REGULATING THE VERY LARGE SYSTEM 13/19 



first card is led has been selected partly by the bids of the players and 

 partly by chance — by the outcome of the statistically standardised 

 act of shuffling — which has selected the distribution of the cards. 

 (Compare Fig. 12/22/1.) The Rules of Bridge ensure, in fact, that a 

 definite part of the whole determination shall be assigned to chance, 

 i.e. to shuffling carried out in a prescribed way. Such an appeal to 

 chance was frequently used in the past as a method for supplementing 

 selection. The Roman general, for instance, after having made 

 many decisions, would often leave the remainder to be determined 

 by some other factor such as the flight of the next flock of birds, or 

 the configurations shown in the entrails of a freshly-killed sheep. 

 (Supplementation was used earlier in this book in S.4/19 and 12/15.) 



In scientific work the first deliberate use of wholly uncorrelated 

 selectors to provide "random" determination to complete the selec- 

 tion imposed by the experimenter, was made apparently by Sir 

 Ronald Fisher; for he first appreciated its fundamental importance 

 and usefulness. 



(By saying a factor is random, I do not refer to what the factor 

 is in itself, but to the relation it has with the main system. Thus 

 the successive digits of u are as determinate as any numbers can be, 

 yet a block of a thousand of them might serve quite well as random 

 numbers for agricultural experiments, not because they are random 

 but because they are probably uncorrelated with the peculiarities of a 

 particular set of plots. Supplementation by "chance" thus means 

 (apart from minor, special requirements) supplementation by taking 

 effects (or variety) from a system whose behaviour is uncorrelated 

 with that of the main system. An example was given in S.12/15. 

 Thus if a chance variable were required, yesterday's price of a gold- 

 share might be suitable if the main system under study was a rat in a 

 maze, but it would not be suitable if the main system were a portion 

 of the financial-economic system.) 



SELECTION AND MACHINERY 



13/19. Selection by machine. In the preceding sections we have 

 considered the questions of communication involved when a 

 machine is to be selected. Whatever does the selecting is, however, 

 on general cybernetic principles, also to be considered as a mech- 

 anism. Thus, having considered the system 



when L acts so as to design or select the machine M, we must now 

 consider L as a machine, in some way acting as designer or selector. 



259 



