A working group at the Centre for Ethics, University of Tartu, has completed an independent analysis, commissioned by the AI Leap Foundation, of the risks associated with the AI Leap educational programme and its AI-based learning application. The analysis describes the risks involved in using AI in education and assesses how AI Leap, the state, schools, research institutions and technology providers can mitigate them.
The 12 risks examined in the analysis are divided into educational risks, societal risks, and technology and governance risks. Among other areas, they concern the goals of learning and the reliability of assessment, the role of the teacher, the sustainability of the Estonian language and culture, data protection, and the reliability and sustainability of technological solutions.
The central conclusion of the analysis is that AI can strengthen Estonian education only if educational goals guide the use of technology – not the other way round. Both the mitigation of risks and the programme’s actual impact on learning, teachers’ work, educational equality, and the Estonian language and culture must therefore be assessed regularly.
According to Margit Sutrop, Head of the Centre for Ethics and Professor of Practical Philosophy at the University of Tartu and leader of the analysis working group, the analysis brought together the expertise of philosophers, computer scientists, lawyers and social scientists. The AI Leap educational programme was assessed against both the goals and values of Estonian education and the European Commission Expert Group on AI’s Ethics Guidelines for Trustworthy AI.
“This allowed us to ask not only whether the technology is safe and reliable, but also whether it helps achieve the goals of Estonian education. The risk assessment should become a management tool for AI Leap, not a document that is put on a shelf after publication,” Sutrop said.
She emphasised that responsibility is shared between AI Leap, schools and school operators, state institutions, research institutions and technology providers.
The educational risks include the possibility that the use of AI may undermine academic integrity, the reliability of assessment and the teacher’s pedagogical leadership.
The societal risks concern the Estonian language and culture, the values mediated by AI, and human interaction. Large language models developed abroad may not sufficiently reflect the Estonian curriculum, Estonian-language learning resources or the local cultural context.
Technology and governance risks include potential data breaches, rising costs, service outages and dependence on external service providers.
The AI Leap Foundation is already working to mitigate several of the risks described. The pedagogical principles of the learning application have been developed together with Estonian education researchers, and the application is intended to support teachers’ work. Teachers are provided with training, subject-specific learning materials and support from in-school professional learning communities. Together with the Institute of the Estonian Language, AI Leap is assessing the conformity of language models with Estonian-language and cultural criteria.
According to Ivo Visak, CEO of the AI Leap Foundation, the completed independent analysis gives the programme a comprehensive overview, helps identify threats at an early stage and shows where existing activities need to be strengthened. He said the aim is to give teachers, students and parents the knowledge they need to make informed choices about the use of AI, including deciding when not to use it.
“As AI is developing rapidly, we must work with schools, the state and other partners to continuously monitor how these solutions perform in teaching and learning. We must be prepared to adjust our choices so that the Estonian education system continues to provide high-quality education amid technological change,” Visak said.
The risk assessment was prepared by Margit Sutrop, Head of the Centre for Ethics and Professor of Practical Philosophy at the University of Tartu; Katrin Laas-Mikko, AI Ethics Specialist at the Centre for Ethics; Dan Bogdanov, Research Director at Cybernetica AS; Liina Kamm, Senior Researcher at Cybernetica AS; Henrik Trasberg, Legal and Policy Officer at the European Commission AI Office; and Andra Siibak, Professor of Media Studies at the University of Tartu.
The risk assessment and its main recommendations are available on the AI Leap website.