[From Rick Marken (940208.1330)]
Bill Leach (07 Feb 1994 18:38:22) --
I might be "heading toward being a heritic" here but what the h*ll!?
You can be "right" and you can be "wrong" but you can't be a "heretic"
on CSGNet.
What is the result of running experiment "A" two or three times in a row?
(and I recognize that the second disturbance will be different for each
run of the experiment).
The disturbance IS always differnet. If you store all the disturbance
vectors the performance with the replayed vectors is always worse
than performance was with the "live" equivalents. The effect has nothing
to do with practice or familiarity; it occurs becuase the dynamics of
the opposing, low gain control system reduces instability due to
the human control system's transport lag.
What are the results of repeating the experiment "B" runs consecutively?
On every repeat (using the same replayed distrubance each time) control
is worse than it was with the original, live disturbance.
The equation I gave you was wrong, by the way. It should have been:
d(i) := d(i-1) + k*(r-p) not d(i) := d(i) + k*(r-p)
A trial starts with i = 1 and d(0) = 0. k depends on the sampling
rate of the computer -- I used .12 in Hypercard on a IIci. r is a smoothed
random series of numbers. p is the cursor position (in this compensatory
tracking task) at the instant that this program step is executed. This
equation is the entire model of the opposing control system, where k is the
gain. Make k too big and you've got a big conflict and no control.
Do you look at peak control errors as well as RMS?
No. Should I? Why?
I think that I am still fuzzy on where the second disturbance variable is
generated.
I hope the information above clears it up. Of course, I can also
send you the HyperCard stack, if you have a Mac.
it really does not strike me as strange for the results to differ.
Sure stuck me that way too. That's why I (and then Bill P.) checked the
hell out of it to make sure it was happening. It was.
It is also quite reasonable that the reason I am not confused at this
point is that I simply do not understand the situation
You understand the situation fine. That's why you are confused by the
results. You'll have to get used to this in the world of PCT. Again,
if you think the Marken Effect is impossible, I suggest that you read
the paper on p. 61 of "Mind Readings" for something that's REALLY
impossible (from an input-output point of view) -- but true.
Now here is where my heresy starts
Ah haaaa.
You say that there is no information in what the subject perceives about
the cursor movement.
There is no information about anything other than the cursor movement
itself. The perception of the cursor does inform the subject of the
position of the cursor.
I don't think that I "buy" that
Good. I didn't sell it. What I was selling was the idea that, in a
compensatory tracking task, there is no information in the perception
of cursor position about the DISTURBANCE that is influencing the
cursor. The cursor position over time, c(t) is the sum of disturbance,
d(t) and mouse, m(t), values: c(t) = d(t) + m(t). In order to
keep the cursor in a fixed position (c(t) = constant) the subject
must generate mouse movements so the m(t) = - d(t). The subject
does this without ANY information about d(t) in c(t) -- says I (and
Bill P. and Tom B. and Gary C. etc).
the subject does sense the mouse movement as well as cause it to move
And controls the perception of mouse movement too. But this cannot
possibly tell the subject anything about the disturbance influencing
c(t) -- unless d(t) were periodic and the subject were able to detect
this periodicity in his own mouse movements. But even this does not
help the subject control the cursor (I've tested it) . And it's not
necessary to make up these kinds of stories anyway. The control model
behaves just like the subject -- and it knows nothing of d(t) and m(t);
it just keeps the perception of c(t) at the reference level; it (like
the human subject) controls its perception of cursor position.
Bill Leach (07 Feb 1994 22:42:36) --
Of course I was neither trying to be disrespectful nor trying to ignore
the work that IS being done on these matters (and I am not taking your
response as an indication that you thought that I was).
I don't know what I might have said that made it seem that I felt
"disrespected" but I don't. I think it is assumed that every person
who posts on this net is respected. What we don't automatically
respect are ideas. I consider all ideas discussed on CSGNet (including,
if not especially, the ideas of PCT itself) open to the most vigorous
challenge. If an idea is worth anything, it will stand just fine against
any kind of assault. Basic PCT is as robust and real as Newtonian
physics; we welcome and ENCOURAGE the most aggressive challenges (especially
challenges involving real DATA) to the basic tenets of PCT.
I do believe that characteristics identified and quantifiable in a known
information stream by IT can have somewhat predictable effects on
perception by the "normal" person.
In PCT (as in all conventional psychology) perception is assumed
to be a function of environmental events. In a tracking task, this
means that p(t) = f(c(t)) where f() is now the perceptual function
(carried out by the afferent networks of the nervous system) that
turns the environmental variable, c(t), into a neural signal, p(t).
I think that information theory might have something useful to say
about the relationship between c(t) and p(t). What I have been objecting
to is the idea that p(t) contains information about disturbances that
influence the state of c(t).
The IT position has been that p(t) carries information about d(t);
they believe this because they believe that this is a logically
necessary assmption -- or how else would the control system be
able to generate outputs, m(t), that have effects that are opposite
to the effect of d(t) on c(t)? That is, if there were no information
in p(t) about d(t), how could m(t) = -d(t)? When you understand that
you are dealing with variables in a causal loop rather than a causal
sequence, however, you can see that this is just the wrong question.
Again, this may be viewing things in an improper fashion for PCT
discussions...
Again, you can be wrong about PCT, but never "improper".
Unlike control systems that man creates out of (initailly) inanimate
objects, human control systems can choose NOT to control.
True. And that is a control process in itself -- controlling the
level of controlling. In order to do this, the human must have a
perception of the level of something called "control" and be
able to do something to influence that perception (probably have
effects on the gain of systems contributing to the perception of
"control").
In fact, everything that a person does "on purpose" is control, is it
it not, including purposefully controlling or not controlling. "Doing"
means bringing some state of affairs (some perception) to some particular
level that is considered "right" (the reference level). In fact' it's hard
to think of ANYTHING of any significance that people do (other than
accelerating downward at 32ft/sec^2 when jumping from a bridge) that
does not involve control. Can you think of anything?
Thus, while human control systems are CLOSED LOOP, they are anything but
fixed.
Right-o.
Best
Rick