“One of the mysteries of scale is that there’s no such thing as scaling well,” said Clay Shirky, who writes about culture, media and technology. “You can make something 100 times bigger, and if it works, you think you’ve got it licked. But the next power of 10 can kill it. So I don’t know whether or not openness and co-creation are incompatible at Wikipedia scale.”
That's a funny thing to say, as there certainly are systems that scale well. One example would be a relational database system. By using a binary tree index or something similar, it offers logarithmic search times, so massive amounts of data can be queried efficiently, and multiplying the quanity of data by some factor has only a relatively minor effect on performance.
Whether Wikipedia specifically could be designed to be infinitely scalable is another question. I have argued that this could be accomplished using a P2P architecture (http://www.allpeers.com/blog/index.php?p=14). Even things like moderation that require human intervention can be designed in a scalable manner a la Slashdot (http://www.allpeers.com/blog/index.php?p=16). Both of these approaches entail significant technical challenges, but to reject scalability out of hand as some sort of chimera is a bit over the top.
Fair enough, but doesn't the example of Slashdot (which has hundreds of thousands if not millions of users but only a handful of moderators) show that technology can be used to achieve social scale as well? I find a lot of the Wikipedia discussion to be surprisingly technology-free, focusing on manual (and thus unscalable) workflow-related practices to improve quality. Meanwhile, companies like Google and Amazon are practically printing money by using automated techniques to filter content.
I'm convinced that Wikipedia's woes could be solved by using an automatic moderation system. Imagine, for example, that each Wikipedia article is assigned a score from the community based on a weighed average of individual ratings. Once an article reaches a specified threshold, it becomes the canonical version of that article (i.e. the one that is retrieved by default). In the article header, there could be a note that a newer, but lower-rated version exists. If enough people access that new version and rate it up, it becomes the canonical version. If enough people rate it down, the changes are discarded. Already, this would make it possible to draw valuable distinctions between the probable quality of individual articles.
Actually making something like this work would require a lot of tweaking and experimentation, but given the evident high level of interest in the issue of Wikipedia's quality (or lack thereof), I'm surprised that there hasn't been more brainstorming on this topic.
1. Matthew Gertner on January 11, 2005 4:30 AM writes...
That's a funny thing to say, as there certainly are systems that scale well. One example would be a relational database system. By using a binary tree index or something similar, it offers logarithmic search times, so massive amounts of data can be queried efficiently, and multiplying the quanity of data by some factor has only a relatively minor effect on performance.
Whether Wikipedia specifically could be designed to be infinitely scalable is another question. I have argued that this could be accomplished using a P2P architecture (http://www.allpeers.com/blog/index.php?p=14). Even things like moderation that require human intervention can be designed in a scalable manner a la Slashdot (http://www.allpeers.com/blog/index.php?p=16). Both of these approaches entail significant technical challenges, but to reject scalability out of hand as some sort of chimera is a bit over the top.
Permalink to Comment2. joe on January 11, 2005 10:40 AM writes...
Being a database geek I'll have to agree with you Matt.
Permalink to Comment3. Ross Mayfield on January 11, 2005 11:37 AM writes...
Guys, architectural scale isn't the issue -- it's social scale.
Permalink to Comment4. Matthew Gertner on January 11, 2005 1:48 PM writes...
Fair enough, but doesn't the example of Slashdot (which has hundreds of thousands if not millions of users but only a handful of moderators) show that technology can be used to achieve social scale as well? I find a lot of the Wikipedia discussion to be surprisingly technology-free, focusing on manual (and thus unscalable) workflow-related practices to improve quality. Meanwhile, companies like Google and Amazon are practically printing money by using automated techniques to filter content.
I'm convinced that Wikipedia's woes could be solved by using an automatic moderation system. Imagine, for example, that each Wikipedia article is assigned a score from the community based on a weighed average of individual ratings. Once an article reaches a specified threshold, it becomes the canonical version of that article (i.e. the one that is retrieved by default). In the article header, there could be a note that a newer, but lower-rated version exists. If enough people access that new version and rate it up, it becomes the canonical version. If enough people rate it down, the changes are discarded. Already, this would make it possible to draw valuable distinctions between the probable quality of individual articles.
Actually making something like this work would require a lot of tweaking and experimentation, but given the evident high level of interest in the issue of Wikipedia's quality (or lack thereof), I'm surprised that there hasn't been more brainstorming on this topic.
Permalink to Comment