Re: performatives and trust

>From: "pat hayes" <phayes@ai.uwf.edu>
>
> > After all, there IS a
> > truth of the matter, right?
>
>I imagine so .. but nobody can knows it, nor will they ever.  I would hate
>to burn such an unknowable metaphysics into the axioms of our thinking.
>Especially because we don't have any particular reason to.  You calculate
>your truth, I'll calculate mine ...  thank you just the same.

I didnt say there is an absolute metaphysical truth of the matter, 
and Im not talking about burning any assumptions into our thinking. I 
meant it in a much more mundane sense, as when we speak in normal 
life about 'fact of the matter', eg either the motor bearings are 
shot or they aren't. If we all take a relativistic view of truth in 
this sense we will rapidly get to the point where we are trying to 
walk through walls.

> >The world is the way it is, in fact.
> > Logic is quite capable of shrugging its shoulders: think of (P or not
> > P) for example, which is about the purest expression of "I dunno"
> > that one could wish for. (The often-expressed feelings that talk of
> > truth is overly left-brainish, or something, seem to me to based on
> > misconceptions about what logic means by 'truth'.) But in any case,
> > if one really wants to be able to express intermediate truthvalues
> > there are any number of formal logics that will do that for you, at
> > more or less cost in clarity, as you know.
>
>Intermediate truth values are only part of the problem.  How about
>constructing a system that know where to apply logic and where to step away
>and apply some other heuristic.

I might find that more convincing if I had any idea what the other 
heuristic might be. Sure, an agent might do all kinds of things with 
logical sentences. But if they are interested in accessing content, 
its helpful to know that what the sentences claim is that they are 
true, ie that the world conforms to what they say about it. Thats all 
that 'truth' means here; its the distinction between truth and 
fiction, not the distinction between Absolute Truth and mere opinion. 
Its all opinion in this sense, sure.

> >Just because the axioms are located on a website
> > somewhere and accessed via a URI doesn't make them ABOUT a different
> > world. And so why would there be any less of 'truth of the matter'
> > with web-based reasoning than with any other reasoning? That matter
> > is the same, whatever the location of the axioms that refer to it.
>
>I agree.
>
> > I know yours is a very popular view, but I see no actual evidence for
> > it, and retain a very strong cynicism about the more, er, social
> > vision of 'agents'. As far as I can tell, that word 'agent' has a
> > collection of technical meanings which have virtually nothing to do
> > with one another, and the only technical meanings of 'agent' that
> > make sense in the SW/DAML context are completely disjoint from the
> > meanings that have any social relevance (chatterbots, animated faces
> > and other media-labbish stuff.)
>
>Forget about all the hype .. there still is a very good reason that we need
>agents on the web to function in a social context.   The reason is that the
>relevance of information retrieved increases dramatically if it is retrieved
>from an agent that shares a context with the retriever agent.  Recently I
>was helping my son with a science project .. and I needed an easy way to
>make a vacuum.  So I went to all the usual suspects Google, AltaVista,
>Copernic and did the best I could.  I got zip, zilch that I could use.  So I
>went to McClendon hardware where you can ask these hardware guru's,  who are
>right there playing with the kind of tools I needed, how to easily make a
>vacuum in a jar.  Bingo .. got it on first hit.

I can't help noting that this all sounds like you and they are 
agreeing on a notion of truth, at least when it comes to vacuums in 
jars :-)

>So that is primarily the
>reason I want my Internet software agents to be social.  They correspond,
>exchanging bookmarks and triples with people and agents that are working in
>the same area and context in which I am working.
>
>So you see it is not the social life of the agent itself or it's manners
>(you examples are hilarious, Pat) that we are concerned with here;  rather
>we are concerned with how we use the social transactions of the owners of
>the agents to establish context.

OK, fine. I agree with all this: there is a kind of over-arching 
human social context surrounding Web activity of all kinds, and often 
it may be useful or maybe essential to take that into account. But 
that doesnt require software agents to interact *with each other* in 
ways that are anything remotely like social transactions between 
human beings, is my point. And there really are a lot of serious 
people out there who claim the opposite, believe it or not.

Pat

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Friday, 8 June 2001 11:32:02 UTC