Our traditional, even classical view of control has served (and in certain ways continues to serve) us well. In our classical view, control is of a controlled (thing) by a controller. Often there is an implication that the controlled is more powerful than the controller, so that there is amplification. Usually control is towards some goal- a set of conditions, possibly progressive, to be attained- although we probably all know some control freak for whom the goal (or rather, the purpose) is just to control. The controller is seen, in this understanding, as causing the controlled to behave in particular ways.
In the Cybernetics of Cybernetics, we modify this understanding of control. In effect, we ask Juvenal’s question: "But who will guard the guardians?" and we expect an answer! Thus, we ask what it is that controls the controller. For, in a model (and it is a model) that is based on control, in which control has the position of primacy, this is not a question that can be long left unasked: to do so would be a matter of both inconsistency and laziness, and would show a lack of diligence and rigour- indicating a mammoth disregard for our subject.
The general answer is, of course, obvious. The controller is controlled, itself, by that which the controller controls (the controlled). Immediately, the relativity of the roles of controller and controlled become apparent: what we have, traditionally, thought of as the controlled can equally be seen as the controller (controlling)- it just depends on where we are looking from. There is terminological confusion, but it diminishes if we remember that controller and controlled are both roles that we (the even-present observer, but that will have to wait for another column) ascribe.
A couple of examples may help get round these - er’s and - ed’s.
In the familiar example of the so-called "thermostat", we are used to talking of the wall switch we have in our homes as the controller. Indeed, manufacturers sell us these things as controllers. (We also, mistakenly, refer to the switch as the thermostat: but thermo stat is just that: thermally static. So thermostat refers to the whole system.) Anyhow, the switch senses the temperature of the air surrounding it, compares it to some goal and sends a signal to the boiler (furnace) and pump that heat the stuff that is used to heat the air, either leaving the system in the state it is in, or turning it off (or on) to compensate for temperatures measured (by the wall switch) as either too hot or too cold. In this case, it is clear that the switch controlling the boiler and pump is being controlled, itself, by the temperature of the air that is a direct product of the boiler and the pump: ie, the switch (the controller) is controlled by the boiler and pump (the controlled), the controller is controlled by the controlled. To start, normally, from the switch as controller is merely a matter of convention.
Equally, in the computer controlled factory mentioned in the last issue, the computer is of little use unless it can somehow confirm that the controlled robots and other equipment have done the necessary; that is, unless it can sense what is happening and adjust its behaviour according to what it senses (and what it wants). The controller computer is controlled by the (controlled) robots etc it supposedly controls.
As a final example, I cannot resist the story I hope is not apocryphal of BF Skinner. Myth (or rumour) has it that before one lecture his students banded together and determined that whenever he moved to the right in the lecture hall they would smile, but whenever he moved to the left they would look disinterested or frown. By the end of the lecture, Skinner was trapped in his corner on the right of the auditorium. And (which is not normally told) the students had frozen grimaces (smiles) on their faces. The old controller, controlled himself by his controller audience, had lost none of his necessary ability as controller of the students, his controllers but also his controlleds.
Thus, control, in the Cybernetics of Cybernetics, is seen as circular: controller and controlled are (relative) roles, each -er being -ed to the other as its -er.
Of course, Wiener and the other founding mothers and fathers knew that control was circular: that was the whole point of Cybernetics, both generator and result of feedback. But they do not appear to have pushed the (Second Order) argument towards the roles of controller and controlled.
(Gordon Pask, my teacher, who Heinz von Foerster calls "Mr Cybernetics, the Cybernetician's Cybernetician" used to talk of the interactivity that this sort of circularity in control implies by reference to kicking a chair. You kick it. It does not necessarily- or normally- kick back. But your foot may well hurt. Thus, in the most linearly directional of actions there is the potential for circularity of response (ie, interaction). Having hurt your foot you may choose not to kick a chair again, intentionally, unless you are especially thespian. And so, you are controlled.)
And it is hard, once this is understood, to think of control in any other way except when circumstances are exceptional: a crowd (even a nation) held in thrall, bewitched, terrorised and hypnotised by a "great" dictator- having given up its autonomy and its acceptance of responsibility (or, more commonly (thank goodness) and much less terrorisingly, a rock star); a powerdrill being pumped through its full downward range. In effect these are general cases where feedback is very, very weak, and they bear the relationship to our current understanding of control in the Cybernetics of Cybernetics that Newton's Mechanics bears to Einstein’s: a simplification, a special case, and still tremendously useful.
(I once tried an experiment in the exercise of non-circular control in a downtown multistorey car park in Seattle: could I go down the spiral ramp, having calculated how much lock to put on my wheels, without scraping the sides of the car and without adjusting my steering? I thought I managed rather well- nearly three whole revolutions before I had to make the adjustment that avoided me scraping the sides of my hire car. (Note that I assumed that the spiral was regular, virtually perfect). But the point is, I did not manage to descend the ramp without the action of feedback, without the control becoming circular, without the walls of the ramp and the path of the car controlling my action: I adjusted the steering!)
So, in the Cybernetics of Cybernetics, Juvenal's guards will (must) guard the guards- themselves, in a circle of guarding. (Or each guard must guard his/her self alone, which is self-reference.) Otherwise, we would be faced with the buck passing problem (the chain of linear control that has no reason to stop) that "caused" Harry S Truman to state (on his desk) "the buck stops here", a phrase later somewhat downgraded in value through its use by Richard Nixon during the Watergate débacle.
