Hybrides Lernen mit KI: Ein Chatbot für den Unterricht

Vocational education faces the challenge of integrating modern technologies into teaching in a meaningful way, especially where students need individual support that goes beyond mere classroom attendance. Against this backdrop, BBS Burgdorf came up with the idea of combining digital learning support with the expert knowledge of teachers using artificial intelligence.

How can AI support teaching?

The goal of BBS Burgdorf was to develop an AI-based chatbot that would be available to students as a supplementary resource outside of class. Instead of relying on internet sources or social media for answers to their questions, learners would have access to a reliable tool that provides well-founded answers based on the school's teaching materials. The chatbot was to be used in particular in the one-year “BFS dual” educational program, in which, after an orientation phase, a decision is made as to whether students will be prepared for a vocational college level or a technical college entrance qualification. It is during this phase in particular that the need for individual support is especially high.

An AI-supported assistance system for learners

The project was launched in close cooperation with DAISEC for technical implementation. The first step was to design an AI chatbot that can answer questions from students based on pedagogically verified content provided by teachers. What makes this project unique is that all language models used were operated locally and in compliance with data protection regulations via infrastructures in Lower Saxony, in order to ensure the protection of underage users in particular.

Initially, the prototype ran on the DAISEC server structure at the L3S Research Center, but was later moved to the scalable KISSKIinfrastructure. This solution not only enables data protection-compliant use, but is also designed for a growing number of users—up to 120 in the pilot project.

Technically, the system consists of three components:

Language model: The central model interacts with students in natural language.
Security filter: An upstream model checks all requests for security-critical content in order to prevent abuse, for example.
RAG-Pipeline (Retrieval-Augmented Generation): The system draws on a database of embedded teaching materials provided by teachers. This enables the chatbot to give context-relevant answers without having to “make up” content.

This system combines modern speech processing with controlled access to educational expertise, forming a highly practical prototype for AI-supported learning assistance in vocational schools.

A model for further schools

The collaboration between BBS Burgdorf and DAISEC demonstrates how innovative AI technologies can be used responsibly in education. The chatbot that has been developed offers students low-threshold, quality-assured support in the learning process while also reducing the workload for teachers.

The experience gained from the project now serves as a basis for further development. Initial discussions with other vocational schools, for example in the field of vehicle painting, are already underway. The aim is to transfer the existing system to other disciplines and educational programs in a modular fashion by adapting the underlying knowledge databases.

In the long term, the concept could be further developed and expanded in a multi-year research project. DAISEC actively supports this process and sees great potential in the approach for integrating AI into the education system in a sustainable, data protection-compliant, and practical manner.