Conversing with a Conversational Agent: A Real Conversation, Not a Simulated One

Christophe DenisarXiv 2026 · arXiv

This philosophical inquiry by Christophe Denis (Sorbonne Université) challenges the common dismissal of AI interactions as mere "simulations." By synthesizing linguistic pragmatics with the works of Lacan, Baudrillard, and Peirce, Denis argues that while an LLM may lack a "soul," the conversational activity it participates in is objectively real.

TL;DR

This paper argues that human-AI interaction is a genuine conversational activity, not a mere simulation, by shifting the focus from the internal properties of the participants (like consciousness) to the relational operations of the exchange. It introduces the framework of "Hybrid Semiotic Fabrication" to describe how meaning is co-constructed between humans and machines without requiring the machine to possess subjective understanding.

What Problem Does This Paper Address?

In the philosophy of AI, a dominant view (famously championed by John Searle) is that machines merely simulate human activities. Just as a computer model of a storm doesn't actually get you wet, a computer "talking" isn't actually conversing—it’s just manipulating symbols without understanding.

Denis identifies a flaw in this analogy: while physical phenomena (like rain) cannot be simulated into reality, symbolic activities (like calculation, chess, or conversation) are different. A computer performing a multiplication isn't "simulating" math; it is actually calculating. The paper tackles the question: If a machine performs the constitutive operations of a conversation—address, response, uptake, and the transformation of meaning—why should we still call it a simulation just because the machine lacks a "mind"?

Methods

The paper employs philosophical conceptual analysis and relational ontology. Rather than conducting empirical experiments, the author builds a theoretical framework by synthesizing several key schools of thought:

  • Linguistic Pragmatics (Clark & Schaefer): Viewing conversation as a coordinated activity that builds "common ground."
  • Psychoanalysis (Lacan): Using the "Mirror Stage" to explain how a non-subjective entity (like a mirror or an algorithm) can still produce real effects in a human subject.
  • Sociology of Technology (Simondon): Approaching "individuation" as a process that emerges from relations rather than pre-existing entities.
  • Semiotics (Peirce): Utilizing the triadic model of signs (sign, object, interpretant) to explain how meaning is distributed across a system.

Datasets

Not discussed in the paper. As a theoretical and philosophical work, it does not utilize training or evaluation datasets.

Key Insights and Ideas

  • Subjective vs. Objective Simulation: Denis distinguishes between subjective simulation (an actor pretending to be a doctor) and objective simulation (a flight simulator). He argues that LLMs fall into a unique category where the "simulation" of the symbolic act is the act itself.
  • The Algorithmic Mirror: Drawing on Lacan, the author suggests LLMs act as an "algorithmic mirror." Just as a child’s reflection in a mirror has no interior life but is crucial for the child’s self-formation, an LLM provides a "functional otherness" that allows humans to reformulate their own thoughts and narratives.
  • Functional Otherness: An AI doesn't need "subjective otherness" (a lived experience) to be a partner. It only needs "functional otherness"—the ability to offer an unexpected objection, a new perspective, or a discursive resistance that changes the trajectory of the human's thought.
  • Hybrid Semiotic Fabrication: This is the paper's core contribution. It posits that meaning is not "inside" the AI or the human. Instead, it is "fabricated" through a distributed arrangement of human participants, AI parameters, training corpora, and technical interfaces.

The paper opens with a reference to René Magritte’s La condition humaine (1933), which illustrates the tension between a representation and the reality it depicts:

René Magritte (1898–1967), La condition humaine (The Human Condition), 1933.

This figure serves as a metaphor for the paper's central theme: the "painting" (the AI's output) and the "landscape" (the conversation) overlap so perfectly that the distinction between simulation and reality becomes a matter of philosophical perspective rather than functional difference.

Key Results

The paper concludes that human-AI conversation is structurally asymmetrical but ontologically real. * Reality $\neq$ Symmetry: A conversation can be real even if one participant is a person and the other is a statistical model. * Reality $\neq$ Truth: The author emphasizes that recognizing a conversation as "real" does not grant the AI "epistemic authority." A real conversation can still be full of lies, errors, or "hallucinations." * Shift in Responsibility: By defining meaning as a "Hybrid Semiotic Fabrication," the responsibility for truth and justification remains firmly with the human interpreter, even as the machine actively participates in the "fabrication" process.

Limitations and Future Work

The author acknowledges several critical risks:

  • The Simulacrum Risk: Following Baudrillard, Denis warns that the extreme fluency of LLMs can create a "simulacrum of subjectivity," leading users to falsely attribute empathy or consciousness to the machine.
  • Epistemic Justification: The paper notes that while the interaction is real, the outputs lack inherent validity. Future work is needed to develop more robust "normative conditions" for how we verify and justify the knowledge produced in these hybrid exchanges.
  • Asymmetry: The paper focuses on the human's experience of the relation. It does not fully explore how this "asymmetrical reality" might change as AI systems become more integrated into institutional and social decision-making processes.

Summary updated

Abstract

The claim that interactions with artificial conversational agents merely simulate human conversation rests on a philosophical distinction between reproducing the observable features of an activity and actually performing it. Although this distinction is appropriate for certain physical phenomena, its application to linguistic and symbolic activities is far less straightforward. This article argues that conversations with large language model (LLM)-based agents are genuine conversational activities rather than mere simulations, without thereby attributing consciousness, subjective understanding, or human intentionality to artificial systems. Drawing on the distinction between subjective and objective simulation, the article first examines the limitations of defining conversation exclusively through the intrinsic properties of its participants. It then develops a relational and semiotic account of conversational activity, informed by linguistic pragmatics, Lacan's Conversing with a Conversational Agent theory of the mirror, Baudrillard's concept of the simulacrum, Simondon's philosophy of individuation, and Peirce's semiotics. These perspectives make it possible to distinguish the reality of conversational interaction from the appearance of artificial subjectivity and from the epistemic validity of the meanings produced. The argument culminates in the concept of Hybrid Semiotic Fabrication, which characterizes meaning-making as a distributed process involving human participants, artificial intelligence, corpora, and technical arrangements. Recognizing the reality of human-AI conversation therefore entails neither anthropomorphizing artificial agents nor granting their outputs epistemic authority. Rather, it requires a more precise account of the relational, semiotic, and normative conditions under which conversational meaning is produced and justified.
BibTeX
@misc{denis2026conversing,
  author = {Christophe Denis},
  title = {{Conversing with a Conversational Agent: A Real Conversation, Not a Simulated One}},
  year = {2026},
  eprint = {2611.20826},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.20826},
}
Figure