But I am getting a bit ahead of myself. For we still have to consider a little more the why of control, and how it relates to cause.
"Why control?" was examined, to some extent, in the last issue. We discovered that there are at least two reasons we like to exercise control. Wishing to amplify some power (in the extreme, any power- that is, the power-freak’s wish to control control control) and the essential centrality of error in our world (which I argued was desirable). We organise control so that we can harness some quality or power that we, in ourselves, don’t have enough of (physical strength; precision of movement; and, from the examples given above, heat generation; coordination of multiple inputs; and teacher control): and we compensate for (universal) error that we find afflicting our predictions (made through models and understandings) of the world into which we construct ourselves (variations in outside temperature and slow response times; flaws in materials and the breakdown of machinery). But what of cause. How does cause enter in here?
When we speak of control being exercised, we also talk of cause. In a control system, the controller is considered to have caused the behaviour of the controlled. Control and cause can, in this universe, (always?) be used interchangeably. Thus, controllers cause behaviours; carrying out a control action causes a particular output (behaviour); the controlled behaves as and at the effect of the controller’s causing.
If control and cause are interchangeable, is it any wonder that Cybernetics has become the science of circular causality? Yet for years we understood that causality could be circular (for that is what feedback indicates) while we seem not to have understood what this intended for control. Even though control is present in Wiener's definition (neither cause nor feedback are), we seem not to have looked (explicitly) at control as circular, although I guess we understood it intuitively and that is how we came to accept circular causality.
Of course, this amalgamation of control and cause (causality) has considerable implications, which we will look at on another occasion. For we might want to try applying the understandings we have applied in each to the other, a sort of vast logical extension, amongst other things: and our understanding may then be enriched, more powerful and, possibly, more far-reaching and revolutionary. Control without the controller? Cause without the causative agent? Revolution without the revolutionary? Surely not. But that is, of course, the heart of the Cybernetics of Cybernetics.
With cause goes blame, if we want it. And we often do. Blame is our way of addressing cause: you caused this (to happen to me), you are to blame (I am not responsible). In the last World War, the excuse was often given that we were not responsible for our actions, we were controlled. I do not mean a moral lecture: I believe I would have behaved no differently, I am afraid to say. I mean to show the connection between control and cause, and between cause and blame (and between blame and responsibility). We live in a world that runs on blame. We have a legal system making fortunes out of attributing blame, even when it is quite inappropriate (uttering "accident" is to say that there is no blame). Truman’s assertion (and Nixon’s blind) attest to the dominance of blame and to the directionality of cause, in our constructions. Yet cause (control) and hence blame are circular, as we have seen (and as many a therapist will tell us), which might cause us to revise our approach. It would be to our benefit to do so: our (mental) health, our peace (of mind). And the way this happens is in adopting circularity, which means responsibility. I am responsible for what I am responsible for (and the accident is responsible for being an accident). We need to get rid of our simplified notions and enter the richness of circularity! The buck’s stopping is a way of doing this: where the buck stops, there is- if there is control, cause, blame- circularity in the stopping.
I once needed to explain to a group of spouses of Cyberneticians what the difference was between Control Theory, and First and Second Order Cybernetics (the Cybernetics of Cybernetics). The best I could do was to offer the following defensible (but still refutable) metaphor. In Control Theory, you turn on the switch and insist the light has gone on regardless; in First Order Cybernetics, you turn on the switch, and, if the light has not gone on, you replace the light bulb; in Second Order Cybernetics (the Cybernetics of Cybernetics) you turn on the switch and, when the light doesn’t go on, you think of something interesting to do in the dark (you redesign your response). For me this metaphor does at least get near to explaining the meaning of control and the value of the three approaches. I choose the third, the Cybernetics of Cybernetics, as mine: but then, I have worked all my life with design. But I am aware that this third option, unless operated within with rigour and determination, can lead- as does so much Second Order Cybernetics- to the development of a talking-shop and to solipsistic and quasi-autistic wishful thinking, claiming that anything goes and so doing nothing much. It can be an easy way out: but it should be the hardest and most demanding.
There is much still to be excavated here, but I will leave it. I did not discuss the difference between governing, controlling and regulating‹and perhaps we will come back to these in due course. The difference is, anyhow, modal (although Watt’s governor is a very nice example of a self-regulating system, giving us a way into that discussion). Control, cause and blame may also merit further reflection. Nor I have really introduced communication: I will do that for the next issue.
But I should like to end with a truly Cybernetic understanding concerning this circularity of control. Ross Ashby argued that every controller of a system (a controlled) must have at least as much variety as the controlled (system). Otherwise, it (the controller) must necessarily limit the behaviour of the controlled. In a wonderful paper (1972), Mike Robinson demonstrated how this happens in classrooms ("Classroom Control- some Cybernetic Comments on the Possible and the Impossible", Instructional Science vol. 8). But the point that must be taken from the discussion in this column is that the variety in both controller and controlled must be exactly the same, for each is the controller to the other's controlled (and each is the controlled to the other’s controller). For system designers, this means great care must be taken, a care we did not previously realise we had to take, to match variety exactly. The variety of the controller and the controlled must, for effective control to take place, be the same.
Return to the content of this issue
Return to the Cybernetics and Human Knowing Homepage