Autonomously acquiring declarative and procedural knowledge for ICAT systems

The construction of Intelligent Computer Aided Training (ICAT) systems is critically dependent on the ability to define and encode knowledge. This knowledge engineering effort can be broadly divided into two categories: domain knowledge and expert or task knowledge. Domain knowledge refers to the physical environment or system with which the expert interacts. Expert knowledge consists of the set of procedures and heuristics employed by the expert in performing their task. Both these areas are a significant bottleneck in the acquisition of knowledge for ICAT systems. This paper presents a research project in the area of autonomous knowledge acquisition using a passive observation concept. The system observes an expert and then generalizes the observations into production rules representing the domain expert's knowledge.