🚀 go-pugleaf

[From Bill Powers (990422.1755.MDT)]

>If engineers who design control systems don't really understand control
>theory (i.e., they use an S-R model and dismiss PCT), how do they manage to
>design control systems that work.

I think Bruce Gregory answered this question, as did you. We don't need
correct understandings to make things work. All we have to do is stick with
it, patching and fixing and fiddling around, until we achieve the result we
want. We're control systems, and that is what self-reorganizing control
systems do.

Some engineers, of course, have a perfectly sound understanding of control
theory. My understanding came from their writings, or the writings of their
mentors in the 40s and 50s, not just from my own experiences. These
engineers design control systems that work properly, and they understand
their workings as negative feedback control systems. They find nothing
alarming about the idea  that control systems control their own inputs. Not
being interested in psychology, most of them don't care, either.

It's been my experience, however, that many engineers are simply not
innovators or pioneers (which could be said about members of practically
any profession). They learn what they are taught, and if they become
teachers, they pass it on as the truth. If you tell them something
different from what they were taught, they become upset and say you are
wrong. This happens even if you tell them the same things they were taught
but in different words. If they happen to come under the influence of a
teacher trying to make a name by coming up with something new, it will
never occur to them that the teacher might not speak for everyone in the
field. So we come up with "fuzzy controllers" and "modern control theory"
and other similar trendy stuff, invented and promoted, often, by one
venturesome person or a small group and adopted and defended as the latest
hot poop by crowds of faithful followers.

In the meantime, most of the control systems that do the world's work and
quide our rockets into space work in the good old fashioned way, without
fuzziness or Kalman filters or neural nets. And a lot of control systems go
right on working even though they are horrible kludges, cost way more than
they needed to, and what with all the patches and fixes, are too complex
even to think about.

Don't forget that half of the control engineers have skills that are below
average.

Best,

Bill P.












  Presumably, whoever designed the
>thermostat mechanism that is cited so often on CSG didn't know or care
>about PCT.  Yet, from my miserable layperson's perspective, the darn things
>works just fine.  Ditto for the weapons systems on board ship.  The
>engineers associated with those systems didn't know squat about PCT either,
>yet the weapons systems they built worked.  That leads me to ask a couple
>of stupid questions:
>
>        If an understanding of PCT isn't needed to design workable control
>systems, what value does PCT offer?
>
>        If all PCT does is offer an alternative explanation for phenomena
(e.g.,
>the thermostat and shipboard weapons systems) that are adequately accounted
>for by other explanations (namely, those in the minds of the designers and
>builders of such systems), where is the "competitive edge" of PCT as the
>would-be theory of choice?
>
>
>>Rick Marken (990420.1810)]
>>
>>>Bruce Gregory (sometime 990420) --
>>
>>> That's certainly true. But Rick is not talking about student
>>> misconceptions. He is saying that engineers who design control
>>> systems don't really understand control theory and that only
>>> "we" truly understand control theory.
>>
>>Real engineers who design real control systems have used control
>>theory to study humans as S-R devices; these same engineers have
>>dismissed PCT because they know that control systems control
>>objective variables in the environment, not their own perceptions.
>>These engineers are not students. Do they understand control
>>theory?
>Regards,
>
>Fred Nickols
>Distance Consulting "Assistance at A Distance"
>http://home.att.net/~nickols/distance.htm
>nickols@worldnet.att.net
>(609) 490-0095
>
>