Playbook EN Backup

0. About the DUAL.AI.TEACHer Playbook

Playbook Objectives

This playbook is aimed at teachers, school principals and university lectures who wish to acquire the necessary knowledge about using Artificial Intelligence (AI) in schools and about changing their institutions for (better) AI use. AI in school and in teacher education could be used for a variety of purposes. With regard to teaching, we distinguish ‘teaching about AI’ and ‘teaching with AI’ to impart subject-specific knowledge. 

The development of Artificial Intelligence is advancing at an extremely rapid pace, and its use is increasingly permeating our society, entering aspects of everyday life as well as education. This pervasiveness brings with it multiple and often interconnected implications, spanning ethical, social, economic, legal, and cultural dimensions, touching on issues such as data protection, fairness and non-discrimination, algorithmic transparency, technical safety, and accountability. Precisely because of the speed of this diffusion and the scale of its consequences, there is a growing need for shared guidelines, regulatory frameworks, and governance tools capable of steering technological development in a responsible and sustainable direction.

Playbook Structure

To address all these aspects, we have structured the handbook as follows:

How to Navigate the Playbook

Gemini_Generated_Image_t1spb4t1spb4t1sp.jpg

Fig. 1 "The pathways to expertise" (AI-generated image, Google Gemini)

For Teachers

There are different ways you can use the Playbook. 

Option 1: You’re looking for ideas for your next school day to teach students about AI or to use AI to teach subject-specific content.

In this case, go directly to Chapter 2, where we’re continuously adding new best practices. We hope you’ll find what you’re looking for. If not, please email us so we can develop a new resource to address the gap you’ve identified.

Option 2: You want to learn about the legal framework governing the use of AI.

In this case, go to Chapter 1.1.

Option 3: You would like to gain the qualifications needed to teach about AI effectively and incorporate AI into your subject-specific instruction. 

In this case, we invite you to take our initial test and then follow our qualification modules. By earning micro-credentials, you can pursue comprehensive professional development tailored to your expertise.

For School Administrators 

Are you ready to set out with your school on an exciting journey to bring AI into your professional world? In Chapter 5, we chart a clear roadmap to guide you every step of the way. Our qualification programme empowers you to equip your colleagues for the adventure that lies ahead.

For University Faculty

AI is making its mark on teacher education, sometimes in bold strokes, sometimes in subtle ways. Viewing teacher education as a journey that begins at university and builds over time, Chapter 5 offers you a roadmap to inspire and support student teachers as they explore the evolving landscape of AI.

1. AI in Education

1. AI in Education

1.1 Potentials & Risks

If you remember only one sentence form this page: AI has made its way into schools like a medicine that comes without a patient information leaflet. As a teacher, you don’t know when to use AI, you lack dosage instructions, and there is no information on the side effects. This chapter serves as the patient information leaflet for the Playbook.

Should AI come with a leaflet like a medicine?

You know that slip of paper tucked inside every medicine box? The one you likely toss aside without a glance? It may seem dull, but this little paper is a quiet triumph of the last century. That leaflet spells out four essentials: what the medicine treats, how much to take, the surprises it might bring, and who should steer clear.

Behind every line on that leaflet stands a whole machinery: clinical trials, regulators, and a responsibility to reveal mistakes. No one believes the leaflet itself heals you. Its real power is making the medicine safe for anyone. That is the magic we are missing when we let AI teach without guidance.

The AI tools now quietly settling into your school inherit a problem that is older than AI: educational technology has rarely come with instructions. In the UK, a review found that only 7 out of 100 education technology companies had put their products through a randomised trial, and only 12 out of 100 had sought outside certification. Schools rarely ask for more. In a survey across 17 American states, only 11 out of 100 teachers and administrators had requested peer-reviewed evidence before adopting a new tool.

The result, in many cases, is wasted money. In the US, on average two-thirds of education software licences went unused, and 98 out of 100 were never used intensively. The EdTech Genome Project, also in the US, looked at about 7,000 teaching tools worth 13 billion dollars and found that 85 out of 100 were either a poor fit or implemented incorrectly.

This chapter may look like the dry, necessary section, but it reveals the hidden stories behind the rules. It tells what really happened, who was affected, and shines a light on the side effects, even those that quietly reached children who never signed up.

Potentials of AI

Most studies on AI's potential focus on university students. School pupils have so far featured relatively rarely in these studies. And most studies relate to short-term interventions lasting from a few hours to a few weeks. Consequently, the so-called novelty effect may have positively influenced the results. A large review of 68 experiments found exactly this pattern. The shorter the study and the smaller the number of particpants, the larger was the reported benefit of AI.  

(1) Let AI explain a new topic and keep the practice for your lesson

Imagine this. Tomorrow, when a pupil misses a topic or needs that tricky concept explained yet again, let an AI tutor handle the first explanation. Reserve the practice for your lesson, where real learning comes alive.

Why this matters. In a well-controlled trial, an AI tutor taught new material at least as well as a good lesson. In a Harvard study, 194 physics students learned two topics. For one topic, they used an AI tutor developed for this purpose. For the other topic, they learned in a well-run active-learning class – with discussion and problem-solving in pairs, not a lecture. Each student completed both approaches, so no one can claim that the stronger group received the better treatment. In a test conducted immediately afterwards, the students taught by the AI achieved significantly higher results. The researchers recorded a median time of 49 minutes to complete the task, compared with the 60 minutes allocated for the lesson in the timetable. The authors are cautious in their interpretation of these results: the material was new to the students, and much of the task simply involved explaining it well. They explicitly decline to extend their result to tasks that require pulling several ideas together. And these were undergraduate students, not eighth-graders.

A large review pooling the results of 228 studies points the same way, with one important limit. Generative AI produced larger gains in knowing and understanding than older forms of AI such as intelligent tutoring systems and adaptive practice software. But this was the only one of the four outcomes the review measured where generative AI had the edge.

Go deeper. Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: An RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15, Article 17458. Free access, DOI 10.1038/s41598-025-97652-6. The article describes how the tutor was designed and controlled. This is useful for anyone wishing to develop an AI tutor themselves.

(2) Making better use of lesson planning time

Imagine this. Save some of the time you need for lesson planning by using AI to draft your first lesson plan and then revise it thoroughly.

Why this matters. No one disputes that teachers’ workloads are heavy. Therefore, one potential benefit of AI is using working hours more efficiently. A trial in England investigated whether targeted use of AI for lesson planning can reduce teachers’ workload. Researchers randomly assigned 259 science teachers at 68 schools either to a group that used ChatGPT together with an online guide to using it well, or to a group that was asked not to use generative AI at all. Teachers in the ChatGPT group spent about 56 minutes a week planning science lessons for Years 7 and 8 (ages 11 to 13), compared with about 81 minutes in the comparison group. A panel of subject experts assessed the quality of the lesson plans. They did not know whether the plans had been created with or without the aid of AI. The experts found no evidence to suggest that the quality of the lesson plans differed. However, two points must be highlighted. (1) The teachers used the tool less and less as the weeks went on. The evaluators read this as teachers moving from a learning phase into routine; it may also mean that the saving does not sustain itself without support. (2) The trial measured the time spent by teachers on lesson planning, not their pupils’ learning outcomes.

Voices from the field

"Everything comes back to time. Even teachers who see the value of AI rarely have the time learn how to use it."

Go deeper. Roy, P., Poet, H., Staunton, R., Aston, K., & Thomas, D. (2024, 12 December). ChatGPT in lesson preparation: A Teacher Choices Trial. NFER, commissioned by the Education Endowment Foundation and the Hg Foundation. Free at educationendowmentfoundation.org.uk.

(3) Providing better support for learners who do not benefit fully from standard lessons

Imagine this. Your school decides how it wishes to spend its funds. It wants to purchase an AI chatbot. How likely do you think it is that the chatbot providers will provide you with evidence of the bot’s benefits for different subjects?

Why this matters. The provider may point to research on learners with disabilities – one of the longest-studied fields of AI in education, and one that is easily overlooked when funds are allocated. In an analysis of 29 studies involving 41 groups of learners with disabilities, researchers found moderate benefits across a wide range of settings. But read the small print. In seven out of ten of these studies, ‘AI’ meant a robot, usually a humanoid one, often steered by a researcher out of the child’s sight. Only one in five studies used software. So this is no proof that a chatbot will help an autistic pupil in your class. Nor could the reviewers say what makes these interventions effective. Based on their research, they could not identify why the studies that worked differed from those that did not.

Go deeper. Zhang, L., Carter, R. A., Jr., Liu, Y., & Peng, P. (2026). Let's CHAT about artificial intelligence for students with disabilities: A systematic literature review and meta-analysis. Review of Educational Research, 96(1), 215–257. Free, DOI 10.3102/00346543241293424. The real gem is the table of included studies: it instantly reveals just how little of this field actually focuses on the tools currently being marketed to schools.

(4) Teaching learners how AI fails

Imagine this. Ask your class to create ten images of 'a doctor' and ten of 'a nurse.' Tally up who appears in each set, then spark a discussion: where did these patterns come from, and who made those choices?

Gemini_Generated_Image_mfaur1mfaur1mfau.jpg

Fig. 1: One shot prompt: a doctor, a nurse (Image generated with NanoBanana 2) 

Why this matters. The classic version of this lesson feels outdated, and that is exactly why it deserves attention. For years, the demonstration was predictable: ask for a doctor, and a man appears. In a test by the non-profit AlgorithmWatch, Midjourney almost always pictured doctors as white men, while Adobe Firefly made visible, if imperfect, attempts to diversify. A peer-reviewed study of 2,400 AI-generated images of surgeons found the same split between products: two of three popular generators showed almost no women, while the third came close to the real proportion of women in surgery.

Try the same prompt today, and you might see a female doctor or a male nurse appear (Fig. 1). But do not be fooled into thinking the bias has vanished. A 2026 analysis of 1,344 images from three popular generators revealed the opposite: when asked for a 'competent person,' one system showed no women.

What changed is not the bias but the layer it lives in – and that is the important lesson. Human decisions stand behind any of these pictures: what the system was trained on, and what its makers afterwards decided it should show. The second is adjustable, invisible, and made by a company.

The first layer of human decisions is what the system has absorbed. Millions of images, crafted and labelled by people, each one echoing the world as it already was. This is the layer that the classic bias lesson reveals, and pupils grasp it easily: the machine became a reflection of what it was fed.
The second layer is shaped by what a company chooses to show you next, yet this influence leaves no trace in the image itself. Three things unfold here, each one described by the creators in their own words.
So when a pupil types “a doctor” and a woman appears, the image offers no clue about which layer brought her into being. That uncertainty is not a flaw in the lesson. It is the lesson itself.
The old story claimed the machine simply soaked up society’s prejudices. The new story cuts deeper and lasts longer: A company made a choice! You cannot spot that choice in the outcome, and it might shift on a Tuesday, quietly, without warning. That truth endures through every new model. The stereotype lesson fades away.

This is the lesson that deserves a student's attention, and it will outlast the next model update: the image is not a simple reflection of reality, and it did not happen by chance. Someone made a choice.

Zeichenfläche 1-100.jpg

Fig. 2: Why the picture based on a prompt like "draw a doctor" looks like it does and why the picture itself won't tell you (Image generated with NanoBanana 2) 

This lesson can start early. When 209 Finnish students in 12 classes, grades 6 and 7, explored this topic, the number who could explain the bias using data jumped from about 7 out of 100 to 4 out of 100. The honest reservation: the study measured what children could explain, not what they would do later.

Voices from the field

"Bias is not unique to AI. Even textbooks can contain one. What matters is knowing about chance, about the data behind an answer & about possible distortions."

Go deeper. Vartiainen, H., Kahila, J., Tedre, M., López-Pernas, S., & Pope, N. (2025). Enhancing children's understanding of algorithmic biases in and with text-to-image generative AI. New Media & Society, 27(9), 5342–5368. Free, DOI 10.1177/14614448241252820. The study had no comparison group – read it as a promising classroom design, not as proof.

Weinmann, H., Messingschlager, T. V., & Appel, M. (2026). Gender bias in text-to-image generative artificial intelligence: Neglect and stereotypical presentations across three popular platforms. New Media & Society. Online first. Free, DOI 10.1177/14614448261435197.

Ali, R., et al. (2024). Demographic representation in 3 leading artificial intelligence text-to-image generators. JAMA Surgery, 159(1), 87–95. DOI 10.1001/jamasurg.2023.5695.

The vendors’ own statements: OpenAI (2022), Reducing bias and improving safety in DALL·E 2; Depue, W. (2023), What’s new with DALL·E 3?, OpenAI Cookbook; Raghavan, P. (2024), Gemini image generation got it wrong. We’ll do better, Google. 

How solid are the research results on the potential of AI at this very moment?

The most important thing. If someone quotes a study to you, ask two questions before agreeing: "Who were the learners, and how long did the study last?"

Why this matters. In 2025, a meta-analysis of 51 studies reported a large benefit of ChatGPT for students’ academic performance. The article was read hundreds of thousands of times and served as the basis for numerous confident statements in staff rooms. On 22 April 2026, the journal retracted the article. A retraction means that a journal officially withdraws something it has published; the article remains visible, marked with the note ‘RETRACTED’, so that anyone who has cited it can recognise this. Two researchers who had not been involved in the study had identified discrepancies in the analysis. The journal's editor wrote that the problems ‘undermine the editor’s confidence in the validity of the analysis’. The authors did not respond to correspondence on this matter.

This does not mean that AI has no impact. There are reputable studies that demonstrate moderate benefits. This field of research is also not yet well defined because it is relatively new and highly dynamic. It means that the figure that was frequently cited was incorrect, and that it took a year and two external experts to realise this.

Another reservation is worth mentioning. In mathematics, at least, the closer the evidence gets to school, the smaller the effect. Across subjects the picture is mixed: some recent reviews find larger effects in schools than at universities – but the school studies are still few and small.

Go deeper. The retraction itself is one page, free, and worth reading precisely because it is short: Wang, J., & Fan, W. (2026). Retraction note: The effect of ChatGPT on students' learning performance, learning perception, and higher-order thinking. Humanities and Social Sciences Communications, 13, 528. DOI 10.1057/s41599-026-07310-z.

The retracted original was HSSC 12, 621 (2025). Still standing: Wu, X., Zhu, P., Zhang, J., Yin, M., & Wang, Y. (2026). ChatGPT's impact on student learning outcomes: A meta-analysis of 35 experimental studies. Humanities and Social Sciences Communications, 13, 684. Free, DOI 10.1057/s41599-026-07019-z.

The school-only figure for mathematics: Yi, L., Liu, D., Jiang, T., & Xian, Y. (2025). The effectiveness of AI on K-12 students' mathematics learning: A systematic review and meta-analysis. International Journal of Science and Mathematics Education, 23(4), 1105–1126. DOI 10.1007/s10763-024-10499-7.

A broader, more recent pooled analysis: Han, X., Peng, H., & Liu, M. (2025). The impact of GenAI on learning outcomes: A systematic review and meta-analysis of experimental studies. Educational Research Review, 48, 100714. DOI 10.1016/j.edurev.2025.100714.

Risks of AI

(1) Doing something is not the same as learning

Imagine this. Before you set a homework task a chatbot can complete, think about the purpose of the task. Might it be a good idea to change the task?

Why this matters. Do you learn faster with a chatbot? Around a thousand pupils in Years 9 to 11 practised maths in three groups: one using a standard chatbot, one using a version programmed to withhold the answer and ask a follow-up question, and one without a chatbot. Whilst using the chatbot, the group with the standard chatbot completed 48 per cent more practice exercises than the group without one. Afterwards, everyone sat the same exam without a chatbot. The result, which seemed surprising at first, was that the group using the standard chatbot performed 17 per cent worse than the pupils who had not used a chatbot. The group that had worked with the chatbot which withheld the answer showed no drop in performance in the test compared to the group without a chatbot, but practised more. The pupils using the standard chatbot were neither lazy nor did they cheat. They worked harder and learnt less, because what the tool eliminated was not the effort, but the cognitive challenge that constitutes learning. 

Voices from the field

“AI makes it very easy to get a solution without understanding it.”

Go deeper. Bastani, H., Bastani, O., S ngu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. *PNAS, 122*, e2422633122. DOI 10.1073/pnas.24 2633122.

(2) AI Detectors punish the wrong pupils

Imagine this. You receive a student’s essay to grade. Curious about its authenticity, you feed it into an AI detector. The result flashes: the essay was almost certainly written by AI.

Why it matters. You use the AI detector to ensure fairness, yet for a particular group, the detectors have exactly the opposite effect. A detector does not recognise machine-generated text. It measures how predictable the wording is. And that is precisely where the problem lies, for example, when a pupil is writing in a foreign language, is unpractised in writing, or has a limited everyday vocabulary. Their writing is predictable. When seven widely used detectors analysed 91 essays written for the TOEFL English test by non-native speakers – collected from a Chinese online forum – they wrongly flagged them as machine-generated in around six out of ten cases. Essays by American eighth-graders writing in their native English were almost all classified correctly. A tool that triggers false positives for exactly one group is not neutral. Furthermore, this group is the least able to defend itself.

Go deeper. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. Free, DOI 10.1016/j.patter.2023.100779.

(3) Where on earth will the pupils' data be?

Imagine this. You use an AI tool your school offers. Try to find out where the pupils' data go.

Why it matters. Can your school ensure that the data is not passed on to a third party? A third party is a company that is neither your school nor the app’s developer. Typically, these are advertising or analytics companies that receive data whilst the pupil is working. No one at the school chose this company. When researchers carried out a technical analysis of 163 learning products recommended by governments during the pandemic-related school closures, 145 of them processed children’s data in a way that jeopardised or infringed their rights by transmitting data to 196 third-party companies or granting them access, predominantly from the advertising industry. Of the 42 governments that developed their own products rather than purchasing them, 39 had developed products with the same problem. The study covers products from 2021, and some may have changed since then. However, we cannot rule out that the same applies to AI applications.

Go deeper. Human Rights Watch (2022). "How dare they peep into my private life?: Children's rights violations by governments that endorsed online learning during the Covid-19 pandemic. Free at hrw.org. The country annexes let you look up what was endorsed where.

(4) AI  arrives last where it is needed most

Imagine this. How much training did you already receive on learning about AI or subject-specific teaching with AI?

Why this matters. In one experiment, adults with and without a university education solved a work-style business task. With an AI assistant, about three-quarters of the gap between the two groups disappeared. When the researchers then took the assistant away, a sizeable part of the gap came back. The authors conclude that the AI did much of the work, and that only part of the gain survived once it was gone.

A survey of American school principals adds the second half of the picture: the more disadvantaged a school’s pupils, the less the school had done to integrate AI – fewer policies, less teacher training, less guidance on pupil use. That is an association, not proof of cause.

Taken together: a good tool can narrow a gap while it is in the learner’s hands. But it tends to reach the schools that need it least, and only part of its benefit stays behind when it is removed.

German figures show the same pattern. Among 1,590 young people aged 14 to 20, 80 per cent from the most affluent families but only 55 per cent from the least affluent saw AI as an opportunity for school. Usage follows the same gradient: among young people who use AI tools, 70 per cent of Gymnasium pupils use them for school or homework, compared with 58 per cent at other secondary schools.

What about teacher professional development? A survey of more than ten thousand teachers in England found that 45 per cent of teachers at private schools had already received formal training on AI, compared with 21 per cent at state schools. Within the state system, the figure ranged from 26 per cent at schools with the most affluent intakes to 18 per cent at those with the least affluent.

Go deeper. Cruces, G., Fernández Meijide, D., Galiani, S., Gálvez, R. H., & Lombardi, M. (2026). Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment (NBER Working Paper 34851). A working paper, not yet peer-reviewed. Campos, C., & Singleton, J. D. (2026). AI diffusion gaps: Unequal integration of AI across K-12 schools (NBER Working Paper 35347). Latham, K., & Montacute, R. (2025). Artificial advantage? AI in the classroom and the inequality gap. The Sutton Trust. Free at suttontrust.com. Vodafone Stiftung Deutschland (2024), Pioniere des Wandels; mpfs (2024), JIM-Studie 2024. Both free, in German.

(5) What improves when learning with AI is not necessarily what you see

Imagine this. You introduce a new chatbot into your teaching and the pupils are enthusiastic about using it.

Why this matters. A large review of 228 studies looked at four different things AI might improve: knowing and understanding; applying what one knows; the affective side, such as motivation and engagement; and the ability to plan, monitor and correct one’s own learning. Three of the four improved. The last one did not, as far as the studies could tell.

Only a handful of the results came from randomised trials, and the enthusiasm you see in the first weeks is exactly what short studies capture best.

Here’s a quick side note: in this review, the greatest gains were found in the arts and humanities.

Go deeper. Yeo, G. H., & Lansford, J. E. (2025). Effects of artificial intelligence on educational functioning: A review and meta-analysis. Educational Psychology Review, 37(4), article 110. DOI 10.1007/s10648-025-10085-5.

Go deeper. Yeo, G. H.  & Lansford, J. E. (2025). Effects of artificial intelligence on educational functioning: A review and meta-analysis. Educational Psychology Review, 37(4), article 110. DOI 10.1007/s10648-025-10085-5. 


1. AI in Education

1.2 Teaching About vs Teaching with AI

Take Home message

If your remember only one sentence from this page: AI is a satnav. It saves the driver who already has a map in her head and quietly stops the driver who hasn't from ever building one

The satnav problem

To drive a London taxi, you must first pass a knowledge test: 320 set routes across the 113 square miles within six miles of Charing Cross, for which Transport for London allows up to two years before the examinations even begin.

Examined drivers also use a satnav, and nobody thinks less of them for it. The knowledge test did not become worthless when the device arrived. The satnav became the thing that makes the drive safe. A driver with the city in her head uses the satnav for what she cannot know: the accident on the bridge or tonight's traffic. And she notices when it routes her down a street that has been dug up since April.

Now put a newcomer in the same taxi with the same satnav. She arrives at every address. For a while, she is indistinguishable from the experienced driver. But following the satnav’s instructions, she is building nothing. After a year of turning left when instructed, she still cannot cross the city without the device, and she cannot tell when it is wrong.

C1_2_Fig_1.png

Fig. 1: The outcome depends on the driver. On the left side the driving begins with a map in the head. The the satnav adds what the driver cannot know. This is called "beneficial offloading". On the right the driver has no map in the head. The device is the only orientation for the driver. This is called "detrimental offloading" (Image generated with GoogleGemini).

Same satnav, two drivers, opposite outcomes. The difference was never in the device. We talk about more than an image. Studies of drivers find that people who rely heavily on satnav form a poorer mental map of their surroundings. A study over three years showed that heavier use went with a steeper decline in spatial memory. And psychologists have known for two decades that the same support which helps a beginner can get in the way of someone who already knows the subject. The same happens the other way around. The effect of an AI tool depends far more on what the user already knows than on the tool itself.

This is what researchers call a schema, the organised prior knowledge of a person. Where it is in place, handing work over to a machine frees you up. Researchers call this beneficial cognitive offloading. However, when knowledge needs to be developed, the same process can prevent it from forming. This is called detrimental cognitive offloading.

A study with around a thousand secondary students shows this happening in a mathematics classroom. One group practised with an unrestricted chatbot, another without. While they had it, the chatbot group scored 48 % higher on the practice problems. Then came an exam with the chatbot taken away. Now, the same students who solved practice problems so well scored 17% worse than the group that had never used the AI chatbot. A third group used an AI version built to withhold answers and ask questions instead. That group lost nothing.

That third group is the whole design question. A satnav that names the destination and helps you find the route teaches you the city. A satnav that says turn left in 200 metres does not, however good its data. Researchers call the difference guardrails, and the design decides whether you help your students learn the city's map or simply find the destination.

Please remember: You have the knowledge. Your students, in the subject you teach, do not yet. That asymmetry is the whole chapter.

Voices of the field

“Right now the tools students reach for are built to solve problems, not to teach, and this is the core difficulty in my work currently.”

Two floors, not two topics

In academic literature, these two areas are considered distinct fields. You could think of them as two floors of a building, with most teachers located on the ground floor and no staircase leading to the first floor.

In a study with 90 teachers at Bavarian vocational schools who had signed up for AI training, teachers rated their knowledge of AI at 3.4 out of 5 on average, but their ability to use it in their teaching much lower, at 2.6. Knowing and doing had come apart. A trial across five European countries then tested whether training could close that gap. 736 teachers in France, Ireland, Italy, Luxembourg and Slovenia took part. The teachers were split up in two groups. One group received an online course and support during the trial, the others received the same online course after completing the trial. The results showed that knowledge and the ability to judge what a tool could do, increased. But in most countries the regular classroom use of AI did not change: they tried tools out and went back to what they had been doing before.

This is the knowledge–action gap. Courses produce knowledge, but changing practice needs accompaniment, trialing and colleagues to think with. Transfer has to be designed, not hoped for.

Voices of the field

“The only drawback of this webinar is that we cannot try it out directly.”

What teachers are actually doing

The most robust finding across countries is quietly surprising: teachers use AI intensively for themselves and hardly at all with their students. In Germany, 52 % of teachers use AI tools several times a month or more. But of those who use AI, only 29 % let their pupils work with it in class. So the public debate about "AI in the classroom" is largely a debate about something that is not yet happening at scale.

C1_2_Fig_21.jpg

Fig. 2: Difference in AI use for preparation compared to usage in the classroom (Image generated with Google Gemini).

What is happening is workload relief, and here the news is good. In a trial in England, 259 teachers across 68 schools were given ChatGPT and a guide for planning science lessons for 11- to 13-year-olds. They spent about 56 minutes a week on planning, against 81 in the comparison group. Independent assessors who did not know which group a lesson came from found no evidence of a difference in quality. In a large US survey, teachers who use AI at least weekly estimated saving almost six hours a week – their own estimate, not a measurement.

And one pattern is worth pausing on. In the same US survey, schools with an AI policy had more teachers using AI – seven in ten against six in ten – and so saved about a quarter more teacher time in total. For the individual teacher who uses AI every week, the difference was small. That is the teachers' own account rather than a measurement, and a survey cannot show which way the arrow runs: well-organised schools may simply do both things well. But it is the opposite of what most people expect – clear rules and heavy use appear together, not in opposition. 1.4 Compliant Use of AI

Voices of the field

“It's important to have it as a supplement, but definitely not let ChatGPT run the whole lesson.”

Why the tool is not the point

If you wanted a ranking of the best AI tools, the research will disappoint you and then, on second reading, set you free. When researchers gather many separate studies and ask what separates the classrooms where AI helped a great deal from those where it barely helped at all, the answer is never the brand of software. It is what the students were asked to do with it. The following bullet points summarise important research results. But all results build on only a few studies.

Pedagogy is more important than the technology used. Where researchers have sorted the studies by what students were actually doing, the activity mattered more than the tool. Interestingly, the same pattern was found for the previous generation of digital tools in secondary schools. The results largely depend on whether teachers had been trained to use them. 

