[From Rick Marken (2009.09.01.0930)]
I’ve retitled this thread in an effort to organize this discussion. Let’s see how the cat herding goes.
I’ve been resisting the temptation to reply to this and the “Got Data?”
thread, because like Shannon, I simply don’t have the time right now.
But suffice it to say that Bill’s idea of creating a model of the
economy populated by PCT-based agents is exactly what I hope to do for
my dissertation. How successful I will be remains an open question. But
Martin’s post below raises an important issue that is also narrow enough
that I thought I could prepare a reply that was both quick and cogent.
Wonderful. The problem I see with the agent approach to modeling the economy is deciding on what I would call the “resolution” of the model. Do you model all 300,000,000 agents in the US economy (if that’s the economy you plan to model) with all their roles activities. Do you model every escrow transaction, every purchase and sale. Or do you just model at the macroeconomic level (as I did), dealing with variables like GDP, inflation, productivity. There is always a choice to be made becuase you can’t really model things in complete detail, but then you might miss something important if the resolution gets too coarse. Bill’s economic modeling has been oriented toward a fine level of detail; mine has been more coarse. I like my approach better, of course, but I think Bill’s is great too. I think they are just oriented toward different things. Bill’s gives a nice picture of how PCT agents interact in a market environment; mine gives a better picture of how the macro economy works, assuming that the market takes care of itself (in terms of determining the price of goods and labor). Anyway, I look forward to seeing what you come up with.
I was going to comment on some of your ideas but I think it would be better to save my comments until I see what you actually end up implementing as a PCT based model of the economy. I think you’ll learn a lot once you set yourself to putting ideas into programming code (assuming that you plan to implement the model as a computer program).
Best regards
Rick
···
On Tue, Sep 1, 2009 at 6:53 AM, Frank Lenk FLENK@marc.org wrote:
On the one hand, Martin is right - maximizing subjective value can be
thought of as simply another way of saying we have things that are
important to us, we have purposes that are known only to us and that we
are constantly trying to achieve. Then trying to achieve them in the
face of environmental disturbances and constraints is eerily close to
maximizing subjective value subject to a budget constraint, which is the
heart of microeconomics, as long as you conceive of reference levels as
being set at values that are somehow optimal for the organism.
On the other hand (and now you can tell for sure I’m an economist),
economics treats these critically important values we are trying to
maximize as not only subjective but unexplainable. The values are
assumed to be given from outside the economic system. As economists, we
are not allowed to ask where they came from or what makes them change.
PCT, it seems to me, opens the values door to analysis. If values
define what we think is important, then they are the dual to purposes
since purposes essentially define the same thing. PCT identifies
purposes with reference signals - the things we are trying to come close
to. Some values may then indeed be given and fixed as economics wants
to assume, such as those associated with the reference signals linked to
intrinsic errors. But others are learned, both from experience and from
others, and form some of our higher-level reference signals. Which
reference signals/values we learn, how we learn them and what causes us
to change them are at least legitimate avenues of study under a
PCT-based approach, whereas they are not typically the concern of
economists. And yet it is these reference signals economics is all about
maximizing our proximity to. To me, this is an untenable myopia. Others
may disagree.
In some sense, the economy itself is simply a set of behaviors that we
undertake to control the amount of food in our bellies. Will a model
composed of PCT-based agents yield different results than traditional
economic models? Perhaps not. It is a hypothesis worth testing. But my
gut tells me it a significant difference is likely.
In part this is because I believe (but expect that Bill may disagree)
that PCT creates agents that are inherently social, because we learn
from others at least some reference signals and even which perceptual
signals are important to pay attention to. Standard economic models
treat individuals as isolated - assuming as the foundation of homo
economicus a Robinson Crusoe-like “state of nature” where, to quote
Hobbes, life is “nasty, brutish and short.” But as we evolved from apes,
and lone apes don’t appear to survive very long, we are much more likely
to be inherently social animals. There is much research in behavioral
economics that says people are at least as motivated by reciprocity and
fairness as profit and “utility.” This is how Adam Smith modeled us, as
inherently Sympathetic to each other, in his Theory of Moral Sentiments,
a book he thought more important than Wealth of Nations as shown by the
fact that he worked on it both before and after WON’s publication in
- It is this Sympathy that is the Invisible Hand that guides
markets to social optima, not simply free market competition on its own.
By allowing agents who use others for references, (especially, it seems,
the wealthy or powerful - something also identified by Adam Smith in
Theory of Moral Sentiments as well as others such as Thorstein Veblen in
his Theory of the Leisure Class) PCT has the potential of the modeling
of agents who act more like real humans.
At least, that’s my hope.
Sorry for the length of the reply - it turned out not to be quick, and
probably not completely cogent either.
Frank
Frank Lenk
Director of Research Services
Mid-America Regional Council
600 Broadway, Suite 200
Kansas City, MO 64105
816.474.4240
816.701.8237
-----Original Message-----
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Martin Taylor
Sent: Monday, August 31, 2009 9:07 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: [CSGNET] Conservative Controlled Variables
[Martin Taylor 2009.08.31.22:00]
[From Bill Powers (2009./08.31.1734 MDT)]
At 05:45 PM 8/31/2009 +0000, Martin Lewitt wrote:
From: “Bill Powers” powers_w@FRONTIER.NET
Sent: Saturday, August 29, 2009 9:00:31 AM GMT -07:00 US/Canada
Mountain
Homo Economicus is different under PCT than under most economists’
theories. How do the human control systems involved show up in the
model? Should that not make a difference?
The original microeconomic models assumed that each economic decision
maker was acting in his own self interest to maximize his subjective
values. As they came to explicitly deal with the fact that humans
may not always be rational, or perform the maximization calculations
properly they found it didn’t make much difference in their models,
since they also didn’t have knowledge of their subjective values, but
could only infer them from the exchanges and other decisions they
made. What is it about PCT that would mean that the same
microeconomic models would not also be robust to whatever difference
PCT would make?
You tell me. Does PCT say that human beings maximize anything? Do
economic theories generally include the concept of “enough” of some
good?
I know this isn’t what you are getting at, but wouldn’t it be fair to
say that minimizing the error in a controlled variable, bringing its
quantity closer to its reference quantity, is maximizing its subjective
value? One could say that PCT is all about maximizing subjective value.
As for whether economic theories generally include the concept of
“enough” of some good, that’s a different kettle of fish, and one I
won’t attempt to fry.
Martin
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com