Artificial Intelligence (AI) systems significantly accelerate the speed of discovering nuanced patterns in data, making complex decisions, and executing repetitive yet critical tasks. However, these AI systems rely on Machine Learning (ML) algorithms that often fail to fully capture end user context. When deployed in real-world applications, these AI systems are used as black boxes by non-experts who have limited understanding to interpret or influence system outcomes. These limitations make it challenging for users to adopt AI systems within their complex workflows. In my doctoral research, I address these challenges by studying, building, and deploying systems that incorporate Machine Teaching (MT); an emerging paradigm that helps users integrate their own knowledge into the AI system through a human-machine dialogue, much in the way that a teacher works with a learner. In this dissertation, I propose a human-centered framework of Machine Teaching, and describe how it can enable AI non-experts to 1) interactively transfer relevant knowledge to an AI system, 2) actively engage in the teaching process and 3) refine machine learners to suit their workflow. I systematically investigate the impact of human factors, sensemaking, conceptual models and human values on the teaching process, experience and outcome.These systems enable AI non-experts like journalists, data-story writers, and healthcare professionals, to effectively transfer their domain expertise to the AI system, critically evaluate AI performance, and develop strategies to improve their teaching outcome. This dissertation directly contributes to the theoretical and practical knowledge for designing MT tools for AI non-experts and building intelligent, context-aware AI-based Information Systems that positively impact human decision-making on complex tasks. My research helps to solve the broader challenges emerging in the field of Human Computer Interaction (HCI) pertaining to usability, reliability and sustainability of AI systems.