Bayes and Shannon

[from Tracy Harms (2008 4 3 11:16 Pacific)]

Martin Taylor wrote:

[You can derive] the probability distribution that
defines uncertainty … from a model or from
experience with similar situations. However it is
done, the result is a subjective measure (i.e. a
perception), for both Bayes and Shannon.

Although the works of Bayes and Shannon can be used to apply (or promote) subjectivist methodologies and/or philosophical stances, the mathematics of either theory does not require such interpretation. Even if subjectivist interpretations were to be presumed, moreover, it would not do to draw an equivalence between “a subjective measure” and “a perception” as seems to occur here. Consider, for example, how any objective measure could also be said to be “i.e. a perception”; thus subjectivity, per se, provides no increased applicability
to PCT.

I am, for the record, interested in the theories of both Bayes and Shannon, both in abstract analysis and in application. My evaluation is that what is routinely described as “subjective” in each theory is simply a matter of content that has been premised, but that is distinct from the premises of the theory itself. The theories apply to situations where additional premises are involved, such as bodies of collected statistical data or messages to be sent through signal systems. Such premises are highly abstract, and have roughly no dependency on the sort of thing known as perception within PCT discussions.

Tracy Harms

P.S. I sent an earlier post on this topic, which does not appear to have made it through the list-serv.

···

You rock. That’s why Blockbuster’s offering you one month of Blockbuster Total Access, No Cost.

[From Bill Powers (2008.04.03.1556 MDT)]

Tracy Harms (2008 4 3 11:16 Pacific) --

Both copies of this post got here.

Excellent point that "it's all perception", whether the perception is labeled "subjective" or "objective."

It might pay to recognize that all the mathematics of theories is done in the vicinity of the program level. So it's subjective, too. However, if everything is subjective, then pointing that out becomes a bit useless. The only time it is relevant is when someone starts talking or acting as if in contact with external reality in some direct way, a la Gibson. Then a reminder is called for.

You're pretty sharp these days, Tracy.

Best,

Bill P.

(Gavin Ritz, 2008.04.04.17.52NZT)

The question begs is there any such thing as an external reality "whatever
that may mean".

Of course when one talks of something objective we are alluding to something
quantitative, process, skill, as distinct from quality, structure, internal
knowing,. Ie the being-becoming paradigm.

The Copenhagen Interpretation tells us it's all a dream.

A brief aside.

What are the three experiments (like the three that more or less prove, the
Big Bang theory), that proves the PCT theory (conjecture).

···

-----Original Message-----
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@LISTSERV.UIUC.EDU] On Behalf Of Bill Powers
Sent: Friday, 4 April 2008 11:02 a.m.
To: CSGNET@LISTSERV.UIUC.EDU
Subject: Re: Bayes and Shannon

[From Bill Powers (2008.04.03.1556 MDT)]

Tracy Harms (2008 4 3 11:16 Pacific) --

Both copies of this post got here.

Excellent point that "it's all perception", whether the perception is
labeled "subjective" or "objective."

It might pay to recognize that all the mathematics of theories is
done in the vicinity of the program level. So it's subjective, too.
However, if everything is subjective, then pointing that out becomes
a bit useless. The only time it is relevant is when someone starts
talking or acting as if in contact with external reality in some
direct way, a la Gibson. Then a reminder is called for.

You're pretty sharp these days, Tracy.

Best,

Bill P.

[From Erling Jorgensen (2008.04.04 1125 EDT)]

Gavin Ritz, 2008.04.04.17.52NZT

A brief aside.

What are the three experiments (like the three that more or less prove,
the Big Bang theory), that proves the PCT theory (conjecture).

"Proves" is a stronger word than I would use, because with any theory I
think the most we can hope for is "not yet disproven," in the sense of
falsified. However, I do think the concept "proof of priniciple" is an
apt one, in showing that a theory, properly implemented, can deliver on
what it proposes.

