AI strategy in educational processes

Background and Problem

In autumn 2022, the release of language models based on artificial intelligence (AI) triggered a broad debate about the potential benefits and risks of AI, including in vocational education and training.

However, there is hardly any empirical evidence on this technology in the field of education, so most publications discuss potential developments based on assumptions or on rather abstract pedagogical considerations. On the other hand, looking at the state of the debate, a consensus can be observed that AI already has, and will increasingly have, a significant influence on occupational work and on teaching and learning in VET – but whether this influence is expected to manifest itself as a risk or as an opportunity often depends heavily on the positions and convictions of the respective authors.

Against this background, a transnational consortium (ES, SI, PT, DE) decided to carry out a project on the potential of AI guided by the following research question:

Research question

• To what extent can AI help to prevent drop-out in electronic learning (e-learning) through individualised support?

Approach

Drop-out rates in e-learning were very high. Among the various reasons for dropping out, there was one that could be influenced by (human or virtual) tutors: if learners got stuck at a certain point, tailored, individualised support could, where appropriate, put them back on a path conducive to learning.

For the project, an e-learning course on additive manufacturing was selected, among others. The target group were trainees in technical occupations. Participation was voluntary and completely anonymous; participants logged in on devices at their vocational training centres using functional e-mail accounts (‘User 1’).

The trainees could choose freely from nine modules, start their learning process at any point in the course and work on any number of modules. Each module took around one hour and concluded with a short multiple-choice test. Navigation within the modules was likewise left to the learners; there was a suggested sequence, but it was not necessary to follow this suggestion. Finally, each module offered additional optional materials, for example links to explanatory videos. In order to collect data that was as informative as possible, the e-learning modules were designed with the maximum possible number of degrees of freedom.

The learners' navigation patterns between and within the modules were documented via the internal tracking function of the learning management system (LMS) and handed over to an AI developed by the project partner for further analysis.

Results

Navigation patterns and dwell rates in the e-learning course that made it possible to detect potential drop-outs and to prevent them through individualised hints and tips.

Dissemination and Transfer Activities

ECER 2024 Nicosia

European final conference, 7 May 2025, Basque Country

German final conference, 28 May 2025, Bremen