Cliff Joslyn writes:
> ... I formally defined a hierarchy as any partially
>ordered structure. These admit to descriptions in terms of levels, while
>not being so strict as to require that each element have only a single
>parent. The latter are also hierarchies, but I call them "strict
>hierarchies". Graphically, the two classes of hierarchy correspond to
>directed acyclic graphs and trees respectively. ...
>
>As you say, the essential aspect is the absence of cycles. Whenever a cycle
>is present, the nodes which participate in the cycle are amalgamated into a
>single new meta-node, and the (loose) hierarchical structure is recovered.
>If the cycle goes from a leaf to the root, then clearly under this
>procedure the whole structure collapses into a single meta-node.
In the real world that the model reflects it is difficult to imagine how
any hierarchy or heterachy can continue to exist through time without both
the interaction of the parts (which may be simultaneous or at least in
parallel) and retroaction (feedback) across levels. Structures do not so
much collapse into unity or metasystems as they become subject for the
moment to the information derived from particular systems/agencies, and
this indeed may vary from moment to moment.
For example, the pituitary "meta-node" functions variously depending upon
the feedback from other agencies in the endocrine orchestra. The leadership
of a country may shift back and forth, depending upon many subsystems and
unpredictable events; etc. etc.
The particular contribution of cybernetics may be an emphasis upon the
dynamics of change which can make otherwise static models relevant to
problems of interest. (Of course I would accept that, if the situation is
seen as a teaching one, rather than an exchange and evolution of views, one
might well be justified in beginning with the simplest possible forms, to
which the complications may be added later..;-).)
Catharina Kennedy -
>I think the only way to capture the concept of "heterarchy" is to consider
>the simultaneous interwovenness of different views of the operation of a
>system.
And perhaps, in addition to simulaneous views, to utilize cybernetic
principles to try to include observations of sequential, recurrent and
evolving behaviors?
Jeff Prideaux -
>If human intelligence involves the ability to jump out of a
>system, change the context, and resolve something that wasn't
>previously resolvable, then it seems to me that an artificial
>device would also have to be able to do this if it was to be
>considered intelligent. . . .
>
> ...After the creative act of jumping out of the
>system, changing the context, and resolving some
>issue...one can then (after the fact) go back
>and formalize the new situation..
O.K., but what might it mean to "jump out of a system"? It seems to me
that intelligence and creativity must relate to impacts of the real world,
to one's own internal models or constructions of one's relation to that
world, and to the utility of such models/constructions as practical guides
to action - action that in turn is successful according to some criteria in
terms of which it can be deemed intelligent.
To "jump out of a system" may mean not only to observe the system and its
relations from the point of view of some metasystem(s). It might also mean
a change to the values or parameters of an existing model. On occasion it
might possibly mean a change in the elements which constitute the model
itself, and the relationships among those elemtns, a modification of the
model which might lead to better management of the actual system.
The most interesting case might be to:
(1) recognize the provisional character of any theory or systems model,
(2) consider its real world (e.g. observational and experimental) referents
and analogues,
(3) consider those amenable to purposeful change (perhaps guided by real
world questions or purposes/values),
(4) make the desired change, and then
(5) modify the abstract/conceptual systems model accordingly, and
(6) observe the behavior according to the modified system/model, and
(7) observe the effects of new interventions in relation to the higher
level or real world purposes/values ((3) above).
Where questions are undecidable within the original model such a procedure
might make it possible to take into account various possible answers and
scenarios.
If this is what may be involved it is difficult to imagine how the changing
values required to meet the problems in the system environment, involving
as they may questions undecidable within the existing system, could
possibly be programmed into any artificial device.
Cheers and best wishes.