A lot of academics in educational research have been unable to link up with usage of
artificial intelligence towards the actualization of quality education (SDG 4) goal due to a
gap on the technical knowledge required for AI development and deployment in
educational assessment. Therefore, researchers carried out a study titled “application of
Item facility indices in development of CAT test adaptive logic following a supervised
machine learning technique in educational research” This study employed a Research
and Development (R&D) design to create a Computer Adaptive Test (CAT) utilizing
achievement tests and a concise decision tree based on item facility indices. The
sample consists of 50 participants, with a reliability index of 0.74. The X-calibre using
IPLM was used to ascertain and categorize test items based on the b-parameter,
representing item difficulty. Items were classified as low difficulty (b < -1), moderate
difficulty (-1 ≤ b ≤ 1), and high difficulty (b > 1). Easy items included Items 1, 4, 5, 6, 7,
9, 21, 27, 30, 39, 41, 44, and 45, while moderate items comprised Items 3, 8, 10, 11,
13, 14, 15, 16, 22, 23, 24, 26, 32, 38, 40, 42, 43, 46, 47, 48, and 50. Hard items
included Items 2, 12, 17, 18, 19, 20, 25, 29, 34, 35, 36, 37, and 49. Facility indices were
used to construct a decision tree for adaptive logic, categorizing b-parameters as easy
(-4 ≤ x ≤ -1), moderate (-0 ≤ x ≤ 0), and hard (+1 ≤ x ≤ +4). The system dynamically
adjusts item difficulty based on examinee responses, tailoring the test to match
individual abilities. This approach exemplifies how facility indices can be leveraged to
develop an AI-driven adaptive assessment tool, improving testing efficiency and
accuracy by personalizing the experience for each examinee. The results highlight the
potential of AI technologies in revolutionizing educational assessment systems.

Keywords: Computer Adaptive test, Artificial intelligence, Facility indices & CAT design