PSYCHOMETRIC PROPERTIES OF AN AI-GENERATED REMEDIAL ENGLISH TEST COMPARED WITH A TRADITIONAL TEST ACCORDING TO CLASSICAL & MODERN MEASUREMENT THEORIES

Abstract

The fast inclusion of artificial intelligence (AI) in the educational evaluation has posed crucial questions about the psychometric standards of AI-generated tests in contrast to the traditional forms of assessment. The present research examines the psychometric characteristics of an AI-generated remedial English test as compared with a human-created traditional test, through both of the Classical Test Theory (CTT) and the Item Response Theory (IRT) in the context of the modern measurement theory. The research was carried out within the setting of Palestinian universities where there are constant evaluation of assessment practices issues based on political instability, excessive instructional workloads and frequently changing examination material. The students in the university who took remedial English courses were used to collect data, and both the test forms were evaluated based on reliability, validity, item difficulty, discrimination, and guessing parameters. The use of CTT indices to test internal consistency and basic item characteristics and the Three-parameter Logistic (3PL) IRT model to give a more comprehensive analysis of item functioning and measurement accuracy at varying levels of abilities was done. The purpose of the findings is to identify whether AI-generated evaluations prove psychometric similarity in comparison to conventional tests and whether there is a reliable usage when applied to diagnostic and remedial situations in the English language. The research provides empirical findings to the accumulation evidence on AI-aided assessment and presents effective implications on the introduction of AI tools into education measurement without jeopardizing psychometric integrity and equity.