WHAT IS CYBERNETICS? — MacKAY 405 



tors, and rudders, where the input, A, determines the form of the 

 output, B without supplying all the energy of B; and (2) devices 

 such as transmission lines, levers, springs and gear trains, where the 

 energy of B is totally provided from the energy of A. In the first case, 

 the energy of A is at least partly devoted to altering the structure 

 through which the energy for B is channeled — altering the coupling 

 between the output, B, and its internal energy supply. In the second, 

 no analogous process occurs. In the first case a cybemetician would 

 say that A exerts "active control" over B. In the second (if we wish) 

 we may speak of "passive control"; though to some of us it would 

 here seem clearer to speak simply of action and reaction. 



The important point is that in cybernetics we are concerned with 

 the action of /orm- upon form rather than of force upon force. The 

 rigorous theory of such processes is still in its infancy, and a good 

 deal of what is offered today under the aegis of cybernetics necessarily 

 has little behind it but sanctified common sense (and not always that !) . 



"WTiat light do all these developments throw on our understanding 

 of biological processes? As current scientific literature shows, they 

 have suggested fruitful questions right down to the level of the 

 components and chemistry of the individual cell, where both informa- 

 tional and cybernetic notions crop up. With the still more complex 

 structure of the brain, our problem is to find any set of manageable 

 abstractions whose interaction may be studied with profit. 



Here the role of cybernetic models is often misunderstood. It is 

 not a question of finding some artificial system that will behave ex- 

 ternally in the same way as the brain. Superficial resemblances of 

 this sort can be a curse to the theoretical neurologist. As in all scien- 

 tific research, the role of a model is to serve as a kind of template, 

 which we hold up against the real thing in order that any discrepan- 

 cies may stand out more clearly, and guide us towards the making of 

 a better one. We judge a model to be useful, therefore, not merely 

 by its predictive successes, but also by the clarity with which its 

 failures can be interpreted, and lead to its refinement. Only the un- 

 expected yields fresh information ; and even this is informative only 

 when we know what to make of it — hence the crucial importance of 

 disciplining our models as far as possible by the structural realities 

 of the system we want to understand. The so-called "black box" ap- 

 proach may serve well enough in "human engineering" ; but, especially 

 if we want our models to account for pathological as well as normal 

 conditions, our progress in science is soon halted, if not totally mis- 

 directed, unless we work hand-in-hand with those who lift the lids and 

 peer inside. 



Finally, what of the future of "cybernetic machinery"? Already, 

 we know enough to say that any pattern of behavior which can be 

 precisely specified — including the sorts of behavior that we would 



720-018—64 27 



