Academic integrity in the assessment system: regulating the use of generative neural network

Abstract

The article examines the theoretical and methodological foundations of modeling artistic and design processes in the context of developing intelligent educational systems. The aim of the article is to analyze the theoretical and methodological framework for ensuring academic integrity in the learning outcomes assessment system amid the proliferation of generative neural networks, and to substantiate approaches to regulating their use in education. Methods. The study employs a set of interconnected methods. The theoretical foundation comprises systemic and structural-functional analysis, comparative analysis of scientific approaches to the use of artificial intelligence in education, and content analysis of scholarly publications and regulatory documents, alongside generalization, induction, and deduction methods to develop conceptual frameworks. Results. The transformation of academic integrity under the influence of generative models (GPT, LLaMA, Codex, DALL·E, MidJourney, GPT-4, Claude) is analyzed. Key risks are identified, including decreased autonomy in task completion, difficulties in establishing authorship, reduced assessment transparency, and the emergence of new forms of academic dishonesty. Approaches to regulation are substantiated, including the declaration of AI use, defining the boundaries of its application, adapting assessment formats, and developing digital and ethical competence among participants in the educational process. Conclusions. The necessity of comprehensive regulatory, methodological, and ethical governance of the use of generative neural networks is substantiated. It is established that effective assurance of academic integrity requires updating assessment approaches and fostering a culture of responsible AI use.