[From Bill Powers (2000.08.18.1442 MDT)]
Bruce Nevin (2000.08.18.1045 EDT)--
"The normal operation of the hierarchy"? What is "the normal operation of
the hierarchy" in the living organism? We infer that it is the same as the
normal operation of the hierarchical model. This inference is justified to
the extent that the model behaves as the organism does. Has the hypothesis
about emotion ever been modelled?
By "normal operation" I mean only that controlling perceptions by means of
action backed up by the appropriate somatic states is how I conceive of the
hierarchy as normally working. I don't think we need to posit any special
state of being to account for emotions. Not yet, anyway. It's still all
perception.
An invokation of "the
normal operation of the hierarchy" is weak and maybe even a bit dishonest,
with yourself if with no one else.
Gosh, Bruce, that sounds pretty bad. Unless, of course, you misconstrued
what I meant. I think I have said exactly in what ways the "normal
operation of the hierarchy" is involved. Of course, as you point out, I am
referring to my understanding of how the human system works, which may or
may not resemble how it does work.
There's some suggestion that a strongish rate effect is involved [in
emotion],
so that sensations that go with increasing error feel disproportionately
bad, and sensations associated with error going away feel much less bad or
even positively good.
Has change of rate been observed, or is this an inference from the
subjective changes?
You may not be familiar with the term "rate of change." An example of rate
of change perception can be seem in muscle spindles that produce signals
proportional to muscle length AND to rate of change of muscle length.
Suppose the muscle length (measured from resting length) rises from 0 to
100 millimeters in one minute (very slow). The proportional signal will
rise from 0 to 100% of the final value in one minute. The rate of change
signal will rise immediately to some small value such as 1 impulse per
second (out of, say, a possible 100), remain constant at 1 per second
during the whole rise time, and then fall back to zero when the stimulus
ceases to increase and becomes constant. If the stimulus rises from 0 to
100 millimeters (the same total change) in one second instead of one
minute, the proportional signal will also increase from 0 to 100% in one
second. The rate of change signal, however, will now rise immediately to a
value of, say, 60 impulses per second, and remain constant at 60 per second
all during the rise time. When the stimulus becomes constant, ceasing to
increase, the rate of change signal will again drop to zero, indicating
zero rate of change. Thus the magnitude of the rate of change signal is a
measure not of the magnitude of the stimulus, but of the rate at which the
stimulus is increasing.
There are many examples of rate-of-change signals in all modalities: the
phenomenon is often referred to as "adaptation" or "habituation" because
after a change in the stimulus, the perceptual signal which rose with the
stimulus returns, quickly or slowly, to zero. In all such cases what we
experience is mainly a _change_ in the stimulus, the more rapid the change
the more marked the perception.
With these examples in mind, it seems reasonable to guess that somatic
sensations might also have rate components which indicate changes more than
actual magnitudes. This phenomenon holds for descreases as well as
increases. A perception that decreases, when added to a rate perception,
can actually drop to zero even though the magnitude of the stimulus has
dropped only, say, 10 percent. Of course as soon as the stimulus has ceased
to decrease, the (negative) rate component will become zero again, and the
total stimulus will rise to its proportional value, here 90% of the value
before the decrease assuming linear perception..
Assume there is in fact a rate change. In the model, what makes the rate p
change?
It's not the rate that changes but the magnitude of the stimulus (and the
proportional perception). If you're feeling sick and then the unpleasant
sensations are reduced by half, you might well conclude that you're well,
if the rate component of the perception is weighted more heavily than the
proportional component. This doesn't mean that you feel the rate component
separately. You simply experience sn exaggerated diminution of the "sick"
signal because the sensors are affected by rate of change. They report an
increasing variable as being larger than it is, and a decreasing variable
as being smaller than it is.
1. In our control diagrams, a change of qi results from a change of the cv
and/or of influences on it (d and qo). The change in qi becomes a change
in p.
2. Reaching beyond what is usually modelled, p can change because of a
change in input functions (sensitivity of receptors, change in weighting of
signals in higher-level input functions).
3. With emotional arousal, the pupils of the eyes may dilate, and I suppose
that would affect rate of visual signals; but on the other hand with
negative emotions the pupils may contract.
4. Are there relevant neurochemical changes that pervade the environment of
nerve cells in the nervous system or in a region of it? This soupy aspect
of neurobiology does not appear in a control diagram.
The control diagram does not indicate many design details, such as the
linearity of the perceptual input function, delays in transmission of the
signals, or dynamical effects in the output function such as time
integration. The input function might combine proportional and
rate-of-change sensitivity; that detail would not appear in the diagram,
but it might well be included in a simulation based on the diagram.
If you want to indicate effects on parameters such as sensitivity,
linearity, delay, or dynamics, then you must draw a diagram showing the
source of the effect, and the source of effects on the source, and so on
until all variables are completely accounted for one way or another. The
variables that are givens are the independent variables of the system and
define its input boundary. All other variables are dependent variables. So
if you want to propose for example that some physiological state associated
with an emotion affects a system parameter such as sensitivity, you must
either propose that the physiological state arises because of independent
variables outside the system, or show how the physiological variable is
proposed to depend on other variables within the system. The reason for the
latter requirement is that if there are any closed loops in the system it
is impossible to predict the effects of any variables without properly
taking feedback effects into account.
How would we model the rate change associated with emotions? Prior to that,
if rate change is inferred from subjective impressions of the greater
"amplitude" of somatic sensations (for lack of a better word), don't we
have to have some data on rate changes before we can model them? Are there
such data?
The term is "rate of change", not "rate change." It means the rate at which
some variable is changing, like the rate at which the pressure in a tire
drops after a puncture, which can be fast or slow.
Are there data relating specifically to the kinds of bodily sensations I
assume are associatecd with emotions? Possibly, but I don't know of any.
The place to look for such data would probably be, as mentioned, under the
heading of habituation or sensory adaptation. Does one become habituated to
an elevated level of adrenaline in the bloodstream (in other words, does a
constant amount of elevation sooner or latter feel like no elevation)? If
so, then we might sense the feelings of anger or fear more strongly while
they are increasing, since the level of circulating adrenaline (or rather,
the effects it has that we can sense) increases in such states. And
conversely, a decrease in the level might be exaggerated in perception,
creating an impression that the feeling is gone when it will actually soon
return and has not actually dropped to zero.
Hope I've manage to address at least some of your actual questions.
Best,
Bill P.