Beyond Automation: A Conceptual Framework for AI in Educational Assessment

This paper synthesizes the main applications of artificial intelligence in educational assessment through a systematic review following PRISMA guidelines. Our comprehensive analysis yielded 60 studies that revealed five key areas of AI-assisted educational assessment: assessment design, automatic grading, data analysis, performance prediction, and feedback provision. Based on identified patterns and implementation challenges, we pro-pose a novel three-dimensional pedagogical framework for AI in educational assessment. Within this framework, we develop the Processual Assessment Integration Model (P-AI-M) to address the first dimension, distinguishing between assessment de-sign/development and implementation/ utilization phases. The complete framework integrates: (1) this processual dimension operationalized through P-AI-M; (2) a stakeholder dimension mapping the distinct roles and responsibilities of researchers, policy makers, school leaders, teachers, and students; and (3) a cognitive-taxonomic dimension aligning AI capabilities with revised Bloom’s Taxonomy levels. The model is grounded in established educational theories including assessment for learning, constructive alignment, and sociocultural perspectives on evaluation. By addressing recurring gaps between technological capabilities and pedagogical integration, our multidimensional approach provides educators and researchers with a structured framework for understanding where AI can most effectively enhance assessment while preserving essential human expertise. The framework offers both theoretical grounding and practical guidance for implementing AI assessment tools in pedagogically sound, ethically responsible, and equitable ways across diverse educational contexts.