AI-driven personalized learning paths: the National Project of Estonia

Sander Tars
October 21st, 2021

student learning at home

This project aims to implement AI-driven solutions to personalize students’ learning paths through using their and other students’ data points created throughout their studies.

It has been calculated that the cost of students not advancing into specialized or higher education levels costs Estonia 1.4% in GDP (Centar 2011). 

Students study in different ways and at different paces. However, they are distributed in classes, courses, etc. mainly according to age and taught using largely identical learning paths. This generalized approach does not suit a considerable number of students. With 33% of students claiming lack of study motivation, the way to solving the issue lies in personalizing studies based on students’ strengths and interests.

Unfortunately, the traditional way of personalizing means a heavy additional workload for teachers and is thus not a feasible solution. Therefore, we explore AI-driven methods to personalize students’ studies while minimizing teachers’ workload.

Learning path project goals

The Main goals of the learning path project are to:

These goals are the first stage of a longer framework and are designed to be extended into more study scenarios.

Student’s helper

personalized education path

The aim of the student’s helper is to suggest next study items to students based on their study profile to fulfil student’s goals in the best manner.


The ML model suggests the study items based on maximizing the expected study results for a specific task and student profiles.


Student’s helper is designed in a similar fashion to chat window (e.g Facebook Messenger) to feel familiar for students to use and to encourage students to take more charge and responsibility of their own studies.


Teachers’ helper

Teacher’s helper gives the teacher an overview of their students’ progression, planned learning path, and suggests which students need what kind of help at the stage they are in their studies.

The ML models and the process is designed such that it minimizes the workload of the teacher in the personalization process. Below is a sample of what the designed Teacher’s helper looks like as an app.

personalzied learning ui

Next steps

The data-based experiments have shown promising results and thus the AI-based Students’ and Teachers’ helpers will be tested on real students and teachers in the real study process. 

Student’s and Teacher’s helper are the first steppingstones in transforming the Estonian education system into a highly personalized learning ecosystem. This cannot be achieved solely with technical solutions but needs to involve materials, teachers, students, and all the supporting infrastructure. This means that in addition to AI models there is extensive analyst work to be done in developing the learning ecosystem to ensure its ease of use and attractiveness – a process which has already been started in designing Students’ and Teachers’ helpers and which will be carried on into upcoming projects.

Takeaway notes

When undertaking such projects, like AI in the public sector, or implementing AI in education pay special attention to the following points:

  • Consult with the external AI development company to make sure that the data you have is sufficient. Some additional data likely needs to be collected and some data collection principles changed. Therefore, it is likely that data processes can take up to 6 months until the data is fully usable.
  • Have a couple of thorough goal-setting workshops with the external ML partner to make sure that the end-goals are feasible ML-wise.  The initial ideas will likely undergo some changes due to technological limitations.
  • Make sure that teachers and students are well included in the process. They are the ones that will be using and benefiting from the solutions the most after all!
  • Have a set of partner schools and classes ready to live test the applications.
  • Have a clear roadmap where you want the development process to take you (and be prepared to alter the roadmap when necessary!)

ai plan execution