Last year, the AI Leap education program was launched at the initiative of President Alar Karis, with the goal not merely of giving schools another digital tool, but of creating a framework for the meaningful use of artificial intelligence.

AI Leap aims to develop young people's learning skills, improve the digital competence of students and teachers and prevent artificial intelligence from being used merely to obtain answers and simulate learning. This is a highly necessary goal because there is a real risk that the convenient use of AI will weaken independent thinking, the willingness to put effort into learning and the sense of academic integrity.

Socratic AI leading through questions

The AI Leap learning application being developed together with OpenAI is based on the Socratic model. Like the ancient Greek philosopher Socrates, the application does not simply provide answers, but instead asks questions that help the learner arrive at the answer independently. It encourages, acknowledges and guides the learner to think further. It reminds me of teacher Laur from "Spring" by Oskar Luts, who told Toots: "If you can't finish the whole problem, do half — but do it yourself."

Teacher Laur was a good teacher not because he knew the answers, but because he saw every student as a person, recognizing their strengths, weaknesses, fears and potential. He guided students toward growth, noticed their needs and created a safe yet demanding learning environment.

On the surface, an AI-based learning application may resemble teacher Laur. It is patient. Questions can be asked endlessly. It never tires of explaining. It helps unpack difficult texts, offers another example and yet another explanation. Such support can be extremely important for learners who are afraid to ask questions in class, who need more time or whose support at home is weaker. AI can help make learning more personalized and accessible.

At the same time, the use of AI-based learning applications in schools also involves risks, which our working group at the University of Tartu Center for Ethics has been analyzing.* At the conference "Artificial Intelligence in Education: Opportunities and Challenges," held in the Riigikogu on May 13, we presented the preliminary findings of our working group and outlined ten risks. Below, I will examine some of the risks related to education.

Will the teacher's authority stand?

With the arrival of artificial intelligence, a parallel pedagogical authority emerges within the learning process — a system that explains, guides, provides feedback and influences learning outside the teacher's field of vision.

Yet a teacher's role is not limited to transmitting knowledge. A teacher designs the learning process, notices misconceptions, supports motivation, shapes values and creates the space and human relationship necessary for learning. If some of these functions are transferred to a technological system, it is not only the method of learning that changes — the role of the teacher changes as well.

A pedagogical problem arises when learning shifts beyond the teacher's view. A teacher may still see the final result, but no longer the learning process itself: where the student made mistakes, what kind of help they received, what they actually understood and what misconceptions emerged. The teacher may no longer know whether the student solved the task independently or whether AI guided them step by step to the answer.

This weakens the teacher's ability to provide formative feedback, identify shared difficulties within the class and adapt instruction according to actual needs.

The AI Leap project has been designed so that students interact directly with the learning application, while teachers and schools do not have access to those conversations. Such a solution justifiably protects the student's privacy. A student may discuss insecurities, health concerns or other personal matters with AI and they must have the opportunity to do so without constant monitoring.

At the same time, however, a new question emerges: how can a teacher remain responsible for a student's development if part of the actual shaping of the learning process takes place within a system over which the teacher has no substantive control? In education, responsibility and oversight cannot be fully separated.

A second risk concerns the teacher's professional role. If teachers increasingly delegate explanations, feedback, the creation of learning materials or recommendations about learning strategies to a technological system, their own didactic competence and professional judgment may gradually weaken as well. The teacher's role may shift from pedagogical leader to manager of a technologically driven learning process. This concerns the future of the teaching profession as a whole.

What about the student–teacher relationship?

A third risk concerns the learning relationship between teacher and student. AI may appear more patient, more quickly accessible and less judgmental than a human being. A student may experience it as a more comfortable interaction partner than a teacher. This can be beneficial in the short term, especially for those who are afraid of making mistakes or asking questions.

But learning involves more than acquiring knowledge — it also means communication, responsibility, tolerating criticism, pushing beyond one's limits and developing a trusting relationship with a teacher, that is, with another human being. If students begin to prefer AI to teachers, the human learning relationship may weaken, even though it plays an important role in shaping motivation, self-confidence, values and social development. The humanization of AI may lead to students becoming alienated from the world and unable to form trusting relationships with other people.

