[From Rick Marken (930407.1000)]
Richard Thurman (930407.1220) --
I for one would really like to see how PCT can
be used in research.
I won't forget this. I just had to do a quick response to Allan
Randall (see below) and then I'll try to get into the research
discussion.
Rick..... I was curious about the 'control of sequence' study you
outlined last month (Marken 930316.20000) so I put together a Hyper-
Card stack that would do roughly the same thing that your QuickBasic
program was doing.
You're too good to be true!
I ran several subjects and found that in every instance the subjects
were faster at restoring the original sequence when the letters were
from the 'different' set (STUV). Presumably this is because subjects
are now controlling a perception at the configuration level rather than
at the sequence level.
Yes, I think so.
I was also curious about Gary's questions concerning doing the same
study in an auditory modality. So I programmed the stack to play
sequences of sounds instead of showing letters. In this case I could
control sequences no faster than 300 milliseconds but could control
configurations as fast as 250 milliseconds.
So the reaction time for controlling sound sequences was about
300 msec and that for vision was 500 msec?
Rick do these reaction times jive with what you found?
I didn't measure the results in terms of reaction time; I measured
control (in terms of number of correct elements in a series of N) as
a function of rate of presentation. I think my control was nearly perfect
when the letter sequence was about 3/sec. So that imples a reaction time
of about 300 msec. So my visual rt's are about the same as your tone rt's.
But you did collect your tone and visual sequence data under the same
conditions so maybe there is a modality difference. I haven't done
the tone study -- maybe I will.
I'm also curious to find why perceptual modality may have to do with this.
Let's first make sure that there really is a difference. But, if there
is, that's fine.
If 'a sequence is a sequence' then why do different (auditory vs visual)
modalities seem to play a part?
I think we have to think really hard about how we can be sure we know
what the subject is actually controlling in each of these experiments.
Then, when we convince ourselves that we know what the subject is
controlling, then maybe we can try to model this control process (which
will mainly involve, I imagine, modelling f(qi), the perceptual function
that transforms some aspect of the outside world -- as modelled by physics
-- into the controlled variable, p. As we test the model and get it honed
to the level where we know that the model and subject are behaving virtually
identically, then maybe we can start trying to make some sense of any
possible modality difference for sequences, if there is one.
Allan Randall (930407.1600 EDT)
Why must we produce a "very high" correlation with
the disturbance?
Because we had reached the point where Martin proposed measuring
the information in perception using his ?Mystery? function -- which
is just the output function of the control system. We all agreed that
if there were useful information about d in p (that is, information that
would let the system compute the appropriate o values based on p)
then the o values generated by the Mystery function would correlate
highly (and negatively) with the disturbance. In the last two days
Bill P. and I have posted data showing that the Mystery function can
lead to o values that have almost NO relationship (correlation)
with the disturbance -- yet there is still control. This is proof that
there is no USEFUL information in p; ie. not information that a
control system can use to reliably to generate the outputs that produce
control.
I have tried to emphasize more than once that no one
is claiming there is 100% of the disturbance information in the percept.
Agreed. But there must be enough information in p to generate o?s that
compensate for d. If there is not enough information in p that allows
you to do this, then the information in p (regardless of how much is
there -- and I?m inclined to say zero) is USELESS; this information
cannot be responsible for the o values that lead to control.
However, the control system's only
source of information is via this percept, so perfect control is not
possible. I still don't think you've responded to this point. The
percept is the only information source the control system has. I would
really really like to know how you view this.
Here's the really, really answer. A controlled perception,p, is NOT
AN INFORMATION SOURCE. IT JUST IS! The perception is
the only input to the control system but it has no information about
nuttin'. You keep assuming that something (the perception, for example)
MUST bring information into the system about the disturbance (for
which the system?s outputs compensate). But that is your point of
view as an observer of the system; you can see that there is a
disturbance influencing p and that the outputs of the system are
counteracting that disturbance. You are assuming that the system
must be able to get information about d in order to ?know how? to
?do the right thing?, o. But this is a stimulus - response view of
control. You are looking at control from the ?outside - in? and
it looks like this:
d --> p -->o
It looks like p is a channel between d and o. Since the system is
generating o values that are nearly equal to -d (at least, in the special
case where p = o + d) you are assuming that the system MUST BE
getting information about d via p in order to compute the appropriate
o values. At least, this is what I have understood your position to be.
My point is that the diagram above does NOT represent the way things
work in control. There is a closed loop. The way this looks from
the ?inside - out? is more informative about how control actually
works.
r --> r-p<--p
>
CL
All the system knows is the deviation of p from r. The perception, p,
is ?informative? only about itself. If p = 5 then the system knows that
p = 5; that?s it. Any difference between p and r (r-p) is ?sent out? by
the system into the closed loop (CL). The well-understood ?magic? of this
CL continuously turns r-p into effects that keep p = r. The system itself
has no idea ?why? this works. It has no idea ?why? it needs to ?do?
anything. It has no idea that it is generating outputs that are compensating
for a disturbance, compensating for a varying feedback connection between
itself and p and/ or compensating for changes in its own properties that
continuously change the characteristics the output function that converts
r-p into effects on p. The only quantity in this loop that could possibly
qualify as information (from my point of view) is r-p, which is information
to the loop itself about how close it is keeping p to r.
Where do you think the
information is coming from to allow the system to control against
the disturbance?
This question shows that you are looking at control from an ?outside
in? S-R perspective (top diagram). I don?t think information is coming
for ANYWHERE to allow the system to control against disturbance.
There is NO information coming in AT ALL. All that is coming in
is perception, p.
Do you think the system needs *any* information
at all about the world in order to control?
Unequivocally, NO.
If not, I just have to shake my head. Magic.
It is the head-shake of conventional psychology. You are definitely
not alone in your head shaking. PCT IS magic from the conventional
perspective. Why do you think people get so put-off by PCT; and so
mad at it that they would rather ignore PCT than deal with it straight on.
That?s why I am VERY impressed by what you (and Martin) are doing.
You are looking straight into the jaws of the monster (PCT) and trying
as best as you can to conquer it with SR tools (I know that you guys
think you are trying to save it from foolish people like me, but it works
out the same; at least you are TESTING -- or getting us to test -- your
claims). Most people are content to shuffle off and mutter ?What
crazies; PCT couldn?t really mean THAT?. I?m afraid that it does.
Along with Rich Thurman, I welcome and celebrate your (and
Martin?s) tenacity and bravery.
Best
Rick