Good feedback is not the same as feedback that works. In one experiment, 70 university students wrote an argumentative essay and revised it a week later. They were assigned at random to three groups: one got feedback from an experienced teacher, one from a chatbot given a plain instruction, and one from the same chatbot prompted to reason through the essay step by step. Marked against the same rubric, the step-by-step chatbot produced the best feedback of the three. The feedback was better than the plain chatbot and better than the teacher. Then came the revisions, and all three groups improved by much the same amount. The best feedback in the study brought no more improvement than the lowest-rated feedback. Better comments did not produce a better essay.

This result challenges a common assumption that better feedback leads to better learning. Obviously it does not on its own. Feedback only does anything if it is taken up: read, understood, acted on, and understood well enough that the student could get it right unaided next time. Uptake is the bottleneck, not quality.

More isn't always better. In a second study, with 60 Year 11 students and 240 pieces of writing, the AI produced way more content-related comments than the teachers did. But the students acted more often on the teachers' comments. When confronted with thirty precise corrections, a student might either stop diagnosing and start triaging or just comply. If a student follows the rules, they're basically saying that the revision was unnecessary. It's the satnav again, but on a different level. It gives you really detailed feedback, like 'Turn left in 200 metres' for your own writing.

It matters who says it. We're not really sure why teacher comments are taken up more often. This is all about interpretation, not measurement, but the most likely explanation is that feedback is seen as an act between people. What someone who knows what you did last term and will see what you do with it has to say is different from the same sentence generated on request.

What follows for practice. A pooled analysis of studies on automated writing feedback that was done before today's chatbots existed came to similar results. Feedback did more when students had to apply it to a new piece of writing than when they only revised the text in front of them, and more when it was used over several sessions rather than once. And one more thing should be said fairly: compared with no feedback at all, AI feedback does help. In a study by a German team with 459 upper-secondary pupils, AI-generated feedback modestly improved their revisions and their motivation.

And the same split shows up at scale. One team pooled 228 studies: AI clearly and consistently improved what students could produce, while any gain in their ability to plan, monitor and correct their own learning was small enough to have been chance. Programming shows it sharpest. Students with an AI assistant were measurably faster, and no better at the end than those without one. Getting more done and getting better are not the same thing, and AI separates them.

Europe as mirror

Across Europe, surveys of teachers' AI use run from about nine in ten in a Czech sample to about one in seven in France. The surveys differ in method and timing: the Czech figure comes from a small 2025 survey of 317 teachers and counts any experience with AI whereas the French figure comes from the OECD's international teacher survey TALIS 2024, in which just under half of Czech teachers reported using AI in their teaching. 

France built one of the continent's largest state AI programmes: adaptive practice software in maths and French, free for every pupil in the first year of upper secondary school. Nevertheless, French teachers use AI rarely and  only 9 % of French teachers have received any training in AI, compared with 38 % on average across OECD countries.

Norway has written the satnav problem into its policy (at least as a recommendation) the way a taxi driving school would do. Learn with a map first, and afterwards add the device. In June 2026 Nowrway recommended that pupils in grades 1 to 7 should, as a rule, not be given access to AI. In grades 8 to 10 AI may be introduced gradually and carefully, once teachers have the competence to guide it. In  upper secondary school pupils should learn to use AI purposefully. 

The blind spot

One risk of AI is rarely named, and the image reaches it too: knowledge about the London city map is not permanent. A driver who follows the arrows for five years does not keep the city in her head. It fades, and she stops noticing the dug-up street. Your expertise is maintained by doing the very things AI is best at absorbing: writing the task, anticipating the misconception and drafting the explanation. Where AI takes over the work through which judgement forms, relief turns into deskilling. In a German panel of e-learning experts, 65 out of 100 expected generative AI to cost people important skills, up from 44 the year before. In medicine the effect has already been measured: after AI support became routine in colonoscopy, doctors working without it detected fewer precancerous growths than before.

Use the satnav. Just keep the map.

1. AI in Education

1.3 Responsible Use of AI

Take Home message

If you remember only one sentence from this page: The dams are breaking. Teach swimming.

The reading journal, second look

Back to our ninth-grade student and his suspiciously polished reading journal. The instinct is almost universal, and it is the instinct of a good teacher: find out whether he used AI. Build a better dam.

This chapter is about why that instinct, on its own, leads nowhere. We will come to what works instead.

Start with the uncomfortable part. Teachers cannot reliably tell AI text from student text. In two studies by a German team, 89 trainee teachers and 200 experienced teachers were given a mix of student essays and chatbot essays and asked to sort them. Neither group could do so reliably but both groups were confident they were able to. Experience helped only a little: the experienced teachers still missed most of the AI texts. At one British university, researchers submitted chatbot-generated answers to real unsupervised online exams through the normal channels, without telling the markers.

Voices from the field

“You may use it, but you must be able to explain the result in your own words.”

The detector that fails the wrong students

Researchers tested seven of the most widely used AI detectors on TOEFL essays written by people learning English as a foreign language, collected from a Chinese online forum. Alongside, they put essays by native-speaking American eighth-graders. The native speakers' work was classified correctly almost every time. Of the foreign-language essays, more than six in ten were wrongly flagged as machine-written.

The mechanism is not mysterious. Many of these tools do not detect the work of large language models. The tools measure predictability. Text that uses common words in expected orders scores as artificial. Writers in a second language, with a smaller vocabulary and more rigid sentence structures, can produce exactly that signal. Meanwhile, detectors trained for one specific test, and studies in other languages, show smaller differences. Up to now there are no detectors on the market that are reliable enough to accuse anyone of willful deception.

The tool's mistakes fall systematically on the students who are already carrying the most. Institutions have done the arithmetic. Vanderbilt University pointed out that even if such a tool were wrong only 1 % of the time, that would still mean roughly 750 papers a year wrongly flagged out of its 75,000 submissions. Consequently, they switched it off. A guide written by researchers for Germany's federal education ministry is blunt: the technology is "not sufficiently reliable to provide legally secure proof." An independent European test of fourteen detectors reached the same verdict: "neither accurate nor reliable."

Much of Europe has moved on from this debate. France's education ministry and its data protection authority advise schools not to use detectors, the ICT agency Kennisnet in Denmark calls them highly unreliable. In higher education individual universities, such as the ZHAW in Zurich, forbid using them for grading. In England, the exam boards allow detection software only as one element of a judgement that rests on the teacher, who "knows their students best". A data-protection point is worth carrying home as well: pasting a student's work into an online tool means processing their personal data.  For this purpose, a legal basis and a contract with the provider is required. → 1.5 Compliant Use of AI

It seems that the cheating panic is thinner than the noise around it. Researchers surveyed students at three US high schools before ChatGPT existed and again afterwards. Cheating stayed broadly where it was. A follow-up survey of more than 4,000 students at six other US high schools a year later found the same.

The real question was never, "How do I catch him"? It is: what does a homework task still measure, once its product can be generated on demand?

Voices from the field

“At first I saw this as cheating. Now I see it as a signal that the task itself has to change.”

Two things that will not simply be fixed, so they must be taught

Confident invention is built in. These systems generate plausible continuations of a text, and plausible is not the same as true. Reviews of the research agree that no technique has yet removed the problem, although it can be reduced and partly detected. The reason is, that the standard tests used to judge large language models reward a confident guess over admitting uncertainty, so as long as they are measured that way, a certain rate of fluent, well-formed falsehood will persist. 

The slants in the training material are built in too. Moreover, they do not shrink as the models improve. A European study of gender stereotyping across languages found that bigger models, even after the extra training intended to make them safer and fairer, sometimes stereotype more rather than less. If a property cannot be engineered away, it becomes curriculum. And it can be taught: in one Finnish programme, the share of students who could explain where an image generator's bias came from rose from 7 % to 44 %. The consequence for teaching must be training error analysis instead of error avoidance. In other words: the machine's failures become the lesson.

Voices from the field

“Let the learner check the chatbot's solution as if they were the teacher. Looking for the mistakes is how they finally learn the method.”

What works: design, not control

The evidence points to one simple principle that's easy to remember: think first, then AI.

Two limits are worth knowing before you rely on this. 

The first: the popular advice that teachers should simply instruct a chatbot to act as a Socratic tutor may not survive a long conversation. Early evidence, from a study with simulated students that has not yet been peer-reviewed (October 2026), suggests that under persistent pressure such tutors drift back into supplying answers. Reliable limits have to be built into the system, which makes them a purchasing decision rather than a prompting skill.

The second: making an AI explain its reasoning sounds like the obvious safeguard, but explanations do not reliably prevent over-trust. In one study they left people's acceptance of wrong advice unchanged, and explanations that are too technical or too simplified can increase misplaced confidence. Explanation is only a condition for learning.

Attitude: two failure modes, not one

We worry that teachers are placing too much trust in AI. However, we should be equally concerned about the reverse. Even experienced people can be caught out by over-trusting. In a small pilot in US schools, four in ten of the teachers surveyed judged an AI grading tool's feedback not useful, mostly vague, or even incorrect. Research on automation bias shows that people tend to go along with a machine's suggestion even when they have reason to doubt it. Spotting an error and correcting it are different steps.

The opposite pole is concealment. Teachers hide their own use of AI from colleagues and students, forbidding them to do what they do themselves: 77% use AI privately, while 57% say students should never use it to generate ideas. Meanwhile, a study of academically able students in grades 6 to 8 found that between a third and two-fifths already said the AI knows more than their teachers. A profession that hides its practices cannot model responsible practice, and transparency is not just a courtesy. It is the mechanism by which teaching by example works. 

Does AI level the field or tilt it?

In learning tasks, it does the opposite. One study found no average benefit at all, but a widening distance between students who already knew a lot and students who did not. Another looked at students writing in a foreign language: the weaker writers received more corrections from the AI and successfully acted on fewer of them, and several found the flood of feedback dispiriting. Feedback without the ability to sort it does not help; it buries.

The gap also exists before anyone touches a tool. In Germany, 80 % of students from the most advantaged homes see AI as an opportunity, against 55 % from the least advantaged. In Britain, the divide runs through the staffroom rather than the device cupboard: 45 % of teachers at private schools have had formal training in using AI, against 21 % at state schools. Within the state sector, wealthier schools again outpace poorer ones. A 2026 survey of more than 25,000 children across 17 European countries finds the same social split in how much and how variedly they use AI [PRÜFEN: SES-Befund am Bericht]. One commentator's line captures the whole problem better than any figure: "the rich have access to both technology and people to help them use it, while the poor have access to technology only."

So the bottleneck is which learner has a teacher who has been trained. AI compensates when three things hold together: secure access, a tool designed for learning rather than answering, and a teacher scaffolding its use. Remove any one, and it amplifies instead. Since all three tend to be present in the same schools and absent in the same schools, amplification is the default and equalisation is something you have to build. 

1. AI in Education

1.4 Compliant Use of AI

Take Home message

If you remember only one sentence from this page: Compliance is a house, not a wall.

What am I allowed to do?

Keep these four steps in mind before typing, pasting or uploading anything. Please use the following overview as a general guide. Requirements in your country may differ from those listed here. Therefore, always check the regulations specific to your country and school. 

Fig 1_4_1.jpg

Fig. 1: ... (Image generated with ...)

Voices from the field

“We are not even sure what we are allowed to do, least of all when it comes to data protection.”

The reading journal, third look

Our ninth-grade student, his reading journal, his teacher's suspicion. In Chapter 1.2, it was a question about learning; in Chapter 1.3, about professional judgement. Here it is a court file.

In December 2025 the Administrative Court of Hamburg, Germany, ruled on an urgent application in the case. The English teacher had noticed that the reading journal was written at a far higher level of spelling, grammar and vocabulary than the boy's class test on the same book. He later admitted using ChatGPT, and the school graded the journal as an attempt to deceive. His father applied to the court for urgent relief; the court refused. Its reasoning goes beyond this case: pupils must assume that they are to do their work entirely on their own unless an aid has been expressly allowed – so undisclosed AI use counts as deception even without a written rule against AI. (In this case, the teacher had in fact told the class that ChatGPT was not allowed.) One caveat to keep in mind from the outset: this is a single court, in urgent proceedings, and the decision is not final.

This is a good moment to address the most common misconception in this field. There is no legal vacuum. There never has been.

The house has three floors

Most teachers envisage AI law as a single wall under construction somewhere in Brussels. However, it is more useful to picture a house that you already live in. The foundation is data protection. It is load-bearing. It was laid years ago. It governs your Tuesday morning. The upper floor is the EU AI Act, Europe's AI law, which categorises systems by their potential for harm and imposes stricter obligations as risk rises. It is partly occupied and still under construction, with the completion date just pushed back. The doors and windows are copyright and examination law — small and specific, and touched daily without noticing.

Fig 1_4_2.jpg

Fig. 2: Compliance is a house (Image generated with ...)

Ground floor: data protection, clearer than many think

Handling someone's personal data always requires legal permission. As a public body, a school's permission essentially comes from its statutory duties under school law. Being a teacher does not come with a general licence.

Three consequences follow, forming the practical core of this chapter.

Removing names does not anonymise data. In 2024, Germany's data protection authorities jointly stated that if someone could still be identified from the context, the data is still personal. This is particularly problematic in schools because the types of text that teachers most want help with — such as essays, free writing and learning journals — identify their author by style.

The dividing line is the contract: a provider may handle your data if it does so strictly on your instructions and there is a written agreement to this effect. However, the moment it uses what you type for its own purposes, the arrangement collapses because it is no longer working on your behalf. So the practical test is this: Is there such an agreement, and does it exclude training on your inputs? If not, no personal data should be entered.

Two footnotes worth noting: data relating to health and disability is subject to even stricter protection under EU law. It does not belong in an AI tool. And where teachers use private devices, at least two German states (North Rhine-Westphalia and Bavaria) make clear that the data-protection responsibility lies with the school and its head, who must approve the use.

Upper floor: what the AI Act does and does not require yet

Here is the correction that makes most existing guidance out of date. In July 2026, with an amending regulation known as the Digital Omnibus, the EU postponed the heaviest set of AI Act duties for education which are those attaching to systems classed as high-risk, meaning systems used to decide admissions, to assess what pupils have learned, or to monitor them during examinations, from August 2026 to 2 December 2027. A great many handouts written in 2025 and early 2026 still print the old date. If a document in your school does, it needs a footnote.

The more useful correction runs the other way: what already applies is underestimated, and what is endlessly discussed does not apply yet.

Two things have been binding since 2 February 2025.

Read that second duty alongside Chapter 1.3 and something clicks. The gap between knowing about AI and being able to teach with it is no longer only a professional problem. Since February 2025, schools have been legally required to work on it.

Since 2 August 2026, chatbots must tell users that they are AI, and providers must mark AI-generated content in a machine-readable way; schools themselves must label deepfakes, and AI-written texts they publish to inform the public. A worksheet a teacher has checked needs no label. From 2 December 2027, the duties for high-risk systems apply – systems used to decide who gets admitted, to assess what pupils have learned, or to monitor them during examinations. Whoever uses such a system then takes on duties of its own, including assigning oversight to people with, in the Act's words, "the necessary competence, training and authority".

Doors and windows: copyright and examinations

Counter-intuitively, what comes out is the easy part. A text or image that an AI generates on its own has no human author in the sense copyright requires, so it is usually free of copyright itself; where a person shapes it through their own creative choices, the result can be protected. It is not, however, guaranteed to be clean: in a widely noted ruling, a Munich court found that a chatbot had memorised song lyrics and reproduced them almost word for word. The court placed the liability on the provider, not on the person who typed the request. The judgment is not final.

The exception that lets teachers copy material for their classes covers exactly that: copying a limited share of a work, for that class. Uploading a protected work into a commercial AI system is a different act, and the exception does not address it. Material produced specifically for school teaching, such as textbooks and workbooks, is in principle excluded from the exception, as long as licences for it are readily available.

How much should that worry you? Less than the gap suggests, and more than nothing. The question is genuinely unsettled: no higher court has ruled on it, and the specialist literature regards the existing exceptions as a poor fit for the situation. Where a rights holder objects, the claim would be a civil one, and in practice, this is a matter for school policy and procurement rather than something an individual teacher can settle. What is clear in practice is the narrow case. It is the line the educational publishers draw, and it is the one worth remembering: school textbook and publisher material should not go into AI tools at all — neither scanned nor retyped, not even to generate a worksheet. 

Case law in this field is currently inconsistent across Europe. In February 2026, the Administrative Court of Kassel, Germany, decided two university cases the same way, and sharply: even though the university had no explicit ban on AI, the line was crossed "already with a single undisclosed use of generative AI". The judgments are not final. In the same month, a Paris administrative court, in urgent proceedings, stopped a university from sanctioning a student because it had set out no rule on AI use in advance [PRÜFEN]. A Swedish appeal court had reached a similar result earlier: a student had handed in three invented quotations, but the course rules allowed "all aids" in take-home work, so there was no rule to have broken. In the Netherlands, by contrast, the country's highest administrative court upheld a sanction in July 2025 against a law student who had submitted four non-existent sources: the fraud was proven beyond reasonable doubt, and whether he had meant to deceive did not matter.

So the pattern across jurisdictions is not about countries. It is about whether a rule existed beforehand, and whether the deception was actually shown. 

Three rules follow for practice.

  1. Detectors do not carry the burden of proof. The school has to show the deception. Courts have accepted conclusions drawn from concrete facts: sources that turn out not to exist, a clear gap between what a student writes and what they can explain orally, sudden jumps in level within the same text, an admission. The student can then explain. Writing style alone is a weak basis – leading German legal commentators doubt that it can carry a finding at all. Detector output is, at best, a reason to look more closely. The failure rate reported in Chapter 1.3 makes it unusable as evidence.

  2. Rules must be clear before the assessment, not after. Neither a blanket ban nor a blanket permission holds up. What works is a declaration of independent work with a clause about AI, plus a duty to state what was used. Disclosure instead of banning.

  3. Grades stay human. People have the right not to be subject to decisions with legal effect that a machine takes alone, and schools have no legal basis for making such decisions by software; a grade is one. Germany's Standing Conference of the Ministers of Education and Cultural Affairs puts it even more strongly: assessment is a task only teachers can perform. Crucially, formally waving through an AI's suggestion does not count as a human decision. So the finding that people often go along with flawed AI feedback is a legal problem as well as a pedagogical one.

Voices from the field

“Everyone is experimenting on their own, so the rules change from one lesson to the next.”

Law, ethics and didactics therefore converge on the same answer: assess the process, not just the product.

The real address of responsibility 

In response, Germany has introduced state-provided school AI environments. These are vetted platforms with negotiated contracts, rather than leaving teachers to sign up to whatever is free. While these solve problems that individual teachers cannot solve alone, they are still criticised for how they are governed, whether teachers actually accept them and whether the assessments they support are sound. Because the rollout differs from one federal state to the next, it will take time to see how teachers are faring with the platforms.

What is distinctive is narrower than it may initially appear, and it is worth stating precisely. Germany is the only system we found in which a public body – the FWU, owned by the federal states – develops and runs the platform itself, with technical contractors. France provides adaptive practice software for an entire year group, but commissioned it: one private company wrote the software, another hosts the data. Several other countries have bought access to commercial chatbots; most of the rest offer guidance only.

Building it yourself does not automatically give you more control. The German platform runs on the same commercial language models as everyone else's on European servers, as an industry association has pointed out. By contrast, the French system is hosted in France under the highest security certification issued by the French state. One country built the surface and rents the engine, while the other bought the software and keeps the keys to the building. Neither option is clearly the most sovereign choice, which is worth bearing in mind before using the term.

The Netherlands took a different approach, and it is worth knowing about because it offers a more honest answer to the same question. Rather than building or buying, Dutch schools and universities joined forces and had the major providers formally assessed. When they examined Microsoft's AI assistant, they found two issues that could not be resolved: the criteria for an automatic content filter that the provider would not disclose and the retention of eighteen months of diagnostic data that could not be justified. They negotiated. In March 2026, however, the provider declined to change. The assessment therefore remains amber, and Dutch schools are advised to exercise caution and use their own judgement on a case-by-case basis rather than granting clearance. One country removed the risk by building around it, while another took a closer look and decided it was not ready yet. Both approaches offer solutions to the problem set out at the start of this chapter.

The comparison also settles one more thing. The complaint that teachers prefer the free chatbot to the official one is not due to a failure of implementation in Germany. In Estonia, where the state negotiated access for upper-secondary students, barely more than a third were using the official chatbot weekly seven months after the programme began. It turned out that most of them had been using free tools before it started. Everywhere, the vetted environment competes with the more convenient one. Providing the official tool is necessary. But this is not sufficient. 

However, the structural point stands: compliance is an enabling condition, not a hindrance. Remember what was said in Chapter 1.2? Schools with clear rules had more teachers using AI – and saved more time overall. Rules did not slow use down

1. AI in Education

1.5 Current State of AI in Education

Take Home message

If you remember only one sentence from this page: AI is already present in educational contexts but is not regulated. Structured frameworks are needed to support educators' professional development improving their knowledge and letting them teach with AI and about AI.

Current Uses of AI in Educational Contexts

Nowadays, artificial intelligence (AI) is embedded in everyday educational practice, well beyond pilot projects. Common applications include adaptive learning platforms, intelligent tutoring systems, automated grading, and AI-assisted lesson planning for teachers (OECD, 2026). The most visible shift, however, is the direct use of general-purpose tools by students and teachers alike.

Teaching With AI

AI is increasingly used by teachers as a support tool across multiple stages of their work, from lesson planning to grading. According to OECD's TALIS survey, 37% of lower secondary teachers reported using AI in 2024, with 57% saying it helps them write or improve lesson plans (OECD, 2026). Beyond lesson design, teachers are using AI to differentiate content for different skill levels, generate practice materials, and support the grading process. At the same time, adoption is not without friction: 72% of teachers expressed concern that AI could enable students to pass off AI-generated work as their own, reflecting a broader tension between AI's practical benefits for teaching and its implications for academic integrity (OECD, 2026).

Teaching About AI

Alongside using AI as a teaching aid, there is a growing need to teach students about AI itself: how it works, its limitations, and how to use it responsibly. This includes building a basic understanding of language models as prediction engines rather than databases, which explains their tendency to hallucinate, as well as practical skills such as fact-checking AI output and recognising fabricated citations (UNESCO, 2024). It also involves developing critical thinking around prompting, shifting students from simply following instructions to formulating them, and instilling awareness of the ethical and legal dimensions of AI use, including data privacy and the EU AI Act. UNESCO's AI Competency Framework for Teachers (2024) frames this as a core professional competency, arguing that teachers themselves need training in these areas before they can effectively teach them to students.

What is missing?

Alongside institutional frameworks and adoption data, it is important to understand how these dynamics play out in practice. A series of workshops was conducted in the context of the DUAL.AI.TEACHer project. Numerous education stakeholders were called to gather first-hand perspectives on the current use of AI in teaching. These workshops surfaced a set of concrete needs and concerns expressed by educators, covering practical, pedagogical, and training-related aspects. The following sections outline these needs in detail.

Voices from the field

“Anyone who teaches with AI must be able to verify every answer it gives, because errors happen. Without solid subject knowledge you will not spot them.”

1. AI in Education

1.6 Legislation

Take Home message

If you remember only one sentence from this page: Use AI to support your professional judgement, not replace it—stricter rules apply as soon as AI helps make decisions about grades, educational pathways or people.

The EU AI Act

The EU AI Act (Regulation (EU) 2024/1689) is the EU’s single, risk-based rulebook for artificial intelligence. Instead of regulating specific products, it classifies AI systems by the risk they pose and attaches obligations accordingly — from an outright ban for a small set of practices, through extra duties for “high-risk” uses, to no special rules at all for everyday tools. It applies directly in every member state, to public and private organisations alike.

The Act phases in step by step. It has applied since 1 August 2024; since 2 February 2025, the banned practices below and the duty to build AI literacy (Art. 4) are already binding; from 2 August 2026, most of its remaining provisions apply. 

Fig1_1_6.jpg

Fig. 1: The "AI roadmap" (Image generated with ...)

Who is responsible for the use of AI?

If a school introduces and controls an AI system, the school or school authority is usually the deployer (Art. 3(4)). The teacher is normally not a separate deployer when using the tool under the school’s instructions, but must still follow those instructions and all relevant school and data-protection rules.

Teachers need to know what they are using

Schools and school authorities must take appropriate measures to support the development of AI literacy among those who use AI systems on their behalf (Art. 4). The AI Act does not specify a particular level of competence, certificate or training format. These rather depend on the AI system used, how it is used, and who may be affected.

AI practices banned outright for schools

The AI Act bans eight AI practices outright (Art. 5); four of them are directly relevant to schools:

  1. Using AI systems in educational institutions to infer emotions or intentions from biometric data (such as students' facial or vocal characteristics) is generally forbidden. Narrow exceptions exist for medical or safety purposes (Art. 3(39); Art. 5(1)(f)).
  2. AI systems that exploit someone’s vulnerabilities due to age (as with children), disability, or social or economic situation, in a way that distorts their behaviour and causes them harm, are banned regardless of intent (Art. 5(1)(b)).
  3. Building or expanding a facial-recognition database through untargeted scraping of images from the internet or from CCTV footage is banned. This is relevant if a school considers a facial-recognition tool for attendance or exam monitoring (Art. 5(1)(e)).
  4. Using biometric data to categorise people in order to infer protected traits such as race, political opinions, religion or sexual orientation is banned. This is relevant for any biometric-based tool, such as exam-proctoring software, that goes beyond its stated purpose (Art. 5(1)(g)).

Does AI-generated content always need a label?

If students interact directly with an AI system, such as a chatbot, they must be informed that they are interacting with AI. Usually the system provider is responsible for making this clear (Art. 50(1)).

Schools and teachers must disclose deepfakes, or AI-generated text published on matters of public interest, when these are used in the classroom (Art. 50(4)). This requirement does not apply where the AI-generated content has undergone effective human review or editorial control, and a person or institution assumes editorial responsibility for its publication. An AI-assisted worksheet reviewed by a teacher therefore does not automatically require a label.

When AI in education becomes high-risk

Ordinary writing, translation, research and lesson-planning tools are not high-risk simply because they are used for work. Classification depends on the system's intended purpose, not simply on which product is used (Art. 6 · Annex III).

AI systems may be classified as high-risk when they are intended to:

For Teachers For Schools
  • decide about students' access, admission or assignment to educational or vocational institutions and programmes;
  • evaluate students' learning outcomes or steer their learning processes;
  • assess the appropriate level of education a student will receive or can access; or
  • detect students' prohibited behaviour during examinations.

(Annex III, point 3)

  • recruit or select teachers, including screening applications;
  • decide on contract terms, promotion or termination of a teacher's employment;
  • allocate tasks based on a teacher's behaviour or personal traits; or
  • evaluate teachers' performance.