Sense of criticism and honesty are important

Monitoring conducted within the AI Leap program shows that students are increasingly turning to AI for consultation and asking teachers for help less often. Based on his research, Jaan Aru has written that students trust what AI tells them uncritically. As a result, they may fail to detect misinformation or AI hallucinations, since AI can provide highly convincing answers even when they are actually false or poorly connected to sources.

We should ask what effect this may have on students' character. Until now, we have encouraged people to honestly admit when they do not know something and we have considered such honesty worthy of recognition because it motivates people to seek knowledge and grow. AI does not set such an example. If you tell it that its answer is incorrect, it apologizes and simply offers a new answer.

Academic integrity is also a concern because AI may provide a text in response to a student's prompt without indicating where the material came from, or it may present false, fabricated citations.

Recently, one of my colleagues from Iceland wrote to say that they had discovered an article citing one of their books. The title was closely related to their field of research and it seemed plausible that they could have been the author. Unfortunately, they had not written it and no such book even existed. This is why it is extremely important to develop critical thinking and diligence so that people do not use AI-generated text uncritically.

It is also extremely important to motivate students to think for themselves and to write independently. If learners become lazy and allow AI to write texts for them, we abandon thinking itself and may unintentionally reproduce someone else's ideas and words without knowing who the real author is. Presenting another person's text as one's own is dishonest and unfair to the original author.

The purpose of education

Estonia's Education Development Plan 2021–2035 states that the goal of education is to shape people with the knowledge, skills and attitudes needed to fulfill themselves in personal life, work and society. According to the Basic Schools and Upper Secondary Schools Act, the purpose of upper secondary education is not only to prepare students to enter the labor market, but also to develop creative, responsible, socially mature and ethically minded individuals.

If a machine can write a text, then the human ability to evaluate that text's meaning and truthfulness becomes more important. If a machine can provide an answer, then the human ability to judge whether that answer is reliable and fair becomes more important. If a machine can simulate empathy, then genuine human contact becomes more important.

The OECD's new future-of-education framework, Education for Human Flourishing (2025), likewise emphasizes that in the age of AI, education must strengthen human meaning, self-determination and ethical decision-making.

Education must develop the ability to understand the world, evaluate values, solve problems and act responsibly. The most important competencies are considered to be adaptive problem-solving skills and ethical reasoning.

The field of education and society as a whole face many questions that require serious reflection.

How can AI be used in a way that makes people intellectually greater rather than smaller? In practical terms: what kinds of assignments should students be given when AI can generate a text, summary, code, image and analysis within seconds? How should grades be assigned when the final result no longer reliably shows what a student can actually do independently? How can academic integrity be preserved without creating a culture of fear and concealment?

How can student privacy be protected while still understanding what is actually happening in their learning process? How should teachers be trained so that AI does not simply become yet another additional burden for them?

A few proposals

First, assignments must be redesigned so that students cannot simply submit AI-generated work, but instead must demonstrate their own line of thinking.

Second, assessment methods must change. Greater emphasis should be placed on the process itself, justification, oral defense, source criticism and the ability to explain honestly how AI was used in the submitted work.

Third, teachers must be given not only technical training, but also new didactic support: how to teach in a situation where students constantly have access to a system capable of answering questions, providing explanations and offering encouragement.

And finally, one principle must be stated very clearly: AI must support a teacher-led learning process, not become a parallel pedagogical authority. A recent document adopted by the Council of Europe describing the teacher's role in the age of artificial intelligence likewise emphasizes that AI must serve as a tool for teachers, not replace them.

AI Leap is a bold educational innovation for Estonia, but its success should not be measured by the usability of AI. We should measure success by whether students become better learners, whether they learn to ask better questions, evaluate the reliability of answers, justify their positions, use AI honestly and take responsibility for their own learning.

AI can provide answers, but schools must teach students how to understand whether an answer is correct, what to do with it and how to act well and fairly on the basis of that knowledge. AI must support, not replace, the human learning relationship. AI Leap must become an opportunity to rethink what Estonian schools teach, how they teach and for what purpose.

* Project: "Analysis of the Ethical, Social, Legal and Technological Risks and Value Choices Related to the AI Leap Learning Application and Education Program." Duration: January 22–June 30, 2026. Commissioned by: AI Leap Foundation. Conducted by the working group of the University of Tartu Center for Ethics (Margit Sutrop, Dan Bogdanov, Katrin Laas-Mikko, Liina Kamm, Andra Siibak, Henrik Trasberg).

--

Follow ERR News on Facebook and X and never miss an update!