SGER: Adapative Competition and Emergent Control

点击翻译按钮获取中文摘要

Social and biological systems are subject to design and control at various levels. In particular, systems may be subject to centralized hierarchical control (top-down), or to control emanating inherently from the dynamics of agents whose actions in the aggregate constitute the system under consideration (bottom-up). In many such systems, agents will seek to act in ways that place them in a minority position. By doing so, agents can often optimize their ability to obtain resources. Examples include traders in a financial market, firms bringing innovative products to market, and packets of information seeking the least congested route on the internet. In such systems, good control of the system, as measured by a effective system-wideutilization of resources, can be sometimes be effected by an emergent coordination of agents' actions, without any top-down centralized control. This Small Grant for Exploratory Research (SGER) provides funding to study a class of simple models that contain the crux of these dynamics, and in which the emergence of good system-wide control is evident. The research will seek to elucidate the precise nature of the control, and the necessary and sufficient conditions for its appearance. The consequences of success in this research are far-ranging. At the theoretical level, this work can elucidate some of the unifying fundamental dynamics that undelie a wide range of complex adaptive systems in the social and biological sciences. In more practical terms, understanding the nature of the observed emegent control can lead to important advances in a number of areas. Most immediately, there are strong implications for the control of internet traffic. By exploiting more directly the distributed, heterogeneous nature of the internet, it may be possible to improve information flow considerably. Similarly, there are implications for vehicular traffic control. Another area of potential application is the financial markets. The emegent control observed in simple models depends crucially on the amount of information presented to the agents. Best control is obtained when the agents use a moderate amount of information to make their choices. This suggeststhat it may be possible to put in place regulations for financial markets so that the effective amount of information used by traders is near the optimal value for good emergent control. These are only a few examples of the kinds of problems that can be illuminated by this research.