[foaf-dev] FOAF-based whitelisting project

R. Steven Rainwater srainwater at ncc.com
Mon Mar 12 16:17:02 UTC 2007


On Mon, 2007-03-12 at 08:36, elw at stderr.org wrote:
> A unidimensional trust metric is often not particularly useful.
> <...>
> Better would be to define terms that allow you to specify terms
> and degrees of trust, as relates to particular domains of 
> discourse or domains of agreement, such that I can trust your
> opinions or attitudes in domain #1 but choose to think you're
> full of hoo-ah in domains #2, #3, #4.  I think this is a better
> fit with FOAF/RDF/SW technologies in general, though it does
> demand a little more definition of vocabulary terms....

One thing we learned on Advogato over the years was that a
unidimensional trust metric is about the most complex thing an average,
uninterested website user will actually use. When presented with forms
for inputting lots of different topical trust data about lots of people,
most users just ignore it and move on to things they find more
interesting. 

We also found that negative trust always got abused by people who
realized it could be used to punish those they didn't like or didn't
agree with. We never found a way to make a negative trust value useful
without the false mistrust ratings being harmful. A number of people
will also abuse positive trust, but the only result to date is a
relatively harmless problem we call "cert inflation" on Advogato (e.g.
someone who should probably be rated at apprentice level gets rated at
master level). 

I'm not saying it wouldn't be interesting to have lots of complex
topical trust data, just that you shouldn't expect average website users
to understand the idea, or be willing to manually input the data. 

In any case, with regard to creating a whitelist, the only domain of
trust I'm interested in is whether or not I can trust that a given email
address does not belong to a spammer.

-Steve




More information about the foaf-dev mailing list