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

Exercise 10.1
Construct unsound arguments against computationalism
which trade on the other possible misconceptions we have
considered. In each case, explain why the argument is
unsound.
If there is a false premise (attributing a claim to the
computationalist) explain why the premise is false in light of a
correct understanding of computationalism.
If there is an invalid inference in the argument, give a
counter-example to the validity of the logical form employed
and explain how a particular misconception would lead one
to believe that the argument instances a valid form.
10.2 SOFTWARE AND WETWARE
Computationalism is often described as a ‘software’ view of the mind.
The human brain is seen as providing the biological computational
hardware – or wetware – which confers on humans the capacity to
have a mind. Having a mind, on this view, is a matter of having the
right program running in one’s wetware.
This provides us with a solid methodological framework for investi-
gating mentality. What we should be interested in, if computationalism

99

---

is correct, is determining the program(s) for various kinds of mental
processes with a view to building up [MIND].
Computationalism has been widely embraced in the cognitive dis-
ciplines and plays a large role in informing research programmes.
Each of the empirical cognitive disciplines approaches the overarch-
ing goal of investigating mentality in distinct fashion, commensurate
with their disciplinary methods and assumptions. In each case, how-
ever, a commitment to computationalism confers a broad methodol-
ogy for pursuing these questions. Researchers who endorse the
computational hypothesis will aim to deliver accounts which are in
principle computationally implementable. That is to say, they will aim
to develop accounts of mental processes as e
ffective procedures. Very
often, this will involve collaboration with computer scientists in devel-
oping computational models of mental phenomena.
This methodology has sparked o
ff the research tradition known as
artificial intelligence. There are weaker and stronger interpretations of
‘artificial intelligence’. The weak artificial intelligence research pro-
gramme simply involves aiming to construct artefacts capable of
instantiating particular functions which are held to be (albeit weakly)
constitutive of intelligence. This is the kind of ‘artificial intelligence’
which is often used to sell white goods.
The strong artificial intelligence research programme is of signifi-
cantly more interest and of central concern in this volume. It commits
to, and pursues, the possibility of developing artefacts which have
minds in the sense that we take ourselves to have minds.
There are certain mental capacities which appear to be unique to
human minds. These include the ability to reason complexly and
abstractly about such things as mathematics, logic and metaphysics,
and the ability to use language. Both the rational capacity and the lin-
guistic capacity implicate a number of what we might call lower-order
cognitive processes, such as the abilities to discriminate, to learn and
to remember. These lower-order processes are achieved to a greater or
lesser extent by other animals. The higher-order cognitive functions
of abstract reasoning, language production and language compre-
hension, however, are uniquely human and, as such, will serve for our
purposes as prime determinants of the kind of intelligence we
attribute to humans.
Consequently, in the following chapters, we are going to concen-
trate on various methods of attempting to develop computational
devices with rational and linguistic capacities.
We will also have much more to say about the conditions under
which we might attribute mentality to an artefact in the final section
100
  

---

of this chapter when we discuss the Turing test. Before we get to that,
however, there is more to be said here about computationalism. To
begin with, we can draw out some advantages of computationalism
from our discussion so far.
Firstly, and most obviously, computationalism fleshes out the func-
tionalist framework. Recall from Chapter 6 that we found the
functionalist account somewhat wanting as a ‘black box’ view of
mentality. Computationalism tells us what is going on inside the black
box, namely computation. Consequently, computationalism confers a
clear methodology for investigating mentality – we should be aiming
to provide computational accounts of cognitive capacities.
Secondly, computationalism allows us to specify the relation
between the mind and the brain by employing a useful wetware/soft-
ware distinction. On this view, minds are what brains do. In other
words, brains provide the computational resources to run [MIND].
We might also note at this point, that computationalism retains the
substrate independence which functionalist theories enjoy and, con-
sequently, is compatible with a purely material view of the mind
without falling prey to either the multiple realisability objections
which frustrate Australian materialism or the methodological vacu-
ousness of token physicalism.
Computationalism, then, appears to enjoy the strengths of other
theories of mind without being subject to the worst of their weak-
nesses. There are, however, a number of objections which we might
mount against the view that all mental operations are computations.
In the following sections, I will consider a number of prima facie
objections – the kind typically raised against computationalism on
first presentation. In each case, I will demonstrate how a computa-
tionalist might defend the theory against the objection in question.
We will leave consideration of more sophisticated philosophical argu-
ments against computationalism until Chapters 17 and 18.
10.3 VARIATION
The objection from variation runs as follows. Computationalism says
that humans have minds by virtue of implementing [MIND], but
human minds vary greatly. How can this be, given all minds are held
to be isomorphisms of the same formal system?
It is certainly the case that there is considerable variation among
individual minds. We all have di
fferent beliefs, desires, emotional
responses and mental capacities. A computationalist will, of course,
admit this – to deny it would be foolishness. This does not answer the

101

---

objection, however. To do so requires further reflection on the notion
of isomorphism.
Formal systems involve variables which are assigned values in any