AI Self-Efficacy in Chinese Multilingual Learners’ Informal English Learning with Large Language Models: A Mixed-Methods Study

The rapid diffusion of large language models has made artificial intelligence (AI)-mediated informal digital learning of English (AI-IDLE) an emerging form of self-directed learning beyond formal classrooms. Although previous research has examined learners’ emotions and self-efficacy in AI-IDLE, less is known about how confidence in using AI relates to participation through emotions and ethical awareness specific to AI use. This explanatory sequential mixed-methods study combined a cross-sectional survey of 1955 Chinese undergraduate English majors also studying an additional foreign language with interviews involving 30 respondents. Adapted five-point scales measured AI self-efficacy, enjoyment, boredom, ethical awareness, and AI-IDLE. Structural equation modeling identified a positive direct association between AI self-efficacy and AI-IDLE, alongside significant indirect associations through greater enjoyment, stronger ethical awareness, and lower boredom arising from AI use. Thematic analysis of the interviews contextualized these patterns through participants’ experiences of redirecting unsatisfactory exchanges, extending enjoyable practice, withdrawing from repetitive interactions, and setting boundaries around privacy and authorship. Drawing on quantitative and qualitative insights, this study identifies complementary affective and ethical-evaluative associations linking AI self-efficacy to AI-IDLE in an English-major sample with additional foreign-language learning experience. Learners’ accounts of evaluating AI assistance connect these participation patterns with the application of human intelligence.