Authenticity problem of educational content created by university students

this article addresses the current problem of recognizing texts generated by artificial intelligence (AI) in academic environment. With the widespread adoption of generative artificial intelligence (GAI) technologies, such as large language models and chatbots, in the educational process, teachers in higher education are faced with the need to distinguish original author’s content from AI-generated works. The article tells how to recognize text generated by a neural network from the one written by students independently, without the use of specialized software. The most popular tools for tracking AI-generated texts are also presented. This article presents a detailed analysis of the characteristic features of texts generated by neural networks, including limited use of synonyms, frequent word repetition, an abundance of clich?s, information incoherence between paragraphs, a lack of emotional depth, humor, and personal experience, inappropriate use of metaphors, and the potential presence of logical errors and fictitious facts, as well as the absence of spelling and punctuation errors. Using real written works of students majoring in non-linguistic specialties at South Ural State University (SUSU), we demonstrate methods and techniques for determining whether the author's content is original or generated using a neural network. We also discuss the educational processes in which the use of AI is justified and facilitates the students’ learning process, and those in which AI-generated learning outcomes are unacceptable. A conclusion is drawn concerning the appropriateness and extent of using neural networks in text generation.

natural intelligence, artificial intelligence, text generation, neural network, large language model, higher education teacher, students, original author’s content