Using Machine Learning to Advise a Student Model

* Final gross prices may vary according to local VAT.

Get Access


Human learning is complex, dynamic, and non-monotonic. Currently it cannot be accurately modeled or measured, and present-day student models are too simplistic and too static to reason effectively about it. This paper explores several machine learning mechanisms which might enhance the functionality of a student model. Human learning experiments are described demonstrating the spontaneous nature of learning, for which action-oriented student model components are needed. An existing student model, built as part of a physics tutoring system, is described which begins to handle non-monotonic reasoning, makes little commitment to a static model of student knowledge, and uses a Multi-layered representation of inferences about student knowledge. The paper asks how a learning mechanism might inform such a student model and represent the dynamicism and spontaneity of human learning.