Abstract for AISB 2000: How to Design a Functioning Mind
AUTHOR: John Fox
Imperial Cancer Research Fund Laboratories
TITLE: Making a mind: a cognitive engineering approach
After the 5th Generation Computer Project stimulated an explosion
of interest in AI in the 1980s the field fragmented into
competing schools of thought (e.g. symbolicists vs
connectionists, logicists vs bayesians, experimentalists vs
formalists, scientists vs engineers). For reasons which are
unclear, but perhaps just a weariness with unproductive
factionalism, AI may now be trying to recover the scientific
unity which it lost in the nineteen eighties. Proposals for
"unified theories of cognition", "intelligent agents", "theories
of consciousness" - and this conference - suggest a desire to
return to the intellectual roots of the subject; to understand
the principles of mind(s), both natural and artificial.
Our own route to an understanding of mind has been through the
study of cognitive science applied to practical problems in
medicine and the life sciences, seeking insights into how
intelligence is manifested and constrained in complex, real-world
settings. A central result of this programme of work is a unified
theoretical framework for a class of intelligent agents that
subsumes a range of mental capabilities, from reasoning about
beliefs and goals, to decision-making and planning, perceiving
and acting. This framework is embodied in the PROforma knowledge
representation language and a systematic methodology for
constructing intelligent systems (Fox and Das, 2000).
PROforma has been strongly influenced by the focus on
understanding medical cognition and expertise, and the language
and method have been validated in a wide range of practical
applications. Some of the apparent power of the method comes from
applying classical software engineering e.g. PROforma is a
formal, compositional specification language and the method uses
a range of CASE tools for assembling software agents from
standard components. But other capabilities appear to originate
in general theoretical commitments inspired by psychology and AI
(notably the core concepts of beliefs, goals, arguments,
decisions and plans).
We are now in a position to establish a systematic empirical and
theoretical programme to investigate the range of agent types
that can be constructed using PROforma and similar frameworks.
The aim of such a programme would be a deep understanding of this
class of intelligences, their strengths and weaknesses,
capabilities and failure modes. We believe that such a programme
would provide insights into the nature and limitations of a
significant class of natural as well as artificial minds.
The presentation will briefly summarise the central components of
the PROforma theory and method, drawing parallels and
distinctions with classical perspectives, such as Newell's
Unified Theory of Cognition and Minsky's Society of Mind. It
will particularly focus on the question of whether PROforma truly
represents a method for constructing and experimenting with a
simple but interesting class of minds, or just an anthropopathic
Fox J and Das S Intelligent agents: from cognitive science to
cognitive engineering, AAAI and MIT Press (in press).
Head of Advanced Computation Laboratory
Imperial Cancer Research Fund
P O Box 123
Lincolns Inn Fields
London WC2A 3PX
John Fox trained in cognitive psychology at Durham and Cambridge
Universities, and then held postdoctoral fellowships in psychology and
Artificial Intelligence at Carnegie-Mellon and Cornell Universities
between 1973 and 1975. After returning to the UK, he was a Research
Scientist with the Medical Research Council where he initiated
research on clinical decision support systems and cognitive modelling
(themes which continue in his current research.
Since 1981 he has been responsible for directing the Advanced
Computation Laboratory, supervising a wide programme of research in
theoretical and applied AI. He is currently collaborating on a number
of projects, including COGENT, CADMIUM II, and RAGs. A notable
recent achievement by Fox and members of the ACL has been the
development of PROforma, a method and technology for authoring and
enacting clinical guidelines and protocols at the point of care.
Fox was managing director, until February 1999, of InferMed Ltd.
This company, a joint venture between ICRF and Integral Solutions Ltd,
has been set up to commercialise a range of clinical applications
based on technologies developed by the ACL (PROforma in particular).
He is now founding editor of the Knowledge Engineering Review
(Cambridge University Press), having been editor from 1984-97. He has
published some 100 papers and edited 3 books. A new book, "The
Systematic Development of Intelligent Systems" by John Fox and Subrata
Das, has been accepted for publication by MIT Press.