The Rapport-Aligned Peer Tutor (RAPT) is an algebra tutoring virtual agent with the mission of becoming familiar and friendly teacher to the student.
Much of a student's success in academics is dependent on the individualized attention and time spent with them by teachers, instructors, and tutors. The more time a student spends on education, the more exposure and practice they receive; the more individualized, the more precisely a student can be challenged at the proper level and learn at the proper rate. Teachers, instructors, and tutors however cost a lot. It is impossible for every student to get the individualized learning experience that they deserve.
RAPT's goal is to provide an individualized tutor for each student who can get to know the student personally as well as cater material difficulty level. To that end, I worked on creating a machine learning framework that could allow for the RAPT tutor to quickly and efficiently create a user model for each student. The tutor can draw upon the on-topic (in this case Algebra) and off-topic models to increase their rapport with a student by calling upon past events, shared memories, etc. and use on-topic models to finely adjust the difficulty of the material presented to the student.
A writeup of my initial findings and implementation can be found here.