Background: Critical thinking (CT) has been identified as one of the key skills in the modern world. Despite a variety of CT assessment systems for secondary school and higher education, there is a shortage of CT assessments for children. Aims: To fill the gap, we developed a computer-based performance assessment (CPBA) of CT with automatic scoring. In the study, we analysed: (1) To what extent is the internal structure of the critical thinking assessment tool confirmed? (2) What are the patterns of students' behaviour within the CPBA? (3) To what extent is the convergent validity of the CPBA supported? Sample and Methods: The sample consists of 1689 students, aged 9-11 years. To answer the research questions, the following quantitative data analysis methods were selected. The Bayesian Network was applied to investigate the internal structure. Latent Class Analysis was used to investigate the patterns of students' behaviour. To verify the convergent validity, the students were asked to solve the standardized math and language literacy tests. Results and Conclusions:The theoretically expected internal structure of the CPBA was confirmed: the two-factor structure (the Analysis of information and the Making inferences subskills of CT) describes the data appropriately. Moreover, CT was described not as a list but as a sequence of cognitive operations. Four different strategies that lead to different results of CT were revealed. A weak positive correlation between the results of the CPBA and the achievement tests confirmed the test validity.

<p>This article discusses the problem of task comparability with the help of scenario-based tasks for metacognitive skills. Using the data of &ldquo;&laquo;4С&rdquo;&raquo; tool for measuring critical thinking (N=500), the comparability of two scenarios within an identical digital environment with one set of indicators was investigated. The main difference in the scenarios lies in the contextual characteristics. The measurement invariance analysis of the instrument using confirmatory factor analysis was conducted. The results show that even with the equivalent construct structure and tasks&rsquo; characteristics, the context of the scenario has an effect on the student`&rsquo;s performance. The main differences in results were recorded for tasks involving interaction with the environment, where the test-taker created an object with elements. Tasks involving working with text in a digital environment can be considered comparable in case of elements content change. The possible reasons behind the observed differences in scenarios are discussed.</p>

No abstract

No abstract