This dissertation studies how information technologies, such as automatic question answering (QA), can add interactivity into a multimedia-based e-learning system and turn it into a "virtual mentor" to provide students with interactive, one-on-one instruction. It also explores the key factors of making such a "virtual mentor" as effective as a real mentor.; Based on a review of multiple learning theories and technologies, an exploratory model for studying the effectiveness of interactive e-learning, named "Learning with Virtual Mentors (LVM)", is proposed and a prototype system is developed to implement the LVM model. A series of studies, including controlled experiments and surveys have been conducted to explore the relationships among the core constructs of the LVM model: learning phases, system interactivity, learner characteristics, learning activity and learning outcomes.; Findings indicate that learning phases and some learner characteristics such as learning style affect students' behaviors, performance, and perceptions in e-learning partly as we expected. Furthermore, the virtual interaction impacts student behaviors, encouraging students to interact more and increasing student satisfaction with the learning process. However, the correlation between virtual interaction and actual learning performance is limited. Consequently, the LVM model needs to be further explored and developed. |