Real Time Organizational Modeling

[Carl Tollander]

>
> Carl: do you think policy modeling, and category theory in general,  
> could handle encoding an organization?
To date, the efforts I'm aware of speak to a Category Theory (CT)
environment
where agents equipped with identities navigate and create additional
model structure
among categorified data.  We might consider organizations to be
black-box generators of such
categorified data, but we don't yet model the organizations themselves
explicitly.  The
current work is thus about interoperability and mixing of models and the
theories about
them.   I think that this work is necessary to figure out how to more
easily explore
and calibrate spaces of models, but that general organizational modeling
might be a
bit ambitious for the approach at this point.

Here's why.  Organizational modeling is usually difficult  because (1)
the organization
is not always cognizant about what the salient features are (it looks
for things that are
more easily quantifiable), (2) there are usually different factions
within the organization
that will tell you (alas, over the duration of contract) that the
'actual' features are different
from those you were told about by those other guys last month, and (3)
what is actually
salient may be different depending on what the organization is trying to
do at any
given time and the environment it is currently situated in.   My
(admittedly limited)
experience with a machine learning project seems to reveal to me that the
feature-extraction area is still full of unsolved problems.  While these
problems aren't
totally intractable (my friends deal with them all the time), for me (a
small organization)
they are distracting,  so, accordingly, I'm trying to keep the focus
more on policy space
than organization space,  at least for now.

The questions one asks of models synthesized by CT-like approaches will
be more of the
'how does a group of models extend itself when we do X or modify the
group's model mix',
as opposed to the 'how does a static model (however nonlinear it may be)
react when we do X'.
Given the difficulties in modeling organizations I mentioned above, I
tend to believe that the
former approaches will eventually come to bear more fruit.    However,
'eventually' probably
doesn't mean 'tomorrow', and anyone new to complexity modeling
approaches should keep
in mind that these approaches will likely help extend the reach of
current modeling practice,
not compete with it.

Carl