Minds and Computers : An Introduction to the Philosophy of Artificial Intelligence

particular mental state. However, unlike Australian materialism, this
is not a type identification. The neural state is token-identified with the
mental state by virtue of being token-identified with a particular func-
tional role which itself is type-identified with the mental state.
In other words, types of mental states just are types of functional
roles. So mental states are type-identical to functional categories. Any
state which is apt to carry out that function just is the mental state but
not by virtue of any intrinsic properties of the state. This is why the
identity is only a token identity. Anything at all could stand in place of
the neural state in Figure 6.1 if it carries out the appropriate function.
Yet another way of saying this is that the neural state, in Figure 6.1,
happens to be identical to the mental state in this instance – it happens
to be the thing which is the mental state – but not by virtue of being
any particular type of neural state. Any type of neural state (in fact
any type of state at all) can happen to be identical to the mental state
if it happens to carry out the requisite function.
Unlike the token physicalism we briefly considered at the end of the
last chapter, functionalism is not methodologically vacuous with
46
  
Figure 6.1
Type/token identity.
neural state X
token identification
mental state X
functional category
type identification
tok
en identification

---

respect to empirical investigation. Quite the contrary. By virtue of the
type-identification between mental states and functional categories,
we know precisely how we should investigate mentality empirically.
Psychological inquiry, on the functionalist account, is a matter of
determining and investigating the characteristic functions of parti-
cular types of mental states.
This is one clear advantage of the functionalist framework – it
directs psychological inquiry in just the way we require from an
empirically adequate theory of mind. To better understand mentality
is to develop an account of the mediation by particular types of
mental states of their relations to characteristic inputs, outputs and
other mental states.
A further advantage of functionalism lies in its preservation of the
intuitions which underwrote precursor theories. The behaviourist
intuition that mentality crucially involves relations between stimulus
and behaviour is preserved and the Australian materialist intuition
that mental life is to be accounted for in terms of neural activity is
accommodated.
It is important to appreciate, however, that while one is at liberty
to identify neural states as those things which serve the role of mental
states, one is not committed to doing so simply by virtue of commit-
ting to functionalism.
It is an important feature of functionalism that it is substrate inde-
pendent and, hence, ontologically neutral. Anything at all – including
a state of non-physical substance – can be a mental state on the func-
tionalist account, provided it carries out the requisite function. This
substrate independence is precisely what allows the functionalist to
accommodate the multiple realisability of mental states without
succumbing to methodological vacuousness.
A corollary of this substrate independence is the avoidance of the
species chauvinism inherent in Australian materialism. As far as the
functionalist is concerned it is an open question whether or not
non-human entities – biological or otherwise – have mental states.
Functionalism allows for the possibility of dog minds, cat minds,
Martian minds and – crucially – man-made artefacts with minds.
The functionalist framework allows for the possibility of artificial
intelligence.
As well as enjoying substrate independence, functionalism is also
mechanism independent. It says nothing about the actual mechanism
by which mental states carry out their function in mediating relations
between inputs, outputs and other mental states. For this reason,
functionalism is often called a ‘black box’ theory of mentality.

47

---

Mental states, on the functionalist account, are akin to black boxes.
We know neither what they are made of, nor what goes on inside
them. While this confers the theoretical advantages we have
described, there is also a sense in which one is left wanting by the func-
tionalist account of mental states. One wants to know more, in par-
ticular, about the details of the mechanisms which facilitate the
mediation held to be characteristic of mental states. This is why I refer
to functionalism as a theoretical framework. Di
fferent ways of speci-
fying the mechanism in question result in various fully-fledged func-
tionalist theories.
A prime candidate for a mechanism which is apt to carry out pre-
cisely such mediation is computation. Fleshing out functionalism with
a computational account of the mediating mechanism will deliver us
the theory we are centrally concerned with – computationalism.
In order that we might do so responsibly and accurately, we are
going to need a rigorous formal account of just what computation is.
This will be the target of the next three chapters.
Before we move on to this formal material, however, let’s briefly
consider a couple of standard philosophical objections to the broad
functionalist framework.
6.3 QUALIA OBJECTIONS
The two objections we will raise here target not any particular kind of
functionalism but, rather, the claim at the heart of the functionalist
framework. These are objections to the contention that there is
nothing more of importance to know about mental states beyond
their function and that carrying out such a function is su