This paper summarizes some of the current approaches to student modelling in Intelligent Tutoring Systems (ITS) and proposes ways in which learning algorithms can be applied in ITS. The role of machine learning in tutoring systems is described, including student modelling, teaching strategies, and collaborative learning. The PIXIE system and the Automated Cognitive Modeller system are examined as examples of systems using machine learning for student modelling. Dynamic student modelling is explored in the context of bounded user modelling technique, which not only identifies the student's knowledge at any one time, but also identifies the way in which the knowledge is acquired. Also discussed is the latest area of research in machine learning, namely, "explanation based learning" (EBL), in which explanations are constructed from domain knowledge and examples of the goal concept. The paper concludes with an outline for a learning algorithm for deriving concepts from subconcepts. (23 references) (DB)