Artificial intelligence has drastically changed the way we work with the Internet. Just a few months ago, conducting research online meant searching for content using keywords. The information found was then processed manually by analyzing and comparing the texts and data. In contrast, generative AI has completely transformed this way of working with text. This is evident in the fact that today we no longer search for keywords but instead “prompt” the AI assistant—that is, ask questions in natural language—which it then answers. The result is a text that has already analyzed and answered everything that used to be the focus of a web search. In the best-case scenario, users receive some references to the sources on which the AI’s answers are based. These answers can then be gradually refined by using new questions to deepen or expand the search results. This new relationship with an AI assistant is often described as a collaborative partnership and dialogue between AI and humans. In this understanding, humans are no longer the sole creators of knowledge and information; rather, the AI assistant acts as a co-creator at their side. One of the bots within the LLM language models itself describes this as a form of collaborative partnership between humans and AI. What this means will be described in more detail below:

From a two-digit to a three-digit work process online

Until now, the relationship between information and the person producing it had been two-way: The producer or creator would engage with the information made accessible through Internet searches and use it, for example, to develop an academic text supported by evidence from books and information found online. However, even this stage of incorporating the Internet had already led to some confusion. For decades, it had been frowned upon to cite Wikipedia in scholarly works because the knowledge there was considered too unreliable and problematic.

With artificial intelligence, however, this relationship between humans and information now involves a third party. An AI assistant—interposed between them—now appears on the scene as a co-creator. This role of AI, however, is highly controversial in discussions about artificial intelligence. If we look at journalism, for example, the use of artificial intelligence to write texts is generally rejected in many quarters. There are fears that the actual work on texts will be handed over to AI models, which will then ultimately decide the content of media articles themselves. The situation is similar in AI music: instead of composing their own songs, producers are handing over lyrics and music to AI modules like Suno. The co-producers are becoming the actual music producers, who are beginning to displace all those musicians who still create and perform their own music.Heise onlinewrites about figures from the music streaming service Deezer: “On the streaming platform Deezer, the majority of music newly uploaded each day was AI-generated for the first time in June, the company announced. Specifically, the service recorded an average of 90,000 AI-generated tracks per day in June.” The fears associated with this are then further fueled by statements from major tech corporations—such as when Elon Musk repeatedly claims that AI will be more intelligent than humans in five years (as he recently stated in*The Economist*).”

But what are we to make of this partnership between artificial intelligence and human creativity? On the one hand, it really does seem that the productivity of human creators can be enhanced by the use of technical co-creators. Let me give an example from my experiments with AI music. For instance, I could never have written a song in the Berlin dialect, as I attempted to do with“Krawall bei Klärchen.”I had previously recorded this piece in the Zurich German dialect and set itin the Zurich Oberland. However, I then realized that the song fit very well into a major German city setting. So “Lanigan’s Ball” now exists in both dialects, even though I don’t speak the Berlin dialect myself.

Even when I’m looking for rhymes for a song, LLM models can be helpful because they were trained on—and have stored—a huge number of song lyrics. For tasks such as finding a suitable rhyme or crafting a line of text, the support provided by generative AI can be productive and useful. However, nonsensical and unhelpful suggestions—perhaps even “hallucinations” on the part of the AI—often creep in when AI assistants attempt to modify a section of text in a newspaper article.

In this vein,Mirko Lange emphasizesin Zurich’s *Tages-Anzeiger* that we must use machines while remaining fully aware of their weaknesses. But that also means new work—for example, to prevent a text from being “muddled” or “mush-ified” by AI. As creators, people must not simply hand over the work—whether in journalistic writing or in composing songs—to the “co-producer,” who would otherwise become the actual creator without being asked. As evidenced by the flood of AI-generated songs on Deezer, purely artificially created productions usually turn out to be pitiful and pathetic.

The “Co-creation” of Human and Artificial Intelligence

When we speak in this context of a “co-production” between human and artificial intelligence, there is another misunderstanding that needs to be clarified. This is because these are not true dialogues. The form of a dialogic pattern itself is part of the technical framework: The technology uses neural networks to conduct human-like, dynamic text-based conversations, which can span several steps. It looks something like this: Right at the start, the AI assistant says “Good morning” and asks what you want to know. It responds to the user’s prompts and often asks additional questions right away to keep the conversation going. When the assistants try to keep users on the platform for as long as possible, this is ultimately not for substantive reasons—such as the bot finding the conversation interesting. Rather, it’s due to the monetization model underlying these platforms.

