Before everyone freaks out about this, the bots that we're talking about are janitorial, doing things like add dates to wikitags added by human editors.
You can see this right now; go find a page to add a (legit, please) {{clarify}} or {{who}} or {{where}} tag to. Add it without a date, and wait 5 minutes; Smackbot will date the tag for you.
This is a good thing for obvious reasons.
Similarly, I'm guessing the "reversion" stat here is misleading. Reversion isn't deletion; anything that moves content around the page is likely counted as well. Write a graf or two on any well-trafficked WP page and you'll probably see it moved around pretty quickly.
More importantly, reversion isn't permanent. On any popular article, someone is bound to disagree with the merits of your edit. The normal way of things when I was wasting time at WP was, edit, revert, talk page, re-post copy.
And that's genuinely valuable; it's helpful to know that someone's thought that this has been unclear for 2 days versus 2 years. The compressed diff will be quite small, but Wikipedia incrementally gets better.
of editors with a lot of trust in the Wikipedia editing community who were abusing that trust. I don't find it easy in the culture of Wikipedia to gain trust by referring to reliable sources and checking what those sources say.
As I was manually going through the top 50 list, "Darius Dhlomo, 920829" seems like someone who was banned, based on significant copyright issues.
Also, plagiarism is difficult, because there are a lot of articles based on the kernel of public-domain texts, of which citation is not required. And sock puppetry is difficult, because some people keep bots that do productive work, but if human beings have trouble sensing that, I imagine it'd be much more code than the hundred lines of code that it took to process this the first time: https://github.com/lsb/ugc-contributors/blob/master/mh-diffs...
The biggest editing bots tend to do template maintenance (so correcting formatting issues and adding dates to templates). Bots simply do tedious work (that would have to be done anyway) faster - so hence a big edit count.
My bot (which notifies English Wikipedia of image deletions on Wikimedia Commons) edits about three times more than I do each month (and I am somewhere around #500 activity wise).
The very top bot ClueBot NG does more than double the edits of anyone else per month; that is an anti-vandalism bot, it catches about 50% of vandalism on Wikipedia each month. So pretty crucial :) Cydebot (#2) does some sort of category maintenance (not sure what, never seen it before). SineBot (#5) goes around and signs peoples comments when they forget (handy beast).
Basically it is just infrastructure; but because of the nature of the Wiki it is more visible than other forms of infrastructure :)
Some bots auto-revert changes that are likely to be vandalism or spam, like if an anonymous user adds a link to a blog on a free blog hosting site. (Specifically, see http://en.wikipedia.org/wiki/User:XLinkBot .)
"Over three quarters of contributions from registered users are from someone who's had a contribution reverted."
Well this makes sense entirely since it is peer reviewed and the articles are a living ecosystem which are built off of and on top of prior submissions and versions.
I approached that question from a purity standpoint: if you only limited Wikipedia to people who did things "right", where would you be? And the answer turned out to be that you wouldn't be very far.
I think urban planners have found that neighborhood cohesion and community watches are far better than gated communities and lawless knifefights outside, and the online textual equivalent of community dynamics agrees with that.
Agreed, wonderful analogy. Also in response to the source article, what defines a post as a "bad contribution" does this mean it was totally inappropriate or just needed slight tweaking. Has the author ever tried to make a legit new wiki post? I've never successfully posted anything without having even the slightest peer review. Who cares if the posts are almost entirely contributed by bots, just about everything is becoming automated these days (stocks, banking, traffic, etc.) and with the peer review base Wikipedia has I doubt the doomsday of robots are going to smarten up and misinform us...
if you only limited Wikipedia to people who did things "right", where would you be?
Although it's notionally possible that a first editor who got something "right" (factually correct based on reliable sources) was later reverted by someone edit-warring to score a point. Sometimes a particular edit makes Wikipedia worse rather than better. If NONE of my edits had ever been reverted (several have been), I would think that I wasn't trying hard enough to check sources and correct commonplace errors. However, it's not clear at the moment whether the editors with the lowest percentages of reverted edits are those who do the best edits, or just those whose prejudices (and topic choices) avoid scrutiny by the mob of other Wikipedia amateur editors. I am conscious of this issue because I was a professional editor and researcher to make my living long before most members of the general public had ever heard of the Internet, and because most wikipedians are so young and so devoid of editing experience that they readily mistake good edits for bad.
