Learners’ perceptions of chatbot-mediated EFL conversation practice: Evidence from post-hoc questionnaires

This thesis is part of the UNITE – Universally Inclusive Technologies to Practice English project, which investigates the pedagogical potential of chatbots for English as a Foreign Language (EFL) practice and aims to build and analyze a learner corpus of written interactions between Italian university students and conversational AI systems. The present study focuses specifically on learners’ perceptions of chatbot-mediated interaction, drawing on post-interaction questionnaires collected during controlled experimental sessions, where participants completed two written conversational tasks (small talk and role-play) with either ChatGPT or Pi.AI. The study combines quantitative analyses of Likert-scale responses with a qualitative and descriptive thematic analysis of open-ended feedback, which was manually tagged. The research examines (i) the most recurrent themes and evaluative tendencies in learners’ comments, (ii) whether thematic patterns vary by task type, (iii) whether self-reported English proficiency influences learners’ evaluations, and (iv) whether chatbot type affects task preference and motivation to continue using chatbots for language practice. Results indicate that learners’ perceptions are predominantly positive. Task type was associated with distinct thematic patterns: role-play was perceived as more engaging, whereas small talk was more frequently linked to positive evaluations of topic choice and comfort. English proficiency significantly affected perceived understandability and autonomy, with differences mainly observed between elementary and intermediate levels. Although chatbot type was significantly associated with task preference, it did not significantly influence learners’ motivation to continue using chatbots for EFL practice.