Between Clones and Snow-Flakes: Personalization in Intelligent Tutoring Systems

摘要

This work improves intelligent tutoring systems by combining the benefits of online personalization of contents with methods that have strong non-personalized long-term optimized policies. Our hypothesis is that students are very diverse but they are not all completely different from each other. We will generalize previous algorithms by creating a new approach that (1) creates profiles of students based on historical data, (2) in real time is able to recognize the type of student that is being encountered, (3) personalizes their experience taking into account the information of similar students. We perform several simulations to study the impact on teaching of the amount of data, the diversity of students, and errors in the estimation of parameters.