Subjective value (was Modeling an Economy)

CSGnet Archive CSG2009
1

[Martin Taylor 2009.09.02.12.43]

[From Rick Marken (2009.09.01.1055)]

Martin Taylor
(2009.09.01.08.09) –

I don’t think “experience” is the same as “value”. You can have
experiences that you don’t value at all highly.

To me (and I think to PCT), experience means “conscious perception”. A
“valued” experience would, thus, be a reference signal. An experience
that is not valued is just a perception that is not in the state
specified by the reference.

There are three sentences here, not one of which follows from its
predecessor. Reference signals don’t relate to conscious perception any
more than to non-conscious perception. To relate “subjective value” to
the quantitative value of a reference signal seems to me to be a
perverse way of mushing up two difference uses of the same letter
string, rather akin to trying to enter a value into a cell of a coffee
table rather than a spreadsheet table. I hope the perversity was not
deliberate, done with the intention of disturbing some perception you
think I might be controlling. And finally, to say an experience has
value ONLY if the perception is at its reference level seems remarkably
outside the normal thinking in PCT, which usually considers variables
to have continuous ranges of possible magnitude. You say it has a
magnitude (though how assessed is unclear) when the corresponding
perception is exactly at its reference value, but has no magnitude,
neither high nor low, when the corresponding perception is not at its
reference value. I find that very strange, coming from one who is
usually quite strict in following classical PCT thinking.

However, after posting my
comment that minimizing error was equivalent
to maximizing the subjective value. it occurred to me that subjective
value also has some element of gain about it.

You seem to be making up stuff that is not part of the PCT model.
“Subjective value” is not part of PCT. It’s an informal term that could
refer to several different things that are already part of the PCT
model.

I do wish you wouldn’t use “the PCT model” as though the Bill Powers
strict hierarchical model were the only permissible PCT model. I
acknowledged that my suggestion was “heretical”, by which I meant that
it did not conform to the received religious truth. I don’t think,
however, that it conflicted in any way with PCT. It suggested the
possibility of perceptions based on the values of error signals and
gain parameters, which is not part of Powers HPCT. However, Powers HPCT
does use the values of error signals which are sensed by the
reorganizing system, and Powers has from time to time alluded to
control by gain variation, from which it is not a long step to
suggesting that gain may be sensed by a system that controls gain.
Maybe my suggestion wasn’t so heretical after all! Perhaps “subjective
value” is a perception relating to reorganization? Reorganization would
be unhelpful if it eliminated something of high subjective value, and
helpful if it enhanced the likelihood of attaining perceptions of high
subjective value.

One doesn’t perceive much
subjective value in something about which one does not care much, even
if it is vry close to its reference value.

Are you talking about perceiving the subjective value of a perception?
Or is subjective value itself a perception?

Yes, subjective value is clearly a perception, often a conscious
perception. It is a perception about the present quantitative value of
some quite different perception. An artist, for example, might have a
high subjective value for a particular level of “redness” at a specific
location in a picture, but a low subjective value for either more or
less red. The level of redness that is perceived (controlled for) by
the artist is a quite different perception than the subjective value
perceived when that level of redness has been attained.

I would suggest that we agree on what “subjective value” means
in terms of PCT. I suggest subjective value means the same as reference
value. That’s it.

Or would be, if that suggestion had made any sense, rather than
requiring the use of the same letter string to refer to two radically
different concepts.

So now we can dispense with the notion of subjective value. It
seem lieka useless term to me.

Yes, I realize you are a minimalist, but remember the other half of
Einstein’s doctrine, that constructs do need to be complex enough to
accommodate the real world.

I would say it this way:

I have a reference for perceiving more people able to lead a good life.
I also have a reference for being wealthy enough. I don’t see those
goals as being in conflict so I don’t know whether one has more “value”
to me (in the sense that I control for it with higher gain) than the
other. They are both equally valuable (or valueless). They just are
goals I have. This, though, makes me realize that when I do have the
experience of evaluating “subjective value” it’s only when I’m in a
conflict.

Right, perhaps this is a minimalist way of doing something analogous to
MOL. You have noticed that you can compare the subjective value of two
radically different perceptions. The economists also have noticed this,
when they say that for a buyer the subjective value of the thing bought
is greater than that of the money used to buy it, whereas for the
seller the opposite is true. If there is no sum of money for which
those two inequalities are satisfied, then there can be no trade freely
agreed. Maybe if you ask yourself what led you to this thought, and
what other implications follow from it, interesting ideas might emerge.
You said: "You are developing model to account for a phrase
“subjective value”. I would rather develop models on the basis of data.
" Good for you, but wouldn’t you then be following in the footsteps of
the behavioural economists, who do just that – though without the
benefit of a grounding in PCT? You could gather all sorts of data for
controlled variables that might or might not differ in subjective
value, as determined by the person’s preference when those perceptions
are put into conflict (as with the kids who can have two candies if
they wait, or they can eat the one in full view now). Bill P hinted at
the outline of such an experiment when he discussed modelling [From
Bill Powers (2009.09.01.0838 MDT)].

