[From Bill Powers (951104.0500 MST)]
Martin Taylor 951103 19:30 --
WTP:
the cat
would compensate by changing its position during the "marking" phase,
assuming that marking requires contact.
MMT:
That last "assuming" is where the cross-purposes start. I'm
assuming that marking does NOT require contact, but the box
configuration along with whatever else the cat perceives just
happens in this case to generate contact as a side-effect. The
notion of its being a side effect was the core of my posting.
...And I assumed that the cat had no perception of rubbing against
the stick, and was controlling for completely different things.
In that case, if I moved the stick so that the cat no longer brushed
against it, the cat would do nothing to restore contact and I would
reject contact as part of the controlled variable.
If the cat stops
marking the stick, the box will no longer open. Our disturbances won't
change that. Please explain.
That's the problem I was trying to ask about. The cat has to do
something else, such as miaow while looking at the picture on the
East wall, which happens to brush the stick, again as a side
effect.
If the cat just happens to brush the stick as a side-effect, then if I
move the stick so the cat brushes against it less, the cat will not
alter its action so as to restore the original amount of brushing.
I think what you may be trying to set up is a case where the action of
controlling one variable _necessarily_ has some side-effect directly and
proportionally linked to the variable the cat is actually perceiving and
controlling, but it is the side-effect that has the desired secondary
effect. In other words, the cat misunderstands what it is about the
first control process that is having an effect on the second one.
Perhaps a better example would involve the cat's controlling for a skin
sensation created by rubbing against the stick, as if it believes that
this skin sensation is causing the door to open. In that case, the
deflection of the stick would not be the controlled variable, but if we
assumed that it was, the Test would, by your reasoning, be passed.
Indeed, that phase of the test would be passed. But we also require that
the sensory link be identified. If the deflection of the stick is the
first guess at the controlled variable, then the cat must somehow sense
it. We recognize, in PCT, that in the final analysis it is a perception,
not the objective variable that we notice from outside the cat, that is
actually under control. So we would find that the cat detects our
version of the controlled variable by means of skin sensations, and this
would result in changing the definition of the controlled variable to
"the skin sensations resulting from brushing against the stick." We
would then see the deflection of the stick as being a side-effect of
controlling for the skin sensations -- an irrelevant side-effect with
respect to the first control process, if not to the second.
Can you think of any case where this would not resolve your problem?
It's never necessary for the cat to understand why controlling one
perception is a means of controlling another one.
The second aspect of the problem I think you're trying to bring up is
superstitious behavior, a word you used. This means controlling for more
than one variable as a means of producing a secondary effect, where only
one of the variables is actually linked to the secondary effect and
control of the others could be omitted without losing the secondary
effect. Would the Test then be failed because it would not reveal the
unnecessary controlled variables?
The answer is no, if we simply asked what it was about the cat's actions
that was actually having the required effect. However, if we noticed the
extra actions, we could enquire about what variable they might be used
to control. Looking at the effects of controlling the extra variables,
we would find (in the case of true superstitious behavior) that none of
the secondary effects of controlling those variables was under control.
Of course this could just mean that we have not found the secondary
(higher-level?) controlled variable. But parsimoniously, we would have
to conclude that those actions were not part of a process of controlling
anything else but the immediate effects of the actions, such as
kinesthetic or other sensations. An example would be observing that
whenever the cat brushes against the post, it also emits a little meow.
Not being able to find any secondary controlled variable that is
affected by the meow, we would conclude that controlling for this sound
is not part of controlling anything else, and that it is probably
superstitious behavior.
I think what may be bothering you is the fact that the cat can make the
door open without any knowledge of why controlling one set of perceptual
variables has the desired effect on the perception of the door opening.
You may be thinking that the Test should reveal the _actually effective_
behavior rather than just the perceptions the cat is controlling. But in
fact the cat does not have to know what is actually effective about its
actions in order to solve the problem. When you transfer this fact to a
human being, the implications are somewhat more unsettling.
When we decide that pressing down on a button is what causes a light to
light, we can convince ourselves that this is the actually effective
action when it is not. In other words, our experimental manipulations
can lead us to misunderstand how the world works. The button may in fact
be part of a capacitive-bridge switch which we unbalance merely by
becoming electrically connected to it, so the effort of pressing is
completely superfluous. Only if we think to vary the pressing will we
realize that only the merest contact is required, and that pressing down
any further is wasted effort. If we don't do that, we will weave the
idea of exerting pressure on the button into our understanding of how
the electrical system works and never realize that we made a mistake in
the initial observation. The effects of this mistake may be innocuous,
or they may propagate throughout the whole logical system we build up,
making it invalid but in a way that we will never detect unless we go
back to the elementary initial observation and do some more tests.
To forestall an obvious rejoinder, yes, this can happen when we apply
the Test. Even more important than the Test is to treat the outcome of
any experimental determination, including the reasoning we build upon
it, with skepticism. Most of what we think we know about the world
contains an element of superstition. For all we know, _every_ piece of
knowledge includes elements that are actually unnecessary, because all
we know about the real world is what our perceptions tell us and what
effects controlling some perceptions has on other perceptions.
The Test is merely a way of attempting to reduce the amount of our own
superstition in our interpretation of the behavior of another organism.
If we see another organism apparently responding to a stimulus, the Test
can (eventually) reveal that the particular stimulus we see is not
actually the important perceptual input to the organism; that it is
actually a disturbance of some other variable, and that the apparent
response to the stimulus is really an attempt by the organism to keep
that other variable from being affected by the disturbance.
This does not mean that we have arrived at the uniquely correct way of
defining the controlled variable, but it does mean that we have to
change our interpretation of the stimulus and the response. Further, we
may find that the stimulus is complex, and that only one aspect of it is
really causing a disturbance. The other aspects of the stimulus are
really "superstitious" aspects, because it is not actually necessary for
them to be present in order to get the same effect. This is like
discovering that the rat is responding to the click of a relay and not
to the light that is being turned on by the relay. Only when you think
of unscrewing the light-bulb do you uncover your own superstition.
Misinterpreting empirical data is a phenomenon familiar to
psychologists. I don't know why the idea that stimulus and response may
be misinterpretations is so hard to swallow.
I guess I kind of got off the track there (heh-heh).
···
---------------------------
How many psychologists actually do what Bridgeman recommended? When a
psychologist publishes a report that some stimulus or situation causesa particular change in behavior, where are the converging operations
that test the causal assertion?
Usually it is another psychologist who sees the situation in a
different way an performs the converging operation: "If what Smith
says is right, then X should happen in situation Y that he never
thought of."
Wouldn't it be more efficient if the same psychologist tried to to both
parts of the process, before publishing? Instead of science being
conducted like a trial, one psychologist acting like the lawyer for the
defense and the other like the prosecution, one single scientist could
play both parts and save a great deal of paper.
I think what tends to be lacking is the same as what is lacking
when a newspaper prints an erratum. It doesn't get to the same
audience, and doesn't correct the original impression. Add to that
the likelihood that a negative result (does not) get published, and
maybe your original comment is somewhat justified. I think it's a
sociological problem, though, not a psychological one--I mean the
sociology of the science, not the science of sociology:-)
Man, you have said a mouthful. Refutations are ineffective; it's as
though you can say anything authoritatively enough, and some people will
go on remembering it for generations no matter how often it is refuted.
After all, we know that feedback is too slow, don't we?
-----------------------------------------------------------------------
Best,
Bill P.