Machines and Societies 
of behaviour that we know how to specify explicitly can in 
principle be mechanized (given unlimited computing capacity), 
using only processes now understood5. Thus any argument that 
begins “You will never get a machine to do such and such”’ 
evaporates as soon as the behaviour required is explicitly 
specified. What does not exist is a standard procedure for 
making explicit the behavioural specification that would satisfy 
us, and our fellow-citizens, in such a functionary. Even if the 
personal touch were to be dismissed as an optional extra, there 
are thus good grounds for scepticism. 
I am not, of course, saying that such mechanical substitutes 
for human decision making will never be used. Even today 
millions of our contemporaries accept and even welcome the 
impersonal arbitrations of Ernie in the distribution of Premium 
Bonds; and to persuade such a society to accept the ministra- 
tions of some similar but more rationally disciplined ready- 
reckoner in matters of law, for example, might not be 
impossible. What I am arguing is that such a system would be 
fundamentally inadequate as a substitute for human judgment, 
and that nothing in our present stage of knowledge offers any 
hope that it could be otherwise. 
MODELS AS PREDICTORS 
Our discussion of societies as self-regulating mechanisms 
suggests yet another possible function for machines, as predictive 
models. The efficiency of government action in complex 
situations, for example, could be greatly increased if even a 
crude information-flow model were available on which to try 
out a change in policy. The more refined the model, presu- 
mably, the more smooth the implementation of government 
policy (though of this more anon). 
At election time, a reliable model of voting behaviour would 
obviously be of supreme value to all who have the power to 
manipulate determining factors—unless, indeed, the possession 
of adequate models by all were to cancel out their value to 
each. 
By a model in this context one means typically a large 
161 
