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

physical facts there are to know about human colour experience.
Mary, however, has never actually had a colour experience herself.
Up until now, there has been a device implanted in her brain which pre-
vents her from seeing in colour – her entire world to date has been a
world of black and white experiences. As a reward, however, for finally
learning everything there is to know about the physical world, this
device is remotely disabled, allowing Mary the capacity to have colour
experiences. The first thing Mary sees after the device is disabled is a red
rose. Mary has an experience she has never had before – the experience
of seeing the colour red. And she learns something new. She learns
something she didn’t know before, even though she knew all the phys-
ical facts about the world. She learns what it is like to see the colour red.
This thought experiment was originally intended to support an
argument against any theory which seeks to account for mentality in
purely physical terms. The argument runs along the following lines.
Mary knew all the physical facts. Yet Mary learned something new
when she had a colour experience. So there is more to know about
mental life than is provided by all the physical facts. Hence, a purely
physical account of mentality is not a complete account.
What is left out, it is contended, is an account of what it is like to
be in a mental state. These privileged subjective qualitative aspects of
mental states – what it is like to have the experience of being in a par-
ticular state – are termed qualia. We touched on qualia briefly in
Chapter 3 when we discussed the hurtfulness of pain as an objection
to behaviourism. The given argument against physicalism makes the
scope and substance of that objection explicit.
It is undeniable that there is something that it is like to be in any
given state that can only be known by having the first-person experi-
ence of being in the state. What is contentious is precisely what
explanatory burden, if any, is conferred on theorists of mind by qualia.
The thought experiment above was originally described – albeit
in a slightly di
fferent form – by the Australian philosopher Frank
Jackson in his 1982 article Epiphenomenal Qualia. The argument
against the explanatory adequacy of purely physical theories of mind
we have presented here – which is also found therein – is known as the
knowledge argument. There is an enormous literature surrounding
this argument which I do not intend to summarise here. Once again,
I refer the interested reader to the suggestions for further reading.
Rest assured, however, that this is certainly not the last we will see
in this volume of the qualia issue. For the moment, though, it is time
to turn our attention to an account of functionalism.
 
43

---

C H A P T E R 6
FUNCTIONALISM
There is very little explanatory work to be done in this chapter and,
consequently, it will be comparatively short. The reasons for this are
twofold.
One reason is that we have done much of the required setting up
for functionalism in developing the causal theory of mind. This was
one of the motivations behind presenting an account of the causal
theory in conjunction with Australian materialism.
The other reason is that functionalism is, strictly speaking, a
theoretical framework which requires fleshing out into a fully-fledged
theory of mind. One of the ways of fleshing out the functionalist
framework gives us the theory which it is the central concern of this
volume to develop and evaluate – computationalism. Before we fully
develop computationalism, however, we are going to suspend our dis-
cussion of philosophical theories of mind and work up a rigorous
technical account of precisely what computation is.
For the moment, we will be satisfied with making the clear the struc-
ture of the theoretical framework which we will later develop more fully.
6.1 FUNCTIONAL DEFINITION
We begin – as we did with the causal theory – by reflecting on the
defining characteristics of certain terms.
Many terms in our language are defined by the characteristic func-
tion of their referents. A paradigm example here is the term ‘carbu-
rettor’. A carburettor is a (now largely obsolete) device in internal
combustion engines whose function is to mix fuel and air in precise
ratios for maximally complete combustion. Carburettors can be made
of metal alloys, or ceramic, or some other material. They can employ
butterfly valves or sliders or other mechanisms to regulate their inputs
and output. They can have one chamber or multiple chambers. They
can be any colour one chooses. None of these characteristics,
44

---

however, have any bearing on whether or not something is properly
called a ‘carburettor’.
Something is a carburettor i
ff it serves the function of a carburet-
tor – i
ff it mediates between fuel and air inputs and a combustion
mixture output. Anything at all which can serve this function counts
as a carburettor.
The functionalist, as you have no doubt guessed by now, holds that
mental state terms are precisely such terms. What makes a state a
mental state is not, according to the functionalist, anything intrinsic
to the state but, rather, its function in mediating relations between
inputs, outputs and other mental states. Mental states are held to be
functional states.
Understanding any particular type of mental state is, on a func-
tionalist analysis, simply a matter of understanding its function. Pain,
for instance, is held to be a functional state which in humans is char-
acteristically caused by bodily trauma and which characteristically
causes distress and reasoning aimed at alleviating the pain, as well as
characteristically causing behaviour which is aimed at seeking relief
from the pain.
In other words, pain is a functional state which mediates relations
between characteristic pain-inducing inputs, pain-alleviating reason-
ing and behaviour. Anything at all which is apt to serve this function
– to mediate relations in such a way – just is a pain state.
No doubt you are thinking this is sounding very much indeed like
the causal theory. If you are, then you are certainly correct – the
causal theory is, in fact, an early form of functionalism.
You may also be thinking that this is rather reminiscent of behav-
iourism. Once again, you would be correct. There is a sense in which
functionalism is the new behaviourism. The functionalist account
preserves the important connection between stimulus, mentality and
behaviour. The crucial distinction, however, is that the functionalist is
not an eliminativist about mental states. In fact, the functionalist
holds that any adequate description of mental states contains an ineli-
minable reference to other mental states.
According to the functionalist, the characteristic function of
mental states is to mediate relations between inputs, outputs and other
mental states.
6.2 A BLACK BOX THEORY
In order that we fully appreciate the structure of the theoretical
framework, it is useful to represent the three levels of description – and

45

---

the identities that obtain among them – diagrammatically (see
Figure 6.1).
As we can see, there are both type identifications and token identi-
fications involved in the functionalist framework. A particular state –
I’ve used a neural state for the sake of example – is identified with a