Articulating Reality

Causal loop diagrams provide a language for articulating our understanding of the dynamic, interconnected nature of our world. We can think of them as sentences which are constructed by linking together key variables and indicating the causal relationships between them. By stringing together several loops, we can create a coherent story about a particular problem or issue.

The next page includes some suggestions on the mechanics of creating causal loop diagrams. Below are some more general guidelines that should help lead you through the process:

• Theme Selection. Creating causal loop diagrams is not an end unto itself, but part of a process of articulating and communicating deeper insights about complex issues. It is pointless to begin creating a causal loop diagram without having selected a theme or issue that you wish to understand better. “To understand the implications of changing from a technology-driven to a marketing-oriented strategy,” for example, is a better theme than “To better understand our strategic planning process.”
• Time Horizon. It is also helpful to determine an appropriate time horizon for the issue — one long enough to see the dynamics play out. For a change in corporate strategy the time horizon may span several years, while a change in advertising campaigns may be on the order of months.

Time itself should not be included as a causal agent, however. After a heavy rainfall a river level steadily rises overtime, but we would not attribute it to the passage of time. You need to identify what is actual driving the change. In computer chips, $/MIPS million instructions per second) have been decreasing in a straight line over the past decade. It would be incorrect, however, to draw a causal connection between time and $/MIPS. Instead, increasing investments and learning curve effects are likely causal forces.

• Behavior Over Time Charts. Identifying and drawing out the behavior over time of key variables is an important first step toward articulating the current understanding of the system. Drawing out future behavior means taking a risk — the risk of being wrong. The fact is, any projection of the future will be wrong, but by making it explicit, we can test our assumptions and uncover inconsistencies that may otherwise never get surfaced. For example, drawing projections of steady productivity growth while training dollars are shrinking raises the question “If training is not driving our growth, what will?” The behavior over time diagram also points out key variables that should be included in the diagram, such as Training Budget and Productivity. Your diagram should try to capture the structure that will produce the projected behavior.
• Boundary Issue. How do you know when to stop adding to your diagram? If you don’t stay focused on the issue, you may quickly find yourself overwhelmed by the number of connections possible. Remember, you are not trying to draw out the whole system – only what is critical to the theme being addressed. When in doubt about including something, ask “If I were to double or halve this variable, would it have a significant effect on the issue I am mapping?” If not, it probably can be omitted.
• Level of Aggregation. How detailed should the diagram be? Again, the level should be determined by the issue itself. The time horizon also can help determine how detailed the variables need to be. If the time horizon is on the order of weeks (fluctuations on the production line), variables that change slowly over a period of many years may be assumed to be constant(such as building new factories). As a rule of thumb, the variables should not describe specific events (a broken pump); they should represent patterns of behavior (pump breakdowns throughout the plant).
• Significant Delays. Make sure to identify which (if any) links have significant delays relative to the rest of the diagram. Delays are important because they are often the source of imbalances that accumulate in the system. It may help to visualize pressures building up in the system by viewing the delay connection as a relief valve that either opens slowly as pressure builds or opens abruptly when the pressure hits a critical value. An example of this might be a delay between long work hours and burnout: after sustained periods of working 60+ hours per week, a sudden collapse might occur in the form of burnout.
  

  TOOL BOX: GUIDELINES FOR DRAWING CAUSAL LOOP DIAGRAMS

  

The post Guidelines for Drawing Causal Loop Diagrams appeared first on The Systems Thinker.

I strongly believe that stock and flow diagrams offer a deeper understanding of a system than do causal loop diagrams. Nevertheless, in the past, I found it difficult to get more than a small handful of clients to develop the facility to build them. Despite its obvious benefits, the rigor of the stock and flow language comes at a price—it is more difficult to learn. While companies that create simulation software have made enormous advances in their products, vastly simplifying the model-building process, we still have a long way to go in learning how to help people develop the facility to create even simple models.

High Performance Systems, the creators of the ithink® software, have stuck firmly to their belief that a true understanding of the dynamics of any system requires an appreciation of the underlying stock and flow structure. For that reason, their software does not provide the facility to build CLDs. The best they offer are “Loop Pads,” which they describe as “. . . simple pictures that identify the cause and effect processes that work to generate dynamic behavior patterns.” To display these pictures, you have to build the stock and flow model first.

The challenge is to find more effective ways of helping clients develop an understanding of the structural dynamics of the system they are studying, while acknowledging that they usually find CLDs an easier place to start than stock and flow diagrams. To that end, I have developed an approach to model building that uses ithink in a slightly unorthodox way to start clients at a relatively easy place and move them quickly to a more sophisticated understanding of a given system using stocks and flows. Paradoxically, this technique capitalizes on the software’s unwillingness to let users draw CLDs.

From Feedback Loops . . .

To follow this process, you must use version 6.0 of the ithink software, which allows you to minimize the size of the converter icon. Start by changing the defaults to set the converters to small. Doing so lets you use the converters as you would the variables in a CLD. Then use the text box facility, which is one of the objects on the menu bar, to create the polarity signs (, “+” or “-,” which correspond to “s” and “o”). For example:

So far so good. Up to this point, we are simply building a CLD in which “resources allocated to process improvement” are influenced by current “performance” and “desired performance.” However, when we try to create a causal link between “resources allocated to process improvement” and “process errors,” an error message appears indicating that such an action would create a circular connection.

The nature of the mathematics that underlies the stock and flow language means that the software is unable to calculate the value of any converter or flow when they loop back on themselves. As the help files state: “In drawing connector linkages, you may encounter an alert which tells you that circular connections are not allowed. Mechanically, this alert means that you have attempted to create a chain of converters or flows, such that one converter or flow ultimately depends upon itself. The software cannot resolve the resultant simultaneous equations.”

