Lorenzarius]

The current version of an article is changing, but a particular past version is static. If you really need to reference Wikipedia, you can go to the page history page and choose one of the version. They actually have a page on citing Wikipedia.

Re:Can't reference Wikipedia because it changes

[Xzzy]

Shouldn't be citing encyclopedias to begin with. When I was in school, I had teachers that would mark student's work down if they used an encyclopedia as a source.

To my eyes their only legitimite use is for someone new to a subject getting a concentrated overview to get them started with real research.

Re:Can't reference Wikipedia because it changes

[shaitand]

Yes but you are talking about a printed encyclopedia. Printed encyclopedias are intentionally brief and not comprehensive. The sheer volume of material prevents comprehensive coverage. Wikipedia is not designed in this manner. Authors include as much information as possible on Wikipedia.

For an example compare the entries on winemaking in the two. One encyclopedia explains how to make wine, the other merely defines it.

The Nature of the Errors count.

[apberman]

It's not the number of errors, it's their nature. Equating an error in birthyear vs. an error in, oh, say, claiming that someone was involved in the Kennedy Assasination, is just stupid.

Wiki has it all....

[Himring]

No other encyclopedia or would-be encyclopedia covers as many topics as Wikipedia. I've used it to do everything from research SOX regulations for my job, to understanding my favorite online game, DoTA to name it. And they even have a page on mail order brides. Not that I've ever looked into that (god they're hot, and they all have the same name, Elena...).

How are they quantifying "error"?

[kalidasa]

If the Britannica article misspells 2 words, and the Wikipedia article is based upon an assumption that light travels through the medium of ether, does that mean that Wikipedia has half as many errors as Britannica? This is a lot more complicated than the kind of statistical error analysis these folks are trying for.

Re:How are they quantifying "error"?

[Raul654]

They had a seperate category for egregious errors like the latter - of which, (from TFA) 4 were found in Wikipedia and 4 in Britannica

Did ANYONE RTFA?

[hkhito]

Slashdot summary: 42 articles compared, but Oh! Wiki is 2.6 times longer on average.
TFA (first paragraph on the page): 50 articles compared, and articles selected with very similar lengths, and some material removed (e.g. references) if necessary to make them same lengths.

I challenge an assumption

[benhocking]

You (and implictly the submitter) are assuming longer == more content. Typically, better writers can say more with less words. Of course, more credentialed != better.

Re:I challenge an assumption

[mrogers]

Unless someone here examines the articles in question, this argument is pointless.

What an accurate and concise summary of Slashdot - you should work for Wikipedia.

Re:I challenge an assumption

[orgelspieler]

Length vs. content was brought up in the Wiki article. Basically the idea was that Wikipedia has inefficient (nonexistant) copyediting, and consequently it fits less information in more words. They called it "filler." Also brought up was the fact that WP and EB split up their articles in different manners, so there's really no way to be sure that comparable information, length, and errors are being measured, just because the articles are titled the same thing. Basically, this has about the same scientific weight as a Slashdot poll.

Back to what you said though, since more credentialed != better, you end up with an interesting situation with a "real" encyclopedia. You have experts that write on a topic, and they pass that on to a copyeditor who may or may not be an expert in the field. Her job is to make the article more efficient; she is the "better" writer you are talking about. Unfortunately, even the best editors make mistakes. (IANA encyclopedia author/editor, so anybody with more info please correct me.) Now with Wikipedia, you often end up with the opposite, where a non expert will get the stub started, and self-declared experts later fill in the blanks. So my guess is that even though the number of errors is comparable, they get introduced to the articles in different manners.

Re:I challenge an assumption

[iocat]

But... this is the slashdot comments section. If we required arguments to be based in fact, or backed up with links to legitimate sources (that is, not wikipedia), where would we be?

Anyway, while Wikipedia may have fewer errors per word, it is possible to say that EB is probably written more concisely, and therefore may have a greater fact density per word, rendering the comparsion invalid.

More importantly, though, I want to know about the QUALITY of the errors. Are the Wiki erros typos, and the EB errors the result of new data not being incorporated fast enough into the books? Or are the Wiki errors pointless political tangents, while the EB errors tend towards not always crediting everyone involved in a discovery. Just saying "errors" doesn't really give a good enough analysis, but it's interesting nonetheless.

Re:I challenge an assumption

[iocat]

IWAE (I was an editor) and I can say that when writing about a technical subject, it's rare to find a copy-editor (proof reader) who is as technically knowledgeable (for a variety of reasons). This inevitably results in small errors -- and sometimes large ones -- entering the text, as the copy editor tries to make the orginal text flow more in line with the english language, hit certain word counts, avoid widows and orphans, etc. If the author, or a technical editor, doesn't have time to carefully reread the text (which is almost always the case), you end up with errors.

