Appendix 4

Coverage of Wikipedia in the popular press veers between two extremes.

It's either the admirable heroics of plucky amateurs - it's Neasden FC winning the FA Cup - or it's the latest net threat to civilization. This week a survey by Nature gave sympathetic news editors the chance to revisit the first view.

Accuracy of Wikipedia matches Britannica, review shows, boasts CBC. Wikipedia as accurate as Britannica on science trumpets CNN's website. Business Week, which wants to be the house journal for Web 2.0 badgers, has no doubts. A Vote of Confidence in Wikipedia it shouts.

But what's the real story?

Science journal Naturechose 50 science articles from both Encyclopedia Britanica and gave peer reviewers a blind test to find mistakes. That gave the free-for-all web site a fighting chance - as it excluded the rambling garbage and self-indulgence that constitute much of the wannabe "encyclopedia" social science and culture entries. 43 reviewers replied, and this is what they found.

Britannica turned up 123 "errors", and Wikipedia 162.

In other words, the quality of information coming from Wikipedia was 31 per cent worse.

Peer reviewers also regarded the Wikipedia entries as poorly written and structured. "They need a good editor," Britannica's Tom Panelas told the BBC.

Many Wikipedians will say that this reflects poorly on Britannica. Indeed so, although without a closer look at the reviews, no one should draw firm conclusions. There are errors and there are errors, and some of the Wikipedia science entries were real clangers. It incorrectly described how Mendeleev's work related to that of British scientist John Dalton, and one peer reviewer asked, in amazement, "Who wrote this stuff? Do they bother to check with experts?"

For anyone concerned about the quality of information, this is depressing reading.

The McDonalds-ization of street food means you can go anywhere on the planet on get a terrible, bland meal of unhealthy junk food. Is the same thing happening to knowledge?

Wikipedia, like McDonalds, wants to be ubiquitous. If Nature is correct, the future is going to be a third more unreliable than it is today - a depressing prospect, we think you'll agree. If ever needed proof that a glut of information doesn't mean better information, we now have it.

Who could possibly hail this as good news? Two camps, we think. People with a real chip on their shoulder about authority, as we saw earlier this week. People with a contempt for learning, many of you say. But more broadly, only someone more obsessed by process than by the end result can regard this as any kind of victory - something all the popular press missed in their anxiety to gives us an upbeat, good news story from Planet Wikipedia yesterday.

[A Wikipedia mailbag special follows later today - in English, not Klingon.] ®

---

Update: Readers Ernest Trurro and Barry Kelly have had a brainwave. Why not count the "error rate"?

"You didn't care to compare the average article lengths to find the actual error _rates_, which is what really matters, did you?" huffs Ernest. "I thought not."

"It seems you left out the fact that Wikipedia articles are on average longer than the respective Britannica entries," says Barry, who adds, darkly: "Whether that's by malice or incompetence I don't know."

Let's put to this to the test.

Here's a hypothetical entry, containing two serious errors.

Sir Isaac Newton was born in 1462 and published the Theory of Relativity.

We can see that it is 13 words long: an "error rate" of one every 6.5 words.

Now here's a longer version.

Sir Isaac Newton was born in 1462.

Badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger badger snake

He published the Theory of Relativity.

This version is 114 words long, and contains only 2 errors - an "error rate" of one every 57 words. That's almost nine times more accurate - and very much proves Barry and Ernest correct.

We unreservedly apologize, and once again, must hail the power of "collective intelligence".