non-adaptive control; Illusion of control; reinforcement problem

[Martin Taylor 950606 20:10]

Bill Powers (950606.1440 MDT)

I was simply asking whether there might not be a
way to account for the fact that human beings do not lose control
entirely when some inputs are lost, by supposing that perceptions can be
composed of non-obvious as well as obvious types of input. One
explanation for why they don't lose control altogether is they they're
using an internal world-model. I was suggesting an alternate explanation
for at least some cases.

I may have been "preaching to the choir," but you appeared not to see
that I was offering an alternate explanation.

Another case of omitting to say the obvious, I suppose. Your alternate
explanation is/was what I would conceive of as the natural one. The "model"
approach is what I would (previously) have called the alternate one. My
discussion originally was meant explicitly to ignore the "natural" structure
and see what could be done within one control system whose input might be
cut off completely.

    If the ECU is made less "E"lementary, and provided with a second
    degree of freedom by way of a "perceptual validity" signal, it
    could shut itself off.

That's too much intelligence to put into an elementary control system,
for my taste. I would much rather try to think of a structurally simple

higher-order system that would do the equivalent.

Fine. So would I. As I said, I'm not proposing that this is what happens.
I'm only (perhaps in "lawyer mode") pointing out the possibility that control
of whether the control system controls (:wink: when the validity of the
perceptual signal is low need not be done from outside the control system.
I even included a query as to whether the validity COULD be assessed from
within the single control system, because if it can't, then the notion is
moot in any case. But whichever way it happens, the validity and the
perception itself are two different degrees of freedom, requiring (as you
say) separate perceptual functions--or at least separate outputs from some
perceptual function that "begins to bulge with ad-hoc complexity."

I try not to lose sight of the fact that systems that seem nicer, simpler,
more probable, are not NECESSARILY more correct than other systems that
might accomplish the same functions. One needs other evidence, and one
stays perhaps a bit more skeptical if one keeps the other possibilities
in mind, even while thinking that they are unlikely. Your criticisms
often seem to be of the form "the standard structure is simpler so it must
be the only one we should consider." I disagree with that approach. But
I do agree that the first consideration should be whether and how well the
standard system will do whatever it is we are considering at any time.

Well, your idea of an "exact" analogy is not mine.

I shouldn't have used the term "exact," because "exact analogy" is a classic
oxymoron. "Exact" was intended to apply to those aspects of the analogy
relevant to the discussion. If you introduce other elements, the
exactness vanishes. At least my analogy is more exact than choosing
whether or not to blow the horn when you are told that your car is going
to be pushed off a cliff! And it is real and personal and painful.


Incidentally, I think it must be almost a uniquely US notion among the
democracies, that you could so easily come up with "Nobody is forcing you
to vote for anybody." Nobody is physically forcing anyone to vote, true.
But how on earth do you expect to keep your democracy if you don't? I've
never missed a vote since I was old enough to be allowed to, except when
I was out of the country for both the election and the advance polls. And
I don't intend to miss one either, whether municipal, provincial, or federal.
What proportion of the eligible US population voted your current bunch of
radicals into Congress?

Leaping into the "having control" discussion (and quickly out again):

I seem to remember some studies way back, in which two monkeys were in
similar cages, both getting shocked at the same times and the same amounts.
But one monkey could turn off each shock as it came, and that control
determined the cessation of the shock to both monkeys. The monkey without
the control got stomach ulcers, but the one with the control didn't.

Or something like that. Anyone remember what I'm talking about, so that
they can report it correctly, because I'm sure my statement is only a
rough approximation? It must have been 20 or 30 years ago, when people
were less concerned about the ethics of such experiments.


<[Bill Leach 950607.02:13 U.S. Eastern Time Zone]

[From Bill Powers (950606.1440 MDT)]

I 'kinda' felt a need to quote this, as it is yet another fine example of
a way to state a fundamental PCT concept:

... That's how the test for the controlled variable works: you can't
prove beyond doubt that a particular variable is under control, but you
can very quickly show that it's not under control.

Also, this "model based control" thing is likely to remain "muddy" for
a long time to come -- at least until a significant amount of work in the
higher levels of the postulated hiearchy has actually been done.

While I don't know that such is a practical use, as far as living control
systems are concerned I suggest that any time there is a reference for a
perception that can not actually be a current perception and the organism
is actually attempting control of said perception and there does not
exist a massive error condition then some form of model based control
exists (regardless of whether the model is correct or not, if not the
error will occur later - at such time as current perception control is