Dear Colleague
May I draw your attention to a recent book published in July 2000 by AAAI
Press and MIT Press called "Safe and Sound: Artificial Intelligence in
Hazardous Applications" (foreword by Herbert Simon). I am taking the
liberty of writing to you because the book emphasises the potential risks as
well as benefits of AI technology and argues that quality and safety are
critical research issues for our field. An overview of the book is included
below.
Thank you for your consideration.
John Fox
--------------
Safe and Sound: Artificial Intelligence in Hazardous Applications
By John Fox and Subrata Das
AAAI and MIT Press
326 pp., references, index, illus., $40.00 hardcover
ISBN 0-262-06211-9
Computer science and artificial intelligence are increasingly used in
hazardous and uncertain situations where small faults or errors can spell
human catastrophe. This book describes a theory and a technology for sound
medical decision-making and safe patient care from the perspectives of the
AI researcher and the practical software developer. It champions the
achievements of the logic programming and AI communities and presents a
design framework for constructing intelligent systems based on a logical
agent model.
The book grew out of a programme of research into AI and basic cognitive
functions like reasoning, problem solving, risk assessment, decision-making
and planning. These are well-established research topics in cognitive
science but the work described here is unusual in its focus on the
integration of these functions into a unified method for building
intelligent agents, and in the need to measure success in practical as well
as theoretical terms.
The developers of practical AI systems can take a purely engineering
approach, but some of the medical challenges discussed in the book demanded
new concepts and new computational techniques. Logic and logic programming
provided the necessary foundations for these. The authors were attracted to
logic programming because they work in a major cancer centre and their
applications are to be used in real patient care. Since such applications
are safety-critical they were interested in tools that had a sufficiently
strong foundation that they could provide some guarantees of quality and
integrity.
While logic programming is extensively used for teaching and research in
academia Fox and Das were surprised, and disappointed, that it has not been
more widely adopted in medical informatics or in the software industry
generally. They suggest that one reason for this lack of acceptance may be
that in the early years the academic community focused on the technical
elegance and theoretical virtues of logic rather than addressing practical
engineering needs. Whether this is a correct diagnosis or not the method
presented in the book sets out to combine insights and ideas from logic and
cognitive science with lessons from software engineering. The result is
PROforma, an agent specification language formalised in classical and
non-classical logics, and a versatile development toolkit for designing,
testing and delivering PROforma applications.
The book is divided into three parts. The first two parts are written in an
informal style, beginning in Part 1 with the medical background and
motivations, technical challenges, and a presentation of the PROforma
method. Part 2 provides a wide-ranging discussion of intelligent and
autonomous agents, with particular reference to safety and hazard
management. The final part provides a formal presentation of the PROforma
language and other aspects of the agent model developed in the book.
Although the application focus is medicine, the method and the underlying
ideas may be applicable in many other domains.
Logic programming could be due for a resurgence of interest in the practical
world, not least because complex software is increasingly reaching into
technical realms that are mission- and safety-critical and the world
increasingly demands guarantees of soundness and reliability. If this is so
then Fox and Das provide evidence of its practical power in a field that
directly affects us all, and present a set of engineering techniques that
could bring logic programming and AI to a wider audience.
--------------
John Fox is a principal scientist at the Imperial Cancer Research Fund and
Professor of Intelligent Systems at University College, London (UK). Subrata
Das is a principal scientist with Charles River Analytics Inc. Cambridge,
Mass (USA).