This isn't just a problem with encyclopaedias, of course. Most PhD dissertations are riddled with errors, some very obvious, even though the author may have spent years on the document. (I mean errors that result from trying to convey information, not intentionally included wrong information -- missing words that change the meaning of a sentence to the opposite of what the author intended, dates the contradict other dates on the same page, etc.) The world's an imperfect place.

Re:I challenge an assumption

[shotfeel]

There's also one other strength of Wikipedia that often gets ignored -links. This makes it an even better starting point than a print dictionary in that more authoritative and in-depth information on anything from the definition of a word to a complex theory is often only a click away.

IOW its not just the information provided, its the linking to more information -something the web was designed for.

Very nice.

[blair1q]

I get 33% more errors, it takes me 160% more time, and random lusers on the Internet say it's a good thing...

what about now?

[krappie]

The real question is.. did the experts reviewing the articles click "edit this page" and correct the mistakes?

Either way, I'd like to see a repeat of the same test. They listed the articles they reviewed. Im sure the wikipedia articles are full of "0" errors now.

Re:Wikipedia

[Raul654]

Look for yourself at the abortion article. It's a properly referenced, neutral article on abortion. The people who wrote it were clever, in that they forked off a seperate article on the "Abortion controversy" (thus moving the debate elsewhere).

Can't we all just get along?

[typical]

Other than as a willy-waving metric, it seems that the error count in a tiny sampling of articles isn't useful at *all*.

I mean, it's pretty clear that both Britannica and Wikipedia are useful references. They have different strengths and weaknesses, but neither is gong to be unilaterally better.

Now, I personally use WP exclusively; It's available from anywhere with a web browser, it's free, it covers the sorts of things that I deal with frequently (tech, pop culture, people) and I'm a fan of the open source mentality. For my particular needs, WP is better suited. However, I don't see a need to claim that one is *better*. There are going to be WP articles that are *chock full* of errors on some points or link to sketchy sources, and there are going to be Britannica articles that just don't exist compared to WP or are simply outdated. It doesn't take people very long to figure out which is more appropriate to their uses, because aside from the initially surprising fact (to me, at least) that WP works and doesn't simply fall prey to vandalism, the strengths of the two aren't that hard to figure out. I'm not going to use WP as a primary source for a research paper, but it's going to be the very first reference that I turn to when I want an overview of a topic.

I think that WP still has some challenges to pass -- WP contains articles on specific *products*, which Britannica completely lacks, and at some point, marketers are going to start expressing interest in the ability to freely edit Wikipedia articles on their products. But people that claim that WP is not useful are so clearly demonstrated wrong by a short while of using WP that there isn't any point in even arguing the point. It would be like someone claiming that Google isn't useful because it can return results to pages that aren't peer-reviewed.

Right now, there's a lot of noise over the Seigenthaler incident, but that's a tiny ripple in a vast ocean -- people will find a way to solve problems like this (if not in WP, then in a competing, derived system), just because it's so useful to do so. Reputation systems, a second system that blocks admission of changes until someone reviews them, whatever. We haven't even scratched the surface of systems like this, and their value is clearly phenomenal. I have read far more history and computer science on WP than I've been motived to read about elsewhere for quite some time. I've looked up a number of things that I always wondered about (what "grunge" actually *is*, for example), because WP is so quick to access, so vast, and so readable.

The best thing about all this is that WP is something that nobody (or very few people, at least) were making noise about until recently. The Internet solves problems (communication, latency, ability to provide links to other content, ease of collaboration, access to everyone to try out new system ideas) that allow incredible new systems that have never existed before in humanity's existence, and the number of new (as of yet raw perhaps, unpolished) systems is *exploding*. Search engines are the only thing that was an immediate and obvious application to me when the Web came into being, and even the mechanisms of something like Google were certainly not obvious. In the past few years, we have seen ideas like del.icio.us, yahoo's bundle of services, free webmail, Wikipedia, and so forth come into being. What's even more incredible is that these things are *enabling* technologies. Each one is a tool that allows people to more easily communicate or deal with things, which makes us even *more* powerful and makes it even easier for us to make new tools. If I can freely collaborate without long-distance phone charges with people in Sweden, I expand the number of people that I can share knowledge with. If I can read, at least in a rudimentary fashion, the languages that I can read through use of Babelfish, I have hugely increased the number of documents available to me. If I can take advantage

Failure modes

[fermion]

As in many things, I feel that failure modes are much more interesting than instances of success. One can have a process that is very succesful when working, but in the failrure mode is catastrophic, then the perhaps that is not such a good process. The focus on success instead means of failure is a big reason why we have so many bad processses, and is a key method to psuh really harmful things onto unsuspecting population.

