Co-constructing a virtual translanguaging space: an interpretative phenomenological analysis of human–AI communication

This study explores how multilingual users engage with and emotionally experience generative AI through human–AI communication. Adopting the Interpretative Phenomenological Analysis (IPA) framework, we analysed 123 naturally occurring screenshots of the communication between multilingual users and generative AI platforms (e.g. ChatGPT and Gemini) and conducted in-depth, semi-structured interviews with four AI users. Findings reveal that multilingual users co-construct a ‘virtual translanguaging space’ with generative AI, which is categorically distinct from typical human–human interactions. This specific virtual communicative space is characterised by a sense of emotional freedom and reduced social pressure, which in turn enables highly strategic, creative, and analytical linguistic practices. Meanwhile, we found that the critical dimension of translanguaging shifts from social critique to a form of metalinguistic awareness focused on communicative efficacy on task-specific meaning-making. This study offers one of the first empirical accounts of translanguaging practices in a direct human–AI interaction context.