Re: Knowledge Databases LO1927
Tue, 4 Jul 1995 04:49:17 -0400

Replying to LO1903 --

To the members,

In my intro (Intro - Bo Newman LO1909) I went on at some length about the
need for knowledge management. At the time, I had only received a handful
of postings and had not been able to get to the archives. After further
readings I see that I was preaching to the choir. Please excuse my
ramblings. After talking to so many deft ears for so long, these are very
welcome discussions.

As I interpret Michael McMaster's remarks on knowledge data bases ("Re:
Intro -- Ian Hosking LO1866"), I take his position to be that (1)
Knowledge is the integration of information into whole patterns that are
(2) interconnected with other patterns, and (3) that it is context
dependent as related to it's ability to support actions. In reply, (Re:
Intro -- Ian Hosking LO1878), Phillip Capper suggested that we,
"...consider the processes by which knowledge becomes contextualised (and
also RE-contextualised - that is transferred between contexts by the
possessor of the knowledge), situated and distributed in an organization."

To expand on these thoughts, I would suggest that it is possible to
represent these integrated interconnected patterns through such constructs
as networks of connected acyclical graphs. In doing so it would be
important to insure that the model is fully dynamic and not dependent on a
priori definitions of the knowledge structure, and second that it be
domain and language neutral. In addition, it is should be possible to
link such a model/representation to knowledge utilization events through
applicable business processes to cause actions or, to reflect changes
resulting from these actions in the knowledge network.

Phillip opened his remarks saying that thinking about knowledge as
something that can be captured in a "knowledge base" seems to be the
antithesis of systems thinking. I would ask that if a system of knowledge
representation did in fact support knowledge definition through relational
integration and contextualization, could it not serve as a tool useable in
the scientific study of how people interact with knowledge?

In Jim Michmerhuizen's reply (Knowledge Databases LO1903) he pointed out
that there are deep logical incoherencies in the notion of using current
[relational] database technology to support a "knowledge" database and
that it is logically impossible, in such a structure, to represent
emergent knowledge. I whole heartedly agree, but it is my suspicion that
a hybrid of object orientated representation and context sensitive
predictive agents could.

Recently, I have had the privilege of working with some folks who have
imbedded mnay of these features in working software and I have been very
impressed. They have even be able to address the contextualization issue
by including definable context based subsetting rules in the generation of
user specific views of the knowledge. To date they have not be able to
demonstrate the recogition of emergent knowledge but I can't help but
think that it is only a matter of time.

Bo Newman
Cogito Inc., Richland, WA (USA)
(509) 943-5500