free will

[Jeff_Vancouver1]

[From Jeff Vancouver 940725.1200]

My fear was realized. There was more response to my "out on a limb" post
than my free will post. I will respond to the "out on a limb" and other
posts about MMPI and personnel managers, but not for a week or two.

I am trying to reconcil PCT with the work of others. Were it is different
and where it is the same. I have found a pattern as I
interact with the net and the psychology literature. First, I think there
is much overlap. THen I post a question or talk of some work by
psychologists. PCTers react that the difference is great. I think it
clarifies
differences, thinking the PCTers are correct. I continue to look at the
psychology literature and PCTer responses and realize it is mostly words,
not concepts that are at the core of the differences.

For example, I am not sure what the technical meaning of feedforward is
anymore. It seems that what I was describing as FF, PCT describes as
planning. Given that FF has a specific technical meaning, planning is
probably the better word. Many in psychology use the term FF when they
mean planning. This is probably because, like me, they are not aware of
the technical meaning. We could have spent, and did spent, much time and
thought arguing about FF and planning when if we understood the words
we would have realized we were talking about the same thing.

I am reposting the post that I need PCTers to address. Bill L.
[940721.2314] was
the only one to respond, and as he noted, he does not represent core PCTers.
Thank you for responding Bill L., but I can't use it unless other PCTers
confirm that it adequately represents the PCT model.

Tom B. [940721.1803] responded to a post that preceded the post I am
repeating below. In it he asks what Locke said. To quote, PCT is
"neobehaviorism." He is very unclear what he means by that, as I discuss
below. This is from his article in Motivation and Emotion (I have a
preprint, so I don't know exact cite).

[from Jeff Vancouver 940721]

First I want to apologize for my inability to read all the posts
on the net. I just skimmed it for references to my posts. I
found two [Marken 940719.1400 and Powers 940720.0745]. To which
I respond below. Meanwhile I am continuing to develop a strategy
to respond to my reviewers. Sometimes this takes my questions to
different levels, as you will see below.

[Marken 940719.1400]

Study of control by Psychologists:

Campion, M. A., & Lord, R. G. (1982). A control systems
conceptualization of the goal-setting and changing process.
Organizational Behavior and Human Processes, 30, 265-287.

Kernan, M. C., & Lord, R. G. (1990). Effects of valence,
expectancies, and goal-performance discrepancies in single and
multiple goal environments. Journal of Applied Psychology. 75,
194-203.

Hollenbeck, J.R. (1989). Control theory and the perception of
work environments: The effects of focus of attention on affective
and behavioral reactions to work. Organizational Behavior and
Human Decision Processes, 43, 406-430.

···

___________________
Can you give me the reference for your "long post on the nature
of control?" Most of your posts seem to fit this criteria

By the way, I was reading your Degrees of Freedom paper. You
cite neural networks as promising for modeling perceptions. Have
you changed your mind or did I misunderstand something you said?
_________________________________
On reinforcement [both Marken & Powers]:

I want to take this to a higher level, but I need to respond to
Marken & Powers. You are correct that I cannot give a counter
model. I do not how to model! I sympathize with your inability
to get your detractors to work with you on such a model. That is
the most one can ask. Your efforts are impressive. I no long
think that you "straw maned" reinforcement theory. Instead, I
think your description of E.coli _as you model it_ can be
interpreted as reinforcement theory. I should reread Powers
response to the behaviorists in earlier writings. But let me
quote Bill P's post (940720.0745):

"In this model, E. coli senses the time rate of change of
concentration of the attractant..., and varies the delay before
the next tumble according to whether the _current_ sensed rate of
change is above of below a reference setting."

Forgive me, but cannot that be restated "E. coli senses the time
rate of change of concentration [stimuli] of the attractant
[reinforcer]..., and varies the delay [a response] before the
next tumble according to whether the _current_ sensed rate of
change is above of below a reference setting."

