[Top] [All Lists]

Re: [ontolog-forum] Role of definitions (Remember the poor human)

To: "John F. Sowa" <sowa@xxxxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 14 Feb 2007 12:35:13 -0600
Message-id: <p06230904c1f90156a68b@[]>
>Pat, Chris, Kathy, and Barry,
>The problem of stating necessary and sufficient conditions
>for defining anything is nontrivial, even in mathematics.
>For phenomena in nature or the results of typical human
>behavior, definitive statements are problematical, to say
>the least.
>Belief revision systems, database systems, and knowledge-based
>systems distinguish levels of "entrenchment" (whether or not
>they use that term), and I believe that an ontology should also
>make such distinctions at the metalevel.    (01)

Suppose we were to agree.    (02)

Allow me to sketch what would would then happen.    (03)

First, we would have to decide what the right set 
of "levels" should be. Should there be 5 of them, 
as you suggest? Or 3, as Cyc has chosen? Four? 
Twelve? Perhaps the whole idea of a total order 
is wrong: can the levels be *partially* ordered? 
(There's a PhD thesis lurking there.) After 
interminable debate, most of it informed more by 
raw intuition than anything substantive, we 
might, possibly, more from exhaustion than 
anything else, decide to settle on some smallish 
number. Say five. That decision could easily take 
six months of weekly debate, assuming leadership 
by a strong chairman. If left uncontrolled, the 
debates could be endless.    (04)

Then, we will have to discuss what these levels 
really mean. Some will urge that they should be 
understood operationally: others will insist 
that, being semantic distinctions at root, they 
must be defined semantically or at the very least 
connected to the semantics. The net effect of 
this particular kind of debate is often that the 
few people in the community who are competent in 
formal semantics can hold the others hostage, if 
they choose to be obstructive, or can block all 
progress, if they quarrel amongst themselves (I 
have been in WGs in which both scenarios have 
played out). But suppose that we manage to 
somehow resolve these issues in a way that all 
parties can just accept, with various kinds of 
gritted teeth. This will take another 6 months at 
least, and will probably be like a volcano 
waiting to erupt again at the slightest 
subsequent change or debate.    (05)

Then software must be written which actually uses 
this stuff, involving discussions among a 
community of implementers about the practical 
strategies and permissible interpretations of the 
new standards. This will take another year or so.    (06)

Then every ontology writer has a huge extra task 
to perform. It is no longer enough to simply 
axiomatize one's knowledge - itself a task so 
daunting and tricky to get right that it 
represents the single most solid barrier to wide 
deployment of the technology - but one must also 
provide an elaborate meta-theory of it. People 
will not do this consistently, which will create 
new problems of interchange and interoperation. 
And it solves none of the old problems, which we 
already have.    (07)

Finally, and worst of all, there are clear 
pressures to not use this markup in the way 
intended (even if we can somehow get this clear), 
since if an honest person publishes an ontology 
which is carefully and accurately meta-marked-up, 
their work can be nullified (on an open network) 
by someone else who adopts an elementary 'cuckoo' 
strategy of publishing a rival ontology in which 
the meta-markup declares everything to have level 
1, thereby causing truth maintenance engines to 
destroy other ontological content with which it 
is inconsistent. The possibilities for hostile 
takeovers, moles and 'semantic viruses' are 
clear. In fact, they are so clear that I venture 
to predict that if this ever were deployed, it 
would immediately be hidden behind security 
firewalls so opaque that it would be invisible.    (08)

Pat    (09)

>  Following are some
>"levels of entrenchment" in descending order of strength:
>   1. Type hierarchy.  The classical tree or partial ordering
>      introduced by Aristotle and first drawn by (or attributed
>      to) Porphyry.  It's useful in every field, it's not going
>      away, and we should recognize it as the minimal requirement
>      for an ontology.
>   2. Necessary distinctions.  The differentiae that split any
>      type into two or more subtypes.  If the split is binary
>      (A or not-A), then it is both necessary and sufficient for
>      distinguishing the two subtypes from one another, but the
>      conditions for characterizing the supertype might not be
>      necessary and sufficient.
>   3. Constraints.  Additional statements that characterize the
>      types or the interactions of entities of various types.
>      The constraints are necessary relative to the ordinary
>      facts in level #4, but they might not be considered
>      defining characteristics.
>   4. Ordinary facts.  Ground-level assertions that must be
>      consistent with statements at the above levels, but they
>      may violate defaults at level 5.
>   5. Defaults and probabilities.  Statements that are usually
>      true of entities of a given type or types, but they are
>      at the bottom of the entrenchment pole.  A probable
>      statement is a default with an associated value that
>      indicates its likelihood or frequency of occurrence,
>      given the occurrence of some other condition.
>Systems of entrenchment levels along such lines are widely
>used and should be supported.  Cyc, for example, has 3 levels:
>True, true by default, unknown (and the negations -- false
>by default and false).  But I think that Lenat would agree
>that a privileged level should be added for some of the
>axioms, especially ones that define the type hierarchy.
>A declaration of which level a particular statement belongs to
>would not be part of the first-order theory, but it would be
>a metalevel statement that should definitely be considered
>part of the ontology.
>Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/ 
>Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/ 
>Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
>Shared Files: http://ontolog.cim3.net/file/
>Community Wiki: http://ontolog.cim3.net/wiki/
>To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>    (010)

IHMC            (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.    (850)202 4416   office
Pensacola                       (850)202 4440   fax
FL 32502                        (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (011)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (012)

<Prev in Thread] Current Thread [Next in Thread>