[Martin Taylor 2015.11.20/16/36]
[From Rick Marken (2015.11.20.1150)]
I made a critical mistake in the first sentence of my
reply; I left out the word “not”. Here is the corrected
version pf the post:
The answer is NO.
Now let's see why. Firstly because you ask two separate questions.
The answer to the first is “An observer might so assume, but there’s
no way for the observer to know”. Only if the observer knows what
perception was being controlled can the observer even see the state
of the environmental referent of the controlled perception, and only
if control is perfect and the observer knew that control was perfect
would the observer be able to see its reference state. Bill covered
the first point in the paragraph to which you referred when he said:
“once behaviour has been defined in terms of an appropriate
variable”. As soon as he needed that precursor condition, he was
into theory, not observation. At the end of this message, I refer
you to a recent (yesterday, I believe) statement of yours which
argues the same point.
Bill did not mention my second and third points, perhaps continuing
to assume that the observer was omniscient, but I think they are
equally important. So my answer can be restated as: “If I had to bet
on it, I’d certainly bet that the leg extension would have been
controlled, but I wouldn’t have bet that the reference level was
‘fully extended’ unless the driver was a rather short lady with the
seat set for an average man”.
The answer to your second question is almost always "No" in the real
world where the Test for the Controlled variable is seldom
available. Once the driver has depressed the clutch, he won’t want
to do it again unless he was unsuccessful in getting the car into
gear. If you disturbed the “depressing the clutch” control system
by, say, lifting the pedal further than its resting position (if
that is mechanically possible) the driver won’t resist your
disturbance until he wants to change gear again, which could be a
very long time if he’s on the highway. You didn’t try using that
disturbance on the occasion that was observed, but it’s still a
pretty good assumption that the driver depressed the clutch because
he wanted to perceive the clutch as depressed, and the action to
achieve that was a controlled extension of the leg to a reference
value determined by the success of the clutch depression. A pretty
good assumption is all it is, and it’s an assumption that can not be
tested experimentally. It’s not an observable fact.
Let's just suppose that the reference value for leg extension was
“fully extended”, but the clutch pedal got in the way and prevented
the leg from being stretched out the way the driver wanted it (to
relieve some ache?). The clutch was now depressed, so the car was
out of gear. Did the driver do something about that disturbance to a
presumed controlled perception of the car being in gear, and in a
particular gear at that? You have to ask about the driver’s actions
with regard to the gear-shift lever. If the driver did move the
gear-shift lever, does that mean the original leg extension was done
in order to change gear, or was it because the leg extension put the
car out of gear and that created a disturbance to the perception of
what gear the car was in? We are presented with an indefinitely long
recursion that needs a whole lot of assumptions, most of them highly
reasonable, none of them observed directly.
In my _***theoretical*** _ way of looking at the situation,
the leg extension is a behaviour that occurred because the output of
a hypothesized clutch position control system was distributed to the
reference input of the leg angle control system (through, I would
presume, a few intermediate stages). That’s all there is to it, but
it is theoretical. The observer could have seen the behaviour of leg
extension, and the behaviour of clutch depression, but that’s all.
The observer could not have seen the states being controlled in the
environment, nor if those could have been seen, their reference
values. Again in my theoretical view, if the leg had been paralyzed,
some other behaviour would have served the pedal position control
function, such as pushing the pedal with a hand-held stick or asking
someone else to push it. The clutch depression behaviour would have
been the same, but the leg extension behaviour would not. But you
could never test that presumption observtionally or experimentally,
because on another occasion on which you experimentally paralyzed
the leg, the driver might not want the clutch pushed – and maybe he
didn’t on the occasion you observed, though he probably did.
Now I ask you to answer the points re-made in the message to which
you responded. What is the reference and what perception is being
controlled in the hammering scenario? Is the controlled perception
something about the relationship between the nail and the board, or
a sound that fit beautifully into the percussion music he was
improvising, or creating a pattern of nails to symbolize the plight
of Syrian refugees, or finding a way to reduce a feeling of great
anger? Which of these (or some other) is easy for the observer to
see and to observe the reference value for?
You refer to Powers, LCS I, p175, para 2. So do I, and I repeat a
rather important phrase that you omitted: “once behaviour has been
defined in terms of an appropriate variable”. That definition (as
you often observe, most recently to Bruce Abbott in [From Rick Marken
(2015.11.2240)]) is not available to the observer. I quote:
···
[From Rick Marken (2015.11.20.1130)
Martin
Taylor (2015.11.19.23.36)–
MT: Correct, and for good reason, as I
explained in [Martin Taylor 2015.11.15.12.54]:
-----quote----
(Rick's suggestion of getting the hammerer
to hammer nails into many different boards
won’t tell that getting the nail flush in
that particular board is why he is
hammering this particular nail. He could
be doing it because he is angry with his
wife and wants to hit something other than
her, and has no interest in how flush the
nail is. He just wants to hit it. And
maybe he is using the protruding nails in
the other boards as an art exhibit.)
---end quote---
RM: The fact that you are looking at the
reference states of controlled variables when
you look at behavior is not based on
knowing exactly what variables are being
controlled at what reference states. It is based
on recognition of the fact that what we see as
behaviors are consistent results produced in the
face of unpredictably (and often invisibly)
varying disturbances. This is why I referred you
to Bill’s example of “extension of the leg” as
the controlled variable in the behavior
“Depressing the clutch” in Table 1, p. 172 of
LCS I. The point was to see whether, despite
excluding them from your definition of behavior,
you at least agreed that the reference states of
controlled variables exist as observable
phenomena (a point Powers made rather explicitly
on p. 175, para 2 of LCS I), being that, as
Powers notes, p. 175, para 3, ** “In these
reference states we have the heart of the
problem to which control theory is addressed.”**
As I said in my previous post:
RM: So do you agree
that “extension of the leg” is a controlled
variable with a reference state “fully
extended”? And that the fact that this is a
controlled variable can be objectively
(experimentally) determined. A simple “yes” or
“no” will do. And if the answer is “yes” then
why not include them in the definition of
“behavior”?
RM: I would really like to hear what your
answer is to this.
RM: You have never included
the ** reference states of
controlled variables** as
being among the “results of the
control of many layers of control
systems” that are observed.
