3 ms·

> As a purely end-user of Wikipedia I just want to chip in only to say that I have not seen any observable decline in Wikipedia over the last 15 years. I am gen

by gwern 7y ago

> As a purely end-user of Wikipedia I just want to chip in only to say that I have not seen any observable decline in Wikipedia over the last 15 years. I am generally very happy with the results I get when I use it. What you are missing is the 'seen and the unseen'. Wikipedia has stagnated badly, and it captures an ever more vanishingly small slice of the world. The world keeps getting bigger, things keep happening, and even the core topics were never actually covered in the detail they could and should be covered. You simply cannot keep up with everything with a core of, what is it now, less than 30k regular editors? That's basically enough to fight off vandals, do basic maintenance, and cover the occasional hot topic like coronavirus or Donald Trump to a reasonably adequate degree. It's not enough to cover even the Anglosphere, much less seriously improve the back catalogue, or realize a fraction of WP's original aims. Whenever I look at WP articles on any topic I know, like deep learning/deep reinforcement learning/behavioral genetics/decision theory/darknet markets etc, I find that they tend to be hilariously out of date, short, incomplete, or not covered at all. They're usually not outright wrong, but what is not there is much more important than what is usually there. (Let's take a moment to consider how WP's AlphaGo article, which is one of its best articles on DL/DRL because AG is globally famous, does not even mention MuZero or the term 'expert iteration'...) Typically, most of that would be 'notable', assuming an editor tenacious enough to fight for it - but such editors are precisely what have been lost. And yet, WP is more important than ever! Isn't it ironic that when WP was at its healthiest, when it was most keeping up with Notable topics, it was hardly used, and now that it's been severely ailing for a decade, search engines and social media networks depend abjectly on it for their knowledge graphs and factcheckers and disinformation fighting efforts? (Well, not ironic. Just bad. For everyone.)

bscphil 7y ago
> Whenever I look at WP articles on any topic I know, like deep learning/deep reinforcement learning/behavioral genetics/decision theory/darknet markets etc, I find that they tend to be hilariously out of date, short, incomplete, or not covered at all. The same is true for the stuff I know anything about, too. But I wonder if there's an alternative interpretation of this. The article argues: > And it’s casual users who matter. We lost the credentialed experts years ago, if we ever had them. I can see the truth in this. But at the same time, couldn't you say that to the same extent, the more "casuals" you have editing, the worse the articles that really need an expert's oversight (like machine learning articles) are going to be? I recently made some huge improvements to an article that was basically unreadable. Then I started looking through the history, and found that the article had been basically okay (though not great) about 8 months prior. Tracing the edits, I found that a professor at a community college had made it a semester long assignment for each student to pick a Wikipedia article to work on, in order to learn ... something. In any case, the student assigned to work on the article in question couldn't even write coherent English, let alone make high quality edits to this article. They had systematically ruined the article with a new account over about 4-5 months, with no oversight, and none of their commits were reverted by any editor. So I find it rather hard to say for certain that what we need are "more" editors, if those editors are largely going to be casual editors. As an expert (in a small number of fields), I don't (usually) edit Wikipedia for two reasons: 1. The starting quality of most of the articles I would work is so abysmally bad that I'd honestly rather tell the students I work with to just totally ignore Wikipedia as a way of learning about stuff in my field. Making two or three articles passable would not resolve that fundamental problem. (The problem is that the quality of an article not overseen by a very good editor is at best going to approach an average quality essay by a college student - in other words, about a 2 out of 5.) 2. I don't mind strict editing rules (see above), in fact I think they may often be necessary. What I've encountered when trying to edit Wikipedia, however, is that long-time editors know how to use and distort the rules to their own advantage. This doesn't mean the rules are misguided, necessarily, it just means that Wikipedia's culture is too insular and the top editors tend to have their own way with things.

by gwern 7y ago

bscphil 7y ago

> Whenever I look at WP articles on any topic I know, like deep learning/deep reinforcement learning/behavioral genetics/decision theory/darknet markets etc, I find that they tend to be hilariously out of date, short, incomplete, or not covered at all. The same is true for the stuff I know anything about, too. But I wonder if there's an alternative interpretation of this. The article argues: > And it’s casual users who matter. We lost the credentialed experts years ago, if we ever had them. I can see the truth in this. But at the same time, couldn't you say that to the same extent, the more "casuals" you have editing, the worse the articles that really need an expert's oversight (like machine learning articles) are going to be? I recently made some huge improvements to an article that was basically unreadable. Then I started looking through the history, and found that the article had been basically okay (though not great) about 8 months prior. Tracing the edits, I found that a professor at a community college had made it a semester long assignment for each student to pick a Wikipedia article to work on, in order to learn ... something. In any case, the student assigned to work on the article in question couldn't even write coherent English, let alone make high quality edits to this article. They had systematically ruined the article with a new account over about 4-5 months, with no oversight, and none of their commits were reverted by any editor. So I find it rather hard to say for certain that what we need are "more" editors, if those editors are largely going to be casual editors. As an expert (in a small number of fields), I don't (usually) edit Wikipedia for two reasons: 1. The starting quality of most of the articles I would work is so abysmally bad that I'd honestly rather tell the students I work with to just totally ignore Wikipedia as a way of learning about stuff in my field. Making two or three articles passable would not resolve that fundamental problem. (The problem is that the quality of an article not overseen by a very good editor is at best going to approach an average quality essay by a college student - in other words, about a 2 out of 5.) 2. I don't mind strict editing rules (see above), in fact I think they may often be necessary. What I've encountered when trying to edit Wikipedia, however, is that long-time editors know how to use and distort the rules to their own advantage. This doesn't mean the rules are misguided, necessarily, it just means that Wikipedia's culture is too insular and the top editors tend to have their own way with things.