Feedback

Often our analysis can lead to the identification of feedback. Sometimes this happens quite quickly with only a few or no intermediate variables. For example, X affects Y affects Z affects X. Sometimes this happens very slowly with many intermediate variables and delays. For example, X affects Y affects Z affects A affects B affects C affects D affects E affects…affects X!

In the same way that a chain of relationships can be supporting or opposing so can a feedback loop of relationships. Feedback loops that are supporting are called Reinforcing loops. Feedback loops that are opposing are called Balancing loops. Reinforcing loops reinforce the direction you start with. Balancing loops change the direction you start with.

Reinforcing loops are sources of growth and decline. Their behaviour is exponential. This means that the size of the change doubles each repeat of the cycle. Let’s try it…

1, 1x2=2, 2x2=4, 4x2=8, 8x2=16, 16x2=32, 32x2=64, 64x2=128, 128x2=256, 256x2=512, 512x2=1024, 1024x2= 2048, 2048x2= 4096, 4096x2=8192, 16384, 32768, 65536, 131072, 262144, 524288, 1048576 ………

In this example, the initial changes are relatively small but then suddenly huge changes are manifest. If we don’t appreciate that a structure like this is at work we could either abandon an activity that is on the way to working or fail to recognise the slow development of a huge problem.

Figure 6 A Thinking Frog

I am told that if you place a live frog in a pan of boiling water he jumps straight out. If, however, you place a live frog in a pan of cold water and bring it slowly to the boil the frog cooks…He becomes the victim of an exponential behaviour pattern that he had not appreciated.

Imagine that I could fold a piece of A4 paper 40 times. It starts off very thin and with an A4 surface area. After one fold the surface area halves and the height doubles. I repeat this folding 40 times. How tall do you think that the paper tower would be?

Answer: The paper tower would reach to the moon as the growth in height is exponential. If you don’t believe me get a calculator and work it out for yourself!

Also, I find it incredibly interesting that on the 39th fold I would only be half way there.

In contrast, balancing loops constrain or control growth and decline. They are the source of the statement “nothing grows forever”.

Feedback is vital to all systems. Balancing loops, in particular, are a force for equilibrium; they prevent things getting out of control. Biological balancing loops are geared to survival. For example, when it gets too hot I will sweat which in turn cools me down. When we humans wish to control (or manage the performance of) something we design balancing loops. Usually these loops have a variable we want to influence. We set a target value for that variable and we monitor the gap between the actual value and the target. When the actual is below target we take action to improve the actual value. When we are on-target we reward the achievement in some way; this may mean that we simply cease acting!

In practice this is nowhere near as easy as it sounds.

We are juggling with many balancing loops in this way when we ride. Unfortunately we often don’t know what we need to work on; what our target should be; how much is too little; how much is too much; how much is just right; when to act.

Figure 7 Balancing Loop with Target

Things get even more complicated when there are delays between cause and effect. A good example is when we first try a new shower in a hotel. Perhaps the water is too cold and we turn up the thermostat. There is often a delay before the water temperature adjusts. If we are impatient for our shower, we often push the thermostat up again. It feels just right and we jump in…then seconds later the water is scalding! If we were to carry on in this way and overreact we will see chaotic behaviour in the system. If we understand the delays in the system we will be more cautious with our intervention.

We can often find such effects at work when we ride. And we are juggling with many more ingredients. For example, I feel I need more energy so I pick up my whip and ask for canter. This can give me too much energy and I lose my ability to control it. So I concentrate on quiet work in walk. Perhaps I overdo it and lose my forwardsness again. So I may canter again and so on. I am always trying to find the right balance of ingredients. This requires great self control – both mind and body.

Things get more complicated still when it is not the variable itself but an accumulation of past variables causing the effect. This is the source of our famous saying, “the straw that broke the camel’s back”. Examples from the horse world include feeding poor quality and/or unsoaked hay or riding on a bad surface. Like the boiled frog we don’t see the effect immediately so carry on regardless because we either don’t see or choose not to see the structure behind it. Once the horse has COPD it is too late. Once the horse has bad legs it is too late.

Can you think of any other examples which have affected you?

We can view our diagrams either as causal loop diagrams (CLD) or we can rearrange our causal loop diagrams and view them as driver trees. We can select a variable from our CLD and construct a driver tree showing a hierarchy of the variables affecting it. So for example taking the CLD in figure 7 and converting it into a driver tree format for the variable “Actual performance” gives us the following:

Figure 8 Driver tree

This says actual performance is affected by effort to improve. In turn, effort to improve is affected by the size of the gap. The gap is greater if target performance increases. The gap is smaller as actual performance increases.

Note that Figure 7 (CLD) and Figure 8 (driver tree) are exactly the same. They are just different representations of the same structure. Some people prefer to work with a CLD and others the linear representation of a driver tree.

In a similar way we can also construct an effects tree of all the variables impacted by a selected variable.

Figure 9 Effects Tree

This says that actual performance affects the size of the gap. As performance increases the gap decreases all other things equal. The size of the gap affects effort to improve. The larger the gap, the greater the effort.

We can view these trees to different levels; from just one level of only the variable directly impacted, to a complete listing of the variables impacted whether directly or indirectly.

We have to think very carefully about how the relationships work when we build our diagrams. Quite often effects we think of as direct are actually indirect. Let me give you an example. A class of students is asked what drives milk production. The list they generate includes cows, grass consumed, milk produced by each cow, temperature, health, number of calves and so on.

We then need to ask how does each of these factors affect milk production? It can feel tempting to see them all as direct effects. However, of the list above, only cows and milk per cow have a direct effect. Every other factor works through these to drive milk production. That is to say, they are indirect drivers.