When deviation or disturbance in a system is amplified instead of reduced, this is called positive feedback (deviation-amplifying feedback). A familiar example is the microphone squeal caused by turning up a public address amplifier excessively. A piercing whine develops and continues until the volume is turned down or the speakers blow out.
Why does this happen? First the amplifier/speaker system (an effector) receives signals from the microphone (a sensor). The amplifier makes the sound louder, so the sound comes out of the speakers at a higher volume.
If the microphone is near the speaker, it picks up the higher volume sound, then the system amplifies it again, making it even louder. With this loop of activity, the sound gets louder and louder until it reaches the upper limit of the system.
Feedback always involves a loop.
Positive feedback, also called deviation-amplifying feedback, can always be diagrammed as a loop. The disturbance causes more of the same disturbance, so it builds on itself. Often such chain reactions are self-terminating because the system breaks apart or consumes all the available energy.
Explosions are rapid, self-terminating positive feedback events. The explosion of gunpowder in a firecracker builds upon itself very rapidly, but only for a split second. The gunpowder is consumed and the system simultaneously blows itself apart.
Positive feedback was such a big problem for early amplifier systems that negative feedback was invented to control it. Bell Telephone Laboratories engineer Harold S. Black invented the first negative feedback circuit in 1927 to control amplifiers and prevent positive feedback. His circuit turned down the gain (amplification) when it caused positive feedback.
Nuclear bombs result from a positive feedback reaction. To trigger the blast, a non-nuclear explosion is used to produce intense energy next to radioactive material inside the bomb, breaking the atoms apart (atomic fission).
A fission reaction in a group of atoms generates so much energy that it causes neighboring atoms to undergo fission, if the right material is employed. The chain reaction builds in a split second to a powerful explosion.
In 1957, Walt Disney appeared in a documentary titled, "Our Friend the Atom." Disney used a physical metaphor to illustrate the positive feedback of nuclear fission. A room was filled with mousetraps, each loaded with a ping-pong ball.
When a mousetrap was triggered, its ping-pong ball popped into the air. The floor was covered with similar mousetraps, each loaded with a ping-pong ball, so when each ball came back down, it triggered several other traps.
Walt Disney tossed a single ball into the room, and the room exploded with ping-pong balls in about two seconds. That was dramatic enough that I remember seeing it as a 6 year old. The original Disney video as well as other replications of the ping-pong ball fission demonstration are on YouTube.
Fire is a positive feedback reaction. The disturbance (combustion) must be started, for example by lightning, and then the heat of combustion triggers more combustion in neighboring areas. A fire builds on itself as long as there is fuel and oxygen available.
When driven by the wind, a forest fire hitting a stand of dry trees can resemble an explosion. However, even a slow-
The fire at one point on the candlewick produces enough heat to start a fire at the next point. The fire continues until it reaches the end of the wick or is snuffed out, cutting its oxygen supply.
Positive feedback is invariably present in situations where there is an explosion, or where a disturbance builds upon itself. Our examples have involved feedback of energy, but the same thing can happen with information. It is "amplified" by being spread around.
Explosive growth of an internet meme is an example of positive feedback. Each person who likes or shares a meme causes other people to be exposed to it, and then they might like it or share it, spreading it further.
Wang and Wood (2011) wrote an article in the journal Applied Mathematical Modeling titled, "An epidemiological approach to model the viral propagation of memes." They showed the same mathematical model could be used to describe a viral epidemic and the spread of internet memes.
As mentioned earlier, Anatol Rapoport co-authored an article for the Bulletin of Mathematic Biophysics in 1953 pointing out the same thing. The spread of rumors could be modeled with the same equations that modeled contagion or the spread of disease (Landau and Rapoport, 1953).
In general, positive (deviation-amplifying) feedback occurs when some disturbance (deviation from a goal, or deviation from an earlier equilibrium state) causes further deviation to occur. This is what we mean by saying a disturbance builds upon itself. The phrase chain reaction conveys the same thing.
If a system leaves an equilibrium state, there must be some disturbance to the system (otherwise it would remain in the equilibrium state). If the disturbance itself causes more of the same disturbance, then it is a positive feedback loop.
No positive feedback process goes on forever. It stops when the system runs out of available resources or explodes.
