What a shocker - Orlowski misses the real story
I'm not sure that Orlowski would know what a fact was if it started stalking him. In an all too predictable hit piece on the accuracy of Wikipedia, he has a lot of fun at Jimmy Wales' expense. I guess he didn't see this piece on the overall accuracy of Wikipedia versus Britannica.
The reality is, Wikipedia's issues with accuracy has far more to do with the controversies on topical issues than it has to do with anyting else. I think I mentioned that awhile back.
Comments
Err... facts, please?
[cdegroot] December 17, 2005 19:33:14.009
Errr... I don't know where Wired gets its conclusion from, but if you do the numbers they say that Wikipedia has significantly more errors (I think I got p=0.15) than Brittanica. So "a toss" it certainly is not. Closer than I would have expected, yes.
Read a little closer...
[MK] December 18, 2005 15:12:32.070
The Wired article states that Wikipedia has more errors on average per article than Britannica (3 vs. 4, I believe), but also that Wikipedia aricles are on average 2.5 times longer than the respective Britannica article. Clearly, Wikipedia therefore has fewer average errors per article.
Encyclopedia's vs. Software
[ Alan Knight] December 18, 2005 19:56:56.364
Comment by Alan Knight
This all makes me curious. So, we have Wikipedia, which is basically an Open Source encyclopedia. We've heard at length about Jim's skepticism on Open Source software, particularly as it applies to the boring detail work, which I assume fact checking would fall into as well. I'm just not clear why these issues only apply to software, and not to content.
We've also heard about how hard it is to do meaningful comparisons, such as benchmarks. Counting the number of "errors per article" strikes me as one of those apples to oranges comparisons. I don't know what the errors involved were, but I'm reasonably certain that nobody who was annoyed with Brittanica ever managed to get them to write up the death of their CEO.
A few things
[ Troy Brumley] December 18, 2005 21:54:32.326
Comment by Troy Brumley
@MK...Could you find an article count for Britannica?
@Alan...You expected Jim to be intellectually consistent? ;)
Open Source seems to work best for big things that a lot of people want. I don't think it would work well for smaller applications. I am quite comfortable getting a web server (Apache) or IDE (Eclipse) via Open Source, and feel equally comfortable relying on something like Wikipedia. But, I would not want an Open Source mailing list manager or a resume or an information pamplet.
Wikipedia ~= software
[ Steven Kelly] December 19, 2005 3:53:07.590
Comment by Steven Kelly
Alan: a few reasons why Wikipedia might be easier to get right than software:
1) each article is largely just a collection of facts, where changes in order and structure do not create an error. Anyone who knows the topic can just read the article, and spot errors on the first read through. A software function on the other hand has a multiple of possible paths through, and changes in order and structure generally create an error. Trying to read a reasonable length function to determine that it is without error is hard.
2) a Wikipedia article has a very low coupling to other articles - there may be links, but they are narrow in terms of bandwidth of information passed. The only real possibility of error in linking is that you link to the wrong one of two synonyms. In comparison, Software has far higher coupling - almost every word is a "link", with quite a few having a high bandwidth for information coupling and hence possibility of errors.
3) people read Wikipedia articles, compilers read source code. People come with built in correction for obvious errors, and correct resolution of ambiguities. E.g. "after his father died, he left home" - a person parses "he" to be the son, without even noticing a problem; a compiler would have more difficulty, and could end up with a surprisingly active corpse.