Nowadays many softwares such as STELLA, Powersim, and Vensim etc. enable us easily to build a diagram, but we have to convert our thought in a natural language into a System Dynamics diagram. During this process, most parts of model-building processes rely on art even today.
Generic models (Richmond) and Archetypes (Kim, Senge) give us
very useful building blocks for building a model as well as learning
an event. However it is important to notice that there is a continuum
of a human image about an event, its expression in a natural language,
its causal loop, its diagram, and its equation model. If a model-building
process is considered as translating each stage into the next
stage of the continuum, it would be possible even for a beginner
to finish from one end to the other quickly. Toward this aim,
this paper will show a simple prototype of modeling as the continuum
of translating each stage, using some simple examples. As a result,
causal loop (Coyle) will also be examined for its extension.
A Continuum of Modeling Process
Modeling process starts from an image about an event . This image is associated in words using a natural language at Step 1 in Figure 1. The words are converted into causal loops using a causal loop at Step 2. These causal loops are converted into a diagram using graphic symbols at Step 3. Finally the diagram is converted into an equation model for computing at Step 4. These processes are recognized as a continuum of translating the former expression into the latter one sequentially.
1) As an example of positive feedback, let's think about a thing which grows by itself. At Step 1 this image can be converted as the Compounding Process in a natural language: "When a thing increases, this growth leads to further developmental changes." This is translated into causal loop I of Figure 2.1. This loop is refined into causal loop II by adding Action to Thing. This new loop II is recognized as causal loop III using Stock and Flow for the System Dynamics. In the case of a bank balance, Stock is a bank balance and Flow is an interest income in causal loop IVa. An interest rate is not a part of a loop but just a supplementary variable to an interest income (IVb of Figure 2.1). Then, causal loop IVc of an bank balance growing is easily converted into a diagram D of Figure 2.1.
2) As an example of negative feedback, let's think about a thing kept constant. At Step 1 this image can be converted as the goal seeking process or the Stock-Adjustment Process. In everyday speech this can be expressed as: "Setting a goal." or "Setting a goal for a thing." This is translated into causal loop I of Figure 2.2. Further this loop is refined into causal loop II by adding Difference, Goal and Action to Thing. This new loop II is recognized as causal loop III using Stock and Flow for the System Dynamics. In the case of thermostatic control, Stock is a temperature and Flow is an adjustment in causal loop IV. Then, causal loop IV of keeping a temperature at a goal is easily converted into a diagram D of Figure 2.2, in which Adjustment is translated as Bi-Flow.
3) As an example of positive & negative feedback, let's think about a thing that grows and declines at will. At Step 1 this image can be converted as the limit to growth in a natural language: "As a thing grows there is a natural limit in which it stops developing." This is translated into causal loop I of Figure 2.3. This loop is refined into causal loop II by adding Action to Thing in positive feedback, and by adding Difference, Goal and Action to Thing in positive feedback. This new loop II is recognized as causal loop III using Stock and Flow for the System Dynamics. In the case of some condition, Stock is Condition in causal loop IV while Flow in positive feedback is Growing Action and Flow in negative loop is Slowing Action triggered by Difference from Limiting Condition. Then, causal loop IV of this condition controlled by the two feedback loops is easily converted into a diagram D of Figure 2.3 using the above two kinds of prototypes.
Three simple prototypes show that each causal loop has to be refined until the causal loop is clearly written using Stock and Flow. During this process an image about an event and an expression about it become explicit. So this process of Figure 1 is revised to a continuum of modeling process shown in Figure 3.
Through feedback at Step 2 in Figure 3 many kinds of variables
will be added to causal loop. Characteristics of this kind of
developing complex causal loop is necessary to be studied further
for the aim of automating modeling processes.
ISDC '97 CD Sponsor