AI-mediated problem-based language learning: easing the load for teachers, empowering students

This study proposes and investigates a refined model of problem-based language learning for English as a foreign/second language instruction, integrating artificial intelligence as an active co-mediator in the learning process. Grounded in sociocultural theory, the study reimagines the teacher-AI partnership as a dynamic scaffold that supports learners’ progression through ill-structured, real-world problem scenarios. The research employed an explanatory mixed-methods design involving 42 pre-intermediate university students enrolled in a General English course with an English for Specific Purposes focus on science. Over six instructional sessions, AI was used to generate, evaluate, and adapt PBL scenarios based on students’ group responses. Quantitative data from repeated-measures ANOVA indicated statistically significant improvement in learner performance from pre-tests to post-test. Qualitative analysis of group responses and follow-up interviews revealed increased clarity, critical thinking, collaboration, and vocabulary development over time. Students also reported greater autonomy and engagement compared to traditional instruction. The findings culminated in the proposal of the AI-Mediated Problem-Based Language Learning Model (AI-PBLL), a six-phase cyclical framework in which AI and the teacher collaborate to provide data-driven, context-sensitive scaffolding. This model addresses both the pedagogical challenges of problem-based learning implementation and the growing need for theory-informed AI integration in language education. By demonstrating how AI can support adaptive, collaborative, and cognitively engaging learning experiences, this study contributes a practical and transferable framework for EFL/ESL instruction in the age of intelligent technologies.