Preface: Thanks to Ken for clarifying the role and nature
of the phd-design list---very useful---and to Dick for also
confirming that this thread is in the right place.
Warning: Sorry, this too is a bit long. About 170 lines.
Dear Terry,
Thank you for your thoughtful and thought provoking reply.
To continue with just two things, you say that
"... My take on this is to regard designing as the
core activity of addressing non-routine situations."
I like this. I think it fits with the concise
characterisation of designing I use to tell people (in as
few words as possible) what I'm interested in. It goes
like this:
We design things when there is a need or a desire for
some part or aspect of our world to be different, and
we cannot immediately specify how it should or could
be changed.
Which I could shorten to:
We design things when there is a need or a desire for
some part or aspect of our world to be different in a
non-routine situation.
Further on, you say:
"... The difference between our perspectives is between
'knowledge as object' and 'knowledge as an a human
activity/state'. Your model utilises 'knowledge as a
capacity or potential for rational action' (an object).
My approach takes knowledge as a dynamically evolving
activity existing in a uniquely individualised biological
process that includes reflexively related cognitive and
feeling sub-processes."
Actually, I don't think that there is this difference
between us, though I agree that it can, and perhaps does,
look like there is.
I am quite sure that for Newell and Simon, knowledge at the
Knowledge Level is to be thought of and treated as an
objective concept: knowledge is something agents have but
which can properly be treated as having an independent and
externalised existence. This is also the predominant view
in AI and Knowledge Engineering, and to a lesser extent in
some quarters of Knowledge Management: people or agents are
necessary as the carriers of knowledge, but their knowldge can
be treated separately from them.
Thus knowledge, for these people is (was, in the case of Newell)
like the concept of energy. You cannot have any energy, in any
of its forms, chemical, electrical, heat, or mechanical energy, etc.
without some physical source, but we can and do theorise about
energy, without having to include any theory of the physical
processes that can provide a source of energy.
However---and I realise that this is a bold thing to say---I think it
is a category mistake to understand knowledge in this way,
even knowledge conceived of as a capacity or potential for
rational action, as we have it from Newell and Simon.
I hope Klaus won't mind me borrowing his phrase "any
understanding is always someone's understanding"
(Krippendorff, 1997) to say that any knowing is always the
knowing of someone, or of some particular agent or agency.
Furthermore, any talk of that knowing must come from the
agent that is the knower: it must be subjective talk. It
cannot come from some other agent. That would be an agent
talking of the knowledge it has of the knowing of another
agent, or second-order first-person talk of some
first-order first-person knowing, to use Sid Newton's
terms. In other words, any knowing must be in the
'head/body' of an agent, and any knowing inside the
'head/body' of one agent is not directly accessible to any
other agent. One agent can come to know of the knowing of
another agent through observation and/or communication,
but this is a different knowing; a second-order knowing,
again to borrow form Klaus.
So, to summaries, the concept of knowledge we have is a
proper abstract notion, but the knowing it refers to is
always a subjective matter.
How, and to finally get to the cause of the category
mistake, we often form representations or models of
knowledge. Indeed we must do so in order to communicate it
or about it. Modeling knowledge is central to modern
Knowledge Engineering methods, and building computable
representations is what knowledge-based systems technology
is all about. Building a model or representation of some
knowing involves, on the part of the agent doing the
modeling, reflection and externalisation. Notice that an
agent can only reflect upon and externalise its own
knowing. An agent cannot reflect upon or externalise the
knowing of some other agent. Most often of course, we are
interested in modeling or externalising the knowledge of
some other agent or agency, but doing this necessarily
requires obtaining a knowledge of the knowing of the other
agent first. It is this second-order knowledge of the
knowing of the other agent that is then modelled. The
knowing of the other agent is not modelled or represented
directly.
Externalised models and representations of knowledge are of
course objects, things that any and all other agents can
access, at least in principle. The category mistake is to
think that because models and representations of knowledge
are objects and objective, then the knowing that they are
models or representations of is an too.
The process of obtaining knowledge of the knowing of
another agent, is what is called knowledge acquisition,
and we often see this described and talked about as some
kind of transfer process, of the knowing of the expert to
the knowledge engineer who does the modeling. This is
knowing as an object talk and an example of the category
mistake in action.
It is quite easy to see that the reality is not like this
because any knowledge modeling always changes the knowing
of the agent whose knowledge is (supposedly) being
modelled. In other words, if agent X obtains, by
communication with agent Y, say, knowledge of some knowing
of agent Y, and agent X then forms a model of its
(second-order) knowledge of what agent Y knows, which
agent X then presents to agent Y, to see if agent Y
approves of it as a model of its knowing, agent Y ends up
knowing something else, which, in turn means that what
agent Y now knows is not quite captured in the model. This
process is not one that typically converges, so going
round the loop several times does not necessarily result
in a situation in which agent Y gains no new knowledge as
a result of being presented with the new model.
So, again, to summamrise, models and representations of
knowledge are objective constructions, but the knowing they
model or represent is not. The knowing is and remains a
subjective aspect of the agents that do the knowing. This
is, as I am sure you fully appreciate, quite different
from the situation we, as scientists or technically
educated people, are used to, where our objective models
and representations are of objective things in the world.
I should stop here to see what you, and others, make of
this. I would simply add, however that this critique and
the idea that knowing is and can only be a subjective
aspect of agents is being swept away by the rush to
establish the knowledge economy, to manage knowledge as an
asset, to engineer knowledge systems, etc.
Agur,
Tim
CEIT, Donostia / San Sebastián
Reference
Klaus Kkrippendorff, 1997. Human-Centeredness: A paradigm
shift invoked by the emerging cyberspaces, Annenberg
School for Communication, University of Pennsylvania.
[http://www.asc.upenn.edu/usr/krippendorff/CENTEREDNESS.html]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
