Education is a special sector: Why we need ethical regulation of Artificial Intelligence and Education (AI&ED) and how we benefit from it

______________________________________________________________________ Please cite as: Stracke, C. M. (2025). Education is a special sector: Why we need ethical regulation of Artificial Intelligence and Education (AI&ED) and how we benefit from it (submitted, in peer-review). Preprint: https://doi.org/10.5281/zenodo.17279562 Education is a special sector: Why we need ethical regulation of Artificial Intelligence and Education (AI&ED) and how we benefit from it Christian M. Stracke 1,2 * [https://orcid.org/0000-0001-9656-8298] 1 University of Bonn, Germany 2 The European Institute for Learning, Information and Cooperation * [email protected] Abstract. This paper analyses the ethical dimensions for the design and evaluation of Artificial Intelligence (AI) and Education and provides related recommendations. It starts with a literature review and discussion of the relationship between AI and Education (AI&ED). We present that education is a special sector for three main reasons: Education is (1) a human right, (2) a complex ecosystem and (3) self-learning. Thus, it requires particular rules and regulations for the AI introduction and use, leading to general benefits. Ongoing developments of AI regulations are currently missing a recognition of this important demand and of the vulnerability of the educational sector. Human rights, democracy, and rule of law as the key global values should guide the ethical debate about AI&ED in combination with specific educational requirements. Based on our literature analysis and discussions, we propose to transfer and adapt the three generic learning quality dimensions to the AI&ED design and evaluation leading to the three dimensions: AI&ED objectives, AI&ED implementations and AI&ED results. In future AI&ED research, they can be validated to focus educational purposes for AI selection. Finally, we recommend three ethical imperatives for the AI&ED design and research directions: (1) impact-oriented, (2) technology-independent and (3) societyfocused to guarantee positive achievements and benefits by AI&ED for all three educational levels and involved stakeholders. All stakeholders in education should call for broad discussions to gain consensus on AI permits and prohibitions and to facilitate regulation of AI&ED reflecting the requirements of the special sector and our societal future. 1. Keywords: Ethical Artificial Intelligence and Education (AI&ED), AI use in education and AI literacy, special sector education, AI&ED regulation, global human rights, theoretical framework for ethical AI&ED, impact-oriented, technology-independent and societyfocused AI&ED design & evaluation. 1 Introduction Artificial Intelligence and education (AI&ED) are a combination that requires in-depth analysis and reflection due to the broad implications of both concepts. This paper discusses the ethical dimensions that are relevant for AI&ED and provides first recommendations how to address them in future research, design and evaluation. In section 2, we present a literature review on AI&ED as background. Afterwards, we discuss the special conditions of the educational sector and its required AI regulation (section 3) and related ethical questions in general (section 4) and current AI&ED challenges (section 5). Based on our findings, we propose a theoretical framework for ethical AI&ED research for future validation (section 6). In the conclusions (section 7), we discuss how AI&ED regulation, research and practices can support our society. 2 Background: Literature review of (ethical) AI&ED The concept of Artificial Intelligence (AI) was coincided in the 1950s (McCarthy et al., 1955) and is controversially debated since this origin (Aiken & Epstein, 2000, Chaka, 2022; Huang et al., 2023). At the start of our millennium, many branches and sectors have started to implement AI (Borenstein & Howard, 2021). It is stated that the ethical dimensions of AI need to be addressed given its anticipated high impact (Kazim & Koshiyama, 2021; Larsson & Heintz, 2020). Many organisations, initiatives and researchers started to focus and define ethical and trustworthy AI (European Commission, 2022; HLEG on AI, 2019) leading to the first global standard adopted by UNESCO (2021). First reflections are already discussing the ethics of ethical AI (Hagendorff, 2020). The concept of education exists much longer than AI (since antiquity): The AI&ED research already started more than forty years ago with key focus on AI use in school and higher education while AI literacy research came later (Hrastinski et al., 2019; Kent, & du Boulay, 2022; Luckin et al., 2022; Pinkwart, 2016). AI&ED research concentrated on intelligent