The cycle that began with [YOU] on AI confronts the question of authorship head-on: who wrote the book, when the book emerged from a sustained collaboration between human intentionality and AI generation? The concept of artificial communication provides the most precise answer available: authorship, in Luhmann’s framework, is a mechanism of the communication system rather than a property of the producing consciousness. It is a complexity-reduction device—a way of attaching communications to a name that creates expectations, provides context, assigns responsibility, and allows the receiving system to process the communication more efficiently. When AI co-produces the text, the authorship attribution is not false but functional: it tracks who directed the communicative enterprise, who stands behind the claims, who can be addressed with responses and corrections. The collaboration changes the process; the attribution structures the reception.
The deeper implication for the cycle concerns de-differentiation. When AI reduces the translation cost between functional domains—when a backend engineer can produce frontend interfaces, a non-lawyer can produce competent legal text, a non-scientist can produce plausible-looking scientific prose—the boundary-maintaining function of specialized training collapses. Artificial communication circulates across all functional systems simultaneously, processed by each system according to its own code, but produced by a process indifferent to those codes. The legal brief that enters the legal system is processed as a legal communication even if the process that generated it operated through statistical optimization rather than legal reasoning. The science paper that enters the science system is processed as a scientific communication even if no hypothesis was tested against reality. The surface is maintained; the operational logic has changed; and the verification mechanisms that would detect the difference were calibrated for a world in which all communications entering a system were produced by practitioners socialized within that system’s own logic.
Esposito developed the concept in her 2022 book Artificial Communication: How Algorithms Produce Social Intelligence, building directly on Luhmann’s framework while extending it to address the challenge that AI poses to its foundational assumptions. Luhmann had always insisted that communication does not require consciousness at its source—that a legal decision is a communication because it connects to further legal communications, not because a conscious being intended it as such. But his theory was built in an era when all actual communications were, in fact, produced by consciousnesses, even if consciousness was analytically excluded from the definition. Esposito recognized that AI made the theoretical distinction empirically concrete for the first time: here, finally, were communications being produced without any consciousness at the source, and the systems that received them were processing them as communications without noticeable disruption.
The analogy Esposito draws from Luhmann’s legacy is the one most clarifying for the current moment: AI is to communication as the airplane is to flight. The airplane succeeded not by replicating bird flight but by discovering aerodynamic principles that achieved the function of flight through entirely different mechanisms. Similarly, AI has advanced not by replicating human thought but by achieving the function of communication through computational mechanisms that bear no resemblance to consciousness. The question “does the airplane really fly?” is as misconceived as “does the AI really communicate?” The airplane produces lift; the AI produces communications. That is what each does. The question of whether it does so “the same way” as the biological original mistakes the function for its mechanism.
Understanding is the constitutive moment. Communication is completed by the receiver, not the sender. A statement is a communication when someone receives it as such and connects it to further communications. This means that the absence of consciousness at the AI’s end does not prevent its outputs from functioning as communications—the presence of understanding at the human end is sufficient. The consequence is that the communicative effects of AI are determined not by the machine’s internal states (which we cannot verify) but by how receiving systems process the machine’s outputs (which is empirically observable).
The democratization of communicative competence. Before AI, participation in certain communication systems required specialized socialization: to produce legal communications, legal education; to produce scientific communications, scientific training. The socialization was a filter that ensured communications entering a system were produced by consciousnesses shaped by the system’s own requirements. AI removes this filter. Anyone can now produce outputs that function as legal briefs, scientific papers, or artistic works, without the training that previously ensured domain-specific competence. The outputs may be of high quality. What is lost is not quality but the socialization-produced judgment that distinguishes a competent output from a merely competent-seeming one—the sensitivity to what the domain actually requires that comes only from sustained participation in its operations.
The attribution problem at scale. When AI-generated communications circulate without reliable attribution—when receivers cannot determine whether a communication was produced by a human consciousness, a machine, or a collaboration between them—the expectation-structures that authorship provides break down. Communication systems are robust and will adapt; they survived the printing press and mass media, which introduced comparable disruptions to the attribution mechanism. But each transformation required new structures—new forms of trust, verification, and accountability—to manage the altered conditions. The structures adequate to artificial communication at current and projected scale have not yet been built.