The AI potential of model management and its central role in decision support

摘要

The paper stresses the general idea that ‘intelligence’ may be viewed to a great extent as the ability to model relevant parts of reality and to draw relevant conclusions from such models. Consequently, future software systems should be able to adequately handle a significant body of models for specific domains together with associated algorithmic tools. With respect to decision making and decision support, which require a high degree of cognitive sophistication, this leads to the quest for integrating into DSS results from model-oriented research in fields such as stochastics, statistics, decision theory, operations research and business applications. Based on such modelling capabilities, a DSS should be able to take a more active, normatively based role in aiding a decision maker. This kind of support requires strong interactive capabilities, driven by online computätional results and based on parallel problem exploration with partial models, incomplete information and robust solution methods. Additionally, such multi-level simultaneous use of a great number of interdependent models and associated algorithmic tools requires in itself an increased sophistication in model management. This should include a dynamic, performance-driven and adaptive use of the available algorithmic tools which actively addresses issues possibly overlooked by the user. In establishing this kind of sophistication, extensive use of available AI techniques will be in order. The paper tries to establish some guidelines for advanced system designs aiming at such sophisticated, highly integrative solutions.