AI and early language learning: A scoping review

Abstract

Abstract In recent years, researchers have started exploring possibilities of AI in early language education to personalize young children’s language input. This study tries to investigate these explorations from five aspects: (1) common AI tools in early language education, (2) theoretical foundations adopted by these studies, (3) the effectiveness of AI for children’s early language and literacy development, (4) moderating factors in these AI interventions (e.g., AI type and outcome measures), and (5) limitations of current studies on AI and early language learning. A scoping review search was conducted using six databases: PsycINFO, Academic Search Complete, Computer Source via Ebscohost, ProQuest: ProQuest Dissertations and Theses, Education database, and EBSCOHost Eric. 1296 records were initially identified through applying our search string to the databases and after the entire screening process, 12 papers remained. The results demonstrate that AI tools for language learning, particularly social robots and unique applications, are primarily developed based on concepts from developmental and educational psychology, as well as human–computer interaction. These tools are useful for language learning, regardless of the intervention language, the type of AI (general or narrow) used, or the outcome measures studied. Key factors influencing their effectiveness include personalization and the role of the robot as a peer. However, current research has limitations related to study design, interventions, and sampling. Altogether, understanding the five aspects can guide future research to address current gaps in the field and to be mindful of potential pitfalls due to issues arising from AI tool use and research design.