Alex Halavais tried an experiment in which he slightly defaced 13 Wikipedia pages. To his surprise, all were fixed within hours. On Alex's blog, I noted that Wikipedia makes a point of looking for damage, and tries hard to fix it quickly. I proposed a mechanism (the "recent changes" list), and I'd like to expand on that a little.
In my comment, I've been recently praised for invoking the magic of RSS as the tool for Wikipedia self-correction. This comes out of a series of conversations about whether Wikipedia is trustworthy or not: see discussions at Techdirt and Many2Many.
I was actually going to blog about something else, but I've become absolutely fascinated1 with the set of tools that the Wikipedia community has developed for social control and monitoring misbehavior. The virtual worlds people may have been right in seeing strong analogies between maintaining social control in the Wikipedia and maintaining control in virtual worlds.
What they perhaps didn't realize is that Wikipedia manages this with a stunningly large number of implicit rules, mores and activities that just aren't externally visible: the Wikipedia "backstage" (to use Goffman's term) is large, complex, and requires a great deal of work.
Update, 9/4/04. Another Wikipedia experiment with subtler changes that weren't noticed. Please do not try this at home!
This article is, I think, still in progress; feel free to comment upon it.
Let's take a look at how Wikipedia does self-correct. While I'm pretty sure that participants in the Wikipedia community know this, I don't think I've seen the process written out much before. So I'll take a stab. I certainly welcome more careful Wiki participants to clarify this discussion...
Start from the beginning. A Wiki is a sort of an online editing space2. Anyone can go to any page and modify it, or create a new page. This becomes a useful tool for brainstorming, for maintaining todo lists or FAQs, and generally for keeping around stuff that other people can fix freely. It's used by classes, by software development teams, and by other groups of people who want to track their collective knowledge.
Of course, it is (by its nature) unstructured. Because they don't have the advantages, and disadvantages, of the temporal order of blogs and discussion boards, Wikis need people to wander through and clean up periodically. This ranges from refactoring portions of articles to fit them into categories, to adding correct links between them, to--yes--correcting damage.
---
Wikipedia is a free online encyclopedia that is basically working on the principle that people reading it will catch errors and fix them. Largely built from the ground up (I don't know what sort of money is behind it), a sizable army of volunteers have written and edited articles that largely seem to explain many issues thoroughly and clearly. It's a pretty good place to look for useful information.
It's also a huge project built by a mob of volunteers, and thus (like much of the open source community) is a large steaming mass of well-organized social capital. Now, I haven't figured out what makes Wikipedia tick--why this particular Wiki Works so very well (and so many fine other wikis don't). There's something to be said for having a compelling concept, and a place for people to show their specializations and interests. But that's not really what I'm into at the moment.
Take it as given, then, that people are there. Some are writing new articles, some are editing articles and refactoring sections of text, and some are correcting errors. Still, the many eyes do not make the problems go away themselves.
---
Wikipedia shows that there are mechanisms for collectively repairing problems and for adding new content. It has the interesting property that any article may be wrong at any moment (because someone may have just damaged it) but the encyclopedia, as a whole, is too big to really thoroughly attack. The WikiGnomes (to use the C2 vocabulary; Wikipedia calls them the RC Patrol )
Most of the time, then, most pages of the Wikipedia is at a (fairly) steady state--that is, they aren't changing at all.
Sometimes, a page comes to attention. World events bring place into focus4, or someone finds a page of particular interest and decides to edit it. It gets tagged as the collaboration of the week, or gets bumped to the front page as a featured article. Or maybe someone just happens to read it, and makes a change. Making changes triggers the recent changes list; that, in turn, drives edits and modifications, and so the page may change rapidly.
At first pass, then, this is a punctuated equilibrium model: catacylysmic changes follow long periods of silence. [Hm. This is statistically testable fairly easily with a couple good SQL queries on change history...]