In a well-regulated system like a candle, the feedback reaction is controlled by limiting the amount of fuel (melted wax traveling up the wick) available at a given time. Rocket launches are similar: they are a controlled explosion produced by consuming the rocket fuel at a carefully regulated pace.
The principle of positive (deviation-amplifying) feedback: Positive feedback is a change or disturbance in a system that amplifies itself until limited or interrupted.
As you have seen, negative feedback is not the same thing as criticism, and positive feedback is not the same thing as praise. When the terms are used incorrectly, the erroneous word usage is propagated, which is itself a positive feedback process.
Incorrect meanings can overwhelm original meanings, if they are propagated enough. So far that has not happened to the cybernetic terms. An internet search for negative feedback or positive feedback shows the original cybernetic meanings listed first.
Criticism could be negative feedback in the cybernetic sense, if it moved a person or a system closer to a goal. That is exactly what constructive criticism should do.
In the following quotation, Elon Musk described criticism as negative feedback. That could be correct in this case, because he is talking about eliminating wrong ideas, taking a person closer to goals.
And then you also should seek negative feedback... from your friends and from people who are knowledgeable.... You need to understand where you're wrong.... A lot of times people can look at what you're doing and they know that it's wrong, but they don't want to hurt your feelings, and that's why they don't tell you. (Shahan, 2014)
When a student takes lessons from an expert, criticism is intended to move the student closer to a goal of competence or near-perfection. However, if criticism makes a student do worse, that is positive feedback of a harmful variety.
In other words, criticism could lead to positive feedback, and praise could lead to negative feedback. A quiz item can highlight the difference.
Your dance instructor praised your successful execution of a difficult dance move. This moved you closer to your goal of being a professional dancer. What would this be called in cybernetics?
A careless student will pick "b" because praise sounds positive. However, a student who understands the cybernetic definitions will pick "a" because deviation from a goal was reduced.
A clever student who missed this question could argue, "I was thinking of my old abilities as an equilibrium state. The praise moved me further away from that state. It was deviation-amplifying feedback." Such a student might receive credit for showing ingenuity and awareness of cybernetic concepts.
In general, a system is showing negative feedback if the distance from a goal state or equilibrium is being reduced. It is showing positive feedback if a disturbance or change is causing more disturbance or change.
Information processing systems such as brains and computers sometimes pass information ahead to another part of the system to prepare for action. In a self-
In psychology, the word association indicates a feed-forward relationship. If an experimenter shows the word "red" to a person in an experiment, that partly activates (in the cognitive network) any words related to red, such as other color names. A person will then be slightly faster to read a word like "orange."
Studies of reading show that skilled readers unconsciously process the next three or four words after the focal point (the word currently being read). This warms up the next few words before they enter attention, making reading faster.
Warm-up effects like this are called priming in psychology. There are tens of thousands of studies of priming (discussed here).
Equally familiar within psychology is classical conditioning, the oldest and simplest form of learning. Classical conditioning is all about anticipatory responses triggered by signals.
In the classic example, Pavlov's dog learned to salivate when hearing a signal indicating food powder was about to be puffed into its mouth. Any such anticipatory response is a feed-forward process.
A signal is sent forward in time, so to speak. That really means a signal is sent to another part of the system to prepare for later action.
Feed-forward processes can be used to take pre-emptive action against dangers or undesirable outcomes. If there is a hint (based on prior experience) that a system will have trouble later, one can take action now to prevent trouble later.
In the educational system, an example of a feed-forward process is identifying at-risk students and giving them helpful attention to increase academic success later. The Head Start program in the U.S. is intended to do just that.
Landau, H. G. & Rapoport, A. (1953) Contribution to the mathematical theory of contagion and spread of information. Bulletin of Mathematics and Biophysics, 15, 173-183.
Shahan, Z. (2014, December 21) "4 Elon Musk Interviews, Tons of Awesome Quotes" Retrieved from: https://cleantechnica.
Wang, L. & Wood, B. C. (2011) An epidemiological approach to model the viral propagation of memes. Applied Mathematical modeling, 35, 5442-5447. https://doi.
Write to Dr. Dewey at firstname.lastname@example.org.
Copyright © 2017 Russ Dewey