PROUDfOOT | 303
was Turing a behaviourist?
There are three reasons to reject the traditional interpretation.
First, Turing’s own words repudiate behaviourism. He said that the concept of intelligence is
‘emotional rather than mathematical’ and that judgements of intelligence are determined ‘as
much by our own state of mind and training as by the properties of the object’ (see the next
section).^8 We can assume that mere behaviour—what a machine (or human) simply does—does
not depend on the observer. A machine’s mere behaviour is one of the ‘properties of the object’
rather than being determined by ‘our state of mind’, to use Turing’s words. It follows that intel-
ligence is not simply a matter of behaviour.
Second, the Turing test does not test machine behaviour. Instead it tests the observer’s reaction
to the machine (see the next section). The goal of the imitation game is that the interrogator
be ‘taken in by the pretence’ and a machine does well in the computer-imitates-human game
if the interrogator in that game is fooled no less frequently than the interrogator in Turing’s
man-imitates-woman game.^9 Why would a behaviourist test the interrogator rather than the
machine? The behaviourist must surely say: if the interrogator is fooled, we can infer that the
computer’s behaviour is appropriately human-like. However, this strategy makes the Turing test
a test of machine behaviour only by making it unnecessarily circuitous. Moreover, the inference
employed is invalid many critics have pointed out, we cannot infer from an interrogator’s being
fooled that the computer’s behaviour is equivalent to that of a human being—the interrogator
may simply be gullible or the programmer of the machine may be lucky.
Third, behaviourism does not explain the structure of Turing’s imitation game. In the game,
Turing said, the computer is ‘permitted all sorts of tricks so as to appear more man-like’ and
indeed ‘would have to do quite a bit of acting’.^10 But why would a behaviourist base a test on
deception rather than merely give the machine a series of cognitive tasks? Moreover, even
allowing this deception, why would a behaviourist include a human contestant rather than
merely hide the machine?^11 On the assumption that Turing was a behaviourist, some com-
mentators label his imitation game as ‘strange’. Many theorists ignore Turing’s own structure
and treat his game as a two-player game, with a human interrogator in one room and a machine
contestant in another.
In sum, the behaviourist interpretation requires us to ignore Turing’s own words and to say
that the brilliant Turing gave us a strange, circuitous test that is both easily counter-exampled
and better understood by modern critics than Turing himself! The cost of the traditional inter-
pretation is high.
If Turing was not a behaviourist, what was his concept of intelligence? According to some
modern theorists, Turing did believe that it matters how a machine produces its behaviour: in
his view, they say, thinking is an inner (brain) process and a computer’s success in the imitation
game is evidence of this crucial process. But again Turing’s own words make it clear that this
was not his concept of intelligence. He said that if we identify the ‘cause and effect working
out in the brain’ we regard an entity as not intelligent (see the next section); no brain process,
therefore, can constitute thinking. Nor did Turing see his imitation game as testing for an inner
process. As Max Newman, director of the Royal Society Computing Machine Laboratory at
the University of Manchester, said in a BBC radio discussion with Turing: ‘if I have understood
Turing’s test properly, you are not allowed to go behind the scenes and criticise the method [by
which the machine arrives at its answers], but must abide by the scoring on correct answers’.^12
How the brain (human or electronic) produces its behaviour is irrelevant.^13