Beruflich Dokumente
Kultur Dokumente
Cartesian dualism
Identity theories
Functionalism
Turing machines
Searles Chinese Room argument
Conclusions
Introduction
What is it to have a mind? Were certain that anyone reading this
handout has a mind. But what are the special properties we consider
minded beings to have, and are these properties shared by other animals,
or even infants? In this lecture I introduce some of the approaches
contemporary philosophers have taken to the question of what it is to
have a mind.
Some terminology: I will use the term mental state to refer to any
mental phenomenon, e.g. thoughts, emotions, sensations. So the thought
that beaver dams are cool, and the joy I feel when I see beavers building a
dam are examples of mental states that I can have.
Further reading
Clark, A. (2001). Mindware. Oxford University Press. (Introduction and
chapter 1)
1
Physicalism is the view that all that exists is physical stuff, that is, stuff
which has extension. Therefore, whatever our explanation of mental
phenomena is, it cant go around citing immaterial stuff! This gets around
the problem of causation, because if mental phenomena consist in
physical stuff, just like our bodies, then they can interact with our bodies
to cause various behaviours.
On one view of physicalism mental phenomena, like thoughts and
emotions etc., are identical with certain physical phenomena, e.g. a
cocktail of chemicals in our heads. This is known as the identity theory.
The identity theory of physicalism with regards to mental states claims
that having a mental state consists in being in a particular physical state.
For example, being in pain is identical to your C-fibres firing and there is
nothing else to it than that. Identity theory is a reductive account of
mental states: it is reducing them to physical processes.
There are two ways of spelling out the identity theory, which rely on the
token/type distinction.
The token/type distinction
How many dogs were at Crufts last year?
This question could be interpreted in two ways. The questioner might be
asking about how many types of dogs (i.e. breeds) were at Crufts last year,
and the answer would be around 300. Alternatively, she could be asking
about how many individual dogs were at Crufts last year, in which case
3
is thus multiply realisable: there are lots of different things that are
currency in different cultures, but they all share a common role.
iii. Functionalism
Putnams insight had a considerable impact on contemporary philosophy
of mind. He was saying that rather than thinking about mental
phenomena in terms of what they might be made of physically (because
this leads to all sorts of problems when it comes to non-humans) we
should be thinking about them in terms of what they do. This led to the
functionalist account of mental states. Functionalists claim that trying to
give an account of mental states in terms of what theyre made of is like
trying to explain what a chair is in terms of what its made of. What
makes something a chair is whether that thing can function as a chair: can
it support you sitting on it; does it have support for your back; does it
raise your sitting position up from the ground? Chairs can be made of lots
of different things, and look completely different, but what makes them
identifiable as chairs is the job that they do.
Putnams big claim was that we should identify mental states not by what
theyre made of, but by what they do. And what mental states do is they
are caused by sensory stimuli and current mental states and cause
behaviour and new mental states.
The belief that tigers are dangerous is distinct from the desire to hug a
tiger in virtue of what that belief does. The desire to hug a tiger would
cause me to rush towards the tiger with open arms, and it might be
caused by the belief that tigers are harmless human-loving creatures.
Whereas the belief that tigers are dangerous is caused by my previous
knowledge that tigers eat people and that creatures with big teeth are
dangerous, and causes running away behaviour as well as new mental
6
states such as the dislike of the person who let the tiger into the room in
the first place. To make the contrast with type-identity clearer: on a typeidentity view what makes the belief that tigers are dangerous distinct from
the desire to hug a tiger is the different chemical cocktails which those
states consist in. But functionalists say that this is wrong: what makes
each of these states distinct is their different functional roles. They might
also be made of different chemicals, but thats by-the-by. The interesting
difference lies in what causes them and what they do.
Part two: Mind as Computer
Functionalism provides the segue to the next aspect of this lecture,
namely, that it has become very popular to think about the mind as
analogous to a computer. Computers are information processing
machines: they take information of one kind, e.g. an electrical pulse
caused by the depression of a key, and turn it into information of another
kind, e.g. displaying a number on a screen. And one could argue that
minds are also information processing machines: they take information
provided by our senses and other mental states which we have, process it
and produce new behaviours and mental states. We individuate mental
states by the processes that they can engage in, processes which require
certain starting conditions (particular mental states and sensations) and
result in end conditions in the form of new mental states and behaviours.
The similarity goes further: what allows us to identify something as a
computer or a mind is what that thing does, and not what it is made of.
In this next part of the lecture, well look at some of the consequences of
the view that minds are computers.
iv. Turing machines
Computers come in varying degrees of complexity. There is a computer in
my washing-machine which controls the various cycles. There are also
computers which can generate complex probabilistic models which we
use to predict all kinds of phenomena: weather cycles, biological
degradation, wave formations etc. If minds are computing machines,
then how complex does an information-processing system need to be
before it counts as a mind? In his landmark paper Computing Machinery
7
and Intelligence (1950) Alan Turing (1912 1954) proposed the following
thought experiment as a response to this question.
