Schema theory-based computational approach to support children's conceptual understanding
Researchers acknowledge the difficulty faced by children in understanding new concepts. Explaining new concepts to children requires supporting their reasoning based on concrete objects and ideas. Human tutors normally use some dialogue to introduce new concepts and tailor the explanations to the prior knowledge of the children. There is a lack of interactive pedagogical agents that guide children's reasoning and adapt explanation to their cognitive state. The design of such agents can be based on learning theories that explain how children understand new concepts, as well as on studies of how human teachers support children's conceptual understanding. The goal of this research is to develop a computational framework to inform the design of a pedagogical agent capable of engaging in a dialogue that supports children's conceptual understanding, the thesis proposes an approach for Schema Activation and Interpersonal Communications (SAIC) to support cognitive tasks that occur when a child is learning new concepts through one-to-one interaction with a computer agent. The approach is based on schema theory that explains how meaning-making occurs and stresses the importance of prior knowledge, and on the results of an experimental study that identifies strategies human teachers use to support schema-based cognitive tasks. A novel architecture of a pedagogical agent whose behaviour is based on schema activation and modification is described. The architecture addresses three important issues: describing the process of activation and modification of relevant prior knowledge to be used in introducing new concepts; defining the reasoning and decision making of the agent to promote schema-based cognitive tasks; and providing adaptive explanations tailored to the child's relevant prior knowledge. The schematic knowledge of the SAIC agent is represented as frames, the dialogue is planned as a sequence of dialogue games, and the interaction language is implemented with linguistic templates extracted from a study with teachers. The applicability of the SAIC approach is demonstrated in a multimedia educational system 'Going to the Moon', as an integrated as an integrated component in a reading session. An experimental study with the multimedia system has validated the SAIC design approach and has examined the usefulness of the agent in supporting children's conceptual understanding in terms of improving their schematic knowledge. The thesis makes original contributions to the fields of Artificial Intelligence in Education by defining reasoning and decision making based on the principles of schema theory, and by designing a schema-based pedagogical agent to support children's conceptual understanding; Education by demonstrating the application of learning theories to inform the design of intelligent tutoring systems; and Knowledge-based systems by demonstrating the feasibility of frames as the representation formalism in Intelligent Tutoring Systems, and by proposing some original mechanism for using frames to design pedagogical agents.