Feedback is a circular process of influence where action has effect on the actor.
For example, a thermostat that controls the temperature in a house uses feedback. It contains a controller that turns the furnace on and off which heats the room. The thermostat also measures the temperature in the room to determine when to turn the furnace on and off. The control of the furnace changes the temperature in the room, which is measured by the thermostat which controls the temperature. The goal, in this case, is a reasonably uniform temperature which is specified by the position of the temperature dial.
We might think about designing a system that would measure the temperature outside the house, rather than inside the house, and use this to calculate what fraction of time the furnace should be running. Such a controller would have no feedback, because what would be measured would not be affected by the controller itself. Experience shows that this is not effective. Why? Well, we might realize that the temperature is also controlled by how well the furnace is working. How well the furnace is working depends on the temperature of the furnace flame. So we would have to measure also the temperature of the furnace. If windows in the house can be opened, we would have to measure whether the windows are open, and the rate of air flow (wind speed and direction). Also, we would have to measure the amount of sunlight coming through the windows. Clearly, for a thermostat, it is much better to measure what we are trying to control than the many factors that might be affecting it. Just like the feedback in a thermostat, many processes in a biological or economic systems use feedback to maintain a desirable state of the system.
This kind of stabilizing feedback is called negative feedback because an increase in temperature leads the thermostat to turn off the furnace, reducing the temperature. In some systems there can also be positive feedback, which leads to runaway increases (or decreases) in some parameter of the system. Examples of positive feedback include a population explosion: more parent rabbits results in more baby rabbits, result in more parents. Similarly, a panic, stampede or arms race all involve positive feedback.
Feedback is essential in most systematic ideas about the actions of a system in its environment; learning, adaptation, evolution, all require feedback. It is also a key part in cascading crises and
Most of our thinking about the world, and particularly the concept of logic which is designed to model thinking, is based upon a sequence of steps (inference) that separate the action from the acted upon. If logical statements are allowed to refer to themselves, paradoxes result. As in: "This statement is false." Avoiding self-reference simplifies various problems and it is a common feature of many models in different branches of science and engineering.
The existence of feedback in the world, which is a form of self-reference, and the problems with self-reference in logic suggest that logic is not a complete way of understanding the world. Interestingly, even in the context of describing the properties of natural numbers, it is possible to prove (Godel's theorem) that logic is not complete.
The paradoxes in logical thinking with feedback extend to the design of computers. Computers are designed based on logic. However, despite the best efforts to avoid self-reference, the existence of infinite loops has been proven (Halting problem) to be unavoidable in computers. From a scientific perspective, we might say that it is fortunate that computers do not exclude feedback, because computer models can simulate systems with feedback.
The study and use of feedback was one of the major goals of the field of cybernetics in the 1950s as part of the understanding of control and regulation (homeostasis) in artificial and biological systems. The field of cybernetics was built on the work of Norbert Wiener. It focused on various aspects of feedback, and the role that feedback plays in a system's response to the environment. For example, automated machines require feedback to pursue goal-directed activities.
Because feedback is an important part of learning, cybernetics was also concerned with how we know what we know and the limitations of knowledge. Thinking about various types of systems, one is immediately impressed by the difficulties that feedback causes any kind of analysis. How can we think clearly about what is happening when actions affect the actors? Note that "think clearly" has the flavor of logic which doesn't have feedback built directly into it. Thus we need new ways to think about and understand feedback.
Since the 1960s, the idea of feedback loops has been developed in the field of System Dynamics to describe and simulate elaborate feedback networks modeling the properties of social and economic systems.
Related concepts: control, nonlinear dynamics
Copyright © 2011 Yaneer Bar-Yam All rights reserved.
Feedback: A influences B and B influences A
A thermostat creates negative feedback between the room temperature and whether the furnace is on or off.
A rabbit population explosion results from positive feedback between the number of parents and babies.