[From Bruce Abbott (2009.05.07.1240
EDT)]
Bill Powers (2009.05.06.0906 MDT) –
Bruce Abbott (2009.05.05.1830 EDT)
BA: We’ve had a discussion
about this before, although it’s been a while. I do believe that we have
emotion systems, which are a part of our inheritance from our evolutionary
past. As I see them, they are innately organized control systems
“designed” to deal with particular situations often faced by our
ancestors over the course of evolution. Certain inputs act as disturbances to
one or another of these systems; the behavioral and physiological
manifestations of a given emotion are the actions that often have been
successful in the evolutionary past in opposing those disturbances.
BP: That sounds convincing until you start asking for details. I won’t argue
against the idea that there are some inherited control systems; at the lowest
levels there are a lot of them. But the concept of a separate set of
“emotional” control systems just doesn’t hold water. How can a
primitive inherited control system know whether there is danger or not? How
does it know anything about the outside world or its relationship to the
organism? Considering all the things we get emotional about, you have to give
these inherited systems all the perceptual levels that are in the PCT
hierarchy, and all the abilities to send reference signals to other control
systems that the hierarchical systems have. When you get done you have two
hierarchies, one learned through reorganization and variable in its properties,
and the other inherited in a fixed form, and interacting in suspicious detail
with the first hierarchy.
I
hadn’t thought of these systems as lying outside the normal control
hierarchy. The question then becomes how they got there, inside the
normal control hierarchy. From my perspective, the cores of these systems
are inherited rather than emerging from random reorganization. Many
animals have a host of innate ways of dealing with challenges in their
environments. Their perceptual systems include mechanisms that recognize when
specific kinds of input are present and generate appropriate actions. For
example, in hungry newly hatched robins, any appropriately sized object seen
looming overhead induces a stereotyped gaping behavior (beak wide open and head
turned up). Normally the object looming over the chick its parent and the
consequence of this action is that the parent drops food into the chick’s
mouth. Here is a fully functional control system that gives the infant bird
some degree of control over its nutrient intake. (In this case it’s a
temporary one that fades away as the chick matures and develops more
independent ways of getting its food.)
As
I see it, emotion systems evolved rather early and have been elaborated, but
not replaced, over the course of evolution. As higher-level cognitive processes
developed, the range of inputs that could serve as disturbances to these
systems enlarged as well as the kinds of actions available as outputs. An
infant experiences fear in a limited range of circumstances; later, experience enlarges
that range. The infant’s actions when error occurs in the fear system
also become more varied and flexible. But the basic system is still there and
still activating the same physiological changes needed to support the vigorous
physical actions that characterized the output of the primitive system. So, for
example, a student must give a talk and interprets this situation as
threatening. (What if I perform badly and look foolish in front of my peers?)
This interpretation of this situation as threatening acts to disturb the fear
system, which among other preparations changes the reference levels of various
physiological systems, producing a racing heartbeat, increased breathing rate,
and tremors in the limbs and vocal apparatus, among other effects. The person
has difficulty focusing on the speech. The result is that these changes
make it far more difficult for the student to deliver the speech smoothly and
competently.
So
here is an example of the actions of an emotion system interfering with other
systems of the control hierarchy. Such effects are well known – people
often experience such effects as “taking over” their behavior and,
consequently, we learn techniques to suppress these actions when they interfere
with our ability to maintain control over other variables.
BP: What my proposal about emotion does is to eliminate all the problems of
duplication of and conflict between functions that are raised by the idea of an
inherited emotion system. We don’t have to explain how the emotion system can
recognize a lion, see that it is in a cage, and not be afraid, or see that it
is loose and cause the legs to run like mad, presumably wresting control of the
legs away from the hierarchical systems that had been using them. If you
propose that the emotion system reaches into the hierarchy and alters the
reference signals for running, you have to explain why the learned systems
don’t simply cancel out that disturbance. You’re trying to fit an old idea
about emotions into the PCT system, and it just doesn’t work.
In
my proposal, as I’ve already noted, these systems are part of the normal
control hierarchy. In PCT, each level in the hierarchy contains a myriad of
control systems, all of which, if they produce observable behavior, must ultimately
produce changes in the skeletal musculature. If two systems controlling
different variables are organized to use the same (or an overlapping) set of
muscles to maintain that control, then conflict results. To function
effectively, the systems must reorganize so as to use different sets of muscles
or be employed sequentially rather than simultaneously. This potential for
conflict between systems is an inherent property of the hierarchy. Introducing
emotion systems alongside the other systems occupying multiple levels in the
hierarchy does not create any additional problems not present a purely learned
hierarchy, so far as I can see at present.
BA: For example, either innately or because of prior experience we may
interpret certain conditions as threatening. This perception of threat is
opposed by what might be called a “threat control system” whose
reference is to keep perceived threats close to zero. When threat perception
rises above this reference, this error produces behavioral outputs (e.g.,
running away from the source of the threat) and physiological changes to
support vigorous behavior such as running or self defense. (The latter outputs
are generated by changing the references of appropriate lower-level systems.)
