Comparative analysis of neural network capabilities for practical tasks in regulatory policy
Irina I. Rakhmeeva, Dr. Sc. (Econ.), Associate Prof., Head of Economic Theory and Applied
Sociology Dept. Ural State University of Economics
Alexander S. Ikrin, Graduate Student. Ural State University of Economics
Abstract
In all areas of management, there are high hopes for artificial intelligence in terms of improving the efficiency and speed of both decision-making processes and their subsequent implementation and control. At the same time, neural networks have significant limitations and shortcomings that pose critical threats to their use in public administration. The article aims to compare the quality of generated texts and identify the risks of using large language models in solving standard tasks of regulatory policy. Methodologically, the study rests on the intersection of public administration, lean regulation, and LegalTech theories. The key research method is an experiment, which involves conducting a critical comparative analysis of solving real-world problems using some of the most popular LLMs. The result, presented as a systematisation of the capabilities, limitations and recommendations for applying various neural networks to specific regulatory processes, can be used both in the daily work of civil servants and in the development of policy documents aimed at creating favourable conditions for economic activity and reducing the administrative burden.
Keywords: state regulation; lean regulation; digitalisation of public administration; LegalTech; neural networks in management; artificial intelligence; AI in regulatory policy; large language models.
For citation: Rakhmeeva I. I., Ikrin A. S. Comparative analysis of neural network capabilities for practical tasks in regulatory policy. Digital Models and Solutions. 2026. Vol. 5, no. 1, pp. 18–31. DOI: 10.29141/2949-477X-2026-5-1-2. EDN: COXDQK.