Turing asks us to imagine three people, a questioner and two
respondents, a man and a woman. The questioner is in a different room
to the respondents, and can communicate with them via an instant
messenger style set-up: the questioner types questions which appear on
screens in front of the responding man and woman, who in turn can type
messages back. The task set to the questioner is to determine which of
the respondents, labelled only as X and Y, is the man, and which is the
woman. The mans task is to mislead the questioner into believing that he
is the woman, and the womans task is to help the questioner.
The next stage of the game is very similar, except that the man is replaced
by a computer, and the questioners task is to determine which of the
respondents is the human, and which is the computer. As before, the
computers task is to mislead the questioner into believing it is the
human, and the human-respondents task is to help the questioner.
Turings hypothesis was this: if a computer can consistently fool the
interrogator into believing that it is a human, then the computer has
reached the level of functional complexity required for having a mind.
(For a cinematic interpretation of the Turing Test, take a look at Ridley
Scotts Bladerunner ).
There are, as ever, objections to the hypothesis. It might be possible, for
example, that a machine with an extremely large database with a powerful
search engine passes the test. Thus, when asked what 84 13 is, the
machine whizzes to its set of files labelled possible subtraction sums,
pulls out the file labelled 84 13 (perhaps it is nestled between the files
labelled 84 14 and 84 12) and displays whatever it finds in that file.
And it does the same for questions like do you prefer your martinis
shaken or stirred or what are your views on Tarantino films? We would
be reluctant to label such a machine a thinking machine. This suggests
that Turings test is not sufficient for finding thinking machines, because
it does not take into account the internal structure of the machine. This
example is intended to show that the internal structure of a processing
machine matters when it comes to determining whether it is minded.
8
In addition to this concern, one might wonder if the Turing test is too
limited: surely there might be beings who cannot persuade the questioner
that they are human but who we nevertheless want to count as minded.
The Turing test relies on language, and it sets very narrow criteria for
minds, namely, that they must be like human minds. But it is not a very
big stretch of the imagination to conceive of aliens who appear to act
intelligently but who would not be able to pass the Turing test.
Further reading
Turing, A. (1950). Computing Machinery and Intelligence. Mind, 59, 433
460. There are also lots of free versions of the paper available on-line, and
it pops up in lots of philosophy of mind anthologies.
v. John Searles Chinese Room
The idea that the mind is a computing machine is certainly an attractive
one. However, there are problems with the view, and to finish Id like to
point to some of these using John Searles Chinese Room thought
experiment (1980). Searle asks us to imagine the following situation: you
are in a sealed room whose walls lined with books containing Chinese
symbols. For the sake of the experiment, we shall assume that you do not
understand any Chinese at all, in fact, you are so ignorant of Chinese that
you do not even know that the patterns in the book are linguistic symbols.
There is a slot leading into the room, through which occasionally come
pieces of paper with patterns on. You have a code-book which contains a
set of rules (written in English) which tells you what to do when particular
patterns are posted through the slot; usually this means going to one of
the books in the library, opening it to a particular page, and copying the
pattern you see there onto the piece of paper you have received, and
posting it back out through the slot. The code-book covers all possible
combinations of patterns that you might receive.
Now lets suppose that outside of the room is a native speaker of Chinese.
Unbeknownst to you, she is posting questions in Chinese through the slot,
and you are giving her coherent answers to these questions. Although
believes she is conversing with someone who understands Chinese, you
9
actually do not understand any Chinese at all, you dont even know that
you are engaged in a communicative act!
Searle is pointing to a fundamental issue facing the view that the mind is a
computing machine. Computers work by processing symbols. Symbols
have syntactic and semantic properties. Their syntactic properties are
their physical properties, e.g. shape. Their semantic properties are what
the symbol means, or represents. Thus, if I were to say let be the
symbol for start dancing, then its syntactic properties are that it has four
right angles and four equilateral sides, and its semantic property is that it
represents the instruction start dancing, and that in certain contexts
when we perceive this symbol we should start to dance.
Calculators, computers, etc. are symbol manipulating machines. But
importantly, they are only sensitive to the syntactic properties of symbols.
We program machines with rules that operate on the syntactic structure
of the symbols it receives. One rule might be If input A and input B
produce output A&B. The computer can do this operation just by
looking at the physical structure of the shapes.
The problem is that the computer does not know that it is manipulating
symbols that have semantic content any more than the person inside
Searles Chinese room knows she is manipulating Chinese characters.
This leads to a fundamental issue with the claim that the mind is a
computing machine: what part of the machine understands the symbols
that it is manipulating? With a computer it doesnt matter that the
machines processing has nothing to do with the semantic content of the
symbols, because it is the humans who use the machine that have this
information, we are the ones who give meaning to those symbols. But if
the mind is just a processor which operates on the syntactic properties of
symbols, then how can it produce a being who can understand the
meaning of the symbols? How can a mind think about dogs, when all it
recognises are the syntactic properties of that symbol? Where is the
programmer who deciphers the meaning of all the symbols?
10
11
12
Further reading
Clark, A. (2001) Mindware: an introduction to the philosophy of cognitive
science O.U.P.
Crane, T. (1995) The Mechanical Mind Penguin
D. Hofstadter and D. C. Dennett (Eds.) The Minds I: Fantasies and
Reflections on Self and Soul Basic Books
13
15