On Mon, Feb 16, 2004 at 12:46:16AM -0800, Daniel Mayer wrote:
Brion wrote:
On the contrary it's very easy to game: create
a bunch of
sock puppet accounts to give each other feedback. On E-Bay
you'd have to go to the trouble of faking some auctions to
yourself, but nothere...
Yes - that would be bad. Here is a simple solution that goes along with what I
thought trust metrics would do:
There are a bunch of people I trust on Wikipedia - I never check their edits
for vandalism, bad edits, or overt POV because, in my experience, they have
very rarely if ever done any of those things. There are others who I trust in
the same way expect in some categories and on particular articles. And then
there are the people who I don't trust at all.
What I would like is the ability to say that I trust user x. Presto! Edits by
user x are either no longer displayed in a special recent changes list that
only I can see or their edits are turned into small gray text on my regular RC
list (I like this second option better). Once a category system is up and
running I would like to refine that for some users (who I do not trust only
when they edit certain categories of articles). I would also like the ability
to explicitly say that I don't trust somebody - their edits on RC would be
bolded.
Then to create a web of trust I would like to tell MediaWiki that I trust the
opinions of user x. Presto! Any edits by a user on user x's white list will be
either removed from my special RC list or made small and grayed out on my
regular RC list (and watchlist for that matter).
There is very serious problem with this approach - the trust is not transitive
at all, especially wrt POV issues.