Dozhdikov A.V. Political System as a Machine Learning Model. Technologies of Socio-Humanitarian Research. 2024. No. 2 (6). Pp. 9-24.

Текст статьи/выпуска на сайте журнала URL: https://lunn.ru/sites/default/files/media/upr_NIR/izd_NGLU/TSGI/zhurnal_2_6_tehnologii.pdf (дата обращения 03.12.2024)

Abstract

The article is devoted to the issues of improving models of the political system through quantitative methods and approaches from the field of natural sciences. The main result of improvements should be associated with an increase in the speed and quality of political decisions made. Classic models by D. Easton, G. Almond and K. Deutsch have been considered in the context of machine learning. The concept of “data sampling” has been introduced to train the decision-making model. Data source: feedback from the external and internal environment and political communication. The author examines the problem of balancing political traditions and innovations has been found. Innovations (validation sample) in the process of operation of the political decision-making machine become traditions (training sample). A disproportion towards the validation sample will lead to “undertraining”, which means errors and crisis development of the political system. The disproportion of the training sample because of isolationism, traditionalism, and conservatism leads to “retraining” of the decision-making model, accumulation of errors and other risks. Balancing “traditions” and “innovations” is carried out by political ideology, the purpose of which is to increase the accuracy of decisions made and the adaptability of the model. The role of the internal chaos of the political system for the selection of hyperparameters is noted: model training is associated with a decrease in the values of the loss function. Because of interpreting the political system as a machine-learning model, it will be possible to expand the arsenal of political science methods. Firstly, clarifying the function of the elite in constructing meanings and generating training samples. Secondly, modeling the processes of political communication in a digital society. Thirdly, the creation of dynamic models of the development of the political system, taking into account the cyclical nature of global geopolitical processes. Fourthly, the formation of requirements for the state political ideology, which adjusts the “hyperparameters” of the decision-making model or triggers its “upgrade”.