Abstract
In order to support human learning, knowledge representation in intelligent tutoring systems must support: a constructivist approach to knowledge formation, dynamic additions and deletions of knowledge, the ability to reuse knowledge in multiple contexts, the ability to represent abstractions, and the ability to represent expertise. In order to support these requirements, we propose a network architecture that is fundamentally different from both neural and semantic networks [1]. It is comprised of two fundamental elements: nodes and links. A node represents a fundamental knowledge building block, or knowledge quantum, or a certain abstraction that is fundamentally derived to represent a collection of knowledge quanta. Nodes can belong to one of possible abstraction layers. Links relate nodes within a given layer or in different layers. A fundamental principle underlying our network architecture is that of local connectionism: each node is only aware of its rules of interactions with possible neighboring nodes, whether these nodes lie within the same layer or in other layers. The creation of inter-layer and intra-layer links leading to the formation of the knowledge base must be achieved using learning algorithms that lend themselves to a local connectionism approach, such as spreading activation [2]. Once knowledge is rooted on such a fundamental premise, we believe that it can be readily applied to multiple learning and disciplinary contexts. Fundamental issues that are being addressed by our research include those pertinent to fundamental knowledge representation, knowledge abstraction as well as the nature and development of expertise.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have