Community Research and Development Information Service - CORDIS


AMD Report Summary

Project ID: 337122
Funded under: FP7-IDEAS-ERC
Country: Israel

Mid-Term Report Summary - AMD (Algorithmic Mechanism Design: Beyond Truthful Mechanisms)

The field of algorithmic game theory (AGT) is a relatively new interdisciplinary field of research that emerged in the late 1990s to study the interaction between algorithms and incentives. It lies at the crossroads of computer science, game theory, and economics, a combination needed to address many of the challenges posed by the Internet. Under the impulse of internal intellectual excitement the field has matured rapidly and is already influencing intellectually its three parent disciplines, as can be seen by its large presence at computer science, economics, and game theory venues. It also has significant implications for the Internet, and has become instrumental in shaping many of the most significant computational domains, including online advertising markets, matching markets, cloud computing, social networks, electronic commerce, and more.
The project embraces a broad segment in AGT, and focuses primarily on two themes: designing mechanisms with provably good performance for complex settings, and measuring inefficiencies that emerge due to strategic behavior in existing settings. The above two themes differ greatly in nature: the former seeks to engineer new protocols, the latter to analyze given ones. Naturally, they also employ a different set of tools in their analysis. Together they compose a coherent interdisciplinary area that has proved itself theoretically insightful and practically useful.
One of the important themes in the project is simplicity, which is considered to be a key concern for practical mechanisms. Indeed, it is common in practice to forego truthfulness and use simpler mechanisms for the sake of reducing the cognitive burden on the participants. Illustrative examples include ascending price auctions in high-stakes auctions (e.g., electricity, gas, and radio spectrum auctions), generalized second-price (GSP) auctions for online advertising, the National Resident Matching Program, which is used to assign physicians to hospitals in the US, and simultaneous item auctions, such as the ones the ones run by eBay. The principle of simplicity plays an important role in both parts of this project. From a mechanism design perspective, we aim to design simple mechanisms with good performance. With respect to measuring inefficiencies, we quantify the inefficiency of simple mechanisms that are used in practice.
In what follows I describe a sample of the recent work emerging from this project, which introduces new directions in pricing mechanisms, inefficiency of equilibria, and market equilibria.
New Directions in Algorithmic Mechanism Design: Algorithmic mechanism design (AMD) combines ideas from the computer science field of algorithm design and the economic field of mechanism design. It establishes a framework for designing computational mechanisms with proper incentives for cooperation in complex algorithmic settings. Algorithmic work has enriched the field of mechanism design and has become the arena of productive cross-fertilization between the above fields. This project focuses on the design of simple auctions for complex settings, and on the design of pricing mechanisms that give good welfare or revenue guarantees.
New directions in the price of anarchy: The price of anarchy (PoA), defined as the ratio of the worst-case cost of a Nash equilibrium of a game and that of an optimal outcome, quantifies the inefficiency of selfish behavior. The last decade has seen remarkably good bounds on this measure for a wide range of applications. Our work includes two main parts: (a) introducing new frameworks for studying inefficiencies of equilibria in realistic settings (e.g., the strong price of anarchy, which considers coordinated actions of agents, and the price of anarchy in large games), and (b) improving existing price of anarchy results by developing new proof techniques.
New Directions in Market Equilibria: The concept of price equilibrium is a beautiful economic notion. In a competitive equilibrium every item is associated with a price, and every single buyer receives a bundle that maximizes his utility. Unfortunately, such a strong equilibrium notion is guaranteed to exist only in a narrow class of preferences. We develop new equilibrium notions that are more amenable to Internet and computational settings.

Reported by

Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top