Data-Driven Probabilistic Framework for Student Learning

Understanding the cognitive and behavioral aspects of student learning in a principled manner can enable educators and psychologists to improve the state of education. With this in mind, we propose a novel data-driven probabilistic framework to model student learning over a period of time. Our framework provides a means to quantify and track student learning as a function of critical factors such as student skill level, quality of instruction and the amount of prerequisite understanding. We evaluate our proposed model on a real dataset of student responses and show that it achieves good accuracy in predicting responses for previously unseen questions.