Exploring the Use of Blended Assessment in Evaluating Undergraduates’ Learning Performance

The Fifth Industrial Revolution (5IR) emphasises the harmonious collaboration between technology and human-centred approaches, revolutionising educational assessment.Traditional standardised tests often lack adaptability, leading to inefficiencies and biases.This study examines the effectiveness of Item Response Theory (IRT) and Computer Adaptive Testing (CAT) in optimising assessment processes by improving efficiency, precision, and reliability.Despite growing interest in adaptive testing, research on its large-scale applicability in education remains limited, highlighting a critical gap this study addresses.A simulation-based quantitative methodology was employed, utilising Monte Carlo techniques to generate 1,000 examinee responses modelled through a four-parameter logistic (4PL) IRT model.Two test conditions-fixed-length CAT and variable-length CAT-were implemented to compare their effectiveness.Item selection followed the Maximum Fisher Information (MFI) criterion, while Bayesian Maximum A Posteriori (MAP) was used for ability estimation.The results reveal that variable-length CAT significantly reduces test length by approximately 30% while maintaining high measurement precision.Adaptive testing demonstrated lower estimation errors and higher reliability than fixed-length assessments, confirming its effectiveness in modern educational evaluation.Additionally, item parameter analysis provided insights into test design optimisation.These results underscore the advantages of integrating CAT in large-scale assessments, particularly in enhancing fairness, personalisation, and engagement.The study concludes that CAT is a viable alternative to traditional testing methods, aligning with the 5IR's emphasis on technological and human synergy in education.Future research should explore AI-driven CAT enhancements to further refine assessment accuracy and accessibility.