I'm not sure the following short list fills what you are seeking. But I
can think of three early working demonstrations of PCT that led to an
"aha" sense for me.

1) Powers' series of 4 articles in the Byte journal (I think it was in
1979). In one of them, he explained & showed a "toy model" demonstration
of a hierarchical program controlling in X and Y dimensions, and a
weighted sum of them (similar to a notion like "muscle tone"). To me it
clearly showed how all variables were changing (and being controlled)
simultaneously, and how the hierarchical linkage worked. Those were
pivotal insights for me in realizing PCT could lead to actual working
models of physical processes.

2) I think another pivotal experiment, accompanied by a working model
which closely mimicked human performance in a similar task, was the one
that Marken & Powers produced where the polarity of the linkage was
reversed part way through the experiment. It was a standard tracking
task, but with room for a higher level control system. When the polarity
reversal occurred, initially a runaway feedback situation was created,
where the cursor went increasingly away from the target and up off the
screen. But very quickly, for both the human runs & for those implemented
by the model, the higher level control system reversed the sign of its
own output, so that negative feeback control was restored. The match
between the model and the human performer was quite striking, and it
showed me that this hierarchical arrangement of control systems had a
lot of rigor.

3) I can think of a third demonstration that showed another foundational
aspect of PCT. It's the Crowd demo, that Powers produced (with input
from the sociologists). A number of "agents" are modeled, with extremely
simple abilities, such as follow a given target and avoid collisions in
the process. The "sensory" capabilities are obviously condensed in the
model, but nonetheless, each agent can control its own position via the
rudimentary program statements. The "aha" part of it comes with observing
the emergent patterns that happen within the "crowd," despite none of
the agents controlling anything at that level of perception. It shows
that even seeming sophisticated sociological patterns could just be the
emergent side-effects of controlling much more rudimentary perceptions.
At the very least, it suggests we should see how much mileage we can get
out of more rudimentary levels, before invoking higher level explanations.

I'm sure other people can come up with a different short list of
foundational PCT experiments / demonstrations. But these were three
that came readily to mind for me, when reading your query.

All the best,
Erling

[From Rick Marken (2008.04.04.0930)]

(Gavin Ritz, 2008.04.04.17.52NZT)

A brief aside.

What are the three experiments (like the three that more or less prove, the
Big Bang theory), that proves the PCT theory (conjecture).

I don't believe that a theory can be proven (in the mathematical
sense). However, if by "prove" you mean "critically test" (which is
one of the dictionary meanings of "prove") then my vote for the three
experiments that prove PCT are:

1. Basic compensatory tracking experiment (reported in the final
installment of Powers' 1979 _Byte_ series) that shows a low
correlation (~0) between input and output, a low correlation between
disturbance and input and a nearly perfect correlation (~.99) between
the undetectable disturbance and output. Supplement this with my
"repeated identical disturbance" version of the compensatory tracking
task (reported on p. 61-67 of _Mind Readings_) showing that the same
outputs are produced on different trials with completely different
inputs, and you have "proof" that an open loop, causal model cannot
account for output variations in a closed loop task. But the way,
these are the first two java demonstrations at my website:
http://www.mindreadings.com/demos.htm.

2. Compensatory tracking experiment where a participant secularly
changes the controlled position of cursor (reported as Experiment 2 in
Powers 1978 _Psychological Review_ paper), showing that it is the
controller, not the input display, that determines the target position
of the input. This is a very simple but powerful demonstration that
behavior is control of -- not by -- input. A variation of this is my
"Mind reading" demo when the computer is able to determine which of
three inputs is being moved "on purpose" (controlled) by the
participant. A current version of this experiment can be found at
http://www.mindreadings.com/ControlDemo/Mindread.html.

These first two experiments prove (critically test) the first "law" of
PCT which is:

(1) i ~ r

That is, input variations, i, depend on variations in reference
specifications for that input (intentions), r, rather than input being
the cause of output, o.

