Knowledge Databases LO1903 was: Intro -- Ian Hosking

Jim Michmerhuizen (
Sat, 1 Jul 1995 00:05:31 +0059 (EDT)

Replying to LO1866 --

On Thu, 29 Jun 1995, Michael McMaster wrote:
> If the "knowledge data base" is attempting to get at this in a
> pragmatic way, then there are some things that we can do. I think
> we'll have to use principles of complex adaptive systems to be
> effective and these are far from well developed. However, there are
> some directional possibilities. One of these is to build a system
> that creates pathways to possible patterns of knowledge based on the
> request/demand for that knowledge rather than its "already"
> existence.
Like threads in a discussion group...

> We need to turn our attention to design of connections and what calls
> the knowledge forth without having any idea of what that knowledge is
> or what it will be when called forth.

Yes. There are deep logical incoherencies in the notion of using current
database technology to support a "knowledge" database. A relational
database - literally - stores "relations". The foundation theories are
those of predicate calculus and N-adic relations; a good analogy is an
english sentence with blanks substituted for some of its words. The
sentence then becomes a template for a whole host of possible statements,
and the corresponding relation stores, in each of its records, a possible
set of values for the sentence.

But it is logically impossible, in such a structure, to represent emergent
knowledge. Either the database fields will be used to represent merely
syntactic features of the knowledge (in the guise of "how it connects"), or
the knowledge itself will get chopped and diced to fit into the procrustean
bed of the relation as defined.

     Jim Michmerhuizen
     web residence at
. . . . There are far *fewer* things in heaven and earth, Horatio,  . . . .
 . . . . .       than are dreamt of in your philosophy...        . . | _ .