For example, when I’m deciding “should I get the doughnut or the
muffin” I guess I experience higher “value” for the side of the
conflict that is currently “winning”. But otherwise I don’t really
experience “subjective value” except when I am getting the perceptions
I want, which is nearly all the time. So I value all my perceptions
(when I am controlling successfully) of the same “value”.

Ah. It was a fleeting insight, summarily dismissed. Maybe it will
return, with luck :slight_smile: Do give it a chance!

Martin

2

[From Rick Marken (2009.09.02.1250)]

Martin Taylor (2009.09.02.12.43) –

I do wish you wouldn’t use “the PCT model” as though the Bill Powers
strict hierarchical model were the only permissible PCT model. I
acknowledged that my suggestion was “heretical”, by which I meant that
it did not conform to the received religious truth.

Modeling is not a religious practice. Models are built to account for data. A good model is one that gives the best fit to data in the most economical way. There is nothing “heretical” about changing a model as long as the change is done in order to account for data that cannot be accounted for by the existing model. I would have no objection to your suggested changes to PCT as long as I knew what data motivated the changes.

Yes, subjective value is clearly a perception, often a conscious
perception. It is a perception about the present quantitative value of
some quite different perception. An artist, for example, might have a
high subjective value for a particular level of “redness” at a specific
location in a picture, but a low subjective value for either more or
less red.

But that behavior (the artist’s selection of a particular red over another) is already accounted for in PCT as a higher level system selecting the reference for a lower (in this case sensation) level system.

The level of redness that is perceived (controlled for) by
the artist is a quite different perception than the subjective value
perceived when that level of redness has been attained.

How do you know? Where’s the data that demonstrates this to be the case?

The economists also have noticed this,
when they say that for a buyer the subjective value of the thing bought
is greater than that of the money used to buy it, whereas for the
seller the opposite is true.

This is not something the economists “noticed” (observed). It’s something they imagined to be happening; it’s a theory. Bill Powers described the PCT model of this same phenomenon and it the PCT model works (explains the data) without any “subjective value” being involved.

You said: "You are developing model to account for a phrase
“subjective value”. I would rather develop models on the basis of data.
" Good for you, but wouldn’t you then be following in the footsteps of
the behavioural economists, who do just that – though without the
benefit of a grounding in PCT? You could gather all sorts of data for
controlled variables that might or might not differ in subjective
value, as determined by the person’s preference when those perceptions
are put into conflict (as with the kids who can have two candies if
they wait, or they can eat the one in full view now).

What you are suggesting is developing experiments to test a theory (subjective value) of a phenomenon (choosing X over Y) which is already explained by a theory (PCT) that involves no concept of “subjective value”. I don’t think I’m going to waste my time on such experiments.

For example, when I’m deciding “should I get the doughnut or the
muffin” I guess I experience higher “value” for the side of the
conflict that is currently “winning”. But otherwise I don’t really
experience “subjective value” except when I am getting the perceptions
I want, which is nearly all the time. So I value all my perceptions
(when I am controlling successfully) of the same “value”.

Ah. It was a fleeting insight, summarily dismissed. Maybe it will
return, with luck :slight_smile: Do give it a chance!

I did and I’ve concluded that I don’t experience subjective value at all; I just experience a release of tension (when the conflict is resolved). I suppose I could verbally describe the choice in terms of value but that’s because I know that the word “value” refers to something I preferred. So if I said I value the doughnut more (if I selected it) all I’m describing is the fact that I did select it. There are emotions associated with a conflict like that; maybe it’s the experience of those emotions (like the good feeling you get when you finally make the choice) that you call “subjective value”. But I’ve given it a chance and I really can’t say that I experience more of something called “value” when I compare perceptions, such as the perceptions of Bush and Obama. All I experience are emotions (revulsion in one case, happiness in the other; guess which goes with which;-)

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

3

[Martin Taylor 2009.09.03.00.40]

[From Rick Marken (2009.09.02.1250)]

Martin Taylor
(2009.09.02.12.43) –

The level of redness that is
perceived (controlled for) by
the artist is a quite different perception than the subjective value
perceived when that level of redness has been attained.

How do you know? Where’s the data that demonstrates this to be the case?

What on earth kind of data do you need in order to convince you that
the perception of the value of something is different from the
perception of the something. Do you need data to demonstrate that the
perception of the configuration of clouds is different from the
perception of the beauty of the sky?

You really do go to extraordinary lengths to control your perception
that anything I suggest must be wrong, don’t you? Why is it so
important to you that this must be so?

Martin

4

[From Rick Marken (2009.09.03.2215)]

Martin Taylor (2009.09.03.00.40)–

[From Rick Marken (2009.09.02.1250)]

Martin Taylor
(2009.09.02.12.43) –

The level of redness that is
perceived (controlled for) by
the artist is a quite different perception than the subjective value
perceived when that level of redness has been attained.

How do you know? Where’s the data that demonstrates this to be the case?

What on earth kind of data do you need in order to convince you that
the perception of the value of something is different from the
perception of the something.

I don’t know. This idea that there is a perception of the value of something that is different from the perception of the something is your idea. What data leads you so believe this is true?

Do you need data to demonstrate that the
perception of the configuration of clouds is different from the
perception of the beauty of the sky?

Of course. At least, when were talking about a model.

