Application of artificial intelligence tools for data search and preliminary analysis in applied research of education systems

Abstract

The article outlines the problems of using popular artificial intelligence tools GROK 3 and ChatCPT 4 for collecting data on educational systems available on the Internet, their pre-processing and analysis. A significant proportion of recent publications is devoted to the development of new artificial intelligence tools, the study of algorithms of its work, methodo-logical and ethical problems of its application in various fields, the impact of artificial intelli-gence in general or of specific tools on academic success and learning outcomes at different levels of education. At the same time, applied research has not paid enough attention to as-sessing the validity, reliability, and relevance of results obtained using artificial intelligence, in particular such tools as GROK 3 and ChatCPT 4.
For this study, several problems of educational systems researches, which can be con-sidered as relevant also un broader contexts, were selected. The main attention was paid to the issues of collecting and preliminary statistical analysis of data on the number of students in higher education institutions, funding of higher education, and indicators of youth in-volvement in higher education.
It has been shown that both tools significantly speed up and simplify the implementation of routine research stages, but GROK 3 allows to obtain more complete, reliable and high-quality data, which may be due to both the ability y to access current data on the Internet and differences in algorithms. ChatCPT 4 is usually limited to providing a small amount of the requested data present in its database, which is sometimes very outdated. The advantage of GROK 3 is also that it collects not only the requested numerical data from various types of sources, but is also able to independently calculate simple statistical characteristics of sam-ples, such as mean values, medians, standard deviations, minimum and maximum values, from the primary data without additional questions, including the ability to make plausible estimates from incomplete, highly asymmetric samples. Nevertheless, in this case too, the general methodology should provide for verification of the results of artificial intelligence, in particular by comparing them with other known data.