15.04.2026
AI in schools: From solution generator to digital learning coach
A new policy paper from the Hector Research Institute of Education Sciences and Psychology examines the interplay between artificial intelligence and self-regulated learning. The key message for education policymakers is that AI must not replace learning, but must instead foster pupils’ agency.
Artificial intelligence, particularly generative systems such as ChatGPT, has long since become part of everyday school life. A team of researchers led by Dr Tim Fütterer highlights the technology’s potential to make education more personalised, but warns of the risks of using it in a purely results-oriented manner. There is an urgent need for action within the education system.
Performance vs. sustainable learning
So far, many young people have mainly used AI as a ‘service provider’ to find information quickly or complete their homework. However, this focus on pure performance (the rapid generation of results) often comes at the expense of actual learning.
The consequences:
De-skilling: Existing skills may atrophy as tasks are outsourced to AI.
Skill-skipping: In some cases, basic skills may not be acquired in the first place.
Superficial understanding: Studies show that unreflective use of AI leads to reduced cognitive activity and poorer memory performance.
Self-regulation as a key skill
The authors emphasise that self-regulated learning – that is, the ability to independently plan, monitor and reflect on one’s own learning process – is becoming a crucial skill for the future in the age of AI. AI should therefore not be used as a ‘shortcut’ to a ready-made solution, but rather as a learning coach that prompts reflection and teaches strategies.
Three scenarios for the use of AI
The paper identifies three ways in which AI can be used in the classroom:
1. AI as a service provider: The AI performs the tasks; the learning outcome remains limited.
2. AI as a teaching assistant: Provides adaptive support in the acquisition of subject knowledge.
3. AI as a learning coach: Specifically promotes learners’ metacognitive skills and personal responsibility.
Recommendations for practice
To ensure a successful transition from dependence to ‘agency’ (the ability to act), the research team sees an urgent need for action at several levels:
- Pupils must learn to use AI as a sparring partner that they can critically scrutinise.
- Teachers require further training to orchestrate AI in a way that simultaneously strengthens subject knowledge and self-regulation skills.
- Policy-makers and industry are called upon to promote evidence-based and quality-assured educational AI tools that follow pedagogical principles rather than pure efficiency.
This policy paper is based on an analysis of the latest empirical research into the use of AI in an educational context, as well as the systematic incorporation of the perspectives of over 100 national and international experts from academia, educational practice, educational administration and the education sector.
The policy paper was funded by the Vodafone Foundation Germany.
Publication:
Fütterer, T., Steinhäuser, R., Udvardi-Lakos, N., Fabian, A., Gerjets, P., Nuxoll, F., Bock, C., & Trautwein, U. (2026). KI in der Bildung: Von Abhängigkeit zu Agency! Künstliche Intelligenz und selbstreguliertes Lernen von Schülerinnen und Schülern (Policy Paper). Hector-Institut für Empirische Bildungsforschung der Universität Tübingen / Vodafone Stiftung Deutschland.
Further interview with Dr Tim Fütterer in German:
https://evido-magazin.de/artikel/selbstreguliertes-lernen-und-ki-zwischen-effizienter-assistenz-und-folgenreicher-abhaengigkeit
Link to the study:
https://link.springer.com/article/10.1007/s10648-026-10133-8
Press contact:
Kristina Laube
presse@lead.uni-tuebingen.de
Share
