Optimizing Human Learning using Reinforcement Learning

Education is a field greatly impacted by the digital revolution. Online courses and MOOCs give access to education to most parts of the world, and many assessments are made online as they are easier to evaluate. This creates an important collection of learning analytics that can be used to provide and generate personalized content, which is essential to keep learners engaged and to have increased learning gains. The purpose of this thesis is to see how machine learning algorithms can be used to learn better knowledge representations of learners, and consequently to recommend learning tasks (exercises or courses) tailored to a student’s needs. We are learning instructional policies from student data so that we can understand how students learn and which lessons/exercises in a course have a strong impact on learning for which students.