BP: Doesn’t that sound very much like something the learned hierarchy of
control systems does? Why do you need a built-in system to recognize threats
when you already have a learned one that can do the same thing – and how do
you explain the fact that most threats are not initially reacted to as threats?
Children happily trot out into traffic in ways that would scare them to death a
few years later. Do those inherited emotional control systems have to learn
about traffic, about the bogeyman, about strangers offering candy, about sharp
knives in the hands of maniacs, and so on? And if that is true, aren’t we just
talking about the one hierarchy of perception and control that is already
posited in PCT? Exactly what does the idea of an emotion system add?
It’s
very much like something the learned hierarchy of control systems does. The
question of why you need a built-in system to recognize threats and organize
actions appropriate to dealing with them can be addressed in at least three
ways. First, at least a primitive version of the system allows the organism
to have an organized and potentially adaptive way of dealing with certain kinds
of threat from the outset, before there has been much opportunity to learn.
(Infants in the crawling stage, for example, have an innate ability to
recognize sharp drop-offs and will refuse to crawl off the edge of one, as
shown many years ago by the “visual cliff” experiments (http://encarta.msn.com/media_461547576/visual_cliff_experiment.html
). Second, it may provide a kind of scaffolding on which a more capable system
can be elaborated. Third, as I see it, the evidence strongly indicates that
such a system is present and operates in much the same way across individuals
and even to a large degree across species. (So, even if it weren’t “needed”
on theoretical grounds, I believe that it is needed on empirical ones.)
BP: I think the idea of a separate emotional system arose because awareness is
not always connected to all control systems at all levels. Control systems
operating outside awareness, automatically, can still react to disturbances by
adjusting reference signals in lower system, so when awareness is engaged
elsewhere, it would seem that the reactions were caused from outside somewhere.
If a change in biochemical state accompanied the automatic motor reactions to
disturbances, the first the systems currently in the conscious mode would know
about it would be a set of bodily sensations and awareness of some automatic
action taking place. Of course attention would quickly go to the problem and
all the “reactions” would then appear as normal control behavior.
Doesn’t that cover all the evidence anyone has ever come up with for the
existence of an independent emotion system?
Yes,
but this suggestion does not argue against the sort of system I’m
suggesting, which is part of the normal control hierarchy but whose
organization depends more on innate factors and less on random reorganization
than some other parts of the hierarchy presumably do.
One thing I believe is important
to keep in mind about the process of evolution is that it builds upon the
products of previous evolution. A future engineer with enough knowledge about
the human brain might be able to design an artificial brain that could do all
the same things the human one does. Almost certainly, the designed brain would
have a far more rational structure than the natural one. The natural brain is
more like an early clock whose inventor started with a basic design which was
then found to perform poorly. So the inventor starts adding components to compensate
for various factors. After many additions, the clock now keeps accurate time
despite changes in temperature, movements, and so on. Some of the
additions are there to compensate for deficiencies created by some of the
earlier additions. Based on the lessons learned in building this clock, the
inventor could now build a simpler watch that would achieve the same accuracy. Our
brains, as products of a long evolutionary history, are more like that early
clock than the cleaner design of the inventor. From this perspective, it is
more likely that we will find that real brains have built upon and modified
existing prior innately-programmed structures rather than developing purely
under the direction of a random reorganization process (beyond the first
levels).
BA: A detailed model of all this
would be more complex than this condensed description suggests. For
example, what is perceived as threatening may involve a complex calculation
that includes one’s own capabilities, availability of weapons, etc.
Because learning is involved, the actual behaviors engaged in may be quite
diverse. (I could generate other descriptions suitable for other emotion
systems, such as those involved in the experiences of anger, grief, curiosity,
love, and so on, but this example gives the general flavor of the approach.)
BP: But now you’re not talking about an emotion system: that’s just the normal
hierarchy of control systems you’re describing. I’ve taken care of the feeling
component by proposing that motor actions are accompanied by adjustments of the
endocrine and other control systems, so what is left to account for?
Surely you’re not proposing that an inherited system separate from the PCT
hierarchy can perceive capabilities, availability of weapons, and other
things that an inherited system could not possibly know about, are you? With
what eyes would such a system see weapons? With what brain would it assess
capabilities? How could it seize control of reference signals for lower
systems without coming into conflict with the learned systems?
As
should be clear by now, that’s not what I’m proposing.
BA: So, what do you find objectionable about this conception of emotion
systems? It’s still turtles, er, control systems all the way down.
BP: All I object to is the idea that you need something other than the one
hierarchy of control that we talk about in PCT. In my proposal, what we experience
as and label as emotion arises naturally in the course of normal behavior, as a
consequence of error signals and the actions that follow from them. I have
added only the idea that when actions are called for, the state of the body is
also adjusted as appropriate to the action, and that we feel the result of that
as a change in feeling state.
It
looks to me that we may be much closer together on this than may have been
apparent at the beginning of this discussion.
Bruce
A.