Selection of artificial intelligence tools for scientific research in the field of education depending on the research tasks

- Вакалюк, Тетяна Анатоліївна (orcid.org/0000-0001-6825-4697), Семеріков, С.О. (orcid.org/0000-0003-0789-0272), Спірін, О.М. (orcid.org/0000-0002-9594-6602), Олексюк, Василь Петрович (orcid.org/0000-0003-2206-8447), Сіренко, Остап (orcid.org/0009-0006-4489-2110) and Осадчий, В. В. (orcid.org/0000-0001-5659-4774) (2027) Selection of artificial intelligence tools for scientific research in the field of education depending on the research tasks Ceur Workshop proceedings. ISSN 1613-0073 (In Press)

Text титулка для бібліотеки.pdf Restricted to Repository staff only until 19 December 2026. Download (376kB) | Request a copy

Abstract

The rapid spread of artificial intelligence (AI) tools in the practice of research in the field of education has outpaced the development of methodological norms for their application: the choice of a tool is mostly carried out spontaneously, without correlation with the research design and without assessing which cognitive operations are delegated to the algorithm. The purpose of the article is to substantiate the system of criteria and indicators for the selection of AI tools for scientific research in the field of education and, on its basis, to formulate recommendations for their application depending on the research tasks. The unit of selection is defined as the pair “research task - tool”; eight research tasks are identified - from the analysis of the state of research of the problem to the design and publication of the results. Specific and integral criteria (a total of 31 criteria) are formulated for each task, divided into blocking, significant and auxiliary. Using the expert evaluation method (12 experts; ranking consistency according to the Kendall concordance coefficient W = 0.79), 29 AI tools were evaluated on a four-point scale: statistical environments, qualitative analysis platforms, dictionary and NLP tools, semantic search engines, research agents, video analysis platforms, and generative language models. It was found that the same tool obtains different solutions in different tasks, and the share of unconditionally recommended tools decreases with the increase in the research autonomy they assume. The most discriminatory criterion was the linguocultural correspondence: only one of the eight text data analysis tools demonstrated a high degree of its manifestation for Ukrainian-language material. For multimodal data analysis, no tool was unconditionally recommended due to the lack of published data on the accuracy of automatic markup in pedagogical contexts. The proposed procedure should be used to justify the choice of tools in the methodological sections of publications, prepare declarations on the use of AI, and annually update institutional lists of recommended tools.

Item Type:	Article
Keywords:	artificial intelligence, educational research, tool selection, criteria and indicators, peer review, research autonomy, academic integrity, linguistic and cultural relevance.
Subjects:	Science and knowledge. Organization. Computer science. Information. Documentation. Librarianship. Institutions. Publications > 00 Prolegomena. Fundamentals of knowledge and culture. Propaedeutics > 004 Computer science and technology. Computing. Data processing > 004.4 Software
Divisions:	Institute for Digitalisation of Education > Department of Open Education and Scientific Information Systems
Depositing User:	професор Т.А. Вакалюк
Date Deposited:	23 Sep 2026 06:35
Last Modified:	23 Sep 2026 06:35
URI:	https://lib.iitta.gov.ua/id/eprint/750356

Downloads

Actions (login required)

View Item