Re: Ashby's Law of Requisite Variety

Bruce Edmonds (b.edmonds@MMU.AC.UK)
Mon, 20 Jan 1997 12:20:05 +0000


Andreas Garcia:
> Ashby's Law of Requisite Variety, which I'm becoming more and more
> convinced is not a "law" at all.

I agree with Andreas. I have long been disatisfied with it, at least in
its present form.

Let us take a simple example: a thermostat coupled with heat scource and
air conditioning. The control system has (essentially) two (or maybe
three) states.

One could attribute as many states to its envorinment as one wants -
this is open ended. Yet, given enough power the control system can
indeed keep the temperature within given bounds.

One may be tempted to say "Ah, but only a few enviromental states are
*relevant* to the control system", but this then begs the question as
one has pre-selected the number of states one would attribute to the
environment to be limited to the abilities of the control situation. In
this case the "law" has been reduced to a tautology and does not tell us
anything about the systems.

Alternatively on might argue that it is the number of goal states
compared with the action control states that is relevant. This is also
not the case: a person who can ony blink can control a typewriter (with
the right technology) - blinking is binary, a typewriter has many more
states. If one insists that the binary blinking must be considered over
time to count the states, then all varying things have an infinite
number of "states" so the law becomes meaningless again.

Finally one might argue that it is the number of internal states
compared with goal states that is relevant. Either this effectively
reduces the "law" to the statement that to specify a goal one
necessarily needs as many states as it, this is a sort of law of
description/information - to specify one of n states one needs
information of at least n. If not there must be some mechanism such
that even given one has less than n internal states one can effectively
specifiy actions to reach a goal describable with information n.

Anyway you look at it, it is undesirable. Versions of it may be
acceptable (it may be true that for efficiency one needs about as many
actions as pertubations?????), but I have yet to see one.

---------------------------------------------------
Bruce Edmonds,
Centre for Policy Modelling,
Manchester Metropolitan University, Aytoun Bldg.,
Aytoun St., Manchester, M1 3GH. UK.
Tel: +44 161 247 6479 Fax: +44 161 247 6802
http://bruce.edmonds.name