LearnLab: Adaptive Data Driven Learning Space

Screenshot of the program 'Karel'

The LearnLab student project aims to port a popular open-source programming learning environment called Karel to the browser and optimize the learning curve using usage data. The aim is to measure the learning speed and motivation of users throughout the course. This makes it possible to identify problematic lessons and iteratively revise them, for example by making overly difficult lessons easier. Furthermore, the curriculum could later be automatically adapted for users if their learning speed deviates significantly from the average in certain areas.

The project aims to address the shortage of skilled workers in the field of computer science and meet the need for basic programming skills in various disciplines. Learning programming can be challenging for beginners, and the lack of teachers makes the learning process even more difficult. The project offers an innovative way to design evidence-based learning materials while adapting them to the individual abilities of users. By adjusting the level of difficulty to the learning speed and prior knowledge of the learners, the amount of supervision required can be reduced and a motivating learning experience can be guaranteed.

The challenges here are both content-related and technical in nature. Meaningful metrics must be defined and their significance validated, and the learning application must also be adapted to our purposes from a didactic perspective. A not insignificant part of the work is therefore the design of learning content. At the same time, a software application is created, which in itself brings with it many challenges. The learning application itself must work, the data must be sent to a server and then evaluated. Later on in the project, the evaluation process may be automated. The data will initially be analyzed using frequentist statistical methods, for example to check for significant deviations from the average learning speed. In the future, however, machine learning is also conceivable as a method for developing a more complex recommendation algorithm for suitable curriculum content.

And here the group in their own words (German):