It is based on the item response theory, and the sequential decisions are performed by designing special cases or generalizations of Markov Decision Processes (in particular, a Stochastic Shortest Path problem or a Partially Observable Markov Decision Process) as to be able to model different Computerized Adaptive Testing formalisms, in which questions are put forward by the examiner one at a time to the student.
Importance of the indicator for the students’ learning methodology:
The estimation of the latent trait can be done with a Bayesian estimator or with a Maximum Likelihood estimator, given a sequence of answers to several questions put forward by the examiner. An optimal examiner can be obtained by solving a constrained optimization problem. As the latent trait is sequentially estimated, the new questions are specifically taylored to the estimated knowledge of the student.
P. Gilavert, V. Freire. Computerized Adaptive Testing: A unified approach under Markov Decision Process. In O. Gervasi, B. Murgante, S. Misra, A. Rocha, C. Garau (eds.), Computational Science and Its Applications--Proceedings of the ICCSA 2022 Workshops, LNCS 13375, ISBN 9783031105210 (July 2022), pp. 591-602. http://link.springer.com/chapter/10.1007/978-3-031-10522-7_40