This project examined how AI can support university teachers in a teaching qualification course and master students in a Technology Enhanced Curricula course when designing STEM (Science, Technology, Engineering, and Mathematics) learning activities. Participants used AI tools to brainstorm ideas, structure activities, align learning objectives with activities and assessment tools, and generate materials such as assessment rubrics, while reflecting on ethical, didactical, and privacy considerations. They viewed AI as a helpful support system and creative partner for preparing class activities, yet stressed that it cannot replace the pedagogical judgment and subject expertise required for accurate, responsible decisions in education design.

Background information

The project was initiated due to the urgent challenge and concurrent opportunity presented by the integration of GenAI into education. Educators needed practical guidance on how to responsibly design didactically sound and meaningful learning activities given GenAI’s capabilities. The initial context was the lack of clear insight into what can and cannot be done with GenAI in pedagogical design, especially concerning the critical need to align its use with regulations, policies, ethical standards, and safety. The main goals were to:

• Explore GenAI’s potential to support both experienced university staff and prospective teachers in designing STEM learning activities.
• Gain insight into critical perspectives and ethical considerations (e.g., regarding privacy issues, environmental impact, biases, misinformation) regarding the use of AI in education design by engaging in a series of hands-on, collaborative activities.

The intervention

Our 1.45-hour intervention was delivered to 5 master-level students and about 12 university-teaching-qualification participants and focused on: integrating generative AI into course and lecture design, effective implementation and best practices, and key challenges, critical considerations, and future directions (. It introduced instructional strategies, learning theories and educational design based on key academic literature for STEM education design during the lecture and links for further reading. These included Constructive Alignment (Biggs, 1996), Constructivism and Constructionism (Ackermann, 2001)), and best practices for developing didactically sound, meaningful learning activities, with concrete learning goals and assessment methods. Generative AI (GenAI) was intentionally used as a meta-cognitive and reflective evaluation tool and not as a tool to generate outputs that could be uncritically used in design. Participants leveraged GenAI outputs to brainstorm and critique their own design choices (e.g., prompting GenAI to analyze the alignment of their planned assessment with their objectives), or generate ‘foundational critiques’ on which they based their refinements on their developed learning activities, or parts of those. The design rationale to use GenAI as an evaluative/reflective tool in our intervention was about maintaining human agency and critical thinking while adhering to best practices and ethical regulations regarding AI use in teaching and learning.

The results

The outcomes strongly validated the concept of hybrid intelligence, which is a combination of human and artificial intelligence, showing GenAI acts as a powerful supportive tool, not a replacement for teacher expertise. Even though most participants used AI systems before for various tasks, they reported learning gains regarding using AI for specific, refinement and time-saving education design tasks: brainstorming ideas for designing learning activities, structuring lesson components (e.g. learning aims, materials, procedure), and adjusting language in terms of accessibility and inclusivity. This efficiency allowed teachers to dedicate more time to pedagogical refinement (e.g. refining activities for diverse student needs).

Some teachers were resistant in the use of GenAI due to ethical reasons and did not mind spending longer time on more ethical education design in terms of environmental and societal impact.

Additionally, challenges arose at the initial design stage where GenAI often produced generic suggestions, failing to align with individual teaching styles or specific contextual needs. Consequently, teachers stressed that they would not trust AI to generate final learning material. Instead, they leveraged its output as a foundation, for example, generating rubrics or detailed feedback which they then critically refined and customized based on their professional judgment and ensure it is aligned with their educational contexts. For example, the AI system might provide elements in the rubric that were not applicable for the specific context or generate too vague items that were too difficult to assess.

These results were not surprising, as they reinforced the intervention core hypothesis: GenAI might be a valuable tool as a metacognitive aid that enhances educators’ analysis, reflection, and iterative design. It may inform judgment, not replace it. Overreliance would risk superficial thinking and diminish professional autonomy but when used critically, GenAI can strengthen and not substitute the expertise of teachers.

The next step

Building on our initial design for our intervention, our next phase focuses on formalizing and expanding reflective GenAI practices across our teaching community. The reflective character of our intervention (during hands on activities, assignments and plenary reflection in the end) created many opportunities for refinements on the material, especially for more criticality when using GenAI in similar tasks. For example, we expanded the slides and readings for critical considerations when using GenAI, and now introduced them earlier on, before participants engage with GenAI activities. Other refinements include updated content on prompt engineering, UU and EU policies, responsible data use, and critical engagement with GenAI’s societal and environmental implications.

We hope that these refinements will further support educators and prospective teachers aligning AI-generated outputs with their pedagogical intent, moving beyond generic responses to foster creativity, efficiency, and integrity in teaching practice. This way, they can use GenAI to generate initial ideas or critiques on their designs, but the emphasis will remain on human judgment, encouraging them to question, refine, and ethically evaluate AI output contributions and their impact.

We hope that this approach will not diminish academic agency or responsibility. The core challenge is sustaining genuine critical engagement as GenAI use grows. More resources and earlier guidance help, but they don’t ensure educators will consistently question AI outputs or avoid overreliance. Preserving agency, reflective habits, and pedagogical intent amid escalating automation remains difficult.