REPLY Characterizing System Dynamics (SD5638)

[Justin Lyon]

J.J.

I tend to think of structure in a system dynamics
simulation as a hierarchy of objects that provide
insight into the interrelationships of the various
objects.

System Dynamics 'Structure' is a hierarchy of objects
assembled in the following process:

A. You define the boundary of the system.
     1. You identify the feedback loops.
          a. You identify the stocks
          b. You identify the flows
                   1. You articulate the policies as
flow equations

Note that two variables, stocks and flows, capture ALL
the aspects of the system under study. And, the flows
capture all the information on the policies of the
system under study.

These processes are the building blocks of
simulations.

Everything in the universe is a process. EVERYTHING.

Simulation Science is commercially very valuable
stuff. Think billions of dollars not millions.

In business, we create objects that are assembled into
representations of workflow processes with mapping to
data warehouses for real-time information access.

In nature, like in a biochemical process, we can
create a simulation model that captures with some
degree of fidelity the processes that we wish to
better understand (like production of an amino acid in
bacteria).

Stocks, such as a stock of customers, are
accumulations of many autonomous humans (agents). The
inflow of new customers captures all the information
on the policies (such as marketing) that drive the
accumulation of customers. The outflow of customers
captures all the information on the policies (such as
poor customer service) that drive the depletion of
customers.

Steve Guerin at Redfish refers to these as 'aggregated
super-agents.' Dr. Suarez at Trinity University refers
to them as 'upper-level agents.' Others refer to them
as 'stocks.' Others refer to them as 'levels.' I call
them 'bathtubs.'

Behavior over time EMERGES from structure. I use
EMERGENCE as a concept because all too often the
behavior of a system is counter-intuitive to what we
actually observe; it arises from nothing, yet is
clearly visible. Some explain, where possible,
behavior over time as resulting from the interaction
of multiple feedback loops and the concomitant
switching loop dominance.

That is, I think that the ABM folks (at Santa Fe,
Trinity and other spots around the world) and the SD
folks (at MIT, LBS, WPI, LSE and other spots around
the world) have their own particular jargon which, in
my not so humble opinion, all means something similar
? life is not reducible to a simple set of
deterministic and solvable equations.

That is, I argue that the concept of 'emergence' is
related or perhaps even the same as the SD concepts of
'behavior over time' as a result of structure?

Of the interesting systems for study in the world,
99.99% are not solvable by humans without the aid of
computers. And, many of those 99.99% of non-linear
systems are not 'solvable' with mathematics known to
humans.

This should really irritate a lot of people because
everyone keeps trying to find some Newtonian-like
super-equation for biology that will 'answer' all our
questions, but no one will ever find it. It is a
will-o-wisp.  

We can only solve the problems of the business,
nations and the world by the close and intimate
interaction between humans and computers. This
interaction must be governed by an appreciation of
reality as understood through the lens of Simulation
Science, that is, complexity science, chaos theory,
nonlinear science ? choose the jargon that you like.

That is, this must always be kept in mind, ALL models
are false, some are useful and some are dangerous.

There are NO point predictions. No ULTIMATE answers.
No TRUTH.

There is only foresight into plausible and probable
future scenarios that may or may not happen.

More and more I'm thinking that system dynamics,
aggregated super agents, upper level agents, etc. ad
nausea, are the 'physics of biology;' the
'meta-engineering' of life through the wonderful, yet
blind and simple algorithmic process of evolution by
natural selection of agents that replicate.

Why is this important? Well, Simulation Science is
incredibly well-suited to studying human-designed
systems like global businesses, surfacing mental
models and enabling humans to re-design and manipulate
these systems to achieve goals -- like making more
money.

Before the 21st century, physicists and mathematicians
were the 'gods' of science. Think of Newton, Leibniz,
Kelvin, Fermi, Einstein. The 21st century 'gods' of
'science' are the biologists, computer scientists,
artists and engineers ? people who are well-versed in
the power and manipulation of emergence (behavior over
time) in everyday life. Think of Darwin, Andronov,
Forrester, Kaufman

-Justin



--- "Jean-Jacques Laubl? jean-jacques.lauble          
   wanadoo.fr" <system-dynamics at VENSIM.COM> wrote:

> Posted by  =?iso-8859-1?Q?Jean-Jacques_Laubl=E9?=
> <jean-jacques.lauble at wanadoo.fr>
> Hi everybody
>
> Jim Hines writes
> < the only defining characteristic of SD work is an
> interest in how
> < structure creates behavior.
>
> What does creates mean?
> Does it mean
>   1.. Has an influence
>   2.. Has a determinant influence
>   3.. is the only influence on the behaviour
>
> What does structure mean?
> A simplistic example
>
> A business sells widgets and the number of widgets
> sold in a period of time
> is the demand for widgets constrained by the
> production capacity.
>
> What is the structure in this case?
> Is it only the fact that the widgets sold depend on
> the demand and the
> production capacity?
> Is the level of the demand of the production
> capacity included in the
> structure?
>
> If I am interested in the behaviour of the turnover
> of the business, knowing
> that it depends on the demand and on production
> capacity is not enough,
> especially if the future demand is unknown.
>
> The difficulties of all my models were not to find
> the causal structure of
> the model but the values of the parameters and the
> behaviour of the
> exogenous data needed by the model.
> And depending on these values, the behaviour of the
> results could change
> drastically, whatever the structure.
>
> Regards.
> J.J. Laubl? Allocar
> Strasbourg France.
> Posted by  =?iso-8859-1?Q?Jean-Jacques_Laubl=E9?=
> <jean-jacques.lauble at wanadoo.fr>
> posting date  Sat, 26 Nov 2005 13:45:26 +0100
>