tutoring systems (ITS) some decades ago (Demerval et al., 2018; du Boulay, 2016), followed by AI&ED research on several use cases including feedback functions (Benotti et al., 2018), automatic grading (Yang et al., 2019) and the overall relation between artificial and human intelligence (Baker, 2016). Several systematic literature reviews provide an overview of the current AI&ED research (Bond et al., 2024; Chen et al., 2020; Chiu et al., 2023; Crompton et al.; 2022; Deng et al., 2025; Kurdi et al., 2020; Memarian & Doleck, 2023; Sanusi et al., 2022; Stracke et al., 2024; Tang & Su, 2024; Zawacki-Richter et al., 2019). Today, international institutions including UNESCO are developing policies for AI&ED to facilitate sustainable education (Miao et al., 2021; Miao & Holmes, 2023). Many promises and advantages are declared for AI&ED, in particular by international stakeholders and organisations such as G20 (2023), OECD (2021, 2023) and World Economic Forum (2024) including: • Personalisation and motivation through diversity and adapted pathways • Individual support by AI chatbots, agents and intelligent tutor systems • Equal chances for all learners independent from human teachers • Inclusiveness for learners with special needs and opportunities • Learning risks' identification and reduction of drop-outs • Automatic assessment and grading of knowledge and skills At the same time, many problems and disadvantages are pointed out for AI&ED (Bozkurt et al., 2023, 2024; Holmes, & Tuomi, 2022; Xiao et al., 2025) including: • Traditional pedagogies, teaching and sequencing • Lack of emotional, social and collaborative aspects • Anthropomorphism and dumb trust in AI and its results • Biases and inequities due to non-representative data sources • Social scoring of learners and their comparisons and selections • Data theft and commercialisation by pure revenue interests 3 AI&ED and the special case of education Why is the educational sector a special case? There are three main reasons why education has to be carefully reflected and to be differently treated compared with other sectors: Education is (1) a human right, (2) a complex ecosystem and (3) self-learning (Figure 1). Figure 1. Three main reasons why education is a special sector. First of all, we need to underline that education is a global human right (United Nations, 1948). And it is approved and focused as the Sustainability Development Goal no. 4 (SDG 4) demanding special recognition and efforts from the public authorities (United Nations, 2015). SDG 4 calls to "ensure inclusive and equitable quality education and promote lifelong learning opportunities for all" (United Nations, 2015, 21). It includes any special needs of all pupils to be addressed what is even more challenging for learners handicapped by brain malfunctions or damages. According SDG 4, each learner should have the same educational opportunity and chance for quality education defined through the ten targets of SDG 4 and related indicators (United Nations, 2015). Second, education is not a sector like any other ones but a complex ecosystem: It does not deliver a product like all manufacturing sectors but it offers a promise for future learning processes that cannot be predicted and pre-assessed (Stracke, 2019). That is true in particular in formal school education with the mandatory participation for pupils and their age as minors. Most of the school pupils are not of legal age, cannot decide by themselves alone and require special attention and care: Thus, the UN Convention on the Rights of the Child guarantees, protects and defines education as a global right and determines and secures mandatory teaching and learning of young children (United Nations, 1990). Furthermore, the Convention on the Rights of Persons with Disabilities requires equal chances and inclusiveness in education (United Nations, 2008) independent of the age what includes school education and causes additional demands. Therefore, next to the teachers of the pupils, the school leaders, authorities and policy-makers of the educational sector have particular responsibilities to respect and guarantee the legal and ethical rights of the pupils and their parents as well as to facilitate quality education for all. Third, learners are self-constructing their knowledge, skills and competences in constant communication and interactions with their environment. This self-learning adds to the special situation in the field of (formal school) education. Next to the difficulties of evidence-based research in education, Hattie (2008) has demonstrated the extraordinary relationship between teachers and learner as well as the importance and impact of (good) feedback. On the one hand, teachers cannot force learners to learn because learners have always to build, develop and