Wikimedia Deutschland is a non-profit association for the promotion of free knowledge. We regard equal opportunity of access to knowledge and education as a human right. Our vision is a world in which all people can participate in, use, and augment the totality of human knowledge. In our view, knowledge is free only if it is available, changeable, and reusable free of charge for everyone at any time.

However, we observe that current developments in the area of artificial intelligence (AI) disregard fundamental principles of free knowledge, namely: transparency, free access, and participation.

Transparency: The most frequently used generative AI in education is ChatGPT. However, little is known about the proprietary large language model, its training data, or filters. Therefore, it is very difficult for external, independent persons to assess it. That is unacceptable in an area as sensitive as education, which is a public task. That is why we are calling for more transparency of the AI systems used in this area.

Access: Private companies charge high licence fees for generative chatbots. If public bodies wish to enable these systems to be used by everyone, they must pay high licence fees. Thus, public funds flow directly to private companies without the State or the citizens having the possibility to have a say in the technologies. It would be better if these public funds could be invested in the development of open, public interest systems.

Participation: Proprietary and commercial AI applications are contrary to the principles and convictions of open educational practice and modern pedagogy. For these principles and convictions also include being able to jointly negotiate and shape the selection of content and the use of applications such as AI tools. To achieve this, priority must be given to open and transparent applications that are understandable and can be adapted to pedagogical needs.

Within the framework of the collaborative event series Forum Offene KI in der Bildung (Forum on Open AI in Education), educational practitioners and researchers as well as people from civil society addressed the challenges and needs in relation to AI applications in education. This resulted in 10 recommendations for action for open, public interest AI technologies in education that focus on the relevance of AI in the context of education from a perspective of pedagogical openness.

We recommend the publicly funded development of public interest, open AI technologies for education with equal access opportunities for everyone. All AI systems used should be independently tested prior to their use in education. To achieve this objective, various measures are needed that should be implemented in the short or medium term.

Infrastructure and Access

• Short term: Federal states should create access to AI systems for everyone.
• Medium term: The Federal Government and the federal states should promote the (further) development of open, public interest AI alternatives for education.
• Medium term: The Federal Government and the federal states should provide public digital infrastructures for the hosting of open AI systems.

Open Educational Practice

• Short term: The Federal Government and the federal states should make all publicly funded learning and continuing education resources for AI available under a free licence.
• Short term: Federal states should promote the development of openness competencies in the context of digital and media education.
• Medium term: Federal states should promote a new culture and structure of continuing education with open formats for peer-to-peer and self-learning.
• Medium term: Federal states should support change in the learning and examination culture.

Fundamental Rights in the Digital Space

• Short term: The Federal Government should take regulatory measures to ensure the understandability of specific AI applications for users.
• Short term: The Federal Government and the federal states should ensure data protection when AI applications are used in education.
• Medium term: The Federal Government and the federal states should establish an independent testing body for AI.

We consider the present paper “Offene KI für alle!” (Open AI for all!) to be a work in progress, and thus a working paper. We therefore invite others to continue the collaborative process that we have initiated. In this spirit, we call on all political and pedagogical actors to take up the recommendations for action, disseminate them more widely, discuss them with colleagues, and develop them further - for example in the wiki version of the recommendations for action: https://meta.m.wikimedia.org/wiki/Offene_KI. We hope that this will contribute to stimulating a wide-ranging and differentiated debate on AI systems in education. We call on policymakers to address the challenges raised, to intensify dialogue with educational practitioners, to take up our recommendations, and to actively support and promote their implementation.

Behrens, S., Boos, C., & Waag, A. (2024). Warum Offenheit bei Künstlicher Intelligenz (KI) für Bildung und Wissenschaft wichtig ist. open-access.network. doi.org/10.64395/wya9r-wkw77.