(Annex III, point 4)

A limited checking or preparatory tool may also fall outside the high-risk category if it does not significantly influence a decision. However, an education system that profiles individuals is always considered high-risk.

What high-risk AI means for schools and teachers

Classification of an AI system as high-risk does not automatically rule it out from use in schools. But high-risk systems come with obligations (Art. 26):

These obligations for high-risk AI in education apply from 2 December 2027 — but the prohibitions above and the AI-literacy duty already apply today.

For Teachers For Schools
  • use it only for its approved purpose and follow its instructions;
  • do not treat its output as a final decision;
  • understand its main limitations and likely sources of error;
  • watch for overreliance on apparently convincing results; and
  • be prepared to disregard or override its output.
  • provide competent and authorised human oversight;
  • ensure that any input data under its control is relevant and sufficiently representative;
  • monitor the system and respond to risks or serious incidents;
  • keep automatically generated logs for at least six months where those logs are under its control; and
  • inform students, teachers or affected people in advance when the system assists decision-making about them.

Public bodies and private organisations providing public services may also have to complete a fundamental rights impact assessment before using high-risk AI (Art. 27). Public authorities and organisations acting on their behalf may additionally have to register the system in the EU database (Art. 49). Whether this is the responsibility of the school or its governing authority depends on the national school system.

The AI Act is not the only law that matters

The AI Act does not replace european or national regulations on data protection, education, employment and worker participation, child protection or copyright (Art. 2(7)). For example, the fact that a tool is not high-risk does not mean that you may upload students' work or personal data to it.

A tool may be acceptable under the AI Act but still prohibited by data-protection law or your school's rules.

The GDPR

The GDPR (Regulation (EU) 2016/679) is the EU’s law on data protection. It plays a supporting role alongside the AI Act: while the AI Act asks whether and how an AI system may be used at all, the GDPR asks a narrower question — may this particular data about this particular person be processed this way? Meeting one does not automatically mean you have met the other.

In practice: before feeding real student or staff data into an AI tool, check that your school has approved it, share no more personal data than the task needs, and prefer anonymised or fictional examples when just trying a tool out.

Where to find help

Enforcement is organised differently in each EU country. Because national responsibilities may change, use the European Commission's current list of national market-surveillance authorities. Anyone (a teacher, a student, a parent) can lodge a complaint with that authority if they believe the Act has been breached (Art. 85). Where a high-risk system, such as an admission or exam-monitoring tool, has led to a decision that significantly affects a student, they (or their parents) can also ask for an explanation of that individual decision (Art. 86).

National rules for your country

The AI Act is EU-wide, but enforcement bodies, school law and data-protection practice differ by country. Country-specific notes for the Playbook's partner countries will be added here:

2. Good Practices & Classroom Examples

2. Good Practices & Classroom Examples

2.1 Introduction

Chapter 1 explained when AI helps, where it harms and which rules apply. This chapter shows what that looks like on an ordinary school day.

The examples are worked through subject by subject, starting with biology, English as a foreign language, history and mathematics. More subjects will be added during the project. Each example takes a situation from everyday teaching and shows how AI can be used, what the teacher does and where things can go wrong. They apply three ideas from Chapter 1: AI supports learning only if learners do the thinking first, you need to know a little about AI before you teach with it, and the rules must be clear before the task begins.

Please read "good practice" modestly. The practices teachers report across Europe are promising approaches, not proven recipes, and none of them has been evaluated yet. Treat each example as a starting point: try it, adapt it to your class and check it against the rules of your school. A checklist summarises the questions to ask before AI enters your lesson. 

You can search for different subjects and sort by lessons & professional development with the build in tag system in the top left corner. 

grafik.png

2. Good Practices & Classroom Examples

2.2 Biology: Vibe Coding Simulations in the Classroom

Predator–Prey Dynamics (Lotka–Volterra model)

Subject: Biology  ·  Grade: 9 (age 14–15)  ·  Duration: 1 double lesson (90 min) ·  AI: Learning WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-TL-2b

(AI foundations and applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.”


 

AF-FC-2b

(AI foundations and applications × Facilitating Learners’ (AI) Digital Competence, Level 2 – Reflective Implementation):

 

“Teachers can implement guidance that helps learners developing practical AI skills, adapting the tasks to learners’ prior knowledge and the AI tools available.”

By the end of the lesson, students can:

  • describe the key variables (prey population, predator population) and parameters (growth rate, predation rate, conversion efficiency, mortality rate) of the Lotka–Volterra model;
  • form and test hypotheses about population dynamics using an interactive simulation;
  • extend an existing simulation with additional layers of ecological complexity (e.g. a carrying capacity, a second species, a seasonal factor) by writing structured prompts for an AI language model;
  • critically evaluate the model’s assumptions and limitations (Nature of Science), including why introducing randomness or an additional factor can destabilise it.

Take-home message: Students don't need to code from scratch. Reading and testing an existing simulation, then using an LLM to add one layer of ecological realism at a time, is enough to build genuine understanding of a dynamic biological model — and, sooner or later, to discover for themselves exactly where that model reaches its limits.

Vibe-Coding SImulations.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), August 27, 2026.

Content

Mathematical models and computer simulations are especially valuable tools in ecology, because they can express highly complex systems through mathematical equations. This makes it possible to run experiments in silico and test hypotheses on the model in situations where a real experiment would be too costly, too slow, ethically difficult, or outright dangerous. Manipulating a real predator population, for instance, is rarely an option. At the same time, every model is necessarily a simplification of reality, and this is exactly where its weaknesses lie: real ecosystems are far more complex than any model can fully capture.

The Lotka–Volterra model is a classic example of such a simplification applied to predator–prey dynamics. It represents the system using two state variables (the prey and predator population sizes) and four parameters (the prey’s intrinsic growth rate, the predation-rate coefficient, the predator’s reproduction rate per prey consumed, and the predator’s mortality rate). Under its idealised assumptions, the model predicts sustained, periodic oscillations in which changes in the predator population lag behind those in the prey population. This behaviour arises from a feedback loop: as the prey population increases, more food becomes available to predators, allowing the predator population to grow. Increased predation then causes the prey population to decline, which reduces the predators’ food supply and leads to a decline in the predator population. With fewer predators, the prey population can recover, and the cycle begins again.

To produce this clean cycle, the model makes several strong assumptions. It is deterministic (the same starting values always produce exactly the same curve); it is a closed two-species system that excludes every abiotic factor (temperature, precipitation, season) and every other biotic factor (food plants for the prey, a third species, disease); and it makes each population's growth depend on nothing except the size of the other population.

Real populations rarely meet these assumptions. Births, deaths and encounters are subject to chance, and a stochastic (randomised) version of the same model tends to drift away from the clean cycle and can even collapse (e.g. the predator population dying out) in cases where the deterministic version predicts stable oscillation. Likewise, adding just a single missing factor (e.g. a temperature-dependent growth rate, or a limit on the prey's own food supply) is often enough to break the clean two-variable cycle. This is not a flaw to hide from students, it is the central Nature-of-Science lesson of the activity. A model can be genuinely useful, while still being a deliberately narrow approximation of reality and using it well means knowing exactly which factors it leaves out. In this lesson's main activity, students discover this for themselves by trying to add exactly these kinds of factors.

Lesson plan

Phase Time Activity
Introduction 8 min The teachers shows an unlabelled graph of a real predator–prey data series (e.g. lynx/hare); students describe the pattern and form first hypotheses.
Input 12 min The teacher introduces the model’s variables and parameters using a simple diagram (no differential equations needed).
Exploration 15 min Students test their hypotheses in pairs on a ready-made HTML simulation and record their observations.
Adaptation 35 min In pairs, students work through 2–3 rounds of predict → prompt → test → diagnose (guiding questions below), each time asking the LLM to add one more layer of ecological realism to the simulation (a limit on the prey's food supply, a second species, a seasonal/abiotic factor) until the clean two-variable cycle visibly breaks down.
Reflection 20 min Whole-class discussion, guided by questions such as: At which round did the simulation stop producing a clean, repeating cycle? What does the original model deliberately leave out, and why? What makes it a useful simplification rather than simply a wrong one?

Guiding questions: layering complexity

Pairs work through as many of the following rounds as time allows in the 35 minutes (most manage two; the third is an extension for early finishers). Each round follows the same four-step cycle:

  1. Round 1 — limit the prey's food supply (a carrying capacity):
    • Predict: Before you start prompting, aks yourself: what do you expect to happen to the cycle if the prey's growth slows down once the population gets large, instead of growing without limit? Sketch the curve you expect.
    • Prompt: Ask the AI to add a maximum sustainable prey population to the simulation, and to explain in its own words how it changed the growth equation to do this.
    • Test & compare: Run the new simulation with the same starting values as before. Compare the previous simulation with the new one: how did the variables change? Why is that?
    • Diagnose: Which assumption of the original model (unlimited food for the prey) did you just remove? Is this factor biotic or abiotic?
  2. Round 2 — add another (biotic or abiotic) factor:
    • Predict: If a third species now competes with the prey for food, or preys on it as well, or if prey growth now depends on temperature what do you expect this to do to the original two-species cycle?
    • Prompt: Choose another factor and write a prompt asking the AI to add it as a new variable linked to the existing ones — specify exactly how it should interact with the prey or predator population.
    • Test & compare: Observe the simulation over several cycles. Does the original two-species rhythm survive, get distorted, or disappear?
    • Diagnose: The original model assumed each population depends only on the other. Is that still true after this round? What real ecological relationship does your new factor represent.
  3. Optional round 3 — introduce randomness:
    • Predict: If one parameter (e.g. the predation rate) is no longer fixed but drawn randomly within a range at each time step, what do you expect for the long-term stability of the populations?
    • Prompt: Ask the AI to replace one constant parameter with a randomised value and to keep track of whether either population reaches zero.
    • Test & compare: Run the simulation several times with the same settings. Do you get the same result every time? Does either population ever collapse?
    • Diagnose: Why can the same “average” parameter values sometimes lead to extinction here, when the deterministic model never predicted that?

Closing question (for the Reflect phase): At which round did the simulation stop producing a clean, repeating cycle — and what does that tell you about how many real-world factors the original Lotka–Volterra model has to leave out in order to stay solvable and easy to interpret?

Materials

2. Good Practices & Classroom Examples

2.3 English as a Foreign Language

Communicative Appropriateness in English

Subject: English as a Foreign Language  ·  Grade: 9 (age 14–15)  ·  Duration: 1 double lesson (90 min) ·  AI: Learning WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-TL-2b

(AI foundations and applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.”


 

AF-FC-2b

(AI foundations and applications × Facilitating Learners’ (AI) Digital Competence, Level 2 – Reflective Implementation):

 

“Teachers can implement guidance that helps learners develop practical AI skills, adapting the tasks to learners’ prior knowledge and the AI tools available.”

By the end of the lesson, students can:

  • distinguish between grammatically correct and contextually appropriate English;
  • identify differences in vocabulary, tone and register across communicative situations;
  • formulate structured prompts that specify audience, purpose, context, language level and tone;
  • compare AI-generated language and identify inappropriate or unnatural formulations;
  • revise AI-generated language using their own linguistic judgement;
  • check whether an AI-generated response follows the requirements given in a prompt.

Take-home message: AI can generate fluent and grammatically correct English, but this does not automatically make the language natural or appropriate for a particular situation. By changing the audience, purpose and context of an AI-generated conversation, students learn to evaluate AI suggestions and use their own language knowledge to make the final decisions.

ChatGPT Image Sep 1, 2026 at 03_17_43 PM.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), September 1, 2026.

Content

Communicating successfully in a foreign language requires more than producing grammatically correct sentences. Speakers constantly adapt their vocabulary, sentence structure, politeness and tone according to the audience, purpose and context of communication. The same intention can therefore be expressed in very different ways depending on who is speaking to whom. A request made to a close friend may be short and informal, while the same request addressed to a teacher, employer or unfamiliar person usually requires different vocabulary, more polite structures and a different degree of formality. These differences are part of register, which describes how language changes according to the social and communicative situation.

For language learners, register can be difficult to recognise because a sentence may be grammatically correct while still sounding unnatural or inappropriate in a particular context. Textbook examples often provide clear distinctions between formal and informal language, while real communication is much more flexible. Age, relationship between speakers, cultural expectations, medium of communication and communicative purpose can all influence the language that is used. A message that sounds natural in a conversation between teenagers, for example, may sound too informal in an email to a teacher, while an expression that is suitable for a formal letter may sound unnecessarily distant in everyday conversation.

Generative AI can produce and adapt dialogues very quickly, which makes it useful for exploring these differences. By changing individual elements of a prompt, such as the relationship between speakers, language level, purpose or degree of formality, learners can generate several versions of the same communicative situation and compare how the language changes. At the same time, AI-generated language is not automatically reliable simply because it sounds fluent. A model may produce language that is too formal, repetitive, culturally awkward, inconsistent with the requested level or not fully aligned with the instructions given in the prompt. Generative AI also works probabilistically, which means that similar prompts can produce different formulations.

This limitation is central to the activity rather than something to hide from learners. AI becomes most useful when students treat its output as language material to analyse rather than an answer to accept. In this lesson, students first examine how register changes across communication contexts and then use an LLM to modify one communicative variable at a time. They compare the resulting dialogues, identify which linguistic choices work and which do not, and revise parts of the output themselves. In this way, students practise both communicative competence in English and practical AI literacy: giving clear instructions, checking whether those instructions were followed and using their own linguistic judgement to decide whether the final result is appropriate.

Lesson plan

Phase Time Activity
Introduction 10 min The teacher shows two short messages communicating the same request, for example one addressed to a friend and another to a teacher. Students identify differences in vocabulary, politeness and tone and suggest who might have written each message.
Input 10 min The teacher introduces the concepts of audience, purpose, context and register. Together, students identify features of formal and informal English and discuss why grammatically correct language is not necessarily appropriate language.
Exploration 15 min In pairs, students examine a short AI-generated dialogue for an everyday situation. They identify useful expressions and highlight anything that sounds unnatural, too formal, too informal or unsuitable for their English level.
Adaptation 35 min Students work through 2–3 rounds of predict → prompt → compare → diagnose. In each round, they change one aspect of the communication context and ask the AI to adapt the dialogue.
Reflection 20 min Students compare their versions and discuss which changes affected the language most. They identify where AI followed the instructions well, where it did not, and which expressions they would change themselves.

Guiding questions: layering complexity

Pairs work through as many of the following rounds as time allows during the 35-minute adaptation phase. Most pairs should complete two rounds, while the third can be used as an extension:

  1. Round 1 — Change the relationship between speakers:
    • Predict: Imagine that the same request is made to a close friend and to a teacher. What do you expect to change? Think about vocabulary, greetings, sentence structure and politeness.
    • Prompt: Ask the AI to rewrite the dialogue for a different relationship between the speakers.

      For example:

      Rewrite this conversation between two friends as a conversation between a student and a teacher. Keep the meaning the same, but use polite, natural English suitable for a B1 learner.

    • Compare: Compare the original and new dialogues. Identify at least three linguistic changes.
    • Diagnose: Did the AI change only individual words, or did it also change sentence structure, greetings, requests and closing expressions? Which changes are appropriate? Is there anything you would change yourself?
  2. Round 2 — Change the communicative purpose:

    Students now change what one speaker wants to achieve. For example:

    asking for information → making a complaint

    accepting an invitation → declining politely

    asking for help → requesting permission

    informal conversation → formal request

    • Predict: What language do you expect to change when the purpose of the conversation changes?
    • Prompt: Create a more detailed prompt that includes: speaker + situation + purpose + language level + tone

      For example:

      Create a short B1-level dialogue between a customer and a café employee. The customer received the wrong order and wants to solve the problem politely. Use natural everyday English and keep the dialogue to eight lines.

    • Compare: Check whether the generated dialogue follows each part of the instruction.
    • Diagnose: Select one expression that works well and one that you would rewrite. Explain why.
  3. Optional round 3 — Add constraints and challenge the AI:

    Students make their instructions more specific.

    For example:

    Rewrite the conversation for two 15-year-olds. Keep the English at B1 level. Make it friendly and natural, but avoid slang. Keep the same communicative purpose.

    • Predict: Which parts of the previous dialogue do you expect to change?
    • Prompt: Ask the AI to produce the revised version.
    • Compare: Check the response against every requirement in the prompt.
    • Diagnose: Which instructions did the AI follow successfully? Which did it ignore or interpret differently?

Closing question (for the Reflect phase): At which point did changing the communication situation require more than simply replacing a few words, and what does this tell you about communicating effectively in another language?

Follow with: If an AI-generated dialogue is grammatically correct, how can you decide whether it is actually appropriate and natural for this particular situation?

Materials

2. Good Practices & Classroom Examples

2.4 Social Studies: AI, Media and Reality

Can We Believe Everything We See?

Subject: Social Studies · Grade: 5 (age 10–11) · Duration: 1 double lesson (90 min) · AI: Teaching ABOUT AI

Take-home message: AI can produce convincing text and images, but convincing does not mean true. Before we believe or share something, we need to ask where it came from, what evidence supports it and whether we can verify it elsewhere.

Content

Artificial intelligence is increasingly becoming part of everyday communication. AI systems can generate texts, pictures and other forms of content that may look as if they were created by a person. This creates new possibilities, but also new challenges for how people understand and evaluate information. For students at this age, the important question is not necessarily how AI works technically, but how AI-generated content can affect people and society.

The lesson introduces AI through a familiar social situation: information encountered in everyday life. The teacher presents examples of texts and images and asks students to decide whether the content is real, edited, or AI-generated. The emphasis is not on teaching students to detect AI perfectly. In fact, one of the central conclusions of the lesson is that appearance alone is not a reliable way of establishing whether something is true.

Students instead learn a simple verification process:

STOP → QUESTION → CHECK → COMPARE → DECIDE

The activity connects with the Social Studies curriculum's emphasis on developing critical thinking, analysing evidence, forming and expressing opinions, and researching the social environment. The curriculum specifically includes the role and influence of the media as one of the proposed social issues for Grade 5. The curriculum also explicitly encourages teachers to develop rational and critical thinking through posing questions that require reasoning, investigating evidence, analysing data that support or challenge conclusions, and considering different interpretations. Importantly, students do not interact directly with an AI system during the lesson. The teacher operates the AI tool and presents selected outputs to the class. Students' role is to observe, question, compare and evaluate the content.

Lesson plan

Phase Time Activity
Introduction 10 min The teacher shows two or three images or short texts connected to a familiar situation. Students individually decide: Is this real? How do you know?
Input 15 min The teacher introduces AI in age-appropriate language: what generative AI can do and why AI-generated content can look convincing.
Demonstration 15 min The teacher uses an AI tool live to generate an image or short text based on a simple prompt. Students observe how quickly the content can be created and discuss what the AI actually produces.
Investigation 30 min Groups receive several examples of information/content. They use a verification checklist and decide what they would trust, what they would question and what they would check further.
Reflection 20 min Whole-class discussion: Can we trust an image simply because it looks real? What about a text? Who is responsible if we create or spread false information?

Guiding questions: 

Students work through the following situations with the teacher. The AI is operated only by the teacher.

1. Round 1 — STOP: Don't believe it immediately

The teacher shows students an AI-generated image of a familiar situation — for example, an apparently realistic photograph of an unusual event in Slovenia.

  1. Before revealing anything, students answer:
    1. Does this image look real?

    2. What makes you think so?

    3. Which details make it seem believable?

    4. Is there anything that looks unusual?

  2. The teacher reveals that the image was generated with AI and briefly explains how generative AI can create realistic-looking images.

    1. What made us believe the image was real?

    2. Did we look for evidence, or did we simply trust what we saw?

    3. Could a person use such an image to mislead other people?

Diagnose: Students identify the difference between "This looks real," and "I have evidence that this is real."

Key idea: A realistic appearance is not proof that something is true.


2. Round 2 — QUESTION: Who created it and why?

The teacher presents a short piece of AI-generated news-style text. For example, the teacher asks an AI system to produce a short article about an imaginary event at a Slovenian school.

  1. Students do not interact with the AI. Instead, they analyse the output.
    1. Who wrote or created this?

    2. Where does the information supposedly come from?

    3. Does the text identify a source?

    4. Does it provide evidence?

    5. Can we verify the information?

    6. Why might someone want to publish such information?

    7. Does the text sound certain even though we don't know whether it is true?

  2. Students discover that a text can sound confident, informative and convincing while still lacking a reliable source or evidence.

Key idea: A confident-sounding answer is not necessarily a correct answer.


3. Round 3 — CHECK & COMPARE: Let's verify it

The teacher presents a fictional claim: “A new rule has been introduced in Slovenia: students are no longer allowed to use mobile phones on Saturdays.”

The statement is deliberately simple and plausible enough to prompt students to consider how they would check it.

  1. Students are not asked simply to answer "true" or "false". Instead, they have to design a verification strategy.
    1. Where would you check this information?

    2. Who could you ask?

    3. Would you trust a social media post?

    4. Would you check only one source or several?

    5. What would count as good evidence?

    6. Which source would you trust more?

    7. What would make you change your mind?

  2. The teacher then demonstrates how the claim could be checked using different types of sources.

Students recognise that checking information is a process rather than simply deciding whether something "feels true". 


4. Optional Round — AI vs. Human

  1. The teacher asks an AI system: “Why is it important to preserve cultural heritage?” - The AI generates a short answer.
  2. Students then independently construct their own answer without AI.
  3. The class compares the two.
    1. What did the AI answer well?

    2. What is missing?

    3. Is the AI giving information, an opinion, or both?

    4. Does the AI know our local environment as well as we do?

    5. Which answer would be more useful when presenting our local cultural heritage?

    6. Can an AI-generated answer replace our own thinking?

This connects with the curriculum's emphasis on cultural heritage, students' understanding of the past and present, and their ability to explain and justify their own viewpoints.


Closing activity: Building a classroom rule

At the end of the lesson, the teacher and students jointly create a simple classroom poster:

5 QUESTIONS BEFORE WE SHARE

The poster becomes the main practical outcome of the lesson.

Closing question: If a computer can create an image or text that looks completely real, what should we do before we believe it or share it with other people?

Materials


The teacher's role and the students' role

A central design principle of this lesson is that AI use is teacher-led.

Teacher Students
Operates the AI tool. Observe the AI output.
Creates selected examples. Ask questions about the content.
Demonstrates how AI generates text/images. Identify reasons to trust or question information.
Provides examples and counterexamples Compare different sources.
Models verification strategies Develop their own verification strategy.
Guides discussion Explain and justify their conclusions.

Makes the limitations of AI visible.

Reflect on responsible information sharing.

The students are therefore learning about the capabilities, limitations and social consequences of AI without needing their own access to an AI system.

2. Good Practices & Classroom Examples

2.5 Cross-Curricular: AI-Assisted Lesson Design

AI-Assisted Instructional Design

Subject: Teacher Professional Development  ·  Target Group: Teachers  ·  Duration: 1 workshop (90 min) ·  AI: Teaching WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-DR-2b

(AI Foundations and Applications × Digital Resources, Level 2 – Reflective Implementation):

 

“Teachers can implement prompting, iteration, and refinement strategies to obtain usable AI-generated resources.”


 

AF-TL-2b

(AI Foundations and Applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.”

By the end of the lesson, participants can:

  • formulate a structured prompt for creating a learning activity in their own subject area;
  • evaluate an AI-generated activity for pedagogical relevance, accuracy and suitability for the target group;
  • identify elements of AI-generated material that require human verification or adaptation;
  • improve an initial AI output through prompting, iteration and manual revision;
  • integrate AI into a lesson while maintaining a clear pedagogical role for the teacher and learner.

Take-home message: Generative AI can accelerate lesson preparation, but a usable learning activity rarely comes from a single prompt. Effective AI-supported lesson design is an iterative process in which teachers define the pedagogical goal, generate ideas, evaluate the output, refine it and make the final instructional decisions.

ChatGPT Image Sep 1, 2026 at 04_27_24 PM.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), September 1, 2026.

Content

Generative AI can support teachers with many tasks involved in lesson preparation, including generating activity ideas, adapting materials to different learner groups, producing examples and questions, or suggesting alternative ways of explaining a topic. The Latvian Datorium case study demonstrates this practical approach to teacher AI competence development: teachers first become familiar with AI and its educational applications and then move towards practical experimentation, lesson planning, assessment and differentiated material development. Classroom application is followed by reflection and feedback, making AI use part of an iterative professional learning process rather than a one-time technical exercise.

The usefulness of generative AI for lesson design depends heavily on the information and constraints provided by the teacher. A request such as “Create a lesson about climate change” leaves most pedagogical decisions to the AI system. A more structured prompt can specify the subject, learner age, existing knowledge, learning objective, duration, teaching method, available resources and expected learner output. Adding these elements does not guarantee a good lesson, but it gives the teacher greater control over what the AI produces and makes the resulting material easier to evaluate against the intended learning goal.

Even a detailed prompt can produce material that is factually inaccurate, pedagogically weak, unrealistic for the available lesson time or unsuitable for a particular group of learners. AI may suggest activities without considering classroom dynamics, assume resources that are unavailable, generate tasks that do not actually measure the stated learning objective, or provide content that appears convincing but requires factual verification. For this reason, AI-generated educational material should be treated as a draft rather than a finished teaching resource.

The central learning process in this activity is therefore not simply learning how to write a “good prompt”. Teachers work through a repeated cycle of define → prompt → evaluate → refine → verify. They begin with a real learning objective from their own teaching context, use AI to propose an activity, evaluate the result against pedagogical criteria, refine the prompt where necessary and finally make their own modifications before the activity is considered classroom-ready. This reflects the Datorium case study's emphasis on practical experimentation, co-creation, classroom application and reflection, while keeping pedagogical responsibility and validation with the teacher.

Lesson plan

Phase Time Activity
Introduction 10 min Participants identify one lesson-preparation task for which AI could potentially be useful. The facilitator introduces AI as an assistant/co-creator rather than an autonomous lesson designer.
Input 15 min Participants compare a simple prompt with a structured educational prompt and examine how additional context changes the generated learning activity.
Exploration 15 min Each participant selects a real learning objective from their subject and asks an LLM to generate a short classroom activity addressing it.
Adaptation 35 min Participants work through two or three rounds of prompt → evaluate → refine → verify, progressively adding pedagogical constraints and checking the resulting activity.
Reflection 25 min Participants compare their original AI output with the final classroom-ready version and identify which improvements came from prompting and which required their own professional judgement.

Guiding questions: from AI output to classroom-ready activity

  1. Round 1 — Define the pedagogical purpose:

    Before prompting the AI, define:

    • Who are the learners?
    • What should they know or be able to do?
    • What prior knowledge do they have?
    • How much time is available?
    • What should learners actually do during the activity?