That is, the stimuli "causes" a behavior, where the behavior is
length of delay!

If this argument, or some form of it, has been rebutted in
published work, just cite it - no need to repeat here. Unlike
Paul, I have amassed most of it over the years. Unfortunately,
it is beginning to fade from my mind.
________________
Here is the lasted bottom line:

The problem of "reinforcement theory" in your models is important
(to me) because of Locke's claim PCT is neobehaviorism. If I
understand it, the usual response is that the model includes
internal [the real crux of control] as well as external variables
(e.g., behavior). Since behaviorism did not acknowledge any of
these internal variables, PCT is not behavioristic. This always
worked for me.

But as I try to reconstruct my own counter arguments (the editor
wants me to take the debate further), I am beginning to see a
point that I cannot get around. If the values of reference
signals (higher-order output signals) are arrived at randomly,
which is the central process for reorganization, then your model
does seem - on one level - difficult to distinguish from
neobehaviorism (by which I think they mean S-O-R). Stimulus (S)
leads to response (R) with the organism's (O) perception as a
mediating variable and discrepancy (of perception from reference
signal) and the ultimate reinforcer. (Ironically, Bandura was
one of those who popularized S-O-R).

Adding the O takes a lot of wind out of the PCT "is not
neobehaviorism" argument I had used in my mind. Maybe I have
been too generous in thinking neobehaviorism includes O. If it
does, than I can accept that, argue that PCT is neobehaviorism,
so what? It still serves functions goal theory and social
cognitive theory do not. But I suspect S-O-R is not acceptable
to PCT either. Plus the argument takes us to interesting new
dimensions.

It is Marken's insistence that behavior is out-of-control that
seems to support the neobehaviorism argument. As I understand
it, controlling a perception variable occurs in two ways. Once
the appropriate lower-order reference signals have been
discovered, the loop simple sends those signals. Although this
might result in different behaviors (due to different
circumstance each lower-order loop must deal with to meet their
sent reference signals), the flavor of the response to a stimuli
(which was translated to a perception that sends the focal loop's
perception off its reference signal), it the same. That is, the
response is reoccurring.

I anticipate red flags going off here. PCTers will say, "but the
behavior is different because the circumstances are different.
Only that special case where the circumstances are exactly the
same, or close enough for your statistic analysis, will a
reinforcement model work. And what is the _flavor_ crap?"

This _flavor_ crap is the other way Locke might mean
neobehaviorism. Regardless of Locke or Bandura, however, I can
describe a concept called "flavor of a response" which means set
lower-order reference signals the same way. The result is the R
in S-O-R is now flavor of response. It is completely internal
because lower-order reference signals are internal. (Perhaps
Locke and Bandura would not accept this meaning for R and I can
stop here. But this argument has been made by Bill P. and I want
to take it further. When one takes it further, it gets very
interesting.)

If one can accept the new definition of R, then I can move to the
next way of controlling -- that is, before a flavor of a response
is developed. This is the reorganization process. During this
process varying reference signals are sent to various lower order
loops in a random, but localized fashion. As these random
changes begin to reduce the discrepancy, the set of reference
signals for the lower-order loops begins to be defined. That is,
the probability that a lower-order reference signal will have a
certain value increases as the discrepancy is reduced
(reinforced). Recognize behaviorism?

Now, will a simple adjustment counter the argument? If the set
of reference signals is developed discontinuously (either it
eliminates the discrepancy or it does not. If not, try
completely different configuration), than the phrase cannot be
probabilistic (it becomes: a set of lower-order reference signals
will be adapted if the discrepancy is made zero). But this is
not much of a change and becomes problematic for meeting complex
perceptions. Also, the concept of "localized" reorganization
would need to be examined.

