Assessment and Practice in Educational Sciences In Press 2027 05 01 Developing a Generative AI Governance Model in Higher Education and Scientific Research Developing a Generative AI Governance Model in Higher Education and Scientific Research 1 18 EN Journal Article 2026 07 20 <p>Generative AI systems have entered universities fast. In just a few years, large language models and content-generation tools have reshaped how students learn, how researchers work, and how academic institutions manage knowledge. These tools offer real benefits—personalized learning, faster research cycles, and new analytical possibilities. They also raise hard questions about academic integrity, research originality, and data privacy. Yet most national higher education systems still lack a coherent governance framework. Responses tend to be reactive and fragmented. This study asks a simple question: What dimensions shape a generative AI governance model in higher education, and how can such a model be validated? The study is applied-developmental in purpose and uses an exploratory mixed-methods design. In the qualitative phase, ten experts sat for semi-structured interviews. Thematic analysis followed Braun and Clarke's six-step framework. Five dimensions emerged: academic ethics and integrity (24.4%), digital literacy and skill development (21.1%), strategic policymaking (20.0%), oversight and evaluation mechanisms (17.8%), and technical infrastructure and data security (16.7%). In the quantitative phase, 200 questionnaires were distributed; 155 valid responses were returned (77.5% response rate). Data were analyzed with PLS-SEM in SmartPLS 4. Construct validity was assessed through factor loadings, composite reliability, AVE, discriminant validity (Fornell-Larcker and HTMT), VIF, f², Q², and overall fit indices (SRMR, NFI). Findings confirm that five dimensions underpin generative AI governance: academic ethics and integrity, strategic policymaking, oversight and evaluation mechanisms, digital literacy and skill development, and technical infrastructure and data security. Academic ethics had the strongest effect (β = 0.364), followed by strategic policymaking (β = 0.302). The model explained 62% of variance in higher education effectiveness (R² = 0.62; Q² = 0.41). All five structural paths were significant. Fit indices (SRMR = 0.062; NFI = 0.912) confirmed good model fit. Multi-group analysis found significant differences between public and non-profit universities in technical infrastructure.</p> https://journalapes.com/index.php/apes/article/download/341/333