I think urban planners have found that neighborhood cohesion and community watches are far better than gated communities
It would be interesting to see citations to research on this issue, as well as the latest citations to research on forming online communities in which fact-checking and encouragement of careful scholarship become the group ethos. I live in a crime-free neighborhood with an annual block party and NO crime in a typical year, not gated, with good community cohesion, but also isolated from a lot of passers-by simply because it isn't a shortcut to anywhere else. Wikipedia is so exposed to the outside world, including spamming advertisers and propaganda agents of warring governments, that it may take more vigilance to protect than just an informal consensus among volunteer amateur editors.
A guy I knew at the Media Lab said that Marvin Minsky tried to correct his Wikipedia article, but someone reverted it. The IP address came from inside the Media Lab, so it might have very well been him.
The small-community-far-from-everything is quite similar to Wikipedia in its early years. Look at 2002-2005 and you'll see lots of anonymous Good Samaritans contributing helpful content. (Look for R6 and R7 on http://slightlynew.blogspot.com/2011/05/who-writes-wikipedia... to see a time series of anonymous contributions from 2003 to 2010.)
While, strictly speaking, this is unrelated to this topic, something has been bugging me that I would like to bring to the attention of HN readers:
BOUML is a free UML2 tool for C++, Java, Python and others. The developer, Bruno Pages, has stopped its development due to some issue with Wikipedia administrators.
I attempted to get more information from Bruno but he is upset enough to not talk about it. The following is quoted verbatim from http://bouml.free.fr/ and is all I know:
"Due to the continuous license violations, attacks and insults from people of wikipedia ( the worst of them were the administrators Bapti (commons)/ Bapti (fr.wikipedia)/ Bapti (de.wikipedia) , Dereckson (commons)/ Dereckson (fr.wikipedia)/ Dereckson (de.wikipedia) and Esby (commons)/ Esby (fr.wikipedia) ), I have decided to stop work on Bouml except to fix bugs. Bruno Pagès."
Oh! The number of articles I have read lately where I have gone and looked up actual books and discovered that the wikipedia content is copied verbatim from the books, yet with sentences that do not correlate strung together in a way that results in totally incorrect information.
And then there is the fact that for ages the Jean Claude van Damme article said he had killed 100 men with his bare hands {expletives deleted}.
Many of the mathematics articles contain appalling errors. You post about the errors in the discussion page and the issue is totally ignored. There seems to be no way to work with the community process to get things fixed.
I use wikipedia a lot for things that don't matter, e.g. as a structured skeleton for notes on a particular topic that I am writing for my own benefit and understanding. But as a project, I think it has basically reached its limits.
There's no way to move it significantly past what it is now.
>There's no way to move it significantly past what it is now.
There are tens, if not hundreds of thousands of pages in need of some loving. Even the good ones aren't perfect. There's still plenty of room for improvement, and thousands of people doing just that.
It's not as easy as you say. When the article contains mathematical errors, sourcing your correction is not easy. Wikipedia does not allow personal research. When the article contains "facts" which are in error, which you only know because you can understand the maths yourself, you don't have a source, other than your own research. Finding one, to correct what is to you a blatant error, can be a massive job. You shouldn't need a source to correct an error. The person who wrote the original piece of incorrect information should have to source their claim.
Also, when I edit wikipedia articles without consultation with the community first, it invariably gets shouted at and reverted. Consult the community, and you have to expect that people there actually understand why the article is in error. Mathematics is sometimes very subtle. Errors can be subtle. Trying to explain them to people who didn't understand what they were writing in the first place is impossible.
Yes, there are very many extremely qualified academics adding stuff to Wikipedia. But there are also very many amateur mathematicians who write many almost-but-not-quite-true things, not understanding the subtleties involved.
I already spend a lot of my time teaching people how to do mathematics. Doing so on Wikipedia is a poor venue.
By the way, since when did people start using the down voting feature on HN for posts they didn't personally agree with?
>By the way, since when did people start using the down voting feature on HN for posts they didn't personally agree with?
I think it was more the implied laziness of thinking Wikipedia is a great resource but not being willing to fix it.
As for your issue, if you know something's wrong from your own research, have you published it? There's a secondary source right there.
Wikipedia, by necessity, aims to be verifiable, not true. People will argue till the cows come home about truth, and do so much less civilly than when arguing about verifiability. So yes, there are plenty of errors on Wikipedia that exist because the source is wrong. The only consistent way to deal with this is to leave it in until other secondary sources emerge that allow you to fix it.
Still, you can just run the math, as that isn't OR. If you'd prefer, you can tell me the page and the error and I'll take care of it. Being an established editor can be handy sometimes.
"I think it was more the implied laziness of thinking Wikipedia is a great resource but not being willing to fix it."
