Re: Emergent Learning LO2082

Carol Anne Ogdin (Carol_Anne_Ogdin@deepwoods.com)
13 Jul 95 4:55:11 EDT

Replying to LO2051 --

Replying to LO1984, Grant Harris wrote, in LO2051 --

> > I'll agree that you won't get a database (or DBMS) to *synthesize*
> > that wisdom, but I assert that not only *can* you store this kind
> > of knowledge (once articulated by a human being) in a database...
> > you already have.

> I have to lean in Michael's direction. I would venture to say that
> knowledge is something that happens in a person's mind. Information is
> something that you store in a database. Such information may help to
> create knowledge, but I have yet to see a database system that does
> anything more 'knowledgeable' than store and manipulate symbols.

Well, counter-exampler that I am, let me see if I can conjure an example
of *knowledge* gleaned from a database. What I have in mind is the
PERT/CPM chart for a large construction project we're involved in. It's a
$2 Billion (US) effort, will take 18 months, and the PERT/CPM chart is on
the order of 6000-8000 activities.

(For those of you unfamiliar with PERT/CPM: It's a project planning
scheme where you define all the activities in a network, connected
together by precedence: For example, the activity of erecting the
walls must be preceded by a large activity called "lay floor," and
lay floor has multiple activities like "pour concrete", "let concrete
cure".)

Now the knowledge of how long each activity takes in time, and the other
resources it will require, must come from people. However, through the
process integral to the PERT/CPM model, we glean new knowledge (not just
information) in the form of the critical path, and for each activity, how
far off the critical path it is.

(Again, by way of explanation: The "critical path" is the computed
list of activities that directly affect the schedule; that is, activities
on the critical path that are delayed by a day will delay the final
completion of the project by a day. An activity that is NOT on the
critical path has "slack," the number of days it could be delayed
without becoming part of the critical path. The object of this exercise
is to determine, dynamically, as activities are completed, which
items are critical for management to attend to, and which are of
lesser importance, and which are moving closer to the critical
path and might become critical if we don't attend to them now.)

I contend that this is "knowledge" about the project that is informa-
tive and useful, and of higher precision and with more "audit trail"
that informed judgements of experienced construction project
managers. To simply call it "information" (that is, data with meaning)
I think devalues the fact that we have *discovered* something
about the entire project that is useful and generally not otherwise
available except via that system.

Let's apply Turing's test: If we had access to an expert in the
business of project scheduling who could give us answers with
the same precision, accuracy and speed, would we call that an
"informed" expert, or a "knowledgable" expert? I submit that we'd
use the latter. So, if the "knowledgable" expert produces know-
ledge on demand, what if that expert is in a machine? Do we
now call it mere "information?"

Perhaps we need to distinguish between the "raw" knowledge
(which we assigned to each activitiy when we decided how many
resources it needed) and "derived" knowledge which results from
the PERT/CPM computations. Somehow I don't quite think that
these notions of "raw" and "derived" map to "explicit" and "tacit"
knowledge; they seem to me different in *kind*.

If "information" is "data with meaning," perhaps the easiest way
to think of "knowledge" is "transformed (i.e., processed through
some model) information." This broadens your original asser-
tion (above) that the transformation takes place "in a person's
mind."

On the other hand, perhaps I'm all wet and there's no such thing
as knowledge in a computer database system like this...but my
gut rebels at such a notion. I've see too many unexpected results
from computer models (which are, after all, just active databases)
that imparted knowledge to humans (and proved ultimately to
be correct) to lump them all into "information."

--
Carol Anne Ogdin
Carol_Anne_Ogdin@deepwoods.com