What is System Dynamics: Introduction | Systems Thinking | Dynamic Modeling | Modeling Tools

What is System Dynamics:

The field of System Dynamics (SD) was founded by Dr. Jay Forrester at MIT in the 1950’s. CIESD’s specific approach for applying SD to real world problems is expressed in our “Ladder of Engagement” and our use of a set of specific tools, as we climb that ladder. However, a basic introductory description of SD is in order.

Historically based on feedback systems in mechanical and electrical engineering systems, SD builds on several realities that are generic to all dynamic systems. These can, in broad terms, be divided into three broad and interrelated categories: Systems Thinking (ST), Dynamic Modeling (DM) and Modeling Tools:

What is System Dynamics: Introduction | Systems Thinking | Dynamic Modeling | Modeling Tools

Systems Thinking:

Systems Thinking is the mind-set that allows (and ultimately obliges) us to examine the world by consciously recognizing and formally exploring the systems of which it is made.

1) Systems are fundamentally dynamic in time. A static “snapshot” of a system reveals the size of the Stocks at that instant, but discerning the role of the Flows, or of the feedback loops that control them, requires a period of time during which they will exhibit their influence. It is this evolution of the system in time that is its primary characteristic, not its state at any instant.

2) The behavior of a system is ultimately controlled by its structure, that is by the combination of Stocks and Flows of which it is composed and by the positive and negative feedback loops that control the Flows.

3) Systems are defined by their Stocks and Flows (or by their Accumulations and Rates). Your personal financial system, for example, is defined by a variety of Stocks of money (what's in your wallet, in your checking account, in the National Treasury, etc.) and the Flows that transport money between those Stocks (interest payments into your savings account, tax payments to the IRS, your paycheck moving money from your employer to you, etc.).

4) Controls of systems are circular feedback relationships, not linear chains of cause and effect. Looking at systems means abandoning those simplistic, linear, one-way chains of cause and effect that lead to a final answer, such as:

A ===> B ===> C ===> D

ST teaches that the most interesting causal relationships are, in fact, circular, in that A influences B which influences C which in turn influences the original A again which ...

5) These causal loops (or feedback loops) come in two basic “flavors.”

• Positive feedback loops (perhaps better called “reinforcing” loops) are those where, if you “push” factor A in one direction, it leads to a cycle (causal loop) of subsequent activity that eventually results in A being pushed further in the original direction. A common example is a biological population: A population of organisms grows as new individuals are born into it; with more individuals now in the population, even more are born into it; which further increases the size of the population; which increases the flow of new individuals; which increases ...
• Negative (or “stabilizing”) feedback loops are quite different. Push A in one direction and you set in motion a sequence of events that ultimately leads to A being pulled back toward the position where it started. A simple example: As you fast after your last meal or snack, your hunger increases. This eventually sets in motion events that result in your eating, which reduces your hunger. Hunger began by increasing, but the course of events ended by your hunger being reduced back to its original low level.

What is System Dynamics: Introduction | Systems Thinking | Dynamic Modeling | Modeling Tools

Dynamic Modeling:

Dynamic Modeling provides a means to rationally, objectively, and precisely wrap our minds around systems. Even “simple” systems with relatively few seemingly straightforward components can present complexities, as they evolve in time, that are beyond our brains' abilities to decipher. Modeling -- and here we mean not just computer simulation modeling but a variety of very powerful and useful mental and graphical modeling approaches -- provides a means to define the structure and behavior of a given system and to project that structure and behavior into the future. Such modeling can serve a number of purposes. Your focus with a given type of model depends on the specific scenario and your particular needs:

1) Models provide a common and objective language for diverse individuals to discuss a particular issue. “Raising taxes is stupid!” may be a genuine and heart-felt belief, but it is very hard to come to grips with a statement like that. Oblige that discussion to be framed within the context of a stock/flow diagram, causal loop diagram, or computer model, however, and you are quickly forced to recognize that taxes represent a Flow of money from one Stock to another Stock, and that as those Stocks rise and fall as a result, there are implications on other components of the system. Whether those implications represent improvements or worsenings in your situation or that of your town, state, or nation now can be more objectively discussed.

2) Construction of a model obliges you to be very clear and precise in defining which component influences which and in what manner that influence is wielded. Does an increase in government revenue alter the way in which government spends that money? Is it likely to spend its “windfall” foolishly or inefficiently, or will it spend it in a manner that will increase the common good? You probably have a “mental model” of that; building and communicating that model forces you to explicitly and clearly define that mental model.

3) Construction and communication of a model provides a powerful framework for collaboration between diverse stakeholders in the system in question. Can we identify areas of agreement? Can we clarify and better understand areas of disagreement? Can we resolve the latter?

4) Construction of a model often reveals areas of ignorance where new information must be gathered or discovered through original research. From a previous point: If there is a disagreement on the likely effect of new revenues on government spending, can or should we explore what has historically been the effect of such changes in other similar instances?

5) Precise computer models allow you to play “what if?” games and explore the implications of specific courses of action,i.e. "policies."  Does a tiny change in one component lead to massive changes in the system’s behavior? Does a huge change in another component have negligible effect? This “sensitivity analysis” provides a powerful means to determine where you might get the “most bang for the buck” by identifying the high leverage points where a small manipulation leads to large improvements.

Similarly, a model allows you to test the result of a policy decision before implementing it. By compressing time, you might simulate 100 years of environmental dynamics in seconds. Better to have the computer suggest the danger of continued ozone release than to perform the global experiment and suffer the consequences for several generations!

What is System Dynamics: Introduction | Systems Thinking | Dynamic Modeling | Modeling Tools

Modeling Tools:

As we move up the “Ladder of Engagement” that integrates the Systems Thinking and Dynamic Modeling aspects of our work, we typically utilize a set of tools to support that development:

1) Knowledge: Knowing (or perhaps agreeing on) the behavior of the system is the first, and often quite challenging, step in our analytical process. Here a particularly useful visual modeling tool is construction of a “Behavior Over Time Graph” (or "Reference Mode") that plots the dynamic changes of the system over time. As well, what is the likely, or desired, or feared behavior of the system as we project it into the future?  Hand in hand with use of Behavior Over Time Graphs, on this rung of the Ladder we frequently begin to develop a basic definition of the dynamics that lead to such behavior.  Can we develop a simple “Stock-Flow Map” to define the stock of interest and identify the flows that contribute to or drain from that stock?

2) Understanding: Given a dynamic behavior, what in the system controls it? A basic tenet of SD is that feedback loops are fundamentally important in understanding why a system does what it does. Identifying and understanding those feedbacks is necessary before effective intervention in the system will be possible. Two tools useful for that purpose are “Causal-Loop Diagrams” and “Stock-Flow Maps.”

3) Influence: Here, the objectivity of “Computer Simulation Models” becomes critical. Being able to intuit the interaction of multiple, often non-linear feedback relationships is simply beyond our unaided mental ability. Are our mental models complete? Accurate? Do they result in unanticipated consequences? Undesired consequences? These are the pitfalls that computer modeling can help to avoid.

Refer to:

http://www.clexchange.org/ftp/documents/Implementation/IM2003-12TipsUsingSDTools.pdf for additional suggestions on using Behavior Over Time Graphs, Causal-Loop Diagrams, and Stock-Flow Maps.