You really do go to extraordinary lengths to control your perception
that anything I suggest must be wrong, don’t you?

Not really.

Look, I don’t care if you are wrong or right. I just care about the ideas. This idea of subjective value seems unnecessary to me, given my understanding of human nature in terms of PCT. All I’m doing is telling you what I think. Feel free to try to convince me that subjective value is a necessary concept. Believe me, I respond well to demonstrations and data; that’s what got me sold on control of perception. I’ll happily recognize your correctness about “subjective value” (as a separate property of perception) if you can demonstrate it to me.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

5

Here is a good explanation of subjective value as viewed by the Austrian school of economics;

“Value is the magnitude of the feeling of future well being one expects from taking a particular action; it is clearly established as future-oriented, subjective, and not cardinally measurable. In each choice that one confronts, one must conceive of the value of each alternative, rank them by value, and then chooses the highest-valued of the available alternatives” from http://mises.org/journals/scholar/egger.pdf

Rick, et al, Is it your point that such a rational subjective valuing process is an illusion, and/or that PCT can somehow put numbers on what has been described as “not cardinally measurable”?

The Lucas Critique should perhaps be considered when evaluating Rick’s argument that the models should match the data:

“The Lucas critique, named for Robert Lucas’ work on macroeconomic policymaking, says that it is naïve to try to predict the effects of a change in economic policy entirely on the basis of relationships observed in historical data, especially highly aggregated historical data.” from http://en.wikipedia.org/wiki/Lucas_critique

I know that when I see analyses of the recent and past economic crises, I cringe at simplistic correlation analyses that blame the actors involved, and ignore the distortions in the markets introduced by regulations, taxes and policy decisions, which incentivised the behavior of those actors. It was interesting for instance to see the US consumers criticised for a low savings rate, at the same time analysts were commenting on the savings glut in the rest of the world, and the low rates of interest, and the ever higher risks that savers were taking in search for any kind of decent return. In such an economic climate, the wonder should be that anyone wanted to save at all. The incentive was to borrow at low rates and expect to pay back in cheaper dollars. No wonder the economy became highly leveraged and we had a bubble.

regards,

Martin

6

[Martin Taylor 2009.09.03.10.05]

Martin Lewitt wrote: (apparently 2009.09.03.09.49)

The Lucas Critique should perhaps be considered when evaluating Rick’s
argument that the models should match the data:

“The Lucas critique, named for Robert Lucas’ work on
macroeconomic policymaking, says that it is naïve to try to predict the
effects of a change in economic policy entirely on the basis of
relationships observed in historical data, especially highly aggregated
historical data.” from http://en.wikipedia.org/wiki/Lucas_critique

The validity of that critique depends on what is conceived of as a
“model”. In a PCT analysis, a “model” could be called a “machine”. It
expresses the causal relationships among variables. For example, in the
canonical elementary control loop, there is a causal relationship
between the sensory input and the magnitude of the perceptual variable,
and between the magnitudes of the perceptual and reference variables
and th error signal, and so forth. A PCT-based model does NOT base
expectations on what happened historically. Its predictions are based
on what the machinery will do, given the current values of the
variables. History IS used, but used in setting the parameter values
for the model under test, in much the same way as climate history is
used to stress-test global climate models. If the model is structured
correctly, and the parameters set correctly, then the model should
describe the probable results of an economic policy, which could be
quite different from what happened in a superficially similar past
situation.

Martin

7

[From Bill Powers (2009.09.03.0713 MDT)]

Martin Taylor 2009.09.03.00.40 --

MT: [To Rick Marken]What on earth kind of data do you need in order to convince you that the perception of the value of something is different from the perception of the something. Do you need data to demonstrate that the perception of the configuration of clouds is different from the perception of the beauty of the sky?

BP: I think the basic difficulty here comes from trying to mix informal common language with control theory, which is a technical subject and requires more carefully defined terms. Data will not show the difference between perception of something (at one level) and its value or usefulness in contributing to something else (at a higher level). That is a question for the proper theory to answer. The specialized language of PCT allows us to say more exactly what we mean when that becomes important.

The value of something to its possesser or perceiver can be its role (in the feedback function) in controlling an input, an immediate end in itself (a reference signal specifying how much or little of it is preferred), or its instrumental usefulness (its contribution to a higher-level controlled variable). If we have less of something we value than we want, value can refer to the state of the perception (the actuality) or to the reference level (the wished-for state), or for that matter, to the error signal (that thing is not worth ten dollars). In any case, we try to get more of it if what we have is less than what we want. If we don't want any of something (like pain), the value can be negative, so we try to get rid it. But the value may lie in having just the right amount of it, neither much more or much less, as in the case of spices or salt -- value can mean a specific amount or degree, with either more or less being less valuable. You may give something a high value when evaluating it, and not desire it (that car is far too expensive and ostentatious).

Value is also a verb. To say you value something is to say you are satisfied with its state (you are the proud owner), or to say that you are trying to get more of it than you have. You may also mean that you are trying to have less of it, because you give it a low or negative valuation.

The biggest controversies about value have to do with its nature: is it something objective that anyone can see, or subjective and different for each individual? Does the dollar value depend only on negotiation, or is there some kind of gold standard that will correctly express the real value of things? Are family values the same for everyone, or should they be? Could they be?

