Artificial Intelligence as a Process of Societal Negotiation

Dickel investigates the conditions under which machines are regarded as “intelligent” within contemporary society. His point of departure is the assumption that artificial intelligence is not a clearly defined technical object, but rather a culturally contested category that is continuously reconstituted with each new wave of technological development. Whether a technology is perceived as artificial intelligence thus depends less on its internal technological functioning than on the expectations directed toward it—and on how it appears within, and intervenes in, everyday social life.

Shifting the Boundary Between Humans and Machines

At the core of Dickel’s analysis lies the observation that current AI systems increasingly challenge the traditional distinction between humans and machines. Whereas earlier AI applications primarily operated in the background of technical infrastructures, contemporary systems appear visibly and audibly as communicative counterparts that can assume a significant role within human communication. They expand beyond narrowly defined expert domains into virtually all areas of society. This shift—or further development—opens up an intermediate zone in which machines are no longer merely tools, but increasingly appear as autonomous actors: as collaborators, co-thinkers, and co-decision-makers. As a consequence, previously stable boundary constructions—especially those that normatively distinguish humans, endowed with naturally produced intelligence, from machines—are destabilized.

Communicative AI: Why Interfaces Matter

A central thesis of Dickel’s research is that machines are perceived as intelligent when they appear capable of communication. Communicativization is the key concept here. What is decisive is not primarily what machines are technically capable of, but how they encounter users and how they exert influence upon them. This is where specifically designed interfaces become crucial: chat windows, voice-based dialogues, and visual interfaces create a mediated contact zone in which machines can imitate human modes of expression. Complex algorithmic processes remain concealed, while users encounter only the interface—a form of “communicative resemblance” that brings machines into social proximity and renders them plausible as conversational partners.

The AGI Discourse as a Promise of the Future

Dickel also directs attention to the long-standing yet newly revitalized debate surrounding Artificial General Intelligence (AGI). AGI functions less as a short-term, technologically realizable objective and more as a future-oriented narrative that bridges cultural uncertainty regarding the actual status of contemporary AI systems. This narrative keeps open the notion of “genuine,” human-like intelligence as a future possibility and thereby establishes a horizon of expectation within which current machines can be interpreted as preliminary stages of that future. In this way, the discourse surrounding AGI symbolically stabilizes what remains technically indeterminate: the reality status of AI.

AI Between Fiction and Fact

The Mainz-based social researcher concludes that AI currently occupies an ontologically hybrid position—it is simultaneously a real technical system and a cultural projection. The societal efficacy of AI does not derive solely from its algorithmic capacities, but from the ways in which it communicates via interfaces, consolidates expectations, and challenges established social categories. Precisely because AI oscillates between fiction and fact, its societal status is subject to continuous renegotiation. According to Dickel, it is within this ongoing process of negotiation that the dynamism of the current AI boom is grounded.

The podcast discussion does not shy away from addressing the societal risks and challenges associated with these developments. At the same time, the guest acknowledges the conveniences afforded by contemporary technologies. In everyday life, he makes frequent and varied use of language models and—like many users—regularly encounters both highly useful and decidedly unhelpful results.

The episode takes up key aspects and arguments from Sascha Dickel’s analysis and develops further reflections in relation to current debates on generative AI and the use of language models. The conversation is conducted by Prof. Dr. Marcel Schütz and research associate Philipp Krüger from the KIWIT research group.

Last updated: 5 December 2025