We are pleased to announce the 4th international conference on automatically generated texts in Romance and Germanic languages to be held on September 21th and 22th, 2026 at the Technische Universität Dresden. Speakers presenting a paper are expected to participate onsite.

Organizing Committee (TU Dresden, Chairs of Romance Linguistics): Anna-Maria De Cesare, Franz Meier, Giulia Mantovani, Michela Gargiulo, Paolo Valentinelli, Tom Weidensdorfer

Scientific Committee
• Vahram Atayan (Universität Heidelberg)
• Valentina Bambini (Scuola Universitaria Superiore Pavia)
• Noah Bubenhofer (Universität Zürich)
• Davide Garassino (Universität Zürich, ZHAW)
• Annette Gerstenberg (Universität Potsdam)
• Pierre-Yves Modicom (Université Jean-Moulin Lyon 3)
• Mirko Tavosanis (Università di Pisa)

Keynote Speakers: TBA (please check the conference website)

---

Topics of interest to the AI-ROM-IV (2026) conference include, but are not limited to, four lines of research.

I. AI systems & language varieties

AI systems, in particular large language models (LLMs), are increasingly used to produce linguistic content across a broad range of social and communicative contexts (education, media, political and administrative institutions, science). This development raises questions about the contextual adequacy of the linguistic usage represented by AI-generated content.

• Can AI systems reproduce specific language varieties (diatopic, diaphasic, diastratic, diamesic, diachronic)?
• Can AI systems mimic individual styles, i.e. idiolects?
• Are some AI systems better suited than others at reproducing certain language varieties?
• Does AI-generated output risk establishing an artificial norm detached from speech communities? Are we witnessing the emergence of a new (supra-)standard norm, valid across different language varieties?
• Can AI systems recognize, analyze and annotate texts belonging to different language varieties?

II. AI systems & linguistic diversity

AI systems are trained by aggregating vast quantities of individual linguistic productions and transforming them into abstract statistical representations. This abstraction process challenges the preservation of linguistic nuances and raises concerns regarding the ethical and cultural implications of AI-driven communication.

• Does the abstraction operated by AI systems tend to flatten linguistic diversity? Which features associated with specific sociolinguistic groups are preserved, transformed or erased in AI-generated outputs?
• Are AI-generated outputs linguistically poorer than human-authored texts? If so, at which level? How can we measure these divergences? What are the consequences?
• Can AI systems lead to (individual / group-based) language loss? If so, how? How can this loss be prevented?
• Does the increased use of AI systems reinforce dominant varieties/high-resource languages? How does the increased use of AI systems affect minority, minoritized, and low-resource language communities whose speakers are underrepresented or even absent from training data?
• Can AI systems recognize, analyze and annotate texts belonging to minor language varieties?

III. AI systems & gender representation

AI systems produce output showing biases towards a wide array of social groups, including the representation of females and individuals who do not recognize themselves in a dichotomic gender system.

• How are women linguistically represented in AI-generated output?
• How are gendered minorities represented in AI-generated output?
• Can AI systems support more gender-sensitive communication? If so, how? Which AI-systems are most appropriate?
• Can AI systems recognize, analyze and annotate texts that do not communicate in a gender-sensitive way?
• What happens to different gendered-groups when AI-systems are used to simplify complex texts, e.g. in the context of ‘plain’ language?

IV. AI systems & forensic linguistics

Linguistic analysis in forensic contexts traditionally involves moving between individual linguistic behavior and group-based patterns, e.g., in authorship attribution, speaker identification and credibility assessment.

• How can current linguistic research on AI-generated output inform forensic linguistics?
• Can AI systems serve as a tool in forensic linguistics? If so, which ones are most appropriate?
• What happens when AI systems analyze/check/control/drive the comparison between individual language use and group-level regularities? What happens when individual linguistic practices are evaluated against an artificial collective norm whose sociolinguistic grounding is unknown?

Submission Guidelines. We welcome theoretical, empirical, and methodological contributions from all fields of linguistics. Comparative and contrastive analyses are particularly encouraged. Abstracts should be written in a Romance language (preferably Italian, French, or Spanish), German or English. Submissions should clearly state the research question, the methodology, and present some (preliminary) results.

Abstract submission and notification of acceptance: Abstracts should be submitted by April 10, 2026 to ai.rom@mailbox.tu-dresden.de. Abstracts must not exceed 500 words (references, tables and examples excluded) and should state the name, affiliation, and email of the author(s). Acceptance / Rejection of proposals will be communicated by April 30, 2026.

A selection of the papers presented at the conference will be published in a special issue of the diamond open-access journal AI-Linguistica. Linguistic Studies on AI-Generated Texts and Discourses (Ai-Ling) ai_ling.journals.qucosa.de.