A Graph-based Model for Assessment of Respondents’ Latent Characteristics with Dialogue Simulations

Abstract The object of the study is the application of the maximum likelihood method to psychometric models, as well as simulation of the passage of a dialogue simulator based on an acyclic multiorgraph. The purpose of the work is to determine the effectiveness of the method of estimating the probabilistic model in the IRT, depending on the difficulty of the dialogue simulator. In the process of work, the model of nominal categories, invented by Bok, in R language was studied, based on the TAM package, simulation was performed with a sample of 1000 respondents. The interactive simulator is presented in the form of an acyclic oriented multigraph, for which the distances between the vertices are calculated. To estimate the characteristics of the respondents obtained using the maximum likelihood method, the bias and the standard error were calculated. It was found that the graph-like model for assessing latent characteristics of respondents allows to obtain a result with an error of 12%. The results of the work can be useful in the design of the exam design, analysis and evaluation of tests, improving the quality of pedagogical measurements and improving the interpretation of the results.