AI-Generated Second Language Output: A Comparative Pragmatic and Discourse-Analytic Study of Speech Acts in Human and AI-Mediated L2 Communication

Abstract

This paper is entitled AI-Generated Second Language Output: A Comparative Pragmatic and Discourse-Analytic Study of Speech Acts in Human and AI-Mediated L2 Conversation and explores the question of whether AI-generated speech act in L2 displays similar pragmatic competence to that of advanced human L2 learners and the impact of AI mediation on the decision making of learners. The analysis is based on interlanguage pragmatics, speech act theory, and discourse-pragmatic studies, and is addressed by increasing integration of generative AI in scholarly communication, and the necessity to re-evaluate pragmatic competence in digital mediated communication. The comparative mixed-methods design was used. A total of 60 advanced L2 learners were used to gather data by completing discourse completion tasks and authentic academic email writing tasks, as well as parallel AI generated responses brought up by controlled prompts. Purposive sampling was used to guarantee high proficiency of participants in a EFL academic setting. Thematic analysis and qualitative discourse analysis of learner reflections were used to supplement quantitative pragmatic strategy coding and appropriateness ratings. Results have shown that AI output is highly pragmalinguistically sophisticated and structurally polite, but is based on standardized, risk-averse templates that lack contextual calibration. Sociopragmatic flexibility, cultural embedding and proportional sensitivity are higher in human learners. The research suggests that AI critical pedagogical integration can be used to improve pragmatic awareness without sacrificing learner agency and contextual reasoning during L2 communication.