The Problem: Wikipedia displays new content and vetted content with equal value.
The Solution: Color trust gradients.
Widget production makes up most of the economy of Springfield, with an emphasis on high tech silicon based widgets. Silicon widget sector growth has been in the double digits for the years 1999 to 2005. Springfield is also home to several large breweries, which together create 17,000 kegs per year of beer. Excellent soil provides Springfield Yucca farmers bountiful crops. Springfield is also home to several whitewater rafting firms which generate a total of 19% of the tax revenue for the city.
Widget production makes up most of the economy of Springfield, with an emphasis on high tech silicon based widgets. Silicon widget sector growth has been in the double digits for the years 1999 to 2005. Springfield is also home to several large breweries, which together create 17,000 kegs per year of beer. Excellent soil provides Springfield Yucca farmers bountiful crops. Springfield is also home to several whitewater rafting firms which generate a total of 19% of the tax revenue for the city.
Widget production makes up most of the economy of Springfield, with an emphasis on high tech silicon based widgets. Silicon widget sector growth has been in the double digits for the years 1999 to 2005. Springfield is also home to several large breweries, which together create 17,000 kegs per year of beer. Excellent soil provides Springfield Yucca farmers bountiful crops. Springfield is also home to several whitewater rafting firms which generate a total of 19% of the tax revenue for the city.
To generate the accuracy zones, each block of text is analyzed to determine edit boundaries based on page history. For each edit boundary text block, accuracy is measured by an aggregate of the following data:
The trust gradient is all based on the probability of accuracy (PoA) of any given block of text. It is assumed that even with a 99% PoA the text can still be wrong, but is less likely to be wrong than something with 1% PoA. The PoA of a text block increases as users view the text without editing it. As every user has the chance to edit, a user who doesn't edit gives implied accuracy to the unedited sections.
If a user edits something else on the page, every block that is not edited is considered of higher PoA. Registered users are even more likely to edit errors they find, and even more so with administrators. Each successive view and edit increases the likelihood that an unedited block of text is accurate.
If a page is edited frequently it will always have red (potentially inaccurate) items on it, but these items will improve quickly as other text blocks on the page are edited. Blocks of text that have been in place unchanged for months will be mostly white as hundreds of users will have read the content and approved it.
As with any system there will be people who try to break it. While this will be a never ending war, the following could help maintain the accuracy of the trust gradient system.
Wikipedia is a registered trademark of the nonprofit Wikimedia Foundation. I hope that they take this idea and run with it. I only ask that somewhere there's a footnote with my name on it... :)
Will Shaver lives in Oregon where he works as a Software Engineer, and goes by the alias Woil on most websites. You can contact him easily via email as he owns willshaver.com.
If you like this idea, please tell people you know about it so that eventually it'll get to someone who can do something with it. I've submitted this to digg, so you can digg it if you'd like.
June 13th 2007 - There has been some discussion on the digg.com site around this idea vs the existing Wikipedia grading scheme. I want to make it clear that this scheme doesn't replace the existing page level grading scheme, it provides a sentence level scheme in an automated fashion. I'm still all for the existing scheme, this doesn't replace it. What it does help is to highlight updates that follow the editor review. (You could even have every item on the page get a few extra points of accuracy rating when the page receives a good grade.)