Knowledge Representation in Structural Learning Theory and Relationships to Adaptive Learning and Tutoring Systems
Joseph M. Scandura
This article summarizes major steps in the evolution of Structural Learning Theory (SLT), a comprehensive, parsimonious, precise and operationally defined theory of complex human behavior. SLT covers knowledge representation, methods for constructing same, cognitive processes, knowledge assessment and interactions with external agents (e.g., teachers). The article emphasizes major advances in recent years that make full automation possible. It details and illustrates: a) ill-defined as well as well-defined knowledge, both represented in terms of SLT rules consisting of structural and procedural Abstract Syntax Trees (ASTs), b) how SLT rules can be represented at arbitrary levels of detail and how the higher as well as lower order SLT rules needed to master any given problem domain can be constructed systematically, c) cognitive mechanisms, including empirical data associated with a Universal Control Mechanism (UCM), which controls the use (and acquisition) of all SLT rules, subject only to a fixed processing capacity and speed constraints characteristic of individuals, d) how the knowledge available to any given individual (behavior potential) is operationally defined relative to SLT rules and e) theoretical, empirical and practical implications for building automated tutoring systems. The concluding section shows why the theory makes a difference, how it can be tested and what this implies for building e-learning, intelligent and other advanced tutoring systems.