All these varied meanings of value only show how little is understood about the way people work. Without PCT there is simply no way to understand what people want and don't want, or even what "wanting" means. Intentions, goals, imagination, perception, and even action remain mysterious, giving rise to wildly varying speculations about what they might mean or how they work. Common usages of words reflect that ignorance and confusion.

So it's a mistake to try to convert PCT into common-language terms. It can't be done, or rather it has to be done over again each time the same common term is used in a different context.

As I tried to show in my brief outline of a model of movie-going versus budget, there is no need to base models on common-language terms, and in fact trying to do so is futile if the critical terms are as ambiguous as "value". Value will never be a variable in a PCT model of the economy. Once a PCT model is complete and working, it will be possible to point out many different features in it that could represent different usages of the term value, but because of that, value itself is useless as the basis for building a model.

Best,

Bill P.

8

[From Bill Powers (2009.09.03.0825 MDT)]

Martin Taylor 2009.09.03.10.05 --

A PCT-based model does NOT base expectations on what happened historically. Its predictions are based on what the machinery will do, given the current values of the variables. History IS used, but used in setting the parameter values for the model under test, in much the same way as climate history is used to stress-test global climate models. If the model is structured correctly, and the parameters set correctly, then the model should describe the probable results of an economic policy, which could be quite different from what happened in a superficially similar past situation.

I fully concur with this observation, in all details. I would add only that if regulations and so on have the effects that Martin Lewitt describes, then when they are put into an agent-based model that correctly represents human nature, the results he expects will occur. If they don't occur, then either he or the model is wrong, and we can go on from there.

Best,

Bill P.

9

[From Rick Marken (2009.09.03.0910)]

Bill Powers (2009.09.03.0713 MDT)–

Martin Taylor 2009.09.03.00.40 –

MT: [To Rick Marken]What on earth kind of data do you need in order to convince you that the perception of the value of something is different from the perception of the something. Do you need data to demonstrate that the perception of the configuration of clouds is different from the perception of the beauty of the sky?

BP: I think the basic difficulty here comes from trying to mix informal common language with control theory, which is a technical subject and requires more carefully defined terms.

Exactly. But Martin is proposing the common language concept (subjective value) as an actual theoretical entity. I’ve been trying to explain how the phenomena that he describes as involving “subjective value” can be explained by PCT (without any “subjective value” component added). Maybe your explanation will help him see the problem.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

10

[From Rick Marken (2009.09.03.0925)]

Bill Powers (2009.09.03.0825 MDT)–

Martin Taylor 2009.09.03.10.05 –

A PCT-based model does NOT base expectations on what happened historically…

I fully concur with this observation, in all details. I would add only that if regulations and so on have the effects that Martin Lewitt describes, then when they are put into an agent-based model that correctly represents human nature, the results he expects will occur. If they don’t occur, then either he or the model is wrong, and we can go on from there.

I concur as well. But it seems to me that, before we have such a model (which might take a while to produce), it behooves us to look at the data. Of course, there is no experimental data on this. But there is what your admirer Donald T. Campbell called quasi-experimental data available. This is the macroeconomic time series data that I’ve been looking at for the last few years. It is quasi-experimental because the “experimental” manipulations – policy changes – are varied randomly over time while other factors vary as well, partially confounding the results. So, for example, regulation policies have varied quite a bit over the last 100 years but so have other economic conditions (technologies, workforce education, population, etc). What makes this quasi-experimental research is that you get to see the effect of each regulation policy (say high versus low regulation) at many different levels of the confounding variables.

Looking at the “effect” of a particular regulation policy while averaging over the varying values of the different confounding variables should give a relatively good picture of the “true effect” of that policy. Of course, this is not as good as a true experimental test of a policy, varying the policy while all other factors are held constant. But, of course, we can’t do this with economics and we’ll have to wait for a good model to see how various policies (like regulation) actually affect economic performance. But until then it seems like it is reasonable to use the quasi-experimental data that is available to inform our economic policy decisions.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

11

[Martin Taylor 2009.09.03.12.18]

[From Rick Marken (2009.09.03.0910)]

Bill
Powers (2009.09.03.0713 MDT)–

Martin Taylor 2009.09.03.00.40 –

MT: [To Rick Marken]What on earth kind of data do you need in order to
convince you that the perception of the value of something is different
from the perception of the something. Do you need data to demonstrate
that the perception of the configuration of clouds is different from
the perception of the beauty of the sky?

BP: I think the basic difficulty here comes from trying to mix informal
common language with control theory, which is a technical subject and
requires more carefully defined terms.

Exactly. But Martin is proposing the common language concept
(subjective value) as an actual theoretical entity. I’ve been trying to
explain how the phenomena that he describes as involving “subjective
value” can be explained by PCT (without any “subjective value”
component added). Maybe your explanation will help him see the problem.

