CHAPTER 28

Turing’s concept

of intelligence

diane proudfoot

T

his chapter sets out a new interpretation of Turing’s concept of intelligence (or think-

ing) and of his famous test, based on his overlooked 1948 and 1952 versions of the

imitation game.^1 According to the traditional behaviourist interpretation, Turing

held that thinking is nothing over and above (the capacity for or tendency to) ‘thinking’

behaviour. Yet his own words are inconsistent with behaviourism, as is the design of his test.

So what was Turing’s view? He said that ‘the idea of “intelligence” is itself emotional rather

than mathematical’ and his writings make it clear that whether or not a machine is intelligent

(or thinks) depends in part on an observer’s reactions to the machine. This is what modern

philosophers would call a response-dependence theory of the concept of intelligence. Turing’s

remarks about a machine’s learning to make ‘choices’ and ‘decisions’ suggest that he took a

similar approach to the concept of free will.

you feel fine—how do I feel?

According to the traditional view, Turing’s concept of intelligence is behaviourist—that is to

say, intelligence or thinking is nothing more than (the capacity for or tendency to) what we

call ‘intelligent’ or ‘thinking’ behaviour.^2 For the philosophical behaviourist, the hypothesis

of a hidden inner mind that guides and explains such behaviour is an illusion—the notorious

‘ghost in the machine’. From the 1950s onward Turing’s computer-imitates-human game was

interpreted as providing a behaviourist criterion of intelligence or thinking. For example, in

1952 Wolfe Mays—a University of Manchester philosopher who had created a ‘logical machine’

with Dietrich Prinz, one of Turing’s collaborators on the Manchester computer^3 —referred to

Turing’s ‘behaviourist criterion’. Turing used a ‘definition of psychological phenomena in terms