The primary purpose of decision support systems (DSS) is to help the decision maker develop an understanding of the ill-structured, complex environment represented by the model. This paper concentrates on understanding the modeled environment through model analysis. Specifically, the purpose of this paper is to propose a framework for model analysis based on Perkins's theory of understanding and its basic premise (knowledge as design) and basic components (purpose, models, and arguments). This framework encourages enhanced user understanding in a DSS via the synergistic combination and integration of: (1) cognitive science (theory of understanding), (2) artificial intelligence (machine learning, knowledge extraction, and expert systems), (3) model analysis (deductive and inductive), and (4) DSS (model management, instance management, and knowledge-base management).
After building and validating a decision support model, the decision maker frequently solves (often many times) different instances of the model. That is, by changing various input parameters and rerunning different model instances, the decision maker develops insight(s) into the workings and tradeoffs of the complex system represented by the model. The purpose of this paper is to explore inductive model analysis as a means of enhancing the decision maker's capabilities to develop insight(s) into the business environment represented by the model. The justification and foundation for inductive model analysis is based on three distinct literatures: 1) the cognitive science (theory of learning) literature, 2) the decision support system literature, and 3) the model management system literature. We also propose the integration of several technologies that might help the modeler gain insight(s) from the analysis of multiple model instances. Then we report on preliminary tests of a prototype built using the architecture proposed in this paper. The paper concludes with a discussion of several research questions. Much of the previous MIS/DSS and management science research has focused on model formulation and solution. This paper posits that it is time to give more attention to enhancing model analysis.