| Print-friendly PDF file |  |
|
FEEDBACK |
|
by Piotr Golec (2004) |
|
|
|||||||
|
Contents |
||
|
|
||
|
The mechanism of feedback has a very simple definition: "the return to the input of a part of the output" [1].
This simplicity should however not undermine the importance of feedback mechanisms and their ubiquitousness in our life, both on macro- and micro-scales. In order to further present the concept of feedback mechanisms I introduce a simplifying division between the systemic view of feedback and the decision making view of feedback. From the point of view of a system (understood as "a regularly interacting or interdependent group of items forming a unified whole" [2]), feedback mechanisms have a very important role to play. Through "feeding back" a part of output again into the system, we obtain a perfect regulatory mechanism. This regulation is based on two basic kinds of feedback, namely: positive and negative feedback. Positive feedback mechanisms We call a feedback mechanism positive if the resulting action goes in the same direction as the condition that triggers it. A good example of positive feedback is a turbo-charger fitted to the engines of vehicles. As we accelerate, increasing revolutions of the engine (after crossing some threshold), set the turbo-charger on, that in fact increases the speed even further. Summarizing, a part of output (acceleration) was "fed back" to the process again, causing action going in the same direction (further acceleration). Negative feedback mechanisms A feedback mechanism is called negative if the resulting action opposes the condition that triggers it. To name an example, one might think of the heating systems we use at homes. Very often we set our heaters to maintain constant temperature while we are at home. So the heaters turn on and off as a function of the interior temperature. If the temperature drops below some threshold, the heating system is being switched on, compensating and increasing the temperature again. In other words, a part of output (falling temperature) was "fed back" to the system again, causing action going in the opposite direction (increasing temperature). These two above mentioned mechanisms of positive and negative feedback constitute a basis for system controlling. Furthermore they can be put together in different configurations composing "feedback loops". Feedback
loops We may divide feedback loops into negative and positive. If positive and negative feedback mechanisms are alternating in one system, we talk about the negative feedback loop. This kind of feedback loop has stabilizing properties. As an example one might think of some biological (food-chain) mechanisms. Increasing population of storks causes a decrease in the population of frogs, which in turn causes the decrease in the number of storks until the moment that the number of frogs is up again. In this sense, negative feedback loops are (ceteris paribus) auto-regulating.
Positive feedback loops can go into two directions: they can be either "exploding" or "imploding". If only positive feedback mechanisms are governing a system, this kind of positive loop is called "exploding".
As an example we might mention a positive feedback loop between income and consumption. The bigger the income per capita in an economy, the more people consume, therefore further increasing their income per capita, and so on. Ceteris paribus, this mechanism will continue infinitely. In many industries, success feeds success, i.e. a successful firm makes money (output) which is partly used to improve the same reasons of its success (input factors). Costly innovation, if successful, give rise to profits which allows for further Research and Development, as you can experience with this business game. In consumption, positive feedbacks can be linked to imitation phenomena. External funding (e.g. from banks) is positively linked to own capital commitment, so that loans are given only to healthy firms, boosting them to even higher levels. For the system effects of financial fragility see this paper. Similarly, rich families are usually able to assure to their children a good level of education (e.g. by buying more books, by travelling, by additional education resources), which in turn positively affect the employability and family income. If only negative feedback mechanisms are governing a system, we call this loop "imploding". For a rather exaggerated example, think of a person who loses appetite when preoccupied. Once she starts to worry, she loses weight, therefore being even more preoccupied seeing her state and losing even more weight, and so on. If nothing else stops this vicious circle (e.g. a societal help), imploding positive feedback loop leads to the self-destruction of the system.
This is the dramatic case of denutrition and low individual productivity (thus, income) in many areas of the Thirld World. From the stance of a system, feedback mechanisms are very important for mutual interaction of the system's elements. However this point of view is in a sense very "endogenous", i.e. treating the parts of the system as being inert and not able to vary their behavior depending on the state of the environment. In order to further analyze the complexity of feedback we have to allow for the decision making feature of the system's parts. From the decision making stance, feedback is specifically defined as "the transmission of evaluative or corrective information to the original or controlling source about an action, event, or process" [3]. From now on, I will focus more in detail on the processes controlled by humans (as opposed to the processes controlled by the artificial intelligence, i.e. computers). Human decision making is largely based on the concept of feedback. To mention an example on the individual dimension, managers try to estimate the correctness of their past actions by observing the output of these decisions or managerial indicators (like Balanced Score Cards) and introduce necessary corrections. On the group decision making dimension, democratic government is also an example of a controlling body trying to incorporate information from the past into the process of decision-making in the future. Having said that, we have to notice the qualitative change in the weight of the positive and negative feedback. From now on, feedback is not only a piece of information "fed back" to the system again. Critical consequences of feedback must be now included into the system: its role in motivation, consistency and learning. The most important consequence of feedback information is its influence on the motivation and consistency of decision makers. It is generally agreed that a decision maker receiving positive feedback tends to be motivated and to continue with the previously chosen course of action only slightly modifying it. If provided with the negative feedback she has a tendency to feel demotivated and search for other alternatives of solving the problem. For instance, the consumer's behaviour could be interpreted as a trial-and-error process: to buy a good - say because of a hint of possible use - to try it, to judge the experience and to renew the purchase in case it is pleasant or search for new brands in the opposite case. Brand loyalty is often explained in these terms. It should be however noted that this simplified and generally accepted view of interactions between feedback, motivation and consistency in decision making has been lately seriously challenged on the academic grounds. Explore by your own this issue by playing this business game. The second, equally important, consequence of feedback is its relation with learning. Generally experts (but also simple examples from everyone's life) indicate that no learning would occur if some kind of feedback was not available. One can easily imagine that a driving lesson with driver's eyes closed and ears plugged would almost certainly result in an accident. In our life, feedback seems to be an inseparable part of learning. However, recently some researchers stated that people may learn rules without any feedback whatsoever; they only need more time to do so. For a discussion about the relationship between learning and feedback see this paper. 4. Further dimensions of feedback There are many ways to divide feedback. Below I propose just a few, basing on the: 1.
simplicity of the feedback; simple feedback is generally based
only on one cue and complex feedback on multiple cues.
From the systemic stance as well as from the decision making point of view, feedback has a crucial disadvantage: it refers to the past, or at best, to the present. Feedback control allows the selection of actions on the basis of past or current information about the system. A control system referring to the future is called feedforward. It requires higher situational awareness, in order to choose an action on the basis of the predictions of the future state of the system. Feedforward control is superior to feedback control because of its focus on avoiding problems rather than fixing them, but requires more adequate knowledge and cognitive effort. Neoclassical theory
has based much of its interpretation of human choice process in terms
of feedforward mechanisms, also in the form of inter-personal game theoretic
interaction. Now, a fully new strand of empirical experiments is sheding
light on the complexity of this issue. 6.
Formal models Feedback and feedforward in a simple monopolistic market Human and artificial agents interaction in a race to a new market A Positive Feedback Loop Innovation System (POLIS) applied to the actual economy of Taiwan Positive feedback trading and herd behaviour by pension fund investors: an empirical analysis Positive feedbacks in the economy 7. Experimental results with real people Influence of delayed feedback on learning, performance and strategy search - revisited Macroeconomic indicators from 136 countries: a long term time series
[1] Merriam-Webster dictionary. [2] Idem. [3] Idem. |
||
