| A large body of research has been devoted to building models for assessing knowledge efficiently. The psychometric field is the earliest to tackle this problem. In particular, the Item Response Theory (IRT) is a classic approach to skill modeling introduced over four decades ago.;This study is part of a wider research program to better assess the strength and weaknesses of POKS in comparison with other approaches. It focuses on the comparison of POKS with IRT. The comparison was performed over two simulation studies and within the CAT framework, since this is what IRT was designed for. The first simulation is based on a 34 items test on the knowledge of the UNIX shell commands and the second simulation is based on a 160 items French language test. In both cases, the simulation consists in choosing the most informative item based on the Fisher information and the information gain criteria, and feeding the actual answer to the knowledge assessment technique. The result is then compared with the actual answers and the process is repeated from the first to the last test item and for each subject. (Abstract shortened by UMI.);A probabilistic network approach, named POKS was developed to provide a fine grained assessment while providing a simpler framework than Bayesian network (BN). |
