PHILOSOPHICAL FOUNDATIONS OF ARTIFICIAL INTELLIGENCE

A course for undergraduate AI students

Aaron Sloman
http://www.cs.bham.ac.uk/~axs

This document is available in two formats.

http://www.cs.bham.ac.uk/research/projects/cogaff/misc/philosophy-of-ai.html
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/philosophy-of-ai.pdf

This course is not currently available. The information is provided

for potential developers of courses linking Philosophy and AI.
Original version: 2003
Partial Update 31 Jan 2020

CONTENTS
INTRODUCTION
ASPECTS OF PHILOSOPHY
THE RELEVANCE OF AI TO PHILOSOPHY
A PROVISIONAL PLAN FOR THE COURSE
REFERENCES

INTRODUCTION
There are many ways of teaching philosophy. A standard way is to ensure that students know what
various philosophers have thought and written.

In the days when I taught philosophy the time available was too short, so my aim was to try to
ensure that students learn two main kinds of things.

(a) A family of new concepts that are useful for formulating and discussing philosophical questions.

(b) How to do philosophy. That is very hard to teach. I don’t really know how to communicate it
except by doing philosophy and helping students to do it, hoping that they will somehow pick up the
techniques by imitation and practice. I tried to outline some of the techniques in my 1978 book The
Computer Revolution in Philosophy, which is now out of print. However describing the process of
philosophising does not seem to be good a way of communicating what it is about to those who
have not yet learnt to do it.

To that extent, learning to do philosophy is something like learning to play the violin, or ride a
bicycle. It’s very easy to play out of tune, or lose your balance.

As this is a single module course you are expected to do on average about 6.5 hours a week of
work on the course, including all the time spent in lectures or classes, reading, writing notes,
thinking, etc. A significant amount of that time should be spent in the main library or using the
internet, finding things out for yourself, e.g. initially by looking for encyclopaedias and dictionaries
of philosophy, so that you learn where to find things out quickly, reading books that give overviews,

1
and later on reading recent articles which analyse some of the problems we’ll discuss.

You may find it useful to start with an introduction to some fairly ’traditional’ ways of doing
philosophy. When switching from maths and physics to philosophy as a graduate student, over half
a century ago, I found the two books by Bertrand Russell, enormously helpful, even though I did
not always agree with him. Or try to find one of the introductory books listed below, e.g. by
Campbell, Hospers, Magee (on Popper), Mitchell, or one of the collections of readings in
philosophy by Edwards & Pap, Goldman or Lycan. There are very many more books that give
introductory overviews, and you will find others by looking in the library, or online, for introductions
to philosophy, to metaphysics, to epistemology, to philosophy of mind, to ethics (moral philosophy)
or philosophy of science. For the first two weeks of the course, simply try to do as much general
reading as you can, and bring to the classes any questions you have about the arguments, or
concepts, or theories you find. But before you read anything try thinking about the questions.

The course is assessed by an essay submitted after the course is over. In order to be able to write
a good essay you will need practice. So each student will be expected to introduce a discussion of
one or two topics this term, and to write a sample essay on which you can get critical feedback.
Details will be arranged later.

ASPECTS OF PHILOSOPHY
There are many different ways of dividing up philosophy. I like to introduce beginners to philosophy
by asking them:

“Can a goldfish long for its mother?”

“If not, why not?”

Note:
There are many variants on the questions about a goldfish. For instance you could ask about other
mental states, or about about other animals. Here are some examples (in each case add the the
question: “What does this question mean?” “What techniques of enquiry or analysis are relevant to
deciding whether one answer is better than another?”

1. Could a mouse desperately hope that her children will do well in life?

2. Could a tadpole hope that it will survive to be a frog and make more tadpoles?

3. Is a fly aware of my fly-swatter coming down to flatten it? (If not, why does it always escape?)

4. Is the fly afraid of being hit by the fly-swatter? (If not, why does if fly away?)

5. Can have robots ability to have hopes, awareness, beliefs, fears, ... How? If not, why not?

These questions relate to several different aspects of philosophy:

Metaphysics: What kinds of things exist? (E.g. fish, material things, mental states, relationships.)

Epistemology (or theory of knowledge): What can we know and how do we know it? In
particular how could we tell whether a goldfish does or does not long for its mother? Can we ever
know about the contents of “other” minds? For that matter can we know about anything other than
our own mind and its contents?

2
Note: The question “How can we know X?” has at least two very different interpretations. On one
interpretation it asks about the causal history of our knowing: did we see the evidence, or hear it,
read about it, or infer it from some experiment, use our own tests or belief what someone else said,
etc. On the other interpretation it is a question regarding justification: is the means by which you
came to know X sufficient justification for claiming that it is true, or that you know it is true?

