[From Bruce Abbott (970929.1415 EST)]
Bill Powers (970929.1109 MDT) --
Bruce Abbott (970927.0855)
Don't put words in my mouth. I neither said nor implied that the organism
was "something passive that is simply acted upon by the food pellets." What
I said was that the effects of a variable on other variables in this complex
system depend on the state of the system. As I recall, any physical
determinate system can be defined completely by its variables, their
functions, and an initial state (e.g., at time t0), from whence one computes
all subsequent states at time t1, t2, etc. Or are you now claiming that a
control system is an indeterminate system?
A living control system is, to some extent, actually an indeterminate
system, and it is also, in practical terms, indeterminate. To the extent
that reorganization involves random processes, it is actually
indeterminate. To the extent that its organization is a product of
evolution and its behavior a function of unobserved variables, it is for
all practical purposes indeterminate: it is certainly not determined by the
present state of the world we can observe.
So what? Neither quantum indeterminacy nor indeterminacy in the sense of
lack of full knowledge have anything to do with the determinacy at the macro
level of the system with which we are concerned. If there is any
indeterminacy in the following system equations I fail to see it:
p = f(i)
e = r-p
o = ke
i = o + d
We don't know what the value of d is going to be at any given moment, but
that fact does not make the effect of d on i indeterminant.
Determinism is a matter of faith, and so is irrelevant in any scientific
discourse. We can observe and infer the states of variables, and discover
relationships among them. But there are always variables of which we know
nothing, and for all we know they are manipulated by a malevolent God. To
say anything at all about them is a waste of breath.
Classic Bill Powers straw-man argument.
Skinner's philosophical position was basically an argument from
determinism. He made two major mistakes in this argument, and they were
both in the premises. First, he assumed that the _present_ environment,
during the lifetime of an organism, determined behavior.
False. Do you make this up as you go along?
And second, he
assumed that behavior without a prior causal event is impossible (despite
his assumption that initially, operant behavior is simply "emitted"). The
first mistake was simply an oversight, but the second came from ignorance.
If Skinner had known anything about electronic systems, he would have
known that some organizations of matter can produce systematic behavior
patterns spontaneously, simply because of their internal connections and
without any external stimulus driving them.
Also false. Skinner was not an S-R theorist, explicitly or implicitly. You
have a great imagination, Bill.
You want to talk about the effect of the food pellet on the error signal.
But the appearance of the food pellet is a function of the error signal.
The content of the food pellet affects the error signal and the error signal
affects the rate of delivery of the food pellet. That does not mean that
the error signal affects the content of the food pellet.
What sets my teeth on edge is mixing the two ways of describing things, so
you treat the food pellets as if they were independent variables.
It is the delivery of the food pellets that reduces error and thus slows
reorganization, under your original scheme. As a result, the action that
produced the food pellets tends to stay selected. The food pellets were
delivered because that action was currently selected and was connected
through the environment function to pellet delivery, and because there was
error between the CV and its reference. It _stays_ connected because the
action led to reduction in error. It would not have happened except that
consumption of the pellet increased the level of the CV. That is the
crucial link explaining why lever-pressing comes to be the mode of output
through which the rat controls its nutrient level. The observed changes all
result from either the system being disturbed (deprivation) or its reference
being changed, the only two independent variables to be found in the system
as it runs. But the reduction of error in the system that results from the
food pellets is crucial; without that effect you would not see
lever-pressing increase at the expense of other activities.
If you
can show some other variable downstream that is a function of food pellet
intake ALONE, then you can speak of this little segment of the system in
those terms; you're only describing one component of the system, which will
always remain the same. But when you start extending the range of the
"effect" so it includes not other other subsystems, but other variables
that are independent of the one you started with (like the reference
signal), this is no longer legitimate.
More making stuff up. I haven't even _suggested_ that food pellets affect
the reference signal.
If you know that y = f(a,b), you can't speak of "the effect" of a on y. You
can speak of the partial derivative of y with repect to a, which assumes
that b is held constant at some known level, but in general that way of
defining an effect of a on y will depend on the level you choose for b. If
your objective is to predict y, knowing a alone is simply insufficient.
Evidently we understand the word "effect" differently. Normally an effect
of one variable on another is determined by holding other variables constant.