Expert Systems in Business and Finance: Issues and Applications. Edited 
by Paul R. Wilkins and Lance B. Eliot. John Wiley and Sons.

This book is a collection of 18 chapters contributed by a total of 28
authors. The editors state that the book is intended to present issues
and identify opportunities "that are or should be of interest to those
who are evaluating, researching, developing or have a particular
interest in applied artificial intelligence, especially expert
systems." The book fails to fulfill its aim for several reasons: it
lacks a clear thematic organisation; many of the articles are
unfocussed and difficult to follow; and basic principles are poorly
presented.

The editors claim to have identified four major themes around which
they have organised the book. Two of these themes are somewhat
coherent, but two are simply rag-bags of topics that have little in
common.

The first section consists of five chapters on "Issues in the design
and integration of expert systems." Four of these chapters concentrate
on issues in the integration of expert systems with conventional
information-processing systems; the fifth (actually chapter 3) looks
at the design of the user interface component of computerised
financial advisors, and has little in common with the other four
chapters.  I was disappointed in all these chapters.  None of the
authors discussed what features of expert systems might make them
difficult to integrate with conventional systems, and there was little
awareness that many of the problems discussed actually arise when
trying to integrate any set of different software packages or systems,
whether or not any of them are expert systems. Nowhere is there a
clear definition of what is meant by "expert system," which appears to
be taken as synonymous with rule-based system, thus ignoring other
possible architectures such as case-based reasoning or blackboard
systems.

There are three chapters in the second section, which is entitled
"Issues in the modeling of human judgment for expert systems." Two of
these chapters lack a clear focus: after reading them I had no clear
idea of what points the authors were trying to convey. The third
chapter, by Mallory Selfridge and Stanley F. Biggs, was the first
chapter in the book that I found worthwhile. It gives a clear
description of the types of knowledge used by auditors when making
going-concern judgements, and would be useful reading for anyone
designing or building a system to be used in the making of similarly
high level financial judgements.

The book's third section, on "Issues in validation, auditability and
security of expert systems," contains a mixed bag of six chapters.
The first is a useful survey of work in verifying and validating
expert systems by Daniel E.  O'Leary. It suffers slightly because of
its lack of discussion of why these processes are different for expert
systems and conventional systems, but provides some useful
pointers. The next chapter is a discussion of the use of expert
systems in auditing; it is followed by two on the auditing of expert
systems and two on the use of expert systems for computer
security. The matters under discussion in this section thus jump
around from looking at expert systems in general and looking at expert
systems that perform specific functions. The common theme appears to
consist only in the words used to describe the aspects of expert
systems considered (verification, validation, auditability) and those
used to describe the tasks performed by expert systems (auditing,
security).

The final section of the book is called "Legal issues, training
effects, and financial and resource planning applications of expert
systems." It is obvious from the title that the section lacks a clear
theme, and this is borne out by the four chapters that comprise it.
The chapter on legal issues considers only the situation in the USA,
and is thus of little interest to readers from other countries.  The
same applies to a survey of expert systems used in the practice of
public accounting.  The survey chapter on expert systems for resource
planning, by Carol E. Brown and Susan Athey, has a useful
bibliography.

My overall impression of this book is that it has no obvious
audience. It assumes too much knowledge of expert systems to be useful
to managers or IT specialists who would like to learn about them, and
is too abstract for those who know about expert systems and would like
to learn about specifically financial and business applications.