Philosophy of mind: What are mental states and processes? What’s the relationship between
mind and body? Are certain material states sufficient to produce mental states? Can mind and
matter interact causally? Could longing cause swimming? Can sadness cause weeping? If so,
how?

Ethics or moral philosophy: How should we think about the rights of a goldfish? Do we have the
right to kill them? To cause them pain? To take them from their mothers? How can such questions
be decided? Can ethical questions have right or wrong answers? If so, why is there so much
disagreement about them: shouldn’t people who disagree be able to find out which side (if any) of
the disagreement is right? Compare disagreements about how far the sun is from our planet, or
what atoms are composed of.

Philosophy of science: If someone thinks it is a scientific question whether the goldfish feels pain,
then we can ask what the difference is between science and other types of knowledge, or
knowledge-seeking? What are: scientific theories? Explanations? Evidence? Can theories ever be
proved, or refuted, and if so how? What’s the relationship between the development of new
concepts and the development of new theories?

Conceptual analysis: It soon becomes clear that we are not sure what question we are asking?
What does it mean to say that a goldfish longs for something? What does it mean to say that it can
think about its mother? Or that it has a mother? (Could a tree or a rock have a mother? What about
a battle?) There are many concepts we use outside of doing philosophy, which are extremely
difficult to analyse. Examples of such concepts are: mind, matter, meaning, truth, causation,
experience, freedom, goodness, concept. knowledge, explanation, science, intelligence, emotion,
and many more.

A great deal of modern philosophy attacks old problems by showing that the questions were
confused because the concepts used were full of muddles, like “Where is the universe and which
way is it moving?”

I believe that learning to do conceptual analysis is one of the most important aspects of learning to
do philosophy: everything else hangs on it. E.g. people who are unclear about the concepts they
use can argue at cross purposes, or flounder in interminable debates (for examples see most of the
discussion in comp.ai.philosophy).

But it is also relevant to being a good scientist. One of the most spectacular examples was how
Einstein’s attempts to analyse the our concept of simultaneity led to the special theory of relativity.

Last year you had a brief introduction to conceptual analysis when we talked about similarities and
differences between: embarrassment, shame, guilt, regret and related concepts. To help you get
back into doing philosophy have a go at trying to write down what these states have in common
and how they differ.

3
There is a fairly terse introduction to conceptual analysis in Chapter 4 of The Computer Revolution
in Philosophy (1978) now available online here
http://www.cs.bham.ac.uk/research/cogaff/crp/chap4.html

THE RELEVANCE OF AI TO PHILOSOPHY

It is fairly evident that philosophy is relevant to AI, e.g. helping to set its goals and clarify many of
the concepts it uses, such as intelligence, perception, learning, memory, understanding, etc.

Equally, AI and Computer Science are relevant to philosophy because they provide a host of new
concepts and forms of explanation, as well as raising new questions relevant to old philosophical
problems, about metaphysics, about what we can know, about the relationship between mind and
matter.

For example: what is a virtual machine? What’s the relationship between virtual machines and
physical machines? Can virtual machines enter into causal relationships? Can a “software event”
like the creation of a new data-structure (e.g. a new list), cause physical events to occur, or is it
only physical things that can enter into causal relationships? Do computing systems have
“emergent” properties? How do computational machines differ from previous sorts of machines?
What are machines? Are connectionist machines significantly different from symbol manipulating
machines? What are symbol manipulations? Isn’t changing the weight on a neural link a sort of
symbol manipulation?

A PROVISIONAL PLAN FOR THE COURSE

It is impossible to plan a philosophy course without knowing the philosophical capabilities of the
students well. So here is a provisional list of topics, which we may pursue in rough chronological
order, though if appropriate we can change the order. In fact there is no right order for learning
philosophical concepts and theories: like learning a new town you have to go round and round
getting to know things better by learning their mutual relationships. So here’s a possible sequence
for the course.

(1) Philosophical concepts and jargon. Make the acquaintance of some terminology, e.g.
notions like epistemology, metaphysics, ontology, monism, dualism, reductionism, physicalism
(materialism), behaviourism, phenomenalism, idealism, epiphenomenalism, interactionism,
intentionality, “derivative” intentionality, concept, proposition, rationality, concept empiricism,
knowledge empiricism, supervenience, the analytic/synthetic and empirical/apriori distinctions,
deductive nomological explanations, the design stance, the intentional stance (Dennett),, and
various theories about the relation between mind and body.

(2) Computer science, AI and Philosophy. Try to get a feel for some of the philosophical
problems raised by computing and AI. E.g. What is a machine? What is computation? What’s the
relationship between computational processes and physical processes? Can a software event
cause a physical event? What is intelligence? Can there be a behavioural criterion for intelligence?
What’s the status of the Turing test? Can computational processes support semantics? Are neural
processes in some fundamental way different from processes on a digital computer? (E.g. can they
discover impossibilities or necessary connections in geometry or topology -- as ancient
mathematicians did?)