[download]:http://download.wikimedia.org/
A lot of what I am saying here can be seen quickly in the HistoryFlow diagram labelled "Iraq on Wikipedia - spaced out by time." The authors of the work seem to prefer "spaced out by change" displays, which are very useful -- but lose the niftiness of just how fast Wikipedia corrects. In the "spaced out by changes" display, every change is given equal distance on the X axis, so it's hard to tell when the changes occur, or how often.
In contrast, if you look at the image labelled "spaced out by time", you realize that an entry just sits there for a while -- and then someone does something to it. This often triggers a flurry of activity: revisions, fixes, edits, discussions, which converge on a newer, better page.
This can be largely explained by the existence of an army of readers following the Recent Changes list: and, indeed, Ward Cunningham has discussed how Wikis manage to be fairly resilient because of the recent edits list. Curious readers want to know what's new -- and anything that a malevolent person adds can be caught.
Now I haven't found the Recent Changes RSS Feed for the Wikipedia, but I can't imagine it's too far away. (Or do people just bookmark it and read it regularly?)
I'm being, incidently, horribly unfair here. An overview of everything that's changing might work for a smaller Wiki, but Wikipedia is big. Really big. Three-hundred-odd-thousand articles. The revisions log scrolls 25 changes in two or three minutes. You can't just expect that to all be caught, checked, double-checked3.
The RC patrol and Village Pump make it clear that there's a substantial backstage (to use the Goffman sense) to this Wikipedia thing: it's not just that a smart person wanders by and cleans up, but that a community of smart people are actively arguing and discussing what should happen.
In addition, it's not like the community only watches present events. Here's a couple of the internal tools that are used for internal cleanup on the Wiki:
So who does the correction, or watches the watchers? It's internal. The people who are correcting are other members of the community; roughly, they seem to trust each other to make decisions. Perhaps because everything is logged in the changes log, there is a strong ethic of explaining decisions; the Wikipedia--it seems--would prefer to err on the side of discussion than overaction, and on the side of retaining informaiton than losing it.
The theme I think is emerging is that the recent changes log is a form of internal accountability--but so are a number of other tools, such as user names. In the Wiki panopticon, good behavior occurs, and is seen, in part as a result of its being continuosly visible and commentable. When every page, including every personal user page, has a meta-page (labelled "Discussion"), there is room to comment on anything that happens.
I'm reminded, then, of David Brin's book, The Transparent Society, in which he defends the idea of a society in which no one has privacy: instead, he suggests, bad behavior is minimized when anyone can see anyone else. (Yes, it takes a radical restructuring of social norms).
Has anyone tried creating a panoptic virtual world. where there are no dark shadows or hidden corners? I'm curious what that would be like.
----
(Like everything on this blog, this article is © 2004 Danyel Fisher. Click the "Creative Commons" button to see the copyright notice.)
----
1 _This is turning out startlingly long. Perhaps I could consider upgrade it in a hurry to a CHI paper? It's not like I have anything to do between now and September 13th... and it's dead-on for the theme: "Technology, Safety, and Community" fits VERY WELL into the question of self-regulation for an online community. Too bad I don't have any data. ponder _
Incidently, Alex Halavais has now followed up his experiment with a brief writeup that points to his editing history and the conversation that then happens around him.
2 To blatantly self-promote, one chapter of my book contains an interesting article by Andreas Dieberger and Mark Guzdial entitled "CoWeb - Experiences with Collaborative Web Spaces." In that chapter, they discuss a Wiki in practice.
3 It's probably worth contrasting Cliff Lampe's worth with Paul Resnick on Slash(dot) and Burn which shows, among many other things, that it takes a lot of eyes to keep Slashdot running. An article is read by thousands, moderated by dozens -- and then each of those moderations is metamoderated by five or six. More people are watching the watchers than are doing the watching themselves
On the other hand, on Slashdot, there is only a small number of featured stories at once: the collective attention of Slashdot examines a story for a day or two (the slashdot effect before it wanders away).
4 I would love to see hit counters for different pages of Wikipedia over time. What's the Wikipedia Zeitgeist? (Some of this is in the Wikipedia stats page).
August 30, 2004 11:55 PM | TrackBack | in Data and Documents