AI-enabled adaptive game-based language learning: looking back and looking forward

Although AI-enabled adaptive game-based language learning (AI-AGBLL) has emerged as a prominent approach to adapting to learners’ diverse personalized needs, there is a notable shortage of extensive evaluations detailing its development, implementation, and effectiveness. To bridge the gap, drawing on the framework of adaptive learning models, this study synthesized literature from 39 empirical studies on AI-AGBLL during 2010–2024 in terms of learner model, game model, instructional model, and evaluation model. It revealed that (1) AI-AGBLL was favored among English-as-a-foreign-language (EFL) primary school learners, and learners’ in-game performance, time on task, and help-seeking behavior were the most frequent parameters. (2) Tutorial games and gamification were the most popular game genres, with a predominant focus being placed on learners’ language (reading and vocabulary) skills and affective (motivation/self-efficacy) perceptions. (3) Intelligent agents, natural language processing, and data mining were the most popular techniques, and adaptive feedback/prompts, adaptive language contents, and adaptive dialogues were the most prevalent instructional supports. (4) Quasi-experimental designs were used in most studies, which showed typically beneficial effects, despite the fact that challenges such as technical issues, subpar AI algorithms, and cognitive overload were also discovered. This research imparts to a detailed comprehension about the current trends of AI-AGBLL study and puts forward useful implications for teachers, designers, and researchers.