No, the real problem is the method of reference signal selection.
Let us not consider hardwired, which some consider, but has not
been modeled. Hardwired will not work for controlling most
complex perceptions anyway, so lets only concern ourselves with
the only other process considered by PCT advocates - random.
This is the only central difference between PCT and the Locke and
Bandura self-regulation models where the Locke & Bandura
arguments make some sense to me. (Another difference is that
they talk of controlling behavior, which is clearly wrong on
their part, so we need not discuss it further with regards to
Locke & Bandura).

Here is where we get really philosophical. Locke & Bandura are
arguing that reference signal selection is our source of free
will. We consciously choose our goals (reference signals). I
think they would concede that this does not happen all the time,
but at least some of the time. They claim conscious choice
translates to free will, because it is not completely predictable
or determinable. I don't buy the free will argument for a second
(although lack of predictability and determinability is much
easier to accept). The free will argument requires that humans
are fundamentally different from the rest of nature (Sappington,
1990?, Psych Bull). A conclusion I cannot accept. (Ironically,
Sappington uses Bill P. and Bandura to argue for free will, but
it seems Sappington was assuming that reference signal selection
was ultimately non-random.)

What I do accept is that conscious processes are often involved
in selecting the (set of) reference signals. I am a soft
determinist, which means consciousness enters the process of
behavior selection (which strictly speaking is better said
reference signal selection), but that conscious choices are
determined by some other factors. This is what I like about
Bandura's model, the other factors are abstractly described
(e.g., self efficacy). Of course, as Bill P. pointed out (1991,
Amer Psych), the specifications are somewhat problematic if one
were to try to model them. I agree. What I don't know if these
_types_ of factors are considered in PCT? Specifically, Bill P.
says "beliefs about one's actual effectiveness in achieving a
given goal [Bandura's self-efficacy]" is a perception (1991). By
that do you mean self-efficacy is just another controlled
variable? If is it a controlled variable than 1) what is "F" in
your model and 2) what is F2? If F is an internal variable
related to the focal goal and if F2 affects references signals
related to the loop (goal/task) under consideration then we have
just resolved a major conflict among the models.

One more problem and I will have resolved two of my biggest
conflicts with PCT. If a function type for F2 includes using an
external address for a reference signal, then outside influences
are available for constructing a hierarchical control system. Is
it possible?

The messy models in psychology begin to describe many of these
influences. Together with the structure of PCT and the rigor in
your methodology, I think psychology can make great strides.

Postscript: I think one can persuade others that PCT has
something to offer. The exact function is debatable, but I have
moved closer and closer to PCT as I have come to understand more
and more of it. Bill P. claimed most people take 2 years. Ed
Ford and Dag report making converts slowly. These data, anecdotal
though they are, indicate the process is not completely
discontinuous, although certain leaps may occur on the way. I
suspect that Locke and Bandura will only be convinced by PCT when
they perceive rewards are contingent upon accepting it, which
will probably not happen in what remains of their careers. But I
will first seek to convince myself, then the reviewers and
editor. If I am not convinced, I will take what I am convinced
about and make clear the distinctions between PCT and my view. I
would appreciate Bill P.'s sanction on the manuscript to assure
our respective views are properly represented.

A parallel question is can I convince PCTers that psychology has
something to offer PCT? The argument is parallel. Some are
already convinced. Others will never be. But, unfortunately, in
my system, that loop has little gain. PCTers won't grant me
tenure.
_____________________________

So ended the high gain post.

A couple of things that I want to add from weekend readings.

1) Bill P., in responding to Bandura, you suggested (AP, 1991) that importance
and commitment (Bandura's imprecise language) might be translated as gain.
That makes sense for importance, but commitment seems to translate better
as the lag of the higher-order loop that set the focal reference signal.
Possible?

2) I was read parts of D. Ford's living systems framework. He is clearly
talking about the same model, using the same methodology, etc. He cites
Powers but not say Marken. He also attributes his model to others. I am
still not sure about Bill P.s understanding of the epistomology of
Ford's theory. Bill P. have you had any contact with Ford?

