Geospatial expert systems

Tony Moore

The division of computational science that has come to be known as expert
systems (ES) has its origins in the broader discipline of artificial intelligence
(AI), where it still resides. Put very simply, the broad aim of artificial
intelligence is to simulate human reasoning (Laurini and Thompson, 1992).
Expert systems are the most mature products to emerge from this field (Raggad,
1996), dating back to the mid-1960s. Since that time, when researchers at
Stanford University developed a program that used chemical expert knowledge
to automatically deduce molecular structure (Durkin, 1996), a plethora of
definitions for the emergent technology have been put forward. The following
gives an indication of how the use of expert systems has expanded to encompass
nearly every scientific discipline in that time (Cress and Diesler, 1990).
‘Expert systems are computer systems that advise on or help solve
realworld problems requiring an expert’s interpretation and solve realworld
problems using a computer model of expert human reasoning
reaching the same conclusion the human expert would reach if faced
with a comparable problem.’
(Weiss and Kulikowski, 1984)

In the literature, expert systems are also known as knowledge-based systems
(Skidmore et al., 1996), reflecting the physical computer manifestation of what
the expert knows rather than what is actually known by the expert: ‘…developed
for representing “knowledge” about some domain and for supporting procedures
for deriving inferences about the domain from some knowledge base.’ (Smith
and Jiang, 1991). The knowledge base can also be called a logistical base and
comprises rules governed by the inference engine (an integral part of an expert
system), which is a set of procedures for undertaking some kind of reasoning
(Laurini and Thompson, 1992).
In addition, Robinson and Frank (1987) have said that expert systems should:
interact with humans in natural language; function despite some errors in the
data and uncertain judgmental rules; contemplate multiple competing
hypotheses simultaneously; and explain their reasons for requesting additional
information when needed. These expert system elements and processes will be
returned to and elaborated on in more detail in the course of the chapter.
In a survey conducted by Durkin (1996) it was estimated that about 12500
expert systems have been developed. Couple this with the assertion that expert
systems have received a great deal of attention in the professional literature,
computing literature and government agencies (Robinson el al., 1986) and one
begins to picture just how extensively expert systems have been embraced.
Cheng et al. (1995) even go so far as to say that expert system research has
emerged as an identifiable field for scientific discourse, i.e. it is seen as a
distinct field of study in itself.
Despite the large number of developments, only an estimated 10% of medium
to large expert systems actually succeed (Keyes, 1989). Raggad (1996) gives a
similar figure for medium-to-large expert system success, despite
acknowledging the growing interest in expert systems. There may be any
number of reasons why expert systems fail; here are a few of them as quoted by
Oravec and Travis (1992): the ‘Tower of Babel’ syndrome; lack of feasibility
prototyping (Raggad (1996) expands on this as the failure of developers to
analyse carefully user needs and the context within which the system will be
used); weak inference engines; hard, slow and tedious required knowledge
formulation; and inadequate knowledge (i.e. uncertain, incomplete,
controversial, spatial and temporal, hypothetical, self-knowledge,
metaknowledge etc.). Raggad (1996) adds that problems arise because the expert
is treated as the end user, which rarely happens.
On balance, Durkin (1996) has remarked that progress since the inception of
expert systems has not lived up to the initial successes and resultant hopes. This
said, expert systems have come far, and still have enormous potential. Some
signs of this optimism can be observed when regarding the status of expert
systems in geography. Fischer (1994) has noted that artificial intelligence has
received an ‘explosion’ of interest in the last few years. Furthermore, Fischer
asserted that there was no longer any question that expert systems (and neural
networks) would be integral in building the next generation of intelligent
geographical information systems (GIS). The reason why there is plenty of
scope for use of expert systems in this subject area is that GIS without
intelligence have a limited chance to effectively solve spatial decision support
problems in a complex or imprecise environment. This air of optimism has been
present since the early days of expert system application to geography.
‘…benefits to be accrued from taking the AI approach far outweigh
any possible disadvantages.’
(Fisher et al., 1988)

In general, this chapter reviews and assesses the application of expert systems to
the geospatial disciplines. It aims to be a picture of the status of expert systems
in geography. Firstly, the history of expert systems is outlined, before exploring
the differences between expert systems and conventional systems. Then aspects
relating to the physical form of the expert system are outlined and their coupling
to GIS explained. Next there is a review of the current status of expert systems
related to geography. Topical issues such as knowledge acquisition are
examined, before a final consideration of the practical aspects of building expert
systems. This is followed by an examination of further opportunities for expert
systems in the geospatial sciences. Finally, selected examples illustrating the
elements, processes (building, coupling), tasks (knowledge representation) and
structure (object orientation) of the expert system are summarized. There is also
a further reading section, structured by application.

This entry was posted in Articles, Gis-RS Books and tagged , . Bookmark the permalink.

One Response to Geospatial expert systems

  1. This is all very new to me and this post really opened my eyes.Thanks for sharing with us your wisdom.