Here's one about 7 months old:



Here's one about 2.5 years old:



Here's one about 3.5 years old:

(As if central Indiana has a music scene!)


That's all I have for you today, loyal Wikipedians. Get to work! I'd help you, but I'm banned!

Jimbo wants an encyclopedia, Jimbo can do it himself

To be fair to WP apologists, did anyone ever claim that vandalism to obscure articles is caught "within seconds"?

Usually when someone explains WP to the uninitiated, they explain that vandalism to prominent subjects such as the President of the United States are promptly reversed.

FWIW, central Indiana has quite a significant music scene; the huge number of students in Bloomington, outside easy travelling distance of anywhere else, makes enough of a captive market to keep a very noisy indie scene afloat, and Indianapolis's location halfway between Nashville and Chicago means pretty much every touring show stops off there.

... or even at all. Perhaps an agument for not having articles that nobody except the author and maybe the subject give a flying fuck about.

As opposed to vacuum-tube computers and sticking ferrets up your trousers? "Nobody gives a fuck" is subjective; as someone once pointed out, the strength of Wikipedia's model (and it does have strengths, despite the obvious weaknesses) is that it allows subjects that are important to the small group interested in them, but wouldn't be covered in a print book. Often the "nobody gives a fuck" articles are in better shape, as they don't attract the hordes of crackpots.

It's usually either caught within a few seconds or not noticed for a long time, outside of very popular articles, due to most "vandalism patrollers" just checking Special:RecentChanges. I'd be curious if anybody has ever actually claimed that vandalism is "reversed in seconds" (outside of media statements), so I can laugh at them.

You mean articles like the 100 for the United States senators?

Deliberate errors on those articles are fixed, on average, in 24 hours, a study proved.

And, yes, I think I could find about a dozen different Wikipedia apologists who claim that vandalism to all Wikipedia articles is caught very quickly. You must not know the Yahoo! Answers community very well.

You say that like it's a bad thing

Early computers, witches, and sticking ferrets down your trousers are important topics, and I would notice vandalism.

... by which i meant that if someone cares about an article they will have it watclisted. The red flag articles are those that no active editor is watching, but of course which they are is secret. "Security by obscurity" has never worked.

Sometimes vandalism lasts a day or two. Here's a recent example that really should have been caught by a new change patroller:

http://en.wikipedia.org/w/index.php?title=...oldid=330998932



How on earth would anyone know which they are? Surely it would take a developer to find out exactly who is watching a given article then it would be laborious to find ot which of them are active. And even if an active editor has something on his watchlist what if he doesnt in fact care any more and has just not bothered to unwatch it?

Umm, Special:UnwatchedPages. There's also http://toolserver.org/~mzmcbride/cgi-bin/watcher.py

Note, however, that the watcher tool only works for pages with more than 30 watchers (unless you have special access), which more-or-less eliminates most of the mainspace.

Quite. The unwatched pages a'e a secret because if we knew what they were, we'd vandalise them, knowing that nobody would notice. As I said, security by obscurity just doesn't work.

Or more accurately, the unwatched pages are a secret because Special/UnwatchedPages crashed in 2007 and Brion never got round to fixing it. (You can see the first 5000 entries just fine, but if you're interested in any article further down the alphabet than Adrenergic receptor you're shit out of luck.)

I disagree - I think most mainspace pages have less watchers, or active watchers, than you think, or you would imagine they should have simply by virtue of being in the mainspace.

Not scientific, I know, but I just picked the contributiosn of an editor chosen at random from recent changes and looked at the number of watchers of the last ten articles they edited. 6 of the ten were BLP, and of those four had less than 30 watchers. The other 2 had 38 and 58, or thereabouts. Of the 4 non-BLP articles, 3 had less than 30 watchers, and the other 1 had around 32 watchers.

Clearly not a scientific thing, but going further I think the number of watchers an article has doesn't necessarily equate to the same number of people being aware of changes to it - either (i) they've got a load of articles on their watchlist - some people automatically add every article they edit to it - in which case they won't necessarily notice or care about a change to an article on their watchlist, or (ii) they simply don't look at their watchlist, perhaps because they no longer edit Wikipedia.

I guess what I'm saying that number of watchers of an article doesn't necessarily equate to number of people actually watching changes to an article.

I'm pretty sure you're actually agreeing with Julian there - his point was that most mainspace articles have fewer than thirty watchers.

Could be, working my way through the tautology of it, in which case I'm glad my non-scientific sampling could prove both his points and mine at the same time!

It's also interesting to note that absolutely no action has been taken on any of the three articles I pointed out. This is good. Hasten the day.

It's not really going to hasten whilst you're talking about articles which have each received a grand total of 8 viewings this month (ignoring the one that Alison's just deleted), though, is it. It just proves that the speed of vandalism's capture is a factor of its impact...

Dude, I think you've just spectacularly missed Greg's point here...