    Generate an initial activity and compare it with the intended learning objective.

    • Diagnose: Does the activity actually help learners achieve the objective, or does it merely relate to the same topic?
  2. Round 2 — Add classroom constraints:

    Refine the prompt by adding realistic conditions such as class size, available materials, lesson duration, learner level, teaching method or accessibility requirements.

    Generate the revised activity and compare it with the first version.

    Compare: Check whether the generated dialogue follows each part of the instruction.
    • Diagnose: Which changes improved the activity? What classroom conditions has the AI still failed to consider?
  3. Optional round 3 — Verify and take back control:

    Evaluate the final AI-generated activity for:

    • factual accuracy;
    • alignment with the learning objective;
    • suitability for the learner group;
    • realistic timing;
    • clarity of instructions;
    • accessibility and inclusion;
    • teacher and learner roles.

    Participants then make at least one manual change without asking AI to revise it.

    The final comparison is therefore not “Which prompt produced the best AI answer?” but rather “What did the teacher have to contribute to turn an AI-generated idea into a usable learning activity?”

Closing question (for the Reflect phase): At which point did the AI-generated activity require more than simply improving the prompt, and what does this tell you about the teacher’s role in AI-supported lesson design?

Follow with: If an AI-generated learning activity appears complete and well structured, how can you decide whether it is actually pedagogically appropriate and ready to use with learners?

Materials

2. Good Practices & Classroom Examples

2.6 History: Reconstruction of the past

A Day in Medieval Riga: What Can Sources Really Tell Us?

Subject: History Grade  ·  Target Group: 7 (age 13–14)  ·  Duration: 1 double lesson ·  AI: Learning WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-TL-2b

(AI foundations and applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.”


 

AF-FC-2b

(AI foundations and applications × Facilitating Learners’ (AI) Digital Competence, Level 2 – Reflective Implementation):

 

“Teachers can implement guidance that helps learners develop practical AI skills, adapting the tasks to learners’ prior knowledge and the AI tools available.”

By the end of the lesson, participants can:

  • extract information about medieval urban life from written, visual and archaeological sources;
  • distinguish between evidence, reasonable inference and speculation;
  • explain aspects of trade, crafts and everyday life in medieval Riga;
  • combine evidence from several sources to con 

    struct a historical recon struction;

  • formulate prompts that require AI to work from supplied historical evidence;
  • analyse information to understand the importance of reliable sources.

Take-home message: What is shown is not necessarily what happened. Historical reconstructions combine evidence, reasonable inference and speculation. Understanding the past means recognising the difference between what we know from sources, what we can reasonably conclude, and what has been imagined to fill the gaps and to tell a better story.

ChatGPT Image Sep 2, 2026 at 01_30_35 PM.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), September 2, 2026.

Content

Historians cannot directly observe everyday life in the Middle Ages. Instead, they reconstruct it from traces that people left behind: written documents, buildings, objects, images, archaeological finds and other evidence. Different sources answer different questions. A trade record may reveal what goods entered a city, while an archaeological object may tell us something about what people owned or used. Neither source alone can recreate an entire day in someone's life. 

Medieval Riga provides a useful case study. Riga developed rapidly after the beginning of the 13th century and became an important centre connecting trade between western Europe, Livonia and lands further east. By the late 13th century, it had become a major Hanseatic trading city. Trade and crafts supported the growth of merchant and craft organisations such as guilds. 

Modern media can make the past feel much more complete than the surviving evidence actually allows. AI-generated history content such as Chloe VS History, for example, presents vivid “time-travel” scenes in which viewers appear to visit historical settings such as London during the Black Death. Clothing, streets, buildings, people and everyday activities are presented as if they were being directly observed. Yet these scenes are reconstructions: some details may be based on historical evidence, some may be reasonable inferences, and others may be speculative or inaccurate. 

The same issue appears in historical films. A Knight's Tale creates a recognisably medieval world through armour, tournaments, clothing, architecture and social roles, while also deliberately introducing modern music, language and behaviour. It therefore provides a useful reminder that something can look convincingly historical without representing exactly what happened. 

Generative AI works in a similar way. It can combine fragments of historical evidence into a coherent reconstruction of a day in medieval Riga, but it may also fill gaps with details that were never present in the sources. Some additions may be historically plausible; others may belong to another region, century or social group.

Students therefore need to distinguish between three levels of certainty: 
Evidence – what historical sources directly support.
Inference – what can reasonably be concluded from that evidence.
Speculation – what has been imagined to fill the gaps.

Lesson plan

Phase Time Activity
Hook 10 min Watch a Chloe VS History medieval AI video. Students identify what makes it believable and question which details are actually supported by evidence.
Input 10 min Introduce evidence – inference – speculation and practise distinguishing them.
Exploration 15 min Students analyse sources about medieval Riga without AI and establish what can genuinely be known.
Adaptation 1 20 min AI reconstructs a day in medieval Riga from limited evidence. Students identify where it fills gaps.
Adaptation 2 20 min More sources and stricter prompting are added. Students compare the two reconstructions.
Reflection 25 min Students transfer the same critical-thinking framework to a medieval historical film.

Guiding questions: from AI output to classroom-ready activity

Materials

2. Good Practices & Classroom Examples

2.7 Informatics / Natural Science: AI for a Real School Problem

Can AI Help Us Sort Waste at School?

Subject: Informatics/Natural Science  ·  Target Group: 7 (age 13–14)  ·  Duration: 1 double lesson ·  AI: Learning WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-TL-2b

(AI foundations and applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers understand basic principles of AI and machine learning and can demonstrate them through practical examples.”


 

AF-FC-2b

(AI foundations and applications × Facilitating Learners’ (AI) Digital Competence, Level 2 – Reflective Implementation):

 

“Teachers can guide students in experimenting with AI tools, testing their outputs and critically reflecting on their reliability.”

By the end of the lesson, students can:

  • explain in simple terms how an AI model learns from examples
  • train and test a simple image-classification model
  •  identify cases in which the model produces incorrect results
  • explain how the quality of examples influences AI performance
  • improve the model based on testing; • assess whether the AI solution would be useful and reliable in a real school situation.

Take-home message: AI can help with real-life problems, but it does not automatically understand what it sees. It learns from examples provided by people, and its results must therefore be tested and evaluated.

image.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), September 7, 2026.

Content

Waste sorting is an everyday issue in many schools. Students use paper, plastic bottles, food packaging and other materials every day, but these items do not always end up in the correct recycling bin.

This familiar situation provides a simple way to explore how artificial intelligence works. Students investigate whether an AI system could recognise an object and recommend which recycling bin it belongs in.

During the activity, students create a basic image-classification model with categories such as paper, plastic and other waste. They provide the AI with several examples, train the model and then test it with new objects.

The activity demonstrates an important principle of machine learning: the AI learns patterns from the examples it receives. If the examples are too similar, incomplete or poorly selected, the model may produce unexpected results.

Students therefore do more than simply create a working model. They deliberately test it with difficult examples, investigate its mistakes and try to improve it. For example, they may discover that a crushed plastic bottle is classified differently from a normal bottle, or that the background of an image affects the result.

The lesson connects a basic AI concept with a real school situation. At the same time, it encourages students to think critically about reliability and human responsibility: even if an AI system gives a recommendation, people still need to decide whether the result makes sense.

Lesson plan

Phase Time Activity
Hook 10 min The teacher shows several everyday waste objects and asks students where they belong. The class discusses whether AI could make the same decision automatically.
Input 10 min The teacher introduces image classification and explains that AI learns from examples. The selected AI tool is briefly demonstrated.
Exploration 20 min In small groups, students create categories such as paper, plastic and other waste, prepare examples and train their first model.
Adaptation 1 20 min Students test the model with new objects and record where it succeeds and where it makes mistakes.
Adaptation 2 20 min Students improve the training examples, retrain the model and compare the new results with the first version.
Reflection 10 min Students discuss whether their AI system would be reliable enough for real use at school and what its limitations are.

Guiding questions: from AI experiment to real-life application

Materials

2. Good Practices & Classroom Examples

2.8 Mathematics: Can the Machine Actually Calculate?

A First Encounter with AI in the Primary Maths Classroom - Speech Recognition with the „Recheneule“

Subject: Mathematics   ·   Grade: 2 (age 7–8)   ·   Duration: 1 lesson (50 min)   ·   AI: Learning ABOUT AI

Real-school example: This exemplar is modelled on the documented practice of the Volksschule Hagenberg im Mühlkreis (Upper Austria), an official KI-Pilotschule of the Austrian Federal Ministry of Education. From the second grade, its teachers introduce AI playfully to prepare children for its chances and risks (director Martina Ketterer-Hager). A central element is the digicase / DLPL project, in which children practise algorithmic thinking - including the maths speech-recognition game „Die schlaue Recheneule“, developed with the PH Linz. This lesson builds one class period around that real tool. It aligns with the Ministry’s KI-Initiative and the four-direction compass Verstehen, Anwenden, Reflektieren, Mitgestalten.

Take-home message: Children don’t just use the tool - they watch it. By doing their sums out loud with the „Recheneule“ and seeing when it hears the right number and when it gets confused, second-graders make their very first discovery about AI: it is a helpful but imperfect machine, and the child is still the one who knows the right answer. This lays the foundation for a lifelong habit of checking, not just trusting.

ChatGPT Image 8. Sept. 2026, 13_34_19.png

Fig. 1: Image generated with ChatGPT (GPT Image 2), September 8, 2026.

Content

Young children meet artificial intelligence long before they can explain it: voice assistants at home, cameras that find faces, apps that finish their sentences. The Austrian KI-Basiscurriculum and the Ministry’s KI-Initiative both stress that AI literacy should begin early and playfully, and that it belongs as much in the primary classroom as in later grades. The Volksschule Hagenberg im Mühlkreis, a KI-Pilotschule, does exactly this: from the second grade its teachers introduce AI through games, with the digicase / DLPL project at the centre, so that children build a first, concrete sense of what these machines can and cannot do.

The „Recheneule“ (the “clever calculating owl”) is one of these games, and it fits the maths lesson perfectly. It uses speech recognition - a form of AI - o that a child can say a number out loud and the program tries to recognise it and use it in a calculation. The tool is deliberately simple and, as its makers note, still experimental: whether it hears correctly depends on the microphone, how clearly the child speaks, and how quiet the room is. For a seven-year-old this is not a flaw to hide but the whole lesson. When the owl shows the wrong number, the child can see with their own eyes that the clever machine did not understand — and that they, the child, still know that three plus four is seven.

This is what makes the tool valuable for teaching ABOUT AI at this age. Children are not asked to understand how speech recognition works inside; they are asked to observe it, to make it work better by speaking clearly and being close to the microphone, and to notice when it fails. In the language of the KI-Basiscurriculum, this is the first step from Verstehen (a machine is listening, not a person) to Reflektieren (it doesn’t always get it right, and I can check it). The teacher keeps the mathematics in the foreground: the point of the sum is still the sum, and the owl is a playful partner whose answers the child learns to check.
In this lesson, children predict whether the owl will hear them, try it, look at what it shows, and talk about why it sometimes gets it wrong — a simple, age-appropriate version of the predict → try → check → talk cycle used throughout these exemplars. The aim is a first, cheerful experience of AI as a useful helper that still needs a human to check it.

Lesson Plan 

Phase

Time

Activity

Warm-up

8 min

On the carpet, the teacher asks where the children have heard a machine “listen” to a voice (phone, speaker, tablet). Together they agree a simple idea: a computer can try to hear us, but it is not a person and can make mistakes.

Show

7 min

The teacher demonstrates the „Recheneule“ once on the big screen, saying a number clearly and then mumbling one, so the class sees the owl hear right and then wrong. The class guesses why.

Try (pairs)

20 min

In pairs at a tablet with a headset, children take turns: one says a simple sum answer aloud (e.g. “three plus four” → says “seven”), the other watches what number the owl shows and ticks a smiley (heard right) or a straight face (heard wrong) on a simple sheet.

Check

8 min

For every “heard wrong”, the pair works out the right answer themselves and decides together: did the child say it wrong, or did the owl hear it wrong? They circle who made the mistake.

Talk

7 min

Back on the carpet: What helped the owl hear us? (speak clearly, be close, be quiet). When the owl was wrong, did we still know the right answer? Who is the boss of the sum — the owl or us?

Guiding questions: predict → try → check → talk

Pairs work through these short rounds during the “Try” phase. The wording is kept simple for the teacher to say aloud; younger or faster pairs can do more rounds.

1.    Round 1 - will the owl hear me?

2.    Round 2 - who made the mistake?

3.    Round 3 (extension) - make it hard on purpose:

Closing question (for the Talk phase): When the owl showed the wrong number, did we still know the right answer? So who has to check the sum — the computer, or us?

Follow with: The owl is clever, but it can hear wrong. Where else might a clever machine get something wrong - and what could we always do to check?

Materials

2. Good Practices & Classroom Examples

2.9 Programming: Assistance in selected tasks

Learn to program code without AI writing code

Subject: Programming  ·  Grade: 9 (age 14–15)  ·  Duration: 1 double lesson (90 min) ·  AI: Assistance WITH AI 

AI-related competencies (DUAL.AI.TEACHer Framework) – teacher level Subject-specific learning objectives – student/learner level

AF-TL-2b

(AI foundations and applications × Teaching & Learning, Level 2 – Reflective Implementation):

 

“Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.”


 

AF-FC-2b

(AI foundations and applications × Facilitating Learners’ (AI) Digital Competence, Level 2 – Reflective Implementation):

 

“Teachers can implement guidance that helps learners developing practical AI skills, adapting the tasks to learners’ prior knowledge and the AI tools available.”

 

By the end of the lesson, students can:

  • know how to use AI to do repetitive or supportive tasks in programming, like indenting the code or writing the comments

  • ask an AI model to propose improvements in the code, without giving the code itself, so students can auto-evaluate their knowledge

  • critically evaluate the AI model when asking it how it will optimise the given code

Take-home message: Students need to know the basics of programming if they want to evaluate the results given by a model. So students in the first steps of learning programming must use AI models to do supportive and repetitive tasks related with programming, but not use them to produce code. They can also use the model to ask for possible improvements in the code, so students can test their programming abilities.

imagen.png

Fig. 1: Image generated with Gemini (3.6 Thinking), September 8, 2026.

Content

Programming education is changing in a context where AI systems can generate, explain, and review code. While these tools offer new opportunities for learning, they also introduce an important educational challenge: students need sufficient programming knowledge to evaluate whether an AI-generated suggestion is correct, efficient, or even relevant to the task at hand. Without a basic understanding of programming concepts, students may accept incorrect solutions, overlook errors, or fail to recognise better alternatives. For this reason, learning programming fundamentals remains essential, even when AI tools are available.


In introductory programming courses, AI should be positioned as a learning support tool rather than as a substitute programmer. Students can use AI effectively for repetitive or low-level tasks that do not reduce the cognitive effort involved in learning to program. Examples include formatting and indenting code, generating documentation comments, explaining error messages, summarising syntax rules, or helping students understand the purpose of specific programming constructs. These uses allow students to focus their attention on the underlying logic, algorithms, and problem-solving processes that are central to programming competence.


Another productive use of AI is code review. Instead of asking an AI model to write a solution, students can write their own code first and then ask the model to identify potential improvements. The AI might suggest clearer variable names, better code organisation, reduced redundancy, improved readability, or alternative approaches that the student can evaluate independently. This transforms the AI into a feedback partner, allowing students to compare their own reasoning with external suggestions and assess the strengths and weaknesses of their solutions.


However, AI-generated feedback should not be accepted uncritically. Models may recommend changes that are unnecessary, inefficient, or based on incorrect assumptions about the code's purpose. A central goal of this lesson is therefore to develop students' ability to critically evaluate AI suggestions. Students will learn to ask an AI model not only what improvements it proposes, but also why it considers those changes beneficial. By comparing the model's reasoning to established programming principles, students can judge whether the proposed optimisation genuinely improves the code or merely represents a different design choice.


This lesson treats AI as a tool that supports reflection, self-assessment, and code quality improvement rather than code generation. Students practice writing their own programs, using AI to perform supportive tasks, requesting feedback on their work, and critically analysing the quality of the responses they receive. The ultimate objective is to help students become both better programmers and more informed users of AI systems.


Lesson plan

Phase Time Activity
Introduction 10 min The teacher presents two short code solutions to the same problem, one written by a student and one generated by an AI model. Students discuss which solution appears clearer, more efficient, or easier to understand.
Input 10 min The teacher introduces appropriate uses of AI in programming education, emphasising the distinction between support tasks, code review, and code generation.
Exploration 15 min Students work in pairs on a small programming exercise and identify which tasks could appropriately be delegated to an AI assistant and which should remain their responsibility.

Improvement Cycle

35 min Students write their own solution and complete 2 to 3 rounds of write → ask → evaluate → revise, using AI only for supportive tasks and improvement suggestions.
Reflection 20 min Whole-class discussion about the usefulness and limitations of AI feedback, focusing on how students determined whether suggestions were valuable or not.

Guiding questions: layering complexity

Students work through as many rounds as time allows during the 35-minute Improvement Cycle. Each round follows the same four-step process: Write → Ask → Evaluate → Revise

  1. Round 1 — Using AI for Supportive Tasks
    • Write: Complete a short programming exercise independently.
    • Ask: Request support from the AI for a non-programming task related to the code, such as:
      • improving indentation,
      • generating comments,
      • explaining an error message,
      • describing what a function does.
    • Evaluate: Does the AI explanation accurately describe the code? Are the generated comments useful and understandable?
    • Revise: Incorporate any helpful changes while ensuring that the program logic remains entirely your own work.
  2. Round 2 — Requesting Improvement Suggestions
    • Write: Review your completed program and identify parts that could potentially be improved.
    • Ask: Request improvement suggestions without allowing the AI to rewrite the code. For example:

"Suggest three ways to improve this program's readability and maintainability, but do not provide replacement code."

    • Evaluate: Which suggestions seem valuable? Which suggestions would have little impact on the program?
    • Revise: Implement any improvements you agree with and test whether the program still behaves correctly.
  1. Round 3 — Critically Evaluating Optimisation Advice
    • Write: Select a section of your code that performs a repetitive task or uses loops.
    • Ask: Request optimisation suggestions and require the AI to justify its recommendations.

Example prompt:

"Explain how this code could be optimised and why each proposed optimisation would improve performance or readability. Do not provide the optimised code."

Reflection Questions:

Closing question (for the Reflect phase):

How much programming knowledge is necessary to determine whether an AI suggestion actually improves a program, and what risks arise when users rely on AI-generated advice without understanding the code themselves?

Materials

2. Good Practices & Classroom Examples

2.10 Art: Fingerprints of Style

Whose Brush, Whose Voice? Detecting Artists' Characteristics with AI

Subject: Art & Literature  ·  Target Group: 9 (age 14–15)  ·  Duration: 1 double lesson  ·  AI: Learning WITH AI

 

Content

Art historians and literary scholars rarely have direct access to an artist's intentions. Instead, they reconstruct an artist's “style” by comparing many works: brushstroke, colour and composition in painting, or word choice, imagery and rhythm in poetry. A single painting or poem can hint at a preference, but only a pattern that recurs across several works turns that preference into a genuine characteristic.

Vincent van Gogh provides a useful case study. Over roughly a decade of intense production, he developed instantly recognisable characteristics: thick impasto, short directional brushstrokes, and bold, often complementary colour contrasts. Because these features recur across hundreds of paintings, scholars can describe them as a real stylistic signature rather than a one-off choice.

AI image tools can now generate convincing pictures “in the style of Van Gogh” in seconds. These images often reproduce recognisable surface markers — swirling brushwork, yellow-and-blue palettes — but they may also exaggerate or invent details that were never characteristic of his actual body of work, or blend in features borrowed from other artists entirely.

The same issue appears with text. AI writing tools can imitate a poet such as Emily Dickinson by reproducing surface markers — dashes, slant rhyme, short lines — without necessarily capturing the thematic and structural qualities that scholars consider genuinely characteristic of her work.

Students therefore need to distinguish between three levels of certainty:

Feature – what is directly observable in one specific work.

Pattern – a feature that recurs across several confirmed works and can reasonably be called characteristic.

Attribution – a claim about meaning, influence or authorship that goes beyond what the works themselves can support.

Lesson plan

Phase

Time

Activity

Hook

10 min

Show two AI-generated “in the style of Van Gogh” images — one closer to his actual technique, one exaggerated. Students guess which is more faithful and discuss why.

Input

10 min

Introduce Feature – Pattern – Attribution and practise distinguishing them using a genuine painting detail.

Exploration

15 min

Students study 3–4 authentic works by one artist/poet without AI and log which features actually recur.

Adaptation 1

20 min

AI writes a style description from a single work or limited prompt. Students identify where it invents unsupported characteristics.

Adaptation 2

20 min

More works and a stricter, source-citing prompt are added. Students compare the two style descriptions.

Reflection

25 min

Students transfer the same critical-thinking framework to a contested art-authentication or attribution case.

 

Guiding questions: from AI output to classroom-ready activity

•     Hook — Show two AI-generated “in the style of Van Gogh” images (one closer to his real technique, one exaggerated/stereotyped):

◦     Discussion question: “What makes this look like a real Van Gogh?”

◦     Introduce the problem: “How do we know which stylistic details are actually characteristic of this artist?” Key principle for this lesson (comparing multiple authentic works)

◦     Observable feature → recurring pattern → attribution/interpretation

•     Input — Close look at a genuine work (e.g. a detail from The Starry Night or a stanza of a Dickinson poem):

◦     Provide statements about the work that need to be categorised into feature, pattern, attribution. Example:

◦     Short, choppy brushstrokes are visible in this painting – feature

◦     Van Gogh regularly used short, directional brushstrokes across his late works – pattern

◦     He used these brushstrokes to express his inner emotional turmoil – attribution

•     Exploration — Group work. Analysing authentic works:

3–4 works by the same artist/poet, e.g.:

A) Two or three reproductions of paintings from the same period of the artist's career.

B) Close-up/detail images showing brushwork or handwriting.

C) Short technique or biography notes from a museum or literary source.

D) A contrasting work by a different artist or poet, for comparison.

Students fill out a worksheet for each work by answering: What features do we observe? What patterns recur across the works? What is still just our interpretation?

•     AI writes a style description:

Prompt used by students: Based only on this one image/poem, describe the artist's characteristic style in approximately 150 words. Only mention features that are actually visible or present in the work.

Students annotate the AI response:

◦     Feature

◦     Reasonable pattern

◦     Unsupported / attribution

•     AI revises its work using more evidence:

Improved prompt: Revise the style description using Sources A–D. After every stylistic claim, cite the specific work that supports it, for example [Work B]. If a claim is a broader pattern inferred from several works, write [Pattern]. Remove any claim that cannot be supported.

Compare:

Version 1 → Version 2

Find:

◦     one detail that became more precise;

◦     one detail AI removed;

◦     one new feature supported by evidence;

◦     one claim that still needs questioning.

•     Reflection/Transfer of knowledge

Show a short, classroom-appropriate example of a contested art-authentication or poem-attribution case.

Discussion question:

◦     Based on the key principles from this lesson, what stands out as feature, pattern, attribution?

◦     If an AI-generated pastiche successfully imitates the surface of a style, does that mean it has captured the artist's actual characteristics?

Final individual reflection:

Students complete:

1.   A stylistic feature becomes a real “characteristic” when…

2.   Before trusting an AI's description of an artist's style, I should…

3.   AI can imitate the surface of a style, but…

Materials

•     Projector/interactive whiteboard with speakers.

•     Two AI-generated “in the style of …” images for the hook.

•     Detail image or excerpt from a genuine work by the chosen artist/poet.

•     Feature–Pattern–Attribution mini worksheet.

•     Source pack: 3–4 authentic works (images/poem texts) plus short technique or biography notes.

•     Source-analysis worksheet.

•     Student devices with access to an approved generative AI tool (text and/or image).

•     AI prompt and comparison worksheets for Rounds 1 and 2.

•     Example of a contested authentication/attribution case for the reflection.

2. Good Practices & Classroom Examples

2.11 Engineering: Trust, but Verify the Circuit

Does It Actually Work? Checking AI's Circuit Explanations Against Real Measurements

Subject: Physics / Technology (Basic Engineering)  ·  Target Group: 8 (age 13–14)  ·  Duration: 1 double lesson  ·  AI: Learning WITH AI

 

Content

Engineers and physicists never simply trust a circuit diagram or a written explanation of how a circuit works. Instead, they build it, measure it, and calculate the expected values from first principles — Ohm's Law and Kirchhoff's laws. A diagram can look entirely correct on paper, but only a real measurement confirms whether current actually flows the way it is claimed to.

A simple series circuit — a battery, a resistor and an LED — provides a useful case study. Ohm's Law (V = I × R) predicts exactly how much current should flow for a given voltage and resistance. Because the components' real-world behaviour can vary slightly (tolerance, internal resistance, temperature), the measured current is usually close to, but not always identical to, the calculated prediction — which is itself a useful lesson in the difference between a model and reality.

AI tools can now generate circuit diagrams and explanations of circuit behaviour in seconds. These explanations often sound fluent and confident, and may even include a full Ohm's Law calculation — but they can also contain real errors: an incorrect resistor value, a broken or incomplete loop, a component connected with the wrong polarity, or a plausible-sounding but physically impossible claim about current flow.

This is a well-documented weak spot: language models are not calculators, and they can produce technical-sounding explanations that do not hold up when a circuit is actually built and tested. A confident tone is not evidence that a technical claim is correct.

Students therefore need to distinguish between three levels of certainty:

Measurement – the value a multimeter actually reads on the real, built circuit.

Calculation – the value predicted from Ohm's/Kirchhoff's laws using the known component values.

Claim – what an AI tool states about the circuit's behaviour, which must be checked against the other two before it can be trusted.

Lesson plan

Phase

Time

Activity

Hook

10 min

Show an AI-generated explanation of a simple LED circuit that contains a hidden error (e.g. a wrong resistor value). Students guess whether it is correct and why.

Input

10 min

Introduce Measurement – Calculation – Claim and practise Ohm's Law on a simple worked example.

Exploration

15 min

Students build a simple series circuit on a breadboard and measure voltage, current and resistance without AI.

Adaptation 1

20 min

AI predicts/explains the circuit's behaviour from a diagram or description alone. Students compare its claims to their own measurements and calculations.

Adaptation 2

20 min

Students give AI exact component values and require a step-by-step Ohm's Law calculation. They compare the two AI responses.

Reflection

25 min

Students transfer the same verification habit to another domain, such as an AI fitness or health calculator.

 

Guiding questions: from AI output to classroom-ready activity

•     Hook — Show an AI-generated explanation of a simple LED circuit containing a hidden error (e.g. a resistor value too low to protect the LED):

◦     Discussion question: “Does this explanation sound correct? How would we actually check?”

