1984 - Knowledge and Representation for Decision Support Systems

Durhum, United Kingdom, 24-26 July

Proceedings edited by L.B. Methlie & R.H. Sprague.


Title Pages
Fox, M. S.

Decision making is information intensive requiring the sharing of information among different parts of the organization. How knowledge is represented will determine the extent to which it is communicated and understood. Artificial Intelligence research into knowledge representation provides an answer to the representation problem. Knowledge is represented in layers in which lower levels define the syntax and higher levels define the semantics. These concepts are illustrated using the SRL knowledge representation system to represent knowledge for job-shop scheduling.

View Full Text (PDF)
3-26
Larichev, O. I.

Given the variety of definitions of decision support systems, common to all of them is that these are systems whose basic elements are computers and decision makers. Rapid development of computers, emergence of microprocessors and flexible, rather sophisticated programming languages provide wide opportunities for their application to decision problems. But how these opportunities are exploited, to what extent they can really be helpful in decision making considerably depend on the arrangement of man-machine interaction, on the account taken of specificities and limitations of the human information processing system. The paper classifies different decision making problems wherein decision support systems were used. An analysis is performed for the major classes of the problems relative to commensurability of decision making procedures to the capabilities of human information processing system. Primary information processing operations, whose analysis allows to evaluate validity of respective decision support systems, are identified.

View Full Text (PDF)
27-39
Pulkkinen, K.

The expectations of knowledge in the organization - or in a particular unit of it - differ significantly from each other in the progressive development phases. The impact of organizational development might be seen e.g. in changes in the individual methods for utilizing the available knowledge. Differentiation of work is characteristic of a certain phase. The necessary integration is thereafter implemented on the communicative level in the next phases of development. Then the expectations of decision-markers towards knowledge representation concern its adaptation to the shift from differentiation to integration. A tentative outline for the characteristics of knowledge representation in various structural and communicative development phases of the organization is given.

View Full Text (PDF)
41-53
Stamper, R.
55-77
Bosman, A. and Sol, H. G.

Empirical evidence gathered from the application of different methodologies for information systems design, decision support system and knowledge representation clearly indicates the need for a knowledge based approach for both problem solving and information systems design. Central in such an approach is the construction of a descriptive model of a problem situation. This empirical model should be object-oriented with specifications in a declarative as well as in a procedural mode. Through the process specifications of these objects we can look into the dynamics of decision-making processes and the resulting behaviour. Through an analysis of scenarios, arising from design considerations or from a sensitivity analysis of the descriptive model, one looks for a satisfying solution. Subsequently this solution is to be constructed and implemented. The construction of the descriptive model and the process of finding solutions can be facilitated by a design environment, denoted as an inquiry system, in which various kinds of expertise with different knowledge representations have their place.

View Full Text (PDF)
81-91
Dutta, A. and Basu, A.

In this paper, a design methodology for the model management component of a DSS is presented. Using a logic framework, methods for the machine representation of computational models are developed. In order to manipulate these models as well as the other components of the knowledge base, a two-phase proof procedure is presented, using which sound solutions to a variety of queries can be generated.

View Full Text (PDF)
93-107
Embrey, D. and Humphreys, P.

Under abnormal plant condition effective decision support has to take into account the operator's or other decision maker's mental model of the plant, derived from operating experience. This will be different from the engineering model incorporated in Disturbance Analysis Systems (DAS). Recently developed approaches – for gaining access to the structure of this mental model provide the basis for the development of an interactive computer system capable of representing and exploring expert knowledge concerning inferences about causal patterns, starting from the information available to the operator in the control room. This system has potential application as an interactive diagnostic aid in support of decision making and problem solving during abnormal conditions.

View Full Text (PDF)
109-124
Lee, R. M.

Bureaucracies are organizations whose administration has been formalized in the form of explicit rules and procedures. To this extent, they are a form of artificial intelligence. What, then, is the underlying structure of bureaucratic software? Does experience in building mechanical AI systems offer any insights to the design of bureaucracies?

View Full Text (PDF)
125-132
Svenson, O., Edland, A. and Karlsson, G.

In order to illustrate possible effects of quantification of non-quantified variables in decision support systems, subjects judged the attractiveness of the same decision alternatives (student apartments) presented either numeric- ally (e.g., "size 31 sq. meters") and/or verbally (e.g., "average size"). The results indicated that the form of presentation of the most important attribute was quite important for the attractiveness of an alternative. In this case, numerical information on the most important attribute (travel time to university) was more effective in discriminating between good and poor alternatives. Furthermore, the presentation mode of the most important attribute appeared to affect the order of importance given the remaining attributes. Time stress had the effect of both giving the most important attribute more weight in relation to the other attributes and of making the alternatives appear generally less attractive on all attributes as compared to judgments in a non-stressful situation.

View Full Text (PDF)
133-144
Belew, R. K.

The adaptability of DSSs has always been one of their most salient characteristics. This paper suggests that there are some important forms of adaptation that DSSs can perform automatically. The paper is in three sections. The introduction reviews the basic components of a DSS and introduces two new information facilities, the TEXT-BASE and the RULE-BASE. A basic architecture for DSSs, organizing these informational facilities along a single dimension of information structure, is given in the second section. The third section characterizes two major types of change – the learning user and the maturing problem domain – to which DSSs must be made responsive and describes evolutionary links between information facilities that are suggested by the structural dimension. These links make a DSS responsive both to the changing user and to the changing problem domain.