3) I read Kent McClelland's piece coming out in Sociological Perspectives.
Very nice. Helps with an argument I want to make. One thing, which is
the crux of the discussion above. How does an external agent influence a
control system into adopting a particular reference signal?

Maybe I will have better luck this time.

Jeff

[Tom_BOURBON3]

From Tom Bourbon [940725.1633]

[From Jeff Vancouver 940725.1200]

My fear was realized. There was more response to my "out on a limb" post
than my free will post. I will respond to the "out on a limb" and other
posts about MMPI and personnel managers, but not for a week or two.

Good. It is obvious that you stirred up the swarm with the reply addressed
to my "out on a limb." It's only fair that you jump back into the fray.

I am trying to reconcil PCT with the work of others. Were it is different
and where it is the same. I have found a pattern as I
interact with the net and the psychology literature. First, I think there
is much overlap. THen I post a question or talk of some work by
psychologists. PCTers react that the difference is great. I think it
clarifies
differences, thinking the PCTers are correct. I continue to look at the
psychology literature and PCTer responses and realize it is mostly words,
not concepts that are at the core of the differences.

Fine, Jeff. But please remember that when you tell us about your
interpretations of the psychological literature, you do it with words, not
working models. When we ask you, in relpy, if you have found any specific
quantitative models in the literature, you reply (honestly and directly),
no. You have found nothing (nor have we) which demonstrates that the words
in traditional psychology can be turned into something that behaves the way
the words say.

That being the case, I believe you have a difficult job ahead of you. To
reconcile PCT with other theories, you must either turn PCT into a merely
plausible word-theory, or turn the other theories into quantitative
generative models that behave on their own. Absent one or the other of
those converisons, you would be trying to do something like showing the
musical equivalence of a cheese sandwich and a concert pianist.

All _you_ need to do to show us (and yuor editors?) that there is great
equivalence between PCT and traditional psychology is show us _their_
generative models. Show us where _they_ demonstrate that their ideas
can produce behavior the way they assert that they can. I think it is at
least a little unfair for you to present their word-based arguments, claim
those arguments are equivalent to PCT, and then ignore the fact that they,
not we, bear the burden of proving that their ideas work.

For example, how can I conclude, as you seem to want the PCT modlers to
conclude, that when psychologists talk about feedforaward, they are talking
about the same things we model? Every instance I have seen in the
psychological literature leaves me with the conclusion that people are
simply using a buzz word, feedforward, in a way that unjustifiably implies
a technical distinction between feedforward and feedback. Further, the
literature reveals a general lack in psychology of any general understanding
of "feedback" in a control system. To complicate things for me, as one
who is expected to say there is an equivalence of ideas, I see the
additional attempts by psychologisis to finish what they believe must be a
full "set" of feed-somethings: feedback, feedforward, feedthrough, positive
feedback, negative feedback, and so on. All I see them using are words;
there are no demonstrations that their word-models work.

There is more to it that words over concepts; there is the matter of
demonstrating that one's ideas can work at all, even in simple simulations.

As you say:

For example, I am not sure what the technical meaning of feedforward is
anymore. It seems that what I was describing as FF, PCT describes as
planning. Given that FF has a specific technical meaning, planning is
probably the better word. Many in psychology use the term FF when they
mean planning. This is probably because, like me, they are not aware of
the technical meaning. We could have spent, and did spent, much time and
thought arguing about FF and planning when if we understood the words
we would have realized we were talking about the same thing.

But _they_ claim to be talking about control theory. _They_ are the ones
who say we do not know the theory. _They_ are the ones who tell us control
theory cannot work the way(s) we demonstrate it working. Are _we_ the ones
who should say the ideas are equivalent? A quantitative generative model
is like word salad?

