Research Summary

Abstract:

The arrival of generative conversational agents has introduced into the language classroom an unprecedented kind of interlocutor: permanently available, linguistically competent, indifferent to judgement, and yet devoid of any intention to communicate. This article examines the conditions under which such an agent may be considered a genuine interaction partner in the French as a Foreign Language classroom rather than a text generator. The study was conducted with master's students at a Moroccan public university, in a context where French operates as the working language of higher education and as an instrument of selection, and where linguistic insecurity remains acute. Three working conditions were compared over eight weeks: interaction with the agent framed by explicit pedagogical scripting, interaction with the same agent left to the students' autonomy, and peer-to-peer dyadic interaction. The analysis draws mainly on a corpus of transcribed interactions, an observation journal, and semi-structured interviews. Three regimes of use emerge, ranging from outright delegation to scripted interaction. The clearest finding concerns the scarcity of negotiation-of-meaning episodes in exchanges with the agent: it almost never signals a failure to understand, thereby depriving learners of the friction that makes language work.