Агент-ориентированные модели в изучении коммуникаций экономических агентов

Abstract

Существует достаточно большое количество балансовых моделей, которые с помощью математических выражений описывают процесс обменов и определяют условия выполнения баланса потоков в системе. Одной из наиболее известных математических моделей такого типа является балансовая модель Леонтьева (модель «затраты-выпуск»). Однако все эти модели имеют определенные ограничения: модель Леонтьева описывает закрытую систему, кроме того, невозможно изучить процесс обменов в динамике, а значит выработать рекомендации по его улучшению. Для решения этой проблемы было создано семейство агент-ориентированных программных моделей. В этих моделях действуют экономические агенты, которые вступают в коммуникации с целью обмена товарами друг с другом. Характеристики агентов рассчитаны на основе уравнений статического баланса Леонтьева, дополнительно в модели введены деньги. Агентам может быть назначена одна из 6 возможных стратегий поведения. В данной статье показано, как поведение отдельного агента и всех агентов как группы влияет на протекание всего коммуникационного процесса в целом. Была разработана еще одна версия модели, в которой введены внутренние (виртуальные деньги). Функционирование этой модели доказывает, что эти деньги вращаются в закрытых циклических контурах, которые можно считать аутопоэтическими, т.к. при введении таких денег баланс сохраняется, а коммуникационный процесс ускоряется.

There are many equilibrium models which with the help of some mathematical equations describe the process of exchanges and determine the conditions under which the balance of product and money flows in the systemis established. One of the most famous is the model of Leontieff’s intersectoral equilibrium (input-output model). But all of these models have certain constraints. As to Leontieff’s model it describes a closed system and that is why it is impossible to analyze exchange process in dynamics and, as a result, it is impossible to work out recommendations how this process can be improved. As this problem solution a set of agent-based software models was engineered. Economic agents act in these models, they communicate in order to exchange products with each other. Agent’s attributes are calculated in accordance with Leotieff’s static equilibrium equations. Money is introduced into the models. Simulations show that equilibrium exists in conditions of any small money volume but if there is a money lack a crisis arises that means exchange volumes great decrease, moreover, the less money volume - the deeper and earlier crisis emerges. The question is “Can agents influence the process characteristics (system macroparameters) by means of their different ways of behavior?” It has appeared that they could really do it. Communication process is divided into separate exchange cycles. In an exchange cycle every agent receives a chance to communicate and it makes a choice of contractor and exchange volume in accordance with its predetermined strategy of behavior. There are 5 predetermined strategies (models) of behavior: • The first “list” strategy - all agents act in accordance to the list they are ordered in the system and the exchange volume is the maximum possible; • The second “maximum” strategy - every agent first tries to supply its maximum requirements, the exchange volume is the maximum possible; • The third “uniform” strategy - every agent in the exchange cycle tries to communicate with all system agents whose products he is willing to consume; the agent tries to buy their products in equal volumes; • The fourth “succeeding” strategy — every agent tries to communicate with the succeeding according to list agent: if the exchange act with this contractor is impossible the agent try to choose the contractor further in the list, the exchange volume is the maximum possible; • The fifth “neighborhood” strategy — agent tries to communicate with his neighbor agents, in case of impossible exchanges with them it widens this neighborhood; the exchange volume is the maximum possible. The optimal strategy in the view of communication process speed and crisis effects reduction appears to be the fourth one. The very interesting additional effect of “egoism” has been revealed. It arises when an agent chooses less optimal strategy. It results in its own benefit but this act spoils the overall situation in the system and the final characteristics of agent’s neighbors as well. The second model simulates the work of an open communication system because single system agent is determined to be “the environment”. There are two types of money in the model: the first one is common real money and the second one is internal, or virtual, money. The first type money is used in all exchanges and the second supports the exchanges between internal agents (non-environmental exchanges). Money volumes are restricted and they are in proportion with the product volumes in the system. During the exchanges agents first try to spend internal money and if there is not enough volumes of this kind money than real money are used for exchanges. Experiments show that if the agent attributes are in accordance with Leotieff’s equilibrium than communication process is a success. And if the environmental agent becomes the first agent in the list than virtual money works as the real one accelerating the communication process. We argue that in the system there are closed contours in which internal money flows and we have a possibility to discuss autopoietic patterns which exist in the system. Model was verified on the real economic data collected for a small municipality. Agent-based technology gives a chance to animate static mathematical model and to receive the results that can be used for recommendations working out. These recommendations can be used by economic agents in conditions of crisis. Agent-based technology is known as “the right mathematics for economic and social systems”.