On Thursday 26 May 2005 08:41, David Gerard wrote: At first blush, it would make sense to rate articles with respect to the criteria of what makes a good article as documented on: [1] http://en.wikipedia.org/wiki/Wikipedia:What_is_a_featured_article [2] http://en.wikipedia.org/wiki/Wikipedia:Featured_articles So it would be nice if the ratings captured whether it was well written, an appropriate reading level, an appropriate size, NPOV, the appropriate use of references, etc. However, that said, I don't think it should include anything that could be done by machine. (Lih's (2004) quality (rigor, diversity), Newberry's (2004) mass and luminosity, and Emigh's & Herring's (2005) formality don't seem to apply in this case of users' subjective ratings.) So for example, that would remove the reading level and appropriate size which could be generated automatically. Also, it should be kept relatively simple. A single subjective rating informed by [1,2] might be good enough. On second thought, the question is very relevant to some of my experiences at the W3C. Since we advocated valid HTML, it was embarrassing that some of our pages were not valid HTML. So after time we began generating reports of the most popular invalid pages, and who owned them. This enabled us to drastically reduce the likelihood of the public encountering an invalid W3C page. So, it would be really interesting to see what are the most popular stub articles. (This to could be generated automatically from referrer, but can also be used so as to find the most popular poorly rated articles once we have that data.)