Chatbots respond solely based on probability rules. There is no human thought behind it—only the algorithm. While empathy and sensitivity may seem to be part of the experience due to the way the AI addresses users, they ultimately play no role in interactions with the AI. The appearance of a dialog is further supported by the fact that the AI continuously analyzes user data and incorporates it into the chat. The AI assistant therefore already knows a great deal about me from my previous chats and can even surprise me with this knowledge. For example, it can point out that a certain question is a good follow-up to a problem I had presented to it three months ago. Here, too, however, it is not a matter of empathy or compassion; rather, the bot simply draws on the arsenal of information it has processed. From this perspective, a collaborative dialogue within the framework of cooperation between creator and co-creator appears, at first glance, to be nothing more than a deliberately designed simulation of collaborative communication.

Since the emergence of artificial intelligence, the relationship between technology and society has once again become a topic of intense theoretical debate. In this context, Johannes Harth (2021) draws on the ideas of Bruno Latour (2008) and his actor-network theory. This theory emphasizes the symmetry between human and non-human actors, or actants. The involvement of technology in social contexts must lead to a sociological recognition of technology’s participation. For Latour, however, this is not about understanding in the empathic sense. Following Luhmann’s theory of communication, it is completely irrelevant how or whether consciousness has understood something. When communication gets underway, it forms its own autopoietic system with its own recursively networked operations. This need not mean that the parties involved understand one another; rather, it is primarily about the connectivity of communication. Communication manifests itself as if the participating consciousnesses had a transparent picture of one another (see Harth, p. 9 ff.). However, as AI becomes increasingly powerful, this does not mean that it becomes more capable of communication. Rather, it merely achieves an ever-more-perfect simulation of communication. It remains unclear to what extent this is in any way related to successful communication processes that involve mutual understanding.

Admittedly, the dialogical principle emphasized by technological communication may have some validity, since the primary concern is that, within the framework of prompting dialogues, something resembling follow-up communication actually takes place and that this simulates mutual understanding. However, whether this follow-up communication actually achieves what it promises is highly uncertain and often not even transparent to the participants themselves. In this context, Harth speaks of “perfected simulation,” since the text appears entirely human. This results in the anthropomorphization of AI assistants, whom people believe possess human qualities such as empathy, understanding, and advisory competence. Yet even hallucinating—which leads people down the wrong path—is part of the logic of this system. This is because AI responds based on probabilities and algorithmic patterns; it has no awareness of its own as to whether something is true or false. And just as surely, it has no qualms about fibbing and inventing facts once it has made a false claim.

Conclusions from Media Education on the Relationship Between Creators and Co-Creators

From a media education perspective, however, it seems essential to me that people—including students—retain their role as creators and are consciously aware of it. AI can certainly serve as a stimulating partner who acts as a co-creator. But if an AI assistant claims in a post that a passage of text needs to be written or rewritten entirely differently, then users must verify this themselves and should not simply agree with the AI assistant blindly.

So it’s not primarily about (having to) learn prompting as a technique in order to keep up with the demands of AI. Rather, it’s about empowering ourselves to systematically evaluate the suggestions and information provided by generative AI models. Under no circumstances should the AI’s responses serve as the standard against which learners are measured. In design-oriented learning processes, the creators involved (i.e., students) must decide for themselves to what extent they adopt the AI’s suggestions, where they raise questions, or where they prioritize their own ideas. AI suggestions can be helpful and insightful, but they can also lead to becoming increasingly lost in the thicket of knowledge work.

Evaluating AI suggestions is likely to be easiest in cases where a bot makes specific recommendations for action that can be measured by whether the intended goal is achieved—for example, when the AI develops a small program for a PC or drafts the text for an email to a delinquent customer. It becomes more difficult when it comes to journalistic or scholarly texts. Here, the issue may not be limited to identifying clichéd phrasing; furthermore, the AI may attempt to sneak its own ideas into the authors’ work unnoticed.

However, when the journal *Medien + Erziehung* (merz), in a special issue on artificial intelligence, questions whether it can even still be the goal of media education efforts to turn media consumers into producers, I find this problematic. In this context, Andreas Büsch asks: “We must critically examine to what extent a creative and critical engagement with AI is even possible—or whether it remains merely a more or less creative manipulation of interfaces for practical purposes. However, the consequence of this cannot be stoicism, fatalism, or even abandoning the field of artificial intelligence as a subject of media education. Rather, massive educational efforts are needed to deconstruct the underlying narratives, which range from the promised salvation from numerous problems to the impending extinction of humanity” (Büsch 2025, p. 15).

