Language learning development in human-AI interaction: A thematic review of the research landscape

Abstract

Interaction is an indispensable part of language learning. Artificial intelligence (AI) has been increasingly applied in language learning to promote interaction in the learning process. In response to the paradigmatic shifts in AI application design, this review maps the research landscape of language learning development in human-AI interaction. From the resulting analysis of 49 studies, this study investigates the contextual characteristics by AI-supported interaction type, AI application, target language, educational level, etc. Moreover, three research paradigms are identified in this emerging field, i.e., Paradigm One (AI-directed, teacher-as-facilitator, learner-as-recipient), Paradigm Two (AI/teacher-codirected, learner-as-collaborator), and Paradigm Three (AI/teacher/learner-codirected). The paradigms are induced through analysis of eight constructs: human-AI relationship, learning objective, task type, level of pre-structuring, mode of engagement behavior, knowledge-change process, cognitive outcome, and research focus. The philosophical and linguistic underpinnings for each paradigm are discussed. Additionally, we highlight future research implications including investigating under-researched themes and exploring diverse methodological possibilities and appropriateness among the three research paradigms.