Optimizing personalized learning

Optimizing personalized learning is a huge challenge. With e.g. 1400+ schools using EduLab’s solution there is an immense potential in analyzing the data created by users of EduLab’s website. While existing recommender systems are focused on “one-shot” recommendations, learning mathematics is a long process with many steps. Recommending a personal path through the material and adjusting this path depending on the student’s progress is an important research challenge.

Focus has been on how to optimize students’ mathematical learning in primary school. The solution is an algorithm that gets input from EduLab’s content and relevant user behavior and where the output is "where to go from here". The first algorithm will soon be employed in EduLab’s system.

After formalizing the SuperTrainer's measurement, we model the SuperTrainer's problem, i.e. recommending the right track of questions as a contextual bandit problem. Assume multiple different ways of serving the questions to the students, the correct track of questions can be seen as choosing a learning style that fits the students best, thus increasing engagement and learning outcome. The goal of the algorithm is to find the most suitable learning style at the correct time for a student, with the learning style being chosen based directly on a student’s actual usage of the system.

The project generates knowledge about optimizing personalized learning, in particular with respect to the process of understanding mathematics at primary school level. By trying out various solutions, we are improving the understanding of how to model personalized learning as an algorithmic problem.

Products using such solutions offer a leap in state-of-the-art learning and address key customer demands and will release part of the resources, which today are spent on homework cafés and expensive private education. The solution expands EduLab’s existing platform and is thus creating value for the company.

This project is a joint project between Department of Computer Science, University of Copenhagen and EduLab.