Dozhdikov, A.V., “Degradation of Traditional Representative Democracy in Western Countries under Conditions of Overproduction and Crisis of Elites and Political Transition to Neural Network Democracy,” Technologies of Social-Humanitarian Research, No. 1(9), pp. 48-65, 2025. Dozhdikov, A.V., “Degradation of Traditional Representative Democracy in Western Countries under Conditions of Overproduction and Crisis of Elites and Political Transition to Neural Network Democracy,” Technologies of Social-Humanitarian Research, No. 1(9), pp. 48-65, 2025.ISSN 2949-2599DOI нетРИНЦ: https://elibrary.ru/item.asp?id=83018356Posted on site: 10.11.25 AbstractModern political systems in the US and Europe are facing systemic crises that can be compared to pathologies of machine learning known to data analysts and predictive systems developers. The analogy with a neural network suffering from overfitting helps to identify the main mechanisms of degradation of political systems of representative democracy in recent years. At the theoretical level, the idea of direct democracy faces serious scalability problems, because in large hierarchical communities, the implementation of direct democracy is impossible. Representative democracy, in turn, is often discredited due to corruption, violations of feedback procedures, manipulative political technologies, and even violence against opponents. As a result of the degradation of managerial competencies and the creative function of elites, internal violence in the political system increases, and attempts at external aggression arise, which in turn leads to increased geopolitical tension. The author of the article suggests considering an alternative model - neural network democracy. In this model, all politically active citizens become "neurons" in a common network that makes political decisions. Weighting factors and biases are provided depending on the qualifications and experience of the participants. Feedback and the error backpropagation mechanism serve as the basis for learning, and competition and rivalry form built-in defense mechanisms. Adjustable hyperparameters optimize the decision-making process, which significantly increases the efficiency of political management. As a result, a number of problems associated with the phenomenon of "retraining" of political systems are solved and dependence on overproduction of the political and economic elite is reduced. Some of the findings of this study may be useful for implementation in political decision-making procedures both in the Russian Federation and in other countries outside the usual "Western" political space.