Routing in Congestible Networks

网络构成了现代社会和经济运转的基础设施。具体地说，交通网络允许通勤者、邮件和物质物品的移动，而通信网络(例如，互联网)实现了同样的信息传输。当用户在网络中自主选择自己的路线时，他们的目标是最小化他们的成本(以时间、金钱或能量衡量)，而不明确考虑他们给他人带来的负面外部性。几乎不可避免的是，随着时间的推移，对网络的需求、网络拓扑或两者都会发生变化。我们不知道网络用户是否以及如何通过选择新的路线来适应这些变化，这些用户是有限理性的，容易受到各种偏见的影响，并且可能不会协调他们的决定。在这项研究中，PI将在实验环境中系统地操纵网络结构，以了解高积极性的网络用户--自愿参加网络实验以获取报酬的人类主体--对这些变化的反应。推动研究的理论基准是博弈论中的纳什均衡解概念。在这一解决方案下，群体中任何成员选择的每一种策略都是对其他成员策略选择的最佳回应。PI将在计算机控制的实验室中构建复杂的网络，其中有多条连接预先指定的起点和预先指定的目的地的路线，推导出这些网络的均衡解，将其显示在计算机屏幕上，并要求受试者独立选择将各自成本降至最低的路线。然后，将引起的行为模式与操纵网络前后的纳什均衡预测进行比较，如果有偏离均衡游戏的情况，将使用基于经济学、心理学和组织科学的概念和想法的新理论来识别和解释。就更广泛的意义而言，这些研究的结果将有助于规划网络的扩展以满足需求或网络结构的变化，提高网络的效率，以及确定向网络用户在线显示哪些关于网络状况的信息。

Networks form the infrastructure for the functioning of modern societies and economies. In particular, transportation networks allow for the move of commuters, mail, and material goods, and communication networks (e.g., Internet) accomplish the same for the transmission of information. When users independently choose their routes in the network, their objective is to minimize their costs (measured in time, money, or energy) without explicitly considering the negative externalities they impose on others. Almost inevitably, over time the demand placed on the network, its topology, or both undergo changes. We do not know if and how network users, who are bounded rationality, susceptible to various biases, and who may not coordinate their decisions, adapt to these changes by choosing new routes. In this research, the PIs will systematically manipulate network structure in an experimental setting to learn how highly motivated network users - human subjects who volunteer to take part in network experiments for payment contingent on their performance - react to these changes.To achieve this purpose, six projects will be undertaken. The theoretical benchmark that drives the investigation is the Nash equilibrium solution concept of game theory. Under this solution, each strategy chosen by any member of the population is a best response to the strategy choices of the other members. The PIs will construct complex networks in a computer-controlled laboratory with multiple routes connecting pre-designated origins to pre-designated destinations, derive the equilibrium solutions for these networks, display them on computer screens, and ask subjects to independently choose routes that minimize their individual cost. The patterns of behavior elicited will then be compared with the Nash equilibrium predictions before and after manipulation of the network, deviations from equilibrium play, if any, will be identified and accounted for using new theories based on concepts and ideas from economics, psychology, and organization science. In terms of broader signficance, the findings from these studies will be useful for planning the expansion of networks to meet changes in demand or in the structure of the network, for increasing efficiency in networks, and for determining what information on network conditions to display on-line to network users.