TEXT GENERATIVE NEURAL NETWORKS IN STUDENTS’ RESEARCH ACTIVITIES

Abstract

The emergence of text generative neural networks has demonstrated to students fundamentally new resources for preparing essays, term papers, and even graduate qualification papers. Artificial intelligence (AI) as an alternative to the challenging, painstaking analysis of books, scholarly articles has significantly increased both the opportunities and risks of the use of text-based generative neural networks by students for their research activities. Strong arguments have justified, with certain limitations, allowing students to use artificially generated text in research activities. However, this bold managerial decision has actualized the search for reliable and effective ways to detect AI-generated text. The article presents the results of comparative analysis of ten texts generated by Gemini, Claude and Perplexity neural networks, reflecting the contribution to science of great educators (J. Comenius, K. Ushinsky, A. Makarenko, L. Vygotsky, etc.), as well as the results of detection of these texts by Anti-Plagiarism and Scribbr systems. Scientifically substantiated recommendations for university teachers to improve the quality of evaluation of student research papers containing artificially generated text are formulated.