◦     Introduce the problem: “An explanation can sound confident and still be wrong — what would prove it right or wrong?” Key principle for this lesson (measurement over confident wording)

◦     Measurement → calculation → claim (to be checked)

•     Input — Worked example with a simple series circuit (battery, resistor, LED):

◦     Provide statements about the circuit that need to be categorised into measurement, calculation, claim. Example:

◦     The multimeter reads 9 mA of current through the circuit – measurement

◦     Ohm's Law predicts 9.1 mA for these component values – calculation

◦     The AI states that this circuit “will work reliably in any configuration” – claim (to verify)

•     Exploration — Group work. Building and measuring a real circuit:

Materials and steps for each group, e.g.:

A) A circuit diagram showing a battery, one resistor and one LED in series.

B) A breadboard, jumper wires and the listed components.

C) A multimeter and a short reference sheet on how to measure voltage, current and resistance.

D) A resistor colour-code chart and the LED's forward-voltage rating.

Students fill out a worksheet answering: What do we measure? What does Ohm's Law predict? Where do the two differ, and by how much?

•     AI predicts/explains the circuit:

Prompt used by students: Given a circuit with a 9V battery, a 470Ω resistor and a red LED in series, explain how much current will flow and why. Show your reasoning.

Students annotate the AI response:

◦     Matches our measurement

◦     Matches our calculation

◦     Unsupported / needs checking

•     AI revises its work using stricter requirements:

Improved prompt: Recalculate the current step by step using Ohm's Law, showing the formula and each substituted value. State any assumptions you make about the LED's forward voltage drop, and flag anything that would need to be measured to confirm.

Compare:

Version 1 → Version 2

Find:

◦     one figure that became more precise;

◦     one assumption AI made explicit;

◦     one value that now matches your measurement;

◦     one claim that still needs independent checking.

•     Reflection/Transfer of knowledge

Show a short, classroom-appropriate example of an AI-generated fitness, nutrition or health calculation (e.g. “your recommended daily calorie intake is…”).

Discussion question:

◦     Based on the key principles from this lesson, what here is a measurement, a calculation, and a claim?

◦     If an AI answer sounds confident and uses correct-looking formulas, does that mean the result is trustworthy?

Final individual reflection:

Students complete:

1.   A technical claim becomes trustworthy when…

2.   Before trusting an AI's explanation of how something works, I should…

3.   AI can sound confident about a technical process, but…

Materials

•     Projector/interactive whiteboard.

•     Breadboards, jumper wires, resistors, LEDs and battery packs (one set per group).

•     Multimeters (one per group).

•     Resistor colour-code chart and LED forward-voltage reference sheet.

•     Measurement–Calculation–Claim mini worksheet.

•     Circuit diagram worksheet and source-analysis worksheet.

•     An AI-generated circuit explanation containing a deliberate error, for the hook.

•     Student devices with access to an approved generative AI tool.

•     AI prompt and comparison worksheets for Rounds 1 and 2.

•     Example of an AI-generated fitness/health calculation for the reflection.

3. AI Pedagogical Competency Framework

The DUAL.AI.TEACHer Competency Framework is the result of a cross-analysis of the DigCompEdu, AI Pioneers and the UNESCO AI Framework for Teachers. Rather than creating new structures from scratch, existing frameworks that already address complementary dimensions for teachers’ professional competences in the age of AI were selected.

3. AI Pedagogical Competency Framework

3.1 Dimensions

The DUAL.AI.TEACHer Framework is structured as a comprehensive two-dimensional matrix that outlines a total of 18 core competencies, each of which emerges from the intersection of the two main dimensions along which the framework is organised. The horizontal dimension defines the main areas that characterise teaching and learning practices, capturing the pedagogical dimension of the framework. The vertical dimension maps the key aspects of AI, reflecting the technological and ethical facets involved in its integration into educational contexts. 

This structure enables the framework to systematically link pedagogical competencies with AI-related knowledge and skills, ensuring that each identified competency reflects both an educational objective and a specific dimension of AI understanding or application. 

DUAL.AI.TEACHer Horizontal dimension

The horizontal dimension of the framework is drawn from the DigCompEdu framework, which identifies the key areas that characterise teaching and learning practices in the digital environments. These areas encompass the fundamental pedagogical activities carried out by teachers, ranging from professional engagement and management of digital resources to the design and delivery of learning experiences, the implementation of assessment strategies, and the promotion of learners’ active engagement and digital competencies. 

By taking into account the AI Pioneers Framework, the DUAL.AI.TEACHer Framework adopts this dimension to ground AI-related competencies in the broader pedagogical practices already recognised as essential for effective teaching, aligning them with those practices. The horizontal dimension is organised into the following six fields:

DUAL.AI.TEACHer Vertical dimension

The vertical dimension of the framework is drawn from the UNESCO AI Competency Framework for Teachers, which originally identifies five key aspects of AI competence relevant for teaching practice: human-centred mindset, ethics of AI, AI foundations and applications, AI pedagogy and AI for professional development.

For the purposes of the DUAL.AI.TEACHer Framework, these five aspects have been reduced to three. Two of the original areas, namely the AI Pedagogy and the AI for Professional Development, were removed, as the horizontal dimension already contained those fields, albeit in a slightly different form. Additionally, the intersection between the vertical and the horizontal dimensions generated redundant competencies. Therefore, it was decided to exclude them from the vertical axis in order to avoid overlaps and ensure conceptual consistency across the matrix.

The vertical dimension is articulated across the following three aspects:

Of the three aspects retained, two are considered cross-cutting aspects of AI (Human-centred mindset & Ethics of AI), as they do not pertain to a specific subject area but rather represent transversal dimensions that influence and permeate the entire spectrum of teaching practices, values, and professional conduct related to AI. The third aspect (AI foundations and applications) is considered a content-specific aspect of AI, as it pertains to the conceptual knowledge and specific technical skills that teachers need to acquire to understand, select and apply AI tools in their teaching practice.

3.1-A-competence-matrix-revised.jpg

Fig. 1: Image generated with 

3. AI Pedagogical Competency Framework

3.2 The DUAL.AI.TEACHer Competence Framework

The intersection of the horizontal and vertical dimensions of the matrix yields a set of eighteen competencies, each of which represents a specific point of convergence between a pedagogical area and an aspect of AI. These competencies define what teachers need to know, understand, and be able to do in order to effectively and responsibly integrate AI into their teaching practice, combining pedagogical competencies with values, ethical awareness, and technical knowledge required by an AI-powered education environment.

DUAL.AI

Professional Engagement

Digital Resources

Teaching & Learning

Assessment

Empowering Learners

Facilitating Learners' (AI) Digital Competence

Human-centred mindset

Teachers can maintain a human-centred perspective when using AI in professional communication, networking, and institutional and professional development

Teachers can ensure human oversight and learner agency in AI-supported processes for selecting, creating, and evaluating digital resources.

Teachers can keep learner agency, human relationships, and teacher judgement at the centre of AI-supported teaching and learning.

Teachers can ensure human accountability and fairness in AI-assisted assessment processes

Teachers can use AI to address diverse learning needs and styles, placing human dignity, equity, and the right to education at the core of every inclusive practice

Teachers can model and cultivate human-centred attitudes towards AI as part of students' reflective digital (AI) competence development

Ethics of AI

Teachers can apply ethical principles and regulatory awareness to their own use of AI in professional practice and institutional contexts

Teachers can identify & mitigate ethical risks in AI-generated or AI-enhanced learning resources.

Teachers can integrate ethical reasoning into decisions on whether and how to use AI for teaching strategies & learning activities

Teachers can ensure that AI-assisted assessment practices meet standards of fairness, transparency, and legal compliance, allowing them to reflect on feedback & improve their assessment practices

Teachers can address the ethical dimensions of AI personalisation, including risks of inequity, data misuse, and exclusion

Teachers can educate students in responsible, critical, and ethically informed use of AI technologies

AI foundations and applications

Teachers can apply foundational AI knowledge to inform professional decisions, engage critically with AI tools, lead AI-related professional learning in institutional contexts

Teachers can critically select & use AI tools for creating, adapting & enhancing digital educational resources based on technical understanding of AI systems

Teachers can apply understanding of how AI systems work to make informed, critical decisions about AI integration in teaching and learning, improving their own teaching practices and instructional design approaches

Teachers can apply AI knowledge to design, implement and evaluate efficient and equitable AI-enhanced assessment practices, and systematically improve their assessment practices

Teachers can select pedagogical strategies and AI tools that support personalisation and inclusion, based on informed understanding of how they function

Teachers can build students' understanding of AI technologies and their societal implications as part of AI competence development


3. AI Pedagogical Competency Framework

3.3 Progression Levels and Operators

Making Competence Development Traceable

The 18 core competencies of the DUAL.AI.TEACHer framework are relatively abstract and provide an overarching structure of the competence field. To facilitate self-assessment or assessment by others and to chart individual competency development, each of the core competencies are operationalised through observable learning objectives, differentiated on three progression levels.

The levels build on one of the most widely used models in education: the revised Bloom’s taxonomy (Anderson & Krathwohl, 2001; Krathwohl, 2002). Its core idea is simple: Cognitive processes develop from remembering and understanding, through applying and analysing, to evaluating and creating. What changes from level to level is therefore not the content or topic, but the cognitive depth of thinking and acting a teacher can demonstrate with it

.

Progression Levels and Operators

This cognitive depth is reflected in the wording of the learning objectives. Each objective is built around a verb (an operator) taken directly from the taxonomy (e.g. recognise, implement, critique). The operator tells you what a teacher at that level is expected to do, and makes the learning objective concrete and observable.

Level

In practice, teachers can...

Cognitive processes (operators)

Level 1

Orientational Awareness

develop an initial professional orientation by learning what AI is, what it can do, and what is at stake. They recognise and recall key AI concepts and tools, give examples, classify them, and summarise the main opportunities and risks in their own words.

Remember
(recognising, recalling)

Basic Understanding (interpreting, exemplifying, classifying, summarising)

Level 2

Reflective Implementation

apply and critically evaluate AI-related practices in their own teaching. They purposefully use AI tools in lessons and assessment, explain and compare different approaches and outputs, compare AI-generated outputs with their own professional judgement, and analyse benefits, limitations, and underlying assumptions.

Higher Understanding
(explaining, comparing)

Apply

(executing, implementing)

Analyse

(differentiating, organising, attributing)

Evaluate

(checking, critiquing)

Level 3

Transformative Leadership

responsibly design and further develop AI-related educational practice while taking a leadership role beyond their own classroom. They evaluate AI practices against pedagogical and ethical criteria, create new AI-enhanced approaches, and support colleagues and their institution in the responsible use of AI.

Create

(generating, planning, producing)

The Logic Behind the Levels

The levels progressively build on each other. Each higher level draws on the cognitive processes of the levels below it. However, following Krathwohl (2002), this is a hierarchy of increasing complexity rather than a set of strictly separated stages. Neighbouring categories may overlap, and the boundaries between levels are fluid rather than sharp. 

While Level 1 provides the necessary foundation and Level 3 represents an extended role in innovation and professional leadership, Level 2 constitutes the primary target dimension for teachers. All teachers should be able to integrate AI purposefully into their own practice, assess its outputs using their professional judgement, and critically reflect on its pedagogical benefits, limitations, and underlying assumptions. Level 2 therefore describes the level of competence required for informed, responsible, and reflective teaching practice in an AI-influenced educational environment.

3. AI Pedagogical Competency Framework

3.4 DUAL.AI.TEACHer Learning Objectives

For each of the eighteen competencies identified, a set of three specific learning objectives (a - c) has been defined in order to translate the competency into concrete and practical outcomes. These learning objectives specify the knowledge, skills, and attitudes that teachers are expected to develop to achieve the corresponding competency, providing a more detailed and practical description of what mastering that competency entails in practice. 3.4-B-competency-card-EI-TL.jpg

Fig. 1: Image generated with ...

By breaking down each competency into clearly defined learning objectives at different levels, the DUAL.AI.TEACHer Framework facilitates its practical application in professional development, training design, and self-assessment, offering educators a concrete blueprint for progressively building and demonstrating their AI-related competencies

Level 1 – Orientational Awareness

LEVEL 1

Professional Engagement

Digital Resources

Teaching & Learning

Assessment

Empowering Learners

Facilitating Learners' (AI) Digital Competence

Human-centred mindset

HM-PE-1a

Teachers can recognise common AI use cases in professional communication and networking and the situations where human judgement, presence, or relationship-building must take precedence.

HM-PE-1b

Teachers can recognise their own AI-related learning needs and tools that support their continuous professional development.

HM-PE-1c

Teachers can summarise how AI is used in their institution for stakeholder engagement and organisational development, including the opportunities and risks it creates for human agency, inclusivity, and educational values.

HM-DR-1a

Teachers can summarise how AI tools can support the selection, creation, customisation, and management of digital resources, including which steps require human oversight and pedagogical judgement.

HM-DR-1b

Teachers can recognise typical quality criteria for AI-generated or AI-enhanced learning materials, including the indicators that signal when a resource needs human review.

HM-DR-1c

Teachers can interpret how learner agency and autonomy can be preserved or undermined by AI-generated or AI-personalised resources, including when learners should be involved in decisions about resource use.

HM-TL-1a

Teachers can recognise, across the ways AI can be used in instruction, which pedagogical decisions must remain in human hands.

HM-TL-1b

Teachers can recognise situations in AI-supported lessons in which learner agency, human interaction, or teacher–learner relationships could be strengthened or weakened, along with the indicators of meaningful pedagogical use of AI.

HM-TL-1c

Teachers can classify AI use in teaching as either supporting pedagogical decision-making or substituting for it, and its implications for accountability and learning quality.

HM-A-1a

Teachers can recognise, among the uses of AI in assessment, which evaluative decisions must remain a human responsibility.

HM-A-1b

Teachers can recognise typical risks in AI-assisted assessment for fairness, transparency, and learner well-being, including the signs that human review of AI-generated results is needed.

HM-A-1c

Teachers can classify AI in assessment as either a support for teacher judgement or a substitute for it, and its implications for accountability and learner trust.

HM-EL-1a

Teachers can recognise, among the ways AI tools can support diverse learners and which learner needs they can and cannot address.

HM-EL-1b

Teachers can recall typical risks of AI-supported differentiation for inclusion and equity, including the signs that an AI tool is restricting rather than expanding a learner's opportunities.

HM-EL-1c

Teachers can interpret how human dignity, equity, and the right to education apply to AI use in inclusive classrooms, including when learner agency and belonging must take precedence over efficiency or standardisation.

HM-FC-1a

Teachers can recognise what it means to model a human-centred, critically informed relationship with AI in front of learners, including how their own practices and statements communicate attitudes about AI.

HM-FC-1b

Teachers can classify dispositions that support learners' reflective AI competence and the classroom situations in which these dispositions can be cultivated.

HM-FC-1c

Teachers can recognise the difference in their own teaching about AI as either transmitting knowledge or cultivating human-centred attitudes, and its implications for their role as a teacher.

Ethics of AI

EI-PE-1a

Teachers can recall key ethical principles that apply to their own professional AI use, as well as the significance of each principle in a professional context.

EI-PE-1b

Teachers can recognise the legal and regulatory frameworks that govern AI use in professional communication, collaboration, and development.

EI-PE-1c

Teachers can recognise situations in their professional practice in which an ethical principle or regulatory requirement is at stake.

EI-DR-1a

Teachers can recall typical ethical risks in AI-generated or AI-enhanced learning resources along with their effects on the learners.

EI-DR-1b

Teachers can recognise legal and regulatory frameworks relevant to AI-generated resources and when they apply.

EI-DR-1c

Teachers can classify sources of ethical risks originating from bias, inaccuracy, or inappropriate AI authorship in a given learning resource.

EI-TL-1a

Teachers can recall ethical principles relevant to AI use in teaching and learning, including their implications for instructional choices.

EI-TL-1b

Teachers can recognise ethical issues specific to AI in classrooms and when they arise in a teaching scenario.

EI-TL-1c

Teachers can summarise the steps of an ethical reasoning process for deciding whether and how to use AI in a lesson (principled decisions vs uncritical adoption or default rejection).

EI-A-1a

Teachers can recall the main fairness risks of AI-assisted assessment and how each can affect learners.

EI-A-1b

Teachers can recognise transparency obligations in AI-assisted assessment, such as the disclosure of AI involvement, the explanation of results, and the ability of the learners to contest or request human review of decisions.

EI-A-1c

Teachers can interpret why educational assessment is treated as high-risk under the legal and regulatory frameworks that apply to AI in assessment.

EI-EL-1a

Teachers can recall the main ethical risks of AI-based personalisation and how each can affect learners.

EI-EL-1b

Teachers can classify the types of learner data used by AI personalisation and how sensitive they are.

EI-EL-1c

Teachers can interpret why personalised educational AI is treated as high-risk under the legal and rights-based frameworks.

EI-FC-1a

Teachers can recall what are the key ethical issues raised by AI that are relevant to students, including the environmental and privacy implications.

EI-FC-1b

Teachers can classify which ethical questions belong in which educational stage, in age- & context-appropriate ways.

EI-FC-1c

Teachers can interpret ethically problematic AI use by learners and why each is an ethical issue rather than only a rule violation.

AI foundations and applications

AF-PE-1a

Teachers can summarise the core concepts of how contemporary AI systems function and which common AI tools are used in professional educational practice.

AF-PE-1b

Teachers can exemplify relevant criteria for evaluating an AI tool for a professional task.

AF-PE-1c

Teachers can interpret why an AI tool's limitations make it the wrong choice in typical situations in their professional work.

AF-DR-1a

Teachers can summarise the core concepts of how the main types of AI tools used for resource creation function, providing correspondent examples.

AF-DR-1b

Teachers can classify the reliability of the outcome for the selected tool type for their corresponding tasks, including avoidable mistakes.

AF-DR-1c

Teachers can recall technical criteria for selecting an AI resource tool and what each entails.

AF-TL-1a

Teachers can summarise the core concepts of how the main types of AI systems used in teaching function, providing respective examples.

AF-TL-1b

Teachers can classify the reliability of the outcome for the selected tool type for their classroom context, including avoidable mistakes.

AF-TL-1c

Teachers can interpret how technical understanding informs realistic decisions about whether, when, and how to integrate AI into a lesson.

AF-A-1a

Teachers can summarise the core concepts of how the main types of AI systems used in assessment function, providing typical examples of each.

AF-A-1b

Teachers can classify what each type of assessment tool actually measures, what it is unable to measure, and in which situations it tends to fail.

AF-A-1c

Teachers can exemplify why particular technical questions about an AI assessment tool are worth to be done.

AF-EL-1a

Teachers can summarise the core concepts of how the main types of AI used for personalisation and inclusion function, including typical examples.

AF-EL-1b

Teachers can classify what each type of tool can and cannot reliably do, as well as the areas where its performance is likely to break down for specific learner groups.

AF-EL-1c

Teachers can classify what an AI tool claims to measure or do versus what it actually measures or does, including why this distinction matters for inclusion.

AF-FC-1a

Teachers can summarise the core concepts of AI that learners should understand, including age-appropriate examples.

AF-FC-1b

Teachers can recognise the technical limits of current AI systems that learners need to grasp.

AF-FC-1c

Teachers can interpret practical AI skills that learners should develop and how the mastery of each looks like at a given educational stage.


Level 2 – Reflective Implementation

LEVEL 2

Professional Engagement

Digital Resources

Teaching & Learning

Assessment

Empowering Learners

Facilitating Learners' (AI) Digital Competence

Human-centred mindset

HM-PE-2a

Teachers can differentiate, in concrete AI-mediated communication and networking situations in their own practice, when to use, adapt, or refrain from AI on the basis of human-centred criteria.

HM-PE-2b

Teachers can organise a personal AI-supported professional development plan that integrates AI tools into their learning routines while safeguarding their growth, autonomy, and peer collaboration.

HM-PE-2c

Teachers can explain human-centred principles in institutional discussions or working groups on AI by proposing adjustments to AI-related procedures, policies, or communication practices.

HM-DR-2a

Teachers can implement AI-generated or AI-enhanced resources into their teaching context, in line with pedagogical and human-centred criteria.

HM-DR-2b

Teachers can compare AI-based tools for creating and selecting digital resources in order to choose the one best suited to a specific instructional purpose.

HM-DR-2c

Teachers can implement ways of involving learners in the evaluation and use of AI-enhanced resources that strengthen learner agency and pedagogical quality.

HM-TL-2a

Teachers can implement AI-supported lessons in which their own pedagogical judgement guides when, how, and why AI is used, preserving learner agency and human interaction.

HM-TL-2b

Teachers can explain the impact of AI use on classroom interaction, participation and learning processes, safeguarding meaningful teacher–learner interactions and peer relationships through instructional strategies.

HM-TL-2c

Teachers can implement ways of shaping how AI is used in lessons and how this strengthens learner agency and pedagogical quality.

HM-A-2a

Teachers can implement AI-assisted assessment tools in their own practice, basing the final evaluative decision on a critical reading of the tools' outputs and on pedagogical and human-centred criteria.

HM-A-2b

Teachers can check AI-generated grades, analytics, or feedback for fair, accurate, and aligned with learning goals, adjusting assessment practices to safeguard equity and meaningful learner support.

HM-A-2c

Teachers can explain to learners and other stakeholders how AI was used in an assessment process and the role of human judgement in the final result, in a way that preserves trust and learner agency.

HM-EL-2a

Teachers can implement AI tools to support learners with different abilities, backgrounds, and preferences, basing the choices on equity, dignity, and inclusive pedagogy.

HM-EL-2b

Teachers can explain how AI-supported differentiation affects learners' participation, autonomy, and sense of belonging in their own practice, safeguarding inclusion and meaningful learning for every student.

HM-EL-2c

Teachers can implement ways to involve learners and, where appropriate, families or support staff in decisions about AI-supported personalisation, strengthening learner agency, trust, and educational equity.

HM-FC-2a

Teachers can execute human-centred AI practices in their own visible practice basing the choices on educational and citizenship goals.

HM-FC-2b

Teachers can implement learning situations to invite learners reflecting on their own attitudes, agency, and responsibilities in relation to AI, while strengthen their critical engagement.

HM-FC-2c

Teachers can organise dialogues with learners about the role of AI in their lives, schools and society, that contributes to informed agency and responsible participation in an AI-permeated world.

Ethics of AI

EI-PE-2a

Teachers can implement ethical principles and regulatory requirements to concrete decisions about their own professional AI use, justified in relation to integrity, transparency, accountability and sustainability.

EI-PE-2b

Teachers can infer principled responses to ethical dilemmas arising in AI-mediated professional engagement.

EI-PE-2c

Teachers can explain institutional AI norms, ethically grounded practices, and how their own conduct contributes to a trustworthy professional culture.

EI-DR-2a

Teachers can differentiate which ethical risks are present in AI-generated or AI-enhanced resources in their own practice and how serious each is, using structured criteria.

EI-DR-2b

Teachers can implement concrete strategies to mitigate the risks arising from AI-generated resources.

EI-DR-2c

Teachers can explain transparently to learners and colleagues about the AI involvement and the residual risks in a resource, in a way that strengthens trust and supports learning.

EI-TL-2a

Teachers can implement ethical principles in concrete decisions about AI use in their own teaching, justified in terms of pedagogical benefit and ethical risk.

EI-TL-2b

Teachers can infer principled, defensible responses to AI-related ethical dilemmas in instruction.

EI-TL-2c

Teachers can explain to learners and colleagues the ethical reasoning behind their AI-related instructional choices, in a way that shapes classroom culture and trust.

EI-A-2a

Teachers can critique, in AI-assisted assessment tools and outputs in their own practice, for standards of fairness, transparency, and legal compliance, adjusting their practice accordingly.

EI-A-2b

Teachers can implement transparency practices in their own assessment while strengthening trust and learner understanding.

EI-A-2c

Teachers can infer the changes needed in their assessment practice based on AI-assisted feedback to ensure fairness.

EI-EL-2a

Teachers can critique AI-based personalisation tools and practices in their own context about risks of data misuse, discrimination, and curricular confinement.

EI-EL-2b

Teachers can implement concrete safeguards in their own personalisation practice that protect learners.

EI-EL-2c

Teachers can explain transparently to learners and families how AI personalises learning, what data are used, and what rights learners have, strengthening trust and equity.

EI-FC-2a

Teachers can implement learning activities that develop learners' ethical reasoning about AI, justified by the methodological choices.

EI-FC-2b

Teachers can infer designs to help learners weighing up conflicting values, based on the assumptions arising from the ethical reasoning they demonstrate in using AI.

EI-FC-2c

Teachers can infer learning opportunities in learners' ethically questionable use of AI.

AI foundations and applications

AF-PE-2a

Teachers can compare AI tools for their own professional tasks using foundational AI knowledge to select tools based on their technical capabilities and limitations.

AF-PE-2b

Teachers can implement adjustments to their use of AI tools that reflect how the underlying system actually works and improve the quality of outcomes.

AF-PE-2c

Teachers can explain technical concepts about AI tools to colleagues in accessible language, contributing to informed collegial dialogue about AI in professional practice.

AF-DR-2a

Teachers can compare AI tools for specific resource-creation tasks in their own practice to select tools based on their technical capabilities and limitations.

AF-DR-2b

Teachers can implement prompting, iteration, and refinement strategies to obtain usable AI-generated resources.

AF-DR-2c

Teachers can explain to colleagues when a tool should be replaced by a different tool, approach, or non-AI resource.

AF-TL-2a

Teachers can compare AI tools for specific teaching and learning purposes in their own practice to select tools based on their technical capabilities and limitations.

AF-TL-2b

Teachers can organise lesson elements that embed AI use with clear pedagogical roles and strategies to verify the output elements.

AF-TL-2c

Teachers can implement adjustments to their instructional design approaches and use of AI tools based on their technical performance.

AF-A-2a

Teachers can compare AI tools for specific assessment tasks in their own practice, based on the tool's technical capabilities, limits, and reliability.

AF-A-2b

Teachers can implement assessment workflows that combine AI outputs with structured human review, recognising when an AI output is too unreliable.

AF-A-2c

Teachers can implement refinements to their use of AI assessment tools based on their technical performance.

AF-EL-2a

Teachers can compare AI tools considering their technical capabilities and performance for personalisation, accessibility and inclusion of specific learners involved.

AF-EL-2b

Teachers can implement customised configurations of AI tools for inclusive use, focused on learners who need it most.

AF-EL-2c

Teachers can differentiate how AI tools actually perform across different learners in their practice, switching tools or combining AI with non-AI alternatives.

AF-FC-2a

Teachers can organise learning activities that build learners' conceptual understanding of how AI works.

AF-FC-2b

Teachers can implement guidance that helps learners developing practical AI skills, adapting the tasks to learners' prior knowledge and the AI tools available.

