1. It is important to examine the loops and links in your model to understand the following effects.

Which are information links? Information links convey information, such as the a decision, but not material.

Which are flow links? Flow links convey tangible items that can turn the node into a stock, or a container of value. Examine any flow links for these aspects, which may affect the resulting behavior over time of the loop.

• Rate-to-level. Look for links that increase a stock when positive but do not decrease it when negative (or vice versa).
• Hidden loops. Look for omitted loops that may be important to the behavior under study.
• Net rates. Examine links to determine if they create a one type of loop (for example, reinforcing) when positive and the other type of loop (for example, balancing) when negative.

2. Use “+” and “-” to label links instead of “s” and “o.”

Recent conventions have often advocated using “s” for “supporting” links, which should cause Variable B to move in the the same direction as Variable A, and “o” for “opposing” links where Variable B should move in the opposite direction of Variable A. However, George Richardson, in Problems of Causal Loops Revisited, argues that this convention can be misleading. Professor Richardson notes that, in some diagrams, if Variable A switches directions, Variable B does not. He thus advocates using “+” and “-”, explaining:

When polarities are labeled with positive and negative signs, links can be defined as either additive or proportional influences, and the notation works well semantically in either case. One can say that a positive arrow from A to B means that A adds to B, or, a change in A causes a change in B in the same direction (resulting in a positive correlation or direct variation). For a negative link from A to B one says A subtracts from B, or, a change in A causes a change in B in the opposite direction.

3. Assign the type of loop based on the count of negative links, not the dynamic behavior over time.

Often, loops are defined on the basis of their dynamic behavior. That is, reinforcing loops are defined as the key variable as having an exponentially increasing or decreasing curve, while balancing loops are defined as oscillating around a goal line. However, this can be problematic because in some loops the key variable could exhibit a reinforcing behavior or balancing behavior, depending on the values of variables in the loop. Consequently, it is recommended to classify loops as balancing when they have an odd number of negative links and reinforcing when they have an even number of negative links.

A recent Ahead of the Tape column concerned how a “negative feedback loop” describes what it means when things keep going from bad to worse. But in response to the column, several readers argued that it is technically a “positive feedback loop.”

Here’s why. Negative feedback loops are broadly referred to in engineering and biology as an input into a system that tends to offset the system’s general direction, leading to a state of equilibrium. Positive feedbacks, on the other hand, reinforce the general direction.

One could easily argue that economists should stick with these definitions when talking about the economy. By that logic, interconnected trouble in the economy — say, when weak housing hurts consumer spending, leading to more weakness in housing and spilling into more corners of the economy — is actually a positive feedback loop.

Or not. Economic downturns, and the processes that propagate downturns, are commonly called “negative.” Calling this dynamic “positive” would be confusing for readers unaquainted with the technical language — a view shared by Yale economist John Geanakoplos, who was quoted in the column.

I would argue it is better overall to just skip that discussion altogether and use the terms “reinforcing loop” and “stabilizing loop.”

This will also include information from previous Global Open Space conferences.

Doug Walton, PhD

President

International Systems Institute