Understanding the power and limitations of AI has become crucial for agentive citizenship as well as for the sustainability of democratic societies (Coeckelbergh, 2023; Hintz et al., 2019). Today, many aspects of everyday life, such as our daily interactions, actions, and information of all kinds, are increasingly being tracked, mediated, augmented, produced, and regulated by algorithmic governance (Tedre et al., 2021). The ubiquity of AI-based technologies, coupled with massive-scale data collection, has also given rise to complex legal, ethical, environmental, and social challenges, such as total surveillance, hybrid influencing, behavior engineering, algorithmic biases, exacerbation of social inequities, and diminishing  trust in media, public authorities, and science (Solaiman, 2023; Zuboff, 2015). Despite the grip of AI- and data-driven technologies on people’s everyday lives, most people remain unaware of how AI is mediating and engineering our epistemic environments, knowledge creation processes, and social practices (Coeckelbergh, 2023).

With the rapid proliferation of AI technologies and especially generative AI, the urgency of AI education initiatives has been broadly acknowledged, for example, by UNESCO (Miao & Shiohira, 2024), OECD (2025) and EU AI Act (Article 4).  However, many current approaches in AI education and educational research focus primarily on the instrumental use of AI tools in educational settings, rather than cultivating deeper level understanding of how AI and data-driven systems work, how they shape our everyday decision-making, and what ethical, societal, and environmental impacts they reinforce. Despite the urgency of increasing people’s understanding of their data-driven world and the risks of failing to do so, there is a noticeable lack of research-based pedagogical models and curriculum materials for AI education. Although several frameworks and content listings for AI education have been proposed, they remain untheoretical, lack deep grounding in empirical research, and tend ignore that education is always context-bound, shaped by the complex interplay of individuals, communities, and cultural practices (Vartiainen & Tedre, 2025). 

Many existing initiatives for AI education also follow top-down models and traditional interventionist research. In such linear views, teachers are left to adapt to approaches that ignore their voices, expertise and everyday realities, as well as the individual, communal, and structural contradictions and challenges that evidently arise when transforming existing activity systems and social practices (Engeström & Sannino, 2010). Accordingly, such an approach to educational change ignores the agency of the teachers and the varied context in which they work, making it unlikely to achieve a transformative and durable impact on a long-run (Engeström, 2011). Moreover, previous research has also evidenced that educational reforms are often ineffective because they tend to focus on isolated elements while disregarding the complex interplay of wider educational structures, such as national policies and curricula, and various contextual factors, such as local school practices, goals, and values, which all shape the everyday realities of teachers (Härkki et al., 2021).

This paper presents a three-year educational research and development project in which researchers from multiple disciplines collaborated closely with teachers from 12 Finnish schools to co-design pedagogical models, educational technology, and curriculum materials for contextualized AI education. Grounded in cultural-historical activity theory (CHAT) and formative interventions (Engeström & Sannino, 2010), our approach fundamentally differed from many existing AI education initiatives by aiming to support both students’ and teachers’ transformative agency.  Transformative agency denotes social transformations where people are collectively overcoming the status quo by developing stances and voices, exerting influence, and taking collective action (Haapasaari et al., 2016; Stetsenko, 2019). Instead of top-down educational reforms, our approach was based on sustained participatory design, through which we aimed to co-create transformative learning opportunities and curriculum materials  for the common good.

The three-year (2023-2025) co-design process involved teachers and school principals from 12 primary and secondary schools, well as over 200 students (4th and 7th graders at the beginning of the project). While the co-design process was open-ended and emergent by nature, it unfolded through iterative phases of 1) analyzing contextual factors and defining educational needs, 2) designing solutions, 3) implementing and testing them in school projects, and 4) reflecting on and evaluating both processes and outcomes of it. Through this iterative process, initially vague ideas about AI education were progressively concretized through the interplay and co-evolution of problem analysis (questioning existing practices and framing problems), future-oriented solution generation (imagining alternative possibilities and modeling new solutions), and practical future-making (designing and testing tools, artifacts, and pedagogical solutions in action) (Engeström & Sannino, 2010). During the project, we organized regular meetings, where we shared ideas and expertise coming from computer science as well as from educational research and practice. To support the development of shared goals and understanding, participants were prompted to collectively envision the key elements of a desired activity system for each school project. This included: 1) the higher-level objectives and learning tasks or problems assigned to students; 2) the main phases of the learning activities; 3) the tools, technology, and curriculum materials provided; 4) the forms of social organization (e.g., individual, small-group, and whole-class activities); and 5) the division of labor between teachers and researchers, including the practical coordination of project activities. The school projects were implemented through co-teaching, providing opportunities to observe the emerging learning activities in practice. In addition, various kinds of research data were collected from classroom activities, including pre-, post-, and delayed post-tests, learning tasks, video recordings, and interviews. These longitudinal data were collected following the same students from 12 classrooms over the three-year period. After each school project, joint reflection meetings were held with teachers to evaluate how the activity system surrounding our co-designed activities, tools, and curriculum materials was forming and how it should be improved. During these meetings, preliminary research findings were also presented. This collaborative reflection informed the iterative design of next steps needed for building cumulative AI learning pathways.