improve knowledge, skills and competences by themselves. And on the other hand, the provided education by the teachers and the learning results achieved by the learners cannot be foreseen and guaranteed as they are always depending on many (external and internal) conditions and factors (Stracke, 2019). In consequence, these three reasons explain why education is a special sector as well as why it requires particular ethical considerations, protections and regulations, especially in the school education. Those reasons are most important and the basis for the whole educational sector to safeguard and guarantee the development of responsible citizens for our future society and to ensure global human rights for all (young and adult) learners. And ethical considerations and regulations are in particular needed for AI&ED given the potential huge AI impact and the current situation in public education systems with lack of resources and decreasing investments. Furthermore, they would be beneficial because they can provide clear guidelines and guard rails for all stakeholders of the educational sector what is allowed and what is forbidden. 4 AI&ED and ethical questions When we focus on ethics in general, three main ethical approaches can be distinguished (Kronqvist, & Rousi, 2022): Applied ethics (for specific ethical questions in practice), normative ethics (for ethical concepts and theories), and metaethics (for meta-analysis of ethical theories from normative ethics). In our discussion here, we concentrate on general normative ethical questions as they are most relevant for the introduction of AI&ED: Normative ethics are often controversially debated in research as well as in politics because they connect moral principles with behavior and explain how they are permeated through socio-cultural systems, both discursively and systemically (Gewirth, 1960). In the ethics of AI, five ethical principles are debated in general and globally: transparency, justice and fairness, non-maleficence, responsibility, and privacy (Jobin, Ienca, & Vayena, 2019). And the specific ethics of AI&ED are already in the focus of international institutions and public authorities publishing first AI&ED policies (European Commission, 2019; European Parliament, 2021; Miao et al., 2021; UNESCO, 2021). Their first analysis revealed ten common ethical principles for AI&ED clustered in seven topics, namely: Governance & Stewardship, Transparency & Accountability, Sustainability & Proportionality, Privacy, Security & Safety, Inclusiveness and Human-centered AIED (Nguyen et. al., 2023). In the global AI&ED research and development community, ethical discussions were launched early but they were not gaining attention and continuation (Aiken & Epstein, 2000). The identification of the necessity of ethical AI&ED and the development of community-driven proposals and frameworks for ethical AI&ED took twenty years (Akgun & Greenhow, 2021; Borenstein & Howard, 2021; Chounta et al.; 2022; Holmes et al.; 2021, Holmes et al., 2022b). The need is confirmed by current research revealing that many legal violations and issues are occurring in 89% of international AI&ED products (Lazarus et al., 2022). However, international ethical and binding regulations for AI&ED are not yet existing and still under development (despite the long period of time of AI (as well as AI&ED) research and development before) because the broad use of AI in education started suddenly due to the new and easy usability of generative AI when OpenAI published ChatGPT on 30th of November 2022 (Bozkurt et al., 2023). The European Union started a first international initiative with the AI Act already in the year 2021 to regulate the AI providers and in the meantime, the AI Act is approved and published in the year 2024 (European Union, 2024). In parallel, the Council of Europe started another international law development to regulate the AI use with a special focus on the values of human rights, democracy and rule of law (Council of Europe, 2023a) that is approved and published as the AI Framework Convention (Treaty 225) in the same year 2024 (Council of Europe, 2024). The AI Act addresses AI providers and deployers in contrast to the AI Framework Convention focusing AI users and their human rights. Despite these different target groups, both legal instruments are developed for all sectors and have to applied in all of them (including education). While these two laws are first steps towards an international AI regulation, the global debate on AI regulation ignored education as a special sector with specific conditions and needs as explained above. Thus, the Council of Europe started a separate initiative for two international conventions on ethical AI&ED to fill this gap and appointed an international AI&ED Expert Group (Council of Europe, 2023b). The collaboration between Council of Europe and the AI&ED Expert Group and their common conferences and workshops led to first discussions and research results (Council of Europe, 2023d, 2024b; Holmes et al., 2022a, 2023; Stracke et al., 2024, 2025b). Finally, the argumentation for specific AI&ED regulations was convincing: It became obvious that next to generic AI regulations like the initiatives by the European Union (AI Act) and Council of Europe (AI Framework Convention), it is urgently required to establish binding laws and conditions specifically in the fields of education (Council of Europe, 2023b; Holmes et al., 2022a; Stracke, 2024). Consequently, the Council of Europe got the mandate from all 46 member states for a binding law on AI regulation in education and for a recommendation on AI Literacy (Council of Europe, 2023c). Both legal instruments are currently under development by the Council of Europe and its AI&ED Expert Group to ensure that future AI&ED contributes to the global ethical values guaranteeing, facilitating and supporting the human rights, democracy and rule of law (Holmes et al., 2022a, 2023; Stracke, 2024). 5 Current ethical challenges related AI&ED Today, we are facing dramatic changes in the whole world that are strongly affecting education and learning processes, too. AI is coming on top of fundamental development processes such as globalization, internet connectivity, mobile communication, social media and last but not least the global COVID-19 pandemic with lockdowns and closures of schools and universities worldwide (Bozkurt, 2022; Stracke, 2022a, 2022b). The spontaneous answers on the AI introduction ranged from immediate hype to complete prohibition and accordingly, the guidelines and policies for AI use in education were quite diverse in the beginning (Dabis & Csáki, 2024; Stracke et al., 2025a). Regarding the technological dimension of AI&ED, we have to note that AI systems are mostly targeting cognitive intelligence while Gardner (1983) had already pointed out the potential diversity in his theory of multiple intelligences. Humans have already problems to identify deepfakes in texts, pictures, audios and videos what will increase with the steadily increasing capabilities of AI systems (Groh et al., 2024). The biggest challenge is that AI is the first technology that can mimic human behaviour and that people want to and do communicate with AI systems like with human beings, often with (or despite?) the fact that they are aware of their artificial nature. Related the pedagogical dimension of AI&ED, we need to emphasize first of all, that evidence-based research results on AI use in education with positive effects at large scale are not yet existing despite all efforts while it is reported that generative AI could harm learning (Bastani et al., 2024; OECD, 2023). Thus, it is also difficult to decide and define what teachers and students should learn about AI leading to a plurality of different approaches for AI literacy (Miao & Cukurova, 2024; Miao et al., 2024; OECD, 2025). Related the human dimension of AI&ED, we need to reflect how AI can support the overarching educational objectives of personal development and capacity building to become a successful individuum and responsible citizen of our society (Burgos et al., 2024). In other sectors such as medicine and healthcare, it is suggested to move the ethical debate towards new definitions and decisions what can be considered as acceptable for human oversights across the diversity of AI applications and in particular of generative AI and Large Language Models (LLMs) such as ChatGPT (Haltaufderheide & Ranisch, 2024). In education, the ethical discussions were started later, also due its special sector conditions and lack of resources, and are currently dominated by debates about fairness, accountability, transparency, and ethics with a special focus on the impact of chatbots and future agents (Ifelebuegu et al., 2023; Memarian & Doleck, 2023). First proposals are the promotion of open education for effective and safe use of (generative) AI in education and to benefit from open educational practices as a framework to the AI use in education (Mills et al., 2023; Tlili et al., 2024). Overall, we need clear guidance and policies for AI&ED and specific regulations for the AI use in education as well as for the building of AI literacy to guarantee the global human rights, in particular for the young children and minors in schools. In particular we should ensure that there are no AI experiments with minors in school education. However, it is a current challenge that the AI regulation is generally questioned demanding that existing laws should be softened as it is discussed for the AI Act or even withdrawn as it is announced in the United