4
(3) Machines and intentionality. What does it mean to say that a machine refers to something, or
understands something? Searle’s Chinese room argument, and replies to it. Causal theories of
meaning. Alternatives to causal theories of meaning. Harnad’s “Symbol Grounding Problem.”

(4) Representations. What are they? What is their role in intelligence? How many different kinds
are there? Does logic have a special role? What, if anything, is special about pictorial or
diagrammatic reasoning?

(5) What sorts of machines could have minds? Are there some aspects of mind that are
particularly difficult to accommodate in machines in general, or in computers in particular? Qualia?
Pains and pleasures? Emotional states? Consciousness? Experience? How could this be settled?
Can machines have what Haugeland calls “original intentionality” as opposed to “derivative
intentionality”? The relevance of new architectures to new analyses of old concepts. What’s the
difference between simulation and replication? When is a simulation of a Y a Y?

(6) Freedom of the will and related concepts. What Marvin Minsky called “Dumbell theories”
(everything is either an A or a not-A and there’s nothing in between) and what is wrong with them.
What could it mean for a machine to have its own goals? What does it mean for us to have our own
goals? Do we have a kind of freedom machines could never have? Are we (i.e. humans!) not
machines?

(7) The importance of architecture. When you have an architecture, it defines a collection of
possible states and processes. (Think of how current theories of the architecture of matter define
different kinds of stuff - different elements, different chemical compounds, etc. Compare how
people previously thought of water, iron, air, etc.) How many of our ordinary mental concepts
presuppose specific architectures? What sorts of architectures can support mental states and
processes?

Other possible topics to be decided later. Maybe we should talk about consciousness. There are
many other concepts of ordinary language that we could try to analyse and relate to the possibility
of instantiation in machines, e.g. em sensory experience, learning, desire, pain, pleasure, emotion,
personality.

REFERENCES (Always out of date!)

Philosophy is a very old subject and has spawned a vast and diverse array of books and journals
from many cultures. Here is a tiny subset of pointers into that literature, including some traditional
introductions to philosophical ideas and also some more recent inspired particularly by
developments in AI. I have not yet read all of this myself. Some of it is included on the basis of
recommendations from others.

1.BOOKS
Margaret Boden has been one of the deepest and most influential thinkers/writers linking AI,
Philosophy, Biology, Psychology/Cognitive science, since the 1970s. A sample follows:
M. A. Boden, 1978, Artificial Intelligence and Natural Man, Harvester Press, Hassocks,
Second edition 1986. MIT Press,

5
Margaret Boden (Ed) The Philosophy of Artificial Intelligence (edited.) (Oxford University Press,)
1990.

Boden, M. (1990) The Creative Mind Abacus edition, 1992.

Mind As Machine: A history of Cognitive Science (Vols 1--2) (2006)

http://www.cs.bham.ac.uk/research/projects/cogaff/misc/boden-mindasmachine.html

More by Boden listed here: http://www.ruskin.tv/maggieb/publications.asp

Churchland, P.M. (1984). Matter and Consciousness: A Contemporary Introduction to the

Philosophy of Mind. Cambridge MA: MIT Press.

Campbell, Keith Body and Mind Macmillan, 1970.

A useful little introduction to traditional philosophical theories of mind.

Dennett D.C. (1978). Brainstorms: Philosophical Essays on Mind and Psychology Cambridge, MA:
MIT Press.

D.C. Dennett, (1984) Elbow Room: the varieties of free will worth wanting, Oxford: The Clarendon
Press,
Compare my notes on how to dispose of the free will issue
http://www.cs.bham.ac.uk/research/cogaff/0-INDEX81-95.html#8

D.C. Dennett, (1996) Kinds of minds: towards an understanding of consciousness, Weidenfeld and
Nicholson, London.

Paul Edwards & Arthur Pap (eds) (1957) (and various new editions) A Modern Introduction to
Philosophy Collier Macmillan, New York

Fetzer, J.H. (ed) Epistemology and Cognition Kluwer Academic, 1990,

Haugeland, John, (ed) Mind Design: Philosophy, Psychology, Artificial Intelligence, Bradford
Books, MIT Press, 1981.

Hofstadter D.W. and Dennett D.C. (Eds.) (1981). The Mind’s I: Fantasies and Reflections on Self
and Soul Brighton: Harvester Press.

John Hospers (1973) An Introduction to Philosophical Analysis Routledge and Kegan Paul
(Many students have found this a very useful (if somewhat wordy) general introduction to
philosophy.

