This paper presents the principles underlying our ongoing efforts to develop a knowledge-based exploratory analysis system (KEAS) which supports decision problem structuring (DPS) through an interactive processof consultation between a decision maker (DM) and a KEAS. This research is an outgrowth of the author's earlier experience in developing and using a decision support system to support decision making in a real world situation . In our earlier work, theDSS (consisting of a database, a model base, and a user interface) was helpful in supporting the decision maker's efforts once the decision maker was able to articulate the exact nature of the decision problem that must be solved. However, our experience suggests that the DSS was lacking in its ability to support the decision maker's efforts at recognizing the precise nature of the decision problem; this issue has been referred to as problem recognition  and issue formulation  in the DSS literature. More specifically, we found that: (i) a DM has difficulty specifying •a set of relevant objectives at the outset, •attributes to measure objectives before DMs know the characteristics of the system that best meets their objectives at an appropriate level,and •constraints before DMs know the consequence of imposing these constraints. (ii) a DM seeks insightsduring the problem solving process, (iii) a DM gains insights by •conducting analyses of interest to them, •applying their own substantial knowledgeabout the application domain, and •using the problem-solving knowledge of expert analysts (iv) at the outset, a DM may not have a detailed plan for accomplishing a task; they conduct one or more analyses and then decide what to do next. The underlying premise of this paper is that a DM may not always be able to precisely define a decision problem at the outset; we refer to such decision problems as ill-structured. A decision situation may be ill-structured because of the DM's inability to completely specify its decision elements and measures to quantify them; this may result, for example, from the DM's need to resolve complex issues resulting from conflicting objectives. We believe that, in an ill-structured decision situation, a KEAS must allow the DM toexplore interactively the ramifications of achieving a specified set of objectives or selecting from a specified set of alternatives; such exploratory analyses may result in a deeper understanding of what can and cannot be achieved in a decision situation. We believe that a DM can attempt to better structure the problem through a careful examination of various issues pertaining to the decision situation, aided by supporting analyses provided by the KEAS. In general, an exploratory analysis is the generation of computational results that a DM may seek to examine. We believe that, through a systematic examination of results of such analyses, a DM develops insights into the nature of a decision problem and a better understanding of what can and cannot be achieved. Hence, the KEAS is designed to facilitate a process-oriented approach that assists a DM in arriving at a complete specification of the decision problem through a cognitive process of learning, understanding, and assessment. This specification includes the identification of decision elements (e.g., objectives, alternatives, and evaluation criteria) that characterize a decision situation and of specifying the DM's expectations in dealing with that situation. We refer to the process of arriving at a specification of the decision problem as decision problem structuring (DPS). The KEAS is designed to support DPS; once a decision problem has been structured, established DSS methodologies can be applied to arrive at a
De, Suranjan, "Supporting the identification and structuring of decision problems using knowledge-based exploratory analysis" (1995). AMCIS 1995 Proceedings. 147.