States of America (Schumer et al., 2024). Therefore, it is crucial to raise and increase the public debates and internal discussions in institutions and among stakeholders why AI regulation is required as well as beneficial for individual human rights as well as for the society. We need to define what we want to allow and what we want to prohibit what will also ease the AI use. For the special educational sector, practical guidelines developed by the AI&ED community (such as Stracke et al., 2024) are valuable instruments for achieving consensus on the two main questions how we want to ethically use AI in education and how we learn about AI. 6 Theoretical framework for ethical AI&ED To design and validate future theoretical frameworks in emerging fields like AI&ED, grounded theory research is recommended as a comprehensive strategy (Cresswell, 2014). Adopting this approach, we are applying generic dimensions to a specific sector (in our case AI&ED) to propose a theoretical framework for future research and validation (in our case an ethical framework for AI&ED) that will follow the constructive grounded theory by Charmaz (2006). Our basis is the existing adaptation of the general quality dimensions (potential, process, and result) to the educational sector (learning objectives, learning realizations, and learning achievements) by Stracke (2019). We propose to transfer and define them for the fields of AI&ED as: AI&ED objectives, AI&ED implementations, and AI&ED results. Our literature review and discussion of current challenges revealed that AI is raising and causing many ethical concerns and questions. For the special case of the educational sector, we could identify the specific need and requirements for AI&ED regulation due to the three conditions of the sector (human right, complex ecosystem and self-learning) to ensure education as global value and human right for all. In any design, development and evaluation of AI&ED, the three transferred AI&ED dimensions (objectives, implementations and results) should be considered and addressed. Based on our reflections of the global values and ethical questions related AI&ED, we recommend three corresponding ethical imperatives and research directions for the three AI&ED dimensions: Future AI&ED should be impact-oriented, technology-independent, and society-focused (as presented in Table 1). Table 2. Three AI&ED dimensions and recommended ethical imperatives. Learning quality AI&ED dimensions Ethical imperatives Learning objectives AI&ED objectives Impact-oriented Learning realisation AI&ED implementations Technology-independent Learning achievement AI&ED results Society-focused The first ethical recommendation demands impact-oriented AI&ED to focus potential effects and benefits for all involved stakeholders and institutions while planning AI&ED implementation and objectives. Consequently, policy makers, public authorities and teachers, responsible for AI&ED introduction, should always analyse the potential personal, organisational and social implications and consequences of AI&ED to avoid any AI risks including commercial abuse for individual profits (Xiao et al., 2024). And everybody should become aware of the fact that each AI use in education as well as in any other sector is always initiated and accounted by human beings and human decisions can and should regulate AI&ED for the results in favour of our whole society (Holmes et al., 2022a, 2023). The second ethical recommendation calls for technology-independent AI&ED. There should always be a precise definition and decision of the intended learning design before implementing AI&ED (Bozkurt et al., 2024). Only technology-independent concepts can provide answers on the questions how AI should be implemented and used in education (AIED) and how education about AI (AI literacy) should be designed and provided. In education, technology should not be applied only due its possibility but always as support of and in relation to learning objectives defined in advance (Stracke, 2019). The third ethical recommendation requests society-focused AI&ED to address planned achievements on all educational levels already during AI&ED design and realisation. Learners should get to know the shortand long-term impacts of AI&ED during the AI use in education (AIED) and through education about AI (AI literacy) (Nguyen et al., 2023). That should cover chances and risks of AI use in general and opportunities for building and improving AI literacy and digital competences for all learners and citizens (Bozkurt et al., 2023). These three ethical recommendations for the AI&ED design are also applicable for the evaluation of future AI&ED research to assess and measure the achievements of AI&ED in practice. They are simple but important as design conditions and