Otherwise, (that is if popularity is used to identify content), then no, because it is an approximation by statistics, it will work for some cases, but not all cases where statistics is not there, like the long-tail, ect.

The credibility measure is actually “popularity” by experts’ voting.

Good points. Thanks.

However, I think all five points in this post & a couple more besides (like users being able to share their own understanding of what’s relevant!) can also be addressed by developing popularity-based searching within a more dynamic environment. For instance, by letting information develop context amongst other bits of information dynamically, the total information pool spontaneously self-organizes into smaller-scale chunks, and all five points are dealt with just the same.

On oddflower for instance, the context of a page, news item or person on the site (i.e. its ’semantics’) emerges from the traffic flows on the site, so it’s completely bottom-up & potentially even more dynamic, democratic, high quality, credible and advertizing accurate. A bottom-up popularity approach like this should not be dismissed, because it does something that a top-down one can’t; it allows users to develop & share their ’semantic’ knowledge.