Model-based training of situated skills

Situated knowledge includes contextualised action and decontextualised description of knowledge. This is a significant distinction for developers of intelligent computer-aided learning (ICAL) applications. Most work in ICAL has attempted to decontextualise skills through descriptions such as rules and procedures because these are better understood and more suited to presentation through existing technology than the skills themselves. Here we discuss skills that are resistant to being decontextualised, and examine how techniques (model-based training) developed to teach decontextualised descriptions of knowledge can be used for training in communication and metacognitive skills. To support this approach to learning an ICAL needs a way to describe models of a problem in many different ways, and a way to explore models to learn about their underlying assumptions and their relationships to the problem context. A vocabulary for describing models, which serves as a cognitive tool for practising situated skills, and a model-based training system that supports this approach to learning are described.