3. The third experiment is a demonstration of the "Behavioral
Illusion". It is presented as Exp. 4 in Powers 1978 _Psychological
Review_ paper. This is another compensatory tracking (control) study.
In this case the nature of the feedback connection from output to
input is changed. The study shows that participants automatically
compensate for this change by producing output such that the
relationship of output to input is the inverse of the feedback
function. This demonstrates the second "law" of PCT, which is:

(2) o = -1/g(d)

That is, in order to keep input under control (first law) the
controller must produce output variations, o, that oppose
disturbances, d, to that input; and these output variations must be
inversely proportional to the feedback function, g(), that connects
outputs to disturbance. Bill and I have submitted a paper (still under
review) showing that Stevens' power law (o =d^a) relating magnitude
estimates, o, to stimulus intensity, d, is an example of a behavioral
illusion if the perception of the magnitude estimates is a log
function of those estimates. In this case the feedback function
relating output (magnitude estimates) to input (perception of the size
of those estimates relative to the intensity of the stimulus being
evaluated) is logarithmic and the power function, which is the
observed relationship between i and o, is the inverse of this
logarithmic feedback function. The illusion, if course, is that the
observed relationship between d and o (1/g()) reflects characteristics
of the participant (controller) when, in fact, it is actually the
inverse of the environmental feedback connection between o and i.

Best regards

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com

(Gavin Ritz 2008.04.05.18.40NZT)

[From Rick Marken (2008.04.04.0930)]

(Gavin Ritz, 2008.04.04.17.52NZT)

Thank you for that. I have looked at this before. I'm not that clear on this
at all. Do you have anything a bit simpler I can start with? I would like to
test this for myself. I'm not even sure what you are testing.

I have trundled through all the books (some three times) but still do not
have a good grasp of this model.

Humour me and help me slowly through this I need to build an integrated
perception of this model.

Gavin

[From Rick Marken (2008.04.05.0930)]

Gavin Ritz (2008.04.05.18.40NZT)

Thank you for that. I have looked at this before. I'm not that clear on this
at all. Do you have anything a bit simpler I can start with? I would like to
test this for myself. I'm not even sure what you are testing.

I don't have anything simpler. But I'd be happy to try to answer your
questions about these demos (I assume you are talking about the demos
-- particularly the first few -- at
http://www.mindreadings.com/demos.htm).

I have trundled through all the books (some three times) but still do not
have a good grasp of this model.

That's hard to believe from someone who apparently understand quantum
physics. But if you say so I'd be happy to try to help.

Humour me and help me slowly through this I need to build an integrated
perception of this model.

Again, I'll will be happy to help you if I can (and with humor, if you
find it such). I think the best way to do that is for you to ask
specific questions about the demos. I think such a discussion might
also help many others who are listening in on this list.

Best

Rick

···

---

Gavin

--
Richard S. Marken PhD
rsmarken@gmail.com

(Gavin Ritz 2008.04.06. 10.24NZT)

[From Rick Marken (2008.04.05.0930)]

Gavin Ritz (2008.04.05.18.40NZT)

Okay thank you.

Thank you for that. I have looked at this before. I'm not that clear on

this

at all. Do you have anything a bit simpler I can start with? I would like

to

test this for myself. I'm not even sure what you are testing.

I don't have anything simpler. But I'd be happy to try to answer your
questions about these demos (I assume you are talking about the demos
-- particularly the first few -- at
http://www.mindreadings.com/demos.htm).

I have trundled through all the books (some three times) but still do not
have a good grasp of this model.

That's hard to believe from someone who apparently understand quantum
physics. But if you say so I'd be happy to try to help.

Humour me and help me slowly through this I need to build an integrated
perception of this model.

Again, I'll will be happy to help you if I can (and with humor, if you
find it such). I think the best way to do that is for you to ask
specific questions about the demos. I think such a discussion might
also help many others who are listening in on this list.

