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

despite the fact that the neural substrate which originally supported
these functions is irrevocably damaged.
Before the damage occurred, being in mental state x meant being
in neural state x. After relearning the lost mental functions, however,
40
  

---

being in mental state x means being in a totally distinct neural state y.
This neural plasticity – the ability of parts of the brain to take up
functions that are ordinarily carried out by quite distinct parts of the
brain – is well documented. The demonstrable multiple realisability
of mental states provides a decisive refutation of the type identity
posited by the Australian materialist.
Mental states then are not only multiply realisable across subjects
in such a way that problematises the rubric of ‘type’, they are also
demonstrably multiply realisable in the same subject in such a way that
refutes the claim that types of mental states are identical to types of
neural states.
Furthermore, it is worth briefly noting here that Australian mater-
ialism is prejudiced against the possibility of non-human minds. If
mental states are held to be type identical to states of the human
central nervous system, then it is not possible for dogs and cats, for
instance, to have mental states. I’m quite certain though that our cat
Linus and our dog Mia have mental states. They certainly don’t have
the complexity of mental states or cognitive powers that humans have
– far from it – but it seems implausible in the extreme to argue that
they lack beliefs and desires.
Australian materialism rules, by fiat, against the possibility of
mental states obtaining in non-human biological substrates. Further
– and crucially for our purposes – it rules against the possibility of
artificial intelligence. This is not, in and of itself, much of an objec-
tion but it is certainly what we might consider an untoward – and
unmotivated – consequence of the theory.
In light of the multiple realisability objection, it is clear that one
cannot continue to maintain type–type identity between mental states
and neural states. One possible modification of the theory is to retreat
to a type-token identity. This is simply to argue that whenever one is
in a particular type of mental state, there is an associated token neural
state. In other words, to be in a type of mental state just is to be in
some neural state. Mental states are still taken to be identical to neural
states, but no particular type of neural state is held to be a particular
type of mental state.
While this modification accommodates the multiple realisability of
mental states – both across subjects and within the same subject
across time – it makes for a very weak theory indeed. The theorist who
posits type-token identity is no longer making the kind of identific-
ation that facilitates an intertheoretic reduction. We can no longer
investigate mentality by doing neuroscience since there is no advan-
tage in determining which mental states obtain as which neural states.
 
41

---

If the identity is only a token identity, then this determination will
only hold for the subject under investigation at the time of investig-
ation. The results are not universalisable in the way they were for the
theorist positing type identity.
The token identity theory then – or token physicalism as it is some-
times known – is barely worth entertaining. One of our desiderata
for the philosophical adequacy of a theory of mind is its empirical
adequacy – ideally our theory of mind should direct empirical
investigation.
Fortunately, we now have the makings of such a theory at our dis-
posal and the purpose of the following chapter will be to develop it.
We’re going to do this by preserving the core intuition of the causal
theory of mind in such a way that allows for the motivations which
underpin Australian materialism, but without overcommitting in the
way their substantive type–type identification does.
Before we do so, however, it will serve our purposes to examine a
very well known thought experiment and consider one possible argu-
ment we might draw from it.
5.4 WHAT MARY DIDN’T KNOW
Thought experiment plays an important role in the philosophy of
mind. Since this is the first time we are seeing one in this volume, it is
worth very briefly discussing their role.
Thought experiments aim to prime our intuitions by asking us to
imagine certain logically possible situations. By their very nature, they
typically describe wildly outlandish and implausible situations and
the following thought experiment is no exception. It does no philo-
sophical work, however, to simply object to the physical possibility of
the thought experiment situation obtaining, although this is a
common response when first meeting them. To engage philosoph-
ically with thought experiments is to identify logical consequences of
the situation being described – to argue that such-and-such must be
the case were the situation to obtain.
With that in mind, let’s consider the case of Mary, empirical scien-
tist par excellence. Mary has access to completed physical theories –
not just our current best theories but completed theories. Mary has
been assiduously studying these theories for rather a long time and has
reached the point where she knows all the physical facts. In particular,
Mary knows everything there is to know about colour. She knows all
about wavelengths of light and the reflectance of various surfaces. She
also knows all about human neurophysiology so she knows all about
42
  

---

human sensory apparatus and the visual capacity. She knows all the