The Absent I in AI Literacy: Meaning-making with language generators

With the arrival of Artificial Intelligence (AI), the timeless and deeply philosophical question of ‘what is and what ought to be’ the relationship between humans and technology has acquired renewed urgency.
If we focus on the second part of the question — what ought to be — there appears to be broad consensus, at least within democratic societies, concerning the desired configuration of humans and technology. AI should serve human interests, remain under human control, and uphold human rights and values. The European Commission refers to this ideal as human-centric and trustworthy AI: AI systems should be lawful, ethical, and technically robust. The European AI Act has been introduced as the first comprehensive regulatory framework aimed at governing both the development and use of AI systems.
The first part of the question — what is the relationship between humans and technology, or, in philosophical terms, its ontology — has received considerably less attention. This is not because there is no prevailing conception of that relationship. On the contrary, I would argue that the dominant characterization of the human–technology relation has become so self-evident that it is rarely questioned at all.
Humans and technologies are assumed to be two distinct entities. On the one hand stands the human being: embodied, rational, emotional, social, and, to a certain extent, autonomous. On the other hand, stands technology: applied human knowledge, lifeless and material, an instrument that enables human development and progress or, when poorly designed or misused, threatens it. Of course, there is interaction between the two, and this interaction requires careful governance, which brings us back to the second part of the question. Yet the underlying assumption remains unchanged: ‘here’ is the human; ‘there’ is the technology.

The point of departure for this chapter is that this conceptualization of human and technology profoundly shapes how we imagine and institutionalize desirable human–technology arrangements. In other terms, ontology informs regulation. Drawing on the philosophical anthropology of Helmuth Plessner, I develop an alternative image of the human — the ambivalent human — whose relationship with technology is inherently paradoxical. I argue that this perspective casts current debates on AI, and particularly AI literacy, in a different light.
I develop this argument through one issue that has received considerable attention in recent regulatory debates: AI literacy. Rather than analysing the AI Act as a legal instrument, this chapter treats it as an expression of an implicit conception of the human that informs the current operationalisation of AI literacy. At present, AI literacy is primarily understood as a means of enhancing human control over technology by increasing users’ technical knowledge and understanding.
Through a phenomenological analysis of writing with an LLM application, I will further develop this image of the ambivalent human and demonstrate that current approaches to AI literacy capture only one side of the story. Writing with an LLM involves an ongoing process of interpretation through which we not only make sense of the world around us but also of ourselves. Moments of doubt, hesitation, and uncertainty are integral to such activities and, particularly in this early phase of learning how to live and work with AI, should not merely be overcome or managed away. Rather, these moments should be recognized as crucial anchoring points for moral (re)positioning and meaningful accountability.
Moreover, I will argue that while all meaning-making practices involve moments of self-loss, these moments take on a radical character in the context of LLM-assisted writing because the temporal structure of meaning-making itself is transformed. Traditionally, the author emerges alongside the text through a gradual process of drafting, revising, and reflection. With LLM-generated writing, however, the text comes first. The author — the “I” —is temporarily absent and must subsequently be reconstructed through a process of appropriation and interpretation.
This has direct consequences for how AI literacy is currently being operationalized. Beyond providing knowledge, tools, and protocols for working with AI applications, greater attention should be devoted to not-knowing, uncertainty, and the responsibility that accompanies the interpretation and evaluation of AI-generated outputs. Rather than understanding AI literacy solely as a matter of control, I suggest that it should also be understood as the cultivation of the capacity to dwell, however briefly, in moments of ambiguity and absence.