To this end, we have co-designed, implemented, and empirically evaluated a cumulative learning pathway for contextualized AI education. The pathway began with foundational AI concepts such as data, classifiers, confidence, brittleness, and bias (Kahila et al., 2024). In the second year, the focus proceeded to social media mechanisms encountered in students’ everyday lives, including data collection, profiling, and recommendation systems (Vartiainen et al., 2025). In the third year, students critically examined broader societal implications of AI and social media, such as filter bubbles, manipulation, and intended and unintended impacts. Our empirical findings have shown significant progress in students’ data-driven understanding, including their ability to explain algorithmic bias (Vartiainen et al., 2025) and key social media mechanisms such as data traces and profiling (Vartiainen et al., 2025). The results from our research and development work show that the design of meaningful educational practices, tools, and technologies depends fundamentally on contextualization and long-term collaboration that positions teachers as transformative agents of change. Our three-year collaboration began with a shared exploration of the status quo, including baseline studies and literature reviews that were discussed with teachers to stimulate engagement and collective design. Breaking away from the status quo was actualized by envisioning what could be, negotiating what ought to be, and taking collective action to realize those futures (Haapasaari et al., 2016). By engaging people from different backgrounds and with a variety of expertise in collective action, the larger national research project behind the co-design activities has developed pedagogical models and curriculum materials as well as several educational technologies that are openly available to anyone interested (https://gen-ai.fi/en).

Coeckelbergh, M. (2023). Democracy, epistemic agency, and AI: political epistemology in times of artificial intelligence. AI and Ethics, 3(4), 1341–1350. Engeström, Y. (2011). From design experiments to formative interventions. Theory & Psychology, 21(5), 598–628. https://doi.org/10.1177/0959354311419252 Engeström, Y., & Sannino, A. (2010). Studies of expansive learning: Foundations, findings and future challenges. Educational Research Review, 5(1), 1–24. Haapasaari, A., Engeström, Y., & Kerosuo, H. (2016). The emergence of learners’ transformative agency in a Change Laboratory intervention. Journal of Education and Work, 29(2), 232–262. Härkki, T., Vartiainen, H., Seitamaa-Hakkarainen, P., & Hakkarainen, K. (2021). Co-teaching in non-linear projects: A contextualised model of co-teaching to support educational change. Teaching and Teacher Education, 97, 103188. Hintz, A., Dencik, L., & Wahl-Jorgensen, K. (2019). Digital Citizenship in a Datafied Society. Polity Press. Kahila, J., Vartiainen, H., Tedre, M., Arkko, E., Lin, A., Pope, N., Jormanainen, I., & Valtonen, T. (2024). Pedagogical framework for cultivating children’s data agency and creative abilities in the age of AI. Informatics in Education. Miao, F., & Shiohira, K. (2024). AI Competency Framework for Students. UNESCO. OECD (2025). Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education (Review Draft). OECD, Paris, France. Solaiman, I. et al. (2023). Evaluating the social impact of generative ai systems in systems and society. ArXiv Preprint ArXiv:2306.05949. Stetsenko, A. (2019). Radical-Transformative Agency: Continuities and Contrasts With Relational Agency and Implications for Education. Frontiers in Education, Volume 4-2019. Tedre, M., Toivonen, T., Kahila, J., Vartiainen, H., Valtonen, T., Jormanainen, I., & Pears, A. (2021). Teaching machine learning in K-12 Classroom: Pedagogical and technological trajectories for artificial intelligence education. IEEE Access, 9, 110558–110572. Vartiainen, H., Kahila, J., Pope, N., López-Pernas, S., Valtonen, T., & Tedre, M. (2025). Long-term Impacts of K-12 AI Education Interventions on Social Media Mechanisms. Proceedings of IEEE Frontiers in Education Conference (FIE), Nashville, TN, USA, 2025. Vartiainen, H., Kahila, J., Tedre, M., López-Pernas, S., & Pope, N. (2024). Enhancing children’s understanding of algorithmic biases in and with text-to-image generative AI. New Media & Society, 27(9), 5342-5368. Vartiainen, H., & Tedre, M. (2025). The CEDE Model: A Learning-Sciences Based Framework for Critical and Transformative K–12 AI Education . 25st Koli Calling International Conference on Computing Education Research, Koli Calling ’25. Zuboff, S. (2015). Big other: Surveillance capitalism and the prospects of an information civilization. Journal of Information Technology, 30(1), 75–89.