4 Soft Systems Approach to Expert System Develop- ment
The main problem of expert system develop- ment is due to the dynamic nature of knowl- edge. Firstly, understanding and expressing expert knowledge is always not an easy task. Secondly, the knowledge may be changed from time to time which is subjected to additional constraints that may not be known at the ini- tial stage of the development. This is partic- ularly common to business management prob- lems. The main idea of adopting the soft sys- tems concept is not only to improve the devel- opment of expert system but also try to help the associated parties to clarify their problem situation. Through the soft systems concept, we identify some weaknesses of the traditional knowledge acquisition techniques. The first problem is related to the human re- source involvement. Although all knowledge ac- quisition approaches indicate that the develop- ment of expert system requires the involvement of the knowledge engineer, domain expert and users, most knowledge acquisition methods fo- cus on the interaction between the knowledge engineer and the domain expert. The lack of contributions from the users would reduce the usability of the knowledge based systems [l]. In the light of soft systems concept , all relevant par- ties and their roles can be clearly identified. The problem should also be expressed in a way to in- clude the objectives, needs, resources and con- straints. The second important issue is the iterative process in knowledge acquisition. All existing approaches can be regarded as iterative pro- cesses which aim at verifying the correctness of the knowledge base and validating the adequacy of the implemented system. However, there is no guidance provided for measure of performance of each stage of the whole knowledge acquisition process. The feedback decisions would be based on the arbitrary experience and judgement of an individual. To overcome the above problems, a methodol- ogy for establishing the expert system develop- ment system (ESDS) has been introduced. This approach considers the expert system develop-
82
Figure 2: ESDS
The methodology for establishing
ment environment as a soft system which can be characterised by the following components:
Objectwe: The ESDS aims at developing expert systems which can satisfy require- ments of relevant parties and the environ- mental constraints. More important, the ESDS should have the capacity to handle uncertainty and changes.
Components The components of the ESDS can be divided into three categories: hu- man resources, computing resources, and processed elements. Human resources in- clude software developer, domain experts, users and the system owner. Computing re- sources include software tools and hardware equipment. Processed elements are those related to human knowledge, information and decisions.
Connectivzty: The components which make up the ESDS form a degree of connectiv- ity. They are more closely connected to each other than the components outside the ESDS. The connectivity indicates the infor- mation flow and operation sequence among different system components.
Subsystems: The subsystem of the ESDS may be defined according to the main func- tions of the whole development process. Some commonly accepted subsystems of the ESDS include knowledge engineering subsystem, system development subsystem, and system acceptance subsystem. The
connectivity of the system implies that, the subsystems may have boundaries, across which will be flows of information. and per- haps resources. A more important property is that t.he activities of the system are acti- vated within the boundaries of the subsys- tems.
Real life problems lie between two extreme ends, namely soft and hard. According to Checkland [3], a soft problem is:
“a problem, usually a real-world prob- lem, which cannot be formulated as a search for an efficient means of achiev- ing a defined end; a problem in which ends, goals, purpose are themselves problematic .”
Thus it is impractical to formulate definite procedures to solve soft problems. Checkland proposes a soft system methodology (SSM) [3] for handling this particular type of real-world problems. The soft systems concept provides a thinking framework to represent the problem en- vironment as a human activity system for visual- ising internal and external environment factors as an integrated entity. This concept also al- lows recognition of the proper place and func- tion of all associated elements. The systems within which the different parties must be in- volved are necessarily complex. The soft sys- tems concept is expected to help to dissolve some of the complexity, and it can also contribute to clarifying the complex problems and thereby op- erate within the perceived environment. The construction of a human activity system should consider the following elements:
0 objective
0 components
0 connectivity
0 transformation
0 subsystems and wider systems
0 measures of performance