Towards AI-Powered Data-Driven Education

Educational platforms are increasingly becoming AI-driven. Besides providing a wide range of course filtering options, personalized recommendations of learning material and teachers are driving today's research. While accuracy plays a major role in evaluating those recommendations, many factors must be considered including learner retention, throughput, upskilling ability, equity of learning opportunities, and satisfaction. This creates a tension between learner-centered and platform-centered approaches. I will describe research at the intersection of data-driven recommendations and education theory. This includes multi-objective algorithms that leverage collaboration and affinity in peer learning, studying the impact of learning strategies on platforms and people, and automating the generation of sequences of courses. The paper ends with a discussion of the central role data management systems could play in enabling modern online education.