If you spend your life teaching and doing research, laziness is not something you like to be accused of. Some of my papers and computing projects have been referenced on wikipedia. I feel my time is far better spent writing such things than entering edit wars with people who don't understand the subtleties of the subject they are writing about.
"As for your issue, if you know something's wrong from your own research, have you published it?"
I suspect you aren't an academic and I can see why you might think this would be a good course of action. You cannot publish corrections to blatant errors in Wikipedia and get academic credit for it. We are under continual pressure to publish new and interesting things. That we have found an error in Wikipedia is unfortunately not one of them.
"If you'd prefer, you can tell me the page and the error and I'll take care of it. Being an established editor can be handy sometimes."
I suspect you'd add a note to the page asking for a citation to support the existing incorrect statement because you won't be able to check it yourself. It will remain there for months, and by then everyone will have lost interest and moved on.
Or perhaps the statement has a citation, but for subtle reasons the statement is correct, but nothing to do with the subject of the article. Without understanding the mathematics, you have no way of checking that. So even "verifiability" doesn't help.
You need to understand just how very difficult this is. Go to the page on "Group action". Look at the section "Types of actions". Now look at the bullet point "irreducible". Do you see anything wrong there? Do you see a citation? (Please do not remove it -- I am not stating there is or isn't an error there.) How would you, an experienced editor, check whether this was correct and relevant information?
Edit: I edited the above for tone after someone pointed out that I sounded like I was yelling. Not my intention.
Friendly advice: you'd be more persuasive if you yelled less.
As someone who jumped into WP with the primary intent of improving technical software security coverage, I know where you're coming from. WP can seem hostile to experts, because it is --- unlike the real world --- structured to prevent experts from carrying higher status. Whole WP forks have been created to allow acknowledged experts to contribute without arguing with laypeople.
Having said that, all those projects failed, and WP is perhaps the most impressive resource on one of the top 5† most impressive information technology projects in human history. I don't love WP's attitude towards expertise, but I also don't love tilting at windmills. WP will continue to be successful without your mathematical expertise. That's just a fact. Knowing that, maybe you can understand how people on HN would feel it's unproductive to loudly complain about your experiences on WP.
† Prepared to concede the human genome project as more impressive than the Internet.
Whilst my posts on this subject are all over the place here. I'm not arguing that Wikipedia is not useful, nor impressive, nor successful. It is all of these.
Nor am I suggesting that having "Dr" in front of your name should give you more power on Wikipedia.
What I am alleging is that Wikipedia has a problem with plagiarism, with incorrect information and it has not brought a technological solution to the very difficult problem of verifying information in technical areas (I have also alleged elsewhere that Wikipedia has a problem with deletionism). The general Wikipedia approach works extremely well for general knowledge, but badly, IMHO, with specialised knowledge.
For that reason, some academics prefer to devote their time to writing primary peer reviewed articles (the corpus of which is vast compared to Wikipedia, and much more successful if quantity and quality is the metric). As I have already noted, there are many academics who do spend time adding to Wikipedia. They are not wrong to do so. But I personally see limitations to the approach, with no workable solution, technological or otherwise, in sight.
I understood the point you were making. I was attempting to add to it, saying that these are issues that are going to be especially annoying for academics, who are accustomed to different norms (in particular, respect for expertise earned through study and publication).
Academics like yourself are much better off writing primary and secondary sources outside the encyclopedia. It is the goal of the encyclopedia to aggregate and promote primary and secondary sources. It is not the goal of the encyclopedia to itself be the authoritative source for information on any given topic.
You will find that having written even a simple survey paper that your ideas and critiques are far more readily incorporated into WP than by bloodying your forehead against the brick wall of other editors following what they know to be well-intentioned and, empirically, effective rules.
At the risk of carrying on a conversation... I would like to reply to that. I certainly agree with that and appreciate your clarification.
However, what is happening is that Wikipedia is attempting to deal with very technical topics. As such there is a great divide between the techniques that are empirically observed to successfully collate general knowledge and the tried and tested techniques, developed over hundreds and even thousands of years for collating technical knowledge.
Precisely because academics have so little time to communicate their knowledge to nontechnical audiences there is a dire need for an approach that can successfully deal with this very difficult problem. My claim is that the wikipedia approach is so far not as successful in this area.
On forums where ignorance is king, I frequently advocate the use of Wikipedia to inform. Both aspects need to be understood. There are many people for whom Wikipedia should be a tremendously important tool. In forums like this, I emphasise the other side of things. The Wikipedia approach has its limitations as a means of disseminating precise and technical information.
