Re: [OpenCyc-devel] Plug into the framework

[Murray A. <m.a...@op...>]

Stephen Reed wrote:
> Let me respond with a few simple thoughts and then you can ask for
> clarification where needed.
> 
> As a rule of thumb, commonsense is the knowledge we all have but it not
> necessarily recorded in reference books explicitly.  For example, it is
> somewhat difficult to find an encyclopedia or dictionary that says that
> when a bowl is inverted, its contents run out.  In addition to the broadly
> applicable information that *is* found in reference books, Cyc has many
> hand entered facts required to understand the assumptions that underlie
> human discourse.
> 
> Commonsense representation and reasoning in the Cyc Knowledge Base has the
> goal of avoiding the brittleness observed when scaling up typical expert
> systems to include more knowledge.  Cyc is engineered to eventually have a
> suitable representation for the full range of human expression, so that
> expert knowledge bases can be created by extending from the Cyc upper and
> middle ontology.
> 
> Our methodology assumes that representing new knowledge is much easier
> when a large body of general purpose knowledge is already present.  Cyc
> uses inheritance hierarchies within its major object types (e.g. terms,
> relationships and contexts) to concisely represent knowledge.

Steve,

I can't speak for the earlier questioner, but my gist of that question
had more to do with the *boundaries* of common sense. Your example of
a relatively empirical physical law such as gravity is doubtfully
going to come under much criticism, but there are many things in the
Cyc ontology that might not be considered by all of humankind to be
"common" in their thinking or sensation. Societies differ sometimes
in very profound interpretations of whatever "common" reality may be
there to share.

At the one extreme, common sense might be considered common to all
people. [One might need to constrain this statement to apply only
to "sane" people, but there's the beginning of the rub, eh?]

At the other extreme, it might be stated that common sense exists
only at an individual level, that taken as a whole, unless an
individual agreed with all statements in the Cyc ontology, their
"common sense" disagrees with the "common sense" of Cyc, and any
usage of an unmodified Cyc under those circumstances could be
considered coercive, to the degree of such disagreement. [I'm not
suggesting that Cyc or its developers are coercive, only such a
usage, forced or voluntary. E.g., I use music CDs or software I've
purchased even though I disagree with the licensing of it, since
I often have no choice otherwise except to opt out completely.]

The only way to avoid coercion is to allow for the modeling of
disagreement. So rather than *necessarily* modifying Cyc upon the
acquisition of new content, if such change were to cause any
disagreement within the community of use it seems it would be
better to include both conflicting statements, and include a
relation establishing the disagreement.

I'm curious as to:

   a. the design of Cyc that might allow disagreements to be
      clearly modeled, viewed, discussed.
   b. the way the Cyc team sees its release into a community of
      users, and how they think the team might expedite Cyc's
      multiple-viewpoint usage

Because most of the world's problems arise out of disagreement
or misunderstanding, it seems that systems that can model such
disagreement are profoundly more valuable than those which state
a monolithic view of reality, a "common sense". Acknowledging of
course Cyc's intended use in projects where its more "common"
statements would be still valuable, if Cyc could be used in some
way to assist in discussion in the resolution of disputes, of
where communities disagree, that would be much more valuable (to
the world community at this juncture especially), than "knowing"
about gravity, the containment of liquids, the structure of
taxonomies, etc.

Murray

......................................................................
Murray Altheim                  <http://kmi.open.ac.uk/people/murray/>
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK

    Hunt the Boeing! And test your perceptions!
    http://www.asile.org/citoyens/numero13/pentagone/erreurs_en.htm