Re: SUO: Motion for a joint SUO project (extended)
Matthew West raises an important point regarding
"consistency" that may require more careful
definition of terms.
{Matthew West]
> There is a flaw in your reasoning.
>
> You are (I think) assuming that it is desirable that all theories in some
> SUO product should be consistent with each other. This is not at all the
> case. What is true is that you would want the subset you were using to
> solve some problem to be consistent. However, there are many theories
> that might be inconsistent with each other, but which might be useful
> for different purposes. 3D vs 4D is one that immediately springs to mind.
>
>
It wasn't the intention of the revised motion to create any
preference for logically consistent modules, but to provide a
mechanism for recognizing which are inconsistent with each other.
The proposed motion recognizes that there can be logically
inconsistent modules within a "generalization hierarchy".
I agree with Matthew on this point:
[MW]> What is true is that you would want the subset you were using
> to solve some problem to be consistent.
One of my main motivations in looking for the maximum degree
of "consistency" (see below for a definition) between ontologies
is the problem of communication between them. For a user that
wants to create a solipsistic ontology-driven application that
never talks to another cognitive system, any self-consistent set
of modules from the standard library could be used. When one
wants one's application to interoperate with others, choosing
sets that are consistent with each other would be not merely
"desirable", but perhaps essential. To some extent, the
sharing of research results on systems using representations
of complex concepts should also be more effective when consistent
representations are used. Nevertheless, the comments within
this group make it clear that there will always be some
users who feel that it is necessary to use some ontological
theories that are logically inconsistent with those of others.
That's fine. For me, one purpose of trying to discover the
maximum degree of consistency between ontologies would be
to allow such users to make an informed choice, to be
able to be certain that the benefits of using one ontology
inconsistent with another outweigh the benefits of
increased interoperability between systems.
With regard to the follow-up sentence:
[MW]> However, there are many theories that might be inconsistent
> with each other, but which might be useful for different purposes.
> 3D vs 4D is one that immediately springs to mind.
Matthew considers 3D and 4D views to be inconsistent. What I
had in mind was that 3D and 4D might *not* be "inconsistent" in the
sense that an assertion about a 4D object might not have any
inferences that are logically contradictory to assertions in a 3D
ontology, since there are no assertions about 4D objects in a 3D
ontology. Perhaps "consistent" is not the proper term for such a
relation between ontologies. We may need another term --
"orthogonal" would convey some of the notion that these deal with
different logical things, but is too strong in suggesting that
there is nothing in common.
What I am hoping is that different ontologies that treat the
real world differently in some respect, but do not generate inferences
that are logically contradictory, can be translated into each other
accurately and without loss of information. Depending on the
way they are axiomatized, I think that 3D and 4D ontologies might
be translatable in that way. If they can be translated into
each other, "interconvertible" or "intertranslatable" might
be proper terms. Other suggestions?
I have considered "inconsistent" ontologies as those which
genuinely lead to logically contradictory inferences. These would
have to be treated differently from ontologies that view the world
from a different perspective, but don't create logically contradictory
assertions.
So "consistent" does not appear to be the proper term to
refer to "non-contradictory" ontologies. I'm open to suggestions
about how to refer to these distinctions.
Pat
=============================================
Patrick Cassidy
MICRA, Inc. || (908) 561-3416
735 Belvidere Ave. || (908) 668-5252 (if no answer)
Plainfield, NJ 07062-2054 || (908) 668-5904 (fax)
internet: cassidy@micra.com
=============================================