Lists Home |
Date Index |
Bullard, Claude L (Len) wrote:
>If you ask "What was the answer in Copernicus' lifetime?", you will get
>an answer that is right and correct.
>There is a whole other aspect, too. Suppose that you decide that Fred's
>credentials are really in order, to what extent can you believe what he
>says? A person can be untrustworthy on one or many subjects even though his
>identity is well-established.
Which is where the logic side starts to come in handy.
Thing is, a lot of this trust stuff already applies to the existing web,
and it's a problem that won't go away.
1. I came across an interesting-looking flower and had no idea what it
was. I posted a photo of it to my weblog  and within 24hrs someone
that recognised the plant posted its Latin name. I ran Google Image
Search on that name and all the top hits looked a good match for what I
This case I'd say was a success of the "augmented community" - the
communications side of the web provided a possible answer, the implicit
metadata of PageRank was enough to confirm the result.
2. New Scientist ran a story on RSS that was full of factual errors .
A failure both in journalistic terms and for the web as an information
source. Much of the surface material available on the topic is
inaccurate, mostly a result of selective focussing of certain
commentators. That there were factual errors in this piece is not
controversial, however what the truth of some of the matter is depends
on where you look (e.g. three often contradictory accounts of RSS at ).
In the first case it was pretty uncontroversial information, and
PageRank effectively did a head count. In the second, the material
brought up by Google is likely to be biased one way or the other. In
such circumstances a headcount willl show who's making the most noise,
not who's telling the truth.
Personally I'm optimistic that the Semantic Web approach can make the
first situation more general - I wouldn't expect to be able to get
comparable results if I did this every day. But if it was less effort to
tap into specialist knowledge, dip into a newsgroup or whatever then
there's no reason why this might not work more broadly.
I can't see how the errors in the second case could have been avoided
without some kind of trust-related filtering, and that would require
some kind of distributed logical system that could work with the
existing web. Again, Semantic Web technologies look the best bet.
One final point is that no matter how good the trust and information
system, the actions that result may have little bearing on their truth
or validity. The suggestion of weapons of mass destruction is enough to
justify a war - the evidence is orthogonal.