The nice thing about britannica is that though it is imperfect, I have seen few cases of pervasive campaigns of misinformation. To avaoid this failure mode, an editor should require a writier to be broad and reference a variety of sources. Also, when we are taught to use the encyclopedia, we are taught not to use a a primary source, but merely as a starting point. For instance, few say that the encyclopedia says this or that.

OTOH, the failure mode of wikipedia is potentially catastophic. The winners are often those who have the power to to push thier persepctive of a particular topic. This is not always the case, but since it is a probably failure mode, and since there does not appear to be an effective defense, it makes the wikipedia a much less reliable source of information, on average, than the britannica.

In the end I think the summary is another example of sloppy science. It is not so bad, as it indicates that the wiki can be more or less trusted on the types of topics nature posted, although the wiki did have more erros, though perhaps not statistically significant. The wikipedia process absolutely has to deal with the failure modes, and should encourage authors to point to peer reviewed sources to justify their claims of science and history, and a variety of sources for current events. After all, if everything comes from the weekly world news, we cannot expect much overall accuracy.

It shouldn't matter that much

[Quiet_Desperation]

If you're doing any kind of information/knowledge search, you never rely on a single source anyway. Unless you're a journalist.

Participation

[shaitand]

Did the experts correct the errors? I hope so.

Re:Participation

[IdahoEv]

RTFA: it was a blind review. The experts were given printouts of the text of both articles, and weren't told what the source was.

About the "class action lawsuit"....

[brian0918]

The parent referred to this site, which states that the group is gathering complaints to file a class action lawsuit against Wikipedia.

The problem? The people hosting the site are far from unbiased on the topic. The site is hosted by baou.com, which runs QuakeAID, a bogus "charity" set up after the 2004 Indian Ocean earthquake.

Why are they mad at Wikipedia? After the earthquake, a member of QuakeAID with the username Baoutrust used Wikipedia to promote the QuakeAID article and the QuakeAID website. Apparently, this included listing QuakeAID on the list of charities for the tsunami survivors. When their true nature was discovered, they were removed from the list, and they got pissed. Since then, they've been smearing Wikipedia at every possible chance.

Part of the Problem

[Rydia]

Part of the problem with this study is its subject matter; science-related articles are by and large cut and dry, and only common misconceptions usually are introduced. While one could say this exonerates wikipedia, I'm pretty sure this doesn't say a whole lot. Another problem is that they consider an "omission" an inaccuracy. That doesn't seem like a good standard to hold either publication to.

What about biographies, the pieces more often cited as innacurate? Or political pieces? Or any subject that has any controversy, really.

While it's nice to see that wikipedia is only slightly worse off in science, as the article said, it's still in general poorly written and still contains more errors than brittanica in the least error-prone subject. Hardly a vote of confidence.

Wikipedia needs a disclaimer

[wrook]

I was thinking something like:

In many of the more relaxed areas of the Internet, Wikipedia has long supplanted the great Encyclopedia Britanica as the standard repository of all knowledge and wisdom, for though it has many omissions and contains much that is apocryphal, or at least wildly inaccurate, it scores over the older more pedestrian work in two important respects.

First, it is slightly cheaper, and secondly it has the words Don't Panic! printed in large friendly letters on its cover.

Well, OK... except for the Don't Panic part...

Comparable length entries were judged

[Acy James Stapp]

From the results page at http://www.nature.com/news/2005/051212/multimedia/ 438900a_m1.html

"All entries were chosen to be approximately the same length in both encyclopaedias."

Are you all idiots? I guess I don't really need to ask that question.

Encyclopedia Britannica is much worse.

[Futurepower(R)]

I compared the information about Barbara McClintock, the Nobel Prize winner, in the Encyclopedia Britannica with that found elsewhere on the Internet.

The Encyclopedia Britannica article was not inaccurate. It was, however, extremely misleading. It was worse than worthless, since it gave the idea that Barbara McClintock's achievements were much less valuable and extensive than they actually are. After many years and much progress in Biology, her work is still valuable. A copy of her papers requires 80 feet of shelf space!

The Wikipedia article is far, far better than the one in the full Encyclopedia Britannica.

No space-limited, profit-oriented publication can compare to internet research, for most topics. I don't think that Encyclopedia Britannica has anything against Barbara McClintock, but the company must decide how much paper they want to buy.

Re:Entries the same length...or not?

[geoffspear]

Yes, but you're assuming that the rate of errors per article remains constant when the lengths of the articles vary.

Even if you ignore the obvious bias of the people (identified as "Wikipedians") refuting the Nature study, you have to admit their methodology is flawed. If the original study properly controlled for the lngth of articles, you can't refute it by showing that articles they didn't study might vary in length.

Why just science articles?