No. I have no problem with anything Bill wrote. It simply did not
address my point that one has a perception of the value of something,
a perception to which one can attach a relative magnitude, as when one
perceives X to be of more value than Y, whether X and Y are perceptions
of actual states or reference values for controlled perceptions, or one
of each. When one is making a trade, one perceives that the status
quo has or has not a greater value than making the trade,

My speculation was about what variables contribute to the perception
of value, not about the observed actions or about modelling the
resolution of conflict. It’s a question on the same level as what
variables contribute to a perception of, say, a category or a
configuration, both of which, if the definition is not to be circular,
depend on a common language concept to define what the corresponding
function should achieve and what variables serve as inputs to the
function.

Discussion is not at all helped by the introduction of absurdities such
as assertions that the perception of beauty is another name for the
perception of configuration. Such assertions verge perilously close to
trolling, a practice usually absent from CSGnet. To avoid escalation, I
find good practice is to try to ignore the troll. I don’t always
succeed :slight_smile:

Martin

12

[From Rick Marken (2009.09.03.1045)]

[Martin Taylor 2009.09.03.12.18]

[From Rick Marken (2009.09.03.0910)]

Exactly. But Martin is proposing the common language concept
(subjective value) as an actual theoretical entity. I’ve been trying to
explain how the phenomena that he describes as involving “subjective
value” can be explained by PCT (without any “subjective value”
component added). Maybe your explanation will help him see the problem.

No. I have no problem with anything Bill wrote. It simply did not
address my point that one has a perception of the value of something,
a perception to which one can attach a relative magnitude, as when one
perceives X to be of more value than Y, whether X and Y are perceptions
of actual states or reference values for controlled perceptions, or one
of each. When one is making a trade, one perceives that the status
quo has or has not a greater value than making the trade,

But you are assuming that people have a perception of the value of something which can be given a magnitude. This is based on using ordinary language to explain a phenomenon; the phenomenon of preferring X to Y. We could describe this verbally as X being valued over Y. But that doesn’t mean that this preference is based on comparing a perception of the value of X being to a perception of the value of Y. What you call “preferring X to Y” can be explained in PCT without any reference to the non-technical term “value”. I mentioned one way: it can be explained as X being closer to the reference than Y. Simple as that.

Discussion is not at all helped by the introduction of absurdities such
as assertions that the perception of beauty is another name for the
perception of configuration.

I don’t think I ever said that but, even if I did, I don’t see it as absurd.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com

www.mindreadings.com

13

Are you assuming that one PCT model will be enough to represent all actors in the system? If not, are you intending to infer individual internal PCT models from data on past economic decision making? Would a PCT model inferred in that way, really have a chance of being a good PCT model. Perhaps I am confusing PCT with MOL. If we don’t have access to the internal MOL models of the actors, and must infer them from the historical data, how will that differ from models of actor behavior derived under the subjective values assumption?

If actors are to be modeled individually, I don’t see how a PCT based model is likely to come to a different conclusion than a subjective values based model about the key issue of the day, free markets versus centralized socialist planning. In both cases markets will be better at distributing the decision making to the individual, and incorporating the individuals feedback.

Martin L

mlewitt@comcast.net

···

----- Original Message -----
From: “Bill Powers” powers_w@FRONTIER.NET
To: CSGNET@LISTSERV.ILLINOIS.EDU
Sent: Thursday, September 3, 2009 8:29:26 AM GMT -07:00 US/Canada Mountain
Subject: Re: Subjective value (was Modeling an Economy)

[From Bill Powers (2009.09.03.0825 MDT)]

Martin Taylor 2009.09.03.10.05 –

A PCT-based model does NOT base expectations on what happened
historically. Its predictions are based on what the machinery will
do, given the current values of the variables. History IS used, but
used in setting the parameter values for the model under test, in
much the same way as climate history is used to stress-test global
climate models. If the model is structured correctly, and the
parameters set correctly, then the model should describe the probable
results of an economic policy, which could be quite different from
what happened in a superficially similar past situation.

I fully concur with this observation, in all details. I would add
only that if regulations and so on have the effects that Martin
Lewitt describes, then when they are put into an agent-based model
that correctly represents human nature, the results he expects will
occur. If they don’t occur, then either he or the model is wrong, and
we can go on from there.

Best,

Bill P.

14

Richard Marken wrote:

[Martin Taylor 2009.09.03.13.58]

[From Rick Marken (2009.09.03.1045)]

[Martin Taylor
2009.09.03.12.18]

[From Rick Marken (2009.09.03.0910)]

Exactly. But Martin is proposing the common language concept
(subjective value) as an actual theoretical entity. I’ve been trying to
explain how the phenomena that he describes as involving “subjective
value” can be explained by PCT (without any “subjective value”
component added). Maybe your explanation will help him see the problem.

No. I have no problem with anything Bill wrote. It simply did not
address my point that one has a perception of the value of something,
a perception to which one can attach a relative magnitude, as when one
perceives X to be of more value than Y, whether X and Y are perceptions
of actual states or reference values for controlled perceptions, or one
of each. When one is making a trade, one perceives that the status
quo has or has not a greater value than making the trade,

But you are assuming that people have a perception of the value of
something which can be given a magnitude.

Yes, based at least in part on introspection. I do, so I tend to assume
other people do, too. Of course it’s possible I’m the only one, but
isn’t equally possible you are the only one who doesn’t?

