Andreas Breiter and Karsten D. Wolf on AI in education

What does it mean for education when AI becomes an integral part of learning and teaching processes? Prof. Dr. Andreas Breiter and Prof. Dr. Karsten D. Wolf present the (interim) results of their current projects on AI in education.

As part of the conference of the EU project “AI Pioneers” organized by the Institute of Technology and Education at the University of Bremen, the final conference entitled “AI and the Future of Education” took place on September 22 and 23, 2025.

The conference focused on vocational education and training and addressed the following four questions:

• What research exists in the field of AI in VET and adult education?
• Which theoretical developments could be of interest for VET and adult education?
• What examples of best practice are recognizable in practice?
• What skills requirements arise from AI in the work environment?

In their keynote speech, Andreas Breiter and Karsten D. Wolf talked about theoretical considerations, the historical perspective, and the opportunities and risks involved in implementing AI-based tools in educational processes.

Andreas Breiter began with an overview of the historical paths of AI and emphasized that the hope of using AI (or other educational technologies) to solve fundamental challenges for teaching and learning processes (such as learning effectiveness, individualization, equal opportunities, etc.) was part of the narrative from the very beginning. As early as the 1960s, a “chatbot” called STUDENT was developed, whose rule-based system allowed users to “learn” geography at their own pace. This trend has continued to this day, partly due to the fact that cognitive scientists began studying the functioning of the brain and its simulation by computers at an early stage. The new wave of AI-based tools is particularly communicative AI, as is also being investigated in the DFG research group “Communicative AI.” What happens when AI-based systems become an integral part of teaching and learning-related communication and thus automate it? This is referred to as “hybrid figuration,” and the question arises as to how agency is then (re)distributed.

Karsten D. Wolf then used prototype implementations of various AI-based tools in a pipeline to demonstrate how common open-source systems can be used to provide learners with different, and ideally better, feedback. The basis for this was provided by explanatory videos created by the University of Twente (Prof. Dr. Kim Schildkamp) in the Netherlands as part of teacher training. After a complex pre-definition of the desired categories and an evaluation of the AI-generated output, it is possible to receive immediate feedback on an uploaded video – both in terms of content and presentation. Initial evaluations with users show that there are both positive and negative reactions: the real-time nature and quality of the feedback were appreciated, but at the same time, users questioned the reliability of the AI-generated feedback.

In the concluding discussion, questions ranged from “even more AI” for “better” education to critical questions about training data and the reliability of LLMs. Andreas Breiter and Karsten D. Wolf are in the midst of their joint research project and will continue to actively report on it at conferences.