Split Knowledge Level Modeling

Traditional approaches to building intelligent information systems employ an object model to define a representational structure for the information of interest within the target domain of the system. At runtime, the model provides a constrained template for the creation of the individual object instances and relationships that together define the state of the system at a given point in time. The ontology also provides a vocabulary for expressing domain knowledge typically in the form of rules (declarative knowledge) or methods (procedural knowledge). Agents operating within the system utilize the encoded knowledge to progress the state of the system towards the specific goals indicated by the users. While this approach has been very successful, it has some drawbacks, particularly in regards to the development of agent based decision support systems. Regardless of the implementation paradigm the knowledge applied by the agents is essentially buried in the code and therefore inaccessible to most domain experts. The knowledge also tends to be very domain specific and is not extensible at runtime. This paper describes the use of an explicit knowledge level within the ontology to mitigate the identified drawbacks while reducing both the number of classes and rules required.