The Future of Wikipedia

Up until recently, how to describe Wikipedia's growth was pretty easy--exponential. From its founding in 2001 to the date of its two-millionth article in 2007, Wikipedia's article count nearly doubled itself every year (1a). Plenty of new users joined up and started editing--to hear some of the old vanguard tell it, those were different times.

Nostalgia or no, however, the demographics and statistics of Wikipedia have changed since those early years of unbridled growth. Wikipedia's statistics for article creation roughly follow a logistical growth model, which utilizes several assumptions: first, that more content leads to more web traffic, which in turn feeds more content (a positive feedback effect); secondly, that the more content there is, the less potential content there can be, thus a force for less content; and finally, that the theoretical limit of content is bound by the combined expertise of possible participants (1b).

What this means is that Wikipedia has no real limit. Even bounded by Wikipedia's guidelines that dictate what content is suitable for the encyclopedia, there will always be new albums released, new wars, new politicians, and new items of note. The real and practical limitation is made by the editors who contribute.

Since 2007, Wikipedia's total article growth has been decreasing. While the number of pages continues to climb--roughly 2.9 million articles and almost 13,000,000 pages in total as of this writing--the growth of Wikipedia's early days is a thing of the past. In 2003, for example, the encyclopedia saw a phenomenal 390% increase in articles over the preceding year--an average of 210 new articles written per day. As the wiki's popularity has grown, the number of new articles has continued to increase, but the percent increase has steadily dropped; 2008's growth rate was 24%, the lowest recorded (1c).

Of course, this doesn't tell us too much. Sure, the growth of the encyclopedia as a percentage has been dropping, but new articles are still being made at a tremendous rate, and it's not surprising that the percentage would decrease--in its early years, fundamental articles such as "cat" and "solar system" and their constituent articles were only roughly sketched out. Now the fundamental blanks have been filled in (if not necessarily to the desired quality).

Of course, depending on who you talk to, the decline is still worrisome, and for varying reasons. There are groups who feel that the decline in articles also means a decline in user participation. While that's hard to quantify because many accounts are created that never edit and statistics on who edits what and when are hard to collate for the entire project, massive at it is, there are some indications on how the real world model of Wikipedia works. Though there are millions of registered editors, the number who are active (generally, measured as X number of edits within X number of months) is dramatically smaller--in 2006, for example, it was just over 35,000 (2). And in the months since, the percentage of active editors has been decreasing. Furthermore the content on the encyclopedia that is actually audited for quality is produced by an even smaller cadre of editors who are responsible for the lion's share of featured content.

Put it this way. Featured Articles (FAs) are articles meeting criteria for neutral-point-of-view, comprehensiveness, "brilliant" prose and the like, and are judged by a group of uninvolved editors in a peer review-esque process. According to the FA page there are roughly 2,500 FAs--that's less than 1/1140 of Wikipedia's total articles (3). Adding in Featured Lists, which are counted in total article count but are distinctly different from FAs, and the number rises about 1400 or so... but that's still only .0013 percent of content that is featured (4). As article count increases faster than any group of humans could write, audit, and feature an article, the gulf will only increase. Furthermore, the processes for FA and FL are rather insular--while they are open for everyone to participate in, the real momentum is conveyed by only a handful of reviewers. If one of these editors decides to leave Wikipedia, the impact is much more dramatic than outsiders might expect. This small size means that even if Wikipedia were greatly expanding, the featured content pipelines would remain rather small. Either way you cut it, quality content is not being produced at a expedient rate.

But hey, there can be excellent articles that aren't recognized as "featured". But as many who stumble around on Wikipedia can tell you, much of these articles are dreck. First off, according to generated "lonelypages" statistics, more than 790,000 articles aren't linked to any other articles--these are called "orphans", and generally are on such obscure or non-notable topic they aren't worth saving. That's a full 27% of Wikipedia right there. Many more articles are merely one or two lines that provide very little content, called stubs. The total number of stubs is mammoth, organized into a whopping 7,181 categories and many more thousands of subcategories; the total number is in the hundreds of thousands, if not more. A further 40,000 articles, as of this writing are tagged as needing some kind of cleanup--either they are gibberish, or else badly written, or else unformatted (see cleanup list). But these lists are maintained by volunteers, so in reality there are many many more articles that need to be tagged, but aren't.

To foster collaboration and improve articles, editors form WikiProjects related to certain subjects. Articles under their scope are tagged as such, and generally given a grade that corresponds to their quality: there are stubs, then start-class, C/B/A class, and then audited grades (GA for Good Article, FA for Featured Article.) A B-class is considered relatively free from grammatical errors, more or less comprehensive, and reasonable structured--essentially a "good enough" article. But most articles, even those shepherded by WikiProjects, are dismal in quality. WikiProject Military History has an impressive 416 FAs, 567 GAs and 187 A-class articles. But in their 90,000 article scope, 94% are either stubs or start-class (chart). Go to other large projects, and the results are similar: for video games, 92% of the 22,000 articles fit this category (source), next to WikiProject Food's similar 91.7% (source). WikiProject Religion's percentage is rather respectable in comparison, with less than 80% stubs or start, but that jumps dramatically when you factor in the articles that haven't even been given a rating at all (more than half its 3,000-odd article scope.) In fact, according to the Wikipedia Assessment Department, the 1000+ WikiProjects have only managed to assess around 1,600,000 of Wikipedia's articles by March 2009. That means that 45% of Wikipedia is of questionable importance and quality.

In short, Wikipedia has existed for more than 8 years. And in that time, we can see the end result is a large amount of junk moreso than polished work that is likely to be accurate. But does this mean the Wiki is a failure? Hardly. What it does mean is that it might be necessary to change parts of the underlying structure in order to prioritize quality, however. Currently a proposal called "flagged revisions" is being demoed; these would flag pages that normal visitors would see on clicking to a page, but leave subsequent unflagged revisions in the background, hopefully reducing the visibility of vandalism or removing the impetus entirely (see here). This is one step in the right direction, but it also results in a fundamental change from an encyclopedia "anyone can edit". However with many controversial pages constantly locked from editing, it's not that far of a step and actually allows for more editing than simple page protection.

Ideally, the project needs more dedicated members working in audited content building and reviewing. The more review featured article candidates and the faster consensus to its status has been reached, the smaller the turnaround time, the less backlog of nominations, and the more articles are speedily promoted. Perhaps something more radical as a freeze on new article creation would allow maintenance backlogs to be cleared and a focus on improving content emphasized. Is Wikipedia doomed to be crushed under its own weight, like some suggest? Even with my glum perspective, I would hardly go that far. But it is likely that as its users and structure evolves, the underlying philosophy that governs Wikipedia might be forced to evolve too.