Rethinking Assessment in the Age of Generative AI: A Theoretical Framework for Academic Integrity, Deep Learning, and Responsible AI Use in Higher Education

Abstract

Generative AI has quickly permeated higher education, presenting opportunities and challenges for teaching, learning, assessment, and academic honesty. While generative AI tools can assist students with tasks like explanations, feedback, drafting, translation, and idea generation, they also pose a threat to the traditional assessment systems that are heavily reliant on unsupervised written assignments. While existing literature has explored generative AI as a learning tool, as an academic integrity concern, and as a policy issue, there has been less conceptual work that brings together AI literacy, redesigning assessment, academic integrity climate, teacher AI competence, responsible AI use, and deep learning engagement. To bridge this, the present paper proposes a theoretical framework to understand how responsible use of generative AI can improve deep learning in higher education. The paper is based on the constructivist learning theory, self-regulated learning theory, assessment for learning, and technological pedagogical content knowledge. It presents seven theoretical propositions connecting generative AI literacy, assessment redesign, academic integrity climate, responsible AI use, teacher AI competence and deep learning engagement. The study makes a significant contribution to educational studies by moving the conversation from the prohibition and detection of AI use to ethical integration, evaluation change and learning-focused academic integrity.