Of course Wikipedia is limited for technical topics: it's limited by the quality and accessibility of primary and secondary sources. The way to improve WP's coverage of highly technical topics is to generate credible, reliable, accessible source material for it to incorporate.
This is the great thing about the Internet, and it's something WP management has been saying for years and years. Start your own wiki. Limit it to experts. Make it an amazing source. Wikipedia can then cite it, and use its content to resolve disputes and correct mistakes in its sourcing.
Arxiv has existed for a long time, and yet Wikipedia coverage of MSc-level math is pretty awful, and coverage of research math seems to be mostly nonexistent. eprint.iacr.org and cryptography: maybe a little better, but not that much.
Wikipedia needs "popularized" knowledge, to some extent, not just primary sources. Experts probably don't have a use for those resources, though.
You're doing a better job than I did of explaining the gap in available sources that experts could address to improve WP.
The point I hope I'm making, though, is that the improvement needs to happen outside Wikipedia. The experts can't directly be duking it out on WP pages as their journal articles get published.
Definitely. But you're suggesting a real change: the current model is researchers -> papers -> popularizers -> books -> science journalists -> articles -> Wikipedia editors -> Wikipedia, or something like that. Reducing this to researchers -> papers + wiki -> Wikipedia editors -> Wikipedia would definitely improve accuracy, but I think both sides would need to learn a lot about the other to make it work. There are benefits to that, but I don't see it happening.
Agreed. Both sides would need to learn a lot for that to happen. I think there's a vast chasm in fact.
Moreover, most academics don't have time for the wiki part. They barely have time for the papers part. Some academics get to write graduate texts because they can get credit for that, but that's about it. They do get time to write lectures to give to students though (sometimes barely).
But there's a problem. In mathematics, a result may depend on definitions, e.g. whether a ring has a unit element or not, which definition of natural number you are using, or which definition of compact, whether you allow the axiom of choice, what definition of norm you are using, or how a function is normalised. The actual results may be completely different in different books depending on the assumptions made and the precise definitions given. Language also changes over time. Sometimes the same word is used to mean totally different things, and the meaning depends not just on context, but era!
What happens in Wikipedia is you sometimes get a mosaic view of things which are not compatible. Only an experienced expert in the field can spot that and correct it, and it may be days of work to cleanly demonstrate why it is an error. Wikipedians can add such errors far faster than they can be noticed, corrected and justified.
The issue is that people think of science in a reductionist kind of way as a list of true statements. Books are lists of true statements and you can pick and choose the ones you understand and collect them together however you like. But this is totally false. Maybe the only reason you can even understand the book is because the author took pains to define things in such a way that a pedagogically sound route through the material exists.
In the mean time, academics are marking the same wrong answers to homework assignments and in written projects and dissertations over and over again. So this is a huge problem.
(This is to say nothing of the fact that many students believe that using quotations from Wikipedia without attribution in dissertations is permissible, presumably because Wikipedia allows this for public domain sources or something like that.)
What process or processes can bridge the vast chasm between what academics are trying to do on the one hand (advance the forefront of knowledge and disseminate it through teaching to the next generation) and what is being done on the other hand by wikipedians (collate human general knowledge in a way that is easily accessible to all)? There's a vast gap and it is precisely at the intersection of these two great enterprises that things are going woefully wrong.
Here is a better example than the "Group action" one, and it's a real clanger. Look at the page on "Group representation". There it defines a notion of "irreducible representation" in the section "Reducibility".
Now, notice the phrase "A subspace W of V that is invariant under the group action is called a subrepresentation." Which group action is it talking about? This is completely orphaned.
You might think to click through to the page on "Group action" and see if you can get a clue what might have been meant before an edit orphaned this phrase.
In fact, a module can be thought of as a generalisation of a vector space. So in the case of an irreducible group action as defined on the "Group action" page, if R is a field, the module there is a vector space. That G acts R-linearly means precisely that there is a group homomorphism from G into GL(X), i.e. that you have a group representation into the vector space X. So, now the definition of an irreducible group action is that there is no non-trivial, proper vector subspace of X which is invariant under the group action. The definition of an irreducible group representation on the other hand states that there is no non-trivial, proper vector subspace of V that is invariant under the associated group action.
But how in a million years is anyone supposed to work that out from looking at either page? And if either page changes, the connection is totally lost.
A substantial part of the "Group representation" article would have to be rewritten to fix this. And similarly for the "Group action" article. Then the two would have to be connected so that a revert in one doesn't mess up all the hard work in the other.
Naturally people have pointed out in 2004 that there are problems here. But in seven years it has never been fixed.
