Agent simulation for contents evaluation with user models for hierarchical knowledge relations

To construct an appropriate e-learning system, it is inevitable to evaluate it by testing it with real human learners, but it naturally implies some sacrifices on the part of learners. To minimize the risk, some computational simulative evaluations are desirable. In this paper, we propose a multi-layered multi-agent simulation model for learner´s growth through using an e-learning system. As a typical example of complex learning objects, we employ the task of learning presentation skills for which we are constructing a comprehensive learner support system and learners´ data, and the knowledge model and learner growth model are prototypically proposed. By a preparatory experiment, our attempt, though it is still in its fancy, is shown to be on the right track.