Abstract
In education, much attention is paid to the development and evaluation of universal skills in schoolchildren. At the same time, the assessment of universal skills requires new test formats based on the observed actions of the student in the digital environment. Scenario-based contextual tasks serve as a promising format. However, the contextual diversity of such tasks can make it difficult to compare results obtained from different scenario tasks. This article aims to analyze the role of scenario task context in measuring two universal skills: critical thinking and communication. The work uses the methods of Generalizability Theory, which allows to analyze to what extent the results can be generalized for other contexts of scenario tasks, and how, by changing the number of indicators or scenario contexts, to ensure satisfied measurement reliability. The study is based on data from fourth-grade students who were tested with various scenario-based tasks of the “4K” instrument. The results of the analysis showed that the behavior of the test-takers differs in scenarios with different contexts, while the difficulties of the contexts are almost the same. To achieve satisfactory reliability, it is recommended to use at least two scenarios with different contexts, and the use of three or more scenarios with different contexts will reduce the number of indicators without loss of reliability. Also, the study evaluated the role of context when using alternative scenario-based tasks forms were used. The alternative forms were similar in the main problem and plot of the scenario, but differed in topic (content). Changing only the content of the scenario makes it possible to generalize the results across scenario forms, that is, alternative forms can be used interchangeably. This study demonstrates how Generalization Theory can be used to optimize the development of tasks, taking into account the requirements for measurement reliability.
Publisher
National Research University, Higher School of Economics (HSE)