1.  Why were you initially drawn to computational and/or informational

My Oxford DPhil thesis was about whether Hume was right in saying
all knowledge is either empirical or analytic (where the latter
involves being true by definition and therefore trivial), or Kant
was right in saying that there are kinds of knowledge that are
non-empirical and synthetic (i.e. not based on or requiring
empirical testing, and neither true by definition, nor lacking
significant content). As someone with a first degree in mathematics,
I knew that what Kant said about mathematical knowledge was closer
to the truth. I tried to explain why in my thesis (which I believe
will soon be digitised by the Oxford University Library and so
become freely available) but did not get very far. I continued
thinking about that problem and other problems while I held a
lectureship in philosophy first at Hull university then Sussex

In 1969 Max Clowes, a leading UK researcher in Artificial
Intelligence came to Sussex, in the Experimental Psychology Lab in
the School of Biological Sciences. Somehow we met and started
talking regularly. I found his ideas and the general field of AI
very interesting and decided to sit in on the lectures on AI and
programming offered to students in experimental psychology.
Eventually I became convinced that the best way to make progress in
studying my philosophical problem and many other philosophical
problem was to try to design a working mind or at least working
fragments of minds, such as a mathematical mind.

However, it soon became clear to me that the forms of representation
used in AI, based on logic and other forms of representation that I
called 'Fregean' because they all made use of syntax based on the
application of functions to arguments, were inadequate for the task,
and in particular inadequate for the kinds of geometrical reasoning
done by mathematicians and others. Urged on by Max Clowes I
submitted to the second international joint conference on AI (held
at Imperial College in 1971) a paper on that topic, criticising the
1969 paper by McCarthy and Hayes ('Some philosophical problems from
the standpoint of AI') because it proposed using only logic for all
of AI. I suggested that for some purposes 'analogical
representations' of which diagrams and maps were examples, could
also be useful and that manipulations of those non-Fregean forms of
representation could as much form patterns of valid reasoning as
manipulations of Fregean structures.

As a result I was invited to spend a year at Edinburgh university
where I learnt a lot more about AI and AI programming.

The more I learnt about AI and computing concepts the more convinced
I became that philosophy needed to be redone in the framework of AI,
and said so in a widely ignored book published in 1978 'The computer
Revolution in Philosophy: Philosophy Science and Models of Mind'.
(Since the online version was copied to the ASSC eprints web site,
this seems to be changing.)

One of the consequences of all that was that I started doing
research in vision in an AI framework around 1975. I thought then
and still think that understanding the forms of representation and
computation required for vision was the key to understanding many
aspects of mind. But I have also learnt that that is far beyond the
current state of the art, and that most of the AI research that
claims to be about vision is not about vision but about something
else, e.g. statistics.

2.  What example(s) from your work (or the work of others) best
illustrates the fruitful use of a computational and/or informational
approach for foundational researches and/or applications?

Almost everything I have been doing for the last 35 years would have
to be listed in answer to that question. I've started pulling the
threads together on this web site:

However the ideas are continuing to develop so quickly that I cannot
keep that web page up to date. Recent examples include:

a new view of vision as involving primarily the perception of
processes rather than static structures,

a new analysis of information as a rich and deep concept that cannot
be defined except implicitly in the context of a theory of varities
of information processing

a new view of compositional semantics in both human language and
internal forms of representation as inherently context sensitive,
with implications for Gricean theories of communication and the
nature of language learning as a collaborative creative problem
solving process rather than absorption of an external structure

a view of causation which, instead of attempting to choose between a
Humean (correlational, statistical) notion of causation and a
Kantian (structure-based, deterministic) notion of causation
accmmodates both, while much learning involves a move from Humean to
Kantian understanding of particular types of causation

a new view of conceptual analysis in philosophy, according to which
underlying what Ryle referred to as 'logical geography' there is
something deeper which I have labelled 'logical topography', the
study of which involves theory construction (using abduction) in a
way that brings philosophy much closer to science than most
philosophers allow, or want to allow.

a new view of evolution as involving tradeoffs that lead to design
problems requiring a spectrum of solutions ranging between precocial
(or genetically pre-configured) and altricial (genetically
meta-configured, i.e. based on learning and development controlled
by high level implicit assumptions about the nature of the
environment). This work is being done with a biologist, and is
illustrated by these two papers (one short, one long)

and many more.

3.  What is the proper role of computer science and/or information
science in relation to other disciplines?

There is not just one proper role.

4.  What do you consider the most neglected topics and/or contributions
in late 20th century studies of computation and/or information?

Two things.

Study of the variety of functions of vision and the extraordinarily
complex set of requirements for visual mechanisms.

Study of the implications of the fact that running virtual machines
which are not physical machines, but require physical machines for
their implication and do things, i.e. cause events both in virtual
machines and in other things. Most philosophers, psychologists and
neuroscientists are totally ignorant about this even though they use
several virtual machines in their daily life and work (operating
systems, word-processors, spelling checkers, internet browsers,
email systems, computer games, and many more.) Because current
computing education mostly focuses on the USE of computers rather
than understanding the principles on which they operate and how they
can be designed and extended people studying information processing
systems such as brains and minds lack the conceptual framework
required to make real progress.

A partial remedy is proposed here: a syllabus for teaching AI in

5.  What are the most important open problems concerning computation
and/or information and what are the prospects for progress?




Progress will depend on deep new ideas that don't seem to be
anywhere on the horizon as yet. But I have been working out some of
the requirements for those ideas, and that work is making progress.