AF-FC-2c

Teachers can explain societal implications of AI from a technical perspective and ways the learners can engage with.


Level 3 – Transformative Leadership

LEVEL 3

Professional Engagement

Digital Resources

Teaching & Learning

Assessment

Empowering Learners

Facilitating Learners' (AI) Digital Competence

Human-centred mindset

HM-PE-3a

Teachers can plan new human-centred practices for AI-mediated professional communication and networking within their team or institution, modelling them and mentoring colleagues in their use.

HM-PE-3b

Teachers can generate innovative formats of AI-supported professional learning that strengthen self-direction and collective professional growth across the wider educational community.

HM-PE-3c

Teachers can plan institutional change processes that embed human-centred AI use in professional engagement and influence the culture of their organisation.

HM-DR-3a

Teachers can plan workflows for the AI-supported creation and curation of digital resources that embed human oversight, learner voice, and pedagogical accountability as standard practice, and share them with colleagues.

HM-DR-3b

Teachers can generate quality criteria, review protocols, or guidance documents for AI-enhanced resources and disseminate them within their team, school, or wider professional community.

HM-DR-3c

Teachers can plan collaborative practices that transform how AI-enhanced digital resources are produced, shared, and improved across the educational community.

HM-TL-3a

Teachers can generate innovative AI-supported teaching approaches that explicitly centre learner agency, human relationships, and teacher judgement, trialling them and sharing the outcomes with colleagues.

HM-TL-3b

Teachers can generate pedagogical principles and produce according lesson formats, or classroom protocols for human-centred AI use in teaching and learning, and disseminate them within their team or professional community.

HM-TL-3c

Teachers can plan collaborative practice development that transforms how AI is used in teaching and learning across their school or wider educational context.

HM-A-3a

Teachers can plan AI-assisted assessment practices that explicitly safeguard human accountability, fairness, and pedagogical purpose, and share these designs with colleagues.

HM-A-3b

Teachers can generate principles, review protocols, or guidance for the responsible use of AI in assessment within their team, school, or wider professional community.

HM-A-3c

Teachers can plan collaborative development of AI-supported assessment practices that transform how assessment is conducted in their educational context.

HM-EL-3a

Teachers can generate inclusive AI-supported learning environments that explicitly centre dignity, equity, and learner agency for diverse groups, and share these designs with colleagues.

HM-EL-3b

Teachers can generate inclusive-practice principles, accessibility checklists, or differentiation protocols for AI use within their team, school, or wider professional community.

HM-EL-3c

Teachers can plan collaborative initiatives that transform how AI is used to advance equity and the right to education in their context.

HM-FC-3a

Teachers can generate classroom approaches that explicitly model and cultivate human-centred attitudes towards AI across subjects or year groups, trialling them and sharing them with colleagues.

HM-FC-3b

Teachers can generate principles, classroom rituals, or pedagogical formats for nurturing critical AI citizenship in learners, and disseminate them within their team or professional community.

HM-FC-3c

Teachers can plan collaborative initiatives that transform how learners' human-centred AI dispositions are cultivated in their educational context

Ethics of AI

EI-PE-3a

Teachers can generate ethical guidelines, decision-aids, or codes of practice for AI use in professional engagement, and share them within their team or institution.

EI-PE-3b

Teachers can plan institutional review or accountability mechanisms that strengthen ethical and regulatory compliance in professional AI use.

EI-PE-3c

Teachers can plan the development of shared ethical norms and responsible AI practices within their educational organisation, shaping a culture in which AI use is principled, transparent, and accountable.

EI-DR-3a

Teachers can produce review protocols, risk check-lists, or plan mitigation workflows for AI-generated learning resources, and share them with colleagues.

EI-DR-3b

Teachers can plan team or departmental practices that turn risk identification and mitigation into a routine part of resource development and curation.

EI-DR-3c

Teachers can plan collaborative initiatives that transform how ethical risks in AI-generated resources are handled across their educational context.

EI-TL-3a

Teachers can generate structured ethical decision-aids for AI use in teaching and learning, and share them with colleagues.

EI-TL-3b

Teachers can plan classroom or departmental practices that turn ethical reasoning about AI into a routine part of instructional planning, and disseminate them within their professional community.

EI-TL-3c

Teachers can plan collaborative work that transforms how ethical reasoning is integrated into AI-supported teaching across their educational context.

EI-A-3a

Teachers can produce fairness-, transparency-, and compliance-check protocols for AI-assisted assessment in their context, and share them with colleagues.

EI-A-3b

Teachers can plan structured reflection routines that turn assessment feedback into systematic practice improvement.

EI-A-3c

Teachers can plan collaborative initiatives that transform how fairness, transparency, and legal compliance are upheld in AI-assisted assessment across their educational context.

EI-EL-3a

Teachers can produce ethical review protocols, data check-lists, or safeguard frameworks for AI-based personalisation in their context, and share them with colleagues.

EI-EL-3b

Teachers can plan team or school practices that turn risk identification, data minimisation, and human review into routine elements of personalised AI use.

EI-EL-3c

Teachers can plan collaborative initiatives that transform how the ethical risks of AI personalisation are handled in their educational context.

EI-FC-3a

Teachers can plan longer learning sequences, project formats, or cross-subject units that systematically develop learners' ethical AI reasoning and environmental responsibility.

EI-FC-3b

Teachers can generate age-appropriate frameworks, vocabularies, or routines for cultivating learners' ethical AI agency.

EI-FC-3c

Teachers can plan collaborative initiatives that transform how learners' ethical AI agency is developed across their educational context.

AI foundations and applications

AF-PE-3a

Teachers can plan evaluation procedures, tool-selection frameworks, or onboarding materials that help colleagues apply foundational AI knowledge to professional decisions, and share them within the institution.

AF-PE-3b

Teachers can generate AI-related professional learning formats that build AI literacy across their team or school.

AF-PE-3c

Teachers can generate institutional capacity-building on AI that transforms how foundational AI knowledge informs practice across their educational context.

AF-DR-3a

Teachers can plan technical workflows or guidance documents for AI-supported resource creation and share them with colleagues.

AF-DR-3b

Teachers can generate novel uses of AI resource tools that exploit their actual capabilities rather than novelty, trialling them and disseminating the results within their professional community.

AF-DR-3c

Teachers can plan capacity-building on AI tools for resource design that transforms how AI is used for educational resources in their context.

AF-TL-3a

Teachers can generate technically grounded instructional formats or produce lesson templates for AI-integrated teaching and share them with colleagues.

AF-TL-3b

Teachers can generate novel AI-supported teaching approaches that exploit the actual capabilities of current AI systems, trialling them and disseminating the results within their professional community.

AF-TL-3c

Teachers can plan capacity-building on AI in teaching and learning that transforms how foundational AI knowledge shapes instructional design across their educational context.

AF-A-3a

Teachers can plan technical evaluation protocols, tool-validation routines, or workflow templates for AI-enhanced assessment in their context, and share them with colleagues.

AF-A-3b

Teachers can plan systematic practices for tracking and improving the technical performance of AI assessment tools, and disseminate them within their team or school.

AF-A-3c

Teachers can plan capacity-building on AI in assessment that transforms how foundational AI knowledge shapes assessment practice across their educational context.

AF-EL-3a

Teachers can generate technical guidance, tool-selection frameworks, or configuration templates for inclusive AI use in their context, and share them with colleagues.

AF-EL-3b

Teachers can generate novel uses of AI tools for personalisation and accessibility that are grounded in a realistic understanding of what current systems can and cannot do, trialling them and disseminating the results within their professional community.

AF-EL-3c

Teachers can plan capacity-building on AI for inclusion that transforms how foundational AI knowledge shapes inclusive practice across their educational context.

AF-FC-3a

Teachers can plan longer learning sequences, project formats, or cross-subject units that systematically build learners' technical AI literacy, and share them with colleagues.

AF-FC-3b

Teachers can produce age-appropriate pedagogical resources, analogies, or activity formats for teaching AI foundations and practical AI skills, and disseminate them within their team or professional community.

AF-FC-3c

Teachers can plan collaborative initiatives that transform how learners' technical AI competence is built across their educational context.


4. Certification & Micro-Credentials

4. Certification & Micro-Credentials

4.1 Certification Pathway

What is the certification?

The DUAL.AI.TEACHer certification pathway provides teachers with a structured way to demonstrate their competences in two complementary areas: AI-Enhanced Teaching and AI Literacy. Teachers may pursue either one certification track or both, depending on their prior knowledge, professional role, and development needs.

The certification pathway combines self-assessment, learning and preparation activities, formal assessment, and digital certification.

Choose your track

The DUAL.AI.TEACHer certification consists of two complementary tracks:

Participants may pursue either one track or both tracks of certification, depending on their prior knowledge, professional role, and development needs.

A preliminary self-assessment will help candidates identify the most appropriate pathway.

How does the certification work?

The certification process follows a common sequence for both tracks:

  1. The teacher discovers the DUAL.AI.TEACHer certification.
  2. The candidate registers on the ICEP platform.
  3. The candidate completes a preliminary self-assessment and selects Track A or Track B.
  4. The candidate learns and prepares using the provided content, including the collection of badges in the MOOC.
  5. The candidate books an exam slot.
  6. The candidate attempts the exam.
  7. Upon successful completion, the digital certificate is issued.
  8. The certificate may be shared through Europass, where third-party institutions can check its validity.
  9. The certificate expires after five years.

Registration, self-assessment and track selection.pngFigure 1: Registration, Self-Assessment and Certification Pathway Selection Process

Learning and preparation

After selecting a certification track, candidates enter the learning and preparation phase. They use the provided learning content and may collect badges through the MOOC as part of their preparation for the certification assessment.

The preparation process includes access to learning materials, support activities, self-assessment, practical assignments, and progress tracking. The candidate can continue learning, repeat activities where needed, or proceed towards the certification assessment once sufficiently prepared.

Learning activities.pngFigure 2: Learning and Preparation Process

Assessment and certification

The certification assessment is delivered through an online examination for each certification track.

The planned assessment format includes:

Candidates complete the assessment for the selected track after the preparation phase and readiness self-assessment. Successful completion of the examination leads to the issuance of the corresponding certificate.

The assessment may include the following question types:

These selected-response formats are intended to support objective, consistent, and scalable assessment, while enabling efficient automated scoring.

Issuing certificate.pngFigure 3: Examination and Certification Process between the Candidate and ICEP

Which track is right for me?

The two certification tracks address different professional needs and interests.

Track A – AI-Enhanced Teaching (WITH AI) is intended for teachers who want to use AI in teaching and learning in a pedagogically meaningful, safe, and professionally responsible way. In the Certification Pathway Design, this is illustrated through Maja Novak, a secondary school teacher who already experiments with generative AI for lesson preparation, differentiated materials, classroom activities, and formative feedback. Her main goal is to move from experimental use of AI towards more informed and reflective professional practice.

Dual.AI.TEACHer Certification Pathway - Track A.pngFigure 4: Candidate Journey – Track A: AI-Enhanced Teaching

Track B – AI Literacy (ABOUT AI) is intended for teachers who want to strengthen their own and their learners’ understanding of AI concepts, limitations, risks, and responsible use. This is illustrated through Luka Horvat, an upper-secondary teacher who wants to help students become critical, informed, and responsible users of AI, with particular attention to issues such as misinformation, bias, privacy, transparency, data use, human agency, and the social impact of AI.

Dual.AI.TEACHer Certification Pathway - Track B.png

Figure 5: Candidate Journey – Track B: AI Literacy

Candidates may pursue either one track or both tracks, depending on their prior knowledge, professional role, and development needs. A preliminary self-assessment supports the selection of the most appropriate pathway.

Credentialing and Recognition

Upon successful completion of the assessment, candidates receive a digital certificate for the completed certification track.

The certification model includes:

The certification is intended to support the recognition of teachers’ AI-related competencies and their continuing professional development. It is also intended to support the recognition and portability of these competences across European contexts.

More Information

Further details on the certification structure, candidate journey, assessment criteria, examination procedure, roles and responsibilities, pilot validation, and sustainability are provided in the full DUAL.AI.TEACHer Certification Pathway Design document.

4. Certification & Micro-Credentials

4.2 Micro-Credentials

The Micro-Credentials

Micro credentials are short, focused qualifications that certify specific learning outcomes and skills axquired through brief training or courses. 

Characteristics: 
Key benefits:
  • Flexibility - Allows fast upskilling or retraining without long-term financial commitments.
  • Employability - Provides employers with transparent, easily verifiable proof of practical competencies.
  • Stackability - Can often be combined or "stacked" over time to count towards larger qualifications or degrees

The Micro-Credentials & Certification Architecture

The certification strategy in DUAL.AI.TEACHer establishes a formal, European-wide mechanism for validating teachers' AI competencies.

Dual Certification Pathways:

  1. Pathway A - AI-Enhanced Pedagogy: Validates competence in using AI tools for lesson design, adaptive instruction, automated assessment, and workflow optimisation.

  2. Pathway B - AI Literacy & Civic Education: Validates competence in teaching students computational thinking, algorithmic bias, data privacy, and ethical AI citizenship.

Structural Framework & Alignment:

Technical Architecture & Credentialing Process:

5. AI Integration Roadmap

5. AI Integration Roadmap

5.1 How to Use the Roadmap

This chapter presents the DUAL.AI.TEACHer Integration Roadmap. The roadmap describes nine stages that take teacher education from its first strategic conversation about AI to a permanent structure for it. The roadmap addresses initial teacher education at universities as well as schools and professional development. This chapter is for directors of teacher-training programmes at universities, individuals responsible for teacher professional development, and school principals.

It is deliberately not a single text to be read from beginning to end. Read the part that matches your role.

Stage Start with Then
Anyone 5.2 It takes two minutes and gives you the whole model

A school leader or responsible for professional development in a school

5.2

5.3, written for schools

Working in a teacher education institution (programme direction, teacher educators, quality assurance)

5.2 5.4, the full version, read in sequence

A teacher looking for classroom practice

5.2 You can stop there. Chapters 5.2 and 5.3 are the ones written for you

Reading the full version

Section 5.4 describes each of the nine stages in full: what the stage involves, why it is designed that way, and how you can tell that you have completed it. It is the reference version of the roadmap, and it is written to be used twice. Read it once in sequence, to see why the stages are ordered as they are and which of them depend on each other. Then return to individual stages as you work on them, which is how most institutions will use it in practice.

Readers coming from 5.3 do not need 5.4 in order to act, but any stage described there in brief is set out at length in 5.4. The two sections meet at Stage 6, Embed AI into CPD: it is where teacher education institutions organise the continuing professional development that schools take up.

5. AI Integration Roadmap

5.2 Roadmap (Summarised Short Version)

Our starting point

The DUAL.AI.TEACHer project aims to develop a clear and adaptable roadmap of seven to ten stages that helps teacher education institutions embed AI into both initial teacher education and continuing professional development. The roadmap focuses on institutional change rather than individual initiatives and was developed together with teacher educators.

Two existing models provided particularly relevant foundations.

The first is a six-stage higher education model, summarised as “learning to learn again”. It moves from reframing institutional conversations about AI through trust and governance, staff fluency, curriculum change and the redesign of teaching and assessment, towards institution-wide amplification (Black, 2025). Its main strength is that it treats AI as a driver of institutional transformation rather than simply a tooling question.

The second is a four-phase GenAI implementation roadmap for schools and districts: establish a foundation, develop staff, update what is taught, and assess and progress. Its strength is its operational focus and emphasis on concrete implementation.

Both models do what they set out to do: one addresses universities as a whole, the other schools and districts. Teacher education institutions sit across both settings, and this is where we saw room to extend the existing work. These institutions transform their own teaching while simultaneously preparing future teachers for contexts beyond their direct control. Attending to internal transformation alone would leave AI-related curricula to be delivered by staff with limited practical experience, while attending to future classroom practice alone would modernise the programme without changing the institution that delivers it. The DUAL.AI.TEACHer roadmap builds on both models and connects the two dimensions.

grafik.png

Fig. 1: DUAL.AI.TEACHer Integration Roadmap: Nine stages for the institutional integration of AI in teacher education.

Two axes beneath the stages

Rather than combining the two reference models step by step, the roadmap is built around two parallel axes.

The first is institutional transformation: structures, governance, curriculum and the formal conditions that enable or constrain AI use. This axis follows the reframe-to-amplify logic of the higher education model and is informed by established research. Diffusion of Innovations identifies trialability and observability as characteristics that facilitate innovation adoption, informing the emphasis in Stage 7 on visible pilot implementation (Neal et al., 2018).

The second is human capacity, which serves as the quality axis of the roadmap. Training cannot be reduced to isolated workshops. Research on effective teacher professional development highlights sustained duration, content focus, practice, reflection, feedback and coaching (Lipowsky & Rzejak, 2021). These principles therefore inform the design of training activities throughout the roadmap, from teacher educators in Stage 3 to in-service teachers in Stage 6. A one-off workshop alone does not constitute completion of a stage.

The Roadmap Cheatsheet

Stage Task Done when
Strategic Alignment Assess where the institution actually stands and agree what it wants to achieve, with leadership involved and a cross-functional team in place. The starting position and the goal exist in writing.
Enabling Governance Put institution-wide guidance in place on AI use, data protection, integrity, disclosure and tool vetting, framed around what is possible rather than what is banned. Staff can answer "am I allowed to do this?" without asking anyone.
Teacher Educator Capacity Build the capacity of those who teach, through sustained, content-focused, practice-embedded professional learning with coaching and reflection. Development is ongoing, not only a single workshop.
Co-Design the Curriculum Move AI into the formal programme by revising syllabi and learning outcomes, anchored in subject didactics rather than in a standalone module. AI-related outcomes appear in the regular module documentation of several subjects.
Pedagogy & Assessment Redesign learning tasks and assessment to be authentic and AI-aware, and model AI-integrated teaching so students experience it as learners. Assessment formats have been revised, not just supplemented by a declaration of AI use.
Embed AI into CPD Turn continuing professional development into a structured offer with stackable micro-credentials, securing the in-service arm. Provision has a published structure and a recognised credential.
Pilot in Authentic Contexts Trial the redesigned modules, CPD and assessments with real cohorts and gather feedback systematically. Documented feedback exists and has led to identifiable revisions.
Evaluate, Learn & Adapt Review competency gains, participation, feasibility, equity and impact, and use the evidence to revise. Evidence has changed something, not just filled a report.

Institutionalise, Scale

Convert the initiative into permanent structure: ownership, roles, resources, quality assurance, renewal. The work continues after the people who started it have moved on.

AI for Education. (n.d.). AI adoption roadmap for education institutions.
Black, A. (2025, December 3). Learning to learn again: A roadmap for higher education institutions in the age of AI. enablinginsights.
Lipowsky, F., & Rzejak, D. (2021). Fortbildungen für Lehrpersonen wirksam gestalten: Ein praxisorientierter und forschungsgestützter Leitfaden. Bertelsmann Stiftung.
Neal, J. W., Neal, Z. P., Lawlor, J. A., Mills, K. J., & McAlindon, K. (2018). What makes research useful for public school educators? Administration and Policy in Mental Health and Mental Health Services Research, 45(3), 432–446.

5. AI Integration Roadmap

5.3 Roadmap for Schools

Who this section is for

School leaders, deputy heads and those responsible for professional development in a school. You do not need to have read the preceding sections in full. If you have read the nine stages in 5.2, this section tells you how they apply to a school; if you have not, you can start here.

Where schools enter the roadmap. This roadmap was developed for teacher education institutions, but its change logic is not specific to them, and one stage concerns schools directly. Stage 6, Embed AI into CPD, is the in-service arm: it is where teacher education institutions organise the continuing professional development that your staff attend. That is your entry point. What your institution offers as Stage 6 arrives in your school as professional development, and whether it changes anything in your classrooms depends on conditions within that school. 

Why you cannot start there and stop. Sending staff to a one-off course is the most common approach and at the same time unlikely to change practice on its own. What works is sustained engagement with a content focus, opportunity to practise, and follow-up in the form of coaching or collaboration (Lipowsky & Rzejak, 2021). This has a direct consequence for a school leader. The scarce resource is not the course; it is the protected time afterwards in which teachers try something, discuss it with colleagues and adjust. 

grafik.png

Fig. 1: DUAL.AI.TEACHer Integration Roadmap: Nine stages for the institutional integration of AI in teacher education.

The three stages to prioritise at school level

Stage 6 is your connection point, but Stages 1-3 create the conditions that make it work. 

Stage 1, Strategic Alignment. Establish where your school actually stands. Some of your staff are already using AI, some are avoiding it, and schools may not have a clear picture of current practice. A short, honest stocktake is worth more than a strategy paper. Name what you want to achieve, and be prepared for the answer that AI is not currently your school's most pressing problem; that is a legitimate result of this stage.

Stage 2, Enabling Governance. This is particulary important at school level and easy to overlook. Staff and students need to know what is permitted, how AI use is to be disclosed, which tools have been checked for data protection, and what counts as academic dishonesty. Where this is unclear, two things happen at once: cautious teachers avoid AI entirely, and confident ones use it without any agreed limits. Clear, enabling guidance helps addressing both. It does not need to be long.

Stage 3, Staff Capacity. In the original roadmap this stage concerns teacher educators; in your school it concerns your teaching staff. The same quality conditions apply. Consider starting with a small group of willing colleagues in one or two subjects rather than a whole-staff rollout, and make their results visible to the rest of the staff. Innovations spread when colleagues can see them working in their own setting (Neal et al., 2018).

What is different in a school

Three constraints distinguish your situation from that of a teacher education institution, and the roadmap should be read with them in mind.

You have less curricular autonomy. Stages 4 and 5 assume an institution that can revise its own course content. Much of your curriculum may be set within national or statutory frameworks and your room for manoeuvre lies in how subjects are taught and how learning is assessed, not in what is prescribed. Read those stages as being about task design and assessment practice.

Your staff have less discretionary time. A teacher education institution can assign development work to academic staff as part of their role. In a school, every hour of development competes directly with teaching. This makes Stage 2 disproportionately valuable: guidance costs little time and removes a great deal of friction.

You are closer to the consequences. Schools experience the the practical consequences of questions around equity, student data and access particulary directly. Stage 8 asks explicitly about equity for this reason.


Lipowsky, F., & Rzejak, D. (2021). Fortbildungen für Lehrpersonen wirksam gestalten: Ein praxisorientierter und forschungsgestützter Leitfaden. Bertelsmann Stiftung.
Neal, J. W., Neal, Z. P., Lawlor, J. A., Mills, K. J., & McAlindon, K. (2018). What makes research useful for public school educators? Administration and Policy in Mental Health and Mental Health Services Research, 45(3), 432–446.

5. AI Integration Roadmap

5.4 Full Roadmap for Teacher Education Institutions

Who this section is for

Deans, programme directors, teacher educators and those responsible for quality assurance in institutions that prepare teachers. This roadmap was written primarily for you, and this section assumes you have read the nine stages in 5.1. Unlike school readers, teacher education institutions need to consider the roadmap as a whole, because the stages address different parts of institutional change that depend on one another.

grafik.png

Fig. 1: DUAL.AI.TEACHer Integration Roadmap: Nine stages for the institutional integration of AI in teacher education.

1. Strategic Alignment

From reacting to strategy Most institutions do not start from zero. They start from a patchwork of individual initiatives, informal tool use and unresolved concerns. The first stage makes that starting position visible and turns it into a shared strategic direction. Leadership opens the process rather than delegating it, because later stages require decisions about staffing, resources and curriculum that need institutional support. The institution assesses its readiness, needs and priorities, and establishes a cross-functional steering team that includes teacher educators, IT, students and quality assurance. Involving students is not symbolic: they experience the curriculum directly and, in teacher education, are also members of the future profession.

The stage is complete when the institution can state, in writing, where it stands, what it wants to achieve and who is responsible for taking the work forward.

2. Enabling Governance

Guidance comes early because uncertainty can become a major barrier to adoption. Staff who do not know what is permitted may either avoid AI altogether or use it without shared expectations. The institution develops institution-wide guidance covering AI use, data protection, academic integrity, disclosure expectations and a process for reviewing and approving tools. The decisive design choice is the framing: guidance should explain what is possible and under which conditions, rather than consist mainly of prohibitions. Draft guidance is circulated for feedback before adoption, improving both its practical usefulness and its legitimacy.

The stage is complete when staff can readily find a clear answer to the question “Am I allowed to do this?” and know where to turn when a case falls outside the guidance.

3. Teacher Educator Capacity

Teacher educators cannot model practices they have not had the opportunity to explore themselves. This stage builds their capacity through professional learning that is sustained rather than one-off, connected to content and subject didactics rather than focused mainly on tools, and embedded in practice through collaboration, coaching and structured reflection. These are features consistently associated with effective teacher professional development (Lipowsky & Rzejak, 2021). Competences are mapped against the DUAL.AI.TEACHer Competency Framework and the UNESCO AI Competency Framework for Teachers so that development has both a direction and a reference point. AI integration should build on teacher educators' existing professional expertise rather than treat AI as a separate technical competence. A single workshop does not complete this stage. The stage is complete when teacher educators have sustained opportunities to develop, apply and reflect on AI-related practice in their own teaching contexts.

4. Co-Design the Curriculum

Here AI moves from extracurricular activity into the formal programme. Course and module documentation is reviewed, and learning outcomes are aligned with the competency framework so that what is taught, practised and assessed points in the same direction. A central design principle of this roadmap is that AI literacy and AI-enhanced pedagogy should be embedded within subject didactics rather than confined to a standalone module. A dedicated AI course may be useful as a transitional step, but it risks signalling that AI is a separate topic rather than part of everyday subject teaching. Co-design with teacher educators matters because those who will teach the revised programme need to have shaped it.

The stage is complete when AI-related learning outcomes are visible in the regular documentation of multiple subjects or modules and are connected to the wider programme rather than isolated in a single course.

5. Pedagogy & Assessment

Changing the curriculum on paper is not enough if teaching and assessment remain unchanged. This stage redesigns learning tasks and assessment formats so that they are authentic, multimodal and AI-aware, and so that assessment focuses on what future teachers can understand, justify, create and apply rather than on outputs that a generative system can produce on request. Equally important is the modelling effect: pre-service teachers should experience thoughtful AI-integrated teaching as learners before they are expected to use it as professionals. Teacher education has a distinctive opportunity here because students experience teaching practices while simultaneously developing their own professional practice.

The stage is complete when AI-related learning outcomes are reflected in actual teaching and assessment practice, not merely added to module descriptions or disclosure requirements.

6. Embed AI into CPD

Teacher education institutions do not only prepare future teachers; they also support those already working in schools. This stage develops continuing professional development into a structured offer rather than a series of isolated events. The same quality conditions as in Stage 3 apply: sustained engagement, a clear content focus, opportunities to practise, and follow-up through collaboration, feedback or coaching. In the DUAL.AI.TEACHer model, this work can be linked to the certification pathway developed in T2.6, allowing professional learning to build over time. Stage 6 also creates a two-way connection with schools: institutions contribute professional learning, while in-service teachers bring current classroom experience back into teacher education.

