SUO: *Date 05 Mar 2002 -- Systematic Errors In Ontology
¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤
JA = Jon Awbrey
JS = John Sowa
MP = Mike Pool
Item 1. Systematic Errors In Ontology
I have begun an examination of sytematic errors in ontology.
Here is the outline that I have formulated up to this point:
Ontology Of Systematic Errors Affecting Ontological Projects.
1. Insufficient Reality Testing.
1.1. Lack of respect for the complexity
of the phenomenon or the problem
that one is addressing.
1.1.1. Procrustean Monadomania.
Trying to force relational realities
into non-relative monadic categories.
1.1.1.1. Absolutism.
1.1.1.2. Essentialism.
1.1.1.3. Invariantism.
1.1.1.4. Objectivism.
1.1.1.5. Universalism.
When I refer to a point of view as an "ism",
I do not mean merely that it emphasizes the
importance of the declared aspect of things,
but that it stresses it to the exclusion of
of the complementary dimension of existence.
For example, I have no fixed position myself with regard to whether
absolutes, essences, invariants, objectivities, or univerals exist.
I think it is very likely that they do, I just do not know of any
for sure, and most of those that I have seen put forward do not
stand up under examination. Aside from all that, it appears
to me that such things become evident only toward the end
of an inquiry, not at the beginning, and so I think that
we mostly discover them à posteriori, rather than being
able to claim them à priori. From this point of view,
then, we ought to treat all such categories and all
such categorical claims as theoretical hypotheses,
and yes, that includes this one, too.
Item 2. Content, Form, Function
JA: I meant to get back to these observations that
John Sowa made, because I think that they are
critically important to the proper conception
of our task:
JS: This raises the question of what we can standardize if
we are not standardizing content. The answer is form:
we should be focusing on formal ways of organizing all
the ontological resources that are available, including
SUMO, OpenCyc, IMPS, WordNet, EDR, etc.
JS: IFF is an effort in that direction, but it has not been tested on
actual content and applications. Having at least two rich content
ontologies, such as OpenCyc and SUMO, should bring to the fore the
question of how they can be mixed, matched, and used together.
JS: I would like to open up all the other sources
of content for consideration in this mélange.
JA: I would add these observations.
JA: Drawing the distinction between form and content is important
because it is the line between what is our business and what
is not our business. We simply have no business trying to
redo the work of content experts and domain specialists.
MP: I would appreciate it if you (Jon) and/or John Sowa could flesh out
this alleged content-form distinction a little bit; I'm afraid that
I don't find the word processor analogy sufficiently helpful.
Mike,
I will try to flesh out the way that I see the issues here,
since I don't yet know how well I got what John was saying.
I have to shift over to the Ontology list, but we can save
anything that develops, worth reporting out, for tomorrow.
I have been reflecting on my experiences with "content experts" in a wide variety
of fields, from times when I used to do all sorts of comp/dat/stats work in many
different academic, applied, education, and research settings, and I am trying
to focus on thinking about what it was exactly that those folks really needed,
and what they clearly did not need, and along with my ruminating about that,
I am trying to diagnose the gaps between the varieties of "solutions" that
I learned at school and the actual problems that anybody had in practice.
The Concept Processor ~~ Word Processor ~~ Statistical Package metaphor is
one that I have always used as a way of drawing a line between the "content"
and the "architecture" of formats and functions that make up the system that
one uses to work with this content. The taxonomies that "contented" users or
domain experts already have on hand, or ever plan to fiddle with, are like the
documents that we use a WP to write or the data sets that we use a SP to analyze.
The basic point is that end users do not need us to create the domain ontologies
for them -- they will treat that as so much dumb clip-art -- since they take quite
a bit of fussy and even indignant pride in being the ones to do that for themselves.
The rule here is not "keep it short and simple" (KISAS) but "knowing your stuff" (KYS).
And I can tell you that, when it comes to topic areas that I myself have spent 2 or 3
decades studying, like graph theory or semiotics, the efforts of ontological novices
to dictate axioms is just bound to be somewhere between infuriating and ridiculous.
And as far as some of the vacuous concepts that we have seen being bandied about
the forum here, like "abstract", "entity", "physical", "top", well, ...
So if we are going to have any use at all here, it had better be something else.
One of the things that frequently comes to my mind in this connection is the task
of finding "primitives more primitive than the primitives in previous prehensions".
These "underwhelming primitives" (UP's) would be called "upper predicates" by some,
but to my way of thinking one always finds them by digging deeper, into the grounds
of assumed solidity that most people are presently taking far too much for granted.
But this is a very difficult analytic, critical, reflective task, one that requires
very different kinds of logical utilities than most people would have in mind under
the run of the mill variety of theorem provers running about the logical farm today.
MP: To my mind there's not much in an ontology that is strictly form.
I am guessing that this view can only arise from not
having seen a sufficient number of formal variations.
MP: I would assume that logical operators and equality could
be safely included and without too much controversy, ...
I bite my tongue ...
MP: except from the set theory purists, we could probably include
notions of membership and subset on the form side. We might
also include the most general relationship categories, e.g.,
BinaryPredicate, BinaryFunction, and the vocabulary and rules
for articulating constraints on the arguments or domain and
range of a relation, but I think most would contend that
we've already moved into content here.
I am experiencing a serious disconnect here.
Let me see if I can figure it out later on.
MP: Does your position (and/or John Sowa's) on the SUO simply amount to a claim
that the IEEE ontology standardization effort should cover only the logical
operators, equality and basic notions of set theory? If so, I'll go one
further. These notions aren't in need of standardization, so if you're
correct that we should stay away from content I would contend that we
should just pack in the whole thing.
Okay, by "form" I mean things like "architecture of form and function",
that is, the formats that are made available for representing knowledge
and the functions that we can apply to the expressed models or theories.
In the analogy, like "style sheets" and buttons on the various tool bars.
There is a vast array of conceivable themes and variations in these things
that are hardly yet dreamt of in Horatio's contemporary onto/logical schemes.
Let me know if that's making more or less sense.
Jon
MP: I feel like my question is a bit naive but I really don't know what
the "concentrate on form" proposal amounts to. Perhaps this has been
addressed in one of John Sowa's emails or papers. If so, please point
me there and I apologize for missing it on the first go round.
MP: Maybe your answer(s) to this will help me to understand what
Jon means by "generic functions of conceptual work" below.
JA: To invoke the analogy between "concept processors" and word processors,
our job is analogous to writing the specifications for a word processor,
our is not the business of supplying ready-made term papers to students.
JA: Most of the domain knowledge that actually exists in the world is not yet
in the above short list of "ontological resources" and much of the content
that does reside in some of these sources would be just plain ridiculous to
any specialists actually working in the supposedly covered domains.
JA: So I think that the list of "other sources" and the overall mix of
ingredients in this mélange would need to be widened considerably
beyond what's currently contained in these putative resources.
JA: To advert a potential misunderstanding that I can already see coming
down the road, I am decidedly not saying that there is nothing left
for us "meta, generic, abstract, philosophical" folks to do once the
lion's share of content and domain cutlets have been carved off the
corpus of the world's knowledge beast. What is left is precisely
the form, and I would add the function, and those are the aspects
of knowledge evolution that we should be working to facilitate.
JA: Again the analogy with word processors;
the real design and spec questions are:
JA: 1. What are the generic forms of conceptual content
that concept users need to be supported in using?
JA: 2. What are the generic functions of conceptual work?
Which functions can be given supporting utilities?
JA: I think that those are the sorts of services that
a standard ontology utility ought to be providing.
¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