Donella A Meadows has defined a feedback loop in her book ”Thinking in Systems” as ”A closed chain of causal connections from a stock through a set of decisions or rules or physical laws or actions that are depended on the level of the stock, and back again through a flow to change the stock.” A little simpler definition can be ”A path by which some of the output of a circuit, a system, or a device is returned to the input.”
Balancing feedback loops
These loops are goal or balance-seeking. They try to keep stocks stable and within a certain range. They also resist the change in systems. If stock level gets too high, balancing feedback loop will try to get it lower. If stock level is too low it tries to get it higher. For example, you have a room in which you want to keep a temperature within a certain range. You have a balancing feedback loop that keeps the temperature within the range you want. There are always some losses of heat in the room, when the temperature outside is colder. Your balancing loop adjusts the heating depending on how much colder the temperature is outside. When the temperature outside is much warmer than in the room, the heat outside comes in. Then your balancing loop will cool the air inside.
These balancing loops are not always working properly. The information from them can come too late, be unclear or incomplete or hard to understand. The action they cause can be delayed or ineffective, etc. Stock-maintaining balancing loops must have their goals set for compensating the draining or inflowing processes that change the level of the stock. All these loops have their own breaking point. This is the point where the pull of other loops is stronger than the pull of the balancing loop. When your heating system is not having enough heating power compared to the leaks outside, your room temperature gets too cold.
Reinforcing feedback loops
These loops are amplifying or reinforcing. They can cause virtuous or vicious effects of healthy growth or complete destruction. It creates a bigger or smaller inflow to a stock than is already there. Reinforcing loops enhance the direction of the change in a stock or a system. When a system element can reproduce itself or grows as a constant fraction of itself, reinforcing loops are found in the system. Reinforcing loops produce an exponential growth. Growth gets faster all the time. For example, the bigger the interest in your bank account, the more money you will get into your account every year, because you get the interest for the interest too. These loops may not change the system until the path of least resistance is overcome. For example, the sales of the new product may not start growing faster until a certain amount of product is sold.
Feedback loops are mostly linked together
Single loops are seldom at work in systems. Mostly, there are complex patterns of interconnecting loops. A single stock can have many balancing and reinforcing loops pulling in many directions. A single interconnecting flow can be attached to many different stocks. This flow can increase the ouflow of one stock and at the same time increase the inflows of many stocks. These feedback loops create a system behavior which is hard to predict. When you change the functioning or a goal of one loop, you may create changes in many others. In other words, one change can produce combinations of changes instead of one.
Lets keep things simple and concentrate on one stock system like population with both a balancing and a reinforcing feed back loops. This is one of the most common and important system structures. Lets start by identifying the most important inflows and outflows of the population. The most important inflow is people born into the population. The most important outflow is people dying. This kind of system changes when the relative strength of the loops change. In a normal situation, more people are born than dying. This means that the net effect of these loops is increasing population. It also means that the net effect is self-reinforcing growth in population.
When there is a sudden humanitarian catastrophe or a civil war in some population, the change reverses into diminishing population and the balancing loop dominates the behavior of the system. When these changes in relative strengths of feedback loops happen, the behavior of the system changes. A stock governed by linked balancing and reinforcing loops grows exponentially if the reinforcing loop dominates the balancing one. Stock declines exponentially if the balancing loop is dominant.
Other things to know about feedback loops
You have to remember that the delivery of information through a feedback loop will only affect future behavior. Feedback loop cannot deliver signal fast enough to have an effect in present. There are three typical delays in the real world. First, a perception delay. For example, a shopowner doesn´t react to any small changes in sales. She normally reacts, when the sales have changed for a longer time period, like five days. Second, there is a response delay. A shopowner makes up some part of any shortage with each new order. Third, there is a delivery delay. A subcontractor delivers the goods with delay because he has to process the order and deliver it. Changing the length of the delay will likely cause a large change in the system. When the delay is too short, the system behavior will likely oscillate.
Physical, growing systems are going to encounter limits. Those limits are balancing feedback loops. When the limits are achieved, these loops start dominating the systems by either strengthening the outflows or weakening the inflows. These limits are temporary or permanent. Eventually, the system will adjust to the limit or the limit will adjust to the system. There has to be at least one reinforcing loop delivering th growth and at least one balancing loop limiting the growth in the physical, exponentially growing systems. One interesting and important thing about feedback loops is that systems with similar feedback structures produce similar behaviors. Physically different parts do not really change behaviors.
Donella H Meadows, Thinking in Systems
John H. Miller Crude Look at the Whole