1. médialab Sciences Po
  2. News
  3. Genre and GenAI

Genre and GenAI

During this seminar session, James E. Dobson will explore AI-powered chatbots through the lens of the instructional-conversational genre. His presentation will highlight the norms, ideologies, and expectations embedded in contemporary linguistic architectures.

Event, Research Seminar

salle Goguel, 56 rue des saint pères, 75007 Paris

Abstract

Instruction fine-tuned large language models that power contemporary chatbots should be understood as participating in what I call the instructional-conversational genre. Drawing on rhetorical theory, media studies, and affect theory, this talk situates the instructional-conversational genre alongside everyday genres that structure social action. This genre is constituted by discursive patterns drawn from prior human–machine exchanges, recursively reinforced through massive neural networks. It is instantiated through instruction fine-tuning, the inclusion of a system prompt as a pretext for dialogue, and the formatting of inputs in markup languages that rely on special tokens and role assignments to manage turn-taking. These conventions interpellate users into predetermined positions, shaping how interactions unfold and constraining the apparent openness of chatbot discourse.

Norms, ideologies, and values are embedded within these architectures, training procedures, and the genre itself. Understanding GenAI chatbots through the lens of genre enables a critical analysis of both their possibilities and their limits. The instructional-conversational genre has been so successful, in part, because it meets the needs of a widely distrustful public, eager to receive personalized responses to their inquiries and demands that are simultaneously apologetic, deferential, equivocal, and engaging. In pointing to the production of text from and human interactions with language models as shaped in part by genre expectations and conventions, this talk calls for giving historicized, rhetorical, and theoretical attention to the architectures, applications, and interfaces that shape human–machine discourse in contemporary GenAI.

Biography

James E. Dobson is Associate Professor of English and creative writing, director of the Writing Program, and Special Advisor the Provost for AI at Dartmouth College. He is the author of Critical Digital Humanities: The Search for a Methodology (Illinois, 2019), and The Birth of Computer Vision (Minnesota, 2023) and co-author of Moonbit (punctum, 2019).

Practical information

This seminar will be held in a hybrid format (in-person and online) and in English, on Tuesday, October 7, 2025, from 2:00 to 4:00 PM, in salle Goguel, 56 rue des Saints-Pères, 75007 Paris.

Registration is mandatory via this link.