This is based on using ordinary language to explain a
phenomenon; the phenomenon of preferring X to Y. We could describe
this verbally as X being valued over Y. But that doesn’t mean that this
preference is based on comparing a perception of the value of X being
to a perception of the value of Y. What you call “preferring X to Y”
can be explained in PCT without any reference to the non-technical term
“value”. I mentioned one way: it can be explained as X being closer to
the reference than Y. Simple as that.

OK. Let’s start there. What does it mean for the perception of the
anticipated (imagined) taste of this apple being closer to its
reference value than the perception of the quantity of money available
to me being to its reference value? I know that for me, at this moment,
I would find the value of $1000 to be greater than the value of the
apple taste, and that the value of 10c would be less than the value of
the apple taste. So, if someone were to offer me the kind of apple I
imagine and ask for 10c in exchange, I would accept the offer, but if
that person were to ask for $1000 in exchange, I wouldn’t. I perceive
those values to be arranged on a scale of magnitude in much the same
way as I can perceive the lengths of two lines as one being longer than
the other. You can’t. That’s OK, but please don’t tell me I can’t.

Discussion is not at all
helped by the introduction of absurdities such
as assertions that the perception of beauty is another name for the
perception of configuration.

I don’t think I ever said that but, even if I did, I don’t see it as
absurd.

That really is hard to believe!

Martin

15

[From Bill Powers (2009.09.03.1118 MDT)]

Martin Taylor 2009.09.03.12.18 –

MT: … one has a perception
of the value of something, a perception to which one can attach a
relative magnitude, as when one perceives X to be of more value than Y,
whether X and Y are perceptions of actual states or reference values for
controlled perceptions, or one of each. When one is making a trade, one
perceives that the status quo has or has not a greater value than
making the trade.

I can accept the proposition that many people seem to work that way,
particularly women in supermarkets who have choices within range of
observation, and who even keep track of prices in different stores. I
don’t work that way, and have known women who find it irritating or
puzzling that so many men don’t (other than male price theorists). A man
walks in and says I need two pounds of sugar, buys the first package he
sees, and goes home.
I say “seem” to work that way because I doubt that this
principle is behind the way most people, even women, make their choices .
It just seems implausible to me that any person compares all the
perceptual qualities of each item desired against a similar set of
qualities of all other items as a way of reaching a final decision. I can
imagine this happening when there is a set of identical items with
different prices, but the combinatorials in the general case boggle the
mind, especially when you include the obvious interactions (sauerkraut
does not go well with chocolate syrup).
The same appearance can be generated in an entirely different way.
Suppose there are three different items: F, a food; T, a means of
transportation, and M, a movie. There is also R, a cash reserve, and Y,
an income (fixed, for simplicity).
Imagine that there are four reference signals F*, T*, M*, and R*, and
four control systems trying to control the corresponding variables F, T,
M, and R. The means of controlling F, T, and M consists of spending
money, and the means of controlling R is to work. Also, control of R can
be done by varying the reference signals for F, T, and M (a two-level
model). You could assume a variable amount of working, so as to vary Y,
the income, but it’s simpler to assume a fixed income.
The problem now is how to establish the condition
F = F*
T = T*
M = M*
R = R*
This means that you need four equations, which in our case represent four
control systems. What you will find is that given Y, you can satisfy all
four as exactly as you please over some range of the reference
conditions, but not over all possible parameter values. If one or more of
the reference signals gets too large, or the cost of one or more of the
goods gets too high, or the loop gain in some of the control systems is
too low, the control systems will come to equilibrium with some degree of
error in one or all of them. The amount of error may or may not be large
enough to qualify this as a conflict situation.
At that equilibrium point it might look as if something inside the person
is comparing the “value” of each good with the
“value” of each other good, and solving the equations, or
rather inequalities, for the correct ranking of the goods in terms of
value. The control system equations will also provide such a ranking, in
that after the system comes to equilibrium, the goods will prove to be
ordered in terms of how much money is being spent on each of them and how
many units of each are being obtained per unit time. But the two
mechanisms for generating the result are completely different.
If you explained the behavior of the control system model by referring to
a process of intercomparison, the explanation would be wrong. If we
accept the control system model, the appearance of mutual
intercomparisons would be an illusion; the concept of “value”
is not used at all. We use terms like reference level, output gain, loop
gain, perception, and function.
In the simplest form of the intercomparison model, the result is only a
relative ranking, not a quantitative weighting, so there would be no way
to predict specific values of the variables. A more complex form would
take price into account as well as references for quantity of goods, and
you would have to solve as many equations as there are goods to arrive at
a result, which would, however, be quantitative instead of merely
qualitative. It would also be even less plausible.
You may recall that the late Bill Williams and I worked on a
control-system model for the Giffen Paradox in which there were two goods
and a budget. Bill W. (who was the chief reason for my signing myself as
Bill P. and thinking briefly of him every time I do now) worked out the
basic principle of the Giffen Paradox. The budgetary constraint could
force the purchases of one item to increase when the price of that item
was increased, flouting the law of supply and demand. This was
simply one form that the general control-system solution could take, the
so-called “law” of supply and demand being only one special
case. In a large system of equations representing many goods, raising or
lowering the price of any one good might either increase or decrease
purchases of that good by any amount. Only the control system model can
explain and predict that phenomenon.