I can see it now: It is unrealistic for PCT people to expect others to use
the same methods, and so on. Poppycock! Please notice, Jeff, that I did
not say _you_ would make that complaint; I can easily imagine that your
editors and reviewers would make it. And I have seen it on this net nearly
every time we react to the idea that _we_ should slug it out in the arena of
word games, but that no one else should be asked to demonstrate that their
ideas can be turned into even a very simple working model. That isn't my
idea of what science is about.

I am reposting the post that I need PCTers to address. Bill L.
[940721.2314] was
the only one to respond, and as he noted, he does not represent core PCTers.
Thank you for responding Bill L., but I can't use it unless other PCTers
confirm that it adequately represents the PCT model.

If you mean your interpretation of the E.coli model as a stimulus-response
model, I can't say it adequately represents the PCT model. Bill Powers has
already replied to you on this and I concur with his remarks. (Have you
gotten hold of the issue of _Closed Loop_ that contains "Models and their
worlds," by Bill Powers and me? IF not, let me have your address and I'll
send you a reprint. The article contains some ideas that are relevant to
your attempt to turn PCT into S-R psychology.)

If you are asking whether I agree with Locke's notion that PCT is really
just S-O-R psychology, I don't. From what I have read of Locke, I do not
agree with either his portrayal of PCT, or his construal of S-O-R
psychology. S-O-R is S-R is Cause-Effect is lineal causality is not a
workable explanation for the phenomenon of control. Notice that I did deny
the seeming _plausibility_ of all the many incarnations of lineal causality;
some of them sound remarkably plausible. The problem is, they won't work
when you model them. The criterion we apply to ourselves is workability,
not a nice, plausible sound to the words.

"Neobehaviorism" is simply the new name given to behaviorism after the early
radical behaviorism from the era of John Watson was combined with the
logical positivism that came from Auguste Comte and Ernst Mach. It is
still "the" science of the environmental control of behavior. I'm sorry,
but I can't make the intellectual leap that sees PCT as neobehaviorism. Can
you tell us why you think there is merit to Locke's assertion, or are you
simply expressing your dispair (which we share!) that the editor expects
you to show an equivalence?

···

===============

My mail seems to be hung up somewhere today. Messages I posted in the
morning are still not on the net. I will wait until things seem to be
working again, then reply to some of your other questions.

Later,

Tom

[Tom_BOURBON3]

Tom Bourbon [940805.1702]

[Jeff Vancouver 940804]

. . .

I guess I will address the testing stuff later, except to thank Bill P.
and L. for correcting Tom.

I'm eagerly watching for your next comments about the goodness and
usefulness of psychological testing when it is used as a crutch to justify
major decisions about the lives of people who are rquired to take the tests.

Jeff, you must be referring to the post in which Bill P. corrected what he
mistakenly thought were statements by me about the additive or
mulitplicative properties of variances. As I indicated in
Tom Bourbon [940804.1519], I was talking about the *proportion* *of*
*variance* *"explained"*, not about variances per se. The distance from
proportions to probabilities is not as great as that from proabilities to
variances.

Besides, the onus is still on you to defend the use, by psychologists, of
pre-employment tests that correlate no higher than .3 or .6 with performance
on the job. Above all else, I am curious about why you think testing
as poor as that has anything significant to add to the theory of behavior
in PCT. In behavioral science, correlations that poor should be viewed as
sources of *noise*, not as sources of *facts* to be explained. For
correlaltions that low, there are no "facts" to be explained.

Later,

Tom

PS Copies of the papers you requested are in the mail to you.

[Bill_Leach1]

<[Bill Leach 940806.01:08 EST(EDT)]

[Jeff Vancouver 940804]

Looks like you may have been a bit premature in your thanks. Also, the
disagreement centered more on just how bad not bad or good.

I don't think that PCT explains what you are asking at all and HPCT
probably does not give you a definitive answer.

There is a lot of evidence that the human mind is able to "ignore"
reality at times to stunning degrees.

I think that you are asking how the systems concepts are formed but
really wanting detail on how a particular systems concept can come into
existance.

