[From Rick Marken (960628.0900)]
Jeff Vancouver (960627.16:40 EST)
you do not know the psychological literature as well as you think. But,
like I have said, I do not think your reorganizing system is anywhere near
producing an error here, so never mind
I would reorganize my thinking about that tiny segment of the psychological
literature with which I am familiar if you could convince me that organisms
do not exist in a closed loop relationship with respect to the proximal
causes of their actions. Or, if you accept the fact that organisms do exist
in a closed loop relationship with respect to the proximal causes of their
actions, you can get me to revise my evaluation of the psychological research
literature by showing me why this fact does not invalidate all data collected
within the framework of conventional IV-DV methodology.
You say that you know what I think of conventional research and that you
find repetitions of my rationale for thinking this way boring. That's
fine. But if you want to show me that my attitude toward conventional
research is wrong, you can do this only by confronting my boring beliefs and
showing me why the closed loop relationship between an organism and its own
input makes no difference to the way we study behavior.
You know what I think about this; the essense of what I think about it is
described in the "Blind men..." paper. This boring little paper argues that
observed relationships between independent variables (d) and "dependent"
variables (r) tell us nothing about what a negative feedback control system
is doing (controlling). The paper shows that IV-DV relationships are
irrelevant side effects of the process of control. The psychological
literature with which I am familiar is filled only with reports of (usually
statistical) relationships between IVs and DVs. Based on my analysis of
closed loop systems and my observation that all living systems are closed
loop with respect to their sensory input, I conclude that the research in the
psychological literature describes only irrelevant side effects of control;
it tells us nothing about purposeful behavior.
Your response to this argument has been "that's just your opinion" or "look
at all we've learned from this research" or "you don't know what's in the
literature" or "I'm bored" or "I know what you think but I don't think that".
This doesn't help me understand what is wrong with my analysis of the
situation. I accept the possibility that I might be wrong; it might be that
the results of IV-DV research tells us something useful about purposeful
behavior. Perhaps the conventional research methods texts (including mine)
do not need to be abandoned before psychology can make the shift to PCT.
But, in order to convince me of this, you must show me what is wrong with my
critique of conventional research methods. You can do this by showing me
where my analysis of closed loop systems is wrong or how it does not apply to
real living organisms. The coup de grace would be to pick some piece of IV-DV
conventional research -- such as one mentioned in your Psych Bulletin
paper -- and show what it tells us about the purposeful behavior of the
subjects in the study.
I do not wish to engage you, Rick Marken, in this conversation.
That's up to you, of course. But I would like to find out what's wrong with
my critique of conventional IV-DV research. This could save me a lot of
embarassment at the CSG meeting where I am planning to present a paper on PCT
research that includes a discussion of what's wrong with the conventional
IV-DV approach to the study of the behavior of living systems.
Bill Powers (960627.1610 MDT) --
What we maintained, and still do, is that the control system can counteract
the effects of environmental disturbances on its CEV without any knowledge
of what is causing those effects.
I think this gets at the essence of the problem with information
theory. The problem is that control systems can deal only with a world of
perception; they have no special access to the environment (disturbance and
output variables) on which these perceptions are based. All a control system
knows -- and all it needs to know -- is what it perceives (and that truth is
beauty and beauty truth, of course;-)).
The whole notion of "information in perception" assumes that perception
exists to "inform" or "tell" the system about the world beyond the senses.
The idea is that there is "real stuff" out there that we look at through a
glass darkly -- we see the shadows on the wall of the cave. This is naive
realism; it is the idea that what we see is a representation of what is
_actually_ out there. The modern version of naive realism is that what we see
is a representatin of what physics and chemistry say is out there (sounds
frequencies, photon packets, etc). This leads to the notion that perceptions
"inform" us (with varying degrees of accuracy) about what's _really_ out
there in the world.
When it comes to control, naive realism leads to the idea that a control
system must know what is really going on "out there" in order to control it.
What's "really going on out there" according to PCT, is disturbances -- a
term that refers to variables in the real world that contribute to the
perceptions that the system controls. The control system must act to protect
the controlled perception from any tendency of these disturbances to move the
perception from its reference state. The naive realist takes this to mean
that the control system must have information about these real events
(disturbances) in order to deal with them.
Naive realism is a mistake because control systems (including the naive
realist) know only their perceptions; they have, and need, no information
about what's out there in the real world. We can only guess what's out
there, beyond our senses, and it is only very recently (in evolutionary
terms) that our guesses (the models of physics and chemistry) have been
precisely confirmed by systematic observation (of our perceptions, of
course); but these guesses about (models of) the real world beyond our
perceptions are still tentative; new observations may require substantial
changes to the models (look at relativity).
What Bill, Tom and I have been trying to show is that it is not necessary
for control systems to know (get information about) the "real world" basis of
their perceptions, which is a good thing because they couldn't do it anyway.
A control system doesn't need to know that a 50 knot wind out of the east is
the reason why the steering wheel must be turned 30 degrees counter-clockwise
in order to keep the car on the road. All the control system has is a
perception of the relationship between the car and the raod; a perception
that tends to vary for many reasons, all unknown to the system controlling
the perception. All the system controlling this perception needs to do is
generate output proportional to the difference between the actual and
intended state of the perception. The system has no information about why it
needs to do this (because the wind is blowing); it just does it. And while it
does this it fools the naive realist who is watching the show into thinking
that the relationship between action (steering wheel position) and
disturbance (wind velocity) exists because the system is getting information
about the wind and/or its effects from perception.