My speculation was about what
variables contribute to the perception of value, not about the observed
actions or about modelling the resolution of conflict. It’s a question on
the same level as what variables contribute to a perception of, say, a
category or a configuration, both of which, if the definition is not to
be circular, depend on a common language concept to define what the
corresponding function should achieve and what variables serve as inputs
to the function.

I have no doubt that human beings can perceive something that they label
with the term “value.” What it is, I have no idea, although we
might deduce that it has something to do with preferences or reference
levels, means and ends, or usefulness for achieving other purposes.
Clearly the same term is used to refer to markedly different variables
and relationships, and unfortunately people seldom attach sufficient
modifiers to their sentences to bias the meaning obtained by the listener
toward the intended one. The perception of value depends on the
perceptual input function and what is being observed; just labeling the
perception “value” isn’t enough to tell us what the word refers
to.

The intercomparison theory should really be modeled so as to get an idea
of what processes it really requires and what the results would be. I
suspect that the computational load would be pretty staggering, or the
brain would be staggering under it. We can expect some to be unimpressed
by the complexities; they will just say “Wow, isn’t it wonderful how
much complexity the brain can handle!” These would be people
with the same genes as those who espouse the “inverse kinematics and
dynamics” model of motor behavior. It never occurs to them that
their model is wrong and maybe the brain does not have infinite computing
capacity.

Best,

Bill P.

16

From: “Bill Powers” powers_w@FRONTIER.NET
Sent: Thursday, September 3, 2009 1:39:01 PM GMT -07:00 US/Canada Mountain
*** snip ***

You may recall that the late Bill Williams and I worked on a control-system model for the Giffen Paradox in which there were two goods and a budget. Bill W. (who was the chief reason for my signing myself as Bill P. and thinking briefly of him every time I do now) worked out the basic principle of the Giffen Paradox. The budgetary constraint could force the purchases of one item to increase when the price of that item was increased, flouting the law of supply and demand. This was simply one form that the general control-system solution could take, the so-called “law” of supply and demand being only one special case. In a large system of equations representing many goods, raising or lowering the price of any one good might either increase or decrease purchases of that good by any amount. Only the control system model can explain and predict that phenomenon.

There is some question of whether a true Giffen good has ever existed, and although theorectically it could. Yes, the law of supply and demand can’t explain it, but the income effect, along with two are three other conditions can explain it. (necessity, inferiority, lack of substitutes). in a way that makes sense, and apparently can be modeled. However, even if it couldn’t be modeled, correct modeling of it would seem to be unimportant advantage of PCT. Does the PCT explanation of the phenomenon offer a different insight? The PCT alternative does not seem to offer simplicity or insight, but then neither does quantum mechanics, but like quantum mechanics, it might have the advantage of allowing better predictions. Where would those advantages be expected? What data would need to be available to realize those advantages?

regards,

Martin

···

----- Original Message -----

17

[From Bill Powers (2009.09.03.1340 MDT)]

  Martin Lewitt wrote:

ML: you assuming that one PCT model will be enough to represent all actors in the system? If not, are you intending to infer individual internal PCT models from data on past economic decision making?

BP:There would be one control system per agent in the model; we might be able to handle several thousand agents this way. We have some amazing computing power on our desk tops.

The model can contain a large number of consumers and a relatively small number of producers. Producers interact with populations, not individuals, whereas individuals act more -- well, individually. The general approach would be to give the population some average properties and then assign distributions of the parameters around those mean values to determine individual characteristics. For example, we could have the producers each produce some mix of goods, some overlapping with others. Each consumer would have reference levels for some subset of these goods, with the reference levels for any one good, and loop gains as well, being randomly distributed around a mean.

I recommend, however, that we start as simply as possible and expand the model as we gain understanding.

Would a PCT model inferred in that way, really have a chance of being a good PCT model. Perhaps I am confusing PCT with MOL. If we don't have access to the internal MOL models of the actors, and must infer them from the historical data, how will that differ from models of actor behavior derived under the subjective values assumption?

I don't pretend to know. I wouldn't begin with the historical data -- that's for testing the model when it's finished enough to test. The first task is to construct a model that will actually run, and learn what its properties are. We can try to preserve realism while doing this, but only provisionally. We don't have to be right the first time. The point initially is just to get a working model with no loose ends unaccounted for. Once we have that, we can start comparing its behavior with whatever data or general knowledge we have. That will tell us what we got wrong and probably how to fix it.

If actors are to be modeled individually, I don't see how a PCT based model is likely to come to a different conclusion than a subjective values based model about the key issue of the day, free markets versus centralized socialist planning. In both cases markets will be better at distributing the decision making to the individual, and incorporating the individuals feedback.

That's a prediction from your model, which I assume you don't actually have running by itself yet. There's no reason why you couldn't put the individual and producer characteristics that you believe in (or deprecate) into the model and see what the model does. I have never taken my modeling efforts to the level of predicting market processes based on policies or political beliefs, but see no serious obstacles to doing that. I think we have a lot of work to do before getting to that level of analysis, however.

The last time I used the term "we" in this economic context, I got back the Tonto Judgment: "What you mean, 'we', paleface?" All I can do is keep trying.

