[rdfweb-dev] Paper on Trust Networks on the Semantic Web from MINDSWAP

Jennifer Golbeck golbeck at c...
Sun Nov 24 20:20:17 UTC 2002


> Nice paper -- it's useful to find a range of references to work underlying
> the FOAF idea documented, and a number of other interesting references.

thanks :)

>
> My one issue is the trust metric: there's no evidence I can see that this
> corresponds to actual measures of trust as applied by people. I guess
> that's ultimately a tricky area for social science research; I find I am
> wary of any approach that attempts to reduce trust to a simple scalar
> metric, as that implies a total ordering, which I don't think is the
> case. I think these issues are somewhat acknowledged in your suggestions
> for further work.
>


i *absolutely* agree. a few of the critiques i've received about this
paper describe counterexamples to the metric, places where it could be
broken, etc. this is all true, but my goal here was not to create an
optimal trust metric. rather, i think the trust metric i use works well as
a simple illustration of how compiled metadata can be processed in a way
done for other complex systems.

i'm a PhD student (in Computer Science) and i'm really not interested in
doing a lot of research into social science, the philosophy of trust, or
things of that nature. the space of this that i'll be addressing in my
dissertation is 1) how one can describe non-explicit relationships in a
network, and 2) how effects propagate through such a network. for that i'm
looking at things beyond "trust", describing bigger systems (i.e. food
webs, businesses, and potentially large collections of metadata) with
relationships that are more complex (but less ambiguous) than trust.

at the same time, i think there is more to be done (with metrics in
particular) on the trust side. i had a few others that i wanted to include
in the paper, but they were pretty harshly attacked by a couple people,
so i took them out. hopefully, this one will get accepted at WWW2003 and i
can follow it up with something to really address the math in this
context.


> There's a lot of trust-related work from Morris Sloman's group at Imperial
> College, London. In particular, a useful survey paper they have done at
> [1]. I am also wondering if the techniques you describe could be applied
> to "subjective logic" [2], which some folks are looking at as a way to
> quantify and compose trust.


and i'll definitely take a look at these.

thanks a lot for the comments (and for reading the paper).

sincerely,
jen golbeck




More information about the foaf-dev mailing list