In and of itself, this finding is valid. However, from a media education perspective, every effort must be made to ensure that the role of creators remains with people and is not handed over to technical co-creators. Here, media education would also need to go a step beyond traditional design-oriented media work. It is not merely a matter of working productively with media and incorporating design-oriented approaches within the framework of maker education. Rather, with generative AI, a third party has entered the picture, one that places its own demands on participation in design processes.

To illustrate this in the context of schools, I’d like to briefly share a personal experience. One of the chatbots I’m subscribed to had discovered that I’m a media educator and immediately offered suggestions on how it envisions working with artificial intelligence in schools. It described three options, which I’ll briefly outline here: It envisions “collaborative partnership” with generative AI in a school setting as follows:

"Here are a few different approaches to making this title even more elegant, vivid, or specifically tailored to the school, teachers, and students:

Option 1: Focus on the students’ new role (active & creative)
These suggestions place students at the center as active creators.
Brilliance Together: When Students and AI Become Co-Creators
Co-Creation in the Classroom: How Young People Learn and Create with AI
People & Machines as a Team: The New Art of Co-Creation in Schools
From Consumer to Co-Creator: Creative Work with AI in the Classroom

Option 2: Focus on the Interplay Between Pedagogy and Technology
These approaches emphasize the synergy between human guidance (teachers/students) and digital tools.
Learning in Tandem: Co-Creation Between Humans and AI in Everyday School Life
Creative Symbiosis: How AI and Schools Grow Together as Partners
Co-Creation in Schools: When Technology Inspires the Mind
The Creative Duet: Human Intelligence Meets Artificial Assistance

Option 3: Shorter, snappier, and “magazine-style”
Ideal for blog posts that aim to pique curiosity and have a modern feel.
Team Zukunft: Co-Creation with AI in the Classroom
Co-Creation: How AI Turns Schools Upside Down and Puts Them on a Creative Footing
People. Technology. School. The New Era of Co-Creation»

It will likely be impossible to avoid generative AI becoming one of the tools we use as part of learning processes, even in schools, in the future. Just as it has become impossible in recent years to imagine learning without Google searches, the same will be true with generative AI, based on this new approach of prompting. However, reading through the chatbot’s suggestions gave me a bit of a stomachache, as they read like PR for the interests of AI technology companies.

When it is emphasized that the new art of co-creation in schools transforms students from mere consumers into co-creators, this can be viewed positively because it supports the goals of design-oriented learning in the classroom. Ultimately, however, students as co-creators are on the same level as AI. Here, humans and machines already appear as a team, which implicitly assumes that AI possesses a kind of human capability. In this context, the creative process could quickly come to mean that it is ultimately the AI that formulates the “right solutions.” This also raises the question of why teachers are even needed in this team at all—or whether AI, as a co-creator, will ultimately replace teachers. Overall, it seems difficult to me to understand co-creation between humans and AI as learning in tandem or even as learning in a “creative team”—if the role of teachers in this “creative duet” is no longer clearly defined.

In my view, theories of teaching should focus on the fact that it is about the creativity and expression of students—and that tools such as artificial intelligence should not take center stage, but rather be understood as aids. AI will therefore not turn schools upside down or usher in a new era of co-creation, but it can be enriching and contribute to learning and creativity in schools.

Bibliography

Latour, Bruno: We Have Never Been Modern: An Attempt at a Symmetrical Anthropology, Frankfurt am Main: Suhrkamp, 2008.

Büsch, Andreas: "The End of the Enlightenment Project? AI as a Challenge for Media Education." In: Medien + Erziehung 3, 2024, pp. 10–17

Harth, Jonathan. “Simulation, Emulation, or Communication? Sociological Reflections on Communication with Nonhuman Entities.” In *Intersoziologie: Human and Nonhuman Actors in the Social World*, edited by Michael Schetsche and Andreas Anton. Weinheim: Beltz Juventa, 2021, pp. 143–5 (quoted here from an earlier draft of the article)

In this essay, I will use the masculine form to refer to artificial assistants, even though technology is not actually associated with any particular gender and is generally referred to as “technology.”