Personalized or one-to-one instruction holds great potential for improving the state of education. But due to lack of good teachers this goal can only be approximated with the help of new and innovative computational technology. The research reported in this dissertation attempts to improve the state of education by aiming to support personalized student interaction with online learning environments in formal and informal educational settings by developing methods and techniques that are scalable, domain neutral, and can be used with open ended student inquiries. In particular this research focuses on conceptual personalization of interaction within online learning environments based on student and domain concept maps which are used as the semi-formal knowledge representation scheme. To address the conceptual personalization agenda three activities were carried out during the course of this research: (1) Assessment study, (2) Algorithm design implementation, and (3) Adaptive learning environments prototyping. In the assessment study, cognitive process of human experts were analyzed to understand operations used by human when assessing student understanding by comparing a student concept map to the domain concept map. The synthesized operations were then used to guide the development of algorithms that can automatically assess student conceptual understanding from their concept map. Lastly, the results from the automatic student assessment were embedded into task specific adaptive learning environments.