[ThinkFr33ly]

Seems to me that science articles might not be the place category of articles to use to judge the accuracy of Wikipedia. I suspect that most people contributing to the science articles have a pretty good knowledge of the subjects in question... they're not things that most people know a lot about. Acheulean industry? Kinetic isotope effect? Meliaceae? Huh?

Where I suspect more errors abound in wikipedia is in the articles about things that a lot of people think they know a lot about, but in fact don't have any idea what they're talking about. Or topics in which people have a vested interest in misinforming people. (Political topics, for example.)

Honestly, a better comparison would have been a sampling of 100 or so randomly selected entries. Confining it to just science articles seems like an attempt to misrepresent the accuracy of wikipedia.

Accuracy not an issue with non-controversial topic

[wcrowe]

In fact, accuracy is not really so much the problem as objectivity. With a non controversial topic, such as the scientific topics mentioned, Wikipedia's accuracy is quite good (it would be hard to "spin" gallium, say). And the level of detail you can get with a Wikipedia article can sometimes be overwhelming.

OTOH, when you get into topics that are controversial, most of the people who are driven to write about it feel passionately about the topic one way or another. In this way, objectivity flies out the window, and it is possible for inaccuracies to abound.

It is wrong to make blanket statements concerning Wikipedia's accuracy. Like information on the WWW in general, sometimes it is very accurate, sometimes it is not. Either way, you have to be amazed at how exhaustive it can be... something Britannica will never achieve.

In our current zeitgeist of moral relativism I am surprised that so many people are up in arms over the accuracy of Wikipedia articles.

But did they correct?

[matt me]

Did those experts having found those errors correct them? If not, why not? What Wikipedia needs is more expert contributors. I can add a little to articles I'm researching, but what is most helpful is when someone who knows more than most about a subject can work on those articles.

Researched???

[hellfire]

Exactly how does one research an article on goatse.cx? I don't mean what resources do you look up, I mean how does one stomach it? And how does one keep that from appearing on their tech writing resume?

"My latest wiki contributions include identifying the person who took the picture for goatse.cx."

At least I know how much to trust Wikipedia

[tgibbs]

Given the choice, I'd send a student to Wikipedia over Britanica.

The biggest problem with an "authoritative source" like Britanica, is that people--especially students--are tempted to take it as a final authority. But Britanica is not infallible, and even when it is correct, it is often superficial. People are tempted to settle for predigested opinions instead of forming their own

I think that the vulnerability of Wikipedia is in some respects a good thing, because it inculcates good research habits. I don't take Wikipedia as a final authority on anything, because I know that any given article might have been edited by a crackpot or an ideologue. Quote Wikipedia as an authority in a debate, and people will laugh at you. But I find Wikipedia extremely useful as a starting point for research; I just confirm anything important from primary sources--something that you should be doing this even if you use Britanica.

Instantaneous Content

[dave1g]

Many times when I read a news story and I find it interesting I will check wikipedia to see if I can add some new info from the article to the entry there.

Ony once had the new bleeding edge research not already been nicely integrated into the current article and sourced with a link to the academic paper or article.

"Rife with howling factual errors"

[AFCArchvile]

The Register had a letters section with people citing numerous errors. Among one of them was a former Data General product manager who wrote in his blog about the "howling factual errors" in the Wikipedia entry for the AViiON server line.

Among the errors is the origin of the OS that the servers ran, a System V variant called DG/UX. From the cited incorrect version (22 July 2005):

The first systems in the series were released in the summer of 1989, followed by a series of speed-bumped versions over the next few years. All of these systems ran a version of System V Unix written for them by Santa Cruz Operation, known as DG/UX, to which they added NUMA support.

And in the current version:

The machines ran a System V Unix variant known as DG/UX, largely developed at the company's Research Triangle Park facility. DG/UX had previously run on the company's family of MV/Eclipse 32-bit minicomputers (the successors to Nova and the 16-bit Eclipse minis) but only in a very secondary role to the MV/Eclispse mainstay AOS/VS and AOS/VS II operating systems.

Night and day. And there was more (quote from the Register letters article):

"It's also interesting to observe in the main Data General article how many "futzing around" edits there are. A link polished here, a comma there, etc. Yet this article as a whole is incredibly poorly organized with no real narrative flow. And what storyline exists is wrong in significant ways; it's not even internally consistent," he writes.

"The whole lock-in or no lock-in paragraph is 75% nonsense (it seems to imply that DG went to Unix because it couldn't afford to develop a SQL database? Yet, further down the article correctly notes that DG HAD a SQL database already.) The AViiON section mixes timeframes and contains multiple out-and-out errors, etc. (I suspect that the first couple of sections source their information largely from Soul of A New Machine and seem fairly accurate and cogent, but then it falls apart.) But that would all take work and expertise to fix."

"Easier to twiddle than create," he concludes.