The stage is complete when in-service professional learning forms a coherent and sustained offer with clear progression, rather than a calendar of unrelated single events.

7. Pilot in Authentic Contexts

Redesigned modules, CPD offers and assessment formats are trialled with real cohorts under realistic conditions, and feedback from teacher educators, students and other participants is gathered systematically. Piloting reduces the risk of scaling an approach before it has been tested in practice. It also makes implementation visible: colleagues can see how an approach works in a context similar to their own, discuss the results and adapt it before wider adoption. The purpose of the pilot is therefore not to demonstrate that the design was correct, but to discover what needs to change.

The stage is complete when documented feedback from authentic implementation has led to identifiable revisions.

8. Evaluate, Learn & Adapt

Evaluation asks whether the changes achieved what they were intended to achieve. Depending on the intervention, this may include competency development, participation, feasibility, equity, staff and student experience, and wider programme effects. Equity is considered explicitly because AI integration may reduce some inequalities while creating or reinforcing others, for example across subjects, staff groups or students with different levels of access and support. Evidence is then used to revise curriculum, professional development, assessment and implementation.

This stage also establishes a regular review cycle for the guidance developed in Stage 2, which will need to evolve as technologies, practices and regulatory conditions change.

9. Institutionalise and Scale

The final stage moves successful work from project status into the regular structures of the institution. This means clear ownership, defined responsibilities, appropriate resources and integration into existing processes such as programme development and quality assurance. Scaling should follow the institution's own context, capacity and profile rather than a single generic model. Institutionalisation also means creating mechanisms for renewal: AI technologies, educational practices and institutional needs will continue to change, so the model itself cannot remain fixed.

The stage is complete when the work continues after the people who started it have moved on.

Your specific position

Teacher education institutions carry two mandates at once. You transform your own teaching, and you prepare future teachers for classrooms you do not control, under conditions you cannot fully predict, using technology that may have changed by the time your graduates arrive there. The two mandates are closely connected. Focusing only on internal transformation risks producing AI-related curricula without sufficient connection to future classroom practice. Focusing only on future classroom practice risks modernising the programme without changing the institution that delivers it. The roadmap therefore treats both dimensions together.

Stage 3 enables what follows. Much of the later work depends on the capacity of teacher educators. Their expertise lies in their disciplines and subject didactics, so integrating AI is not simply a matter of adding another tool. Professional development needs to connect AI use to subject-specific teaching, learning and assessment rather than treating it as a generic technical competence. For this reason, Stage 3 deserves particular focus when resources are limited.

Where Stage 4 becomes difficult. The principle that AI literacy should be embedded within subject didactics rather than treated only in a standalone module is easy to state but a lot harder to implement. A standalone module can be assigned to one person, while distributed integration requires many colleagues to reconsider parts of their own teaching. Where a standalone module is useful as a transitional step, it should ideally be connected to a longer-term plan for integration across the programme.

Your distinctive lever in Stage 5. Stage 5 highlights a particular feature of teacher education: your students are learners now and teachers later. How you teach them therefore becomes part of what they learn about teaching. If pre-service teachers experience thoughtful, AI-integrated teaching as learners, they encounter concrete models they can later reflect on and adapt in their own practice. Teaching about AI without modelling its pedagogical use provides a very different experience. This is why Stage 5 goes beyond assessment reform alone.

Stage 6 is your interface with schools. Your continuing professional development provision is the entry point. It extends your institution's reach beyond future teachers to those who are already working in classrooms. At the same time, collaboration with in-service teachers can bring current classroom experience back into teacher education. CPD can therefore function not only as a service provided to schools, but also as a channel for reciprocal learning between schools and teacher education institutions.

What you can do that schools cannot

Teacher education institutions often have greater scope to revise programmes and assessment, access to research capacity, and a mandate to work across multiple schools. This gives Stages 7 and 8 particular value: what is learned through pilots and evaluation can inform not only your own institution but also partner schools and the wider system. Making those experiences visible can also support the uptake of new practices elsewhere. 

Where to begin if everything is urgent. Start with Stage 2 and Stage 3. Governance reduces uncertainty, while staff capacity creates the conditions for meaningful work in Stages 4 to 6. Beginning with curriculum revision before staff have the capacity to implement it risks producing changes on paper that are difficult to realise in practice.

6. Resources & Glossary

6. Resources & Glossary

6.1 Resources

Peer reviewed

Ali, R., Tang, O. Y., Connolly, I. D., Abdulrazeq, H. F., Mirza, F. N., Lim, R. K., Johnston, B. R., Groff, M. W., Williamson, T., Svokos, K., Libby, T. J., Shin, J. H., Gokaslan, Z. L., Doberstein, C. E., Zou, J., & Asaad, W. F. (2024). Demographic representation in 3 leading artificial intelligence text-to-image generators. JAMA Surgery, 159(1), 87–95. https://doi.org/10.1001/jamasurg.2023.5695

Anderson, L. W., & Krathwohl, D. R. (Eds.). (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom’s taxonomy of educational objectives. Longman.

Autenrieth, D., & Schluchter, J.-R. (2026). Künstliche Intelligenz und inklusive Bildung: Empirische Perspektiven auf Einstellungen und Nutzungsmuster von Lehrkräften [Artificial intelligence and inclusive education: Empirical perspectives on teachers' attitudes and usage patterns]. Medienimpulse, 64(1). https://doi.org/10.21243/mi-01-26-24 

Barnes, M., & Tour, E. (2026). Teachers' use of generative AI: A 'dirty little secret'? Language and Education, 40(1), 24–39. https://doi.org/10.1080/09500782.2025.2485935

Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), Article e2422633122. https://doi.org/10.1073/pnas.2422633122

BEKIARIDIS, G. (author) and ATTWELL, G. (ed.), Supplement to the DigCompEDU Framework: Outlining the Skills and Competences of Educators Related to AI in Education, AI Pioneers – Work Package 3, Erasmus+ Programme, European Union, 2024. Available at: https://aipioneers.org/wp-content/uploads/2024/01/WP3_Supplement_to_the_DigCompEDU_English.pdf 

BLOOM, B.S. (ed.), ENGELHART, M.D., FURST, E.J., HILL, W.H. and KRATHWOHL, D.R., Taxonomy of Educational Objectives: The Classification of Educational Goals. Handbook I: Cognitive Domain, David McKay Company, New York, 1956.

Buçinca, Z., Malaya, M. B., & Gajos, K. Z. (2021). To trust or to think: Cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), Article 188. https://doi.org/10.1145/3449287

Dier, M. (2025). KI-Kompetenzprofile von Lehrkräften an beruflichen Schulen: Vom Wissen zum didaktischen Handeln [AI competence profiles of vocational school teachers: From knowledge to didactic action]. Journal of Technical Education, 13(2), 66–85. https://doi.org/10.48513/joted.v13i2.328

Dong, Y. (2026). Generative AI technologies and educational outcomes: A comprehensive meta-analysis comparing traditional and AI-driven approaches. Humanities and Social Sciences Communications, 13(1), Article 559. https://doi.org/10.1057/s41599-026-06903-y

Farrokhnia, M., Latifi, S., Papadopoulos, P. M., Hogenkamp, L., Gijlers, H., Khosravi, H., & Noroozi, O. (2026). Generative AI offers more, but students revise less: Comparing the effects of teacher and AI feedback on student essay revisions. International Journal of Educational Technology in Higher Education, 23(1), Article 6. https://doi.org/10.1186/s41239-026-00579-9

Fleckenstein, J., Liebenow, L. W., & Meyer, J. (2023). Automated feedback and writing: A multi-level meta-analysis of effects on students' performance. Frontiers in Artificial Intelligence, 6, Article 1162454. https://doi.org/10.3389/frai.2023.1162454 ✔ ⚠ Ch. 1.2

Jaggy, A.-K., Jansen, T., Fütterer, T., & Trautwein, U. (2026). Almighty ChatGPT? Development of gifted secondary school students' AI (mis)conceptions. Journal of Educational Psychology. Advance online publication. https://doi.org/10.1037/edu0001063

Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: An RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15(1), Article 17458. https://doi.org/10.1038/s41598-025-97652-6 

Krathwohl, D. R. (2002). A revision of Bloom’s taxonomy: An overview. Theory Into Practice, 41(4), 212–218.

Lee, V. R., Pope, D., Miles, S., & Zárate, R. C. (2024). Cheating in the age of generative AI: A high school survey study of cheating behaviors before and after the release of ChatGPT. Computers and Education: Artificial Intelligence, 7, Article 100253. https://doi.org/10.1016/j.caeai.2024.100253

Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), Article 100779. https://doi.org/10.1016/j.patter.2023.100779

Liu, B., Zhang, W., & Wang, F. (2026). Can generative artificial intelligence effectively enhance students' mathematics learning outcomes? A meta-analysis of empirical studies from 2023 to 2025. Education Sciences, 16(1), Article 140. https://doi.org/10.3390/educsci16010140

Maršálek, R., & Teplá, M. (2026). "Is AI inevitable?" Development of attitudes and practices of Czech teachers between 2023 and 2025. Education Sciences, 16(2), Article 335. https://doi.org/10.3390/educsci16020335

MIAO, F. and CUKUROVA, M., AI Competency Framework for Teachers, UNESCO, 2024. ISBN 978-92-3-100707-1. DOI: https://doi.org/10.54675/ZJTE2084 

REDECKER, C., European Framework for the Digital Competence of Educators: DigCompEdu, PUNIE, Y. (ed.), EUR 28775 EN, Publications Office of the European Union, Luxembourg, 2017. ISBN 978-92-79-73494-6 (pdf). DOI: https://doi.org/10.2760/159770. JRC107466.

Rind, I. A. (2026). Conceptualizing the impact of AI on teacher knowledge and expertise: A cognitive load perspective. Education Sciences, 16(1), Article 57. https://doi.org/10.3390/educsci16010057

Romeo, G., & Conti, D. (2026). Exploring automation bias in human–AI collaboration: A review and implications for explainable AI. AI & Society, 41, 259–278. https://doi.org/10.1007/s00146-025-02422-7

Rowe, J., Klimaszewski, M., Guillou, L., Vallor, S., & Birch, A. (2025). EuroGEST: Investigating gender stereotypes in multilingual language models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (pp. 32074–32096). Association for Computational Linguistics. https://doi.org/10.18653/v1/2025.emnlp-main.1632

Scarfe, P., Watcham, K., Clarke, A., & Roesch, E. (2024). A real-world test of artificial intelligence infiltration of a university examinations system: A "Turing Test" case study. PLOS ONE, 19(6), Article e0305354. https://doi.org/10.1371/journal.pone.0305354

Schemmer, M., Kühl, N., Benz, C., Bartos, A., & Satzger, G. (2023). Appropriate reliance on AI advice: Conceptualization and the effect of explanations. In Proceedings of the 28th International Conference on Intelligent User Interfaces (pp. 410–422). Association for Computing Machinery. https://doi.org/10.1145/3581641.3584066

Tian, Z., Liu, A., Esbenshade, L., Sarkar, S., Zhang, Z., He, K., & Sun, M. (2025). Implementation considerations for automated AI grading of student work. In J. Wilson, C. Ormerod, & M. Beiting Parrish (Eds.), Proceedings of the Artificial Intelligence in Measurement and Education Conference (AIME-Con): Full papers (pp. 9–20). National Council on Measurement in Education. https://aclanthology.org/2025.aimecon-main.2/ 

Vartiainen, H., Kahila, J., Tedre, M., López-Pernas, S., & Pope, N. (2025). Enhancing children's understanding of algorithmic biases in and with text-to-image generative AI. New Media & Society, 27(9), 5342–5368. https://doi.org/10.1177/14614448241252820

Wang, J., & Fan, W. (2025). RETRACTED ARTICLE: The effect of ChatGPT on students' learning performance, learning perception, and higher-order thinking: Insights from a meta-analysis. Humanities and Social Sciences Communications, 12(1), Article 621. https://doi.org/10.1057/s41599-025-04787-y 

Wang, J., & Fan, W. (2026). Retraction note: The effect of ChatGPT on students' learning performance, learning perception, and higher-order thinking: Insights from a meta-analysis. Humanities and Social Sciences Communications, 13(1), Article 528. https://doi.org/10.1057/s41599-026-07310-z

Weinmann, H., Messingschlager, T. V., & Appel, M. (2026). Gender bias in text-to-image generative artificial intelligence: Neglect and stereotypical presentations across three popular platforms. New Media & Society. Advance online publication. https://doi.org/10.1177/14614448261435197

Wu, X., Zhu, P., Zhang, J., Yin, M., & Wang, Y. (2026). ChatGPT's impact on student learning outcomes: A meta-analysis of 35 experimental studies. Humanities and Social Sciences Communications, 13(1), Article 684. https://doi.org/10.1057/s41599-026-07019-z

Yan, D., & Zhang, S. (2024). L2 writer engagement with automated written corrective feedback provided by ChatGPT. Humanities and Social Sciences Communications, 11, Article 1086. https://doi.org/10.1057/s41599-024-03543-y

Yeo, G. H., & Lansford, J. E. (2025). Effects of artificial intelligence on educational functioning: A review and meta-analysis. Educational Psychology Review, 37(4), Article 110. https://doi.org/10.1007/s10648-025-10085-5 

Yi, L., Liu, D., Jiang, T., & Xian, Y. (2025). The effectiveness of AI on K-12 students' mathematics learning: A systematic review and meta-analysis. International Journal of Science and Mathematics Education, 23(4), 1105–1126. https://doi.org/10.1007/s10763-024-10499-7

Yu, H., & Xie, Q. (2025). Generative AI vs. teachers: Feedback quality, feedback uptake, and revision. Language Teaching Research Quarterly, 47, 113–137. https://doi.org/10.32038/ltrq.2025.47.07

Zhang, L., Carter, R. A., Jr., Liu, Y., & Peng, P. (2026). Let's CHAT about artificial intelligence for students with disabilities: A systematic literature review and meta-analysis. Review of Educational Research, 96(1), 215–257. https://doi.org/10.3102/00346543241293424

Preprints, working papers and research reports

Campos, C., & Singleton, J. D. (2026). AI diffusion gaps: Unequal integration of AI across K-12 schools (NBER Working Paper No. 35347). National Bureau of Economic Research. https://doi.org/10.3386/w35347 

Cruces, G., Fernández Meijide, D., Galiani, S., Gálvez, R. H., & Lombardi, M. (2026). Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment (NBER Working Paper No. 34851; rev. October 2026). National Bureau of Economic Research. https://doi.org/10.3386/w34851 

Kalai, A. T., Nachum, O., Vempala, S. S., & Zhang, E. (2025). Why language models hallucinate (arXiv:2509.04664v1) [Preprint]. arXiv. https://arxiv.org/abs/2509.04664 

Lehmann, M., Cornelius, P. B., & Sting, F. J. (2025). AI meets the classroom: When do large language models harm learning? (arXiv:2409.09047v2) [Preprint]. arXiv. https://arxiv.org/abs/2409.09047 

Lodge, J. M., & Loble, L. (2026). Artificial intelligence, cognitive offloading and implications for education [Report]. Australian Network for Quality Digital Education; University of Technology Sydney. https://doi.org/10.71741/4pyxmbnjaq.31302475

Maier, S., Gunzenhäuser, M., Schweisthal, J., Schneider, M., & Feuerriegel, S. (2026). A meta-analysis of the effect of generative AI on productivity and learning in programming (arXiv:2605.04779v1) [Preprint]. arXiv. https://arxiv.org/abs/2605.04779 

Shao, J., Wu, Q., Zhang, H., Sun, S., & Zhuang, J. (2026). Mitigating scaffolding collapse in Socratic tutors via representation alignment (arXiv:2607.19371v1) [Preprint]. arXiv. https://arxiv.org/abs/2607.19371

Paris, A., Labetoulle, A., Bezjak, S., Butler, D., Cardoso-Leite, P., Chesné, J.-F., Kirsch, C., Mori, S., Nardi, A., Nencioni, P., Niewint, J., Rossi, F., Storai, F., Toci, V., & Mirazchiyski, P. V. (2023). Comparative European evaluation report (AI4T – Artificial Intelligence for and by Teachers, Deliverable D3.3). Cnesco-Cnam; France Éducation international. https://cnesco.fr/wp-content/uploads/2024/05/AI4T_WP3_D3.3_Comparative-European-Evaluation-Report.pdf 

Roy, P., Poet, H., Staunton, R., Aston, K., & Thomas, D. (2024). ChatGPT in lesson preparation: A Teacher Choices Trial [Evaluation report]. National Foundation for Educational Research; Education Endowment Foundation; The Hg Foundation. https://www.nfer.ac.uk/publications/chatgpt-in-lesson-preparation-a-teacher-choices-trial/

Staksrud, E., Mascheroni, G., Milosevic, T., Ní Bhroin, N., Ólafsson, K., Şengül-İnal, G., & Stoilova, M. (2026). European children's use and understanding of generative AI: EU Kids Online 2026 (EU Kids Online Reports No. 1). London School of Economics and Political Science. https://doi.org/10.21953/researchonline.lse.ac.uk.00137132

Statistics and reports by organisations

Miao, F., & Cukurova, M. (2024). AI competency framework for teachers. UNESCO. https://doi.org/10.54675/ZJTE2084

OECD. (2025). Results from TALIS 2024: The state of teaching. OECD Publishing. https://doi.org/10.1787/90df6235-en

OECD. (2026). OECD digital education outlook 2026: Exploring effective uses of generative AI in education. OECD Publishing. https://doi.org/10.1787/062a7394-en

UNESCO. (2023). Global education monitoring report 2023: Technology in education – A tool on whose terms? https://doi.org/10.54676/UZQV8501

Surveys and reports by foundations, NGOs ...

Bellio, N., & Kayser-Bril, N. (2024, March 4). Wie Bildgeneratoren die Welt zeigen: Stereotypen statt Vielfalt [How image generators show the world: Stereotypes instead of diversity]. AlgorithmWatch. https://algorithmwatch.org/de/bildgeneratoren-stereotypen/

Human Rights Watch. (2022, May 25). "How dare they peep into my private life?": Children's rights violations by governments that endorsed online learning during the Covid-19 pandemic. https://www.hrw.org/report/2022/05/25/how-dare-they-peep-my-private-life/childrens-rights-violations-governments

Latham, K., & Montacute, R. (2025, July). Artificial advantage? AI in the classroom and the inequality gap. The Sutton Trust. https://www.suttontrust.com/wp-content/uploads/2025/07/Artificial-advantage.pdf

Medienpädagogischer Forschungsverbund Südwest. (2024). JIM-Studie 2024. Jugend, Information, Medien: Basisuntersuchung zum Medienumgang 12- bis 19-Jähriger [JIM study 2024: Youth, information, media]. https://mpfs.de/studie/jim-studie-2024/

Michel, L. P., & Goertz, L. (2025). mmb-Trendmonitor 2024/2025. Weiterbildung und digitales Lernen heute und in drei Jahren [mmb trend monitor 2024/2025: Continuing education and digital learning today and in three years] (Results of the 19th mmb Learning Delphi). mmb Institut. https://www.mmb-institut.de/wp-content/uploads/mmb-Trendmonitor_2024-2025.pdf

Robert Bosch Stiftung. (2026). Deutsches Schulbarometer 2026: Befragung Lehrkräfte [German School Barometer 2026: Teacher survey] (Survey conducted by forsa). https://www.bosch-stiftung.de/sites/default/files/documents/2026-06/RBSG_Deutsches_Schulbarometer_Lehrkr%C3%A4fte_2026.pdf

Trucano, M. (2023, July 10). AI and the next digital divide in education. Brookings Institution. https://www.brookings.edu/articles/ai-and-the-next-digital-divide-in-education/

Vodafone Stiftung Deutschland. (2024, March). Pioniere des Wandels: Wie Schüler:innen KI im Unterricht nutzen möchten [Pioneers of change: How students want to use AI in the classroom] (Survey conducted by infratest dimap). https://www.vodafone-stiftung.de/pioniere-des-wandels/

Walton Family Foundation, & Gallup. (2025, June). Teaching for tomorrow: Unlocking six weeks a year with AI. https://news.gallup.com/poll/691967/three-teachers-weekly-saving-six-weeks-year.aspx

Institutional documents

Børne- og Undervisningsministeriet, Styrelsen for Undervisning og Kvalitet. (2025, August 22). Forsøg med prøverne i engelsk A (stx) [Trial of the English A examinations (stx)] (Case 25/23313). https://www.uvm.dk/-/media/filer/uvm/udd/gym/pdf25/aug/250822-forsoeg-med-proeverne-i-engelsk-a--stx-.pdf 

Commission nationale de l'informatique et des libertés. (2025, June 20). Enseignant : comment utiliser un système d'IA dans le respect du RGPD ? [Teachers: How to use an AI system in compliance with the GDPR]. https://www.cnil.fr/fr/enseignant-usage-systeme-ia 

Department of Education Northern Ireland. (2026, May 6). Givan announces roll-out of AI tools and training for Northern Ireland teachers [Press release]. https://www.education-ni.gov.uk/news/givan-announces-roll-out-ai-tools-and-training-northern-ireland-teachers 

Der Hamburgische Beauftragte für Datenschutz und Informationsfreiheit. (2023, November 13). Checkliste zum Einsatz LLM-basierter Chatbots [Checklist for the use of LLM-based chatbots]. https://datenschutz-hamburg.de/fileadmin/user_upload/HmbBfDI/Datenschutz/Informationen/20231113_Checkliste_LLM_Chatbots_DE.pdf 

European Commission. (2026, September 7). Market surveillance authorities under the AI Act. https://digital-strategy.ec.europa.eu/en/policies/market-surveillance-authorities-under-ai-act 

FWU Institut für Film und Bild in Wissenschaft und Unterricht. (n.d.). AIS.chat (vormals telli). Retrieved October 16, 2026, from https://fwu.de/ais-chat/ 

Haridus- ja Teadusministeerium. (2026, March 8). Europe's top PISA performer showcases AI Leap programme as global summit opens in Tallinn [Press release]. https://www.hm.ee/en/news/europes-top-pisa-performer-showcases-ai-leap-programme-global-summit-opens-tallinn 

Joint Council for Qualifications. (2025). AI use in assessments: Protecting the integrity of qualifications (Effective from 30 April 2025). https://www.jcq.org.uk/exams-office/malpractice/artificial-intelligence/

Kennisnet. (2026, January 15). Toetsing en generatieve AI in het voortgezet onderwijs [Assessment and generative AI in secondary education] (updated April 28, 2026). https://www.kennisnet.nl/ 

Konferenz der unabhängigen Datenschutzaufsichtsbehörden des Bundes und der Länder. (2024, May 6). Orientierungshilfe: Künstliche Intelligenz und Datenschutz [Guidance: Artificial intelligence and data protection] (Version 1.0). https://www.datenschutzkonferenz-online.de/media/oh/20240506_DSK_Orientierungshilfe_KI_und_Datenschutz.pdf 

Kultusministerkonferenz. (2024, October 10). Handlungsempfehlung für die Bildungsverwaltung zum Umgang mit Künstlicher Intelligenz in schulischen Bildungsprozessen [Recommendation for education administrations on artificial intelligence in school education] (Resolution of the Bildungsministerkonferenz; rev. ed. March 2025). https://www.kmk.org/fileadmin/Dateien/veroeffentlichungen_beschluesse/2024/2024_10_10-Handlungsempfehlung-KI.pdf ✔

Ministère de l'Éducation nationale. (2025, June 13). Cadre d'usage de l'IA en éducation [Framework for the use of AI in education]. https://www.education.gouv.fr/cadre-d-usage-de-l-ia-en-education-450647 

Ministère de l'Éducation nationale. (2026, February). M.I.A. Seconde, une approche personnalisée de la remédiation en français et en mathématiques [M.I.A. Seconde: A personalised approach to remediation in French and mathematics]. Éduscol. https://eduscol.education.gouv.fr/6693/mia-seconde-une-approche-personnalisee-de-la-remediation-en-francais-et-en-mathematiques 

Ministerstvo školstva, výskumu, vývoja a mládeže Slovenskej republiky. (2026, January 15). 20-tisíc licencií ChatGPT Edu podporí rozvoj digitálnych zručností budúcich učiteľov [20,000 ChatGPT Edu licences will support future teachers' digital skills] [Press release]. https://www.minedu.sk/20-tisic-licencii-chatgpt-edu-podpori-rozvoj-digitalnych-zrucnosti-buducich-ucitelov/ 

Ministerstvo školstva, výskumu, vývoja a mládeže Slovenskej republiky. (2026, May 12). Ministerstvo chce do konca roka sprístupniť AI nástroje pre všetkých učiteľov na Slovensku [The ministry wants to make AI tools available to all teachers in Slovakia by the end of the year] [Press release]. https://www.minedu.sk/ministerstvo-chce-do-konca-roka-spristupnit-ai-nastroje-pre-vsetkych-ucitelov-na-slovensku/ 

Ministerium für Schule und Bildung des Landes Nordrhein-Westfalen. (n.d.). Dienstanweisung für die automatisierte Verarbeitung von personenbezogenen Daten in der Schule (BASS 10-41 Nr. 4). Retrieved October 16, 2026, from https://bass.schule.nrw/17585.htm 

Bayerisches Staatsministerium für Unterricht und Kultus. (n.d.). Private Endgeräte im Dienstgebrauch [Private devices for official use]. Retrieved October 16, 2026, from https://www.km.bayern.de/gestalten/digitalisierung/datensicherheit/private-endgeraete-im-dienstgebrauch 

Office of Qualifications and Examinations Regulation. (2026, March). SLT briefing pack: AI and coursework integrity (Version 2.5). https://www.gov.uk/government/organisations/ofqual 

Scheiter, K., Bauer, E., Omarchevska, Y., Schumacher, C., & Sailer, M. (2025). Künstliche Intelligenz in der Schule: Eine Handreichung zum Stand in Wissenschaft und Praxis [Artificial intelligence in schools: A guide to the state of research and practice]. Bundesministerium für Bildung, Familie, Senioren, Frauen und Jugend. https://www.empirische-bildungsforschung-bmbfsfj.de/img/KI_Review.pdf 

Statsministerens kontor, & Kunnskapsdepartementet. (2026, June 19). Ny nasjonal anbefaling: Kunstig intelligens skal i all hovedsak ikke brukes i barneskolen [New national recommendation: AI should as a rule not be used in primary school] [Press release]. https://www.regjeringen.no/no/aktuelt/kunstig-intelligens-skal-i-all-hovedsak-ikke-brukes-i-barneskolen/id3166807/ 

SURF. (2026, May 27). Privacy risks Microsoft 365 Copilot remain 'orange' despite improvements. https://www.surf.nl/en/news/privacy-risks-microsoft-365-copilot-remain-orange-despite-improvements 

Transport for London. (2022, March). The Knowledge of London: An introduction to learning the Knowledge of London and the examination process. https://foi.tfl.gov.uk/FOI-0355-2425/Introduction%20to%20Knowledge%20Booklet%20(March%202022).pdf 

Vanderbilt University, Brightspace Support. (2023, August 16). Guidance on AI detection and why we're disabling Turnitin's AI detector. https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/ 

Verband Bildungsmedien. (n.d.). Urheberrecht und KI [Copyright and AI]. Schulbuchkopie.de. Retrieved October 16, 2026, from https://www.schulbuchkopie.de/index.php/urheberrecht-und-ki 

ZHAW Zürcher Hochschule für Angewandte Wissenschaften. (2026). Verwendung generativer KI-Systeme bei Leistungsnachweisen [Use of generative AI systems in assessments] (Richtlinie Z-RL, Version 4.0.0). https://gpmpublic.zhaw.ch/GPMDocProdDPublic/Vorgabedokumente_ZHAW/Z_RL_Richtlinie_KI_bei_Leistungsnachweisen.pdf 

Press sources

AI Leap Foundation. (n.d.). AI Leap (TI-Hüpe). Retrieved October 16, 2026, from https://tihupe.ee/en 

Depue, W. (2023, November 6). What's new with DALL·E 3? OpenAI Cookbook. https://developers.openai.com/cookbook/articles/what_is_new_with_dalle_3 

EvidenceB. (2022, November 30). EvidenceB et Docaposte remportent le marché public pour offrir un service de remédiation en français et en mathématiques pour tous les nouveaux lycéens de France [Press release]. https://evidenceb.fr/evidenceb-et-docaposte-remportent-le-marche-public/ 

OpenAI. (2022, July 18). Reducing bias and improving safety in DALL·E 2. https://openai.com/index/reducing-bias-and-improving-safety-in-dall-e-2 

Raghavan, P. (2024, February 23). Gemini image generation got it wrong. We'll do better. The Keyword (Google). https://blog.google/products/gemini/gemini-image-generation-issue/

Legislation

Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), OJ L 119, 4.5.2016, p. 1. https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng ✔

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), OJ L, 2024/1689, 12.7.2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng — consolidated version of 27 July 2026: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A02024R1689-20260727 

Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI), OJ L, 2026/1744, 24.7.2026 (in force 27 July 2026; corrigendum 29 September 2026). https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng 

Gesetz über Urheberrecht und verwandte Schutzrechte (Urheberrechtsgesetz – UrhG) [German Copyright Act], §§ 2, 16, 44a, 44b, 60a, 97. https://www.gesetze-im-internet.de/urhg/ 

Schulgesetz für Baden-Württemberg (SchG), § 115b. https://www.landesrecht-bw.de/ 

Bayerisches Hochschulinnovationsgesetz (BayHIG), as amended by the Act of 8 May 2026 (GVBl. p. 208). https://www.gesetze-bayern.de/ 

Court decisions

Verwaltungsgericht Hamburg, Beschluss vom 15. Dezember 2025, 2 E 8786/25 (interim relief, § 123 VwGO; not final). https://justiz.hamburg.de/resource/blob/1128534/4a02d061e1d2a47e4ff1449241831117/2-e-8786-25-beschluss-vom-15-12-25-data.pdf 

Verwaltungsgericht Kassel, Urteile vom 25. Februar 2026, 7 K 2134/24.KS und 7 K 2515/25.KS (appeal admitted; not final). Press release No. 1/2026 of 26 February 2026. https://verwaltungsgerichtsbarkeit.hessen.de/presse/unzulaessige-nutzung-kuenstlicher-intelligenz-bei-studentischen-pruefungsleistungen 

Landgericht München I, Urteil vom 11. November 2025, 42 O 14139/24 (GEMA v. OpenAI; appeal pending before the OLG München). 

