International Seminar of Semiotics in Paris

The theme of the 2025-2026 International Seminar of Semiotics aims to discuss certain results that have emerged, and questions that have come to the fore, during the past two years of seminars dedicated to artificial intelligence (AI) and enunciation theory (2023-2024), and to generative artificial intelligence (GAI), intersemiotic translation and creativity (2024-2025). In particular, it has become apparent that the texts produced by GAI require semiotics to reconsider, afresh, the sedimentation of discursive stocks and the dynamics that stabilise, diffuse or renew discursive forms and registers (praxis of enunciation).

From automatisms to stereotypes

The objective is to pursue reflections on automatisms which involve, on the one hand, bodily, psychic and cultural reactions that characterise the human and animal species – studied in detail in biology, psychology and psychopathology – and, on the other hand, linguistic constructions that may stabilise into highly recognisable formulas and formulations (motifs, topoi, or stereotypes). These discursive productions prove to be very useful – even conditioning communicability itself as a “commonplace” – since they are easily transportable and detachable from one discourse to another. Unlike motifs and topoi, which are the starting points of intertextuality and of creative reappropriations circulating across domains, stereotypes have often been relegated to a secondary role. They are frequently reproached for belonging to the folkloric or popular discourse, which places them modestly within the hierarchy of discourses. Nevertheless, they have always played an important role in political and artistic discourse.

When AI replays stereotypes

These stereotypes, generated by collective linguistic automatisms, find an echo in the products of GAI, which themselves stem from a complex history of language and reasoning automatisation, for which linguistics and cybernetics contributed foundational principles. The formalisation and automatisation of language brought with them a new questioning of discursive normativity (how can a machine adapt to changing contexts?) and of the possibility of producing texts that are more or less stereotypical or original.

Automatisation and discursive intention

Language automatisation is crucial today, particularly because current generative models are capable not only of imitation but also of plausibly recombining vast discursive corpora, producing utterances that, while unprecedented, reactivate stereotypical structures embedded within statistical regularities. This challenges our understanding of discursive intention. Furthermore, advances in deep learning in computer vision and natural language processing now allow for the production of visual and audiovisual utterances, revealing typifications and ideological trends within the framework of translation between languages based on different topologies and syntaxes. The case of GAI is revealing of a new way of (re)producing stereotypical discourses, since each response from a generative model differs from every other response that the same model will ever produce, complicating the notion of stereotypy. It thus becomes a kind of extensive magnitude governed by a system of embeddings, itself structured by statistical laws.

Standardisation in question

The quantitative standardisation of content (linguistic, visual, socio-cultural, political…) produces increasing automatisation within the different institutions of meaning and leads to a reification of forms of signification. This standardisation raises questions for semiotics at several levels: at the enunciative level, where automatised discursive production highlights the gap between personal and impersonal regimes; at the tensive syntactic level, where tensions arise between programmatic becoming and the advent of the unexpected. It also raises questions at the level of modes of existence, where one might wonder to what extent automatised semiosis merely reactivates, in each occurrence, the same potentialised heritage of forms.

Open questions

Does each new actualisation reorganise this heritage through a new syntactic and paradigmatic redistribution? At the level of narrative models, how does the question arise of the adequacy between invariant schematisations and singular discursive manifestations? Attention will focus on the evolution of linguistic stereotypes caught between biological, human and cultural automatisms, and the automatisations that characterise machinic modelling of languages.

- - - 

Practical information

• Sessions will take place at Maison Suger – 16-18 rue Suger, 75006 Paris
• Days and times: every Wednesday from 1:45 pm to 5:00 pm

Scientific Committee

Juan Alonso Aldama, Pierluigi Basso (coordinator), Denis Bertrand, Anne Beyaert-Geslin, Jean-François Bordron, Marion Colas-Blaise, Nicolas Couégnas, Ivan Darrault-Harris, Rossana De Angelis, Valeria De Luca, Maria Giulia Dondero, Veronica Estay-Stange, Jacques Fontanille, Didier Tsala-Effa