Iranian Journal of Educational Sociology 2645-3460 In Press 2026 11 01 Development and Validation of a Model for AI-Based Educational Interactions in Secondary Schools in Tehran Development and Validation of a Model for AI-Based Educational Interactions in Secondary Schools in Tehran 1 20 10.61838/jk8pjx58 EN Journal Article 2026 04 07 <p><strong>Purpose:</strong> This study aimed to develop and validate a comprehensive model of artificial intelligence-based educational interactions for secondary schools in Tehran.</p> <p><strong>Methods and Materials:</strong> An exploratory mixed-methods design was employed. In the qualitative phase, thematic analysis of theoretical literature and semi-structured interviews with 14 experts in educational management and communication sciences, selected through purposive and snowball sampling until theoretical saturation, was conducted using MAXQDA. In the quantitative phase, a researcher-developed 91-item questionnaire was administered to 361 secondary school students selected through stratified random sampling from Tehran public schools. Data were analyzed using SPSS and partial least squares structural equation modeling (PLS-SEM) in SmartPLS. Model reliability and convergent validity were assessed using Cronbach’s alpha, composite reliability, average variance extracted, factor loadings, and bootstrapping. The final model was additionally validated by 30 experts.</p> <p><strong>Findings:</strong> One-sample t-tests showed that data-driven school–parent interactions were significantly above the theoretical midpoint (M = 3.6146, t = 11.031, p &lt; .001), whereas all other dimensions were significantly below it (p &lt; .001). PLS-SEM results indicated that all component and indicator t-values exceeded 1.96 and all construct reliability coefficients exceeded .70, while AVE values exceeded .50. The endogenous construct demonstrated substantial explanatory power (R² = .785). Second-order loadings showed that quality of smart educational content had the strongest association with the overall construct (β = .928, t = 46.153, p &lt; .001), followed by evaluation and continuous improvement (β = .867), whereas technical infrastructure and AI platform had the lowest significant loading (β = .770). Expert validation further showed that all evaluated model dimensions significantly exceeded the criterion midpoint (p &lt; .001).</p> <p><strong>Conclusion:</strong> The validated model indicates that effective AI-based educational interactions depend primarily on high-quality smart educational content, continuous evaluation, multidimensional interaction, and teacher support, while technological infrastructure serves as an essential enabling foundation. Accordingly, successful AI integration in secondary education should be approached as a comprehensive pedagogical transformation rather than merely a technological intervention.</p> https://qijes.com/index.php/ijes/article/download/1563/1547