Raad van State (Afdeling bestuursrechtspraak), uitspraak van 30 juli 2025, 202501708/1/A2, ECLI:NL:RVS:2025:3585. https://uitspraken.rechtspraak.nl/details?id=ECLI:NL:RVS:2025:3585 

Tribunal administratif de Paris, ordonnance du 12 février 2026, n° 2600972 (référé). 

Kammarrätten i Göteborg, dom den 14 juli 2023, mål nr 4285-23. Reported in Processrätt (2023, November 4), https://processratt.se/kammarratten-i-goteborg-upphaver-disciplinnamndens-beslut-om-fusk-med-ai-verktyg/

6. Resources & Glossary

6.2 Glossary and Key Terms

Term

Synonyms

Definition

A

Accountability

 

Being answerable for decisions about AI use and their consequences. Responsibility stays with people and institutions, never with the system.

Adaptive learning platform

 

A platform that adjusts content, pace or learning path to the individual learner.

Adaptive practice software

 

Practice software already used in schools that adjusts exercises to a learner's answers. Generative AI chatbots showed larger effects on knowledge and understanding.

Adequacy decision

 

An EU decision that a non-EU country protects personal data adequately, so that data may be transferred there.

AI

Artificial Intelligence

Computer systems that perform tasks associated with human thinking, such as recognising patterns, processing language or generating content, mostly by learning from data.

AI Assistant

 

An AI tool that carries out work on a user's behalf. It can close the gap in output between users, but not the gap in ability.

AI Chatbot

Chatbot

A program that converts in natural language, today usually based on a large language model.

AI Compliance

Compliant use of AI

Using AI in line with data protection law, the AI Act, copyright and examination law. Compliance enables use rather than hindering it.

AI Detector

 

Software claiming to identify machine-written text. What it actually measures is predictability.

AI foundations and applications

 

Conceptual knowledge and transferable skills to understand how AI works and to select, apply and adapt AI tools. The content-specific aspect of the vertical dimension.

AI literacy

 

Understanding of AI concepts, opportunities, limitations and responsible use. Under Art. 4 AI Act, schools must support its development among staff who use AI (binding since 2 February 2025); no specific level, certificate or format is prescribed. Also the name of certification Track B.

AI model

 

The trained core of an AI system that generates outputs from inputs.

AI Pedagogical Competency Framework

DUAL.AI.TEACHer Competency Framework

Structure that sorts AI-related teaching competencies into dimensions & levels and links them to outcomes & assessments, aligning pedagogical, ethical and technical aspects.

AI Pioneers

 

Framework that extends DigCompEdu with AI-specific competences for educators.

AI Tool

 

Any application that uses AI to support a task.

AI Tutor

 

An AI system designed to explain content and guide learners. Can support the acquisition of factual knowledge.

Algorithmic transparency

 

The extent to which it can be understood how an AI system reaches its outputs and who decided how it behaves.

Amplification effect

Amplification vs. equalisation

AI tends to widen existing gaps. It only equalises when three things hold together: secure access, a tool designed for learning rather than answering, and a teacher scaffolding its use.

Anonymisation

 

Changing data so that a person can no longer be identified, not even from context. Deleting names is not anonymising, especially for essays and learning journals that identify their author by style.

Assessment Activity

 

An activity used to evaluate whether learning outcomes have been achieved.

Assessment Criterion

 

A rule or benchmark used to evaluate whether a learning outcome has been achieved.

Automated decision-making

 

A decision based solely on automated processing that significantly affects a person. A grade is such a decision; formally waving through an AI's suggestion does not count as a human decision.

Automated grading

Automated scoring, automated assessment

Using software to score or grade work. In the certification exam this applies to selected-response questions.

Automation Bias

Over-trust

Accepting a machine's suggestion without independently checking it.

B

Bias

 

Slants inherited from the material a system learned from and not reliably reduced by making the system bigger.

Biometric categorisation

 

Using biometric data to sort people in order to infer protected traits such as race, political opinions, religion or sexual orientation. Banned under the AI Act.

Biometric data

 

Data from physical characteristics such as face or voice. It is special category data under the GDPR, and several AI uses based on it are banned in schools.

C

Certificate

 

A formal document recognising the achievement of competences.

Certification Pathway

 

The route through which competences are validated and formally recognised, here as a dual pathway (Track A and Track B).

Certification track

Track A, Track B

One of two complementary certifications: Track A, AI-Enhanced Teaching (WITH AI), and Track B, AI Literacy (ABOUT AI). Teachers can take one or both.

Code generation

 

Letting AI write program code. Beginners should avoid it because they cannot yet judge whether the code is correct.

Code review

AI-assisted code review

Students write their own code first and then ask AI to suggest improvements, instead of letting it write the solution. The suggestions must be evaluated critically.

Cognitive Offloading

 

Handing mental work to something outside your head. It is useful when knowledge is in place, and damaging when it is still forming.

Competence

Competency

What a teacher knows, can do & is willing to do, described so that it can be taught and checked.

Competence code

Learning objective code

Label of a learning objective, e.g. AF-TL-2b: aspect of AI (HM, EI, AF) – pedagogical area (PE, DR, TL, A, EL, FC) – progression level (1–3) and objective (a–c).

Competence Domain

Competence area

A group of related competences within a framework.

Computational thinking

Algorithmic thinking

Approaching problems in a way a computer could carry out, for example breaking them into steps and recognising patterns.

Content filter

 

An automatic mechanism that blocks or alters content a provider considers unsuitable. Its criteria are often not disclosed.

Content Unit

 

The smallest reusable learning component used within lesson plans and modules.

Continuing professional development

CPD, in-service training

Professional learning for teachers already working in schools. Stage 6 of the roadmap turns it into a structured offer with stackable micro-credentials.

Controller

Data controller

Under the GDPR, the body responsible for processing personal data. In schools this is normally the school.

Copyright

 

Legal protection of works. With AI, what comes out is usually free of copyright, but uploading protected works into an AI system is not covered by the teaching exception.

Core competency

 

One of the 18 competencies of the DUAL.AI.TEACHer framework, each at the intersection of a pedagogical area and an aspect of AI.

Course

 

A structured set of training modules designed to deliver a complete learning experience.

Course Run

 

A specific instance of a course delivery with enrolled learners and a defined timeframe.

Cyborg writing

 

Writing in which AI extends rather than replaces the writer's thinking, with students flagging which parts are their own.

D

Data hosting

 

Where data are stored and processed, e.g. on European servers or outside the EU. Data leaving the EU need a safeguard.

Data processing agreement

DPA

The written contract between school and provider that spells out how the provider may handle personal data (Art. 28 GDPR). The practical test: does it exclude training on your inputs?

Data protection

 

The legal protection of personal data. The load-bearing foundation of AI law that has applied all along; in the EU mainly the GDPR.

Deepfake

 

AI-generated or manipulated image, audio or video content that appears authentic. Must be disclosed when used in the classroom (Art. 50(4) AI Act).

Deployer

 

The organisation that introduces and controls an AI system, usually the school or school authority. A teacher using the tool under the school's instructions is normally not a separate deployer.

Deskilling

 

The slow erosion of professional expertise when a tool absorbs the work through which that expertise is maintained.

Deterministic

 

Always producing exactly the same result from the same starting values.

Diagnostic data

Telemetry

Technical data a provider collects about how its service is used. How long it is kept is a data protection question.

Diffusion of innovations

 

Theory of how innovations spread: they are adopted more readily when they can be tried out (trialability) and seen working (observability).

DigCompEdu

 

The European Framework for the Digital Competence of Educators. It provides the pedagogical structure (horizontal dimension) of the DUAL.AI.TEACHer framework.

Digital Badge

Open digital badge

The shareable digital form of a certificate or micro-credential.

Digital Hub

 

A central online platform that bundles a project's resources, courses and community exchange.

Digital sovereignty

 

Control over one's own data, infrastructure and technology choices. Building a platform oneself does not automatically give more control, so the playbook advises using the term with care.

Dimension

 

One of the broad areas the framework is divided into, each grouping competences that belong together in practice.

E

Emotion recognition

 

AI that claims to infer feelings or intentions from biometric data, e.g. attention detection, mood analytics or 'engagement' scoring from a webcam. Prohibited in schools since 2 February 2025, with narrow exceptions for medical or safety purposes.

EntreComp

 

European Entrepreneurship Competence Framework. Contributed creativity and innovation to the framework.

ESCO

 

European classification of skills, competences, qualifications and occupations. The certification is aligned with it.

Ethical Dimension

 

A dimension addressing ethical considerations related to the use of AI in teaching and learning.

Ethics of AI

 

Ethical principles, regulations, institutional policies and practical guidelines governing AI use that teachers need to understand and apply. A cross-cutting aspect of the vertical dimension.

EU AI Act

AI Act, Regulation (EU) 2024/1689

The EU's single, risk-based rulebook for AI. It classifies AI systems by the risk they pose, from banned practices through high-risk uses to everyday tools without special rules, and applies to public and private organisations, including schools. In force since 1 August 2024, phased in through 2027.

EU database

 

The EU register in which public authorities may have to record the high-risk AI systems they use (Art. 49 AI Act).

Europass

 

EU platform through which digital certificates can be shared and their validity checked by third-party institutions.

European Framework

 

A reference framework providing standardised competence definitions at the European level.

European Qualifications Framework

EQF

EU reference framework that makes qualifications comparable across countries. The certification is aligned with it.

Explainability

AI explanation

An AI system showing how it reached its output. Explanations do not reliably prevent over-trust; explanations that are too technical or too simplified can even increase misplaced confidence.

F

Fabricated citation

Invented source

A reference produced by AI to a source that does not exist. Spotting such sources is a core verification skill and one of the typical signs of undisclosed AI use.

Facial recognition

 

AI that identifies people by their face, e.g. for attendance or exam monitoring. Building or expanding facial-recognition databases by untargeted scraping is banned.

Factual knowledge

 

Knowledge of facts and concepts. AI tutors can support its acquisition.

Fairness

Non-discrimination

The principle that AI and its use must not systematically disadvantage people, e.g. AI detectors wrongly flagging non-native writers.

False Positive

 

A correct, honest piece of work wrongly flagged. The error that damages students.

Feedback uptake

Uptake

Whether feedback is read, understood, acted on and understood well enough that the learner could get it right unaided next time. Uptake, not quality, is the bottleneck.

Fundamental rights impact assessment

FRIA

An assessment that public bodies and organisations providing public services may have to complete before using high-risk AI (Art. 27 AI Act).

G

Gamification

 

Using game elements in lessons. Using AI for gamification produced a gain too small to call proven.

GDPR

General Data Protection Regulation, Regulation (EU) 2016/679

The EU's data protection law. While the AI Act asks whether and how an AI system may be used, the GDPR asks whether particular data about a particular person may be processed in a particular way.

Generative AI

GenAI

AI that creates new text, images or other content from a prompt.

Guardrails

 

Limits built into a system so that it supports thinking instead of replacing it.

H

Hallucination

 

Fluent, confident, false. Not a defect awaiting a patch but a consequence of how the systems generate text.

High-risk AI system

High-risk AI

An AI system whose intended purpose makes it subject to strict obligations. In education: deciding on admission, evaluating learning outcomes or steering learning, assessing access to education levels, and detecting prohibited behaviour in exams; for staff: recruitment, promotion, task allocation and performance evaluation. Obligations apply from 2 December 2027.

Horizontal dimension

Pedagogical dimension

The six areas of teaching practice in the DUAL.AI.TEACHer framework, taken from DigCompEdu: Professional Engagement, Digital Resources, Teaching & Learning, Assessment, Empowering Learners and Facilitating Learners' (AI) Digital Competence.

Human oversight

 

The requirement that competent, trained and authorised people supervise a high-risk AI system, understand its limits and can disregard or override its output. Oversight without expertise is not oversight.

Human-centred mindset

 

Values and attitudes teachers need in human-AI interaction to prioritise people's needs, judgement and wellbeing. A cross-cutting aspect of the vertical dimension.

I

ICEP

Institute of European Certification of Personnel

Accredited certification body (ISO 17024) that runs the certification platform and exam and keeps the certification records.

Image classification

Image classifier

An AI task in which a model sorts images into categories (e.g. paper, plastic, other) after being trained with examples, and is then tested with new objects.

Inclusion Dimension

 

A dimension focusing on accessibility, equity, and inclusion in AI-supported teaching and learning.

Initial teacher education

ITE, pre-service teacher education

The university-based education of future teachers.

Integration Roadmap

 

The step-by-step route a teacher or an institution takes from first contact with AI to settled practice; in DUAL.AI.TEACHer, nine stages.

Intelligent tutoring system

ITS

Established learning software that guides learners with tailored tasks and feedback.

Iterative prompting

Prompt refinement

Improving an AI output over several rounds instead of relying on a single prompt: define the goal, prompt, evaluate the output, refine and verify.

K

Knowledge-Action Gap

 

The measured distance between knowing about AI and teaching differently because of it.

L

Large Language Model (LLM)

LLM

An AI model trained on large amounts of text that generates plausible continuations of text; the basis of chatbots. Works as a prediction engine, not a database.

Learner

 

An individual participating in learning activities to acquire knowledge, skills, or competences.

Learner agency

 

Learners' ability to make their own decisions about their learning, including about the use of AI-generated resources. Teachers should preserve it in AI-supported processes.

Learning Outcome

 

A measurable statement describing what a learner is expected to know or be able to do.

Learning Unit

 

A concrete learning activity performed by the learner within a module.

Lesson Plan

 

A detailed plan for a teaching session, including objectives, activities, and integration of AI tools.

Licence 

 

 

LLM chat interface

AI chat assistant

The browser-based chat window through which users work with a large language model.

Locally installed model

Locally hosted model

A language model that runs on a school's own device or server. It is an alternative when internet access at school is restricted.

Logs

 

Records automatically generated by an AI system. Deployers of high-risk systems must keep them for at least six months where the logs are under their control.

M

Machine learning

ML

The approach in which an AI system learns patterns from the examples it receives instead of following hand-written rules. If the examples are too similar, incomplete or poorly chosen, the model produces unexpected results.

Market surveillance authority

 

The national authority that enforces the AI Act. Anyone, including teachers, students and parents, can lodge a complaint with it (Art. 85 AI Act).

Matthew effect

 

Those who already have advantages benefit most: 'to those who have, more will be given'.

Meta-analysis

Review

A study that pools the results of many separate studies on the same question.

Micro-credential

 

A smaller certification covering one focused set of competences instead of a whole pathway; can be stacked towards larger qualifications.

MOOC Platform

 

A digital platform used to deliver courses, manage learners, and host learning content.

N

Novelty effect

 

A temporary positive effect simply because something is new. It may inflate the results of short-term studies.

O

One-shot prompt

Single prompt

A prompt sent once, without refinement or follow-up. (In technical usage, the term can also mean a prompt containing one worked example.)

Online Exam

 

A formal, structured assessment conducted online as part of the evaluation process.

Open Educational Resource (OER)

OER

Teaching and learning material published under an open licence that allows free use, adaptation and sharing.

Open model

Open AI model

An AI model whose components are openly available, as opposed to proprietary corporate models. Recommended as part of publicly funded, equitable access to AI.

Operator

 

The verb a learning objective is built around (e.g. recognise, implement, critique). It shows what a teacher at that level is expected to do and makes the objective observable.

Overreliance

Over-trust

Trusting AI output more than it deserves, for example accepting wrong grades or feedback. Teachers using high-risk AI must actively watch for it.

P

Pedagogical Framework

 

A structured set of pedagogical principles and approaches guiding teaching and learning design.

Personal data

 

Any information about an identifiable person, e.g. names, school email addresses, student work, voice recordings, photos, grades, attendance and learning-progress data. Data remain personal if someone can be identified from context.

Personalisation

AI personalisation

Adapting resources, tasks or support to individual learners with AI. It can support inclusion but also carries risks of inequity, data misuse and exclusion.

Playbook

 

A structured guide that operationalises AI competences into teaching practice through use cases, pedagogical models, and training resources.

Policy Guideline

 

A policy-level document informing the development of competences and teaching practices.

Predictability

 

How expected the wording of a text is. AI detectors measure predictability, not machine authorship, so they wrongly flag plain or second-language writing.

Prediction engine

Probabilistic output

Language models generate a plausible continuation of a text instead of looking facts up in a database. This explains why they hallucinate and why similar prompts can produce different answers.

Processor

Data processor

Under the GDPR, a body that processes personal data on the controller's behalf and strictly on its instructions, for example an AI provider.

Proctoring

Exam proctoring, remote proctoring

Supervising examinations with software, for example via webcam. AI that detects prohibited behaviour during exams counts as high-risk; the certification exam uses remote proctoring.

Procurement

Purchasing decision

The process by which a school or authority selects and buys AI tools. Reliable limits and contractual safeguards are secured here, which is why responsibility sits with procurement, not with the individual teacher.

Profiling

 

Automated evaluation of individuals to analyse or predict their characteristics or behaviour. An education AI system that profiles individuals is always considered high-risk.

Progression level

Level

One of three levels on which each core competency is described: Level 1 Orientational Awareness, Level 2 Reflective Implementation (the primary target for teachers) and Level 3 Transformative Leadership.

Prohibited AI practices

Banned AI practices

Eight AI uses banned outright by the AI Act (Art. 5) since 2 February 2025. Four concern schools: emotion recognition, exploiting vulnerabilities, untargeted facial-recognition databases and biometric categorisation.

Prompt

 

The instruction a user gives an AI system. Its wording, context and constraints shape the output.

Prompt engineering

Prompting

Writing effective prompts as a teachable skill, for example for lesson planning, differentiation and grading.

Prompt rewriting

 

A provider automatically changes a user's prompt before the model receives it, for example to add detail or diversity. The user does not see the change, so what was typed is not what the machine was asked.

Provider

AI provider

The company that develops and offers an AI system. It is usually responsible for informing users that they are interacting with AI; when it processes data for a school, it acts as a processor under the GDPR.

R

Randomised controlled trial

RCT, controlled trial

A study in which participants are randomly assigned to groups, so that differences in results can be attributed to the intervention.

Retraction

 

The official withdrawal of a published article by a journal. The article remains visible, marked 'RETRACTED', so that anyone who cited it can recognise this.

Revised Bloom's taxonomy

 

Model in which cognitive processes develop from remembering and understanding through applying and analysing to evaluating and creating (Anderson & Krathwohl, 2001). Basis of the progression levels.

S

Scaffolding

 

Structured guidance by the teacher that supports learners in using a tool or completing a task they could not yet manage alone.

Schema

 

Person's organised knowledge.

School Administrator

School leader, head teacher

A person responsible for leading a school, such as a head teacher or deputy head. Chapter 5.3 addresses them directly.

School AI environment

Vetted platform

A state-provided, vetted AI platform for schools with negotiated contracts, offered instead of leaving teachers to sign up to free tools.

Server

 

A computer that stores and processes data for others over a network. Where an AI tool's servers are located (e.g. in the EU) affects which data protection rules apply.

Socratic tutor

 

A chatbot instructed to ask guiding questions instead of giving answers. Set up by prompt alone, it tends to drift back into supplying answers in longer conversations, so reliable limits have to be built into the system.

Special category data

 

Personal data needing extra protection, such as health, disability, biometric data or data revealing ethnic origin or religion (Art. 9 GDPR). In principle off-limits without a specific exception.

Speech recognition

 

A form of AI in which a program tries to recognise spoken words or numbers. How well it works depends on the microphone, how clearly a person speaks and how quiet the room is.

Stackability

Stackable

The possibility of combining micro-credentials over time so that they count towards larger qualifications.

Standard contractual clauses

SCCs

EU-approved contract terms that safeguard personal data transferred outside the EU when there is no adequacy decision.

Stochastic

Randomised

Subject to chance, so that the same settings can lead to different results.

Structured prompt

 

A prompt that specifies more than the topic, for example audience, purpose, context, language level, tone, learner age, learning objective, duration and expected output. It gives more control over the result but does not guarantee a good one.

Sustainable AI

AI Sustainable

Developing and using AI in a way that takes its long-term environmental, social and economic effects into account.

T

Teacher

 

An individual responsible for designing and delivering teaching and learning activities.

Teacher AI Competence

 

Teachers' knowledge, skills and attitudes for teaching with and about AI, as structured in the DUAL.AI.TEACHer framework.

Teacher educator

 

Academic staff who teach in initial teacher education. Building their capacity is Stage 3 of the roadmap.

Teaching about AI

Learning ABOUT AI

Making AI itself the subject of the lesson, so that students understand how it works & where it fails.

Teaching exception

 

The copyright exception that lets teachers copy a limited share of a work for their class. It does not cover uploading protected works into an AI system, and material intended for school teaching, mainly textbooks, is excluded.

Teaching Model

 

A pedagogical approach that defines how teaching and learning are structured, including the role of AI.

Teaching with AI

Learning WITH AI

Using AI as a tool for your own teaching work, from planning to feedback.

Technical safety

 

Reliability and robustness of an AI system so that it works as intended and does not cause harm.

Text-to-image generator

Image generator

Generative AI that creates images from a written prompt. What it shows depends on its training data and on adjustments made by its provider.

Third party

 

A company that is neither the school nor the app's developer, typically advertising or analytics companies that receive data while the pupil is working.

Training data

Training material, training examples

The examples an AI system learns from, such as millions of images labelled by people. Slants in this material carry over into the system's output and do not disappear simply because models get bigger.

Training Module

 

A structured unit of training that develops competences through learning activities and materials.

Training on user inputs

 

When a provider uses what users type or upload for its own purposes, especially to train its models. Personal data may only be entered if a written agreement excludes this.

Transparency obligation

Labelling obligation, disclosure duty

The duty under Art. 50 AI Act to tell people when they interact with AI and to disclose deepfakes and AI-generated text on matters of public interest. Content that has undergone human review and editorial control is exempt, so a teacher-reviewed worksheet needs no label.

Tuning

Model tuning, safety training

Adjustments a provider makes after initial training to steer what a system shows or says, for example to make it safer, fairer or more diverse. Tuning is invisible in the output, can miss, and can change without notice.

U

UNESCO AI Competency Framework for Teachers

 

UNESCO framework (2024) defining the principles, values, knowledge and skills teachers need to use AI ethically, effectively and safely. Source of the vertical dimension.

Universal Design for Learning

UDL

Framework for inclusive, learner-centred design. Used to strengthen the inclusive dimension of the framework.

University Faculty

 

University staff involved in teacher education, one of the playbook's target groups.

Untargeted scraping

Scraping

Automated mass collection of images from the internet or CCTV footage without a specific purpose or person in mind.

Use Case

 

A concrete scenario describing how AI is applied in teaching practice.

V

Vertical dimension

AI dimension

The three aspects of AI in the DUAL.AI.TEACHer framework, taken from the UNESCO framework: Human-centred mindset, Ethics of AI, and AI foundations and applications.

Vibe coding

 

Creating or extending software by describing in natural language what it should do and letting an AI write the code. In the playbook, students extend an existing simulation through prompts instead of programming from scratch.