A new task for expert system analysis libraries: the decision task and the HM method

摘要

As a contribution to the search for recurrent abstractions and their formulation in terms of generic tasks, we describe a generic decision task and its decomposition in accordance with the Heuristic Multi-Attribute (HM) method, consisting of the assessment of a set of alternatives in terms of the degree of satisfaction of a set of objectives. This task is not found in the usual KADS libraries of Analysis, Modification and Synthesis tasks. However, as well as being of well-known applicability in fields such as economics or medicine, it is the basis of the intelligent agent model, as introduced by Newell, the reference for many computational paradigms and currently the focus of much of attention in artificial intelligence research. In the analysis of this task we emphasise the importance, from a methodological perspective, of giving precise and implementation-oriented descriptions as well as using a more rigorous approach to the modelling of domain ontologies and domain-specific data than is customary. To this aim we make extensive use of the semi-formal analysis notations CML and UML to describe, respectively, the inference and task level, and the domain level. Of particular note is the use we make of these notations for modelling learning (via UML metaclasses). We produce a semi-formal description of classes which are already implementable and are easily extendible via the inheritance mechanism.