> Well you would be correct in the sense of what we
>call 'emergence', where agents at design time seem ok but at run
>time can do things not envisaged. That would be a good agent with
>bad data. However there is the issue of purposely designed / built
>agents who are "bad" that will lie, cheat, etc to other agents. -
>Just like a virus we see today.


Yes - that would be bad. But who would view such agents? Only other bad agents?

It is a kind of Turing Test.

"Is that agent good or bad?"

>
>In your reply I think you were leaning more towards the actual
>information that agents will seek out and manipulate etc on tghe
>Seamntic Web. In that instance, then yes a bad agent would be a good
>agent with bad data. We would probably use some sort of trust /
>reputation system similar to the rating system we see today on Ebay
>where we give users +1, 0, or -1 points depending on our expereinces
>with them.


Yes. I think we as consumers can be critical c.f. "intelligence is the mind of
the beholder"

Gordo

--
Linux User No. 256022///
http://pobox.com/~gordo/
gordon.joly@...///