RI: Medium: Collaborative Research: Methods for Empirical Mechanism Design

点击翻译按钮获取中文摘要

The design of computational systems that support group decisions, allocate resources to distributed tasks, or mediate social interactions is fundamentally different from the corresponding design problem serving individual or centralized users. When multiple parties, or agents, are involved, the designer's objectives are complicated by the fact that the interests of these parties are rarely, if ever, perfectly aligned. The field of mechanism design offers a theoretical framework that directly addresses the issue of incentives as it relates to the design of multiagent systems. However, this purely analytical approach carries with it inherent practical limitations. The investigators introduce a new approach, empirical mechanism design (EMD), whose premise is to extend the basic foundation of mechanism design with empirical methods such as simulation and statistical analysis. These extensions promise to dramatically expand the scope of mechanism design beyond the small-scale, stylized, or idealized domains to which it has been predominantly limited to date.The project will investigate several concrete EMD problems, within the general theme of market design. Improved market design has significant implications for the public and private sectors. In public policy, market-based approaches are likely to play a major role in, for example, instituting measures to cope with climate change, banking reform and regulation, and adoption of new energy sources. In the commercial domain, new markets for advertising placement, computational services, and other goods will also entail significant mechanism design efforts. Regardless of the sector, design outcomes bear on important social objectives including efficiency, transparency, and stability (e.g., of financial relationships). An empirical basis for evaluating candidate mechanisms will complement existing theoretical perspectives, enriching the tools available to designers and other stakeholders.