Dude, quite possibly, as I seem to be on a roll for that sort of thing this evening. Alternatively you may well have missed mine...dude.

Such distinctions are not often made, and I remember that for some time the PR claim was being made that "half of all vandalism is reverted within four minutes" based on some flawed study someone did.

I believe you're thinking of this much publicized study http://alumni.media.mit.edu/~fviegas/papers/history_flow.pdf. As a matter of fact, the study itself is quite good, but it was used to support statements about how quickly vandalism is reverted that it didn't quite support. It's old, 2004 which is ancient in Wikipedia time, but it's good work.

Thats all very well but what if an article is on 35 watchlists and every one of them is banned or hasn't edited for a year or doesnt care about the article any more or is indeed the vandal. You could create dozens of socks just to watch articles you're vandalising to fool the Mzmcbride tool.

That's certainly true, but in general there's a pretty good correlation between how many people have an article on their watchlist and how many are actively watching it. It's extremely unlikely that an article on zero watchlists is being watched (it's possible that there could be some very committed person who actually checks the page history of a given article frequently, but it's unlikely). Similarly, the odds are that if 31 accounts are watching a page, at least one real person is actively watching it (yes, the situation you describe is possible, but it's unlikely). Certainly, there are many exceptions, but I suggest that there's a very strong correlation between how many watchlists an article is on and how actively it is being watched.

As has been pointed out, the first site doesn't work, and the second doesn't provide the most interesting part of the data, which is PRECISELY how many people are watching pages which have fewer than 30 watchers.

McBride, you read WR. Why isn't that data available?? Is it a secret? It is somehow difficult to get? Is it embarrassing? What gives?

When I originally wrote the script, it allowed full access, regardless of the number of watchers. The Toolserver system administrators felt that this was akin to bypassing MediaWiki security (as Special:UnwatchedPages is restricted to administrators) and they disabled the script altogether. As a compromise, the 30 limit was put in place. (I'll note it wasn't my first choice for a limit and it really sucks for the smaller projects.)

That little "log in" link at the bottom refers to Toolserver/watcher, an access list that has (hackishly) been implemented into the interface to allow listed users to see the uncensored results.

I've been asked by one researcher from Harvard for a full copy of the data set, which I released to him or her privately. I have no problem releasing the data to others, assuming there's a legitimate need and the user can be reasonably trusted not to do harm with it (though I think the data is vastly overrated, personally)....

Arghh. Yes, as I suspected, it's embarrassing. Even if it was only available to administrators (to prevent vandals from having the data to target) there's a risk that the stats would get out, and there's nothing that can really be defended there, is there? I mean, even if only 1% of pages aren't watched by a single person, how is Sue or Jimbo or Godwin going to answer when some media wonk says "What about the 1% of pages nobody watches but your robots (which are stupid can can be fooled?)"

There's really no answer to that except to deflect the questioner to the subject of the pages that ARE watched.

So look, why don't you go to the foundation and say: "This is unconscionable. I propose that all unwatched BLP pages be declared orphan BLP pages, and fully protected against all editing for a week or a month grace period, while they either do, or do not, collect 5 sponsors--- and then either released to editing or else deleted entirely, depending on that."

The same with newly created non-BLP pages--- except that after some grace period, if they aren't watched by a couple of people other than their creators, they ought to be simply sprotected until they do collect the minimum number of stewards.

You do remember I gave you stats specific to biographies, right? http://toolserver.org/~mzmcbride/milton-watchers.txt (data from October 29, 2009, I think)

Someone recently proposed doing what you've suggested, though the overrated-ness of the data becomes more apparent when you consider that you can't tell if the user watching a page is active and checking their watchlist every 20 minutes or... dead. Or a bot. Or hasn't figured out (or never will figure out) what a watchlist because they just started a biography one day while they were bored.

All of this isn't to say that writing such a feature into MediaWiki is impossible. But, y'know, FlaggedRevs needs more development work. Or something. So that would leave people trying to use the available data, and they'd end up just making a mess based on weak (or bogus) conclusions.

The current data is useful for satisfying that "huh, I wonder how many people watch my talk page" feeling, but the overall utility right now (without being able to see the user ID column) is pretty limited.

The proposal, for reference: http://en.wikipedia.org/w/index.php?oldid=..._inactive_users

Uh, somehow I never saw that. Good heavens, so 56,000 BLPs are on nobody's watch list? That's 14% of them. And twice that many are on only one person's list, which means 40% of BLPs have one person or nobody caring for them.

Who's the one guy on 3400 watchlists, twice as many as everybody else? W Bush? Obama?

But that doesn't mean the data is overrated-- merely that the situation is (at best) worse than you describe, for your tool gives a best-possible scenario. We still know that nobody is watching articles on NOBODY's watchlists, right? What we don't know (as you point out) is if articles on a few watchlists are watched at all, because we don't know if those (few) users are active. But AT LEAST 14% of BLPs are orphans. I don't know how that happens. Presumably it happens if the user takes a bio he created off his own list, but how else? If the user deletes their account I suppose all the watched articles on that account decrease by one-watcher count, right?

