FROM CRAMMING TO CONTINUOUS INTELLIGENCE: REFRAMING ASSESSMENT THROUGH ARTIFICIAL INTELLIGENCE IN HIGHER EDUCATION

The rapid integration of Artificial Intelligence (AI) into higher education is transforming assessment from a high-stakes, outcome-oriented practice toward a continuous, process-based model of learning. Drawing on recent systematic reviews, this paper synthesizes emerging evidence on AI-assisted Assessment for Learning (AFL) and AI-supported university assessment to examine how intelligent systems reshape cognitive, metacognitive and emotional dimensions of student development. The analysis highlights a persistent theory-practice disconnect, where pedagogical principles of formative assessment are insufficiently aligned with technological design. AI enables continuous data collection, personalized feedback and “stealth” assessment within digital environments, thereby reducing reliance on cramming and promoting long-term knowledge retention and transfer. At the same time, implementation challenges - including system complexity, user trust, educator TPACK readiness and disciplinary calibration - shape adoption outcomes. Emerging trends such as generative AI, sentiment analytics, attention-aware systems and Human–AI teaming signal a shift toward distributed intelligence and process-oriented evaluation. However, concerns related to explainability, algorithmic transparency, academic integrity and ethical governance require careful policy and design responses. The paper argues that sustainable AI integration in assessment depends on balancing innovation with pedagogical coherence, cognitive enhancement with fairness and automation with human oversight, thereby reframing assessment as a dynamic, transparent and development-centered ecosystem.