Blog
The Neighbourhood
Wed 07 Mar 2007 05:31:33 PM UTC — SamiWhy is it that when you create a blog on one of the popular hosting sites, it already has a reputation even when it's brand new? And why are the reputations clearly bad in some categories? It's because the neighbourhood where you host your new blog also affects its reputation.
Earlier, I discussed how reputations traverse the domain hierarchy. This is exactly what happens here. Your blogging service hosts a huge number of blogs under the same domain, all of which might not be exactly trustworthy. In fact, many of them could be just spam. In our system, the reputations of these thousands of neighbouring blogs affect the reputation of the parent domain your hosting service reserves for its users. And when you create a new blog under one of these domains, it initially inherits the parent's reputation.
In the figure above (see larger) you can see a snapshot of the reputations for some of the popular blog hosting domains. These reputations can be seen as an indication on how the average reader initially views a previously unknown blog. Without any past knowledge of contents or quality, the reader will derive her expectations from earlier experiences with nearby blogs. Obviously, the more popular a hosting service becomes, the more questionable content tends to end up there.
This behaviour in WOT was modelled after the real life. In the physical world you are constantly reminded of your surroundings. It's likely that you'll trust a business located in a good neighbourhood far more than a one in the shady part of town.
Therefore, until we have enough distinct information about your blog, the reputations of all the other blogs under the same domain will continue to contribute to your reputation. Only once we have enough reliable testimonies, we can ignore your neighbourhood. Of course, the best way to avoid this initial prejudice is to register your own domain and host your blog there.


improve real life
"WOT was modeled after the real life", but it's proven that real life is not perfect...
The present rules are good but I think you can have a better service by adding some improvements...
for exemple having 2 different indicators on a domain can be a good idea, the first one indicates the note for the domain and the second the estimated note for 'not-enough-evaluated sub-domains'. This second indicator (more severe) then can have rules like "bad sub-domains evaluations count twice more than good" or an indicator based on the 50% worst evaluations...
maybe i'm wrong but i'm sure we can do better.
evaluate the evaluation
A good idea can be to add some temporal rules to the evaluation of the evaluation...
for exemple, a bad eval on a site 5 years ago and only good since, then the bad eval is probably no more accurate.
we can have also some way to indicate that a sub-domain no-longer exists...