Of course I can't point out the error. I'm asking you to, and provide a source to back it up. I'm no expert on math, but I've been editing for years and I know how to get corrected information included. I'm not suggesting that I, as a layman, could replace your knowledge as an expert. I'm suggesting that I, as an experienced editor, could help you fix a problem in a way that's likely to stick.
When I find errors in Wikipedia articles, I add a section to the Talk page explaining my reasoning, citing sources if I can. Then more "established" people can make the edit.
As far as sourcing errors goes, think about it from the perspective of a generalist. For any given "correction," I think it's fair to assume that the change is more likely to be wrong than correct. The generalist who curates many pages is probably not qualified to determine either way, so the safest thing to do is revert back to the old version. Hence, sourcing corrections can give them confidence that the change is more likely to be correct than the original.
Yes, this is difficult. But I think that's it's inherently difficult. That is, providing a vast source of knowledge covering many topics digestible by non-experts is inherently a difficult endeavor.
Except I find most of my changes get reverted, even if referenced or simply deleting incorrect and unreferenced information. I don't have time to enter a life-long battle keeping an article correct.
Something does not compute there. If your changes are getting reverted, obviously someone cares enough to notice, so it follows that they should care enough to check the discussion. Are you sure you're not just getting hit by a bot? Maybe you did not tag your changes properly? If anything, there might be a problem that the admins should look at. I would like to see some of these changes you are talking about.
My experience with Wikipedia has been hit or miss. Sometimes I'll find an older article and fix it up a bit and it will stay. Other times I get my edit reverted by an over zealous neck beard who can't handle being wrong on the internet. I don't have the time to battle these types of people. This is what sunk, "Wikipedia is not a creditable source" into my brain.
How would you pay for it? Also, the error rate on Wikipedia is comparable to Britannica, which is written by paid experts. Letting anybody edit has worked so far. It may need to be adjusted, but there's no need to entirely throw out the model that's been so incredibly successful.
This is NOT correct. It is highly dependent on the subject matter and what is filtered from Wikipedia before doing the comparison. Take a hard science and take everything that is written and Wikipedia is substantially worse.
Furthermore, take Wikipedia and compare it with authoritative books on the subject matter in the actual articles and you will find that Wikipedia contains many very serious errors.
(Search for Ofqual in that article.) In particular, Wikipedia is about as accurate as general books and press articles. It is not an accurate source of academic information.
>It is not an accurate source of academic information.
It has never claimed to be. It seems a bit silly to say that something is bad because it isn't some other thing that it was never designed to be. My iPhone is a pretty lousy toaster, too.
It'd be cool to see how this was calculated. Though considering the amount of auto-spellcheck and similar that goes on, it's kind of surprising to see that half of the top contributors are human.
All bots have to be registered, and there exist simple tools to get edit counts. You'd just have to copy the list of bots and write a quick script to feed the names into the tools. Sum up the results and divide by the total number of edits.
There are no general spellcheck bots (I would say there are none, but it's possible there are some that spellcheck within a very clearly defined area, under very specific circumstances). Spellcheck bots are actually a fairly frequent request, and are always refused.
The idea that a human and a program would use the same interface is a good one if your goal is to allow the humans to focus on adding domain knowledge and the machines to focus on the boring repetitive tasks. Consider it a design pattern, not an evil plot.
The stat I found interesting is this one [citation needed]: while a majority of edits are done by few editors, most of the content is generated by either a) anonymous accounts or b) people who have contributed 2-3 times and haven't returned since. The editors, like the bots, represent maintenance edits.
Keep that in mind the next time Jimmy Wales gives beaucoup credit to editors, or some super-user gets power-happy.
Also gray-to-black hat spam with links. Just like the content farms that plague Google's results, they'll push in any low-quality content in hopes of trolling some clicks to their ad- or malware-laden sites.
I think most of those are probably officially endorsed bots that do things like correct common misspellings, crosslink between different langauge Wikipedias and do huge one-off tasks like category reorganizations.
You can see this right now; go find a page to add a (legit, please) {{clarify}} or {{who}} or {{where}} tag to. Add it without a date, and wait 5 minutes; Smackbot will date the tag for you.
This is a good thing for obvious reasons.
Similarly, I'm guessing the "reversion" stat here is misleading. Reversion isn't deletion; anything that moves content around the page is likely counted as well. Write a graf or two on any well-trafficked WP page and you'll probably see it moved around pretty quickly.
More importantly, reversion isn't permanent. On any popular article, someone is bound to disagree with the merits of your edit. The normal way of things when I was wasting time at WP was, edit, revert, talk page, re-post copy.