THE ASSESSMENT REVOLUTION: THEORIES AND METHODOLOGIES OF AUTOMATED ASSESSMENT USING MACHINE LEARNING FOR EVALUATING LEARNING PROGRESS

Rapid advances in artificial intelligence and machine learning have fundamentally transformed educational assessment practices, shifting evaluation from episodic, human-centered measurement toward continuous, data-driven monitoring of learning progress. This study aims to examine the theoretical foundations and methodological approaches underlying automated assessment systems that employ machine learning to evaluate learning progress in diverse educational contexts. A qualitative systematic review with an integrative analytical framework was employed, drawing on peer-reviewed studies from international journals across education, learning analytics, and computer science. The selected literature was analyzed to identify dominant assessment purposes, theoretical alignments, data sources, modeling techniques, and validation strategies. The results indicate that most automated assessment systems prioritize predictive accuracy and efficiency, frequently conceptualizing learning progress through performance-oriented metrics while offering limited alignment with established assessment theories such as formative assessment and construct validity. Theory-informed and interpretable models remain underrepresented despite their pedagogical relevance. The findings reveal a persistent gap between technological innovation and educational meaning-making in automated assessment research. This study concludes that the assessment revolution driven by machine learning will remain incomplete without stronger integration of educational assessment theory, methodological transparency, and interpretability. Aligning machine learning methodologies with robust assessment principles is essential to ensure that automated systems support meaningful evaluation of learning progress, instructional decision-making, and educational equity.