. . . to Stocks and Flows

To get past this barrier, we must create at least one stock somewhere in the loop. This process forces us to look more closely at the structure of the loop we are creating and identify one or more stocks. Every feedback loop has an accumulation—it is this accumulation that generates the feedback dynamics. In this example, the issue for the team was how actual performance levels drove “resource allocation to process improvement.” With this in mind, we can now make a simple modification to the CLD by converting the variable “performance” into a stock.

We are now a step closer to gaining a deeper understanding of the feedback processes involved in this structure. We have done so, however, by beginning with a process that clients are familiar and comfortable with and then moving to a structural understanding through one simple step. How we develop the model from this point forward depends on your goals. We could stay with this loop and simply develop the stock and flow structure for each variable.

Going into Greater Detail

We also might want to explore a certain part of the structure in greater detail. For example, we might be interested in the dynamics involved in a process-improvement program. In this case, the team realized that the resource allocation decisions were not only determined by actual performance but by the gap between actual and desired performance:

On the other hand, we may want to develop a loop to explore the impact of process quality. One possibility could be:

Once again, when we try to close the loop by connecting “investing in process quality” with “process quality,” we will receive an error message that circular connections are not allowed. How we respond to this message depends on what we are trying to understand with the model. If we want to examine the financial implications in more detail, we could begin to unravel the structure underlying the variable called “profits.” For example:

The key point is that we always anchor the model development process in something the client is familiar and comfortable with—the development of CLDs. We then force the software to highlight a logical error to provide a stepping stone to unfolding the stock and flow structure. I have found that, using this technique, more clients are able to develop an ability to create their own stock and flow models than before. Prior to using this approach, I found that clients viewed CLDs and stock and flow diagrams as separate and distinct languages. Since I’ve implemented this process, I have noticed that they have begun to see the similarities, rather than the differences, between the two. As a result people are less mystified when working with stocks and flows, seeing them embedded in the feedback loops of CLDs.

David Rees is the director of High Performance Learning Systems, a consultancy firm specializing in applying systems thinking principles and tools in public and private sector organizations. He is also a research fellow at the Centre for the Design of Innovative Systems at UNITEC in Auckland, New Zealand.

The post Anchoring Model Development in Causal Loop Diagrams appeared first on The Systems Thinker.

Causal loop diagrams provide a language for articulating our understanding of the dynamic, interconnected nature of our world.

Articulating Reality

Causal loop diagrams provide a language for articulating our understanding of the dynamic, interconnected nature of our world. We can think of them as sentences that are constructed by linking together key variables and indicating the causal relationships between them. By stringing together several loops, we can create a coherent story about a particular problem or issue.

Following are some more general guidelines that should help lead you through the process:

• Theme selection. Creating causal loop diagrams is not an end unto itself, but part of a process of articulating and communicating deeper insights about complex issues. It is pointless to begin creating a causal loop diagram without having selected a theme or issue that you wish to understand better. “To understand the implications of changing from a technology-driven to a marketing-oriented strategy,” for example, is a better theme than “To better understand our strategic planning process.”
• Time horizon. It is also helpful to determine an appropriate time horizon for the issue—one long enough to see the dynamics play out. For a change in corporate strategy, the time horizon may span several years, while a change in advertising campaigns may be on the order of months.

  Time itself should not be included as a causal agent, however. After a heavy rainfall, a river level steadily rises over time, but we would not attribute it to the passage of time. You need to identify what is actually driving the change. In computer chips, $/MIPS (million instructions per second) decreased in a straight line in the 1990s. It would be incorrect, however, to draw a causal connection between time and $/MIPS. Instead, increasing investments and learning curve effects were likely causal forces.

• Behavior over time charts. Identifying and drawing out the behavior over time of key variables is an important first step toward articulating the current understanding of the system. Drawing out future behavior means taking a risk—the risk of being wrong. The fact is, any projection of the future will be wrong, but by making it explicit, we can test our assumptions and uncover inconsistencies that may otherwise never get surfaced. For example, drawing projections of steady productivity growth while training dollars are shrinking raises the question, “If training is not driving our growth, what will?” The behavior over time diagram also points out key variables that should be included, such as Training Budget and Productivity. Your diagram should try to capture the structure that will produce the projected behavior.
• Boundary issue. How do you know when to stop adding to your diagram? If you don’t stay focused on the issue, you may quickly find yourself overwhelmed by the number of connections possible. Remember, you are not trying to draw out the whole system—only what is critical to the theme being addressed. When in doubt, ask, “If I were to double or halve this variable, would it have a significant effect on the issue I am mapping?” If not, it probably can be omitted.
• Level of aggregation. How detailed should the diagram be? Again, the level should be determined by the issue itself. The time horizon also can help determine how detailed the variables need to be. If the time horizon is on the order of weeks (fluctuations on the production line), variables that change slowly over a period of many years may be assumed to be constant (such as building new factories). As a rule of thumb, the variables should not describe specific events (a broken pump); they should represent patterns of behavior (pump breakdowns throughout the plant).
• Significant delays. Make sure to identify which (if any) links have significant delays relative to the rest of the diagram. Delays are important because they are often the source of imbalances that accumulate in the system. It may help to visualize pressures building up in the system by viewing the delay connection as a relief valve that either opens slowly as pressure builds or opens abruptly when the pressure hits a critical value. An example of this might be a delay between long work hours and burnout: After sustained periods of working 60+ hours per week, a sudden collapse might occur in the form of burnout.

See detailed guidelines for drawing causal loop diagrams.

The post Guidelines for Drawing Causal Loop Diagrams appeared first on The Systems Thinker.