A human-centered approach to integrating positive psychology and artificial intelligence in EFL instruction

This mixed-methods study examined whether integrating Positive Psychology (PP) principles into AI-supported English as a Foreign Language (EFL) instruction was associated with gains in learners’ linguistic achievement and socio-emotional development in three Middle Eastern higher education contexts. Although previous research on artificial intelligence in language education has largely emphasized cognitive performance and technological efficiency, comparatively less attention has been paid to the emotional and cultural dimensions of AI-mediated learning. To address this gap, the study proposed a human-centered instructional framework that combined adaptive AI support with emotionally supportive PP-informed activities. Participants included 300 undergraduate EFL learners from twelve universities across Iran, Oman, and Türkiye. Twelve intact classes were assigned, through cluster randomization, to three instructional conditions: AI integrated with PP activities (AI + PP), AI-only instruction, and traditional instruction. Quantitative data included measures of language proficiency, Foreign Language Enjoyment (FLE), well-being (PERMA), and academic buoyancy, while qualitative data were collected through interviews, classroom observations, and teacher journals. Quantitative findings showed that the AI + PP group demonstrated significantly greater gains than the AI-only and control groups across all major variables. Mediation analysis further indicated that enjoyment partially mediated the relationship between instructional condition and language achievement. Qualitative findings complemented these results by showing that learners perceived the AI system as supportive and motivating, while teacher mediation and cultural adaptation played central roles in shaping learner engagement. Because classes rather than individual learners were assigned to conditions and instructors were not randomly rotated across conditions, these findings should be interpreted as evidence of association rather than definitive causal proof. With this caveat in mind, the findings suggest that AI-supported language learning may become more effective when combined with emotionally supportive and culturally responsive pedagogical practices.