Re: Hierarchies, recursion...

Cliff Joslyn (joslyn@KONG.GSFC.NASA.GOV)
Tue, 5 Sep 1995 15:05:21 -0500


>In view of some complications you and I have encountered to date, I want to
>define what the word "system" means to me: "a complex unity formed of often
>diverse parts subject to a common plan or serving a common purpose; etc.

That's OK. I'm MORE than willing to work with your terminology. The only
thing is that then YOU have to offer definitions like this.

OK, from here on out hierarchy_B (for Boris) will refer to a particular
form of social organization characterized by a strictly hierarchical
(tree-like) system of command and control, in the style of military
organization. I admit that it is a common form of social organization,
although far from universal, especially in societies whose largest unit of
organization is relatively small (e.g. tribal groups, hunter-gatherers).

>It is called "hierarchy," which means "government by priests,"

I thought the original etymology rather referred to the ranking of the
angels, which is reflected in the earthly ecclesiastical ordering.

>>From our experience, we know that people don't appreciate this systemic
>architecture, anymore. It is:
>-- patriarchic and/or autocratic and/or undemocratic

I believe that virtually ALL societies are patriarchic, whether
hierarchical_B or not.

>-- authoritative and potentially abusive

Clearly non-hierarchical_B societies can also be abusive, however.

>-- cumbersome in handling complexity

This is not at all clear to me. In fact, I believe that hierarchies_B can
provide a very elegant way to handle at least a certain kind of complexity.
This is why large, complex organizations use them.

>-- closed and/or highly resistant to change

This seems intuitively reasonable, although I'm really not sure if either
it's universally the case in hiearachical_B societies, or whether it is
either a cause or an effect of hierarchy_B.

>There is nothing new here; this is common understanding in management sciences.

Well, then I'd question at least some aspects of "common understanding".

>What does cybernetics tell us about hierarchical architecture?
>-- that parts or components are arranged in servient to dominant relationship
>-- that information flows from the servient to the dominant, that is, from
>the bottom towards the top

Well, TECHNICALLY information does flow both ways, in that even if only
commands come from the top to the bottom, information transmission is still
required to transmit the command. But I know what you mean: the lower
levels act as "sense organs" for the higher levels.

>-- it has no feedback channel or loop for information exchange

E.g. bosses don't explain themselves or circumstances to their subordinates?

>-- the line of command or management flows from dominant to servient, from
>the top down, and there is no feedback channel, either
>-- the balance or equilibrium is established by imposition rather than
>causal circular feedback
>-- the integrity of the servient is compromised to the commands of the dominant

I generally agree with most of that.

>-- the informational process is reductionist, which is to say that the
>information is continually being attenuated as it moves up the ladder,
>creating problems with requisite variety

"Reductionist" isn't clear to me here. Generally, higher levels will use
more abstract representations than lower levels.

>-- as a probabilistic model, the decision making dominant levels don't have
>the best possible understanding of what is happening at the servient levels

Not necessarily: the quality of an understanding is only RELATIVE TO A
PURPOSE. The same situation perceived by a mid- or lower-level might be
better for a purpose at that level, but a decision made at a higher levels
might be BETTER with a more abstract representation.

>-- both the decision making and control process are highly centralized.

Not necessarily. The top levels make more general decisions (we're building
prodct Y this year), the mid and lower levels more specific (we're hiring
Joe, I've got to call Karen now about that order). But decisions and
control exist distributed throughout the hierarchy. What's centralized is
the PRIORITY: the CEO can fire Joe, and tell me to hang up the phone, but
not vice versa.

>-- the informational process tend to be linear, sequential and closed
>-- the interrelationship of components is framed on dependencies

?

>In other words, hierarchical systems tend to stifle interdependence,
>interrelation, interactivity and interdependence between the component
>parts. Not too helpful in terms of viability.

Yes, walls are placed: if I'm in marketting, and I want to talk to
production, I have to go through our common boss.

>What we are talking about, here, is the relative merits and demerits of
>hierarchial and recursive systemic architecture, particularly as related to
>social systems.

While I'm clear about hierarchy_B, at this point I have only a vague idea
of SPECIFICALLY what you (and/or Beer) mean by "recursive system" (say now
recursive_B). For example, Sales and Advertising are part of Marketting in
a hierarchical_B company, and sales can't talk to Production. But, Sales
and Advertising ARE wholly contained in Marketting, like two Russian dolls.
Furthermore, just by opening a horizontal channels from Sales to Production
doesn't elminate hierarchy_B, since the CEO can still overide everybody.

It would help a lot if you would give either a very specific formal
description of a recusrive_B system, or a small, simple example,
contrasting it with a hierarchical_B system.

O---------------------------------------------------------------------------->
| Cliff Joslyn, NRC Research Associate, Cybernetician at Large
| Mail Code 522.3, NASA Goddard Space Flight Center, Greenbelt, MD 20771 USA
| joslyn@kong.gsfc.nasa.gov http://groucho.gsfc.nasa.gov/joslyn 301-286-2598
V All the world is biscuit-shaped. . .