There is no difference between a person committing suicide and a zelot
dying for a cause as far as "intrinsics" related to continuing life. As
has been stated by others, there probably really is no "intrinsic" that
measures for probability of dying (with which you obviously agree).

In addition, it really does appear that an unsatisfied reference does not
stay at maximum output for long. The question is not how does one ignore
all of the intrinsic signals that are going off during a hunger strike as
much as it is, how does a person set a reference priority (in the context
of your questions).

I don't think that there is any doubt that people can and do set
references for things that can be ultimately destructive to them
personally and control well enough that they indeed do die.

If the time frame were long enough, reorganization would probably disrupt
the control system that is causing the high intrinsic error level (and
indeed might be one of the reasons for some cases of "psychosis").

What I am looking for (P) is mechanisms that 1) allow beliefs to
influence the choice of a reference signal, and 2) mechanisms of
ignoring or toning down error signals (particularly from intrinsic and
higher-order loops). For the first mechanism, I need more than simply
to view beliefs as a kind of perception (Marken). That is, where in the
hierarchy is a belief type loop, and how does it affect output functions.
For it seems that the decision mechanism is at the output function.

The systems concepts is the area where "beliefs" would affect reference
signal choices. However, the difficulty there is that "beliefs" are
perceptions. What you think that you believe is not what drives your
decision making.

I am no doubt on "shakey ground here" but it is not your belief in
gravity nor your "knowledge" of gravity and its effects that keep you
from stepping off the roof of a tall building but rather your experience
with the effects of gravity (incombination with the fact that most of us
are generally controlling for long relatively pain free lives).

-bill

[Jeff_Vancouver1]

[Jeff Vancouver 940804]

Mary, Bill P, & Bill L.

I think that covers it. My e-mail is down, so I am not fully functional
here (as if I ever was).

What I am trying to understand with my free will discussion is these fuzzy
concepts that Bill P. used like "decide" and "want." If ultimately the
organism is attempting to maintain its intrinsic signals, then how can an
organism decide to die? That is, decide to commit suicide?

This is just the pen-ultimate example. For me, this is not a problem in
PCT because, in fact, the intrinsic signals refer to much more specific
things than "life". Nonetheless, the idea that the organism can override
all the intrinsic signals that are going off during a hunger strike is
difficult to explain.

What I am looking for (P) is mechanisms that 1) allow beliefs to influence
the choice of a reference signal, and 2) mechanisms of ignoring or toning
down error signals (particularly from intrinsic and
higher-order loops). For the first mechanism, I need more than simply to
view beliefs as a kind of perception (Marken). That is, where in the
hierarchy is a belief type loop, and how does it affect output functions.
For it seems that the decision mechanism is at the output function. An
error signal goes to an output function that has more than one set of
reference signals available to it (e.g., "I could hit this guy or just smile
and pretend to ignore him"). This condition evokes a thinking mode
(switching output and input gates), the results of which determine the
choice and give us the experience of some level of self efficacy and valuing.
Free
will, if it exists, which I doubt, is at the choice point. At least the
experience of free will (e.g., "I choose not to be the type that hits
people, no matter how offensive they are.")

The all-else equal/doesn't matter condition strikes me as very unlikely.
There are too many loops involved. Every choice one makes will adversely
affect higher-order perceptions, at least for an adult. Hence (2).

In terms of (2) - toning down an error signal, or preventing one from
happening - may relate to the lag of the loops. Is this experienced as "I
just did not think of that at the time?" But what is the experience "I
swallowed my pride and just walked away"?

I am trying to reconcil subjective experience, which influences Locke and
Bandura, with PCT. The subjective experience must be just as explanable
as behavior. It is just a matter of translating PCT to explain this
experience (and the causal apparence of this experience on behavior).

I guess I will address the testing stuff later, except to thank Bill P.
and L. for correcting Tom.

Later

Jeff