Artificial intelligence in computerized adaptive testing to assess the cognitive performance of students: A Systematic Review

Computerized adaptive testing is a form of assessment that uses artificial intelligence to provide cognitive diagnoses of learners, in order to situate them in their learning process and offer them personalized support. "Adaptive" refers to the embedded selection algorithm that optimally selects questions based on a continuous estimation of the learner's proficiency at each step of the assessment. Computerized adaptive tests (CAT) are computer-based tests that offer individualized, equitable questions based on the learner's abilities. It is designed to correctly measure skills in relation to a latent trait (e.g. comprehension, knowledge, personality characteristics). Item selection, ability estimation and automatic item generation represent the challenges posed by CAT. Traditional approaches use psychometric and statistical models for estimation. However, to improve their performance, many new methods have been developed using artificial intelligence and machine learning techniques. In this article, we present a systematic review of artificial intelligence approaches and techniques used in computerized adaptive testing, highlighting the benefits, challenges and future orientations of this field of study.