Wikipedia really needs a way to delete accounts which haven't been used in 6 months or a year or something, so these stats mean something. Yes, I looked as your more moderate suggestion that we don't count them for watchlist stat purposes, and I see you couldn't even get that through, so SNOBALL on what follows.

But keeping "inactive users" as active accounts IN ANY SENSE only inflates "membership figures". It sounds like something a religious or charitable organization would do, but it's not very helpful for Wikipedia. But yes, you won't be able to reach consensus to change a thing.

Oh. My. God.

That proposal should be read simply as a example why nobody thinks "the community" is capable of deciding ANYTHING on Wikipedia. What idiots. What obstructionists. What dunderheaded reactionaries.

What Wikipedians.

(Come on, did you really think they would do something "reformist"?)

So if I don't edit within 30 days, my watchlist will be deleted?

This is the end of the wiki-gnome and the vast majority of Wikipedians. Only the elite Metapedians will be able to edit Wikipedia and everyone else will be shoved out. Mark my words!

Looking at Wikipedia:Database reports/Most-watched pages (sort by the "ID" column), you can see that it's Richard Evans (British author), Barack Obama, and George W. Bush at 3441, 1899, and 1891 respectively. Evans appears to be some sort of database corruption; two entries from the page's history:

And, yes, the stats are depressing. If you look at pages with fewer than four watchers, it's about 295,000 biographies, or nearly 70% of all biographies of living people total.

This just got posted on Akahele.org. If you think it's an important new fact about Wikipedia's carelessness, please Digg it.

That is sick.

While it's not likely vandalism, just a copy-paste issue in the infobox from the get go...

Michael Russell (tennis) indicates that this guy is a six-time Grand Slam champion. Wikipedia's had it that way for just about three years now!

I estimate that the article's been viewed about 14,000 times over that span, and nobody's bothered to fix it. That would be too hard! Those infoboxes are kind of pesky to correct.

How do you come across errors like this?

These 3 steps do the trick.

1. Special:Random
2. Read article
3. Check implausible claims using a reliable source.

I think in this case the steps probably go like this:

1. Find that some misguided individual is calling your website communist
2. Look up said individual's name in WP
3. Check implausible claims using a reliable source.

Bummer. Can you do 100 BLPs a day, 365 days a year? Then it will only take you 11 years. By which you'll have to start over, to say the least. Not to mention that the category will have taken WP over like Kudzu by then (all else being equal and with no changes in present trends).

I haven't heard form Mr. McBride since we noted that his stats show that AT LEAST 56,000 BLPs must be completely unwatched, and surely more.

I think some adventurous person needs to put this on WP:VP and Jimbo's talk page, just to see what kind of completely idiotic and/or childish reaction you get. No, I don't expect any changes to result. This is just for shits and giggles.

Carbuncle's paying attention. Five gold stars to him.

Speaking of doppelgangers, yesterday was not a good day to affirm that my family's surname at one time used to be "Koos".



But seriously... someone who is not drunk on the JimboJuice, with a college education, and a critical eye, should be able to find at least one glaring problem in every non-Featured Wikipedia article. Every one. Without needing to consult any outside reference except one, perhaps two, queries of Google.

Fair enough. Though, I try to avoid reading Wikipedia's articles usually—they're not very good.

What do you mean by "glaring problem"? If you're limiting it to factual errors, I'll take that challenge (for one thing, there are thousands of perfectly accurate one sentence stubs). If you're including things like poor quality writing, you're probably closer to right (though again, there's those stubs).

Agreed, this requires more qualification. A great percentage of articles are two or three perfectly accurate but entirely uninformative sentences (X is a city in Y with a population of Z; John Doe is a musician who played in band X). Now, if we look at any Wikipedia article of substantial length (featured or not) there are errors in it, some more more severe than others.

As a matter of fact, I challenge any one out there to provide me with an example of a single full-length Wikipedia article (including or perhaps especially a featured one) without any factual errors or significant omissions.

I'm right here, dear. I decided to make a log of the number of unwatched biographies. It should update every 30 minutes.

Can your bot replace the contents of each of them with "poooop!" and see how long it takes for anyone to actually notice?

How many biographies are tagged with BLP unsourced? I quit counting after 25,000. How many are both unwatched and not sourced?

I'd equally challenge you to provide an example of a standard textbook that doesn't contain any factual errors or omissions.

I've used Joe Nye's Understanding International Conflicts for years, and to the best of my knowledge there are no errors in it, and according to Google, no one else seems to have found any either.

51,047 non-redirects in the article namespace use "Template:BLP unsourced".

8,062 non-redirects in the article namespace use "Template:BLP unsourced" and are completely unwatched (0 watchers).