Re: clarifying clarifying ontologies

phayes@cs.uiuc.edu (Pat Hayes)
Message-id: <199508071848.NAA10374@eris.ai.uiuc.edu>
X-Sender: phayes@tubman.ai.uiuc.edu
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Mon, 7 Aug 1995 13:48:12 -0500
To: hovy@isi.edu (Eduard Hovy), phayes@cs.uiuc.edu (Pat Hayes)
From: phayes@cs.uiuc.edu (Pat Hayes)
Subject: Re: clarifying clarifying ontologies
Cc: fritz@rodin.wustl.edu (Fritz Lehmann), cg@cs.umn.edu, doug@csi.uottawa.ca,
        srkb@cs.umbc.edu
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

At 11:32 AM 7/28/95 -0700, Eduard Hovy wrote:
>At 4:58 PM 7/28/95, Pat Hayes wrote:
>>
>>I agree that Joe shouldnt need to be bothered by issues like these. But he
>>HAS to be, because if he is careless about them, he is going to get
>>ensnared in inconsistency and (what seems like) paradox. ....  Joe
>>Blow might not want to be concerned with all this, but he doesnt have the
>>option, because even apparently quite ordinary, harmless assumptions turn
>>out to committ him to taking a position in some "rarefied" debate.
>>
>>I think you have the idea that there is a kind of good robust common-sense
>>middle ground which we ought to be able to get clear, and then there is a
>>lot of exotic fussing which is of interest only to mathematicians, and is
>>just a kind of intellectual decoration, angels-on-pinhead stuff. But this
>>isnt how it works. The middle-road stuff DOESNT WORK PROPERLY....
>
>
>It would be nice if there was a robust simple middle ground though, wouldn't 
>it, in which one could operate at a tolerable level of ambiguity, and only 
>descend into the exotic-fussy world on (rare) occasion.  It rather seems to 
>me this is the way I myself work, and this is the feeling behind Fritz's 
>comment.  If your claim is indeed true that operating on the "middle ground" 
>just doesn't work for Joe Blow, then how have the Joe Blows (and me and my 
>family and probably all the other people who ever lived) managed to survive 
>for so many centuries without working out all the fine details?  And why do 
>we find it so difficult to work them out when we try now?  

Maybe we are misunderstanding one another. Just for the record, lots of us
who I imagine you would classify as 'exotic-fussy' (anything to do with
logic, maybe?) have been trying to operate at this midle of the road for
years. Of course it would be nice. Lets see you do it.

This sounds like someone who says that they have been seeing things for
years without any trouble, so why should machine vision be so concerned
with all this geometry and statistics and other fussy exotica? The point
being that it is one thing todo something, but quite another to understand
how we do it. Otherwise AI would be a cinch.


>
>I suppose I am asking: why is it so scary to operate with a system that 
>in its closure is ambiguous/incomplete/tolerant of paradox?  Clearly if 
>the system's paradoxes surface too often the system is more of a nuisance 
>than a help, but if they surface seldom enough I'd be quite happy (taking 
>my cue from Joe Blow and all the rest of humanity).  
>
The problem is, what confidence can you have in any of its conclusions if
it is inconsistent? From a paradox you can correctly infer ANYTHING, and do
so without any explicit contradiction 'surfacing'. 

I agree, deduction does seem too 'stiff' and rigorous as a model of
intuitive reasoning, but I dont know of any alternative which doesnt. (For
the record, fuzzy logic (for example) is even 'stiffer' in this sense.)

Pat

======================================================================
Until September: 
1916 Ivy Lane, Palo Alto, Ca. 94303               phone (415)855 9043

phayes@cs.uiuc.edu