Enhancing governmental policy-making in demographics and migration through multi-agent Deep Reinforcement Learning: A case study with the MADDPG algorithm

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.