Best

Rick

···

---

Gavin

--
Richard S. Marken PhD
rsmarken@gmail.com

[from Tracy B. Harms (2008-04-05 18:05 Pacific)]

Bill Powers (2008.04.03.1556 MDT) wrote:

Tracy Harms (2008 4 3 11:16 Pacific) –

Both copies of this post got here.

Actually, those posts had different contents.

Excellent point that “it’s all perception”, whether the
perception is labeled “subjective” or “objective.”

It might pay to recognize that all the mathematics of
theories is done in the vicinity of the program level. So
it’s subjective, too. However, if everything is subjective,
then pointing that out becomes a bit useless.

That pattern of uselessness is definitely what I was pointing to. I’m tending to find the entire discussion of the role of the mind and/or subjectivity to be leaning heavily in the “useless” direction, frankly. Either I don’t understand what problem is intended to be solved by emphasizing subjectivity, or it is not something that holds my interest as an alleged problem.

I want to point out a shift you made in the quoted post, Bill, which I think was mere carelessness but is the sort of carelessness that can give life extension to topics that ought to perish. You agreed with me that everything that comes under discussion is perception-dependent, and indeed this is what I tried to communicate. Shortly thereafter, however, you moved to the claim that everything is subjective, which is certainly not something I’m inclined to agree with. At first I thought you saw this as what I’d said, but now I’m less sure.
It matters little, since you brought it up only to emphasize that anything that is true about everything does not give us a tool for discernment. Of course, I agree with that. What I don’t agree with is that this idea – “everything is subjective” – is one to which it is worthwhile to concede even in that small way. The whole notion seems to promote, or accompany, a metaphysical position that seems to move discussion away from what perceptual control theory assumes.

I must admit, though, that I do recognize that it can be important to iron such things out. The recent discussion on uncertainty seems to turn, ultimately, on the difficulty of adequately incorporating into our descriptions the subjectivity of all perception. I don’t want to squelch inquiry here, but I do want to propose that there are many ways of dwelling on subjectivity that are counter-productive to the advancement of PCT as it
stands.

Tracy

···

You rock. That’s why Blockbuster’s offering you one month of Blockbuster Total Access, No Cost.

[From Bill Powers (2008.04.06.0401 MDT)]

Tracy B. Harms (2008-04-05 18:05 Pacific)

I want to point out a shift you made in the quoted post, Bill, which I think was mere carelessness but is the sort of carelessness that can give life extension to topics that ought to perish. You agreed with me that everything that comes under discussion is perception-dependent, and indeed this is what I tried to communicate. Shortly thereafter, however, you moved to the claim that everything is subjective, which is certainly not something I'm inclined to agree with.

You're going to have to explain the difference to me. If all we can know about anything is in the form of perceptions, why doesn't that make everything subjective? Perhaps you're using the term subjective in some special sense?

Best,

Bill P

[from Tracy B. Harms (2008-04-06 11:28 Pacific)]

Bill Powers (2008.04.06.0401 MDT) replied to me and asked,

the claim that everything is subjective, which is
certainly not something I’m inclined to agree with.

You’re going to have to explain the difference to me.
If all we can know about anything is in the form of
perceptions, why doesn’t that make everything subjective?
Perhaps you’re using the term subjective in some special
sense?

I don’t think it’s a special sense. Let me return to what Martin Taylor wrote, because I think his topic is a fine area for considering and contrasting subjective and objective, and because that is what I was replying to in the first place.

However [the
probability distribution is defined,]
the result is a subjective measure (i.e. a
perception), for both Bayes and Shannon.

I replied that it will not do to draw an equivalence between “a subjective measure” and “a perception” because any objective measure could also be said to be “a perception”. This reveals that subjectivity vs objectivity does not qualify anything as lying inside or outside the analytic scope of PCT.

