We analyse the conversational and pedagogical behaviour of AI and human tutors in one-on-one task-based, text-based tutoring and measure their respective effects on learners' engagement, perceptions, and writing fluency. The study follows a crossover longitudinal experiment with 50 A2-B1 learners of French and records their interactions with human tutors and AI tutors. The learners' perceptions, engagement, and writing fluency are measured through surveys and keystroke logging.
Bibauw et al. (Wed,) studied this question.