Lycan, William G. (ed) Mind And Cognition: A Reader Oxford: Basil Blackwell, 1990.

Magee, Bryan, Popper Fontana Modern Masters Series.

(A very good short introduction to Karl Popper’s views of knowledge and science.)

Mitchell, David, An introduction to logic Hutchinson, 1962.

(This is mainly an introduction to philosophical analysis of logical concepts rather than an
introduction to formal logic.)

6
William J. Rapaport
http://www.cse.buffalo.edu/~rapaport/
University at Buffalo, The State University of New York,
He has a massive online freely downloadable, frequently corrected/updated book, in PDF format,
here (Last updated January 2020):
Philosophy of Computer Science
https://cse.buffalo.edu/~rapaport/Papers/phics.pdf

Russell, Bertrand The Problems of Philosophy

(An ancient, but very useful paperback book)

Russell, Bertrand A History of Western Philosophy

(When I was switching from mathematics and physics to philosophy over half a century ago, this
taught me a huge amount. Much has happened in philosophy since then, but much of it is not so
useful for beginners to chew on.)

Ryle, Gilbert The Concept of Mind Hutchinson 1949. (A seminal, yet still underrated book).

Searle J. (1984). Minds, Brains and Science: The 1984 Reith Lectures London: BBC Publications.
(An attack on AI)

Sloman, Aaron (1978) The Computer Revolution in Philosophy: Philosophy science and models of
mind, Harvester press.
Out of print but now available online with some afterthoughts added:
http://www.cs.bham.ac.uk/research/cogaff/crp/

2. JOURNALS
There are many journals that include discussion of philosophical issues relevant to Computing, AI
and Cognitive Science. Over the last ten years or so, standard philosophical journals have
increasingly included such articles, including the following:

Mind

The British Journal for the Philosophy of Science

There are also now journals that are explicitly concerned with issues to do with mind, brain and AI,
and include philosophical and non-philosophical articles.

Artificial Intelligence

Artificial Intelligence Review

Computational Intelligence

Behavioral and Brain Sciences

Cognitive Science

7
Minds and Machines

The Monist

New Ideas in Psychology

and many more.

Note added 30 Jan 2020: Since this document was originally produced, a vast amount of additional
relevant information has become available online, much of it concerned with powers of neural nets.

3. CONFERENCE PROCEEDINGS
Several AI conferences include philosophical papers, including, for example, proceedings in the
following series of regular conferences:

The Society for the study of Artificial Intelligence and Simulation Behaviour (AISB)(UK)

International Joint Conference on Artificial Intelligence (IJCAI)

Association for the Advancement of Artificial Intelligence (AAAI)

European Conference on AI (ECAI)

And conferences on Artificial Life.

4. THE INTERNET
There are increasing numbers of World Wide Web sites, some of which include philosophical
references or articles.

David Chalmers and David Bourget have set up an extensive online bibliography with pointers to
many useful philosophical web sites and publications, and philosophers:
https://philpapers.org/
https://philpeople.org/

A site with general philosophical information is The Stanford Encyclopedia of Philosophy

https://plato.stanford.edu/

Many more internet sites giving information about philosophy can be found by giving google some
combination of these words and phrases: philosophy, epistemology, philosophy of mind,
philosophy of language, philosophy of law, introduction tutorial, etc.

There are various philosophical articles included in the Birmingham CogAff Web directory,
accessible as:
http://www.cs.bham.ac.uk/research/cogaff/

Some of my slide presentations including philosophical presentations are here

http://www.cs.bham.ac.uk/research/projects/cogaff/talks/

8
5. USENET NEWS GROUPS - THE INTERNET
(These may now have to be accessed in a different way.)

There are several usenet groups that include philosophical discussion from time to time. Often the
“signal to noise” ratio is not very high.

The easiest way to read news groups is to use Google’s groups facility:
http://www.google.com/grphp?hl=en&tab=wg&ie=UTF-8&oe=UTF-8

You can try these news groups:

comp.ai.philosophy
sci.philosophy.meta
sci.philosophy.tech
alt.consciousness
sci.cognitive
sci.psychology.consciousness
alt.consciousness
comp.ai

There is an “electronic” journal of AI that is accessed via Usenet, the “Journal of AI Research”.

See the two groups:

comp.ai.jair.announce
Includes announcements of new papers available on JAIR.

comp.ai.jair.papers
Includes the actual papers, circulated in compressed, uuencoded form.
(Ask for help if you don’t know what that means.)

This work, and everything else on my website, is licensed under a Creative Commons Attribution
4.0 License.
If you use or comment on my ideas please include a URL if possible, so that readers can see the
original, or the latest version.

Creative Commons License

Maintained by Aaron Sloman

School of Computer Science
The University of Birmingham