Dozhdikov A.V. Enhancing governmental policy-making in demographics and migration through multi-agent Deep Reinforcement Learning: A case study with the MADDPG algorithm. RUDN Journal of Public Administration. 2025. Vol. 12. No. 3. Pp. 366-374. DOI: 10.22363 ...

Òåêñò ñòàòüè íà ñàéòå æóðíàëà URL: https://journals.rudn.ru/public-administration/article/view/46832 (äàòà îáðàùåíèÿ 06.11.2025)

Abstract

The study identifies the main social, political and economic risks associated with the “overproduction” of the elite, the reduction of the middle class, considering uncontrolled migration. To mitigate the risks, a general theoretical approach is proposed to optimize the “hyperparameters” of public administration procedures, “upgrade” the decision- making model using hybrid systems based on machine learning. The experiment was conducted for 7 regions with initially random features (the number of regions can be any). During the experiment with the MADDPG algorithm, the author shows the possibility of implementing a balanced migration, socio- economic and resource policy for an arbitrary number of regions in conditions of instability, chaotic, noise processes and interregional migration for an unlimited period while maintaining the main environmental parameters. Trained AI algorithms in joint activities showed population growth, economic growth and development of territories, rational use of available resources (without their depletion), balanced interregional migration. Further direction of the research involves the inclusion of the external migration factor and detailing the factors of interregional migration, economic growth and resource consumption in the context of the social structure of society. The prospect of application are hybrid human- machine control and decision support systems for the sphere of public political administration. The  study  identifies  the  main  social,  political  and  economic  risks  associated with  the  “overproduction”  of the  elite,  the  reduction  of the  middle  class,  considering uncontrolled migration. To mitigate the risks, a general theoretical approach is proposed to   optimize  the  “hyperparameters”  of public  administration  procedures,  “upgrade”  the decision- making model using hybrid systems based on machine learning. The experiment was  conducted  for  7  regions  with  initially  random  features  (the  number  of regions  can be   any).  During  the  experiment  with  the  MADDPG  algorithm,  the  author  shows  the possibility of implementing a balanced migration, socio- economic and resource policy for an   arbitrary  number  of regions  in conditions  of instability,  chaotic,  noise  processes  and interregional migration for an unlimited period while maintaining the main environmental parameters. Trained AI algorithms in joint activities showed population growth, economic growth  and  development  of territories,  rational  use  of available  resources  (without  their depletion), balanced interregional migration. Further direction of the research involves the inclusion of the external migration factor and detailing the factors of interregional migration, economic growth and resource consumption in the context of the social structure of society. The prospect of application are hybrid human- machine control and decision support systems for the sphere of public political administration.