The effect of generative artificial intelligence on students’ English learning outcomes: A meta-analysis

Abstract

This meta-analysis aimed to assess the overall effectiveness of Gen-AI-based interventions on learners’ English outcomes and to explore potential variations across educational contexts, broad language domains, and specific language domains. A total of 34 independent studies involving 2,579 participants were collected from three academic databases. The pooled estimate analysis revealed a statistically significant effect of Gen-AI on English learning outcomes. However, considerable variations were found across the included studies. This suggested that the effect varied depending on contextual and instructional variables. Further moderator analyses showed that Gen-AI was more consistently effective at the tertiary level and for productive skills, specifically writing skills. In contrast, evidence for other language learning domains and educational contexts remained inconclusive due to the limited number of available studies. Thus, future experimental studies are still needed, particularly to examine Gen-AI’s effects in underexplored areas. Finally, the findings suggest that ELT practitioners can use Gen-AI to support English teaching and augment learners’ outcomes. This meta-analysis offers evidence-based insights for policymakers to produce proper guidelines on Gen-AI use in ELT contexts.

Ähnliche Arbeiten