COMPARATIVE ANALYSIS OF GENDER AND LOCATION-BASED DIFFERENTIAL ITEM FUNCTIONING IN AI-GENERATED AND WAEC ENGLISH LANGUAGE TEST ITEMS IN CALABAR, NIGERIA
https://doi.org/10.83151/62kt-rf92
The research examined gender- and location-based Differential Item Functioning (DIF) of AI-generated and WAEC 2025 English Language multiple-choice test questions among senior secondary students in Calabar Municipality Local Government Area of Cross River State, Nigeria. The study was guided by two objectives, derived from the relevant research questions. The research design was a survey, and stratified random sampling was employed to select a sample of 400 students from Senior Secondary School Two (SSII) in the Calabar Municipality Local Government Area. The research tools were AI-generated and WAEC 2025 English Language multiple-choice tests, all of them comprising 30 items. The scales were tested by scholars in the field of English Language and in the measurement of education, and reliability was determined by application of the Kuder-Richardson formula (KR-20), and the coefficient of reliability was found to be 0.76 and 0.81, respectively. The analysis of data was done with the help of Bilog-MG 3 at the level of 0.05. The results demonstrated that AI-generated and WAEC test items had DIF by gender and location, but the AI-generated test is lower in terms of DIF magnitudes, meaning that it is fairer and more balanced. The AI-generated test items exhibited less bias than the WAEC 2025 test items with respect to differences between urban and rural students. Also, the AI-generated test items exhibited less bias than the WAEC 2025 test items with respect to differences between male and female students. The conclusion was made that although both tests included DIF, AI-created items were fairer and gender or place-unbiased. The paper suggested that adequate human supervision, review of the items, and validation be done to promote fairness in AI-based tests.
Keywords: Differential Item Functioning, AI-generated test, West African Examinations Council (WAEC) English Language
Uchegbue, Henrietta Osayi, Otu, Bernard Diwa, Ita, Caroline Iserom, Ekaette Mfon Useh, Otu, Stelladeborah Bokan, Ogba, Uwaoma Flora, Ekereke, Aidam Benjamin, Beshel, Ignatius Akwagiobe, Emmanuel King
COMPARATIVE ANALYSIS OF GENDER AND LOCATION-BASED DIFFERENTIAL ITEM FUNCTIONING IN AI-GENERATED AND WAEC ENGLISH LANGUAGE TEST ITEMS IN CALABAR, NIGERIA https://doi.org/10.83151/62kt-rf92
The research examined gender- and location-based Differential Item Functioning (DIF) of AI-generated and WAEC 2025 English Language multiple-choice test questions among senior secondary students in Calabar Municipality Local Government Area of Cross River State, Nigeria. The study was guided by two objectives, derived from the relevant research questions. The research design was a survey, and stratified random sampling was employed to select a sample of 400 students from Senior Secondary School Two (SSII) in the Calabar Municipality Local Government Area. The research tools were AI-generated and WAEC 2025 English Language multiple-choice tests, all of them comprising 30 items. The scales were tested by scholars in the field of English Language and in the measurement of education, and reliability was determined by application of the Kuder-Richardson formula (KR-20), and the coefficient of reliability was found to be 0.76 and 0.81, respectively. The analysis of data was done with the help of Bilog-MG 3 at the level of 0.05. The results demonstrated that AI-generated and WAEC test items had DIF by gender and location, but the AI-generated test is lower in terms of DIF magnitudes, meaning that it is fairer and more balanced. The AI-generated test items exhibited less bias than the WAEC 2025 test items with respect to differences between urban and rural students. Also, the AI-generated test items exhibited less bias than the WAEC 2025 test items with respect to differences between male and female students. The conclusion was made that although both tests included DIF, AI-created items were fairer and gender or place-unbiased. The paper suggested that adequate human supervision, review of the items, and validation be done to promote fairness in AI-based tests. Keywords: Differential Item Functioning, AI-generated test, West African Examinations Council (WAEC) English Language