МЕТОД ПРИНЯТИЯ УПРАВЛЕНЧЕСКОГО РЕШЕНИЯ В НЕКОММЕРЧЕСКОЙ ОРГАНИЗАЦИИ НА ОСНОВЕ ИСКУССТВЕННОГО ИНТЕЛЛЕКТА

Abstract:
Non-profit organizations make managerial decisions under a limited budget, non-financial goals and multiple stakeholders, while traditional decision-making methods rely on human judgement and are constrained by the limits of information processing, by cognitive biases and by the ordinal nature of expert assessments. The purpose of the article is to develop a method of managerial decision-making in a non-profit organization based on artificial intelligence as an element of the management system. The hypothesis of the study is that a direct transfer of corporate artificial intelligence solutions to the non-profit sector produces no result, and that the method requires specialization for the mission of the organization, the grant cycle and volunteer labour. The methodological basis includes a comparative analysis of classifications of decision-making methods, a comparison of classes of artificial intelligence technologies by three features, a generalization of the data of an industry survey of 246 organizations from 61 constituent entities of the Russian Federation and structural-functional modelling of the composition of the method. The scientific novelty lies in the development of a method that combines a hybrid architecture of a decision support system with machine learning elements and generative models and retains the final choice for the management subject. The results include a comparison of traditional methods and the method based on artificial intelligence by universal criteria and by criteria specific to the non-profit sector, as well as the composition of the method of five components covering data sources, processing stages, model types, criteria for selecting alternatives and the mechanism of interaction with the management subject. The practical significance lies in the applicability of the method by managers of non-profit organizations when selecting projects and allocating limited resources. Further work involves testing the method on the data of particular organizations.