If this modeling project is to work, we all have to give up whatever beliefs we have about economics and pretend to be ignorant (that's easy for me). The whole point is to find out what is true, not to defend or attack existing beliefs. If we think some relationship between people or characteristic of individuals might exist, the strategy is not to think up all the reasons why we are right, but simply to model the idea and see if the model behaves as we expect when it runs. If it doesn't, we're wrong about something, and even better, we're about to learn something new. If we find that we can't convert the idea into a model, that's also very valuable information: it says we don't have a complete idea yet and need to work on it some more.

Best,

Bill P.

18

[From Bill Powers (2009.09.03.1651 MDT)]

Martin Lewitt; Thursday, September 3, 2009 1:39:01 PM GMT -07:00 --

[ I know that standardized headers are an imposition on autonomy, but it's nice to be able to edit a few characters and show everybody whom you're replying to and which post it is -- you sent quite a few today. It's a bit awkward to use your format, as I have to cut and paste it into the text and type in your name]

BP earlier: You may recall that the late Bill Williams and I worked on a control-system model for the Giffen Paradox in which there were two goods and a budget. ...

ML: There is some question of whether a true Giffen good has ever existed, and although theorectically it could. Yes, the law of supply and demand can't explain it, but the income effect, along with two are three other conditions can explain it. (necessity, inferiority, lack of substitutes). in a way that makes sense, and apparently can be modeled.

It's much easier than that. All you need is a budget limit and two goods which are alternate ways of obtaining some necessity. Bill W. used bread and meat as goods which provide calories to keep a person alive. The meat was preferred, being tastier, but much more expensive per calorie than the bread -- so expensive that it was impossible to obtain all the necessary calories from it without running out of money. So part of the calories are obtained from meat and the rest from bread, in proportions that stay within budget but provide as much meat as possible.

Now if the price of bread goes up, the former quantities of meat and bread would cost too much. Reducing the meat component will free up enough money to buy much more bread and thus maintain the calorie intake at the required level. When you think in terms of controlling inputs and consequences, the Giffen Paradox becomes simply the Giffen Effect. There's nothing paradoxical about it.

However, even if it couldn't be modeled, correct modeling of it would seem to be unimportant advantage of PCT. Does the PCT explanation of the phenomenon offer a different insight? The PCT alternative does not seem to offer simplicity or insight, but then neither does quantum mechanics, but like quantum mechanics, it might have the advantage of allowing better predictions. Where would those advantages be expected? What data would need to be available to realize those advantages?

Well, judge for yourself. The PCT model suggests the conditions under which any good can exhibit the Giffen Effect; there is no such thing as some special "Giffen good" that causes the reversal of the effect of price changes. Any good can be a Giffen Good. Seem to me that this is a new finding.

Bill W. also used as an example a member of the jet set who preferred the prestige of travel by private jet, but whose budget was a little too small to do all traveling that way -- the rest had to be done by commerial carrier. When the price of commercial flights is raised, the jet-setter has to shift some of his flights from private to commercial, so the use of commercial flights increases. So this isn't just a matter of poverty. Relative poverty, perhaps.

When many goods are considered instead of only two, Giffen Effects will probably abound. Prosperity, of course, removes the budget constraint at least for a while, but the poor are always with us, and they are the ones most likely to be victims of this phenomenon. When WalMart raises its prices, many people have to increase their patronage of WalMart and buy less at Marshall Fields. When China raises its labor prices, some manufacturers will be forced to increase the amount of outsourcing to China if no one else underbids them. The law of supply and demand is an oversimplification of the actual relationships.

Best,

Bill P.

19

[From Rick Marken (2009.09.03.1830)]

Martin Taylor (2009.09.03.13.58) –

My answer to this post is: Bill Powers (2009.09.03.1118 MDT). I think you should just continue this discussion with Bill. I’ll just watch. I think my participation would (as usual) just end up being confusing. Bill is presenting my point of view, just (as usual) more lucidly and comprehensively.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

20

[From Rick Marken (2009.09.03.1945)]

Bill Powers (2009.09.03.1651 MDT)–

When many goods are considered instead of only two, Giffen Effects will probably abound. Prosperity, of course, removes the budget constraint at least for a while, but the poor are always with us, and they are the ones most likely to be victims of this phenomenon. When WalMart raises its prices, many people have to increase their patronage of WalMart and buy less at Marshall Fields. When China raises its labor prices, some manufacturers will be forced to increase the amount of outsourcing to China if no one else underbids them. The law of supply and demand is an oversimplification of the actual relationships.

Dare I mention my demonstration of the Giffen effect at

http://www.mindreadings.com/ControlDemo/Economics.html

In this demo, the participant is required to control caloric intake by controlling the amount of bread or meat purchased on a budget. When the budget is large, the normal price demand relationship is observed for the expensive good. When the budget is small, Giffen effect. THe demo uses a human rather than a model for the control system.

I also ran across a study that was done some time ago (I don’t seem to have saved it) showing a real world macro level Giffen effect in Asia where consumption of a less preferred but also less expensive staple (sorghum?) goes up when it’s price goes up.

Best

Rick

···

–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com