Information Technology Reference
In-Depth Information
Chapter 10
Knowledge for Design
“It is to be noted that when any part of this paper appears dull
there is a design in it”
Richard Steel (1672-1729)
10.1
Knowledge-Based Systems
Having dealt early with the issues of intelligence and the dual semantics of computer
programs in Chap. 9, we must now address the issue of how to design a knowledge-
based system.
A Knowledge-based system was coined to distinguish a database system from
one that captured expertise or 'knowledge'. Such systems are a product of Artificial
Intelligence research. They have taken on many roles as people come to terms with
the idea of manipulating knowledge. In the first instance, a knowledge-based system
was considered by many people to be a direct replacement of an expert. The “Turing
test” was often evoked as a paradigm for a knowledge-based or expert system. This
paradigm is where given the limited means of communication with a computer it
should not be possible to tell the difference between the knowledge-based system
and an expert (hence the term expert system). The original designers of such systems
had high expectations that have never really been fulfilled.
The Turing test, as I have suggested in Chap. 1, has always been a misconceived
approach to a specification of the objectives of Artificial Intelligence, since it has
the appearance of a clear definition in that it uses equivalence, but this notion of
equivalence does not include the most important element. This element is the exact
manner in which a user of a knowledge-based system would attempt to make the
distinction between a person and a program to determine the criteria to be satisfied
that a person and a program are in practice equivalent. It is within this set of actions
or this specification of the criteria that the real difficulties lay.
The analogue of knowledge was often considered (although only half seriously by
the practitioners) as a sort of substance that was extractable (mined) from experts. The
extraction process (knowledge acquisition) is accomplished by interviews with the
expert where every utterance, action and gesture (if possible) is recorded on audio
and/or videotape. The script is then analyzed (the 'ore' refined) by a knowledge
engineer, sometimes with the help of a computer 'induction' program. The induction
