Systematic Literature Review on the Integration and Role of Artificial Intelligence in the Development of Computer Adaptive Testing (CAT)

Abstract

The integration of Artificial Intelligence (AI) has sparked a fundamental transformation in assessment systems, shifting from static methods to dynamic and personalized paradigms through Computer Adaptive Testing (CAT). This study aims to map the state-of-the-art AI integration in CAT, identify technological evolution trends, and analyze the contributions of dominant algorithms in optimizing the core components of tests. A Systematic Literature Review (SLR) following the PRISMA guidelines was used to synthesize data from Scopus, WoS, IEEE, ERIC, and Google Scholar databases from 2020 to 2025. Macro analysis was performed with VOSviewer, and micro synthesis with NVivo. The results indicate an evolution trend from basic machine learning integration in 2020 towards automation systems based on Reinforcement Learning and Generative AI by 2025. Algorithms such as Deep Learning and Multi-Objective Optimization have been shown to improve the precision of ability estimation, while empirical findings demonstrate that the use of Model Trees (M5P) can reduce item counts by 85%–93% on clinical instruments without compromising score accuracy. In conclusion, AI is transforming CAT into a smart, efficient, and personalized assessment ecosystem, with challenges in transparency (explainability) and algorithmic bias as key priorities for the future development of evaluation systems.