Attempting Self-Determined Language Education

Inspired by the principles of Montessori education I am conceptualizing and developing a software application for self-determined language learning that uses computational methods to liberate users from the constrained practices of conventional education still found in many modern learning environments today.

Instead of having to learn a language in a strictly linear way, the application allows learners to aquire a foreign language on their own terms: Using the app, users are constantly given choices to determine their personal learning path. To enable this, I designed algorithms that ensure no matter what choices are being made, the structure of the curriculum and the learning content of lessons constantly adapt accordingly. This creates a dynamic learning experience where learners are free to explore, yet new lessons always coherently connect to what has been learned before.

This is a sneak peek into an unfinished and ongoing research project. At present, it takes the form of an experimental software platform for the development of computational methods for non-linear language learning. Usability and user experience questions are only barely adressed at this point.

I started this project before the release of ChatGPT and the succeeding wide availability of powerful large language models. Current developments of AI now enable entirely new ways of exploring digital language learning. Hence I am currently investigating a tighter integration of AI into the core of the application, to improve personalization and guidance for learners, and to explore new ways of learning a language.

Year: 2021

Keywords: AI Natural Language Processsing, LLMs, Node.js, Frontend Javascript

Main interface and dynamic curriculum: At any given point, not only one, but multiple options exist to continue the personal learning path. The available new lessons are selected by the application in evaluation of the entirety of lessons in the course against the individual skills the learner has aquired before. As the learner progresses, their recurring choices of what paths to take continuously generate a unique curriculum tailored to their interests.

Lesson overview: A lesson contains a defined set of words to be taught. On the basis of the skills the learner has acquired before in the app, sentences (used as teaching material) are algorithmically selected from an extensive database. It is ensured ensured that they are only made up of words the learner understands. This enables the generation of unique lessons tailored to topics of personal interest and the individual learning path of the learner. Simply put, every learner experiences a unique learning journey and develops different skills through excerting control over their learning process.

Learning session: In various kinds of learning tasks, the previously selected sentences are used as learning material to teach the subject of the lesson. A context-aware guide on the side can be consulted either for support with the given task, or in case the learner wants to further explore a certain aspect of the subject matter.

Easy Multi-Language Support

This video of an older version of the application demonstrates how due to its procedural and data-driven nature, courses in different languages can be easily created: