Feedback is an important term from cybernetics and control theory which characterizes a process in which part of the output of a system is returned (fed back) to its input in order to regulate its further output. It describes a recursive process. If the output becomes part of the input or changes the further ouput, and the flow of information is closed, a feedback loop has been established. In a feedback loop the output of a system is sent continuously back to regulate the input or output. Causal loops and circular processes are typical for systems with feedbacks. If this feedback is used to control a system (as for example in the Observer-Controller Pattern emphasized by Organic Computing), the feedback signal indicates at the same time the current state of the controlled element.
Feedback is a circular process in which part of the output of a system is returned to its input in order to regulate its further output. According to the NECSI Wiki, "Feedback is a circular process of influence where action has effect on the actor".
The idea of feedback loops originated in the field of cybernetics, control theory and electrical engineering around 1950 and 1960.
W.R. Ashby noticed the importance of feedback in coupled systems for forms of self-organization, which are more than transitions from ‘parts separated’ to ‘parts joined’ (Ashby 1962). He argued that no machine could by itself change its own organization, if the behavior of each part is independent of the other parts' states. He added further that a ‘self-organizing’ machine is only possible by a machine coupled to another machine.
Types and Forms
Cognitive Feedback Loops
The concept of an autonomic manager used in Organic Computing and Autonomic Computing in order to achieve self-* properties is usually based on a feedback loop. This cognitive feedback loop consists of a perception phase, where the state of the managed element is observed, monitored and analyzed, and and a corresponding action phase, where the managed element is controlled through corrective action which is planned and executed.
- perceive-reason-act cycle in autonomous agents
- OC (Observer-Controller Pattern) loop in Organic Computing
- MAPE (Monitor-Analyze-Plan-Execute) loop in Autonomic Computing
Positive and Negative Feedback
Negative feedback means a higher output reduces further input. Negative feedback loops are balancing, regulating and stabilizing. They oppose any direction of change which is imposed on the system. Negative feedback results in robust and stable systems, leveling of differences and equalization or stabilization in general, and can be used to regain equilibrium. In mathematics this kind of behavior towards a limt or fixed equilibrium state is named convergence.
Positive feedback means a higher output of the system increases further input. Positive feedback loops are amplifying, reinforcing, and enhancing. They reinforace a direction of change. They leads to instable systems or polarization in general, rapid amplification of differences, and allows systems to convert graded, uncertain inputs into decisive, all-or-none outputs, see also the entries for Butterfly Effect, Path Dependence and Frozen Accidents. In mathematics this kind of behavior related to unrestricted growth is named divergence.
Iteration is the mathematical counterpart of feedback and recursion. Examples are a CA or a RBN.
Feedback is also related to recursion, self-invocation, recurrence, iteration iterated maps and functions, Iterative Function Systems (IFS) and self-similarity. Recursion and self-invocation mean a function is defined in terms of itself, which causes a function to invoke or call itself during runtime.
Links and References
W.R. Ashby, Principles of the self-organizing system, in “Principles of Self-Organization: Transactions of the University of Illinois Symposium”, H. Von Foerster and G. W. Zopf, Jr. (eds.), Pergamon Press (1962) pp. 255-278.