View Full Text (PDF)
147-160
Brookes, C. H. P.

Progress on a project aimed at building a corporate intelligence system is described. The system has the objective of facilitating the exchange of soft information within an organisation in such a way as to support the decision making processes of the system's users. The system embodies content addressable electronic mail, a dynamic electronic conferencing sub-system, as well as an organisation alert function which are based on an analysis of the flow of soft information.

View Full Text (PDF)
161-166
Coelho, H. and Rodrigues, A. J.

Along the present paper we defend that one way to control the complexity of large knowledge bases is to build architectures in layers of abstractions, and embed partial evaluation techniques in the inference engine. Therefore one can flatten those layers and optimize the results of matching so that the overhead of interpretation is removed. This approach is facilitated by choosing a logical representation and by taking PROLOG as the programming tool for the domain modeling.

View Full Text (PDF)
167-175
Dickinson, D. and Ferrell, W. R.

Fuzzy set methods were used to represent the imprecision and vagueness in the rules given by "experts" that form the knowledge structure of an expert system that can play a large business simulation game. The rules mainly involve magnitudes of variables, relations among them and calculations based on them. A computationally convenient representation of fuzzy quantities was developed which is particularly suitable for use with fuzzy relations. The system performed well and its recommendations incorporate the fuzzyness due to imprecision in the rules and data and the cascading of fuzzy inferences.

View Full Text (PDF)
177-190
Fuglseth, A. M. and Stabell, C. B.

This paper reviews the application of modified and extended versions of the Critical Success Factors Method and the Role Construct Repertory Interview as a basis for mapping and diagnosis of manager information perceptions. Use and limitations of the methods are discussed in the context of an application to the development of a decision support system for marketing managers in four Norwegian banks.

View Full Text (PDF)
191-210
Hawgood, J.

Most decision support systems (DSS) are designed for use in "bureaucracies" (see definition below), but the value of the DSS cannot be defined without reference to the goals of the wider system which the bureaucracy exists to serve. This paper is an attempt to approach this problem by combining Elliott Jaques’ notion of “levels of abstraction” in bureaucracy with Peter Checkland’s “human activity system” concepts. The purpose is to derive same guidelines for the structuring of information for decision support, related to the “system morphology” of the organization (that is, the, structural relation between its subsystems). First Jaques’ notion is outlined, with some comments on the light thrown on it by Checkland’s approach (and vice-versa), then the concept of "supersystem language and subsystem dialect” is introduced as a description of problems of communication between remote organizational subsystems, an important factor affecting the human acceptability of “raw data”, hence a vital design parameter for DSS. Finally, it is suggested that a DSS generator for a given organization should contain a general system morphology module which would guide the construction of each specific DSS for individual managers in the organization by indicating how much detail should be provided about remote subsystems.

View Full Text (PDF)
211-220
Kobashi, Y.

A tables-oriented decision aid is introduced which is free from some restrictions on the course of interaction with a conversational decision aiding program. The use of suggestions is proposed as assistance in structuring a decision making process by means of such an aid. An action grammar would be a useful tool in representing the suggestions.

View Full Text (PDF)
221-225
Lundberg, B. G.

The approach is taken in which an information system is considered as a formal system, i.e. a formal language for which a set of inference rules is specified. This view is applied to traditional data bases as well as to generalized data bases. It is shown that the concepts of derivation rule and integrity constraint can be explained in term of inference rules. Further, the assumed approach is compared with the approach in which a traditional data base is embedded in a predicate logic system and it is pointed out that also for such systems one has to specify the applicable inference rules.

View Full Text (PDF)
227-236
Migliarese, P. G.

Organizational knowledge has evolving and dynamic properties. The paper recalls the role of organizational noise, defined
as unexpected variety which organization's members have to face, as the potential source of new levels of representation and description. Decision Support Systems are viewed, in this organizational framework, as tools for admitting and interpreting this
kind of noise. Some technical features of DSS regarding models management and manipulation are outlined in correspondence with this conception.

View Full Text (PDF)
237-244
Power, D. J.

Developing management expert systems is a difficult task. One approach that could be used to structure knowledge for such systems is the diagnostic reasoning framework of symptoms, problems and treatments explored in this paper. Previous diagnostic management methods, especially checklists and management models, are inadequate for developing expert systems. The diagnostic reasoning approach seems feasible; however, management knowledge is not structured for that framework, and much effort will be needed to restructure our knowledge.

View Full Text (PDF)
245-254
Sernadas, A. and Sernadas, C.

Several semantic abstraction primitives are introduced for capturing knowledge on the dynamic aspects of organizations taken as collections of interacting entities. The state of each entity is identified with the sequence of events which have occurred in its life. An event description includes the specification of its effects, modifiability scope, occurrence restrictions and triggering conditions. The envisaged structure of the knowledge-base for storing such information on the organization dynamics is outlined.

View Full Text (PDF)
255-267