Thus I can be understood to be saying “it is all perception”, but I’m doing this by claiming that objective things (also) fall within the scope of that-which-is-perceptually-dependent. I am not doing this by claiming that there are no objective things. Under that alternative claim, “it is all subjective”, only subjectivity would be actual and to say that anything is in any genuine manner objective would be an error.

My hunch is that centuries of philosophical struggle over objectivity and
subjectivity have deadlocked due to an absence of appreciation of the biological nature of knowledge. PCT is compatible with, and deserves to be integrated with, biologically sophisticated perspectives on knowledge. Through such a perspective there is no longer a risk that to call something objective is to disconnect theory from the fact that knowledge requires individual lives with unique and highly conditional perspectives. We may now focus on subjectivity, including the stark interiority that characterizes what occurs in an active control system, without implying that this precludes plain talk about objective facts.

Some people who like PCT do so because they see it as fitting the idea that “everything” is subjective. For some this may go all the way to solipsism. My liking for PCT involves the way it displaces emphasis on subjectivity with emphasis on perception. Doing so allows us to work with a truth that the
advocates of subjectivity have been trying to get across all along, while disengaging from the adversarial situation whereby objectivity was denied. Errors and conundrums that occur if we say “there is no reality independent of subjectivity” do not occur if we say “there is no knowledge that is independent of perception.”

A distinction between perception and subjectivity may be made by relying on PCT, and developing it is well worth doing.

Tracy

···

You rock. That’s why Blockbuster’s offering you one month of Blockbuster Total Access, No Cost.

[From Bill Powers (2008.04.08.0324 MDT)]

Tracy B. Harms (2008-04-06 11:28 Pacific) –

I’m getting behind, so excuse cursory response.

I replied that it will not do to
draw an equivalence between “a subjective measure” and “a
perception” because any objective measure could also be said to be
“a perception”. This reveals that subjectivity vs objectivity
does not qualify anything as lying inside or outside the analytic scope
of PCT.

OK, I remember that now. Yes, since it’s all perception, what we call
“objective” and “subjective” are just different
classes of perceptions.

My hunch is that centuries of
philosophical struggle over objectivity and subjectivity have deadlocked
due to an absence of appreciation of the biological nature of
knowledge. PCT is compatible with, and deserves to be integrated
with, biologically sophisticated perspectives on knowledge. Through
such a perspective there is no longer a risk that to call something
objective is to disconnect theory from the fact that knowledge requires
individual lives with unique and highly conditional perspectives. We may
now focus on subjectivity, including the stark interiority that
characterizes what occurs in an active control system, without implying
that this precludes plain talk about objective facts.

I like that. It’s always been a problem to me to reconcile the total
dependency on perception with, for example, modeling, which requires us
to pretend that we are making a model of how things actually work. The
nagging question remains, how can we possibly hope to verify anything we
say about real reality? I suppose the answer is that we can’t, so we had
better get used to making hypotheses as carefully as possible and
constructing a human picture of reality that works as well as can be
achieved. That’s “objectivity,” I guess.

I hope we get in touch with an independently-evolved intelligence pretty
soon.

Best,

Bill P.

···

Some people who like PCT do so
because they see it as fitting the idea that “everything” is
subjective. For some this may go all the way to solipsism. My
liking for PCT involves the way it displaces emphasis on subjectivity
with emphasis on perception. Doing so allows us to work with a
truth that the advocates of subjectivity have been trying to get across
all along, while disengaging from the adversarial situation whereby
objectivity was denied. Errors and conundrums that occur if we say
“there is no reality independent of subjectivity” do not occur
if we say “there is no knowledge that is independent of
perception.”

A distinction between perception and subjectivity may be made by relying
on PCT, and developing it is well worth doing.

Tracy

You rock. That’s why Blockbuster’s offering you

one month of Blockbuster Total Access
, No Cost.

No virus found in this incoming message.

Checked by AVG.

Version: 7.5.519 / Virus Database: 269.22.8/1362 - Release Date: 4/6/2008
11:12 AM