Artificial Intelligence–Enhanced Scaffolding and Peer Collaboration in Culturally and Linguistically Diverse Classrooms: Evidence from Asia

This experimental study compares the effects of scaffolding via artificial intelligence (AI) and peer scaffolding on learning engagement in culturally and linguistically diverse classrooms in Asia. It used an explanatory sequential mixed-methods design with data from 280 South Asian students (aged 14–18) from 12 schools in India, Pakistan, and Bangladesh. Quantitative data were collected using the Learning Engagement Scale, AI Scaffolding Effectiveness Questionnaire, and Peer Collaboration Quality Index. Qualitative data collection included semi-structured interviews and classroom observations Quantitative data were analyzed using ANOVA, multiple regression, principal component analysis, k-means, and random forest modelling. PCA analysis indicated components of engagement as AI-Scaffolded Engagement (42.3%), Peer Support Quality (21.1%), and Cultural Responsiveness (16.2%). ANOVA results indicate significant differences between scaffolding conditions (F (2,277) = 24.6). PCA analysis revealed components of engagement as AI-Scaffolded Engagement (42.3%), Peer Support Quality (21.1%), and Cultural Responsiveness (16.2%). ANOVA results indicated significant differences among scaffolding conditions (F (2,277) = 24.67, p < 0.001, η² = 0.15). Based on random forest models, learning engagement was predicted with 89% accuracy using AI feedback frequency and peers’ interaction quality as predictors. Qualitative data emphasised the importance of culturally responsive peer collaboration in supporting multilingual learning. The study shows that AI scaffolding and culturally responsive peer support can optimise engagement with students when taken together, consistent within Zone of Proximal Development.