Free and open education is based on progressive educational ideas from last century and is closely related to the principles of free and open-source software (FOSS) and open science. These approaches share a focus on aspects such as free access, independent control, and collaborative cooperation. Generative AI systems for education, on the other hand, pursue different goals: relief, efficiency, and individualization. What impact do these different educational principles have on attitudes, values, and practices in everyday education? Anne-Sophie Waag advocates edu-hacking and socialized AI infrastructures as a means of counteracting the advance of big tech in education.
*
It has been clear to most people in Germany since the PISA shock of 2001, if not before, that the state of the education system is not very good. Since then, there has been a steady stream of alarming reports about teacher shortages, sluggish digitization, and high school dropout rates.
In this respect, it is not surprising that educational practitioners and policymakers are clinging to any hope they can find. As in other socio-political fields, such as healthcare and sustainability, technological innovations are seen as a source of hope. It would be a great relief if the many complex structural challenges that have come to a head in recent years could be solved with a technological solution – at the touch of a button, so to speak. However, since this has not yet been achieved, there is a widespread assumption that the only thing missing was the “right” technologies.
Here to stay?
The wait for salvation seems to be over because, over two years ago, a technology that was new to most of the world’s population became accessible. Generative AI systems, such as chatbots and image generators, have functionalities that were previously unknown. Unlike previous systems, people now have access to machines that can perform a variety of tasks in real time and simulate human communication. However, these systems function in a way that is anything but human. They draw on huge amounts of data and develop response patterns based on statistical probability calculations.
Rather than raising doubts, this fuels imagination about the potential of generative AI systems in everyday education. Various ideas can be found in public papers and statements from schoolbook publishers, such as Cornelsen; technology companies, such as Microsoft; and even the Conference of Ministers of Education. With comparatively little expenditure of resources, all learners could be provided with a personalized, adaptive tutoring system that is infinitely “patient” and can answer any question at any time. Alternatively, teachers could be relieved of the burden of lesson planning, providing feedback, and making corrections, freeing up time to better care for their students.
Although these promises are still empirically shaky and need further research, technology companies have successfully marketed their products and services to the education market using these narratives. Microsoft recently announced a partnership with the state of North Rhine-Westphalia to launch an AI ‘skilling initiative’ for around 200,000 teachers. Similar news is emerging from the US, where Microsoft, OpenAI, and Anthropic, in collaboration with two teachers’ unions, have announced an ‘AI offensive’ to train approximately 400,000 educators.
Education as a government responsibility – or as a market?
The involvement of private-sector players raises the fundamental question of whose interests are prioritized and the extent to which profit-oriented companies can make decisions in the public interest. Ultimately, we must ensure that these collaborations do not lead to the privatization of educational content or goals. After all, education encompasses more than just the provision of technology or infrastructure; it is also a central venue for transmitting democratic values and facilitating personal development.
In Germany, education is fundamentally the responsibility of the state. Therefore, this issue goes far beyond the question of which software a school uses for teaching or lesson preparation. It concerns the independence of the education system and control over content, data, and learning processes. It is also about the values, attitudes, and practices that are promoted in the education system and the vision of education that we, as a society, want to promote today and tomorrow – in contrast to the tech industry.
Two images of education: efficiency versus emancipation
A comparison made by Daniel Otto in his keynote speech at the “OER im Blick” conference of the Federal Ministry of Education and Research on May 13, 2025, provides an exciting starting point for thinking about different ideas of education. The comparison illustrates the narratives and implicit notions of education and learning, as promoted by the tech industry, particularly in the context of marketing generative AI systems for educational purposes. These are contrasted with findings from educational science on the prerequisites and processes of learning.
It becomes clear that two fundamentally different images of education are colliding. The tech industry’s narratives are certainly seductive; they promise relief, efficiency, and individualized support in a system that has suffered from overload and a lack of resources for years. However, the price is high. If we increasingly outsource education to proprietary, non-transparent systems, we risk becoming dependent on a few corporations and losing central pedagogical practices and values represented by free education.
Therefore, it is imperative for the free education community to not only react to socio-technological developments but also proactively develop and implement its own visions that align with the principles of emancipatory and democratic education, particularly in the context of generative AI in education.
AI-Edu Hacking
The principles of free education include free access to and independent control of materials and systems, as well as collaborative cooperation. When applied to technology, such as generative AI, these principles give rise to various requirements and needs.
First, the principle of free access necessitates open AI systems deliberately designed for transparency, co-creation, and values oriented toward the common good. One way to ensure the provision and maintenance of necessary AI infrastructure is through cooperation between public institutions and civil society. These infrastructures could be financed through cooperative financing models, such as those used in public broadcasting, or through cooperatives, which would ensure independence from private investors. Based on this socialized infrastructure, educational institutions could work with development teams to create their own participatory, needs-based, open AI tools with the support of teachers and learners. These tools would, in turn, be published under free licenses, making them available to other institutions as well. Funding could come from public support funds. These ideas were developed through a participatory process by Wikimedia Germany (2024).
In addition to this approach, which challenges the dominance of non-transparent, proprietary generative AI systems in education, open and transparent AI systems already exist. The free education scene could give these systems much greater attention. Symbolic AI, or rule- and knowledge-based systems, works with explicit rules, logical conclusions, and structured knowledge databases. This is in contrast to the opaque, black-box models of neural networks. This approach is particularly valuable for emancipatory, democratic education because it is more transparent and controllable and enables more active learner participation.
A key advantage of symbolic AI is its traceability. Since the systems are based on clearly defined rules and knowledge bases, learners and teachers can understand exactly how an answer is arrived at. A rule-based tutoring system can deliver a final result and explain each step of the solution in detail, showing which logical principles or mathematical laws were applied. This transparency promotes understanding and a critical attitude toward technology. Learners recognize that AI systems are not magical, but rather, are based on verifiable structures. Rule-based systems also require significantly fewer resources, meaning decentralized solutions can be developed, such as knowledge graphs based on open-source databases like Wikidata or Wikibase. These systems are cost-effective and can be designed to protect privacy because they do not need to collect user data to ‘learn.’
Potential for hybrid approaches
Of course, symbolic AI has its limitations. It is less flexible than generative models when it comes to creative tasks, such as writing texts or generating images. However, this is precisely where hybrid approaches can be useful. Open, generative AI can facilitate creative brainstorming processes, and symbolic systems can offer fact-checking, explanations, and structured exercises. This combination would leverage the strengths of generative AI, such as its ability to paraphrase complex relationships, while maintaining the transparency and control of symbolic systems.
Free education is not a niche concept, but rather a central social necessity, especially in the age of digitalization. The issue is defending education as a public space where people, rather than algorithms, decide on content, goals, and methods. The question is not whether AI will remain in education but how we can shape it to serve the principles of openness, participation, and the common good. The free education movement can play a central role here. It can suggest alternatives, establish spaces for experimentation, and remind us that education is an act of emancipation. So, let’s hack AI for better education!