evaluation criteria for a trustworthy and ethical use and reflection of AI&ED. 7 Conclusions: Recommendations for future AI&ED research The famous report "Learning: the treasure within" underlines that education in general (considered as a fluid relationship between teachers and learners) requires careful thinking and ethical reflections due to the given importance for future developments of individual learners as well as of our upcoming generations of society (UNESCO, 1996). That is even more true for the fields of AI&ED: In general, human beings should be able to know, recognise and reflect any use of AI, its implementation and result. Next to the "standard" ethical questions that are relevant for all sectors (including education), the educational sector (with its teaching and learning processes and their conditions and settings) is raising additional ethical challenges. Thus, there are much more ethical issues related AI implementation and impact in the educational sector than in other sectors (Holmes et al., 2022a; Stracke, 2025). Many open ethical questions are raised while the AI&ED development is continuously progressing and accelerating. We could demonstrate why the educational sector is a special case and why it requires specific AI regulation to guarantee human rights and safeguard the objectives of the public education. And we could highlight that such AI regulations would be beneficial because they can provide clear guidelines and guard rails for all stakeholders what is allowed and what is forbidden. To inform our discussions and decisions about AI&ED regulation, we need evidence-based research at large-scale in safe environments. While research including systematic literature reviews on general AI&ED have already analysed specific research questions as explained above, the focus on ethical aspects of AI&ED is just starting now (Bond et al., 2024; Stracke et al., 2023, 2024). Future research on ethical AI&ED could and should validate (or reject) the ethical AI&ED dimensions of the proposed theoretical framework. Furthermore, it has to critically reflect general challenges related to AI such as increasing monopolism and power concentrations, energy and water consumptions, black box systems (due to hidden system prompts), and investments of resources that are lacking for other (educational and societal) objectives and tasks. In this way, future AI&ED research can support and serve the global and sustainable society as the broad and ambitious objective for future AI developments and uses (Chen et al., 2020; Chounta et al., 2022; European Parliament, 2021; Miao et al., 2021). To achieve this goal, we require in particular society members and citizens with AI literacy and digital competences understanding, facilitating and improving our sustainable society (Holmes et al, 2022a, 2022b; Holmes & Tuomi, 2022; Stracke et al., 2022a, 2022b). We need AI&ED following and realizing the three ethical recommendations to achieve our long-term and sustainable learning objectives and positive personal and societal impacts. Overall, all stakeholders in education should call for broad discussions to gain consensus on AI permits and prohibitions and to facilitate regulation of AI&ED reflecting the demands of the special sector and our societal future. 95. UNESCO, UNICEF, & The World Bank. (2020, October). What have we learnt? Overview of findings from a survey of ministries of education on national responses to COVID-19. http://tcg.uis.unesco.org/wp-content/uploads/sites/4/2020/10/National-EducationResponses-to-COVID-19-WEB-final_EN.pdf 96. UNESCO, UNICEF, The World Bank, & OECD. (2021, June). What’s next? Lessons on education recovery: Findings from a survey of ministries of education amid the COVID-19 pandemic. http://covid19.uis.unesco.org/wp-content/uploads/sites/11/2021/07/NationalEducation-Responses-to-COVID-19-Report2_v3.pdf 97. World Economic Forum (2024). Shaping the Future of Learning: The Role of AI in Education 4.0. https://www3.weforum.org/docs/WEF_Shaping_the_Future_of_Learning_2024.pdf 98. Xiao, J., Bozkurt, A., Nichols, M., Pazurek, A., Stracke, C. M., Bai, J. Y. H., ..., & Themeli, C. (2025). Venturing into the Unknown: Critical Insights into Grey Areas and Pioneering Future Directions in Educational Generative AI Research. TechTrends, 69(3), 582-597. https://doi.org/10.1007/s11528-025-01060-6 99. Yang, Y., Xia, L., & Zhao, Q. (2019). An automated grader for Chinese essay combining shallow and deep semantic attributes. IEEE Access, 7, 176306–176316. https://doi.org/10.1109/ACCESS.2019.2957582 100. Zawacki-Richter, O., Marín, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education – Where are the educators? International Journal of Educational Technology in Higher Education, 16(1), 1– 27. https://doi.org/10.1186/s41239-019-0171-0