Not human, still social: Thai EFL learners’ willingness to communicate with ChatGPT

Abstract

As generative artificial intelligence (AI) becomes increasingly used for second language speaking practice, willingness to communicate (WTC) (MacIntyre et al., 1998) needs to be understood not only in relation to task conditions, but also in relation to how learners experience AI as an interactional counterpart. This study examined how their WTC was sustained across AI-mediated speaking tasks and how their social positioning of AI shaped that willingness. Eighty English as a foreign language (EFL) learners completed five speaking tasks with ChatGPT, each followed by a post-task WTC self-rating and written reflection. Results showed no statistically significant differences in WTC ratings across the five tasks, suggesting overall stability at the aggregate level. However, the reflective data revealed variation in how learners experienced the tasks in relation to their WTC ratings. Four recurring domains were identified: language and cognitive manageability, affective state, task and topic fit, and AI interaction experience. Interview findings further showed that learners’ WTC was linked to four positionings of ChatGPT: as socially present while recognizably artificial, as able to follow intended meaning, as nonjudgmental and at times emotionally sustaining, and as a perceived adaptive listener. Interpreted through a posthumanist lens (Pennycook, 2018), the findings suggest that WTC in AI-mediated speaking emerges relationally through the learner-AI encounter and that interactional affordances are essential to sustaining communicative willingness in AI-supported ELT contexts.